Democracy and Economic Openness in an Interconnected System: Complex transformations

http://www.cambridge.org/9780521491433

DEMOCRACY AND ECONOMIC OPENNESS

IN AN INTERCONNECTED SYSTEM

In this book, Quan Li and Rafael Reuveny combine the social scientific approach witha broad, interdisciplinary scope to address some of the most intriguing and impor-tant political, economic, and environmental issues of our times. Their book employsformal and statistical methods to study the interactions of economic globalization,democratic governance, income inequality, economic development, military violence,and environmental degradation. In doing so, Li and Reuveny cross multiple disciplinaryboundaries, engage various academic debates, bring the insights from compartmental-ized bodies of literature into direct dialogue, and uncover policy trade-offs in a growinglyinterconnected system of polity, economy, and environment. They show that growinginterconnectedness in the global system increases the demands on national leaders andtheir advisors; academicians and policymakers will need to cross disciplinary boundariesif they seek to better understand and address the policy trade-offs of even more complexprocesses than the ones investigated here.

Quan Li is Professor of Political Science and Director of the Program on InternationalConflict and Cooperation (PICC) at Texas A&M University, which he joined in 2008.Previously, he was a faculty member in political science at the Pennsylvania State Uni-versity, where he codirected the Multidisciplinary Seminar Series on Globalization in theCollege of Liberal Arts and served on the inaugural Faculty Governing Council of theSchool of International Affairs. Professor Li served on the editorial board of the Journalof Politics and is serving on the editorial boards of International Studies Quarterly andInternational Interactions.

Professor Li holds a Ph.D. in political science and international relations. His researchinterests focus on the causes and consequences of economic globalization (internationaltrade, foreign direct investment, financial openness, and capital account liberaliza-tion), democratic governance, political violence (interstate military conflict, civil con-flict, transnational terrorism), and macroeconomic policymaking and cooperation. Hisresearch has appeared in numerous journals, including the British Journal of PoliticalScience, Comparative Political Studies, International Organization, International StudiesQuarterly, the Journal of Conflict Resolution, the Journal of Peace Research, the Journal ofPolitics, and Political Research Quarterly. Professor Li is the corecipient of the 2003 BestArticle on Democratization Award from the American Political Science Association.

Rafael Reuveny is Professor of International Political Economy at the School of Publicand Environmental Affairs, Indiana University, Bloomington. His research focuses onthe causes and effects of economic globalization, democracy, international military con-flict, and sustainable development. He is the author and coauthor of numerous articlesand book chapters. Professor Reuveny’s work has appeared in journals such as the Amer-ican Journal of Political Science, the Journal of Politics, International Studies Quarterly,International Organization, the Journal of Conflict Resolution, Ecological Economics, andEnvironmental and Resource Economics. He is the coauthor or coeditor of five books, themost recent of which is North and South in the World Political Economy (2008). He wasalso a guest coeditor of a special issue of International Studies Review (2007).

Professor Reuveny was program chair of the 2006 meeting of the International StudiesAssociation and the North America program chair of the 2008 meetings of the GlobalInternational Studies Conference. Reuveny has won two teaching awards at IndianaUniversity and is the 2007 corecipient of the Award of Excellence in World SocietyResearch, First Place, given by the World Society Foundation, Zurich, Switzerland.Professor Reuveny is also the corecipient of the 2003 Best Article on DemocratizationAward from the American Political Science Association. He holds a double-major Ph.D.in business economics and political science.

x Acknowledgments

Chapter 3 is a thoroughly revised and extended version of the followingarticle: Reuveny, Rafael, and Quan Li. 2003. “Economic Openness, Democ-racy and Income Inequality: An Empirical Analysis,” Comparative PoliticalStudies 36(5):575–601. Copyright c© 2003 Sage Publishing.

Chapter 5 is a thoroughly revised and extended version of the followingarticle: Reuveny, Rafael, and Quan Li. 2003. “The Joint Democracy–DyadicConflict Nexus: A Simultaneous Equations Model,” International StudiesQuarterly 47(3):325–347. Copyright c© 2003 Blackwell Publishing.

Chapter 7 is a thoroughly revised and extended version of the followingarticle: Li, Quan, and Rafael Reuveny. 2006. “Democracy and EnvironmentalDegradation,” International Studies Quarterly 50(4):935–956. Copyright c©2006 Blackwell Publishing.

Chapter 8 is a thoroughly revised and extended version of the followingarticle: Li, Quan, and Rafael Reuveny. 2007. “The Effects of Liberalismon the Terrestrial Environment,” Conflict Management and Peace Science24(3):219–238. Copyright c© 2007 Taylor & Francis.

This book has grown out of our decade-long joint research and collabo-ration, friendship, and many debates since we first met in 1999. The qualityof our joint scholarship has benefited from both our agreements and ourdisagreements, all of which we were ultimately able to resolve with a smile.Without our mutual willingness to listen and compromise, and our con-tinuous mutual support and trust, this book would not have come intobeing.

Finally, our families have been important to the completion of this book.We owe many thanks to our respective parents in China and Israel, Li Maojiand Kuang Juying, and Mordechai and Ora Reuveny, for their unendingsupport. Our spouses and children, Liu, Ellen, and Andrew, and Ronit, Adi,and Noam, have given so much support, encouragement, and meaning toour work that we feel we must dedicate this book to them.

Democracy and Economic Opennessin an Interconnected System

Complex Transformations

QUAN LITexas A&M University

RAFAEL REUVENYIndiana University

CAMBRIDGE UNIVERSITY PRESS

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,

São Paulo, Delhi, Dubai, Tokyo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

First published in print format

ISBN-13 978-0-521-49143-3

ISBN-13 978-0-521-72890-4

ISBN-13 978-0-511-65157-1

© Quan Li and Rafael Reuveny 2009

2009

Information on this title: www.cambridge.org/9780521491433

This publication is in copyright. Subject to statutory exception and to the

provision of relevant collective licensing agreements, no reproduction of any part

may take place without the written permission of Cambridge University Press.

Cambridge University Press has no responsibility for the persistence or accuracy

of urls for external or third-party internet websites referred to in this publication,

and does not guarantee that any content on such websites is, or will remain,

accurate or appropriate.

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org

Paperback

eBook (NetLibrary)

Hardback

http://www.cambridge.org

http://www.cambridge.org/9780521491433

Contents

List of Figures and Tables page vii

Acknowledgments ix

1 Introduction 1

PART I: THE DEMOCRACY–ECONOMY NEXUS

2 Democracy and Economic Openness 23

3 Democracy, Economic Openness, and Income Inequality 62

4 Democracy and Development 89

PART II: BRINGING IN CONFLICT

5 Democracy and Conflict 125

6 Economic Openness and Conflict 158

PART III: BRINGING IN THE ENVIRONMENT

7 Democracy and the Environment 205

8 Economic Openness and the Environment 239

9 Conflict and the Environment 266

10 Conclusion 292

References 309

Author Index 337

Subject Index 344

v

List of Figures and Tables

FIGURES

1.1. Graphical layout of the book page 56.1. Disaggregated bilateral trade equilibrium 167

10.1. Graphical layout of key findings 294

TABLES

2.1. Globalization promotes democracy 282.2. Globalization obstructs democracy 312.3. Globalization does not necessarily affect democracy 342.4. Effects of economic globalization on democracy 39

2.A1. Pooled time-series cross-sectional models of democracy bydecade 54

2.A2. OLS estimates with additional control variables or FreedomHouse data 55

2.A3. Parameter estimates from alternative estimators 573.1. Income inequality, democracy, and economic openness 74

3.A1. Interactive effect of democracy and FDI on income inequality 843.A2. Income inequality, democracy, and economic openness

(all countries) 863.A3. Effects of democracy and economic openness in DCs and

LDCs 884.1. Variables and expected effects in the simultaneous equations 1004.2. Democracy and development, 2SLS 103

4.A1. Democracy and development, 3SLS 1204.A2. Democracy and development, 2SLS-Kiviet 121

vii

viii List of Figures and Tables

5.1. Expectation of direction of effects in the simultaneousequations 139

5.2. Interactions among joint democracy, regime dissimilarity,and dyadic conflict 142

5.3. Probabilities and relative risks of MID involvement 1435.A1. Interactions among joint democracy, regime dissimilarity,

and dyadic conflict, controlling for affinity 1566.1. Trade categories and conflict expectations 1716.2. Effects of bilateral import and export in five sectors on MID

initiation 1766.A1. Equality tests on significant positive and negative effects of

sectoral trade flows 2006.A2. In-sample prediction of MID initiation 2006.A3. Effects of bilateral sectoral flows on display of force and use

of force 2017.1. Effect of level of democracy on environmental degradation 2177.2. Effect of democracy on environmental degradation 2207.3. Effect of autocracy on environmental degradation 2217.4. Effect of political regime type on environmental composites 223

8.1A–C. Summary of causal mechanisms 2498.2. Effects of democracy and trade openness on deforestation

and land degradation 2548.A1. Effects of democracy and trade openness on deforestation 2628.A2. Effects of democracy and trade openness on land

degradation 2649.1. War and CO2 emissions per capita 2759.2. War and deforestation 277

9.A1. War, cubic GDP per capita, CO2 emissions per capita, anddeforestation 289

9.A2. War and CO2 emissions per capita, omitting the laggeddependent variable 290

Acknowledgments

This book would not have been possible without the support of manyfriends, colleagues, and students at the Pennsylvania State University, Indi-ana University, and Texas A&M University. Our many friends and colleaguesin political science and economics also offered numerous useful comments,criticisms, and suggestions regarding the various parts of our project, whichwere presented in earlier forms at many professional conferences, work-shops, and seminars. We thank them all even though it is not possibleto enumerate all those who played a role in some stage of our researchproject. We give special thanks to Jim Eisenstein, Frank Baumgartner, EvanRingquist, and the reviewers for Cambridge University Press for their com-ments and suggestions directed at various parts of our book manuscript.Daehee Bak, Andreea Mihalache, Sam Sniderman, Ashley Allen Peterson,Leslie McDonald, Melanie Arnold, and Matt Warhol provided valuableresearch and editorial assistance, for which we are grateful. We also warmlythank Scott Parris at Cambridge University Press, who is the best editoran author could hope for. We deeply value his guidance throughout thedifferent phases of this project. Scott’s assistant, Adam Levine, providedexcellent editorial assistance. Himanshu Abrol effectively oversaw the pro-duction of our book, and Heather Phillips did a thorough job copyeditingthe manuscript.

Five chapters of our book draw upon our previously published journalarticles. We want to thank Blackwell Publishing, Cambridge UniversityPress, Sage Publishing, and Taylor & Francis for their permission to usematerials from those journal articles for this project.

Chapter 2 is a thoroughly revised and extended version of the follow-ing article: Li, Quan, and Rafael Reuveny. 2003. “Economic Globalizationand Democracy: An Empirical Analysis,” British Journal of Political Science33(1):29–54. Copyright c© 2003 Cambridge University Press.

ix

ONE

Introduction

Two important and complex transformations have characterized the globalpolitical–economic system since the end of World War II in 1945. Onetransformation is the rise and spread of democracy over time and acrosscountries. By the term democracy, we mean a national system of politicalgovernance based on free elections and broad political representation. Weconceptualize this concept as a continuum of political regime types, rangingfrom full autocracy at one end to full democracy at the other end.

The second transformation is the expansion and deepening of globaliza-tion. In recent decades, globalization has been a popular term but is morecomplex to define than democracy. Scholars typically employ this termto describe recent global transformations toward growing cross-nationalinterconnectedness. Although many scholars tend to focus on the intercon-nectedness features of particular interest to their own disciplines, some takea broader perspective. For example, Held et al. (2009) cast a wide net, cov-ering essentially all the transformations that have increased internationalinterconnectedness, including political-legal (e.g., growth of internationaltreaties and institutions), military (e.g., disputes, growth of armies, andweapon proliferation), communication and informational (e.g., Internet,telephone, and media), economic flows (e.g., trade and investments), knowl-edge flows (technology transfers and education), taste convergence (e.g.,consumption preferences), social contacts (e.g., migration and tourism),and environmental developments (e.g., pollution and changes in the bio-sphere), as well as the national and international income inequalities result-ing from these processes. Whereas the scope and meaning of globalizationare controversial, most scholars agree that, at a minimum, economic global-ization implies that countries are becoming more integrated into the worldeconomy, with increasing information flows among them. Greater eco-nomic integration, in turn, implies more trade, investment, and financial

1

2 Democracy and Economic Openness in an Interconnected System

flows, or rising economic openness. We also adopt this conceptualizationin our book.

Questions that pertain to democracy and globalization have attracted theattention of scholars, policymakers, and the public at large. These questionsstand at the center of our book. We believe it is best to describe our analyticalfocus by situating it within a big picture of world affairs.

THE BIG PICTURE AND OUR ANALYTICAL FOCUS

Let us assume we are charged by an intergovernmental organization, com-prising all the countries in the world, with the mission to address twoquestions. First, assuming other things do not change (e.g., public policies,state of technology, and human preferences), what will be the national tra-jectories of key forces such as income inequality, economic development,democratic governance, military conflict, and environmental degradationin the next decade? Second, assuming other things do not change, how willthese economic, political, and environmental trajectories change if we altercertain policies to promote democracy and encourage economic openness?Fundamentally these questions point to the implications of democracy andeconomic openness for the world in which we live.

Answering these questions helps to predict the evolution of the globalsystem. Although no one can absolutely predict the future, we might be ableto say something informative about it if we can explain current and pastdevelopments. Indeed, almost all the predictions in the social sciences usethe present and recent past as a baseline. The logic driving this approachis straightforward. Because we try to say something about the near future,not the very distant future, we and our immediate offspring will probablystill be around. It follows logically that the behavior of the socioeconomic–political–environmental system in the near future is connected to the stateof affairs in the present and recent past.

This book is not about forecasting the future but rather provides a soundanalytical basis for such an exercise. The ultimate goal is to say somethingnew and informative about the direction and magnitude of global transfor-mations, defined in the spirit of Held et al. (2009), that are expected for thenear future. To that effect, we must understand theoretically and empiricallythe interconnections among the relevant forces in a big picture of the world,which is what this book is about.

Going deeper into our purpose in this book, we note that both economicopenness and democracy have expanded dramatically over the past sixdecades. The number of democracies has increased considerably, spreading

Introduction 3

from Western Europe and North America to almost all the regions in theworld. Meanwhile, the world economy has witnessed an unprecedentedexpansion in the volume and patterns of international commerce, cross-border investments, and international production.1 It is reasonable toassume, which is the basic argument we make and evaluate in this book, thatdemocracy and economic openness, as two important and rapidly changingforces, must have left their marks on many other forces in the internationalsystem.

We are particularly interested in the consequences of democracy and eco-nomic openness for three important aspects of life on Earth: the economy,the polity, and the environment. One can study the economic, political, andenvironmental aspects of life on Earth at different levels of analysis, includ-ing individual, subnational, national, regional, continental, and planetary.We focus on two levels of analysis: the monadic state and the dyadic inter-state. Our choice is not arbitrary, for we submit that the nation-state remainsthe most important and influential actor in the international system despitethe rising challenges from various nonstate actors.

What should we study in terms of the economy, the polity, and the envi-ronment? The national economy covers many dimensions, such as the levelof development, income distribution, money, investment, saving, taxation,innovation, and so on, whereas the national polity may concern forces, suchas the government, the party system, civil society, social cleavages, the lawand property rights institutions, and so forth. We could also examine theinternational aspects of the national polity that involve how states inter-act with each other. In the environmental domain, there are many aspectsthat concern nation-states, beginning with how they reach agreements overthe environment, the way they treat their own environments, their abilitiesto provide resources, their levels of food sufficiency (as all types of foodultimately come from environmental sources), and so on. We recognizethat, naturally, we cannot study everything at once. Such a study is notonly overly complex but also largely intractable. We also face considerabledifficulty communicating our findings to the scholarly and policymakingcommunities, toward which we aim our book. Therefore, we concentrateon the consequences of democracy and economic openness for some keyaspects of the national economy, polity, and environment.

1 Information and communication technology (ICT) and immigration are also relevant. Butin this book we emphasize on economic globalization (i.e., flows of goods and capital). Oneshould note that the current globalization wave, which stands at the center of our book, involvesmuch smaller immigration flows than globalization of the late nineteenth and early twentiethcenturies, whereas ICT is highly correlated with GDP per capita, which is part of our analysis.


For the national economy, we focus on the national level of economicdevelopment and the national distribution of income. These two forces areimportant for any country. They inform us about the average standard ofliving in a country and the distribution of this standard across the popu-lation. The importance of economic development for human welfare goeswithout saying, and the effects exceed economics. The lack of economicdevelopment tends to beget grievances, increasing the likelihood of politi-cal instability. Spillovers into the environment also occur as poor countries,preoccupied with the immediate needs of subsistence, spend less on envi-ronmental regulation and cleanups. A skewed income distribution will mostlikely make things even worse. As the poor compare their situation to thatof the rich, their grievances grow stronger, causing the political regimeto break down in some cases. In sum, we are interested in furthering ourunderstanding of how democracy and economic openness influence incomedistribution and economic development.

For the polity, we are interested in the determinants of national demo-cratic governance and interstate military conflict. It is no exaggeration toclaim that these two phenomena are among the most salient issues ininternational relations and in political science in general. From a normativeperspective, democracy and world peace are desirable, for they appeal to thelongings of most human beings. In reality, however, democracy and worldpeace are often difficult to achieve and sustain. The scarcity of resources andthe human desire to dominate others have also been with us for millennia,leading to conflicts of interest among individuals and between nation-states.Democratic governance is an important human-engineered institutionalsolution for resolving conflicts among people and groups living withinnational territorial boundaries. In the current international system, whichoperates on principles of anarchy and state sovereignty, interstate militaryconflict is often employed as an instrument for resolving disagreementsamong nation states. We seek to find out how the forces of globalizationaffect the prospect of democratic governance and how democracy and eco-nomic openness influence interstate military conflict.

For many years, the environment was on the margin of scientific inquiryand policy analysis. As long as the economy was small in size relative tothe environment in which it resided, and as long as the activities of thepolity did not exert a large effect on the environment, one could arguablyaccept this neglect both intellectually and in public policy. Yet conditionshave changed in recent decades. Human activities have increasingly causedenvironmental degradations of various types, impacting human conditionsaround the world and the global climate system. Environmental concerns

Introduction 5

Figure 1.1. Graphical layout of the book.

are no longer marginal issues. In light of these changing conditions, webelieve it is important to expand research on the links between our forcesof interest and the environment and to address such critical questions aswhether spreading democracy, rising trade, and political violence are goodor bad for the environment.

CONCEPTUAL FRAMEWORK, ANALYTICAL APPROACH,AND CONTRIBUTIONS

The previous section described the primary elements of the big picture westudy in this book. This section presents our analytical approach, lays outthe conceptual framework within which we operate, and highlights the newthings we bring to the table. Figure 1.1 displays a graphical blueprint of ourinquiry. Although the interactions we study are in fact more complex andintricate than the figure illustrates, we believe this blueprint is a useful wayto guide and visualize the overall structure of our inquiry.

The forces of democracy and economic openness stand at the centerof the figure, indicating the analytical foci of our conceptual framework.One set of arrows flows from the center to the national economy in the upperpart of Figure 1.1, corresponding to the focus on the economy. The national


economy is represented by the national level of economic developmentand the national level of income inequality. The two-sided arrow flowingbetween democracy and the national economy represents their effects oneach other. We will study the reciprocal effects between democracy and thelevel of economic development as well as the effects between democracyand income inequality. The arrow from economic openness to the nationaleconomy indicates that we will investigate how openness influences incomeinequality and development.

The arrow from economic openness to democracy and the set of arrowsfrom the center to military conflict in the bottom right of Figure 1.1 indi-cate the part of our analysis focusing on the polity. With respect to thepolity, we conduct two analyses. First, we examine how economic opennessaffects democracy. Second, we study the effects of economic openness anddemocracy on military conflict.

The set of arrows that flows from the center to the environment inthe bottom left of Figure 1.1 represents our focus on the environment. Thearrows leading from economic openness and democracy to the environmentindicate that we study how democracy and economic openness affect theenvironment. Finally, we consider the effect of military conflict on theenvironment.

We seek to shed new light on the causal interactions indicated inFigure 1.1, but the channels depicted in Figure 1.1 do not exhaust all possi-bilities. The arrows shown in the figure represent the channels we study inthis book. They involve some of the most important and intriguing ques-tions in the fields of international political economy, international relations,comparative politics, environmental economics, and global environmentalpolitics. Thus, the reader should construe the conceptual layout in Figure1.1 as a road map for the structure of our book, not as a model per se. Eachof the primary causes and effects in Figure 1.1 is addressed in a separatechapter in the book.

We adopt a quantitative and sometimes formal approach to study thecomplex interconnections in Figure 1.1 for several reasons. The variousinterconnections are modeled separately in the respective chapters, assum-ing other things do not change, but the findings are integrated in the con-cluding chapter. In the context of Figure 1.1, each primary cause and effecttranslates into a statistical model in the respective chapter. The effect (i.e.,the phenomenon we seek to explain) is called the dependent variable orthe left-hand-side variable in a statistical model. The causal determinantsof the effect are referred to as independent or right-hand-side variables inthe statistical model. We typically refer to the primary causal factor in a

Introduction 7

chapter as the key independent variable, and we call the other relevant, butsecondary, independent variables in the model control variables. Of course,a key independent variable in one chapter often becomes a control variablein another chapter, and vice versa.

In the quantitative approach, we conduct statistical tests of the hypothe-ses to ascertain whether the effect of each independent variable on thedependent variable is statistically different from zero (where zero means noeffect), and we compute the size of the effect for each relationship that issignificantly different from zero. We employ large samples of real-world dataand, where appropriate, we also evaluate the robustness of our findings insubsamples and with different measurements and statistical techniques. Weseek robust, valid statistical inferences that answer the important questionswe and others raise.

The advantages of the quantitative approach are fully exploited. First, sys-tematic quantitative analysis allows us to study some forces in detail whilecontrolling for many other relevant factors to prevent spurious findings.Second, the quantitative analysis identifies the average cause-and-effect sta-tistical association and allows the analyst to evaluate whether a theoreticallyhypothesized relationship is statistically different from zero in the empir-ical data. This approach is important if we seek to build theories that areempirically valid and generalizable and if we intend to adopt policies thatmay actually work. Third, the use of large samples of real-world data oftenprevents subjective case-selection bias and uncovers general patterns, whichis critical if we want to understand how the world generally works. Fourth,systematic statistical analysis also enables us to evaluate the robustness ofour findings when we vary measurements, samples, and statistical tech-niques. Finally, the quantitative approach offers a valuable opportunity forus to gauge the substantive size of an effect that is statistically differentfrom zero, which provides useful information for policymakers who have tocontemplate the effectiveness of public policies and the various trade-offsassociated with competing policy objectives. Therefore, we believe that thestatistical approach contributes with rigor and clarity new insights to ouranalysis of significant issues.

Of course, the statistical modeling approach is not without limitationscommon to all studies of this type. First, we are aware that our approachmay be more technical and numerical than that in many, though not all, ofthe books on these topics, possibly posing challenges to some readers. Tomake the findings more accessible to the general readership, we write outexplicitly the estimated statistical models accompanied by detailed expla-nations, present one main set of statistical results, and describe the formal


and statistical models and their results in a way that does not require priortraining in statistics or mathematics for comprehension. We relegate all thetechnical details to the chapter appendices. Readers who are interested inthe substance of our findings only need to read the main text; those whoare interested in the technical details can obtain the relevant informationfrom the appendices. We believe that such organization of the presentationmakes the book highly readable for a wide audience.

Second, statistical findings imply probabilistic statements regarding rela-tionships of interest, and they are not intended to depict or forecast eachparticular case. Where appropriate, we highlight some real-world cases forillustration. But in essence, our analysis is statistical and reflects averagepatterns in the real world.

Finally, statistical models typically include a limited number of variables,relegating other possible influences to the error term. In other words, weshould and do make further assumptions concerning which variables toinclude in each model. At the same time, these assumptions are not arbi-trary; rather they are based on theoretical considerations and, to a largeextent, previous studies.

Our book brings at least three new things to the table, which other stud-ies of our topics have typically not attempted to achieve thus far. First, wemodel a relatively large number of forces and topics pertaining to global-ization and democracy. Previous scholarly works on these topics tend tobe compartmentalized, focusing on one or two aspects of the global sys-tem. We acknowledge and model explicitly the economic, political, andenvironmental dimensions of the global system and their interconnections.Taken together, the findings from the different chapters form a relativelymore holistic and integrative view of the inner workings of our world thanprevious research. Second, because we study seemingly diverse topics in aconceptually coherent framework and treat these topics as different compo-nents of a big picture, we are able to uncover interconnections that have thusfar received little attention. Finally, because all the analyses employ statisticalmodeling, we have one coherent methodological approach throughout thebook, which controls for various confounding forces, rigorously tests therelationships of interest, estimates the magnitude of each key relationship,and enables us to uncover general patterns.

STRUCTURE OF THE BOOK

Our book consists of three parts, and democracy and economic opennessrun through all of them. The first part examines the relationships among

Introduction 9

democracy, openness, and the national economy. The second part investi-gates the relationships between democracy and openness, on the one hand,and interstate military conflict, on the other hand. The third part of the bookstudies the relationships among openness, democracy, and conflict, on theone hand, and various indicators of environmental degradation, on theother. Given the complexity of our topics, each chapter typically employs,as an anchor, one of the two core concepts – democracy or openness –relegating the other to the secondary role.

As a preview, we discuss the gist of each chapter, focusing on the orga-nization and flow of the argument and summarizing the key findings. Thebook includes eight main analyses or chapters, each pertaining to a partof the big picture. The organizational structure of each substantive chapter(2–9) is similar. We begin each chapter by stating the research question(s)and the theoretical argument(s) concerning the relationship(s) studied. Onthe basis of the theoretical argument(s), we turn to the empirical researchthat specifies the statistical model, clarifies important research design issues,and describes the primary findings in light of the arguments posited at thebeginning of the chapter. The technical details pertaining to each chapterare in the chapter appendix, which follows the general flow of the discussionin the body of the chapter.

The first and last chapters of the book integrate the eight analyses into alarger picture, but do so with different levels of detail. The current chapterintroduces the book, focusing on the conceptual framework, analyticalfocus and approach, and structure. The concluding chapter of the booksummarizes the key findings by incorporating them into a revised andmuch more elaborate version of Figure 1.1, suggests avenues for futureresearch, emphasizes key policy implications, and evaluates the trade-offsand tensions that result from different policy objectives such as economicdevelopment versus environmental quality, economic development versusequitable distribution of income, democracy versus environmental quality,economic liberalization versus political liberalization, and democracy versusnational security.

Part I: The Democracy–Economy Nexus

Part I of the book focuses on the democracy–economy nexus. We begin ouranalyses by looking at the two key variables in our book, democracy andeconomic openness. In Chapter 2, we study the effects of economic open-ness on democracy, controlling for the important influences of economicdevelopment and income distribution. We ask whether more integration


into the world economy and the wider spread of democratic ideas andnorms across countries lead to a rise or decline of democratic governance.This question has captured the attention of policymakers and academicscholars alike, who have provided various answers and policy recommen-dations. Anecdotal evidence is typically invoked in debates, but systematicevidence is scarce. To answer our question, we examine the effects of tradeflows, foreign direct investment flows, financial investment flows, and thespread of democratic ideas on the level of democracy in a country. Ourstatistical analysis suggests that trade openness and portfolio investmentinflows reduce democracy. The effect of trade openness is constant overtime, whereas the negative effect of financial investment grows stronger.Foreign direct investment inflows promote democracy, but the effect weak-ens over time. The spread of democratic ideas is persistently conducive todemocracy over time.

In Chapter 3, we examine the effects of democracy and economic open-ness on income inequality within countries. Although the issue of incomeinequality was central in classical economics – the body of thought thatemanated from the writings of the liberal philosophers of the nineteenthcentury – it has received relatively little attention in neoclassical economics,the body of thought and knowledge in modern mainstream economics. Webelieve that the issue of income inequality deserves more attention, becausecommercial liberalism, or free market–oriented capitalism, and republicanliberalism, or democracy as a form of political governance, are not easycompanions. Whereas democracy is based on the principles of “one person,one vote” and representative government, capitalism is based on the princi-ples of laissez-faire and private enterprise. Furthermore, democracy is oftenassociated with redistributive policies (e.g., progressive taxation), but cap-italism typically rewards heterogeneous individuals with different levels ofincome. Hence, democracy may suppress income inequality, but capitalismpromotes income inequality. A skewed income distribution under capi-talism could lead to an asymmetric distribution of political power, whichcould undermine democracy and, therefore, its effect on income inequalitywithin a country.

Turning to the empirical analysis of Chapter 3, we argue that the effectsof democracy and economic openness on income inequality should beanalyzed together rather than in separate models that include one forceand exclude the other. Economic openness is measured based on nationaltrade flows, foreign direct investment inflows, and financial capital inflows;income inequality is measured by the Gini coefficient of each country inthe sample. We find that democracy and trade openness reduce income

Introduction 11

inequality, foreign direct investments increase income inequality, and finan-cial capital inflows do not affect income inequality. We also find, however,that democratic governance mitigates the inequality-increasing effect offoreign direct investments in the advanced industrial democracies.

In Chapter 4, we bring development into the story, focusing on therelationship between democracy and economic development. We open thedemocracy–development box, this time relegating economic openness to acontrol variable, and study the interactions between democracy and eco-nomic development. The democracy–development relationship not only iscentral to much academic research but also has significant policy implica-tions. A recent essay in Foreign Affairs argues that “democracies consistentlyoutperform autocracies” in economic development (Siegle et al., 2004: 57).U.S. President George W. Bush, the authors write, supports having demo-cratic governance “as qualifying criteria for countries to receive assistance . . .The U.S. Agency for International Development (USAID), still the majordevelopment actor in the U.S. government, should also offer preferentialtreatment to democracies and target its assistance to help countries under-taking democratic reforms” (2004: 67).

Normatively speaking, democratic governance, which favors equal par-ticipation, equal political rights, and equal civil liberties, is of course worthyof promotion. The argument in the Foreign Affairs essay, however, justi-fies the promotion of democracy as a means to increase economic devel-opment. Does democracy really encourage economic development? TheForeign Affairs essay presents anecdotal evidence to support its argument,but it does not take into account the possibility that the causality betweendemocracy and development could flow from development to democracy.We argue that sorting out the causal direction between democracy anddevelopment is very important. If the goal is to promote democracy but thecausation goes from development to democracy, we should not conditionthe distribution of aid to poor countries on instituting democratic reforms.In this case, we should condition aid on forces such as the quality of projects,market reforms, or investment in education.

The issue of causation between democracy and development stands atthe center of Chapter 4. We specify and estimate a statistical simultaneous-equations model, which allows for reciprocal effects between the level ofdevelopment and the level of democracy. This modeling approach rigorouslyevaluates the nature of the relationship, that is, whether democracy affectsdevelopment only, development affects democracy only, they affect eachother, or they are not related statistically. The empirical results indicate thata rise in the level of democracy reduces the level of economic development,


whereas a rise in the level of economic development promotes democracy.It is not difficult to imagine the important policy implications of thesefindings.

Part II: Bringing in Conflict

In Part II, we introduce into our book a new force – interstate militaryconflict. Interstate political relations behave like a continuum of types ofinterstate interactions, spanning from full cooperation to full-scale war. Inmeasuring the concept of conflict, some scholars consider it a continuum,whereas others focus on the discrete events that involve violence only. Inthis book, we employ the latter approach, focusing on militarized interstatedisputes, for these types of interstate interactions have the most damagingconsequences to human society.

In Chapter 5, we focus on how democracy affects conflict, controlling foreconomic openness. This concerns the famous democratic peace proposi-tion in the field of international relations. Democracies are highly unlikelyto go to war with each other, which suggests that the spread of democracyin the international system implies the spread of peace. This idea has beenaround for many years and can be traced to the writings of German philoso-pher Immanuel Kant. The modern, empirical-statistical incarnation of thisargument takes the form of a simple and powerful claim: democracies rarely,if at all, go to war with one another. This is not to say democracies may notexhibit conflict with one another at times, or that they may never fight withone another. It only means that they are less likely to fight each other thanto fight autocracies, and they are less likely to do so than two autocracies.

The empirical literature concerning the democratic peace propositionhas been arguably the largest in the field of international relations over thepast two decades. The bulk of this literature shares a similar design, employ-ing the dyadic level of analysis and estimating single-equation models. Thedependent variable is typically dichotomous, measuring the presence orabsence of a militarized interstate dispute (MID). The models typicallyinclude a measure of joint (dyadic) democracy as the key independent vari-able and control for various confounding forces. The vast majority of studiesreport that the probability of a MID between two countries declines as jointdemocracy increases. A few studies, however, reject this argument and donot find supporting evidence.

At the same time, another separate, growing body of literature in inter-national relations argues that conflict affects democracy. Although scholarsdebate the direction of the effect of conflict on democracy, they seem to

Introduction 13

agree that some type of effect exists. Conceptually, these studies form a sub-set of the large monadic country-level literature within comparative politicsthat studies the determinants of national democracy. The argument thatconflict affects democracy is intriguing because it implies the possibilitythat results of the effect of democracy on conflict may in fact have merelycaptured the effect of conflict on democracy. At any rate, the democraticpeace literature and studies that address the effect of conflict on democracyhave basically ignored the insights offered by one another. In particular, theeffect of joint democracy on conflict has been treated as unidirectional byinternational relations scholars even though it has been demonstrated else-where that conflict affects democracy. By taking the two bodies of literatureinto account, one can reasonably argue that conflict and democracy affecteach other simultaneously.

Chapter 5 evaluates this possibility empirically. We develop an innovativesimultaneous-equations model of democracy and dyadic conflict for a largenumber of countries over many years. We find that dyadic military disputereduces the level of joint democracy of two countries, and that the level ofjoint democracy reduces the probability of MIDs. Our findings vindicate themain arguments in two separate bodies of literature. But we offer importantcaveats concerning the size of the effect of democracy on conflict, which issomewhat smaller than in previous research.

In Chapter 6, we study how trade, as an important aspect of economicopenness, affects conflict. This involves another salient controversy – theeffect of trade on peace – among academics and practitioners, which hasimportant theoretical and policy implications. We contribute to this debatewith new theoretical and empirical insights.

The intellectual history of competing positions on the relationship be-tween trade and conflict is long. The idea that international trade promotespeace traces back to the great works of the eighteenth-century philosopherImmanuel Kant, French social commentator and political thinker Baronde Montesquieu, English philosopher and economist Adam Smith, andearly-nineteenth-century English journalist and member of the ParliamentNorman Angell. The antithesis that trade begets political disagreement andeven military conflict also has a long history, appearing in the writingsof such important policymakers as Vladimir Ilich Lenin, political scientistKenneth Waltz, and political economist Albert Hirschman.

Among contemporary social scientific studies, the pacifying effect oftrade has two explanations: the liberal argument and the bargaining argu-ment, whereas the conflict-generating effect of trade is often explained froma neo-Marxist view or a neomercantilist derivative of realism. Scholars have


proved formally the logical consistency of the liberal and bargaining argu-ments. However, both formal models consider aggregated or total trade,ignoring variations of trade across economic sectors and flow directions(export and import). Almost all recent empirical studies have focused onthe dyad as the unit of analysis and have employed data on total bilateraltrade flows, computed as the sum of export and import flows in a dyad.Only a few studies examine the effect of bilateral trade in various sectors oninternational political conflict, but they fail to provide a microfoundationformally, they do not focus on military conflict, they have small samples,and they fail to distinguish between imports and exports.

In this chapter, we challenge the prevalent approach and offer a theoreticalformal model to explain how export and import flows in specific economicsectors influence the decision to initiate military conflict, a question almostall previous theories have ignored. Our model indicates that one country’ssectoral imports and exports influence the conflict it initiates toward anothercountry through the expected effects the conflict will have on the prices ofthese trade flows. These effects reflect the sensitivities of sectoral importdemand and sectoral export supply to military violence and, fundamentally,whether a country expects to benefit from initiating conflict in terms of lesscost on imports or more revenue from exports.

We employ these principles to generate hypotheses regarding the effects ofexport and import in specific goods on the likelihood of military conflict. Wetest the hypotheses using a data set covering many countries for almost threedecades. The findings support our expectations on the effects of specificimports and exports broken down along sectors or goods. In particular, weare able to identify traded goods whose export and import flows increasethe likelihood of conflict, goods whose trade flows reduce the likelihood ofconflict, and goods that have no effect on conflict.

Our research in this chapter leads to new insights on a long-standingdebate in international relations, compelling scholars to rethink the logicof how trade affects conflict. The theory encompasses the liberal argumentas a special case and offers an alternative explanation to the bargaining,neo-Marxist, neomercantilist realist, and classical realist arguments. It alsohas various important policy implications.

Part III: Bringing in the Environment

Part III focuses on how the polity and the economy influence the physicaland biological environments of countries or their biospheres. Specifically, westudy the effects of democracy, economic openness, and military conflict on

Introduction 15

human activities that degrade the environment. The impact of these forceson the state of the environment has not received wide attention from polit-ical scientists, economists, or policymakers. We believe that this relativeneglect will likely change as the global environment continues to degradeand environmental concerns and issues acquire salience in the publicdiscourse.

Chapter 7 focuses on the effect of democratic governance in a countryon various aspects of its environment such as land, air, water, and vegeta-tion. In recent years, the democracy–environment relationship has receivedsome attention from theorists and is controversial. We begin the chapter bydiscussing various theoretical arguments on the effects of democracy on theenvironment. Some theorists claim that democracy reduces environmentaldegradation, whereas others argue that democracy has no effect on envi-ronmental degradation at best and may harm the environment at worst.Despite the growing importance of the issue, existing empirical evidence isrelatively scant and mixed. Extant results also do not pertain to the same typeof environmental variables. Some scholars examine government commit-ment to environmental quality in terms of signing international agreementsto protect the environment, others investigate resource scarcity and access toenvironmental amenities such as safe water or sanitation, and a third groupexplores human activities harmful to the environment. Although the focalpoints of these analyses are undoubtedly important, we believe it is essentialto study human actions that directly damage the environment because thebest way to protect the environment is to minimize the damage in the firstplace.

The empirical analysis examines five salient, specific types of human-induced degradation: carbon dioxide (CO2) emissions, nitrogen oxide (NOx)emissions, land degradation, rate of deforestation, and organic pollution inwater. We also look at an aggregate indicator of environmental degradationthat may be interpreted as a measure of sustainable development. Theresults demonstrate that a rise in democracy reduces CO2 emissions, NOx

emissions, land degradation, water pollution, and the aggregate measure ofenvironmental degradation, but it raises the rate of deforestation. The sizeof the effect varies from small to substantial, depending on the particularenvironmental attribute at issue. The findings suggest that one should nottreat democracy as a magic solution to environmental problems.

Chapter 8 examines how international trade and its interaction withdemocratic governance influence the terrestrial environment. As interna-tional trade flows have become more and more important in the globaleconomy, their effects on the environment have attracted more and more


attention. In principle, international trade may directly increase or decreaseenvironmental degradation by changing the patterns of consumption, pro-duction, and investment; affecting the production technology; and influ-encing government regulatory policies. However, scholars disagree on thenature of the overall effect of trade on the environment. In one view, tradereduces environmental degradation and promotes environmental quality.In another view, trade can harm the environment, increasing environmentaldegradation. Most of these studies do not consider the effects of both tradeand democracy, and certainly not their interactive effect on the environment.

The empirical analysis in Chapter 8 studies the effects of trade on landdegradation and the rate of deforestation within countries. These attributesare important because land quality is crucial for food production and defor-estation is important to the water cycle, land erosion, and climate change.We include both democracy and trade in the model, examine the effect oftrade conditional on democracy by including their interactive term, accountfor possible differences across continents, and distinguish between devel-oped countries (DCs) and less developed countries (LDCs). We find thata rise in trade openness reduces deforestation in autocracy but raises it indemocracy. These effects are similar for LDCs and DCs. A rise in tradeopenness reduces land degradation, but the effect is not robust and doesnot depend on regime type. A rise in democracy increases deforestation andreduces land degradation, but these effects are weaker in LDCs than in DCs.The effect of democracy on deforestation is stronger when trade opennessincreases, and the effect of democracy on land degradation does not dependon trade openness.

Chapter 9 studies the effect of warfare on the environment. Many schol-ars in international relations have examined the effect of environmentalfactors on warfare. The expectation is that as climate change progressesand environmental degradation intensifies in the world, countries vie fordepleted or degraded resources or quarrel about who should carry the costsof environmental damages, resulting in more warfare. This possibility hasreceived attention, but the effect of warfare on the environment has largelybeen ignored. In empirical studies, the effect of war on the environment hasvirtually not been studied statistically.

To understand how wars affect the environment, we separate them intothose fought at home and those fought abroad, because they are likely tohave different effects on the environment. We find that the effect of wardepends on the location of the conflict and the environmental indicatorexamined. Wars fought at home or abroad reduce carbon dioxide emissionsat home; wars fought at home speed up the deforestation in a country, but

Introduction 17

wars fought abroad slow it down. War could be a boon for the environmentsometimes, though we obviously do not think that it is a good idea to usethis knowledge to promote war.

Chapter 10 concludes our book by performing three tasks. First, we sum-marize the analyses performed in this book and take stock of our key find-ings. We flesh out the conceptual framework of Figure 1.1 with our findings,explicitly identifying the sign of each cause–effect arrow we have analyzedand found to be statistically significant. Second, we offer some caveatsregarding our analyses and propose some general directions in which onecould usefully extend our research. As we noted earlier, one book-lengthmanuscript cannot address all issues. Despite the richness of our mod-els, we believe room always exists for additional research on the complextopics we study in this book. Finally, we highlight the key policy implica-tions and trade-offs that have been revealed by our analyses. They pertainto economic, political, and environmental policies at both national andinternational levels. We believe that the issues we study and the resultswe obtain in this book should be of great interest to both academics andpractitioners.

SUMMARY AND OUTLOOK

In this chapter, we have presented our book in a nutshell by highlightingthe key questions, issues, motivations, arguments, and findings. What weoffer is a fresh look – either theoretically or empirically, or theoretically andempirically – at some of the most intriguing and important questions incontemporary social sciences. Our analysis in this book is relatively com-plex. We seek theoretical and empirical insights by reaching into a consid-erable number of issue areas, crossing multiple social science disciplines,linking different bodies of literature, engaging various academic debates,and applying formal and statistical social scientific methods. Our coher-ent conceptual, methodological, and organizational approach enables us todemonstrate the interconnectedness of different issues, debates, and disci-plines among domestic and international forces in the economy, polity, andenvironment.

We believe that our book is unique in that it combines the social scien-tific approach with a broad, interdisciplinary scope. In terms of the range oftopics covered, the one book that comes closest to ours is Global Transfor-mations by David Held and his coauthors (1999). This popular text, how-ever, is more of a survey of the state of knowledge, whereas our book is astatement of an ongoing research program that aims at expanding the extant


body of knowledge.2 In terms of the methodological approach, many booksuse the same approach as ours, but their respective substantive coveragesare much narrower, focusing on only one or some of the interconnectedprocesses we investigate here. For example, Democracy, Governance, and Eco-nomic Performance by Yi Feng (2003) studies the interrelationship betweendemocracy, regime type, inflation, and economic growth at the nation-statelevel, ignoring much of the globalization, environmental, and interstateconflict processes. Triangulating Peace: Democracy, Interdependence, andInternational Organization by Bruce Russett and John Oneal (2001) investi-gates the interrelationships among military conflict, democracy, trade, andinternational organizations at the dyadic level, ignoring income inequality,development, environment, and aspects of economic globalization otherthan trade. Growth, Trade, and Systemic Leadership by Rafael Reuveny andWilliam Thompson (2004) explores the long-term dynamics among growth,trade, and a few great powers at the system level, ignoring aspects of glob-alization other than trade, issues of democracy, income inequality, and theenvironment.

These books are important in their own right, but they fail to address theinterconnectedness of several key processes in the global system. We believethat to understand the impact of broad transformations like the spread ofdemocracy and rising economic openness, one has to cross and transcenddisciplinary boundaries. To that effect, our book confronts debates in vari-ous fields and subfields in the social sciences such as international relations,comparative politics, political economy, environmental politics, environ-mental economics, and globalization studies. It brings the insights fromcompartmentalized bodies of literature into direct dialogues and exchangesto produce a better understanding of the inner workings of the internationalsystem.

Our book is academic, scholarly, and policy-oriented in nature. We ori-ent this book toward a rather broad readership that encompasses scholars,graduate students, undergraduate students in upper-division honors pro-grams, practitioners, and policymakers in the areas of international politicaleconomy, international relations, comparative politics, international eco-nomics, sociology, political geography, human geography, environmentaleconomics, ecological economics, and environmental studies, as well asthose who have broad interdisciplinary interests in the topics we analyzehere. Our book ought to help graduate students, scholars, and policymakersget familiar with different debates, literatures, and new ideas in different

2 For another example by Held and his coauthors, see Held and McGrew (2007).

Introduction 19

fields that are interconnected in one framework. It also provides a spring-board from which one can continue to investigate both the linkages we studyand those linkages we do not study. Because we relegate all the technicalmaterials to the chapter appendices, the general readership interested in thetopics and relevant debates can be fully informed of the issues, arguments,and findings simply by reading the main text. Most important, we expectour readers to benefit from the opportunity to confront the various newconnections, solutions, and policy trade-offs one could only identify withina larger picture of the world we live in. In an age of complex transforma-tions, it is important to realize that policies and processes often interact inunexpected manners and produce unintended consequences.

We must note that we do not cover all the possible cause–effect links,nor is it desirable to do so in one book. Yet it is important to discuss themissing links at the outset so that the analytical boundaries of the book aretransparent. The reader may recall the big picture with which we startedand gradually narrowed down to Figure 1.1 that represents the essence ofour book. Naturally, the causal arrows shown in Figure 1.1 do not exhaustall possibilities. Figure 1.1 demonstrates the causal arrows we study in thisbook, not all the possible cause–effect links among those forces.

One may group the causal arrows “missing” from Figure 1.1 into twotypes. One type concerns some channels that we include but treat as sec-ondary in the book. For example, the channels from economic developmentto income distribution (the so-called Kuznets curve) and from developmentto the environment (the so-called environmental Kuznets curve) are dis-cussed theoretically and analyzed statistically, albeit not in the spotlight ofour analysis.

A second type of missing arrows involves possible linkages between forcesin Figure 1.1, but they are neither shown in the figure nor studied here. Mostnotably, income distribution, democracy, and interstate military conflictmay influence economic openness. In fact, we intentionally treat economicopenness as exogenous in this book for two reasons. We do so to focus onthe consequences of globalization, for our interest lies in the effects of eco-nomic openness on democracy, inequality, development, the environment,and conflict, and the interactions among these forces facing globalization.In addition, this keeps our analysis tractable because the sources of eco-nomic openness and globalization are controversial and involve multiplebodies of literature on trade, foreign direct investment, and financial lib-eralization. Although we do not study these and perhaps other potentiallymissing channels in this book, we discuss them again in the last chapter aspossibilities for future research.


In closing, this book has grown out of our joint research program inrecent years. The fact that several of our chapters have been published indifferent forms in high-quality journals and scrutinized by peer scholarsmultiple times speaks to one important strength of our project. But ourbook does not simply replicate our previous work word-for-word. Rather,we build on, extend, and rewrite our various projects from recent years tointegrate them together into one coherent picture and to cater to the broadreadership we have in mind. As such, this book amounts to a product whosetotal size is bigger than the sum of its individual parts. As we explained, theglobal system has experienced rising interdisciplinary interconnectednessamong its different processes and mechanisms. Our previous works, likethose of our peers, have emphasized single parts of the ensuing globaltransformations. We believe that only by putting together separate analysesand identifying their interrelationships can one understand the complexinterconnections of these global transformations.

With the introduction of the book behind us, we proceed to the nextchapter. Chapter 2 begins by observing the growth of globalization forcesin recent decades. The question we seek to answer is whether and how theseprocesses affect democracy within national boundaries. To that effect, weconceptualize globalization as the interconnectivity of a country to the worldsystem in the realms of economic and information flows. As we shall see,this interconnectivity may have significant and controversial ramificationsfor national democratic governance.

PART I

THE DEMOCRACY–ECONOMY NEXUS

TWO

Democracy and Economic Openness

INTRODUCTION

We begin our analytical journey with an analysis of the causal flow fromeconomic openness to democracy, which stands at the center of the concep-tual layout of the book in Figure 1.1. The issue we seek to investigate in thischapter is in fact much broader than the components of openness itself; itfalls under the umbrella of economic globalization. A popular interpretationof this phenomenon is the gradual turning of separate national economies,each operating in its own domain, into a set of national economies heavilyentangled with one another, affecting and being affected by the economicforces operating in other countries in the world much in the same way thatthe economy of the state of Texas is affected by and affects, for example,the economies of the states of California or Florida in the United States. Asnoted, the causal relationship of interest falls under the umbrella of the glob-alization discourse; therefore, we use the terms “economic globalization”and “economic openness” interchangeably.

Two questions stand at the core of our investigation in this chapter. Doeseconomic globalization affect the level of democracy? Is deepening inte-gration into the world economy associated with a decline or rise of demo-cratic governance? These questions have captured the imagination of policy-makers and academic scholars alike. Various answers have been provided,and policy recommendations have been made. Anecdotal evidence is typ-ically invoked in debates, but systematic evidence is scarce. This chapterseeks to fill this empirical lacuna.

The notions of globalization and democracy are widely discussed in theliterature. Most scholars agree that, at the minimum, globalization impliesthat countries are becoming more integrated into the world economy, withincreasing information flows among them (e.g., Held et al., 1999, 2009).

23


Greater economic integration, in turn, implies more trade and financialopenness. A rise in information flow implies, arguably, the spread of ideasand cultural convergence across countries. As we discussed in the previouschapter, most scholars also agree that democracy implies a national politicalregime based on free elections and broad political representation, and this isthe conceptualization of democracy that we continue to use in this chapterand, for that matter, throughout the book.

Before we delve further into this chapter, some clarifications are in order.Although the historical developments of both globalization and democracyhave been long and cyclical, we will not take a long-term historical slantin this chapter, primarily due to the lack of economic performance andopenness data over a long period of time.1 We also note that throughoutthis book, democracy is conceptualized as a level variable, a continuumof regime types, from fully autocratic at one end to fully democratic at theother. Hence, we use the terms “democracy” and “level of democracy” inter-changeably.

Returning to our questions for this chapter, the theoretical literaturepresents conflicting positions on the effect of globalization on the level ofdemocracy: one position claims a positive effect, a second position assertsa negative effect, and a third position argues that globalization does notnecessarily affect democracy.

The position that globalization promotes democracy is entrenched in theviews of recent American presidents, and it has been used as a rationale forpromoting international economic liberalization. For example, U.S. Presi-dent William Clinton (1996) argues that “commerce helps make the worldsafe for democracy” and U.S. President George W. Bush (2001) claimsthat “societies that open to commerce across their borders will open todemocracy within their borders.” This view garners some support in styl-ized observations. Lin and Nugent (1995) argue that Korea, Taiwan, andChile exemplify the positive influence exerted by economic openness on thedemocratic transitions of their authoritarian regimes. As another example,the opening up of Indonesia to the world economy, despite causing somefinancial difficulties in the late 1990s, seems to have facilitated the evolutionof a more competitive political system in that country.

In contrast, history also provides some opposing cases in which economicopenness had little or even a negative effect on democracy. For instance,as noted by Flandreau (2007), the liberalization of trade between France

1 On the historical development of globalization, see, e.g., Wallerstein (1974), Cameron (1997),Held et al. (1999), and Held and McGrew (2007). On the historical development of democracy,see, e.g., Huntington (1991), Potter et al. (1997), and Diamond (1999).

Democracy and Economic Openness 25

and Britain in the 1860s was not associated with a change in democracy inFrance, which was then led by an autocratic regime. The contraction of tradeliberalization in the late nineteenth century did not affect the levels of Britishor French democracy. The opening up of Russia to the world economy inthe late nineteenth century, which involved deepening economic relationswith democracies, was associated with an increase in its level of autocracy.Democratic governance in the United States evolved largely as an internalmatter, not affected much by its increasing contacts with the world in thenineteenth century. Finally, recent increases in the economic openness ofRussia and China seem to have led to a decline in the level of democracyin Russia since the late 1990s and have had little impact on the level ofdemocracy in China.

Of course, the anecdotal evidence does not demonstrate generalizablepatterns across countries. Moreover, historical cases may not be applicableto the contemporary era. After all, one may argue that the globalization ofour times is much more all-embracing than the one that intensified duringthe latter half of the nineteenth century. This argument in and of itself canmake a difference in terms of the direction, not only the size, of the effectof increasing connectivity to the world economy on the level of democracywithin a country.

Progress in explaining the effect of globalization on democracy requiresan empirical evaluation of the three competing theoretical positions. Weseek to detect generalizable empirical patterns in the effect of globalizationon democracy. As noted, democracy is conceptualized as a continuum.Four national aspects of globalization are examined: trade openness, foreigndirect investment (FDI) inflows, portfolio (financial) investment inflows,and the spread of democratic ideas. Our analysis in this chapter covers127 countries from 1970 to 1996 in a pooled time-series cross-sectionalstatistical model.

Our primary findings can be summarized as follows. Trade opennessand portfolio investment inflows negatively affect democracy. The effect oftrade openness is constant over time, whereas the negative effect of portfolioinvestment inflows strengthens democracy. FDI inflows positively affectdemocracy, but the effect weakens over time. The spread of democratic ideaspromotes democracy persistently over time.

The chapter is organized as follows. The next section briefly reviews thestudies on the determinants of democracy. The section that follows discussesthe effects of globalization on democracy, which, as noted and as we shallfurther demonstrate, are highly controversial. This discussion is followed bya section that describes both research design and findings from the statistical


analysis. Understanding this section does not require statistical expertise,for the materials of more technical nature are delegated to the chapterappendix. The last section of this chapter summarizes our main findingsand discusses their implications.

THE LITERATURE ON DETERMINANTS OF DEMOCRACY

This section reviews the literature on the determinants of democracy. Ourreview is not meant to be exhaustive but rather seeks to illustrate howour analysis fits into the larger picture. We categorize this voluminousliterature into three groups. One group consists of detailed case studies. Asecond group includes statistical analyses. These two groups, by and large,focus on domestic political and economic variables and pay relatively littleattention to international factors. A third group (discussed in the nextsection) includes largely theoretical studies of the effects of globalization ondemocracy.

In their synthesis of a number of case studies, O’Donnell, Schmitter,and Whitehead (1986) conclude that the effect of international factors ondemocracy is indirect and marginal. Challenging this conclusion, Pridham(1994) labels international factors as the “forgotten dimensions in the studyof domestic transition,” and Schmitter (1996: 27) argues “perhaps, it is timeto reconsider the impact of international context upon regime change.”Attempting to bridge this gap among case studies, Whitehead (1996) andDrake (1998) argue that international factors such as the diffusion of demo-cratic ideas and global markets are important determinants of democracy.

Among the statistical studies, the dependent variable is democracy,but the independent variables vary. As discussed in the previous chapter,one group of studies argues that economic development positively affectsdemocracy.2 A second group debates the effects of economic crisis, such asrecessions and high inflation (Helliwell, 1994; O’Donnell, 1973; Haggardand Kaufman, 1995; Gasiorowski, 1995). A third group expects positiveinfluences by Christianity and negative effects of social cleavages (Hunting-ton, 1984; Muller, 1988; Gasiorowski, 1995). A fourth group examines theeffects of institutions such as constitutional arrangements, nonfragmentedparty systems, and parliamentary versus presidential systems (e.g., Lijphart,1977; Mainwaring, 1993; Linz, 1994). Finally, some statistical studies

2 See, e.g., Lipset (1959), Dahl (1989), Huntington (1991), Burkhart and Lewis-Beck (1994),Muller (1995), Londregan and Poole (1996), and Feng and Zak (1999).


consider external factors but not globalization-related variables, such ascore-periphery status and diffusion.3

The strengths and weaknesses of case studies and statistical analyses ofdemocracy are debated (e.g., see Przeworski and Limongi, 1997). Briefly,statistical studies focus on the macro conditions that facilitate or hinderdemocracy. Although they can test general theoretical claims and controlfor competing forces, they are less able to explain the micro processes thataffect democracy. Case studies identify detailed microlevel influences, butthey are less able to test general theories or to assess the relative strength ofcausal factors.

Our study in this chapter takes a macro approach. However, we believethat the macro and micro approaches to our question provide complemen-tary insights into the determinants of democracy, much as macroeconomicand microeconomic analyses provide complementary insights into the oper-ation of the economy. Our inquiry adopts the spirit of the important quali-tative comparative study by Rueschemeyer et al. (1992). Like their study, wealso study many countries and emphasize the importance of transnationalpower relations, particularly economic and information flows. In the nextsection, we discuss how these external forces affect democracy.

THE GLOBALIZATION–DEMOCRACY CONTROVERSY

The literature on the effects of globalization on democracy is quite large andmostly theoretical. It posits three competing theoretical positions: global-ization promotes democracy, globalization obstructs democracy, and glob-alization has no systematic effect on democracy. To streamline the presen-tation of these positions, Tables 2.1, 2.2, and 2.3 summarize the argumentsfrom studies supporting each of the three theoretical positions, respectively.

Globalization Promotes Democracy

The first proposition listed in Table 2.1 is that globalization promotes demo-cracy by encouraging economic development. The notion that free mar-kets facilitate democracy can be traced back to the late eighteenth century.

3 For example, see Bollen (1983) and Burkhart and Lewis-Beck (1994), who study position inthe world system. Starr (1991) and Przeworksi et al. (1996) study diffusion of democratic ideas.Gasiorowski (1995) focuses on the effect of economic crisis on democratization but also includestrade openness as a control variable.


Table 2.1. Globalization promotes democracy

Number Argument Discussed in

1 Globalization promotes economicdevelopment.

Schumpeter (1950), Held (1992),Platner (1993), Weitzman (1993),Bhagwati (1994), Lipset (1994),Muller (1995), Im (1996)

2 Globalization increases thedemand of internationalbusiness for democracy.

Kant (1795), Bhagwati (1994),Schmitter (1996), Oneal andRussett (1997, 1999a)

3 Globalization reduces theincentives of the authoritarianleaders to cling to power.

Rueschemeyer and Evans (1985),Diamond (1995), Drake (1998)

4 Globalization reduces informationcosts, increasing contact withother democracies and makingthe prodemocracy internationalnongovernmental organizationsmore effective.

Van Hanen (1990), Brunn andLeinback (1991), Diamond(1992a), Schmitter (1996), Keckand Sikkink (1998), Kummell(1998), Boli and Thomas (1999),Risse and Sikkink (1999)

5 Globalization pushes theauthoritarian states todecentralize power.

Self (1993), Sheth (1995), Roberts(1996)

6 Globalization promotes domesticinstitutions that supportdemocracy.

Roberts (1996), Fruhling (1998),Keck and Sikkink (1998), Stark(1998), Boli and Thomas (1999),Risse and Sikkink (1999)

7 Globalization intensifies thediffusion of democratic ideas.

Kant (1795), Whitehead (1986,1996), Huntington (1991), Starr(1991), Przeworski et al. (1996)

In this view, globalization promotes economic growth, increases the sizeof the middle class, promotes education, and reduces income inequality,all of which foster democracy. Trade, FDI, and financial capital flows aresaid to allocate resources to their most efficient use; democracy is said toallocate political power to its most efficient use. The outcome in both casesrepresents the free will of individuals.4

According to a second view, globalization increases the demand of inter-national business for democracy. Business prosperity requires peace andpolitical stability. Because democracies rarely, if ever, fight each other, com-mercial interests pursue democracy to secure peace and stability. As the

4 See Schumpeter (1942), Held (1992), Platner (1993), Weitzman (1993), Bhagwati (1994), Lipset(1994), Muller (1995), and Im (1996).


economic links between states develop, commercial interests strengthen,and the demand for democracy rises. Authoritarian countries that open theireconomies face greater pressures from international business for politicalliberalization. Furthermore, in economies open to the transnational flowsof commodities and capital, the government and the central bank have to bemore transparent in their procedures and policymaking to attract and main-tain international business. The increased transparency, in turn, implies lesspower for the autocratic regime, which can facilitate democratization.5

A third argument is that globalization reduces the incentives of theauthoritarian leaders to cling to power. Because the state can extract rentsfrom society, losing office implies the forfeit of these rents. Hence, auto-cratic rulers cling to power, resisting democracy. However, globalizationreduces the capacity of the state to extract rents from society by increasingcompetition and weakening the effectiveness of economic policies. It fol-lows that leaders of autocracies whose economies are more open are lesslikely to resist democratization. To the extent that the autocrats believe thatproviding social welfare for the losers of globalization reduces the latter’sincentive to contest their grip on power, the autocratic leaders should bewilling to give their citizens more political rights and civil liberties.6

A fourth view argues that globalization reduces information costs,increasing contacts with other democracies and making the prodemoc-racy international nongovernmental organizations (INGOs) more effec-tive. A prosperous democracy requires well-informed actors. With increas-ing globalization, citizens have access to more information, supplied notjust by their own governments. Economic openness enables the establisheddemocracies, aided by their developed media, to export their values to auto-cracies. Authoritarian regimes now have less control over information. Moreexposure to the media also strengthens the effectiveness of transnationaladvocacy networks and the INGOs, helping them protect prodemocracyforces in authoritarian regimes and promote democracy.7

According to a fifth view, globalization pushes the authoritarian stateto decentralize power. As globalization deepens, states relinquish controlover economic and social progress to the market, which is “inherentlydemocratic” – as if millions of economic agents cast their “votes” voluntarily.

5 See Kant (1795), Bhagwati (1994), Schmitter (1996), Oneal and Russett (1997, 1999a), andDailami (2000).

6 See Rueschemeyer and Evans (1985), Diamond (1995), and Drake (1998). For the effect of socialwelfare spending and globalization, see Rudra (2005).

7 See Van Hanen (1990), Brunn and Leinback (1991), Diamond (1992a), Schmitter (1996), Keckand Sikkink (1998), Kummell (1998), Boli and Thomas (1999), and Risse and Sikkink (1999).


The weakened state also implies the entrance of grassroots groups into thepolitical arena (e.g., business and professional associations, labor unions).Citizens become more involved in the day-to-day governance of the country,facilitating democracy.8

A sixth view argues that globalization strengthens domestic institutionsthat support democracy. Because the efficient operation of the marketrequires an enforceable system of property rights and impartial courts,economic openness compels the popularization of norms respecting therule of law and civil and human rights. The increased involvement of inter-national business and INGOs in the domestic economy further promotesthe transparency and accountability of domestic institutions and reducesstate intervention, all of which are said to facilitate democracy.9

According to a seventh view, economic globalization intensifies the dif-fusion of democratic ideas across borders. Transnational movements ofcommodities, services, and factors of production facilitate and encouragethe exchange of norms and ideas, increasing the diversity of political viewsin a country and indirectly promoting political competition. Scholars arguethat the more democracies that surround a certain nondemocratic coun-try, the more likely it is that this country will become democratic. Becausegreater economic openness is associated with more information flow andtransnational contacts, the diffusion of democratic ideas across borders isexpected to intensify with growing economic integration. The diffusion ofdemocratic ideas and norms will affect the people in the nondemocraticcountries, motivating them to demand democratic reforms.10

Globalization Obstructs Democracy

The first argument in Table 2.2 is that globalization reduces state policyautonomy and brings about public policies that please foreign investorsinstead of the common people. Globalization increases financial capitalmobility across countries and facilitates relocation of the means of produc-tion, which, in turn, reduces the ability of states to implement domesticallyoriented economic policies. Another consequence is that governments nowtry to compete for foreign capital and design their policies to please global

8 See Self (1993), Sheth (1995), and Roberts (1996).9 See Roberts (1996), Fruhling (1998), Keck and Sikkink (1998), Stark (1998), Boli and Thomas

(1999), and Risse and Sikkink (1999).10 Kant (1795), Whitehead (1986, 1996), Huntington (1991), Starr (1991), Przeworski et al. (1996),

and Dailami (2000).


Table 2.2. Globalization obstructs democracy

Number Argument Discussed in

1 Globalization reduces state policyautonomy and brings aboutpublic policies that pleaseforeign investors instead of thecommon people.

Lindblom (1977), Held (1991),Diamond (1995), Gill (1995), Jones(1995), Gray (1996), Schmitter(1996), Cox (1997), Cammack (1998)

2 Globalization produces moredomestic losers than winners, atleast in the short run, and it alsodiminishes the ability of thestate to compensate the losersfinancially.

Drucker (1994), Muller (1995), Beck(1996), Bryan and Farrel (1996), Cox(1996), Moran (1996), Marquand(1997), Martin and Schumann(1997), Rodrik (1997), Longworth(1998)

3 Globalization enables the fastmovement of money betweencountries, resulting in frequentbalance of payment crises andunstable domestic economicperformance.

Im (1987), MacDonald (1991),Diamond (1992a, 1999), O’Donnell(1994), Trent (1994), Haggard andKaufman (1995), Cammack (1998)

4 Globalization deepens ethnic andclass cleavages and diminishesthe national-cultural basis ofdemocracy.

Robertson (1992), Dahl (1994),Im (1996)

5 Globalization enables the stateand multinational corporationsto control and manipulateinformation supplied to thepublic.

Gill (1995), Im (1996), Martin andSchumann (1997)

6 Globalization degrades the conceptof citizenship, an importantprerequisite for a functioningand stable democracy.

O’Donnell (1993), Whitehead (1993),Im (1996), Sassen (1996), Cox (1997),Boron (1998)

7 Globalization widens the economicgap between the North and theSouth.

Wallerstein (1974), Bollen (1983),Tarkowski (1989), Przeworski (1991),Gill (1995), Amin (1996), Cox(1996), Im (1996), Kummell (1998)

investors and firms, who may not act in the best interest of, or be heldaccountable to, the voters. It follows that the level of democracy declines.11

According to a second argument, globalization produces more domesticlosers than winners, at least in the short run, and it also diminishes the

11 See Lindblom (1977); Held (1991), Diamond (1995), Gill (1995), Jones (1995), Gray (1996),Cox (1996), Schmitter (1996), and Cammack (1998).


ability of the state to compensate the losers financially. Domestic producersthat cannot compete internationally lose from more economic openness.Trade and financial liberalization also limit the ability of governments todeploy regulation and risk-sharing policies that can redistribute incomefrom winners to losers. Governments that want to compensate the vic-tims of economic openness confront the phenomenon of footloose capital,which in turn shrinks the tax base and penalizes deficit spending. Conse-quently, governments reduce the scope of welfare programs, and the poorincreasingly feel this pinch. The result is rising income inequality and classpolarization, which serve to weaken democracy.12

A third view argues that globalization enables the fast movement ofmoney between countries, resulting in frequent balance-of-payment crisesand unstable domestic economic performance. In such situations, the lessdeveloped countries (LDCs) are compelled to accept economic reformsimposed by the developed countries (DCs) and international organizations,and those reforms typically involve austerity measures. Economic criseshurt the poor more than the rich, raising domestic income inequality.Social unrest then rises, and support for radical opposition groups grows.In an attempt to reassert power, weak democracies resort to authoritarianmeasures. The electoral technicalities are seemingly retained, but civil rightsand the inputs from the elected legislators are increasingly ignored.13

According to a fourth argument, globalization deepens ethnic and classcleavages and diminishes the national-cultural basis of democracy. Thelosers from economic openness tend to seek a united identity based onethnicity or religion. The winners may promote discriminatory measuresto maintain their edge over the losers. Globalization also induces labormigration across countries. The old comers typically attempt to restrictor eliminate the participation of the immigrants in the political system toreduce their competitiveness. All these actions intensify social cleavages andundermine the consolidation of democracy.14

A fifth argument is that globalization enables the state and the multina-tional corporations (MNCs) to control and manipulate information sup-plied to the public. With the help of new information technologies, the stateand the MNCs feed to the public processed information that represents

12 See Drucker (1994), Muller (1995), Beck (1996), Cox (1996), Moran (1996), Marquand (1997),Martin and Schumann (1997), Rodrik (1997), Longworth (1998), and Dailami (2000).

13 See Im (1987), Diamond (1992a, 1999), MacDonald (1991), O’Donnell (1994), Trent (1994),Haggard and Kaufman (1995), and Cammack (1998).

14 See Robertson (1992), Dahl (1994), and Im (1996).


only certain views. They are also better able to tightly monitor the people’sinformation sources. The result is a disconnection between the way govern-ment decisions are made and the way the public thinks they are made. Thegovernment becomes less transparent and less accountable to the people,and the level of democracy declines.15

According to a sixth argument, globalization degrades the concept ofcitizenship, which is an important prerequisite for a functioning and stabledemocracy. The global market transforms the individual into a common“Homo economicus” who cares more about profits than public and civiccommitments. Individuals pursue their own economic interests, disregard-ing whether governments practice democratic decision making. Since thepublic has less interest in the conduct and content of public policy, demo-cracy gradually weakens.16

A seventh argument is that globalization widens the economic gapbetween the North and the South. Globalization involves mostly DCs, drain-ing capital, technology, and skilled labor from LDCs. With the gap in wealthrising, social unrest increases in LDCs, their elites cling to power, and theirgovernments become less democratic. The dependency story evolves alongsimilar lines. In a world composed of a rich core and a poor periphery, thecore dominates the periphery. The elite in the periphery unite with the elitein the core to exploit the masses in the periphery. MNCs relocate to LDCsto enjoy both lower wages and more lax labor and environmental standardsand then repatriate profits to the core. The penetration by MNCs distortsthe economies in LDCs and sways domestic politics in their own favor, allof which obstruct democracy.17

Globalization Does Not Necessarily Affect Democracy

Table 2.3 shows three arguments that question whether globalization hasany general effect on democracy. The first argument holds that the extentof globalization is greatly exaggerated. The world economy is not as inte-grated as is commonly believed. Most international trade takes place withingeographical regions, MNCs typically have a home-country bias, and mostFDI concentrate in a few advanced countries. Because LDCs generally donot participate in the global economy, the effect of economic openness on

15 See Gill (1995), Im (1996), and Martin and Schumann (1997).16 O’Donnell (1993), Im (1996), Sassen (1996), Whitehead (1996), Cox (1997), and Boron (1998).17 See Wallerstein (1974), Tarkowski (1989), Przeworski (1991), Gill (1995), Amin (1996), Cox

(1995), Im (1996), and Kummell (1998).


Table 2.3. Globalization does not necessarily affect democracy

Number Arguments Discussed in

1 The extent of globalization isexaggerated.

Scharpf (1991), Jones (1995), Hirstand Thompson (1996), Wade(1996), Hirst (1997)

2 Globalization does not renderthe welfare state powerless.

Vernon (1971), Kurzer (1993),Frieden and Rogowski (1996),Garrett (1999)

3 Effects of globalization oncountries vary.

Haggard and Kaufman (1995),Frieden and Rogowski (1996),Milner and Keohane (1996),Armijo (1998), Longworth (1998)

their democracy should not be large to begin with. Because DCs are alreadystable democracies, globalization does not affect their levels of democracy.18

A second argument asserts that globalization does not necessarily renderthe welfare state powerless. Increased national economic openness origi-nates from the deliberate choices of states. Governments still exert consid-erable control over their own economies. Moreover, the modern welfarestate is still effective because it provides important collective goods under-supplied by markets (e.g., social stability, property rights, infrastructure)and compensates the losers of economic openness. By implication, one canargue that the level of democracy does not necessarily have to decline witheconomic openness.19

According to a third view, the effects of globalization vary across coun-tries and depend on government policies, a country’s location in the globalpecking order, the domestic political institutions, the identity of the domes-tic winners and losers, whether economic sectors are privatized or not, andthe current level of democracy. For example, though globalization-inducedeconomic crises may force the authoritarian regime to exit and be replacedby democracy, these crises, if managed effectively, may instead increase thepublic support for the authoritarian leader. Hence, the effects of globaliza-tion on democracy may not be uniform.20

The arguments summarized in Tables 2.1, 2.2, and 2.3 all appear to exhibitface validity in terms of logic consistency, and it is not possible to evaluate

18 See Scharpf (1991), Jones (1995), Hirst and Thompson (1996), Wade (1996), and Hirst (1997).19 See Frieden and Rogowski (1996) and Garrett (1999).20 See Haggard and Kaufman (1995), Frieden and Rogowski (1996), Milner and Keohane (1996),

Armijo (1998), and Longworth (1998).


them based on pure theoretical considerations. In other words, becausethese arguments reach conflicting conclusions, they need to be evaluatedempirically, to which we turn next.

EMPIRICAL MODEL AND ANALYSIS

This section first presents our statistical model for the empirical analysisand then discusses several research design issues. Next, the section presentsthe key results of the empirical analysis. The discussion in this section isself-contained and does not require any specific statistical expertise. Inter-ested readers seeking further details should consult the chapter’s appendix,which discusses various technical issues related to the statistical design andimplementation of the regression-based model presented here. The chapterappendix follows the same structure of presentation as in this section.

Empirical Model

To assess the competing claims about the effects of globalization on democ-racy, we specify and estimate the following statistical model of democracy.We denote variables with uppercase letters and their coefficients with Greeknotations. Each coefficient indicates the effect of the independent variableon the dependent variable – the phenomenon we seek to explain. The nota-tion εt denotes the random error that is not explained by the statisticalmodel. The variable subscripts t and t −1 indicate the time period of thevariable, where t represents the current period and t−1 the previous timeperiod (a lagged variable). To simplify the presentation, we refer to thevariables without their time subscripts t or t −1. The model specification isbased on the preceding theoretical discussions and the literature on democ-ratization. It provides a structure for us to guide the statistically uninitiatedreaders through the empirical exercise.

democracyt = �0 + �1tradet−1 + �2fdit−1 + �3portfoliot−1

+ �4diffusiont−1 + �5inflationt−1 + �6growtht−1

+ �7developmentt−1 + �8prior democracyt−1

+ �9year + �10year∗fdit−1 + �11year∗portfoliot−1

+ �12year∗inflationt−1+�13year∗developmentt−1+εt.

(2.1)

In the model, democracy denotes the left-hand-side variable or thedependent variable (the phenomenon we explain), whose determinants are


specified on the right-hand side of the equation and dubbed the indepen-dent variables. The democracy variable represents the level of democracyin a country in a given year. The level of democracy can be measured quan-titatively over a continuum of national attributes ranging from completeautocracy or dictatorship at one end of the continuum to full democracyat the other end of the continuum. The measure construction, described inthe appendix, includes both democratic and autocratic characteristics, anapproach widely used in the literature.

The independent variables divide into two types. One type involves theglobalization-related variables, which stand at the center of our study. Asecond type of independent variable involves other determinants of democ-racy that have been identified in various other studies. We call the secondtype of factor control variables. Here we discuss briefly the measurement ofthese variables.

We start with how to measure the notion of globalization. As a multidi-mensional concept, the integration of states into the world economy needsto be measured from multiple indicators. In our empirical model, glob-alization is measured by four indicators: trade, foreign direct investment,portfolio investment, and diffusion of democratic ideas across countries.The trade variable is a measure of trade openness – the extent to whicha country depends on trade with other countries. The fdi variable is ameasure of the openness of a country to foreign production capital flowinginto a country, which refers to investments that enter a country for the pur-pose of building plants or acquiring management positions in plants. Theportfolio variable is a measure of the openness of a country to financialcapital flowing into a country, meaning portfolio investments that enter forthe purpose of buying financial securities. The diffusion variable repre-sents the diffusion of democratic ideas into a country, which is measuredhere by the extent to which a country is affected by democratic ideolo-gies emanating from neighboring countries. Our study is one of the firstattempts to assess the effects of globalization on democracy by using mul-tiple indicators in one empirical statistical analysis. Because globalizationproduces competing effects on democracy through many mechanisms, theseglobalization indicators help estimate the short-run net effects.

Obviously, democracy does not only depend on a country’s economicopenness. Control variables must be included to avoid spurious statisticalinferences. Although existing studies of the causes of democracy differ whenidentifying these additional determinants, many studies find that domes-tic economic factors such as economic development, economic growth,


and inflation influence democracy.21 The inflation variable measures theyearly rise in price level in a country. The development variable measuresthe level of economic development in a country and is based on the grossdomestic product (GDP) per capita, which is the value of final goods andservices produced in a country divided by the country’s population. Thegrowth variable is the yearly growth rate of GDP, which captures thevitality and improvement of economic performance over time in a country.

The prior democracy variable denotes the level of democracy in acountry in the previous year. This variable controls for the fact that polit-ical regimes tend to have inertia and change slowing over time. Theoreti-cally, domestic structural variables (e.g., Protestant population, institutionalqualities of the regime, and party fragmentation) and the attributes of theinternational system may also affect democracy. These additional factorsare relatively stable over time, implying that democracy exhibits inertiaover time or that the past level of democracy affects the present.

The year variable is a yearly counter that serves two functions. By itself,year tests whether the level of democracy has a linear trend, and it alsointeracts with other variables that have time-varying effects over democracy.It is possible that the effects of economic globalization on the level ofdemocracy at home may change over time. For example, the scope andtypes of FDI, financial capital investments, and information flows took on arelatively more prominent role in the 1990s than in the 1970s. This changeis not unique to the globalization variables. Gasiorowski, for example, findsthat the effect of inflation on the likelihood of democratization changes overtime. Ignoring this issue altogether may lead to model misspecification. Inour exploratory by-decade analysis of which variables have time-varyingeffects on democracy (presented in Table 2.A1 of the appendix), we findthat the effects of development, fdi, portfolio, and inflation on thelevel of democracy in a country change their signs over time. Hence, ourstatistical model includes interaction terms between year and these fourvariables, respectively.

Research Design Issues

To implement our statistical model, several design issues must be raisedthat require some clarification. Although we address the technical details

21 For the three economic variables, see Bollen (1979), Muller (1988), Lipset et al. (1993), Burkhartand Lewis-Beck (1994), Helliwell (1994), Muller and Seligson (1994),Gasiorowski (1995), Muller(1995), and Feng (1997).


of these issues in the appendix, here we provide a brief overview for thegeneral reader. First, the variable εt is the so-called error term in the statis-tical model, that is, the part of the variations in democracy the model doesnot explain. For statistical inferences to be valid, the error term needs tosatisfy some assumptions. If the error term does not have several desirableproperties, the accuracy of the results and our ability to make correct infer-ences is compromised. We address the related assumption violations usingappropriate econometric techniques.

Second, the relationship between democracy and other variables on theright-hand side may be reciprocal or simultaneous. For example, whereaseconomic integration variables may affect democracy, they themselves maybe affected by democracy. A higher level of democracy may imply lowerpolitical risks and more secure property rights, attracting trade and invest-ments (Olson, 1993). Ignoring this possibility can produce incorrect results.We deal with this risk by using the first lag of the right-hand-side variables,as denoted by the subscript t −1 in the model.

The third issue concerns sample selection. We take the approach of large-N studies, employing a sample for many countries, where each country isrepresented by yearly data. More specifically, our sample includes 127 coun-tries from 1970 to 1996. The unit of analysis is the country year. Such a sam-ple design allows us to assess the effect of globalization on democracy overtime and across countries and to generalize the inferred results across cases.

A fourth issue concerns a complication in sample selection, which hasto do with a possible distinction between DCs and LDCs. Most DCs havealready achieved high levels of democracy at the beginning of the sampleperiod. On the other hand, many LDCs experienced large variations in theirlevels of democracy from 1970 to 1996. Although the effects of globalizationon the level of democracy may be independent of the development level ofa country, it is equally possible that LDCs exhibit patterns in the effectsof globalization on democracy distinct from those of DCs. In fact, severalauthors argue that the adverse effects of globalization on democracy may bestronger in DCs than in LDCs. We investigate this possibility by analyzingthe effects of globalization for all countries and then for a sample thatexcludes DCs as measured by membership in the Organization for EconomicCooperation and Development (OECD).

Empirical Findings

We present the main results of our analysis in Table 2.4 for two samples.One sample includes all the countries for which we have data, whereas the


Table 2.4. Effects of economic globalization on democracy

All countries Non-OECD countries

TRADE −0.0029∗∗∗ −0.0032∗∗∗

(0.0010) (0.0011)FDI 5.1906∗ 5.8359∗

(3.3016) (3.7712)PORTFOLIO 4.7240∗∗∗ 5.3892∗∗

(1.8888) (2.4406)DIFFUSION 0.2437∗∗∗ 0.2922∗∗∗

(0.1058) (0.1219)INFLATION 0.0599∗∗ 0.0537∗

(0.0350) (0.0344)GROWTH 0.0068 0.0069

(0.0084) (0.0087)DEVELOPMENT 34.2597∗∗∗ 36.7637∗∗∗

(11.5024) (16.8662)PRIOR DEMOCRACY 0.9269∗∗∗ 0.9242∗∗∗

(0.0109) (0.0115)YEAR 0.1514∗∗∗ 0.1593∗∗∗

(0.0505) (0.0670)YEAR∗FDI −0.0026∗ −0.0029∗

(0.0017) (0.0019)YEAR∗PORTFOLIO −0.0024∗∗∗ −0.0027∗∗

(0.0009) (0.0012)YEAR∗INFLATION −0.00003∗∗ −0.000027∗

(0.000017) (0.000017)YEAR∗DEVELOPMENT −0.0172∗∗∗ −0.0184∗∗

(0.0058) (0.0085)Constant −301.8202∗∗∗ −317.8469∗∗∗

(100.2552) (132.8935)Observations 2021 1640Adjusted R2 0.93 0.90

Note: Standard errors in parentheses. ∗significant at 10%; ∗∗ significant at 5%; ∗∗∗ significantat 1%.

second sample includes only LDCs. The findings for the so-called controlvariables are consistent with those in previous studies in the democratizationliterature. This general pattern provides support for our statistical modeling.To sharpen our presentation, we delegate the discussion of control variablesto the chapter appendix and focus on the findings of the globalizationvariables instead.

Beginning with the all-countries sample, as shown in the second columnof the table, the effect of a rise in trade (the level of trade openness of


a country) on democracy (the level of democracy in a country) is foundto be negative. The finding for trade is consistent with the well-knowntendency of trade to generate economic winners and losers in the shortrun. Although the overall gain from trade outweighs the cost of trade, theprocess by which the winners compensate the losers is political and slow, andthe endogenous domestic economic adjustment process (e.g., movement oflabor from losing to winning sectors) is also relatively slow. As such, tradecan generate changes in the distribution of income in the short run, givingrise to the type of processes described in the section on the obstructingeffects of globalization on democracy.

The effect of a rise in fdi (the openness of a country to FDI inflow) on thelevel of democracy in that country is positive and statistically significant.However, this positive effect declines over time, as indicated by the negativeand statistically significant coefficient of the interaction term year∗fdi.More specifically, the total effect of a rise in fdi on the level of democracy,which is given by (5.1906 − 0.0026∗year), is 0.055 for 1975, 0.029 for 1985,and 0.001 for 1996. Hence, the total effect of a rise in FDI on the level ofdemocracy in a country is positive, but its size is declining over time.

These results suggest that a rise in FDI exerts both positive and negativeeffects on the level of democracy in a country, as discussed earlier, but thenegative effect grows over time. On the positive side, FDI can promoteeconomic growth and technological progress. On the negative side, FDI canraise domestic income inequality. The tendency of FDI to quickly relocateto cheaper production places in other nations and the tendency of MNCsto interfere in host countries’ domestic politics (see “The Globalization–Democracy Controversy,” this chapter) may also account for the negativeeffect of FDI on democracy.

The direct effect of a rise in portfolio (the openness of a country tofinancial flows) on the level of democracy of a country is also positive. How-ever, once the dependence of this particular effect on time is incorporated, asindicated by the interaction term of year∗portfolio, the total effect comesout negative. Specifically, the total effect of portfolio investment inflowsin the all-countries sample, which is given by (4.7240 − 0.0024∗year),is −0.016 for 1975, −0.04 for 1985, and −0.066 for 1996. Hence, the totaleffect of portfolio on democracy is negative and strengthens over time.

The results for portfolio suggest that a rise in portfolio investmentsalso has both positive and negative effects on democracy, as discussed ear-lier in the chapter, but the negative effect grows to become dominant in theall-countries sample. The positive main effect of portfolio investments ondemocracy may reflect the economic discipline financial markets impose on


governments, which has contributed to greater transparency, better man-agement practices, and stronger property rights institutions. The negativeeffect may capture the dramatic increase in the ability of portfolio invest-ments to move quickly among countries, which at times leads to financialcollapse and devastating sociopolitical outcomes.

An increase in diffusion (the number of democracies in the regionaround a country) raises the level of democracy in that country. Interpret-ing this variable as a proxy for information flows induced by integration of acountry into the world economy, one may argue that democratic ideas andvalues are more likely to transmit and diffuse to geographically proximatecountries. Indirectly, then, this result does not support the claim discussedearlier in the chapter that the increased information flows into a countrywould reduce the level of democracy in that country. The spread of demo-cratic ideas from one’s neighboring countries improves, rather than hinders,the country’s own level of democracy.

The third column in Table 2.4 shows the estimation results obtainedfor a sample that includes only LDCs. These results match well with thosewe reported earlier for the sample of all countries. However, two primarydifferences in results exist between the two samples. One difference involvesthe sizes of the direct effects of trade, fdi, portfolio, and diffusion(our measures of globalization) on democracy. The coefficients of thesevariables appear to be larger in the LDC sample than those obtained in thefull sample of all countries. We interpret this outcome as a demonstrationthat a rise in globalization has a larger effect on democracy in LDCs thanin DCs. In a way, this result is intuitive since, in the period covered by oursample, the democracy levels of DCs are stable for the most part and hoverat, or close to, the highest possible positive level of the democracy indicator.The democracy levels in LDCs, on the other hand, exhibit much largervariations and are more sensitive to the external influences of globalization.

A second difference between the two samples concerns the total effect ofportfolio on democracy. As in the all-countries sample, once one incor-porates the dependence of the democracy effect of portfolio on time, indi-cated by the interaction term year∗portfolio, the total effect (5.3892 −0.0027∗year) is 0.0567 for 1975, 0.0297 for 1985, and 0 for 1996. Hence,the total effect of a rise in portfolio on democracy for the LDC sampledeclines over time, as in the all-countries sample, but it has not yet turnednegative in our sample period.

We think this result is also intuitive. The direct effect of a rise in port-folio on an average country in the LDC sample is larger than the directeffect on an average country in the all-countries sample (5.8359 vs. 4.724,


respectively). A typical LDC gains more from a rise in portfolio investmentthan a typical country in the all-countries sample in terms of a rise in its levelof democracy. Over time, the positive effect of a rise in portfolio is coun-teracted by a growing negative effect that reflects the overall rise in globalcapital mobility and increasing volatility of financial markets. The size ofthis negative component is larger in the LDC sample (−0.0027∗year) thanin the all-countries sample (−0.0024∗year), but because the direct positiveeffect is larger in the LDC sample than in the all-countries sample, the overalleffect in the LDC sample remains positive. This outcome is consistent withthe observation that capital flows into a typical LDC are relatively smallerthan those into a typical country in the all-countries sample. As such, theshare of the growing negative effect of capital flows over time out of thetotal effect is smaller in the LDC sample than in the all-countries sample.

While we have discussed the signs of the effects of globalization on thelevel of democracy, the statistical findings also can inform us about the sub-stantive significance of these effects. In evaluating the substantive effect, weassume that trade, fdi, portfolio, or diffusion exhibit certain increases,and then we compute the changes in the level of democracy induced by thesevariables ceteris paribus. Studies often assume a one-standard-deviationincrease for a certain variable, as identified by the estimation sample, andwe also employ this approach. Using the coefficient of each variable forthe all-countries sample in Table 2.4, we compute the percent change indemocracy relative to its sample mean as the baseline.

Turning to the size of effect, a one-standard-deviation rise in trade(46.83%) leads to a decline of about 8% in the average level of democracyin the sample (1.72). A one-standard-deviation rise in diffusion (0.79)raises the level of democracy by 6.5% relative to the baseline case. A one-standard-deviation rise in fdi (2.04%) in 1975 raises the level of democracyby 6.5% relative to our baseline, ceteris paribus, whereas a similar rise infdi in 1996 raises the level of democracy only by 0.12%. A one-standard-deviation rise in portfolio (2.5%) in 1975 reduces the level of democracyby 2.3% relative to the baseline, whereas a similar change in 1996 reducesthe level of democracy by about 10%. Therefore, the immediate responses ofdemocracy to changes in globalization do not appear to be large in absolutesize; this indicates that the globalization-induced changes in democracytend to happen slowly over time.

What about the long-run dynamic changes in democracy resulting fromglobalization forces? Indeed, the changes noted earlier in globalization alsoaffect democracy in the next period through their effects on the lagged valueof democracy. As detailed in the appendix, we can compute the effect of a


change in, say, trade on democracy in the long run. On employing thoseone-standard-deviation changes, the long-term percentage changes in theaverage democracy level in the sample register a decline by 108% due totrade and a rise by 153% due to diffusion. The long-term effect of fdichange is a rise in the average level of democracy by 89% in 1975 and only1.6% in 1996. With respect to portfolio, its long-term effect is a decline inthe average level of democracy by 32% in 1975 and 131% in 1996. Hence,globalization forces could generate large changes in democratic governanceover the long run.

Sensitivity Analyses

The robustness of statistical findings is important for establishing empiricalregularity, proving the validity of theoretical arguments, and facilitatingpublic policy positions. Do the empirical results reported earlier hold whenother possible confounding control variables enter the statistical model,or when other statistical estimation procedures are used? To answer thisparticular question, we conduct a number of sensitivity analyses to verifythe robustness of the reported results.

In one type of sensitivity analysis, we employ different statistical estima-tors, which are methods to obtain the quantitative effects of variables fromhistorical data. In a second type of sensitivity analysis, we use various mea-sures of democracy based on different types of data. In the third and finaltype of sensitivity analysis, we add various additional control variables to themodel, including income inequality in a country, the level of education ina country, the number of INGOs to which a country belongs, and variablesdenoting whether a country belongs to periphery or semiperiphery of theworld system (yet another indicator of its level of development).

The details of these sensitivity analyses and their results are reported inthe chapter appendix. We can summarize these additional statistical resultsby noting that the statistical findings we have reported, both in terms ofthe effects on democracy of the key globalization variables and the effectson democracy of the original control variables specified in this section, arefound to hold under the large majority of the sensitivity analyses.

IMPLICATIONS FOR THEORY AND PUBLIC POLICY

In general, the effects of changes in globalization on the level of democracyin a country tend to be small in the short run but could accumulate to largemagnitudes over the long run. The effects of FDI, however, tend to dwindle


over time. We believe that these results have important implications notonly for future research on the determinants of democracy but also for thelarger public policy debate about globalization. If maximizing economicefficiency is the only public policy objective, then the negative effects ofglobalization on democracy do not, and should not, matter. However, ifthe democratic form of governance is deemed a desirable public policyobjective, our empirical findings presented in this chapter suggest a policydilemma between economic efficiency gains and democratic decline.

Our analysis implies that globalization erodes the prospects for demo-cracy. The post-1945 model of embedded liberalism served to sustain demo-cratic governance in DCs with open economies. However, this model cannotbe easily implemented in LDCs in the current globalization. The emerg-ing democracies among LDCs lack the financial and managerial resourcesneeded to build social safety nets. As trade liberalization continues, thenegative effect of trade on democracy may increase.

Similarly, the growing capital mobility accompanying globalization pro-duces a political dilemma for governments who want both economic com-petitiveness and democratic political accountability. Footloose capital isgenerally not accountable to the public. Capital mobility reduces the demo-cratic government’s ability to respond to popular demands for social welfareand effective economic management. Our findings imply that under eco-nomic openness, the room for policy maneuvering is obviously reduced.Hence, the threats to democracy from financial inflows and FDI are sub-stantial.

If the current trend in the effect of globalization on democracy contin-ues, our finding that the increasing number of democracies in a region isassociated with greater democracy may reveal a “diabolic” flip side. Thatis, as trade, FDI, and financial flows become increasingly associated withdemocratic decline, the number of democracies in a region may also decline,which in turn may contribute to a decrease in the number of democracies.

In light of our findings, it is only natural that one may ask the followingquestion: If democracy falls due to trade openness, is openness a mistake,say, for Eastern Europe? Openness may generate economic gains. Countriesopen up because they seek the gains, or actors poised to gain are able to gettheir way, but some arguments suggest openness can also reduce democracy,which is what we find for trade. Thus, if both democracy and openness aredesirable, countries face a trade-off. We further discuss this trade-off inthe context of our larger conceptual picture in Chapter 10. In addition, oneshould note that our findings apply to the large-N sample, detecting averagestatistical patterns rather than explaining any particular country or region;Eastern Europe is just part of a larger sample.


How can one alleviate the negative effects of globalization on democracy?We believe there are basically three possibilities. The first possibility is forthere to be one world government, managing one economic system, withoutany barriers to the flows of goods, financial capital, and other factors ofproduction. The trade-off between efficiency gains and democratic declinecan then be effectively managed. But this option is currently impractical.

A second possibility is to enhance the socioeconomic policy coordinationamong governments within the existing international system. For example,concerted tax policies could minimize excessive financial capital volatility inan integrated global financial market. With policy coordination, countriescould continue to benefit from the globalizing economy while minimizingthe harmful effects on democracy. However, one also notices the weaknessof such an option. Historically, international socioeconomic policy coor-dination was generally unstable. Within the existing international politicalsystem, each national government is accountable to its own people. Thepolicy coordination option is difficult to achieve and sustain over time.

A third possibility is to slow down the rate of globalization. For exam-ple, governments could design tax and subsidy policies to compensate thelosers from economic openness, slow down capital movements and reduceexcessive volatility, and curtail by law excessive rent-seeking by MNCs. To besuccessful, this approach would require coordinated market regulation thatinvolves the government, the private sector, and citizen advocacy groups,implying wide consultation with and under close monitoring by the privatesector and citizen advocacy groups. The purpose of such coordination is tominimize the reliance on the government alone to design and implementregulation. Governments, particularly in LDCs, often suffer from ineffi-ciency and inflexibility. Relying solely on governments may suffocate themarket. Where democratic rules are not well established, as in many LDCs,the notion that only the government can correct market failures may causeexcessive government intervention and lower social welfare, and it eventu-ally precipitates a decline in democratic governance. Although this optionmay yield lower social welfare relative to the first two possibilities, it maybe the only practical approach available at present for reducing the negativeeffects of globalization on democracy.

SUMMARY AND OUTLOOK

The theoretical literature presents conflicting expectations of the effect ofglobalization on democracy. These conflicting expectations are also reflectedin the public debate on globalization. In this chapter, we make a system-atic empirical effort to assess the controversial effects of globalization on


democracy. We find that trade openness and portfolio investment inflowsexert negative effects on democracy. The negative effect of trade openness onthe level of democracy is stable over time, whereas the negative effect of port-folio investment inflow into a country strengthens over time. FDI inflowsinto a country are conducive to its democratic governance, but the strengthof this particular effect weakens over time. The spread of democratic ideasflowing into a country from other countries in its vicinity contributes toincreasing the level of democracy in that country persistently over time.These effects are all substantively important, particularly in light of theirlong-term implications for democracy.

These empirical regularities are found to hold for both the all-countrysample and LDCs alone. They are also robust across various model specifi-cations, different measures of democracy, and alternative statistical estima-tors. In sum, the economic aspects of integration into the world economyare beginning to cause a decline in national democratic governance, whichought to cause concern for both policymakers and academicians.

In this chapter, we investigated the effect of aspects of economic openness,or economic globalization, on the level of democracy within countries. Thereader may recall that, in our sensitivity analyses, the distribution of incomewithin countries, as measured by the Gini coefficient, was one of the controlvariables. In the next chapter, we change the focus of our inquiry, makingincome inequality the main variable of interest. What is the effect of a rise inthe level of democracy on the national distribution of income? What is theeffect of our measures of economic openness (i.e., trade, FDI, and portfolioflows) on the national distribution of income? Do these forces make therich richer and the poor poorer, further skewing the income distribution,or do they make the rich poorer and the poor richer, reducing the incomegap? Historical observation suggests that these are important questions inthat they can have political ramifications, particularly for the stability ofthe government and civic life. Extremely skewed distributions of incomeare known to have contributed to the violent revolutions in France in theeighteenth century, Russia in the early twentieth century, and China in themid-twentieth century. Now it is time for us to step into Chapter 3.

APPENDIX


This appendix accompanies the section “Empirical Model and Analysis”in the main text. We provide various details on the measurement of our


variables, their data sources, issues pertaining to the technical design of thestatistical model, and the expected effects of various control variables. Wealso report and discuss the results for the control variables as well as thevarious additional analyses.

Empirical Model

The dependent variable democracy denotes the level of democracy for acountry in any given year, based on the POLITY III data (Jaggers and Gurr,1995; Gurr and Jaggers, 1999). The widely used POLITY III data registervarious attributes of regime type on an annual basis for many countriesfrom about 1800 to 1998, using two indices. The 10-point democracy index(DEMOC) measures the democratic characteristics of the regime. The 10-point autocracy index (AUTOC) measures the autocratic characteristicsof the regime. As pointed out by Oneal and Russett, Londregan and Poole,Mansfield and Snyder, among others, because many governments have bothdemocratic and autocratic characteristics, DEMOC and AUTOC do notprovide redundant information about regime type, and both should be usedto measure the level of democracy.22 We follow these studies and measuredemocracy as the difference between DEMOC and AUTOC, generating anindex ranging between −10 (for the most autocratic regime) and 10 (forthe most democratic regime).

Globalization is measured by four indicators: trade openness, FDIinflows, portfolio investment inflows, and the spread of democratic ideas.The three economic variables are collected from the World Bank’s WorldDevelopment Indicators CD-ROM. The trade variable denotes the yearlylevel of trade openness for a country. As conventional in the literature,trade is defined as the sum of the value of imports and the value of exportsof goods and services of a country with the rest of the world, measured as apercentage of the country’s GDP.23 The fdi variable denotes the yearly valueof net inflows of FDI as a percentage of a country’s GDP for each country.FDI involves either the acquisition of a lasting management interest (10%or more of voting stock) in an enterprise operating in an economy otherthan that of the investor or the creation of a new subsidiary of a firm in aforeign country. It is given here as the sum of equity capital, reinvestmentof earnings, and other long- and short-term capital as shown in the bal-ance of payments. The portfolio variable denotes the yearly value of net

22 See, e.g., Mansfield and Snyder (1995), Londregan and Poole (1996), and Oneal and Russett(1997, 1999a).

23 We also estimated the models using the log of trade/GDP. All the results remain robust.


inflows of portfolio investments as a percentage of GDP for each country.Portfolio investments (excluding liabilities constituting foreign authorities’reserves) cover transactions in equity securities (the sum of country funds,depository receipts, and direct purchases of shares by foreign investors) anddebt securities (publicly guaranteed and nonguaranteed debt from privatelyplaced bonds).

The measure of the spread of democratic ideas is coded using POLITY IIIdata; diffusion denotes the yearly number of democracies within a regionaround each country. The regions include Europe, Middle East, Africa, Asia,and North and South America. We define a country as democratic if thedifference between its DEMOC and AUTOC scores is greater than or equalto six (as in, e.g., Mansfield and Snyder, 1995; Oneal and Russett, 1997,1999a, 1999b, 1999c). Like Starr (1991) and Gasiorowski (1995), we assumethat the type of political regime in a country is more likely to be affectedby the type of political regime of its neighboring countries than by moredistant countries.

In general, the diffusion of democratic norms is said to work throughcontact-based mechanisms, which involve information flows, contact inthe marketplace, communication networks, tourism, and so forth. Geo-graphically proximate countries typically have more contact. Hence, in ourcontext, diffusion may be thought of as a proxy for information and com-munication flows of democratic ideas among countries. It should be notedthat we compute the correlation between a yearly measure of the numberof main phone lines and television sets for each country and diffusion.The correlation (r = 0.45) is statistically significant at the 1% level. We donot include this measure as an independent variable because it is highlycollinear (r = 0.8) with GDP per capita in the model. Hence, it is possibleto argue that GDP per capita also serves as a proxy of the communicationaspect of national integration into the world economy.

Moving to the economic control variables, inflation denotes the yearlygrowth rate of the GDP deflator, which is the ratio of nominal GDP (mea-sured in current prices) to real GDP (measured in constant prices). Severalstudies use inflation as a proxy of economic crisis, the effect of which ondemocracy is debated. Some studies argue the effect is positive, whereasothers argue the effect is negative (Huntington, 1991; Gasiorowski, 1995;Haggard and Kaufman, 1995; Drake, 1998). The development variabledenotes the logged yearly GDP per capita in purchasing power parity–adjusted real international prices. As discussed at length in this chapter,many statistical studies of the determinants of democracy use this variableas an indicator of the level of economic development, either as a central


variable or as a control variable.24 The growth variable denotes the annualpercentage growth rate of real GDP.


Several empirical design issues influence our model specification and esti-mation in this chapter. First, in the main text, we note that, to control forthe effect of inertia, we include the lagged dependent variable (the level ofdemocracy) as an independent variable, a strategy also adopted in otherstudies.25 In addition to the theoretical reason, this lagged variable helpscapture the effects of variables not present in the model. “With such a perva-sive control in place, it is more difficult for spurious effects to be reported”(Burkhart and Lewis-Beck, 1994: 905). It is worth noting that the inclu-sion of the lagged dependent variable might soak up the variations in thedependent variable that could be explained by other independent variables,making it harder for us to find statistically significant results. Hence, ourapproach can be described as conservative.

Second, to address the possibility that the effects of some of the determi-nants of democracy may change over time, we use the systematically varyingparameter approach (Judge et al., 1988: 435; Griffiths et al., 1993: 421). Todetermine which variables have a time-varying effect, we first estimate apooled time-series cross-sectional model for each decade (1970s, 1980s,and 1990s). The decade selection has no particular theoretical justification,but the results provide a general sense about the temporal stability of theeffects on democracy. Variables whose parameters exhibit reversals of signsacross those pooled decades are suspected to have time-varying effects ondemocracy. Interaction terms between these variables and a calendar yearvariable are created and included in the model.26

Third, the error term in the model faces the risks of heteroskedasticityand serial correlation. The estimated coefficients are still consistent withheteroskedastic variance and serial correlation in the error term, but theirstandard errors are not efficient and are most likely biased. To deal with thisproblem, we estimate our statistical model using a variant of the White esti-mator of robust standard errors, which adjusts for clustering over country

24 In addition to sources cited in Chapter 2, see, e.g., Rueschemeyer (1991), Diamond (1992a),Gasiorowski (1995), and Przeworski and Limongi (1997).

25 See, e.g., Bollen (1979), Muller (1988), Burkhart and Lewis-Beck (1994), and Muller and Seligson(1994).

26 Gasiorowski (1995) employs such an interactive model to examine the time-varying effects ofinflation on democratization.


(White, 1980). This estimator yields consistent estimation of the covariancematrix under very general conditions of heteroskedasticity and serial cor-relation (Wiggins, 1999). In addition, as Beck and Katz suggest, the laggeddependent variable models the temporal dynamics and also helps to addressserial correlation (Beck and Katz, 1995a, 1995b).

Fourth, consider the design issue for resolving the simultaneity bias.Ignoring the possible effect of democracy (the dependent variable in thischapter) on the right-hand-side variables in the model (globalization-related and others) implies the risk of simultaneity bias, which generatesmisleading estimated coefficients. The ideal solution to this problem isto estimate a fully specified simultaneous equations system, endogenizingdemocracy, trade, FDI, portfolio investment, and information flows. At thecurrent stage of our knowledge of how these forces interact, it is not feasibleto specify such a system of equations. In contrast, many studies deal withthis problem by lagging the independent variables one period, which inour case amounts to one year (e.g., Muller and Seligson, 1994; Oneal andRussett, 1997, 1999a, 1999b, 1999c). Though this is not an optimal solutionto the issue of simultaneity, we follow the same approach here.

Fifth, we need to consider which statistical significance level to use whenwe interpret the results. In evaluating the effect of globalization on democ-racy, we have discussed three types of theories. One type expects that glob-alization will promote democracy, the other type expects the opposite, andthe third type expects no effect at all. As we noted, the three types oftheory appear logically consistent, given their own assumptions, regardingthe effect of globalization forces on democracy. The empirical model canonly identify the net effect of globalization on democracy. We thereforeonly test the sign of this net effect on democracy against the null hypoth-esis of no effect – or rather the two competing effects of globalization ondemocracy are equal in size – employing the one-tailed test in reporting theresults. Many other studies have employed this approach (see, e.g., Morrowet al., 1998; Oneal and Russett, 1999a, 1999b, 1999c; Li and Reuveny, 2003;Reuveny and Li, 2003).

Sixth, there is the issue of what thresholds to use for statistical significance.In interpreting the results, we employ 1, 5, and 10% significance levels. Wetake this approach for a couple of reasons. Some of our samples are relativelysmall because of limitations on data availability. In addition, the laggeddependent variable in the model soaks up the variations in the dependentvariable to be explained by other explanatory variables, making it difficultto find significant effects.

Seventh, we need to measure the size of globalization’s effect on demo-cracy. In the text, we discuss not only the direct effect of the globalization


variable but also the long-run effect through the lagged dependent variable.For the long-run effect, the impact of a globalization variable on the level ofdemocracy from previous periods is absorbed into the effect of the laggeddemocracy variable, which is also on the right-hand side. The globalizationvariable affects the current level of democracy via the direct effect andcontinues to affect the level of democracy in the next period via the laggeddemocracy variable (Londregan and Poole, 1996). The long-run impactof a change in a globalization variable produces the following change inthe level of democracy in a country: [coefficient of globalization variable/(1–coefficient of lagged democracy)] ∗ (change in globalization variable).We compute the long-run effects for all our globalization variables.

Finally, one may question in particular why we measure the impact of FDIby its flow rather than by stock. As noted, FDI flows may affect democracy,and FDI flows through many mechanisms and generates competing effectson democracy. The goal of the empirical analysis is to measure the immedi-ate net effect of FDI flows on democracy. This measurement requires the useof FDI inflows (incurring during a year) because the stock variable, cumu-lated over time since the beginning of FDI, conflates the short- and long-runeffects of FDI. As explained earlier, we estimate the long-run impact of FDIthrough the inclusion of the lagged dependent variable in the model. Inaddition, the FDI stock data are statistically problematic. Whereas flows arereported by national governments to the IMF, stock data, which are onlyavailable from the United Nations Conference on Trade and Development(UNCTAD), are often estimated and involve large measurement errors.Stock data sum FDI flows and reinvestment earnings since the start ofFDI. Reinvestment earning data are often missing, generating measurementerrors that continue in the stock data in future years. In contrast, the effect ofsuch measurement error in the flow data is limited to the year of the missingdata. Moreover, the initial baseline stock of FDI in a country is oftenunknown for sure, and the year UNCTAD uses as the start of stock accumu-lation is the year in which a country began to report flows, not necessarilythe year when FDI first entered a country. This generates another type ofmeasurement error. Due to these errors, the stock data are not appropriatefor cross-national comparisons.

Empirical Findings

Table 2.4 presents the results for all countries from 1970 to 1996 in thefirst column and for the non-OECD countries in the second column. Theadjusted R2 for both models are 0.93 and 0.90, respectively, indicatingthat the models explain most of the observed variations in democracy.


All the interaction terms are statistically significant, as are all the maineffects of the variables for which an interaction was included. Anothergeneral observation is that all the globalization-related variables are alsostatistically significant. In fact, except for economic growth, all the effectsare statistically significant, which supports our model specification. Becausewe have discussed the results for the globalization variables in detail, herewe focus on the control variables.

In terms of the specific results of the control variables in column 1, priordemocracy has a positive and statistically significant effect on currentdemocracy. Other studies that include this variable report similar results,illustrating that democracy as a social phenomenon tends to have inertia.As Przeworski et al. (1996) argue, it almost seems tautological to arguethat the current level of democracy rises when the past level of democracyrose. However, this result has an important substantive implication. Onesimply cannot expect to find a higher level of democracy by strengtheningdictatorship or authoritarianism ceteris paribus. If the level of democracyfell in the past, one should expect to find a lower level of democracy inthe future. Therefore, international interactions that strengthen an author-itarian regime are likely to hinder the movement toward democracy inthat country in the near future. Similarly, reforms toward a higher level ofdemocracy tend to accumulate over time, leading to democratic consolida-tion.

The effect of inflation is statistically significant and positive, but itsinteraction term year∗inflation is significant and negative. The total effectof inflation on the level of democracy is positive and was the largest in the1980s; economic growth has a positive sign, but it is not statisticallysignificant. As in many studies, economic development has a positiveand statistically significant effect on the level of democracy. The interactionterm year∗economic development is statistically significant but negative,suggesting that its effect weakens over time. The calendar year variable yearis statistically significant and positive, indicating in this sample a partialtendency for democracy to grow over time.

The results regarding the domestic economic effects on democracy repli-cate the spirit of previous studies, which supports our model. To furtherinvestigate the robustness of the results, we estimated the model for a sampleof only non-OECD countries. Compared with the results for all countriesin column 1, the adjusted R2 for the non-OECD countries in column 2is a bit lower because of the smaller sample size and reduced variation inthe independent variables. However, the signs and statistical significancelevels for all the variables in column 2 match well with those in column 1,


indicating that the results in column 1 are not an artifact of the inclusion ofthe OECD countries in the sample.

We also estimate the models in Table 2.4 by excluding the lagged depen-dent variable. Although the inclusion of the lagged dependent variableusually makes other independent variables insignificant, its exclusion fromour model actually renders the results statistically less significant in bothsamples. This suggests that we did not get null results in Table 2.A2 becauseof the lagged dependent variable. These results are available from the authorsupon request.

Now, as noted, to identify which variables have varying effects on democ-racy over time, we first estimate pooled time-series cross-sectional modelsfor each decade for all countries and for LDCs. We report those resultsin Table 2.A1. The effect of trade on democracy is always negative andstatistically significant, and the effect of diffusion is always positive andstatistically significant in four out of six cases. The effect of prior demo-cracy is always positive and statistically significant. The effect of economicgrowth is positive but statistically insignificant in all cases (except for thecase of all countries from 1970 to 1979). Because the signs of these fourvariables are consistent across decades, we include them in the model asmain effects only. The signs of economic development, fdi, portfolio,and inflation change over decades. Hence, they enter the model bothindividually and as interactive terms with year. For the overlapping vari-ables (inflation, economic growth, economic development, tradeopenness), the results generally agree with those reported by Gasiorowski(1995), although his dependent variable measures the event of democrati-zation (a 1, 0 variable), whereas our dependent variable measures the levelof democracy. Overall, these results provide justification for why we specifythe interactive model as such in the main text.

Additional Analyses

Model specification, the measures of democracy, and the estimation tech-niques vary across statistical studies of democracy. To investigate the effectsof these variations in our case, we conduct various sensitivity analyses.Specifically, we add new control variables, use Freedom House data to mea-sure democracy, and apply different estimators. The results are presentedin Tables 2.A2 and 2.A3. Overall, the effects of globalization on democracyreported in Table 2.4 are replicated across the 25 experiments reported inTables 2.A2 and 2.A3. Therefore, we judge our results to be robust. Later wediscuss Tables 2.A2 and 2.A3 in detail. The results for the original control

Tabl

e2.

A1.

Pool

edti

me-

seri

escr

oss-

sect

iona

lmod

els

ofde

moc

racy

byde

cade

ALL

(197

0–19

79)

ALL

(198

0–19

89)

ALL

(199

0–19

96)

LDC

(197

0–19

79)

LDC

(198

0–19

89)

LDC

(199

0–19

96)

TR

AD

E−0

.004

2∗∗−0

.002

1∗−0

.004

1∗∗−0

.004

9∗∗−0

.002

9∗∗−0

.003

2∗

(0.0

022)

(0.0

014)

(0.0

021)

(0.0

025)

(0.0

016)

(0.0

024)

FDI

0.01

41−0

.035

7−0

.018

80.

0110

−0.0

426

−0.0

200

(0.0

178)

(0.0

340)

(0.0

391)

(0.0

192)

(0.0

385)

(0.0

424)

PO

RT

FOLI

O0.

0031

−0.0

211

0.00

21−0

.011

8−0

.024

70.

0092

(0.0

183)

(0.0

196)

(0.0

069)

(0.0

216)

(0.0

210)

(0.0

093)

DIF

FUSI

ON

0.60

34∗∗

0.15

010.

2251

∗0.

9548

∗∗∗

0.19

810.

2378

∗

(0.2

662)

(0.1

559)

(0.1

445)

(0.3

172)

(0.1

777)

(0.1

617)

INFL

AT

ION

−0.0

031

0.00

02∗∗

∗−0

.000

1−0

.004

5∗0.

0001

∗∗∗

−0.0

001

(0.0

028)

(0.0

000)

(0.0

001)

(0.0

029)

(0.0

000)

(0.0

001)

GR

OW

TH

−0.0

019

0.00

670.

0115

0.00

120.

0071

0.01

27(0

.015

3)(0

.012

2)(0

.015

6)(0

.016

5)(0

.012

5)(0

.016

0)D

EV

ELO

PM

EN

T0.

1098

0.40

81∗∗

∗−0

.082

00.

1523

0.53

52∗∗

∗−0

.193

8∗∗

(0.1

143)

(0.0

999)

(0.0

797)

(0.1

509)

(0.1

199)

(0.1

120)

PR

IOR

DE

MO

CR

AC

Y0.

9253

∗∗∗

0.92

53∗∗

∗0.

9202

∗∗∗

0.91

41∗∗

∗0.

9290

∗∗∗

0.91

70∗∗

∗

(0.0

237)

(0.0

161)

(0.0

226)

(0.0

244)

(0.0

170)

(0.0

238)

Con

stan

t−1

.414

2−3

.092

7∗∗∗

1.00

82−2

.299

9−4

.027

7∗∗∗

1.75

65∗

(1.0

956)

(0.8

233)

(0.7

240)

(1.3

985)

(0.9

187)

(0.9

460)

Obs

erva

tion

s35

394

872

027

377

259

5A

dju

sted

R2

0.93

0.94

0.90

0.89

0.91

0.88

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗

∗si

gnifi

can

tat

5%;∗

∗∗si

gnifi

can

tat

1%.

54

Tabl

e2.

A2.

OLS

esti

mat

esw

ith

addi

tion

alco

ntro

lvar

iabl

esor

Free

dom

Hou

seda

ta

Mod

el1

Mod

el2

Mod

el3

Mod

el4

Mod

el5

Mod

el6

Mod

el7

Mod

el8

ALL

LDC

ALL

LDC

ALL

LDC

ALL

LDC

PO

LIT

YII

Ida

taT

RA

DE

−0.0

028∗∗

∗−0

.003

3∗∗∗

−0.0

025∗∗

∗−0

.002

7∗∗∗

−0.0

025∗∗

∗−0

.002

7∗∗−0

.002

3∗∗∗

−0.0

025∗∗

FDI

5.43

10∗

6.37

62∗∗

5.55

93∗

5.71

625.

6112

∗6.

1508

∗8.

4656

∗∗10

.595

0∗∗

PO

RT

FOLI

O4.

1119

∗∗4.

9084

∗∗2.

7475

3.60

582.

6123

3.17

675.

6914

3.46

33D

IFFU

SIO

N0.

2369

∗∗0.

2971

∗∗∗

0.31

41∗∗

∗0.

3915

∗∗∗

0.31

24∗∗

∗0.

3928

∗∗∗

0.28

44∗∗

∗0.

3558

∗∗∗

YE

AR

∗ FDI

−0.0

027∗

−0.0

032∗∗

−0.0

028∗

−0.0

029

−0.0

028∗

−0.0

031∗

−0.0

043∗∗

−0.0

053∗∗

YE

AR

∗ PO

RT

FOLI

O−0

.002

1∗∗−0

.002

5∗∗−0

.001

4−0

.001

8−0

.001

3−0

.001

6−0

.002

9−0

.001

7G

INI

−0.6

336∗

−0.9

099∗

−0.1

091

−0.3

664

0.09

50−0

.226

3E

DU

CA

TIO

N0.

1417

0.11

370.

1343

0.08

790.

2889

∗∗∗

0.24

57∗∗

N20

2116

4016

6212

8116

6212

8117

8313

53A

dju

sted

R2

0.93

0.90

0.93

0.91

0.93

0.91

0.93

0.91

Mod

el9

Mod

el10

Mod

el11

Mod

el12

Mod

el13

Mod

el14

Mod

el15

ALL

LDC

ALL

ALL

ALL

ALL

LDC

Free

dom

Hou

seD

ata

TR

AD

E−0

.001

8∗∗−0

.001

2−0

.001

6∗∗−0

.002

6∗∗∗

−0.0

006

−0.0

007∗

−0.0

007∗

FDI

6.29

54∗∗

8.00

87∗∗

6.83

47∗∗

4.86

32∗

5.84

62∗∗

4.39

57∗

5.32

03∗

PO

RT

FOLI

O4.

5246

∗∗∗

5.71

76∗∗

3.93

81∗∗

∗5.

1121

∗∗∗

2.36

40∗

2.40

35∗∗

1.65

29D

IFFU

SIO

N0.

2576

∗∗∗

0.30

73∗∗

∗0.

1345

0.29

02∗∗

∗0.

1151

∗∗∗

0.09

95∗∗

∗0.

1209

∗∗∗

YE

AR

∗ FDI

−0.0

032∗∗

−0.0

040∗∗

−0.0

035∗∗

−0.0

025∗

−0.0

030∗∗

−0.0

022∗

−0.0

027∗

YE

AR

∗ PO

RT

FOLI

O−0

.002

3∗∗∗

−0.0

029∗∗

−0.0

020∗∗

∗−0

.002

6∗∗∗

−0.0

012∗

−0.0

012∗∗

−0.0

008

ING

OS

0.00

02∗∗

0.00

07∗∗

∗

SEM

IPE

RIP

HE

RY

∗ DE

VE

LOP

ME

NT

0.01

70∗∗

−0.0

035

PE

RIP

HE

RY

∗ DE

VE

LOP

ME

NT

−0.0

046

−0.0

117

EU

RO

PE

0.01

48M

IDD

LEE

AST

−0.7

589∗∗

AFR

ICA

−0.5

427∗∗

ASI

A−0

.159

1N

1949

1568

1879

1826

1705

1784

1421

Adj

ust

edR

20.

930.

900.

930.

930.

940.

940.

90

Not

e:∗

sign

ifica

nta

t10%

;∗∗si

gnifi

can

tat5

%;∗∗

∗si

gnifi

can

tat1

%.O

ther

inde

pen

den

tva

riab

les

not

repo

rted

tosa

vesp

ace.

Mod

els

5an

d6

incl

ude

wh

ileM

odel

s7

and

8ex

clu

deec

on

om

icd

evel

opm

ent

and

its

inte

ract

ion

term

.

55


variables from the models in Tables 2.A2 and 2.A3 are broadly consistent withTable 2.4. Therefore, we do not discuss these results again in this subsection.

Table 2.A2 reports the results from 15 experiments, denoted Model 1through Model 15, which investigated the effects of adding new controlvariables and/or using a different measure of democracy. In Models 1–12,democracy is measured based on POLITY III data as in Table 2.4 whereasin Models 13–15, democracy is based on Freedom House data. As for thenew control variables, gini denotes the level of national income inequality,measured by the Gini coefficient. The variable education denotes thelevel of national education measured by the average number of years ofeducation in the population. For gini and education, we use the samedata and variable transformation as in Feng and Zak (1999). The variablegini comes from the high-quality data from Deininger and Squire (1996).We log and lag the raw data. The missing values of gini are filled withpredictions from estimating gini as a function of GDP per capita, GDPper capita squared, and regional dummies. The variable education comesfrom Barro and Lee (2000) and is also logged and lagged. The variable ingosdenotes the number of INGOs to which individuals or organizations fromeach country belonged in a given year. The data on INGOs come from Boliand Thomas (1999).

The two dummy variables semiperiphery and periphery denotewhether a country belongs to the world-system semiperiphery or periphery,respectively. In the model, they interact with economic development asin Burkhart and Lewis-Beck (1994). Similar to this study, we have also esti-mated the models using only semiperiphery and periphery dummies bythemselves and found similar results; europe, middle east, africa, andasia are dummy variables that denote the geographical region of a country.

Across Models 1–12 in Table 2.A2, the effect of trade is significant in 11out of 12 cases and is always negative. The main effect of fdi is significantin 11 cases and is always positive. The interaction effect of fdi is significantin 11 cases and is always negative. The effect of diffusion is significant in11 cases and is always positive. The main effect of portfolio is significantin 6 cases and is always positive. The interaction effect of this variable issignificant in 6 cases and is always negative. These results strongly supportthose in Table 2.A2.

Regarding the dependent variable, although we measure democracy basedon the POLITY III data, some scholars employ the Freedom House data(e.g., Burkhart and Lewis-Beck, 1994; Diamond, 1999). It is worth notingthat some scholars point out that the Freedom House data are not suitablefor a pooled time-series analysis because the data are not fully comparable

Tabl

e2.

A3.

Para

met

eres

tim

ates

from

alte

rnat

ive

esti

mat

ors

Fixe

def

fect

s(c

oun

try)

Fixe

def

fect

s(y

ear)

Ran

dom

effe

cts

OL

San

dP

CSE

GE

E

All

cou

ntr

ies

TR

AD

E0.

0004

−0.0

027∗∗

∗−0

.003

0∗∗∗

−0.0

029∗∗

∗−0

.002

7∗∗∗

FDI

8.79

6∗∗4.

786∗

5.19

65.

191∗

4.90

3∗

PO

RT

FOL

IO0.

515

3.54

5∗4.

724

4.72

44.

704∗∗

DIF

FUSI

ON

0.19

70.

285∗∗

∗0.

244∗∗

∗0.

244∗∗

∗0.

237∗∗

∗

INFL

AT

ION

0.00

60.

039

0.06

00.

060

0.06

7∗∗

GR

OW

TH

0.00

40.

005

0.00

70.

007

0.00

7D

EV

ELO

PM

EN

T45

.851

∗∗∗

36.5

66∗∗

∗34

.290

∗∗∗

34.2

60∗∗

32.6

34∗∗

∗

PR

IOR

DE

MO

CR

AC

Y0.

797∗∗

∗0.

925∗∗

∗0.

927∗∗

∗0.

927∗∗

∗0.

934∗∗

∗

YE

AR

0.26

2∗∗∗

0.15

7∗∗∗

0.15

1∗∗∗

0.15

1∗∗0.

145∗∗

∗

YE

AR

∗ INFL

AT

ION

−0.0

0000

3−0

.000

02−0

.000

03−0

.000

03−0

.000

03∗∗

YE

AR

∗ FDI

−0.0

04∗∗

−0.0

02∗

−0.0

03−0

.003

∗−0

.002

∗

YE

AR

∗ PO

RT

FOL

IO−0

.000

3−0

.002

∗−0

.002

−0.0

02−0

.002

∗∗

YE

AR

∗ DE

VE

LOP

ME

NT

−0.0

23∗∗

∗−0

.018

∗∗∗

−0.0

17∗∗

∗−0

.017

∗∗−0

.016

∗∗∗

Con

stan

t−5

17.1

72∗∗

∗−3

12.6

5∗∗∗

−302

.064

∗∗∗

−301

.820

∗∗−2

88.8

35∗∗

∗

N20

2120

2120

2120

2120

21A

dju

sted

R2

0.93

0.93

0.93

0.93

LD

Cs

TR

AD

E−0

.000

005

−0.0

030∗∗

∗−0

.003

2∗∗∗

−0.0

032∗∗

∗−0

.002

9∗∗∗

FDI

11.0

75∗∗

5.18

3∗5.

836

5.83

6∗5.

536∗

PO

RT

FOL

IO−1

.311

4.65

2∗5.

389

5.38

95.

882∗∗

DIF

FUSI

ON

0.18

50.

358∗∗

∗0.

292∗∗

∗0.

292∗∗

∗0.

290∗∗

∗

INFL

AT

ION

0.00

60.

024

0.05

40.

054

0.06

1∗∗

GR

OW

TH

0.00

30.

004

0.00

70.

007

0.00

7D

EV

ELO

PM

EN

T24

.572

41.1

39∗∗

36.7

64∗∗

36.7

64∗

35.5

77∗∗

PR

IOR

DE

MO

CR

AC

Y0.

792∗∗

∗0.

922∗∗

∗0.

924∗∗

∗0.

924∗∗

∗0.

934∗∗

∗

YE

AR

0.18

6∗0.

171∗∗

∗0.

159∗∗

0.15

9∗0.

154∗∗

∗

YE

AR

∗ INFL

AT

ION

−0.0

0000

3−0

.000

01−0

.000

03−0

.000

03−0

.000

03∗∗

YE

AR

∗ FDI

−0.0

06∗∗

−0.0

03∗

−0.0

03−0

.003

∗−0

.003

∗

YE

AR

∗ PO

RT

FOL

IO0.

001

−0.0

02∗

−0.0

03−0

.003

−0.0

03∗∗

YE

AR

∗ DE

VE

LOP

ME

NT

−0.0

13−0

.021

∗∗−0

.018

∗∗−0

.018

∗−0

.018

∗∗

Con

stan

t−3

68.0

01∗

−340

.864

∗∗−3

17.8

47∗∗

−317

.847

∗−3

06.1

19∗∗

∗

N16

4016

4016

4016

4016

40A

dju

sted

R2

0.91

0.91

0.90

0.90

Not

e:∗

sign

ifica

nt

at10

%;∗∗

sign

ifica

nt

at5%

;∗∗∗

sign

ifica

nt

at1%

.

57


across years due to changes in methodology and scale (Gates, 2002).27 Weuse these data here because they offer a new take on qualities of democracies,despite their limitations. POLITY III data emphasize constraints on execu-tive political recruitment, contestation, and functioning. Based on the Free-dom House data, Diamond (1999) argues that in recent years some countrieshave become pseudodemocracies, where the rights of contestation are pro-tected but civil rights are precarious. In addition to institutional aspectsdirectly measured in POLITY III, Freedom House data capture aspects thatare only indirectly implied in POLITY III, such as the de facto power of theopposition, freedom from foreign domination, minority rights, freedomof expression and belief, association rights, rule of law, human rights, andpersonal economic rights.

Unlike POLITY III, Freedom House data are available only from 1973onward. In their overlapping period, the cross-correlation between the twodata sets is relatively high: 0.88 from a sample of all countries and 0.83from a sample of LDCs. Indeed, the inferences from the Freedom House–based models in Table 2.A2 do not differ much from the inferences fromthe POLITY III–based models in Table 2.4. Like Burkhart and Lewis-Beck(1994), we sum the two 7-point scales of “political rights” and “civil liberties”in Freedom House. This procedure gives an index ranging from 2 (lowestdemocracy) to 14 (highest democracy). We also recompute diffusion andprior democracy based on this score. Across Models 13–15, the effect oftrade is always negative and statistically significant in two cases. The effectof diffusion is always positive and significant. The positive effect of fdiand the negative effect of its interaction term are always significant. Thepositive effect of portfolio and the negative effect of its interaction termare significant in two cases.

Compared with Table 2.4, the results for Models 14 and 15, based onthe Freedom House data, are weaker in terms of statistical significancefor portfolio investments. We believe this is because Freedom House alsomeasures civil rights and democratic norms that take a relatively long timeto develop. The Freedom House index typically lags behind the POLITYIII index and has a smaller variance.28 Because the Freedom House indexvaries less, the results are less significant. Furthermore, because portfolio

27 For additional issues on the measurement of democracy, see Bollen (1991).28 For example, for Bulgaria, the POLITY score rose from −7 in 1989 to +8 in 1991, a 75% increase

(within a range of 20 from −10 to +10), whereas the Freedom House score rose from 2 in 1989to 9 in 1991, a 53% increase (within a range of 13 from 2 to 14). For another example, for Russiaover the same time frame, the POLITY score rose from −4 to +6, a 50% increase, while itsFreedom House score rose from 5 to 7, a 15% increase.


investments in LDCs are small and concentrated in the elite, their effectcan be large in terms of constraints on the executive, which is the focus ofPOLITY III, but is likely to be smaller in terms of the civil rights, which iscaptured by the Freedom House.

Table 2.A3 presents results for five estimators: country fixed effects, yearfixed effects, random effects, OLS with panel corrected standard errors(PCSE), and general estimating equation (GEE) model. The country fixed-effects estimator assumes that the intercept term varies across panels, theyear fixed-effects estimator assumes that the intercept term varies acrossyears, and the random effects estimator assumes that the error term variesacross panels. The PCSE method assumes that the variance of the error termis heteroskedastic across panels and homoskedastic within panels (Beck andKatz, 1995b). GEE is a population average–based estimator that is frequentlyused for panel data (Liang and Zeger, 1986). All models, including those inTable 2.4, have roughly the same goodness of fit. The estimates in Table 2.A3are generally similar to those in Table 2.4, except for the country fixed-effectsestimator.

Scholars debate the merits of panel data estimation methods. Althoughthis chapter is not about methods, we believe the estimator used in Table 2.4(OLS with White robust standard errors adjusted for clustering over coun-tries) is the most appropriate for our case. Beginning with the fixed-effectsestimators, King (2001) argues that using more elegant statistical techniquesis preferable to using fixed effects estimation, which is the approach takenin Table 2.4. In simulations, fixed-effects estimators are known to soak upexcessive between-country variations attributable to substantive variablesand to reduce degrees of freedom due to the inclusion of many dummy vari-ables. Unless the number of time periods approaches infinity, the estimatedeffects are biased for the country fixed-effects estimator (Green, 2007).Because we have 27 periods, our analysis may suffer from this problem,which could explain why the results from the country fixed-effect estimatordiffer noticeably from those based on the other estimators.

The applicability of the random effects model for a certain data set canbe tested using an appropriate Lagrange multiplier test. The null hypothesisis that the random effect model is not supported by the data (Green, 2007).From conducting this test we are not able to reject the null hypothesis,because we find no evidence in favor of the random effects model.

The OLS with PCSE estimator assumes that the variance of the error termis heteroskedastic and contemporaneously correlated across panels, andhomoskedastic within panels (Beck and Katz, 1995b). The PCSE estimatormay not be appropriate in our case because in Table 2.4 several variables have


time-varying effects, suggesting the possibility of heteroskedastic errors overtime. The estimator we used in Table 2.4 does not assume a certain structureof heteroskedasticity or serial correlation, making it more appropriate forour purpose. One advantage of the PCSE method is its ability to correct forcorrelated errors across panels. However, as Beck and Katz (1995b) note, abetter strategy is to model this correlation theoretically, which is achievedby our diffusion variable.

The results from the GEE method in Table 2.A3 are identical to the resultsin Table 2.4 in terms of hypothesis testing. This outcome further supportsthe claim that our results are generally robust when different estimators areused (except for the country-fixed effects variant).

Turning to the results for the new controls in Table 2.A2, Feng and Zak(1999) argue that a fall in income inequality and a rise in education raise thelikelihood of democratic transitions. Whereas our analysis differs from Fengand Zak (we include economic openness, study the level of democracy, andour sample is larger), our results generally agree with theirs. In Models 1 and2, the effect of gini is negative and significant (a rise in inequality reducesdemocracy). In Models 7 and 8, the effect of education is positive andsignificant when economic development is excluded. As in the study byFeng and Zak, when economic development is included (Models 3–6),the effect of education is positive but insignificant. As they explain, thisis because education and economic development are correlated. Whenincluded with education (Models 5–8), the effect of gini is negative,as in Feng and Zak, but it is not significant, which we attribute to theaforementioned differences between our model and theirs.

The effect of INGOs on democracy is positive and significant in Models9 and 10. These results demonstrate that membership in INGOs increasesthe level of democracy in a member country, supporting the argumentsof Keck and Sikkink (1998) and Risse and Sikkink (1999). The coefficientsof the regional dummies in Model 11 measure the effect on democracy ofbeing in some region, relative to the American continent (i.e., we excludeone dummy to prevent falling into the dummy variable trap). The resultsindicate that relative to the American continent, the level of democracy inthe Middle East, Africa, and Asia is lower but roughly the same as in Europe.As could be expected, when the regional dummies are included, the effectof diffusion (which correlates with regional effects), although still positiveas in Table 2.4, is insignificant.

The effects of the world position variables on democracy are examined inModels 12 and 13. In Model 12, the effect of semiperiphery is positive and


significant, and the effect of periphery is negative and insignificant. How-ever, in Model 13, which uses the Freedom House data as in Burkhart andLewis-Beck (1994), both variables are negative and neither is significant.Our results differ from the results of Burkhart and Lewis-Beck becausethey find negative and significant effects for both periphery and semi-periphery interaction terms. We believe one reason for the difference hasto do with the sample. Their data end in 1989 whereas ours end in 1996, cap-turing post–Cold War democratic transitions in many of the semiperipherycountries (e.g., Bulgaria, Czechoslovakia). A second reason has to do withthe difference between the POLITY III and Freedom House data. As noted,the measure based on POLITY III has a larger variance than that of FreedomHouse. POLITY III focuses on the changes in institutional constraints onthe leaders of the executive branch, whereas Freedom House also covers thegradual changes in civil rights. These are also reflected in the world positionvariables, particularly semiperiphery.

THREE

Democracy, Economic Openness,

and Income Inequality

INTRODUCTION

In the previous chapter, we saw that the different aspects of national eco-nomic openness affect the level of democracy in a country. In that analysis,the level of national income inequality acted as a control variable in thestatistical model of the sensitivity analysis. We found that a rise in thelevel of income inequality could reduce the level of democracy in a coun-try. This chapter changes the causal arrow between income inequality anddemocracy, and we now focus on identifying the causes of national incomeinequality. We seek to study two interrelated important questions pertainingto this phenomenon: How does the level of economic openness of a countryaffect its distribution of income? How does the level of democracy influencea country’s distribution of income? These two questions are very importantfor social scientists because a country with a highly skewed income distri-bution tends to be politically unstable. Political instability often may leadto intrastate conflict and ultimately civil war.

So far, these two questions of interest have been addressed in two separatebodies of literature. The effect of economic openness on the distributionof income has been debated in the literature on economic globalization(e.g., Rodrik, 1997; Held et al., 1999). The issue has also been debated inpolicy and popular circles (e.g., Soros, 1997; Wolf, 2000; World Bank PovertyNet, 2000). As discussed in the previous chapter, although the exact defi-nition and scope of globalization are debated, scholars agree that currentglobalization implies growing economic openness of countries to trade,foreign direct investments (FDIs), financial capital flows, and – to a lesserextent – labor mobility (e.g., Held et al., 1999; Hughes, 2000). Existingstudies offer conflicting theoretical expectations of the effects of economic

62

Democracy, Economic Openness, and Income Inequality 63

openness on the distribution of income, but these studies do not providerigorous empirical analyses of their claims.

Meanwhile, the effect of democracy on income distribution is an impor-tant topic in the literature on democracy (Chan, 1997b). Scholars agree thatdemocracy implies a national political regime based on free elections andbroad political representation (e.g., Diamond, 1999). Existing studies arguethat democracy reduces income inequality, but empirical evidence is incon-clusive. Some studies find a negative effect of democracy on income inequal-ity, other studies find a positive effect, and a third group finds no effect.

These two bodies of literature have remained largely on separate courses.Taking a broader view, one may position economic openness and democracyat the center of liberalism. The literature on liberalism, which emphasizesthe importance of free choice, comprises two main streams (Zacher andMatthew, 1995). Republican liberalism implies that democracy, the pre-ferred political regime, reduces income inequality. Commercial liberalism(classical and neoclassical economics) argues that free market capitalismis the preferred economic regime, and income inequality is an inevitableby-product.1 Although the issue of income inequality was central for clas-sical economists, it has received relatively little attention in neoclassicaleconomics (Ferreira, 1999). We believe the relationship between incomedistribution, on the one hand, and democracy and open market, on theother, deserves more attention. We argue that the effects of these forces onincome distribution need to be studied together.

Capitalism and democracy are not easy companions. Whereas democ-racy is based on the principles of “one person, one vote” and representativegovernment, capitalism is based on the principles of laissez-faire and privateenterprise. Democracy is inherently associated with redistributive policies(e.g., progressive taxation), but capitalism typically rewards individuals withdifferent levels of income and wealth. Hence, democracy may promote in-come equality, whereas capitalism may promote income inequality. Demo-cracy and economic openness may cancel out or mediate each other in affect-ing the income distribution within a society. Since many studies2 also arguethat economic openness affects democracy, empirical studies of incomeinequality that exclude either economic openness or democracy as a causaldeterminant can incorrectly attribute the effect of one force to the other.

1 For a review article focusing on the variants of liberalism, see Zacher and Matthew (1995).2 See, e.g., Im (1996), Whitehead (1996), Drake (1998), Held et al. (1999), and Li and Reuveny

(2003).


Our empirical work in this chapter differs from previous studies in severalways. We study the effects of both democracy and economic openness onincome inequality. We examine how democracy, trade, FDI, and financialcapital inflows influence income inequality measured by the Gini coefficient.Furthermore, we pay particular attention to the possible interactive effectbetween democracy and openness. Using the World Bank data on nationalGini coefficients, we employ a sample of 69 countries over the period 1960–1996, which is larger than the samples in previous studies. Finally, weperform separate analyses for developed countries (DCs) and less developedcountries (LDCs).

Our primary findings can be summarized as follows: democracy reducesthe level of income inequality within countries; trade openness induces moreequitable income distribution within countries; foreign financial capitalinflows do not affect income inequality; and FDI leads to greater incomeinequality, but this effect is mediated and weakened in democracy.

The chapter is organized as follows. The next section discusses the effectsof democracy and economic openness on income inequality. The sectionthat follows describes the empirical research design and findings from thestatistical analysis. The last section of the chapter summarizes our findingsand discusses their implications.

EFFECTS OF DEMOCRACY AND ECONOMIC OPENNESSON INCOME INEQUALITY

Effects of Democracy

The claim that democracy promotes an egalitarian distribution of incomecan be traced back to the late eighteenth century. Many scholars argue thatdemocracy increases the opportunities for political participation, whichallows the poor to demand more equitable income distribution (e.g., Chan,1997b; Boix, 1998).3 As suffrage expands, reelection-oriented leaders areheld accountable to the voters and become increasingly attuned to theirneeds. Democratic governments are inclined to help the lower and middleclasses by adopting redistributive policies such as welfare spending, pro-gressive taxation, minimum wage laws, price subsidies, and public works.Authoritarian leaders, in contrast, are mainly accountable to a powerful and

3 See also Bollen and Jackman (1985), Sirowy and Inkeles (1991), Lappe et al. (1998), and HumanDevelopment Report (2000).


rich minority. They tend to adopt public policies that benefit this minority –policies that tend to maintain or increase income inequality.4

Democracy affects the distribution of income through the process ofcompeting pressures: the government is subject to pressure from interestgroups. By promoting a more equal distribution of political power, democ-racy gives rise to labor unions and political parties that represent the lowerand middle classes, as well as public policies that redistribute income totheir constituents. The more organized and vital these groups are, the moresuccessful they are at influencing policymaking. As described by Lenski(1966), democracy redistributes political power in favor of the majorityand, therefore, leads to policies that reduce inequality.5

Previous empirical studies of the effect of democracy on income inequal-ity exhibited several limitations. Some authors describe a few specific his-torical episodes, arguing, for example, that the expansion of suffrage inWestern Europe in the early 1900s reduced income inequality (Lindert,1994; Justman and Gradstein, 1999). When income inequality was mea-sured directly in statistical studies, the data were limited. Muller (1988), forexample, measured income inequality using the Gini coefficient (as we do),but only for 1970. Other scholars used larger samples but employed indirectmeasures of income inequality.6

The statistical evidence on the effect of democracy on income inequality ismixed. Muller (1988), Moon (1991), and Rodrik (1998) report that democ-racy reduces inequality. However, Bollen and Jackman (1985), Deiningerand Squire (1996), and Gasiorowski (1997) report that the effect of democ-racy on income inequality is statistically insignificant. Chan (1997b) reportsmixed findings, whereas Simpson (1990) argues that income inequality firstrises with democracy and then declines after a certain threshold level ofdemocracy.7

4 Some scholars argue that autocracies suppress wages to industrialize their economies (e.g.,Schamis, 1991). Others note that industrialization in autocracies can also reduce inequality(e.g., Birdsall, 1998).

5 For details, see, e.g., Lipset (1959), Muller (1988), Lindert (1989), Moon (1991), and Gasiorowski(1997). Alesina and Rodrik (1994) formalize these ideas by presenting an economic model inwhich the expansion of suffrage makes a poor individual the swing vote, leading to redistributivetaxes that provide the poor with more income.

6 For example, Chan (1997b) used government expenditures in areas expected to transfer incomefrom the rich to the poor (e.g., health, housing, and education). Gasiorowski (1997) used thegrowth rates of industrial wages to gauge changes in income inequality. Rodrik (1998) usedlabor wages as a percentage of GDP.

7 In their literature survey, Sirowy and Ingeles (1991) report that six studies find the effectof democracy on income inequality is negative and six studies find the effect is positive orstatistically insignificant.


Effects of Economic Openness

In broad terms, economic openness creates winners and losers in a society.The effects of openness on income distribution depend on the identity of thewinners and losers. We discuss these effects in four categories: internationaltrade, FDI, foreign financial capital, and international labor mobility.

TradeAccording to the Heckscher–Ohlin model of international trade, countriesexport goods that use their abundant factors of production intensively andimport goods that use their scarce factors intensively. Building on this model,Stolper and Samuelson (1941) predict that trade would raise the incomesof the owners of abundant factors and reduce the incomes of the ownersof scarce factors.8 Because DCs are relatively well endowed with skilledlabor and capital, their imports are expected to hurt their unskilled laborwhile their exports should benefit the capital owners and skilled labor. Incontrast, LDCs are relatively well endowed with unskilled labor. Therefore,their imports should hurt their capital owners and skilled labor while theirexports should benefit the unskilled labor. Hence, trade should raise incomeinequality in DCs and reduce it in LDCs.

Wood (1994) finds evidence in favor of the Stolper–Samuelson model.Robbins (1996), however, argues that wage inequality rose in many LDCs,contrary to the prediction of Stolper and Samuelson. It is also reportedthat, in many LDCs, trade shifts income to the resource-intensive sectors(Inter-American Development Bank, 1998).9

The Stolper–Samuelson theorem does not exhaust the channels throughwhich trade can affect income inequality. Rodrik (1997) argues that trademakes it easier for firms in DCs to substitute the unskilled labor with cheapimports, reducing their bargaining power and wage rates. But to the extentthat trade reduces the wages of the unskilled labor, it provides incentives forworkers to acquire education and for firms to employ more unskilled labor,which may reduce inequality (Blanchard, 2000). Furthermore, according toBirdsall (1998), trade intensifies economic competition, which reduces the

8 The Stolper–Samuelson theorem deals only with the distribution of income. It is well establishedin the trade literature that economic gains from trade outweigh economic losses.

9 Slaughter and Swagel (1997) attribute wage inequality in DCs to the technological-change biastoward skilled labor, whereas Minford et al. (1997) argue that trade and technological innovationcontribute equally to wage inequality. However, Rodrik (1997) argues that the evidence in favorof a technological-change driver of wage inequality in DCs is weak.


prices of basic consumption goods, benefiting the poor more than the richbecause the poor spend relatively larger shares of their incomes on basic con-sumption goods. The competition also diminishes the monopoly positionenjoyed by the upper class, which may reduce income inequality (Birdsall,1998). Yet another argument is that trade increases labor productivity, whichincreases wages and reduces inequality (Held et al., 1999). Finally, the win-ners from trade could compensate the losers, reducing inequality, althoughsuch compensation typically is not offered voluntarily (Rodrik, 1997;Salvatore, 1998).

Foreign Direct InvestmentFDI typically involves multinational corporations (MNCs), whose effectson income inequality are debated.10 Some analysts argue that FDI raisesinequality. First, MNCs can pressure host governments to cut welfareexpenditures and repress labor unions to reduce wages, both of which hurtthe lower and middle classes. The threat of MNCs to leave the country alsoreduces workers’ wages by weakening their bargaining power (Nafziger,1997; Salvatore, 1998). Second, MNCs are said to repatriate profits fromLDCs, leaving LDCs underdeveloped and hurting their poor (Baran, 1973;Lall, 1974; Jenkins, 1996; Nafziger, 1997). Third, the capital-intensive tech-niques used by MNCs are said to create a dual economy with a smalladvanced sector and a large backward sector, which promotes unemploy-ment among the unskilled laborers and distorts income distribution(Muller, 1979; Lall, 1985; Jenkins, 1996; Robbins, 1996; Nafziger, 1997).Fourth, MNCs are said to pay low wages in labor-intensive industries suchas footwear and clothing, and to push domestic suppliers to follow suit inorder to reduce the MNCs’ purchasing costs (Barnet and Cavanagh, 1994;Held et al., 1999). Fifth, domestic tax systems are not well suited to taxMNCs. The smaller tax base reduces government revenue and, therefore,welfare expenditures, which hurts the poor more than the wealthy (Hatzius,1997; Human Development Report, 1999).

In contrast, several studies argue that MNCs provide LDCs with capitaland technology, improve their corporate governance, and propagate bettermanagement practices. These forces, in turn, raise labor productivity and

10 FDI entails the acquisition of a lasting management interest in an enterprise operating in aneconomy other than that of the investor, or the creation of a subsidiary of a domestic firm ina foreign country. Portfolio investments cover private transactions in securities. For details, seeWorld Development Indicators (World Bank, 2002).


promote economic growth.11 It follows that FDI could eventually reduceincome inequality via the Kuznets (1955) effect, which is discussed in detailin the next section. Dollar and Kraay (2000) also support this view; theyargue that economic growth raises the income of the poor proportionallymore than that of the rich, making FDI a useful tool in reducing poverty(Stiglitz, 1998). In this spirit, Borenszstein et al. (1994) find that MNCspromote economic growth in LDCs, and Blomstrom et al. (1992) find thatFDI promotes technology transfer from DCs to LDCs. Other scholars arguethat MNCs cannot easily relocate from one country to another to reducetheir labor costs because lower wages also are associated with lower laborproductivity. Hence, labor’s bargaining power is not necessarily diminishedby FDI (Lawrence, 1994). Furthermore, the host governments can regulatethe operation of MNCs, thus controlling their effects on host economies(Vernon, 1971; Kurzer, 1993).

Relatively few studies investigate empirically the effects of FDI on wagesin LDCs (Klein et al., 2001). They find that MNCs tend to pay higherwages for skilled labor, indirectly supporting the claim that FDI promotesincome inequality by enhancing the premium on skilled labor.12 Severalstudies in the field of sociology argue that FDI promotes income inequalityby reducing the power of labor unions to pull low-wage workers up theincome ladder, thus distorting the occupational structure of LDCs andmarginalizing workers displaced as a result of the capital-intensive natureof FDI.13

A couple of other studies (Bussman et al., 2005; Sylwester, 2005) findFDI and income inequality are uncorrelated, but these studies appear tosuffer from important data problems. Bussman et al. (2005) use FDI stockover GDP, but as we explained at length in the appendix of Chapter 2, FDIstock data have serious measurement errors, making them inappropriate forcross-national comparisons. Also, using FDI stock, one conflates the short-and long-term effects of foreign investment. Sylwester (2005) uses net FDIinflows over GDP, but because he focuses on the change of income inequalitybetween 1970 and 1990 and his sample is so small (29 observations), it isnot surprising to find insignificant results.

11 See, e.g., Haddad and Harrison (1993), Coe et al. (1994), OECD (1994a), Blomstrom and Kokko(1996), Batra and Tan (1997), and Markusen and Venables (1999).

12 See case studies by Aitken et al. (1996) for Mexico and Venezuela, Feenstra and Gordon (1995)and Graham and Wada (2000) for Mexico, and Mazumdar and Mazaheri (2002) for Africa.

13 See, e.g., Sullivan (1983) and Bornschier and Chase-Dunn (1985). After empirical analysis, Tsai(1995) rejects this argument, whereas Dixon and Boswell (1996) and Alderson and Nielson(1999) support it.


Foreign Financial CapitalSeveral authors argue that to attract foreign financial capital, governmentsengage in liberal reforms (e.g., they reduce taxation and government expen-ditures, privatize state-owned enterprises, and deregulate markets). Thesereforms may hurt labor and increase income inequality (Strange, 1996;Germain, 1997; Held et al., 1999). A related issue is that financial opennesstends to bring about volatile financial flows across countries, increasingthe frequency and magnitude of financial crises. These crises typically hurtthe poor more than the wealthy (Human Development Report, 1999). As thefinancial crisis ensues, the economy enters a recession and the rate of un-employment rises while the tax base erodes, public budgets are slashed, andwelfare expenditures decline. Since the wealthy can better weather reces-sions than the poor, the poor suffer more and income inequality increases.Some statistical evidence (Quinn, 1997) indicates that financial opennessraises income inequality.

A competing view argues that financial openness does not necessarilyimply the end of the welfare state. Financial markets allocate funds to placesthat need them the most, as they pay higher rates of return (Wooddall, 1995).Foreign capital also allows countries to consume more than they produceand to invest more than they save, promoting economic development andreducing income inequality (Nafziger, 1997). Financial openness forces gov-ernments to become more prudent, efficient, and accountable to the public,because financial markets penalize corrupt, deficit-oriented, inflationary,and inefficient governments with higher interest rates and currency crises(Wooddall, 1995). Better governmental policymaking can reduce incomeinequality by improving tax systems, property rights, and public welfareprograms (Held et al., 1999). Foreign money also can be used to fund socialsecurity. Consequently, governments can spend more of their budgets onpoverty reduction (Normand, 1996).

International Labor MobilityLabor mobility across countries (migration) also affects the distributionof income within countries. We consider this empirically complex topicto be outside the scope of our chapter and defer it to future research. Inthe past several decades, international labor flows have generally been fromLDCs to DCs. Accordingly, let us consider a two-country, North–Souththeoretical framework. The North is endowed with more skilled labor, andless unskilled labor, than the South. The wages for skilled labor are higherthan they are for unskilled labor. It follows that unskilled labor migrationfrom South to North reduces income inequality in the South and raises it


in the North (O’Rourke, 2001). Contemporary migration to DCs typicallyinvolves unskilled labor from LDCs. Some immigrants from LDCs havejoined the middle class in DCs. However, immigrants from LDCs typicallytake on the lower-paying jobs in the DCs. As a result, the unskilled workersin DCs can move to better-paying jobs (Held et al., 1999). It follows that theeffect of labor migration on income inequality is theoretically ambiguous.Empirically, contemporary international labor migration generally has beenrestricted and small in size, and its effects on wages have been marginal andmixed.14

Hypotheses

The literature argues that democracy reduces income inequality, as inHypothesis H1, which follows. However, because some empirical studiesreport that the effect of democracy on inequality is not statistically signifi-cant, Hypothesis H1 may not be supported in our empirical test. In contrastto that of democracy, the overall (or net) effect of economic openness onincome inequality is theoretically ambiguous. To focus our analysis, we positdirectional Hypotheses H2–H4 on the effect of openness. Since the com-peting theories on the effects of openness all seem plausible, HypothesesH2–H4 may be rejected in our empirical test.

H1: Democracy reduces income inequality.H2: Trade increases income inequality in DCs and reduces it in LDCs.H3: FDIs reduce income inequality.H4: Financial capital inflows reduce income inequality.


Similar to the previous chapter, this section first presents our statisticalmodel for the empirical analysis and then discusses several research designissues. The section then presents key results from the empirical analysis.The discussion in this section is self-contained and does not require anyspecific statistical expertise. Interested readers seeking further technicaldetails should consult the chapter’s appendix.

14 For a survey of empirical results, see Ichino (1993). For results from the United States, Canada,and Australia, see Borjas (1987, 1990, 1993). Labor migration flows were much higher in thelate 1800s and early 1900s than today. Data from 1870 to 1913 show that inequality fell inthe originating poor European countries and rose in the destination New World countries(Williamson, 1997).


Empirical Model

To test the effects of democracy and economic openness, we specify andestimate the following statistical model of income inequality. As noted,it provides a structure for us to guide the statistically uninitiated readersthrough the empirical exercise.

income inequalityt = �0 + �1tradet−l + �2fdit−1

+ �3portfoliot−1 + �4democracyt−1

+ �5gdppct−1 + �6gdppc2t−1

+ �7past inequalityt−1 + εt . (3.1)

The model specification is based on the theoretical discussion and theliterature on income inequality. As in the previous chapter, we denote vari-ables with small capital letters and their coefficients with Greek notation.Each coefficient indicates the effect of the independent variable on thedependent variable, or the phenomenon we seek to explain. The notationεt denotes the random error that is not explained by the statistical model.The variable subscripts t and t − 1 indicate the time period of the variable,where t represents the current period and t − 1 the previous time period(a lagged variable). To simplify the presentation, we refer to the variableswithout the time subscripts t or t − 1.

In the model, income inequality within countries is the dependentvariable (the phenomenon we explain). The right-hand side of the modelincludes independent variables that explain income inequality. The variableincome inequality represents the level of income inequality in a country ina time period. An ideal measure of income inequality is the Gini coefficient,which is computed on the basis of national income surveys. As noted,previous studies have used alternative measures of income inequality dueto the lack of comprehensive Gini data. We take advantage of the incomeinequality data collected by Deininger and Squire (1996). However, eventhough their data set has the most comprehensive coverage, it contains manymissing values because national income surveys are never conducted yearafter year. To address this problem, we follow the practice of Easterly (1999)and Higgins and Williamson (1999), and we use the decade averages of theGini data for each country. Although the technical details of the Gini dataare presented in the appendix, it is worth noting that the Gini coefficientranges between 0 and 100, where 100 denotes perfect income inequality andzero denotes perfect equality. To apply the appropriate statistical technique,we transform the Gini coefficient into an unbounded measure.


To test the effects of both economic openness and democracy on incomeinequality in the same model, we include four key independent variables:trade, fdi, portfolio, and democracy. Corresponding to the incomeinequality measure, these variables take on average yearly values duringeach decade. As in the previous chapter, we measure economic opennessby using flows of trade, fdi, and portfolio investments. The importanceof these flows to a country depends on their magnitude relative to the sizeof the domestic economy. Accordingly, we divide each of these flows byGDP. The trade variable measures the share of trade within a country’snational economy. The variable fdi reflects the share of foreign productioncapital in the country’s national economy, and portfolio is the shareof foreign financial capital in the country’s national economy. Also, as inthe previous chapter, democracy measures the level of democracy in acountry. It ranges from –10 (the most autocratic regime) to +10 (the mostdemocratic regime) and combines information on both the democratic andthe autocratic characteristics of the political regime.

The model also controls for the effects of economic development andpast inequality. Kuznets (1955) hypothesizes that below some level of GDPper capita, income inequality rises with GDP per capita; above this level,income inequality declines with GDP per capita. This pattern is known as theKuznets curve. Previous empirical results on the Kuznets curve are mixed.Ahluwalia et al. (1979) and Higgins and Williamson (1999), for example,found evidence supporting the Kuznets hypothesis; Deininger and Squire(1998) found no supporting evidence. We use gdppc and gdppc2 to capturethe Kuznets effect. gdppc represents GDP per capita, expressed in purchas-ing power parity – adjusted international dollars – and gdppc2 is gdppcsquared. If the Kuznets curve is correct, the coefficients of gdppc and gdppc2

should be positive and negative, respectively. The level of education and theshare of agriculture in GDP, which can also affect income inequality, areindirectly included in the model. Both variables tend to be highly correlatedwith GDP per capita.

The level of past inequality is measured by the one-decade lagged Ginicoefficient. The inclusion of this variable is consistent with the observed ten-dency of inequality to persist over time. Several theoretical reasons accountfor this tendency. First, wealth concentration typically correlates posi-tively with political influence, generating arrangements that favor wealthowners.15 Second, people tend to marry those from the same socioeconomicgroup. Consequently, the children of the rich (or poor) group remain in

15 For example, in 1950, 65% of all the agricultural land in Latin America was held by 1.5% of thefarm owners. This inequality has increased since (Birdsall, 1998).


the original group, perpetuating income differences across groups. Third,in cases where the poor and the rich belong to different ethnic groups,racial discrimination can institutionalize the current income distribution(Lewis, 1994). Fourth, education can promote upward social mobility, butacquiring education is costly. Poor people tend to have more children thanrich people (Heerink, 1994). Thus, education spending per child tends tobe smaller for the poor, ensuring a vicious circle. The poor remain lesseducated and earn less, and income inequality persists (Dasgupta, 1993).


As in the previous chapter, to justify the validity of our analysis we highlightseveral key design issues that require some clarification. As before, the relatedtechnical details of these issues are relegated to the appendix. First, εt in themodel is the so-called error term, that is, the part of the variations in incomeinequality that are not explained by the model. For statistical inferences tobe valid, the error term needs to satisfy some assumptions. We address therelated assumption violations using appropriate econometric techniques.

Second, the sample includes 69 countries over the period 1960–1996.Since economic openness and the level of democracy are generally higherin DCs than in LDCs, we examine the robustness of the results from the fullsample in separate samples for DCs and LDCs. The DCs in our analysis aremembers of the Organization for Economic Cooperation and Development(OECD), whereas the LDCs are non-OECD countries.

Third, we design our research to measure the net effects of economicopenness and the level of democracy on income inequality. We do notintend to measure the strength of each of the competing forces discussed inthe previous section. To assess the effect of each of the causal mechanismswould require a very large data-collection effort that would include vari-ables such as wages, international technology transfers, government welfareexpenditures, breakdown of fdi and trade by sectors, taxes, governmenttransfer payments, and various attributes of political parties. The collectionof these data is better deferred to future research.

EMPIRICAL FINDINGS

Table 3.1 presents the statistical results for the full sample, the DCs, and theLDCs.16 The models exhibit reasonably good explanatory power, and the

16 The number of countries in our full sample is smaller than the number of countries in theDeininger and Squire (1996) data, due to missing data points for the other variables in ourmodel.


Table 3.1. Income inequality, democracy, and economic openness

All countries LDCs DCs

DEMOCRACY −0.0125∗∗∗ −0.0112∗∗∗ −0.0125∗∗

(0.0033) (0.0037) (0.0056)TRADE −0.0013∗∗ −0.0013∗∗ −0.0026∗∗

(0.0006) (0.0008) (0.0013)PORTFOLIO 0.0074 0.0340 0.0059

(0.0166) (0.0367) (0.0157)FDI 0.0632∗∗∗ 0.0518∗∗ 0.0590∗∗∗

(0.0229) (0.0290) (0.0218)GDPPC −1.91e-06 3.92e-05∗∗ 8.51e-07

(1.08e-05) (2.20e-05) (1.54e-05)GDPPC2 1.90e-10 −2.98e-09∗∗ 2.07e-10

(4.40e-10) (1.28e-09) (4.85e-10)PAST INEQUALITY 0.7181∗∗∗ 0.7163∗∗∗ 0.4901∗∗∗

(0.0590) (0.0693) (0.0755)Constant −0.1307∗∗∗ −0.1898∗∗∗ −0.2718∗∗∗

(0.0402) (0.0510) (0.0682)Observations 142 99 43Adjusted R2 0.69 0.62 0.52

Note: Standard errors in parentheses. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significantat 1%.

effects of the control variables are largely consistent with our expectations.The results provide us with confidence in the models. As before, we leavethe discussion of the results of the control variables to the appendix.

We begin with the result for democracy. As shown in the table, demo-cracy reduces income inequality, and the effect is consistent in all threesamples (all countries, DCs, and LDCs). When using better data of incomeinequality than previous studies of the effect of democracy on inequality –including economic openness in the model – and controlling for the Kuznetscurve and past inequality, our analysis provides empirical support for theexpected negative effect of democracy on income inequality.

As Table 3.1 indicates, trade openness also reduces income inequalityin all three samples. The results support Hypothesis H2, with regard toLDCs but not DCs. As discussed earlier in the chapter, trade generatesboth inequality-increasing and inequality-decreasing effects. Hence, ourfindings can be interpreted as representing the net effect of trade on incomeinequality within countries, which, in turn, reduces income inequality.

In contrast to the effects of democracy and trade openness, fdi inflowsraise income inequality in all three samples (all countries, DCs, and LDCs).


These results reject Hypothesis H3. Again, fdi can generate both inequality-increasing and inequality-decreasing effects. According to our results, thenet effect of fdi is to raise income inequality.

In all three samples, the effect of portfolio inflow on income inequalityis positive. But the effect is never different from zero statistically, meaningthat the effect is too weak to be different from what one would expectdue to chance. These results are consistent with the observation that therise in portfolio investment inflows is a relatively recent phenomenon. Iffinancial market integration continues to deepen, portfolio investmentsmay significantly affect income inequality in the future. In any case, ourresults do not support Hypothesis H4.

How do the statistical findings in Table 3.1 inform us about the sub-stantive significance of the effects of democracy, trade openness, and fdiinflows? Similar to what we did in the previous chapter, we first compute abaseline Gini coefficient by setting all variables in Model 1 of Table 3.1 attheir respective means. We then raise democracy, trade, and fdi by onestandard deviation at a time and compute Gini again. We illustrate the sizesof these effects in two ways. First, we compute the percentage change inGini across the two scenarios, showing the absolute influence of these threevariables on Gini. Second, we compute the ratio of each of the changes inGini over the range between the average of the Gini maxima (42.93) and theaverage of the Gini minima (38.98) across the 69 countries in the sample(see Appendix 3.1 for the country statistics). This computation indicatesthe influence of changes in trade, democracy, or fdi on Gini relative tothe variability of Gini in our sample.

In absolute terms, a rise in democracy by one standard deviation aboveits mean reduces Gini by 1.5% (from 40.56 to 39.06). A rise in tradeopenness by one standard deviation reduces Gini by 1.39% (from 40.56 to39.17). A rise in fdi by one standard deviation raises Gini by 2.17% (from40.56 to 42.73). In relative terms, these changes are 38%, 35%, and 55%over the range of Gini values between the average of the Gini maxima andthe average of the Gini minima in the sample for democracy, trade, andfdi, respectively. These estimated effects of democracy, trade, and fdi onincome inequality are substantial.17

17 Alternatively, one could express the sizes of these effects in terms of their shares out of onestandard deviation of Gini (9.77). This would give 15.46%, 14.22%, and 22.21% for democracy,trade, and FDI, respectively, which are also substantial. A third way is to examine the size of oureffects as shares of the range between the Gini sample minimum (22.46) and sample maximum(65.38). This range understates the size of our effects, because the majority of the observationsin the sample are far from the tails of the Gini distribution.


Our results indicate that trade reduces income inequality, whereas fdiraises it. This result may reflect the fact that trade is often associated withwelfare programs that compensate the losers from trade, whereas fdi typi-cally is not associated with such programs. Additionally, the benefits of tradetend to diffuse throughout the domestic economy, benefiting all consumers.In contrast, the benefits of fdi tend to be concentrated in the industries orthe individual firms (in the case of joint ventures) in which the invest-ment takes place. Such concentration of benefits tends to increase incomeinequality.

In summary, our results support Hypothesis H1 that democracy reducesincome inequality. We find that trade openness reduces income inequality,supporting Hypothesis H2 with regard to LDCs, but not DCs. fdi increasesincome inequality, rejecting Hypothesis H3. Foreign financial capital inflowsreduce income inequality, which does not support Hypothesis H4.

Additional Analyses

We conduct several additional analyses to complement the main resultsreported earlier. As in the previous chapter, we report the details of theseanalyses in the chapter appendix. The first additional analysis is substantiveand policy-relevant. In Table 3.1, we find that fdi increases income inequal-ity and that democracy reduces it. As fdi flows into many democracies,an important policy question is whether democratic governance helps tomitigate the positive impact of fdi on inequality. If democratic institutionshelp reduce the negative impact of foreign capital on the host country, weuncover another benefit of spreading democracy, acquire more confidencein our ability to harness the powerful forces of economic globalization, andbring good news to the many democracies of the world. On the other hand,if democratic governance does not reduce the impact of foreign capital onincome inequality, we should worry about how we could take advantage ofthe benefits of globalization and minimize its drawbacks. The interactionbetween fdi and democracy is thus an important empirical question. Toevaluate this effect, we create an interaction term between fdi and democ-racy and reestimate the models in Table 3.1 with this additional interactionterm.

Table 3.A1 reports the results for these interaction models. The fdi∗

democracy interaction term does not have any significant effect in theall-countries sample or the LDC sample, but it has a significant negativecoefficient in the OECD sample. Specifically, as democracy rises from −1.7to +10 within the OECD sample, the significant total effect of fdi declines


from about 0.30 to about 0.057. fdi increases income inequality across therange of all possible democracy values, but its effect shrinks dramaticallyas we move from Greece and Spain in the early years of the OECD intothe mature democracies in the OECD world. This effect, however, does notexist in many other democracies in the developing world.

Second, we evaluate the robustness of the main results in Table 3.1by using alternative measures of income inequality, excluding the laggedincome inequality from the model, and employing alternative estimators.The results of these analyses are reported in Table 3.A2. Overall, the effectsof democracy and economic openness remain robust. democracy reducesincome inequality. portfolio investment inflows do not produce any sig-nificant effect on inequality in any of the six models. fdi inflows alwaysraise income inequality. trade openness is correlated with less inequalityin all cases, and the effect is weak in all but the case of the country fixed-effects estimator. We believe that this insignificant result reflects the notedlimitations of the fixed-effects estimator.

Finally, one may wonder if the effects of economic openness and demo-cracy on income inequality are different in size between the DC and LDCsamples. We estimate a statistical model to compare the effects of the vari-ables on the two samples; the results are in Table 3.A3. The effect of demo-cracy remains the same between the two samples, as do the effects oftrade, portfolio, and fdi. Interestingly, the effects of the Kuznets curveand past inequality differ between the two samples.


Our analysis demonstrates the need to study the effects on income inequalityof economic openness and democracy together. Better theoretical explana-tions of inequality should consider economic openness, democracy, andtheir interaction. Our statistical findings indicate that trade openness tendsto be associated with more equitable income distribution within countries,but FDI leads to greater income inequality, and foreign financial capitalinflows do not have any significant effect on income inequality. A rise indemocracy reduces the level of income inequality within countries. In addi-tion, democratic governance mediates and weakens the undesirable impactof FDI on income inequality in the OECD world but not in DCs.

One may wonder why trade and FDI, often highly correlated with eachother, exert different effects on inequality. It is important to note that tradeand FDI do not coincide perfectly, because they are generally operatedby different actors and are often driven by different costs and benefits.


Had they perfectly correlated, we could not have estimated their individualeffects statistically. Theoretically, as discussed earlier, trade and FDI ofteninfluence the distribution of income through different causal mechanisms.Also, government policies regulating and responding to trade and FDI tendto be different. Trade adjustment assistance has a much longer history andis much more common and widespread.

What are the public policy implications of our findings? To the extent thattheir goal is to bring about a more equitable distribution of income, inter-national organizations such as the International Monetary Fund, the WorldBank, and the World Trade Organization may condition new member-ships, continued memberships, or various assistance programs on improve-ments of democracy. Industrialized democracies that are interested inreducing income inequality in other countries may also make liberal politicalreforms a condition of foreign aid. In DCs, democracy enables people toappeal to all levels of their governments, which then regulate and makesure FDI is not damaging to income distribution. In LDCs, whereas demo-cratic governance by itself reduces income inequality, it does not restrainthe impact of FDI. People in LDCs are not as effectively organized andempowered politically. As their governments invite MNCs and give thempreferential treatment, it comes at the expense of the poor, causing an evenmore skewed income distribution. Importantly, because most FDI flowsamong DCs and a much smaller part goes to LDCs, the democratic DCsare in a win-win situation: they are able to enjoy the benefits of FDI andneutralize one of its bad side effects (skewing income distribution). Incontrast, LDCs are likely unable to grapple politically with this side effectof FDI because their democratic institutions may be too weak to make adifference.

Both proponents and critics of economic openness should reassess theirclaims if their arguments hinge solely on income inequality. Economicopenness may improve or worsen income equality, depending on its type.Policymakers seeking a more equitable income distribution could promoteinternational trade while compensating the losers from trade openness. Onaverage, it appears to be the case that governments tend to compensate thosewho lose from trade openness, preventing income inequality from gettingworse. Policymakers may also take measures to reduce the negative effects ofFDI. For example, governments may offer subsidies or tax breaks to sectorsor regions without FDI and curtail the occurrence of FDI-induced dualeconomy. At the same time, governments need to be careful not to imposerestrictions that entirely drive away FDI. Although the inequality-increasingeffect of portfolio investments in our results is statistically insignificant, we


believe that the consistently positive coefficient across samples may indicatea need for caution on the part of governments considering liberalization ofthis type.

Finally, our results indicate that income inequality declines with eco-nomic development, but it also tends to persist over time. Governmentsinterested in reducing income inequality need to design welfare programsthat help the impoverished. Financial aid from DCs and international orga-nizations to LDCs should be designed to reduce poverty. Regardless of theexact policies taken, we believe that reducing income inequality is impor-tant. If history is any guide, failure to reduce income inequality may wellresult in domestic and international political instability.

SUMMARY AND OUTLOOK

Scholars have investigated the effects of economic openness and democracyon income inequality in two separate bodies of literature. In studies ofdemocracy, scholars agree that democracy reduces income inequality, butthe empirical evidence is mixed. In the literature on globalization, the effectsof economic openness on income inequality are debated, but they have notbeen examined rigorously. To the best of our knowledge, our analysis is thefirst systematic statistical study of the effects of both economic openness anddemocracy on income inequality. We argue that the two effects should bestudied together. In other words, the exclusion of one variable or the othercan lead to incorrect inferences of the determinants of income inequality.

Our empirical analysis comprised 69 countries over the period 1960–1996. We have focused on the effects of trade, FDI, foreign financial capitalinflows, and democracy on income inequality. We find that a higher level ofdemocracy reduces the level of income inequality within countries. Tradeopenness is associated with more equitable income distribution withincountries, FDI is associated with greater income inequality, and foreignfinancial capital inflows have no statistically significant effect on incomeinequality. All the empirical findings in this chapter are found to be robustacross different measures of income inequality, alternative statistical esti-mators, and model specification.

In this chapter, we investigated the relationship between economic open-ness and democracy, on the one hand, and income inequality, on the other.In this analysis, economic development acted as a control variable. Inthe following chapter, we shift our attention to the relationship betweeneconomic development and the level of democracy. Is economic develop-ment a prerequisite for democratization? Is democratization good for the


economy? These two questions have generated large bodies of literaturefocusing on either economic development or democracy as the dependentvariable. Our interest in the next chapter is to analyze the potentially recipro-cal relationship between democracy and development that has been largelyignored in previous research.

APPENDIX


As before, this appendix accompanies the section “Empirical Model andAnalysis” in the main text of the chapter. We provide various details on themeasurement of our variables, their data sources, issues pertaining to thetechnical design of the statistical model, and the expected effects of variouscontrol variables. We also report and discuss all the results for the controlvariables as well as various additional analyses.

Empirical Model

The dependent variable – income inequality within countries – is measuredvia the Gini coefficient, which is computed on the basis of the Lorenzcurve. In this curve, the cumulative percentage of income held by shares ofsociety appears on the y-axis and the percentage of population that holdsthe particular income share appears on the x-axis. The line that is 45 degreesfrom the origin denotes perfect income equality (e.g., 10% of the peoplehold 10% of national income). However, perfect income equality is neverobserved empirically, and so the Lorenz curve is below this line. The Ginicoefficient measures the area between the 45-degree line and the Lorenzcurve and is expressed as the percentage of the area between the x-axis andthe line. Thus, a Gini coefficient of 100 denotes perfect income inequality,whereas a Gini coefficient of 0 denotes perfect equality.18

The Gini coefficient is an ideal measure of income inequality. As noted,previous studies have used alternative measures of income inequality dueto the lack of comprehensive Gini data. We take advantage of the incomeinequality data collected by Deininger and Squire (1996). However, whereastheir data set has the most comprehensive coverage, it contains missingvalues. To that effect, as in Easterly (1999) and Higgins and Williamson

18 For details, see, e.g., Nafziger (1997) and World Development Indicators (World Bank, 2002).


(1999), all the variables in the model are computed as decade averages.Hence, the unit of analysis is country decade.19

Deininger and Squire (1996) note that although their Gini data generallyare of good quality, differences in data sources and data analysis meth-ods across countries may cause measurement errors in a pooled sample.20

Using these data, Easterly (1999) applies all the available information fromDeininger and Squire (1996) in computing the Gini decade averages, underthe assumption that the measurement errors are randomly distributed andare captured in the statistical model’s error term. We use the data computedby Easterly. To examine the robustness of our Gini-based results, we alsoemploy the share of income held by the top 20% of the national populationas an alternative measure of income inequality.21

Because the Gini coefficient is bounded between 0 and 100, using ordinaryleast squares (OLS) regression could be problematic (OLS assumes that thedependent variable is unbounded). The usual practice is to transform thebounded variable into an unbounded indicator. We transform the Ginicoefficient into an unbounded measure using the formula log[Gini/(100 −Gini)].22 We also assess the robustness of results using the untransformedGini coefficient.

As noted, we measure economic openness by using flows of trade, fdi,and portfolio investments. We divide each of these flows by GDP to capturetheir importance in a country’s domestic economy. The economic opennessdata are taken from the World Development Indicators (World Bank, 2002).As in the previous chapter, democracy is the level of democracy in a countryand is computed on the basis of the POLITY III data set (Jaggers and Gurr,1995; Gurr and Jaggers, 1999). This widely used data set provides twoindices of political regime characteristics. The 10-point democracy index

19 The Deininger and Squire data include 111 countries, covering the period from 1890 to 1996,but the data coverage is very limited before 1960. The Gini decade averages we use werecomputed by Easterly (1999) and cover four decades (1960s to 1990s). These data are availableat www.worldbank.org/research/growth/ddlife.htm.

20 For example, the Gini coefficient may be computed from the level of income or the level ofexpenditure. In the case of income, individual welfare may be measured before or after paymentof taxes. The aggregation of individual levels of welfare may vary across countries in the mixingof individual and household units.

21 Appendix 3.1 provides the mean, minimum, and maximum values for the Gini data for eachcountry in the sample. The country with the highest mean level of inequality is South Africa(Gini coefficient = 62.3), and the country with the lowest mean level of inequality is Hungary(Gini coefficient = 25.17).

22 Log denotes natural logarithm. The transformed variable equals −∞ for Gini = 0 and +∞ forGini = 100. See Pindyck and Rubenfeld (1991) for technical details.


(DEMOC) measures the democratic characteristics of the regime. The 10-point autocracy index (AUTOC) measures the autocratic characteristics ofthe regime. Mansfield and Snyder (1995), Londregan and Poole (1996), andothers observed that many governments may have both democratic andautocratic characteristics. Consequently, these studies measure the level ofdemocracy as the difference between DEMOC and AUTOC, a practice wealso adopt here. The democracy measure ranges between −10 (the mostautocratic regime) and 10 (the most democratic regime). In the model,gdppc is GDP per capita, expressed in purchasing power parity–adjustedinternational dollars.


Several design issues require further clarification. First, we employ a pooledtime-series, cross-sectional research design. The sample includes 69 coun-tries over the period 1960–1996. Statistical models for pooled time-seriescross-sectional data may exhibit heteroskedasticity and serial correlation.Although these problems do not bias the estimated coefficients, they couldresult in inefficient and biased standard errors for the coefficients. To dealwith these potential problems, we estimate the model using OLS regressionwith Huber–White robust standard errors clustered over countries. Theseestimated standard errors are robust both to heteroskedasticity and to a gen-eral type of serial correlation (Wiggins, 1999). The inclusion of the laggeddependent variable in the model helps to absorb temporal dependence inthe data, as shown by Beck and Katz (1995a, 1995b).

Second, in addition to the substantive reasons noted in the main text,the inclusion of past inequality in the model helps to control for the effectof potentially relevant but omitted structural variables, such as the ethnicand demographic structures of society. Many studies adopt this model-ing strategy to prevent spurious findings (e.g., Bollen, 1979; Burkhart andLewis-Beck, 1994; Muller and Seligson, 1994; Muller, 1995).

Third, as noted, we design our research to measure the net effects ofeconomic openness and the level of democracy on income inequality. As istypically done, the statistical significance levels of the estimated coefficientsare investigated with a one-tailed t-test since our hypotheses are signed.

Fourth, we need to clarify how the size of the effect is computed for thestatistically significant key variables. The coefficient sizes cannot be inter-preted linearly from Table 3.1, because the dependent variable is a nonlineartransformation of Gini. As noted, we first compute a baseline Gini valueby setting all variables in Model 1 of Table 3.1 at their respective means.


We then raise democracy, trade, and fdi by one standard deviationat a time and compute Gini on the basis of the following equation:log[Gini/(100 − Gini)] = a + bx, where a is a constant, b is a vectorof coefficients, and x is a vector of independent variables. We solve for Giniin terms of a, b, and x. Next we illustrate the sizes of these effects in twoways, as noted in the main text.

EMPIRICAL FINDINGS

Table 3.1 presents the estimated coefficients and their standard errors fromthe full sample, the DC sample, and the LDC sample.23 All the models inTable 3.1 have a relatively good fit to the data, with the adjusted R2 rangingfrom 0.52 (in the DC sample) to 0.69 (in the all-country sample). The jointsignificance level of the model’s independent variables from the F-test isstatistically significant at a level better than 1% in all the samples, whichsupports our model specification.24

The effect of democracy on income inequality is statistically significantat the 5% level for the DC sample and at the 1% level for the LDC andall-country samples. The effect of trade openness on income inequalityis negative and statistically significant at the 5% level for all the samples,indicating that trade openness reduces income inequality. The effect of FDIinflows on income inequality is positive and statistically significant at the 5%level for the DC and LDC samples and at the 1% level for all countries. Theeffect of portfolio inflow on income inequality is positive in all samples,but it is never statistically significant.

The effects of gdppc and gdppc2 on income inequality are statisticallysignificant in the LDC sample at the levels of 5% and 1%, respectively. gdppchas a positive effect on income inequality, whereas gdppc2 has a negativeeffect. These results support the existence of a Kuznets curve for LDCs. In theDC and all-country samples, the Kuznets curve is not statistically significant.The insignificance of the Kuznets curve for DCs is to be expected. TheWestern European countries experienced a Kuznets curve transformation

23 The number of countries in our full sample is smaller than the number of countries in theDeininger and Squire (1996) data due to missing data points for the other variables in ourmodel.

24 We also check for high collinearity using the variance inflation factor. We find no high collinearityin the sample of LDCs, but high collinearity in the DC and all-country samples. Using the matrixof variance decomposition, we identify the sources of high collinearity to be gdppc and gdppc2.These variables are correlated by construction, a quality shared by all models that includethe Kuznets curve. Nevertheless, this result suggests caution in interpreting the Kuznets curvecoefficients for the DC and all-country samples.


Table 3.A1. Interactive effect of democracy and FDI on income inequality

All countries LDCs DCs

DEMOCRACY −0.0123∗∗∗ −0.0097∗∗ 0.0041(0.0040) (0.0053) (0.0130)

FDI 0.0637∗∗∗ 0.0545∗∗ 0.2680∗∗

(0.0266) (0.0299) (0.1063)FDI∗DEMOCRACY −0.0001 −0.0011 −0.0211∗∗

(0.0022) (0.0027) (0.0105)TRADE −0.0014∗∗ −0.0013∗ −0.0026∗∗

(0.0007) (0.0008) (0.0013)PORTFOLIO 0.0074 0.0352 0.0063

(0.0167) (0.0378) (0.0159)GDPPC −1.92e-06 0.00004∗∗ 2.85e-06

(0.00001) (0.00002) (0.000015)GDPPC2 1.91e-10 −3.07e-09∗∗ 1.41e-10

(4.42e-10) (1.38e-09) (4.71e-10)PAST INEQUALITY 0.7183∗∗∗ 0.7181∗∗∗ 0.4870∗∗∗

(0.0583) (0.0684) (0.0718)Constant −0.1305∗∗∗ −0.1889∗∗∗ −0.4523∗∗∗

(0.0407) (0.0516) (0.1176)Observations 142 99 43

Note: Standard errors in parentheses. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significantat 1%.

in the late nineteenth and early twentieth centuries (Kuznets, 1955). Sincethis period is not included in our sample, we do not detect a Kuznets curveeffect for DCs.

The effect of past inequality on current inequality is statistically sig-nificant at the level of 1% in all samples. The positive sign of this effectcorroborates our expectation that income inequality exhibits inertia. Therealization of equitable income distribution is a lengthy process; hence, oneshould take a long-run view of the dynamics of income inequality. Therelatively large size of the inertia effect in our results suggests that models ofincome inequality that do not control for this effect may omit an importantvariable.

Additional Analyses

The first additional analysis concerns the interactive effect of democracyand fdi on income inequality. Table 3.A1 presents the results. The totaleffect of fdi is (0.0637 − 0.0001∗democracy) for all countries, (0.0545 −0.0011∗democracy) for LDCs, and (0.2680 − 0.0211∗democracy) for


DCs. We test the significance of the total effect of fdi across the differentvalues of democracy for each sample. The total effect of fdi is significantonly in the DC sample throughout the values of democracy.

The second additional analysis focuses on the robustness of our resultsin Table 3.1. We use alternative measures of income inequality, exclude thelagged income inequality from the model, and employ alternative estima-tors. The results for the all-country sample are presented in Table 3.A2.The results for the LDC and DC samples are consistent with those in Table3.A2 and are not reported here.25 Overall, the effects of democracy andeconomic openness in Table 3.1 are replicated and robust across the sixexperiments reported in Table 3.A2.

The models presented in Table 3.1 employ a transformed Gini coefficientas the dependent variable. To assess the robustness of these results, weuse alternative measures of inequality. In Column 1 of Table 3.A2, incomeinequality is measured as the share of national income held by the top20% of the population. The data for this particular measure come from thestudies of Deininger and Squire (1996) and Easterly (1999). In Column 2 ofTable 3.A2, we use a nontransformed Gini coefficient, as in Easterly (1999)and Muller (1988).26

The inclusion of the lagged Gini coefficient in the models presented inTable 3.1 may reduce the variance to be explained by other variables. In col-umn 3 of Table 3.A2, we exclude the lagged Gini coefficient for a robustnesscheck. Recall, however, that the lagged Gini is part of our theoretical modelspecification.

The results reported in Table 3.1 are generated from OLS regression withrobust standard errors. Columns 4–6 of Table 3.A2 present the results fromtwo other estimators: fixed effects with robust standard errors, and randomeffects. We employ two versions of the fixed-effects estimator, introducingcountry or decade dummies into the model. The fixed-effects estimatorcontrols for the possibility that some countries or decades are marked bylarge income inequality. As noted in the literature, the fixed-effects estimatorhas serious limitations. First, the country and decade dummy variables areatheoretical and absorb many of the variations in the dependent variable thatare attributable to the model’s independent variables.27 Second, the countrydummy variables and other country-dependent variables (e.g., GDP per

25 Statistical results for the LDC and DC samples, as well as the by-country statistics for Gini andthe sample descriptive statistics, are available from the authors on request.

26 As noted, the transformation of the Gini coefficient is needed, since it is bounded. However,this transformation may introduce nonlinear effects and so is not used in some studies.

27 As summarized by King (2001), most methodologists agree that using better models, bettermeasures, and robust estimation is preferable to fixed-effects estimation.

Tabl

e3.

A2.

Inco

me

ineq

ualit

y,de

moc

racy

,and

econ

omic

open

ness

(all

coun

trie

s)

Top

20%

Un

tran

sfor

med

Wit

hou

tpa

stC

oun

try

fixe

dD

ecad

efi

xed

Ran

dom

inco

me

Gin

iin

equ

alit

yef

fect

sef

fect

sef

fect

DE

MO

CR

AC

Y−0

.003

2∗∗∗

−0.3

004∗∗

∗−0

.012

5∗∗−0

.011

1∗−0

.013

0∗∗∗

−0.0

147∗∗

∗

(0.0

010)

(0.0

797)

(0.0

073)

(0.0

076)

(0.0

036)

(0.0

043)

TR

AD

E−0

.000

2∗∗−0

.031

6∗∗−0

.001

4∗−0

.002

5−0

.001

3∗∗−0

.001

2∗∗

(0.0

001)

(0.0

153)

(0.0

010)

(0.0

026)

(0.0

006)

(0.0

007)

PO

RT

FOLI

O0.

0041

0.21

96−0

.001

40.

0096

0.00

400.

0057

(0.0

035)

(0.3

847)

(0.0

233)

(0.0

157)

(0.0

151)

(0.0

188)

FDI

0.01

03∗∗

∗1.

4552

∗∗∗

0.08

84∗∗

∗0.

0586

∗∗0.

0566

∗∗∗

0.06

86∗∗

∗

(0.0

029)

(0.5

495)

(0.0

318)

(0.0

335)

(0.0

223)

(0.0

187)

GD

PP

C2.

15e-

06−1

.50e

-05

−4.1

9e-0

5∗−3

.96e

-05∗

2.78

e-06

−5.0

2e-0

6(2

.18e

-06)

(0.0

003)

(2.5

4e-0

5)(2

.62e

-05)

(1.4

2e-0

5)(1

.56e

-05)

GD

PP

C2

−7.4

7e-1

13.

37e-

098.

61e-

101.

27e-

09∗

−4.8

9e-1

13.

17e-

10(9

.12e

-11)

(1.0

5e-0

8)(1

.06e

-09)

(8.1

1e-1

0)(5

.81e

-10)

(6.2

8e-1

0)PA

STIN

EQ

UA

LIT

Y0.

7352

∗∗∗

0.71

41∗∗

∗0.

0290

0.72

66∗∗

∗0.

6033

∗∗∗

(0.0

521)

(0.0

629)

(0.1

839)

(0.0

588)

(0.0

616)

Con

stan

t0.

1166

∗∗∗

11.0

658∗∗

∗−0

.202

1∗∗−0

.139

9−0

.133

7∗∗∗

−0.1

635∗∗

∗

(0.0

252)

(3.2

733)

(0.0

950)

(0.1

323)

(0.0

505)

(0.0

560)

Obs

erva

tion

s10

414

214

214

214

214

2A

dju

sted

R2

0.75

0.68

0.25

0.92

0.71

0.70

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗

∗si

gnifi

can

tat

5%;∗

∗∗si

gnifi

can

tat

1%.

86


capita) tend to be collinear. The random-effects estimator parameterizesthe error term associated with different cross sections and is useful forcross-nationally dominated panels.28

Table 3.A2 suggests that the results from Table 3.1 are robust againstthe exclusion of the lagged dependent variable, alternative estimators, andalternative measures of income inequality. In all cases, the effect of democ-racy is statistically significant and negative, reducing income inequality.The effect of portfolio investment inflows is insignificant in all six mod-els, as in Table 3.1. The effect of fdi inflows is always statistically significantand positive, again as in Table 3.1. The effect of trade openness is negativein all cases, reducing inequality, and is statistically significant in all but thecase of the country fixed-effects estimator. We believe that this insignificantresult reflects the noted limitations of the fixed-effects estimator.

The control variables in Table 3.A2 replicate the spirit of Table 3.1. Theeffect of past inequality is positive in the five cases reported and is signif-icant in all cases except for the country fixed-effects estimator. The effectsof gdppc and gdppc2 are consistent with Table 3.1 in four of the six cases.Columns 3 (without past inequality) and 4 (country fixed effects) are theexceptions. The weaker results obtained without past inequality supportour model specification. An argument against using the lagged dependentvariable is that it causes insignificant results. Our results from the modelincluding the lagged dependent variable, however, are more significant thanthose from a model without this variable. Since we also have theoretical rea-sons for including past inequality in the model, we believe Column 3 suffersfrom model misspecification bias. In Column 4, the effects of gdppc andgdppc2 are significant, but their signs are not consistent with the Kuznetscurve. We believe this weaker result reflects the aforementioned limitationsof the fixed-effects estimator.29

Finally, one may question whether the effects of economic openness anddemocracy on income inequality are statistically different between the DCand LDC samples. To answer this question, we constructed two variables foreach independent variable in the original model. For example, instead of onedemocracy variable, we now have two separate variables: oecd democracyand ldc democracy. We estimated a full model of all variables thus created

28 An alternative estimator is the panel corrected standard error (PCSE; Beck and Katz, 1995b).Because our sample is cross-nationally dominated with only four decades, the PCSE estimatoris not suitable for our application. We thank Neal Beck for this comment.

29 We have also estimated the model in Table 3.1 with a dummy variable for communist regimestatus, because these countries tend to emphasize income equality. The results for the hypothesistesting are similar to those reported in Table 3.1.


Table 3.A3. Effects of democracy and economic openness in DCs and LDCs

CoefficientDCs LDCs equality test

DEMOCRACY −0.0142∗∗∗ −0.0113∗∗∗ 0.26(0.0047) (0.0038)

TRADE −0.0027∗∗ −0.0012∗ 1.02(0.0012) (0.0008)

PORTFOLIO 0.0052 0.0346 0.54(0.0151) (0.0370)

FDI 0.0591∗∗∗ 0.0516∗∗ 0.04(0.0205) (0.0293)

GDPPC −8.36e-06 0.00004∗∗ 5.10∗∗

(0.00001) (0.00002)GDPPC2 5.20e-10 −3.06e−09∗∗∗ 8.74∗∗∗

(4.71e-10) (1.27e-09)PAST INEQUALITY 0.4868∗∗∗ 0.7155∗∗∗ 4.89∗∗

(0.0756) (0.0703)Constant −0.1945∗∗∗

(0.0490)Observations 142Adjusted R2 0.69

Note: ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at 1%.

and report the results in Table 3.A3. We then tested whether the coefficientof, say, oecd democracy equals that of ldc democracy, and we did so foreach variable. These equality test results are in the last column of Table 3.A3.The test results indicate that the sizes of the effects of economic opennessand democracy on income inequality, although statistically significant inboth DC and LDC samples, are not significantly different across the twosamples.

FOUR

Democracy and Development

INTRODUCTION

The effect of democracy on economic development has captivated manythinkers. De Tocqueville (1835) and Schumpeter (1942), for example, be-lieve democracy provides the social securities required for development.Hayek (1944) argues that democracies exhibit relatively less unrest andpolitical instability, promoting development. Friedman (1962) asserts thatdemocracy promotes development by keeping state power in check. North(1990) expects autocrats to prey on their subjects, and Olson (1993) arguesthat autocrats are corrupt and promote their cronies, which are bad fordevelopment. Going a step further, Fukuyama (1992) argues that capitalismand democracy work in unison to promote welfare, a view many observersshare (e.g., The Economist, 1994; Cohen, 2007).

In contrast, Marx (1867, 1871) and Lenin (1911) argued that democ-racy came out of capitalist development as a tool for the elite to depriveand appease the masses. John S. Mill (1860), a supporter of democracy,is skeptical about its ability to promote development. Huntington (1968)argues that democracies exhibit high state expenditures in response to theircitizens’ demands, reducing the excess that is expendable for development.Olson (1982) argues that special interests in democracy shape public policyto promote their own interests, undermining the economy. Wade (1996)argues that strong autocrats can resist popular pressure for policies to savedoomed industries, which is good for development. Barro (2000) and Buenode Mesquita and Downs (2005) argue that autocrats can effectively pursueprodevelopment policies.

Despite the controversy, the view that democracy promotes developmenthas recently become prevalent in the policy circles of the developed world.For example, a number of senior practitioners recently argued that poor

89


countries undertaking democratic reforms outperform poor autocracies indevelopment. The United States should, therefore, target its aid to help coun-tries that attempt democratic reforms. President Bush, the authors note,supports having democratic governance “as qualifying criteria for countriesto receive assistance” (Siegle et al., 2004). Similarly, Bush’s Under Secretaryof State Dobriansky argues that solid democratic institutions are the basis foreconomic development (U.S. Department of State, 2007), and Bush himselfapparently believes that political and economic liberties, both of which pro-mote social welfare, are the natural by-products of each other (Cohen, 2007).

Democracy, as a regime favoring equal political rights and civil liberties,is definitely worthy of promotion. We have seen that many writers andpolicymakers justify the promotion of democracy as a means to encourageeconomic development, but does a rise in democracy really lead to a higherlevel of development? The argument that democracy promotes develop-ment considers neither the possibility that democracy could underminedevelopment nor the possibility that the causality may flow from develop-ment to democracy. In fact, as we show in the next section, the effect ofdevelopment on democracy has received ample scholarly attention.

The issue of causation stands at the center of this chapter. We arguethat sorting out the causal direction between democracy and economicdevelopment is important. For example, if the goal is to promote democracybut the causation goes from development to democracy, we should notcondition the distribution of aid to poor countries on instituting democraticreforms; rather, we should condition aid on forces such as economic reformsthat encourage development or on the quality of the projects themselves.

The relationship between democracy and development raises two salientquestions. Does development lead to democracy? Does democracy encour-age or undermine development? Together, these questions imply that thetwo forces may affect each other in a reciprocal manner. But statistical studiestypically employ single-equation models, which specify either democracy asthe dependent variable and development as the key independent variable,or development as the dependent variable and democracy as the key inde-pendent variable. As several scholars noted recently, these single-equationmodels cannot address the possible reciprocal relationship between democ-racy and development, and their inferences may be misleading (Chan, 2002;Midlarsky, 2002; Przeworski and Limongi, 2003).1

1 Chan (2002: 118) warns that even if some results indicate that democracy boosts develop-ment, “the relationship may well be the other way around [development boosts democracy].”Midlarsky (2002: 672) calls “for the greater use of structural equations modeling” in the

Democracy and Development 91

The development–democracy reciprocity issue has so far received verylimited attention. Burkhart and Lewis-Beck (1994) tested the causal-ity between democracy and energy consumption per capita by usingGranger’s method, but they did not specify a simultaneous equationsmodel (SEM). Przeworski et al. (2000) used dynamic probit to study theeffect of development on regime transition and a selection-corrected esti-mator to assess the effect of regime type on economic growth, but theydid not specify a full SEM. Helliwell (1994) and Feng (2003) employedSEMs, but their analyses raise a number of concerns, to which we willreturn.

Given the significance of and the inadequate attention to the issue, fur-ther analysis of the simultaneity between democracy and development iswarranted. Some readers may view this issue as a methodological nuisance,not a theoretical problem. We consider it to have both theoretical andmethodological implications. If democracy and development affect eachother, unidirectional models are theoretically incomplete and, as Alvarezand Glasgow (2000) noted, will produce biased estimates.

We construct a SEM of the levels of development and democracy. Byemploying this approach, we are able to decide which of the followingfour theoretically distinct possibilities holds: (1) development affects demo-cracy, but not the other way around; (2) democracy affects development,but not the other way around; (3) the two forces affect each other; and(4) the two forces are unrelated. We also identify the signs and sizes of thestatistically significant effects. We estimate this model using a large-N sampleof countries across nearly four decades. The empirical results suggest thata rise in development in a country increases its democracy, whereas a risein democracy reduces development. These findings have important policyimplications.

The rest of this chapter is organized as follows. The next section discussesthe theoretical arguments concerning the relationship between democracyand development, and the following section provides an overview of pre-vious empirical efforts and raises several important but overlooked designissues. The fourth section discusses our statistical model and presents themain empirical findings. The last section summarizes the findings andexplores their implications. As in previous chapters, technical details andadditional results are delegated to the chapter appendix.

development–democracy literature. Przeworski and Limongi (2003: 348) argue that “infer-ences based on standard [single-equation] regression models [of democracy and development]are invalid.”


THEORETICAL ARGUMENTS

This section provides the theoretical context for our analysis rather than anexhaustive review of two very large bodies of literature. We demonstrate thata large number of arguments exist for both claims: economic developmentinfluences democracy, and democracy affects development. Acknowledgingthese plausible arguments, one should entertain the theoretical possibilitythat development and democracy affect each other simultaneously.

Causality from Development to Democracy

The hypothesis that economic development promotes democracy hasreceived wide attention in the literature. Lipset traces this so-called mod-ernization thesis back to the Greek philosopher Aristotle. “From Aristotledown to the present,” he writes, “men have argued that only in a wealthysociety in which relatively few citizens lived in real poverty could a situationexist in which the mass of the population could intelligently participate inpolitics and could develop the self-restraint necessary to avoid succumbingto the appeals of irresponsible demagogues” (1959: 75).

Development arguably promotes democracy through several channels.First, as economy develops, the need for an educated labor force risesand people can afford to spend more on education. Educated people, inturn, often demand greater government transparency, political freedom,and participation in policymaking (Diamond et al., 1987; Lipset et al.,1993), which are all elements of democracy. Educated people also tend tobe more receptive to the idea of resolving political disagreement throughnonviolent, majority-based decision making. They also demand freer flowsof information, more mass media, and communication outlets, which assistthe diffusion of democratic ideas (Lipset, 1959; Dahl, 1989; Diamond andPlattner, 1994).

In a second channel, economic development promotes civic society andsocioeconomic progress, which increase the power of the middle class at theexpense of the ruling elite and constrain authoritarianism (de Tocqueville,1835; Lipset, 1959; Huber et al., 1993; Putnam et al., 1993). Also, as thenumber of urban dwellers rises, the masses can more easily get organized todemand political rights (Lipset, 1959; Lipset et al., 1993).

Third, underdevelopment and poverty cause discontent, accentuategrievances, and polarize societies. These tensions increase the attractive-ness of violence as a means to secure survival and acquire social mobil-ity, which can increase political instability and even lead to civil war. In


trying to assert power, the state may become oppressive, reducing democ-racy. Economic development promotes democracy by alleviating these ten-sions (Im, 1987; Huntington, 1991; Haggard and Kaufman, 1995).

A fourth channel considers the appeal of autocracy to the elite. Politicalleaders constantly contemplate whether to accept the outcome of a demo-cratic competition that mandates the distribution of the national pie or tofight to establish autocratic control so as to acquire a larger share of the pie.Assuming the marginal utility of consumption and the marginal return ofcapital decrease with affluence, the attractiveness of dictatorship is expectedto wane when affluence rises. Also, as the economy and civil society develop,defeated politicians have more opportunities for respectable employment,enabling their acceptance of electoral defeats. Because economic develop-ment increases affluence and supplies more acceptable alternatives, dictator-ship becomes less attractive as a political solution to distributional conflicts(Lipset, 1959, 1981; Przeworski et al., 2000).

Finally, poverty and underdevelopment erode public confidence in theability of nascent democracies to resolve problems, reducing the legitimacyof the regime and increasing the appeal of authoritarianism to the public.When developmental problems become salient and pressing, the democraticgovernment may lose public support altogether, raising the likelihood of aforceful transition to autocracy. Development, in contrast, alleviates thesepressures and enables democracy to take hold and consolidate (Linz, 1978;Seligson and Muller, 1987; Huntington, 1991; Haggard and Kaufman, 1995).

The positive effect of development on democracy, however, is not the onlypossible outcome. Development may delay democratization in two ways.First, economic development may strengthen an authoritarian government(Bueno de Mesquita and Downs, 2005); development provides more finan-cial resources for the autocrat, not only the middle class, particularly whenthe government owns some means of production. Because a financiallystronger government can sustain a larger security force and develop bettermonitoring and surveillance methods, it becomes more capable of sup-pressing the grassroots efforts toward democracy.

Second, as long as the economy develops strongly, the public in an autoc-racy may not always be interested in changing the status quo. If the publicvalues robust economic development more than democratic norms suchas freedom of speech, freedom of association, or freedom of political par-ticipation, economic development may serve to maintain the autocraticregime.

Some stylized observations support the argument that development maynot bring about democratization. China, for example, has been vigorously


developing its economy since the late 1970s, but it remains a one-party state.Since the late 1990s, Russia has achieved considerable economic develop-ment, but various reports suggest its level of democracy is on the decline.Strong developments in Japan from the late nineteenth century to 1945 andin Germany from 1933 to 1945 coincided with highly undemocratic yetpopular regimes. Finally, despite Singapore’s stellar economic developmentrecord, the country is still autocratic.

Causality from Democracy to Development

Theoretical considerations suggest the effect of democracy on developmentcan be negative or positive. Starting with the negative impact, one chan-nel argues democracies are bogged down by special interests and rent-seeking, which hinders development. Special interests use the democraticpolitical process to redistribute wealth from the public to themselves, andrent-seeking allocates resources to unproductive activities. A development-oriented autocrat may restrain the overbearing rent-seeking interests(Olson, 1982; Lal, 1983; Jackman, 1993). In this vein, regimes based onalliances between the military and technocrats may be efficient in promotingdevelopment, because the technocrats devise required policies and the deter-mined military regime implements them (O’Donnell, 1979; Cohen, 1985).

In a second channel, free people may be reluctant to curtail their con-sumption in order to save. Low savings, in turn, leads to low investments,which slows down development. Democratic governments, depending onthe electoral support to stay in office, are often afraid to impose unpopu-lar measures to increase savings and investments. Depending relatively lesson public support, autocrats can implement unpopular policies to boostsavings and investments (Hewlett, 1979; Rao, 1985).

Third, democratic governments tend to spend much on social welfareprograms because the public values them. Massive government expendi-tures, in turn, crowd out resources for development projects. Democraciesmay also have to confront strong populist pressures to redistribute wealthfrom the rich to the poor in various ways, reducing the willingness of thewealthy to save and invest. Authoritarian leaders may be better able to wardoff these pressures, encouraging economic development (Haggard, 1990;Barro, 2000).

In a fourth channel, democracies are less able to suppress social unrestthan autocracies because they have to protect civil liberties. Social unrest, inturn, slows down production, undermines effective business planning, and


scares away investors. Because it is more capable of enforcing social order,an autocratic regime can promote economic development better than ademocratic government (Pye, 1966; Hewlett, 1980).

Finally, less developed countries (LDCs) feature agrarian societies, highincome inequality, youth-oriented demographics, low levels of technology,underdeveloped financial markets, and dependence on international capitaland technology. In such situations, the state necessarily must play an activeand interventionist role in promoting development, suggesting that author-itarian decision making may outperform the diffused democratic process(Chirot, 1977; Cohen, 1985).

In contrast, several arguments indicate that democracy can facilitatedevelopment. First, political freedom leads to economic freedom. Free mar-kets, in turn, allocate financial resources and factors of production to theirmost efficient use, bringing about development (Goodin, 1979; Goodelland Powelson, 1982).

Second, democracy respects individual rights, including private propertyrights. Secure property rights, in turn, are critical for trade, savings, andinvestments, which promote development (Goodell and Powelson, 1982;North, 1990; Olson, 1993; Clauge et al., 1996; Li and Resnick, 2003). Auto-crats may also secure property rights, but democratic leaders are more likelyto do so since they are subject to checks and balances (North and Weingast,1989). Autocrats may abruptly change the rules of the game, increasinguncertainty and, therefore, the willingness to invest. Autocrats may also bemore predatory than democratic leaders, because they have fewer incentivesto consider public welfare, as opposed to their own interests and those oftheir cronies (Goodell, 1985; Findlay, 1990; Olson, 1991).

Third, because democratic regimes are more responsive to public con-cerns and more likely to solve problems through political compromise, theymay experience less domestic unrest than autocracies. Because democraticgovernments represent the public, they may seek to achieve relatively moreequitable income distributions than autocracies. All these forces work todefuse public grievances and reduce the likelihood of social strife and revo-lution. Political stability, in turn, leads to a longer time horizon for economicactivities, increasing the allocation efficiency and investments required fordevelopment (Przeworski, 1985; Barro, 2000; Feng, 2003).

Fourth, autocrats’ military spending needs to be relatively more excessiveto sustain control over their population. Consequently, they need to extractmore resources from society through tax or pillage, which hinders economicdevelopment. In addition, as autocrats allocate most of their energy to


maintaining social order and guarding against possible coups that might bestaged by their opponents or the public, they are less likely to be effective atmodernizing their economies (Nordlinger, 1970; Bienen, 1971).

Finally, the autocrat, who requires the support of the ruling elite, maybe subject to rent seeking, corruption, and patron–client relationships, allof which undermine development. Whereas autocrats have more decision-making power, they still make bad decisions. For example, capital-intensivedevelopment strategies, which autocrats have often chosen, may producedual economies in LDCs, which is detrimental for development. In contrast,rent-seeking in democracy need not be as severe as in autocracy, and demo-cratic deliberation does not necessarily lead to bad public policy (Goodin,1979; King, 1981; Nafziger, 2006).

THE EMPIRICAL LITERATURE

The previous section discussed various theoretical conjectures on the effectsof democracy and development on each other. Taken together, these argu-ments suggest that democracy and development affect each other simul-taneously. Moreover, because all the preceding arguments are theoreticallyplausible, the net effects of development and democracy on each othercould be positive, negative, or insignificant. To the extent that theory can-not definitively calibrate the strengths of the causal forces, the exact natureand strength of the reciprocal relationship between democracy and devel-opment is an empirical issue.

This section overviews empirical findings in the literature and discussesa number of critical design issues. In general, some stylized facts supportLipset’s (1959) modernization thesis. For example, today the developedcountries (DCs) are all democratic, and those DCs with the most recenthistory of autocratic rule – Greece, Portugal, Spain, and Turkey – are also theleast developed among the DCs. In Taiwan, South Korea, and several othernewly industrialized countries, a move toward greater democracy followedeconomic development, whereas some of the worst democratic records arefound in LDCs. That said, stylized facts rejecting the modernization thesisalso exist (e.g., China and Russia) and, more generally, stylized facts maynot generalize across many countries.

Statistical studies of democracy typically employ single-equation models.Development is typically measured by real gross domestic product (GDP)per capita (RGDPPC) and democracy is typically measured as a continu-ous variable. Most of these studies find that a rise in RGDPPC increases


democracy.2 Other studies model democracy as a discrete event wheredemocracy grows above some predefined threshold. The results of thesestudies are somewhat mixed.3

Turning to the reverse causality, analysts study the net effect of democracyon the growth rate of RGDPPC or GDP. The findings are mixed. In Borneret al.’s (1995) review, 10 studies find an insignificant or mixed effect ofdemocracy on growth, 3 find a negative effect, and 3 find a positive effect. InKurzman et al.’s (2002) survey, 19 studies report a positive effect, 6 reportnegative effects, 10 report insignificant effects, and 9 present mixed results.In Feng’s (2003) review, 4 studies find insignificant effects, 6 find positiveeffects, and 6 find negative effects. More recently, Krieckhaus (2004) alsoreports mixed results.

Consider next the issue of simultaneity. If development and democracyindeed affect each other, the single-equation analysis is problematic andlikely produces biased results. So far, only two studies have addressed thisissue using a SEM, but their findings and designs differ. Helliwell (1994)found that a rise in real income per capita promotes democracy, whereasa rise in democracy does not affect economic growth. Feng (2003) foundthat a rise in economic growth does not affect democracy, whereas a rise indemocracy promotes economic growth.

The models of Helliwell and Feng raise three design issues. First, Helliwelluses the 1960 values as instruments for endogenous variables in 1985,but a SEM requires instruments based on all the exogenous variables inthe model. Second, both models are cross-sectional, where variables areaverages over more than two decades. However, as Feng notes, this designcannot study change over time, which is important in the relationshipbetween democracy and development. Third, both studies focus on theeconomic growth rate, not the level of development, but the modernizationthesis stresses the importance of the level of development in democracy,not growth. If one accepts the premise that democracy and developmentmay affect each other, the democracy equation ought to include the level ofdevelopment on the right-hand side, and the development equation shouldemploy this variable as the dependent variable to maintain simultaneity

2 The findings of Lipset et al. (1993), Burkhart and Lewis-Beck (1994), Gasiorowski (1995),Londregan and Poole (1996), and Li and Reuveny (2003) support the modernization thesis. Forcontrary evidence, see Gonick and Rosh (1988) and Colaresi and Thompson (2003).

3 For example, Przeworski et al. (2000) find that development does not affect democratic transi-tion but promotes democratic consolidation post-transition. Boix and Stokes (2003) find thatdevelopment promotes transition and consolidation.


between the two equations.4 Also, we note that growth and development tellus different things. An economy can grow fast and be underdeveloped (e.g.,China), whereas DCs can exhibit slow growth (e.g., Japan). Growth maychange erratically, but development changes slowly. Furthermore, growthand change in development also differ. Using RGDPPCt for development intime t, the change in development is RGDPPCt − RGDPPCt−1, where t − 1is the previous period. Growth is (RGDPPCt − RGDPPCt−1)/RGDPPCt−1.Economies may display the same change in RGDPPC but grow at differentrates due to different values of RGDPPCt−1 in the denominator of thegrowth expression. As we demonstrate clearly in the chapter appendix,inferences based on a growth–democracy SEM differ from those basedon a development–democracy SEM. We believe that any SEM model ofthe development–democracy nexus needs to seriously address these designissues.


Although theoretical expectations indicate a possible simultaneous relation-ship between democracy and development, the empirical evidence obtainedso far is mixed. The designs of previous studies that attempted to addressthe issue of simultaneity raise concerns, but we think their logic in recogniz-ing the need for a SEM is sound. This section presents our statistical SEM,related research design issues, and main empirical findings. As in previouschapters, technical details of the model, measures, and data sources arepresented in the chapter appendix.

Empirical Model

We first need to consider the measurement of democracy as it affects ourmodeling strategy. The issue is not which features of democracy or autocracyshould be recorded – which is important in its own right – but whetherto model democracy as a discrete event indicating that democracy existsonly above some threshold or as a continuous variable that ranges fromfull autocracy to full democracy. The discrete measure leads to models ofthe probability of democratic transition, whereas the continuous measure

4 If Y and Z are endogenous, the Y equation in a SEM needs to include Z on the right-hand sideand the Z equation needs to include Y on the right-hand side. Specifying a democracy equationas a function of development and an economic growth equation as a function of democracydoes not create a SEM.


leads to models of the level of democracy. We seek to study the reciprocalrelationship between the level of development and the level of democracyrather than the event of meeting a predefined threshold for democracy,because development may interact with the whole spectrum of democracy,not just with the discrete event of democratic transition.5

We specify and estimate a SEM to test the net effect of democracy ondevelopment and the net effect of development on democracy. The modeldistinguishes between endogenous and exogenous (or control) variables.An endogenous variable appears on the left-hand side of one equation andthe right-hand side of the other. The exogenous variables appear only onthe right-hand side of each equation. As in the previous chapters, smallcapital letters denote variables, and Greek letters denote coefficients to beestimated empirically. A coefficient indicates the effect of the independentvariable on the dependent variable. The notations �c and �t denote vectorsof coefficients. The Greek letters εt and ut represent the random errors in thetwo equations, respectively. The subscripts t and t − 1 denote the currentand previous time periods (or years), respectively. We lag the exogenousvariables for reasons discussed in the appendix. To simplify presentation,we refer to variables in the text without the time subscripts. Equations(4.1) and (4.2) present the development and democracy equations of theSEM, respectively, specified based on the democratization and developmentliterature.

RGDPPCt = �0 + �1democracyt + �2prior rgdppct−1

+ �3investmentt−1 + �4population growtht−1

+ �5educationt−1 + �6instabilityt−1

+ �7tradet−1 + �ccountry fixed effects

+ �tyearly fixed effects + ut, (4.1)

democracyt = �0 + �1rgdppct + �2prior democracyt−1

+ �3tradet−1 + �4diffusiont−1 + �5inflationt−1

+ �6economic growtht−1 + �7year + εt. (4.2)

The dependent variable in Equation (4.1), rgdppc, is the level ofeconomic development, measured as the logarithm of real GDPPC in

5 For the alternative discrete event approach, see Przeworski et al. (2000) and Boix and Stokes(2003).


Table 4.1. Variables and expected effects in the simultaneous equations

Democracy (DEMOCRACY) equation Development (RGDPPC) equation

Variable Sign Variable Sign

Endogenous EndogenousRGDPPC Positive or negative DEMOCRACY Positive or negative

Exogenous ExogenousPRIOR Positive PRIOR RGDPPC Positive

DEMOCRACYYEAR Positive INVESTMENT PositiveTRADE Positive or negative TRADE Positive or negativeDIFFUSION Positive POPULATION Negative

GROWTHGROWTH Positive or negative EDUCATION PositiveINFLATION Positive or negative INSTABILITY Negative

FIXED EFFECTS Positive or negative

purchasing power parity (PPP) adjusted terms. The dependent variable inEquation (4.2), democracy, is the level of democracy of a country, mea-sured as in the previous two chapters. The specification of the developmentequation follows the economic development literature, and the specificationof the democracy equation follows Chapter 2 (and the comparative politicsliterature).

Table 4.1 summarizes the relationships among the endogenous variablesand lists the expected effects of the exogenous variables. The variable priorrgdppc is gdppc from the previous period; investment is a measureof the amount of physical capital in the country; trade is a measure ofeconomic openness, given by the share of a country’s total trade in thenational economy; population is the population growth rate of a country;education is the level of human capital in a country; instability is a binaryvariable indicating the occurrence or absence of revolution or coup d’etatin a country; and country fixed effects is a group of country indicatorsthat capture country-specific heterogeneity, whereas yearly fixed effectsis a group of year indicators that capture year-specific heterogeneity.

In the democracy equation, prior democracy is the first lag of the levelof democracy; year is a year counter that captures the possible linear trendin democracy; trade is the same variable appearing in the developmentequation; diffusion is the average democracy score of countries withinthe region around each country; economic growth is the yearly percentgrowth rate of real GDP, PPP adjusted; and inflation is the yearly inflationrate in a country.



Several research design issues require clarification; we discuss their details inthe chapter appendix. First, to assess and generalize the overall relationshipbetween democracy and development, we employ a pooled time-seriescross-sectional (TSCS) design. The unit of analysis is the country-year, andthe empirical sample includes 127 countries from 1961 to 1996, includingboth DCs and LDCs.

Second, stylized facts suggest that LDCs have generally experienced widevariations in their democracy levels during our time period, whereas DCsgenerally had consistently higher levels of democracy that were quite stableover time. To evaluate the effect of this difference on inference, we conductanalyses for two samples. One sample includes all the countries for whichwe have data, both DCs and LDCs. A second sample contains only LDCs,excluding members of the Organization for Economic Cooperation andDevelopment, which we identify as DCs.

Third, to assess whether any statistical basis exists for suspecting a simul-taneous relationship between development and democracy, we employ aso-called exogeneity test. This test evaluates if democracy is not affected bydevelopment (thus, exogenous in the development equation) and if devel-opment is not affected by democracy (thus, exogenous in the democracyequation). The results of this test are not meant to replace a full SEM analysisbut to provide some indication of the simultaneity and whether a SEM isnecessary from a statistical perspective.

Fourth, to estimate a SEM, we need to consider the mathematical issueof identification and the statistical issue of inference. We handle the issueof identification by using the familiar exclusion criterion. We handle theissue of inference in the presence of simultaneity between development anddemocracy by employing the two-stage least squares (2SLS) estimator. Fifth,we conduct an additional statistical analysis to evaluate the robustness ofour results when we use a different statistical estimator – the three-stageleast squares (3SLS) estimator, which is more efficient.

Sixth, the economic development equation controls for cross-countryheterogeneity for theoretical reasons. Empirically, one may model such het-erogeneity as random or fixed effects. The choice between the two methodsis based on the so-called Hausman test. Seventh, the economic developmentequation also includes the lagged dependent variable for theoretical reasons,which could lead to biased statistical results when used in conjunction withthe country fixed effects. We check the robustness of our findings by using avariant of the 2SLS estimator, which we refer to as the 2SLS-Kiviet estimatorthat addresses this problem.


Eighth, the error term of a statistical model needs to satisfy certainassumptions for the inferences to be valid. We address the related assump-tion violations by using appropriate econometric techniques. Finally, it ispossible that other right-hand-side variables in the model will be affectedby democracy or development. We handle the possible endogeneity of theother variables by lagging them one year, a widely employed practice in theliterature.

Empirical Findings

We first conduct the exogeneity test, seeking to establish a statistical basisfor our SEM. The results indicate we should reject that democracy is exoge-nous in the development equation or that development is exogenous inthe democracy equation. They demonstrate the need for a SEM setup ifone intends to model correctly the relationship between democracy anddevelopment.

Table 4.2 reports the 2SLS estimation results for two samples: one forall countries and the other for LDCs only. We begin with the results forthe democracy equation. In column 1, for the sample of all countries, thecoefficient of economic development, rgdppc, is positive and significantlydifferent from zero. A rise in the level of economic development promotesthe level of democracy, which supports Lipset’s modernization thesis. Doesthis hold in the LDCs? In column 3, for the sample of LDCs only, the effectof a rise in the level of economic development on the level of democracyremains positive and significantly different from zero. Thus, the positiveimpact of development on democracy is not an artifact of sample choice.

How large is the effect of economic development on the level of democ-racy in a country? The coefficient of rgdppc (the logarithm of real GDPper capita) for the sample of all countries indicates that when the real GDPper capita rises from the estimation sample minimum of 416 PPP-adjustedreal international dollars (a very-low-income country) to the estimationsample mean of 3,994 PPP-adjusted real international dollars (a low-tomid-income country), the level of democracy increases by 0.48 unit on the20-point democracy scale (from −10 to +10). This change in the level ofdemocracy seems small, but it only represents the immediate impact of arise in economic development on democracy.

In our SEM, a rise in economic development in the current period notonly affects the level of democracy in the current period but also con-tinues to affect the next period democracy via prior democracy in themodel. The long-run effect of some change in development on democracy


Table 4.2. Democracy and development, 2SLS

Democracy Development Democracy DevelopmentDCs and LDCs DCs and LDCs LDCs LDCs

RGDPPC 0.2127∗∗∗ 0.1901∗∗∗

(0.0497) (0.0624)PRIOR DEMOCRACY 0.9237∗∗∗ 0.9150∗∗∗

(0.0094) (0.0011)DIFFUSION 0.2863∗∗∗ 0.3586∗∗∗

(0.0888) (0.107)INFLATION 0.00002 8.58E-6

(0.00001) (0.00002)ECONOMIC −0.0058 −0.0060

GROWTH (0.0088) (0.0097)YEAR 0.0195∗∗∗ 0.0286∗∗∗

(0.0040) (0.0054)TRADE −0.0014∗∗ 0.0003∗∗∗ −0.0009 0.0003∗∗∗

(0.0008) (0.00007) (0.0010) (0.00008)PRIOR RGDPPC 0.9537∗∗∗ 0.9558∗∗∗

(0.0096) (0.0065)DEMOCRACY −0.0006∗∗ −0.00059∗

(0.0003) (0.0004)INVESTMENT 0.0016∗∗∗ 0.0019∗∗∗

(0.0003) (0.0004)POPULATION −0.0064∗∗∗ −0.0068∗∗∗

GROWTH (0.0023) (0.0026)EDUCATION 0.0024 0.0021

(0.0023) (0.0033)INSTABILITY −0.0099∗∗∗ −0.0103∗∗∗

(0.0044) (0.0046)Constant −40.6822∗∗∗ 0.3516∗∗∗ −58.5994∗∗∗ 0.0000

(7.6661) (0.0626) (10.1188) (0.0000)Observations 2914 2914 2230 2230

Note: Standard errors in parentheses. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at1%.

accumulates over time; we explain this computation in detail in theappendix. Suppose real GDP per capita rises from 416 to 3,994; the long-run change in democracy amounts to 6.30 units on the 20-point democracy.This increase in the level of democracy is much larger than the immedi-ate impact of development, covering about one-third of the autocracy–democracy range of the democracy scale.

Does a rise in the level of democracy in a country influence its level ofdevelopment? In Table 4.2, the coefficient of democracy in the development


equation is negative and significantly different from zero for both the all-countries sample and the LDC sample. Hence, a rise in the level of democracyin a country reduces its level of economic development, all other things beingequal. The negative effect of democracy on development is not sensitive tothe sample choice.

How large is the effect of democracy on economic development? Con-sider the following scenario. Suppose an autocratic country, whose level ofdemocracy is−10, experiences a rise in democracy to+6, a level that scholarsoften use to indicate a country’s transition into a democracy. How much doesthis democratization affect economic development? Since democracy is notlogged and rgdppc is logged, the percent change in real GDP per capita dueto a change of 16 units in democracy is given by 100 × (Change in demo-cracy) × (democracy Coefficient), which is −0.96% (i.e., 100 × 16 ×−0.0006 in percent). Hence, an autocracy whose level of democracy risesfrom −10 by 16 units can expect to see its real GDP per capita fall by almost1% in real terms in that year, all other things being equal. This immediatechange is substantively not small.

How large is the change in the long run? The change in democracy in thedevelopment equation affects real GDP per capita immediately (computedearlier) and continues to affect it in the next period through its effect onprior rgdppc in the model. The long-run effect of some change in democ-racy accumulates over time. Based on the computation illustrated in theappendix, the long-run percent change in GDP per capita due to a changeof 16 units of the democracy score is −20.73%. That is, a country whosedemocracy rises from −10 to 16 will see its real GDP per capita fall byalmost 21% in the long run, all other things being equal.

The results for the control variables in both the development equationand the democracy equation are largely consistent with the theoreticalexpectations stated in Table 4.1. These expectations, in turn, are based onprevious studies. This increases our confidence in the results from our SEM.We discuss the results of the control variables in the appendix.

Additional Analysis

As planned, we conduct two additional analyses. First, we employ the 3SLSestimator. Second, we employ the 2SLS-Kiviet estimator. Overall, the resultsfrom these analyses are consistent with those in Table 4.2, both in terms ofthe reciprocal effects between democracy and development and in terms ofthe control variable in the two equations. The additional analyses suggestthat development promotes democracy, and yet democracy reduces the level


of development. The robustness of our findings gives us more confidencein our analysis.


Scholarship on democracy and development has called for study of thereciprocal relationship between democracy and economic development.So far, almost all studies have employed single-equation models. The twoimportant exceptions to this practice did not evaluate the possible reciprocalrelationship within the context of a structural SEM that endogenizes thelevel of economic development and the level of democracy in a pooled TSCSlarge-N design.

This chapter is the first to study the reciprocal relationship between thelevels of economic development and democracy in a pooled TSCS researchdesign using a SEM. The structure of the development equation followsdirectly from neoclassical economic growth theory, and the specification ofthe democracy equation resembles the one used in many studies of demo-cratization. The large-N TSCS design captures both cross-sectional andtime-series patterns. The results are robust across subsamples of countriesand alternative SEM estimators. To summarize, in both the all-countriessample and the sample of LDCs only, we find that a rise in the level ofeconomic development in a country leads to a rise in its level of democracy,whereas a rise in the level of democracy in a country reduces its level of eco-nomic development. The democracy–development causality goes two ways.

Our use of the SEM methodology reveals a tension in the democracy–development relationship, which has important policy implications. Ourresults suggest that the wealth level of a country influences the creation andsustenance of democratic governance. In other words, the establishmentand maintenance of democratic institutions are likely to be more successfulin richer countries. If democracy is a desirable political system, promotingeconomic development should be as important, if not more important, atask as promoting democracy directly. Efforts to promote democracy inpoor countries without lifting them out of poverty are likely to fail.

The idea that promoting democracy will lift poor countries out of povertyreceives little empirical support. In fact, democratization presents a dilemmafor policymakers. A rise in the level of democracy reduces the level of eco-nomic development, which in turn undermines democracy. Countries thatset development as their priority may need to delay democratization, at leastfor a while. Authoritarianism may provide conditions that are conducive toeconomic development. Nondemocratic countries can more swiftly choose


prodevelopment policies, avoiding the need to choose a middle-of-the-road approach that would placate and keep all constituencies on board.The drawback of this approach is that the public may have no say at allin the development process, leading to skewed economic development thatfavors the interests of a narrow group in society at the expense of themasses. The current policy approach in countries such as China, Singapore,Russia, and perhaps to a lesser degree Malaysia, and the policy approachof newly industrialized countries such as South Korea and Taiwan in the1970s–1990s, suggest that when faced with a choice between democracyand development, many leaders and nations seem to choose developmentfirst.

Our findings also suggest a significant implication about the outcomeof democratic and economic reforms in LDCs. A strategy of developmentfirst, which postpones democratization until some level of development isreached, can eventually bring about a rise in democracy as developmentpromotes democracy. In contrast, a strategy of democracy first may fail asa rise in democracy reduces economic development, which in turn is badfor democracy. Hence, the former strategy is more likely to improve bothdevelopment and democracy.

Ultimately, choosing between the two goals (i.e., which one should takethe driver’s seat) is a normative decision of leaders and nations, but in doingso the leaders should be aware that they are walking a very fine line. Puttingtoo much emphasis on the goal of economic development at the expense ofdemocracy may backfire. As the country develops economically, pressurestoward democratization in the country will likely rise. If these pressuresare not released in an orderly manner, they may explode in the form ofrevolution or coup d’etat. Our results suggest that such political instabilityin turn reduces the level of economic development in a country, possiblywiping out the economic gains the society worked hard to achieve.

SUMMARY AND OUTLOOK

The relationship between democracy and economic development involvestwo causal questions. One question asks whether economic developmentpromotes democracy; the other asks whether democracy influences develop-ment. Statistical analyses typically focus on one of the two questions, relyingon single-equation models. Taken together, however, these two questionsimply a possible reciprocal relationship between democracy and develop-ment. The few statistical studies that attempted to address this possiblesimultaneous relationship suffer several methodological weaknesses.


In this chapter, we revisit this important issue regarding the possiblereciprocal relationship between democracy and development. We specifyand estimate a structural equations model of the simultaneity between thelevel of development and the level of democracy. This approach allowsus to rigorously evaluate the nature of their relationship; that is, whetherdemocracy affects development only, development affects democracy only,they affect each other, or they are not related statistically. The empiricalresults indicate that the level of democracy reduces the level of economicdevelopment, and development promotes democracy.

This chapter concludes Part I of our book. In this part, we modeled theeffects of economic openness on democracy, the effects of democracy andeconomic openness on income distribution, and the reciprocal relationshipbetween economic development and democracy. As demonstrated in thedifferent chapters, these are important issues. But these analyses ignoredthe possible influences of democracy and economic openness on interstatepolitical relations. Yet we observe every day how interstate political eventshave influenced and been influenced by such significant transformations asdemocratization and economic liberalization. Part II of our book explicitlytackles the relevance of democracy (Chapter 5) and economic openness(Chapter 6) to interstate military conflict. In Chapter 5, we focus on therelationship between democracy and military conflict, which speaks to twoseparate bodies of literature: one on democratic peace and the other onthe causes of democracy. We seek to highlight the previously overlookedreciprocal causal link between democracy and conflict.

APPENDIX

MODELS OF DEVELOPMENT, GROWTH, AND DEMOCRACY

In this part of the appendix, we show in detail that a structural equa-tions model of democracy and economic growth differs substantially froma structural equations model of democracy and economic development;hence, inferences based on the two models are not equivalent.

Let us assume that theory expects the following SEM to be the “true”model:

democracyt = b1rgdppct + c1Z1(t−1) (4.A1)

rgdppct = b2democracyt + b3rgdppct−1 + c2Z2(t−1),

(4.A2)


where, as in the main text, rgdppct is real GDP per capita at time t;rgdppct−1 is real GDP per capita at time t − 1; democracyt is the level ofdemocracy at time t; Z1 and Z2 are vectors of relevant control variables; b1,b2, and b3 are coefficients; and c1 and c2 are vectors of coefficients.

Is a model of the change in rgdppc equivalent to the model of the levelof rgdppc shown in equation (4.A2)? The answer is yes. To see this point,subtract rgdppct−1 from both sides of equation (4.A2), generating equation(4.A3):

rgdppct − rgdppct−1 = b2democracyt + (b3 − 1)rgdppct−1

+ c2Z2(t−1). (4.A3)

Estimation of Equation (4.A3) captures coefficient b2 for democracyand coefficients vector c2 for Z2, as in Equation (4.A2). And coefficient(b3−1) for rgdppct−1 is a simple transformation of b3.

But is a model of the growth rate in rgdppc equivalent to Equation(4.A2)? The answer is no. To see this, rework Equation (4.A3) to havethe growth rate of gdppc as the dependent variable, denoted g_rgdppc,by dividing both sides of Equation (4.A3) with rgdppct−1 and producingEquation (4.A4):

[rgdppct − rgdppct−1]/rgdppct−1

= b2democracyt/rgdppct−1

+ (b3−1)rgdppct−1/rgdppct−1

+ c2Z2(t−1)/rgdppct−1. (4.A4)

This equation can be rewritten as

G rgdppct = b2democracyt/rgdppct−1

+ (b3−1) + c2Z2(t−1)/rgdppct−1. (4.A5)

For the rgdppc growth rate model (i.e., Equation (4.A5)) to be equiv-alent in parameters to the rgdppc level model (i.e., equation (4.A2)),the right-hand-side variables in the level model need to be transformedfollowing Equation (4.A5). Previous scholars, in their typically estimatedsingle-equation model of the rgdppc growth rate, do not transform theirdemocracy and control variables as in Equation (4.A5). Therefore, the coeffi-cients of democracy and the control variables in previous models of rgdppcgrowth rate are not the same as those in the rgdppc level model in Equation(4.A2).

Similarly, for the democracy model in Equation (4.A1), the coefficient b1

for rgdppct is not the same as the coefficient of g_rgdppct. To see this, one


divides both sides of Equation (4.A1) by rgdppct−1:

democracyt/rgdppct−1 = b1rgdppct/rgdppct−1

+ c1Z1(t)/rgdppct−1. (4.A6)

Next, we rewrite Equation (4.A6) to get the growth rate of GDPPC onthe right-hand side:

democracyt/rgdppct−1 = b1[rgdppct/rgdppct−1 − 1]

+ b1 + c1Z1(t)/rgdppct−1. (4.A7)

The expression in the square brackets in Equation (4.A7) is the growthrate of GDPPC. Thus,

democracyt/rgdppct−1 = b1G rgdppct

+ b1 + c1Z1(t)/rgdppct−1. (4.A8)

Hence, for the coefficient of rgdppc in Equation (4.A1) to be the sameas the coefficient of g_rgdppc, one needs to estimate Equation (4.A8). Inother words, the coefficient for g_rgdppc in a regression of democracy ong_rgdppct and Z1 is not the same as that for rgdppc in Equation (4.A1) ofthe true model.


Empirical Model

We have two endogenous variables: rgdppc and democracy. rgdppc isthe real GDPPC adjusted for PPP.6 Data are from the Penn World Table 6.1.democracy is measured from the POLITY IV database (Marshall andJaggers, 2000), as in other chapters. The widely used POLITY data regis-ter various democratic and autocratic attributes for many countries on anannual basis from 1800 to 1999. Our measure of democracy is the compos-ite indicator POLITY from POLITY IV. It is constructed as the differencebetween the democracy and autocracy scores in the POLITY IV data set,ranging from −10 (fully autocratic) to +10 (fully democratic).7 The effect

6 Studies of democracy typically use real GDPPC to measure development (e.g., Diamond, 1992a,1992b; Gasiorowski, 1995; Przeworski et al., 2000; Boix and Stokes, 2003; Feng, 2003; Li andReuveny, 2003).

7 Among others, Londregan and Poole (1996) and Li and Reuveny (2003) also use this measure ofdemocracy. As Marshall and Jaggers (2000) recommend, we code the “standardized authoritycodes” of −66, −77, and −88, which appear in the POLITY data set, into missing for −66,averages of the scores before and after the transition for −88, and zero for −77.


of rgdppc on democracy could be positive or negative, whereas the effectof democracy on rgdppc can be positive, negative, or statistically insignif-icant, in light of the earlier arguments.

The specification of the exogenous variables in the development equa-tion follows Solow’s (1956) theory, which Mankiw et al. (1992) extendedto include human capital in addition to physical capital. This theory pre-dicts that the steady-state level of development rises with education andinvestment in physical capital, falls with population growth, and dependson country-specific determinants. The country-specific effects “reflect notjust technology but [also] resource endowments, climate, geography, insti-tutions, and so on; it may differ therefore across countries” (Mankiw et al.,1992: 6). In statistical estimation, the country effects can be modeled aseither fixed effects or random effects, an issue to which we will return.

Mankiw et al. (1992) assumed that countries are in steady state andtherefore employ a cross-sectional design. As they note, however, at anygiven point in time in the Solow model, the economy is somewhere onthe transition path leading to the steady state, not necessarily in the steadystate. Hence, one needs to model development dynamics. Islam (1995)employs this insight in his dynamic extension of the Mankiw et al. setup,where development adjusts over time. Islam’s extension includes the laggeddependent variable on the right-hand side, as well as country- and period-specific effects.8

The specification of our development equation builds on the setups ofMankiw et al. (1992) and Islam (1995). The right-hand-side variables in ourdevelopment equation include investment in physical capital, populationgrowth rate, and average level of education, following the Solow model. Weinclude, as called for by Mankiw et al. and Islam, country effects and, ascalled for by the Islam extension, the lagged level of development and yearlyeffects. We also control for political instability, as suggested by Feng (2003).

The expected effects of the right-hand-side variables on rgdppc followthese studies. prior rgdppc is the lagged value of rgdppc. It models thetendency of economic processes to exhibit inertia due to frictions such astransportation delays, evolution of tastes, and the time it takes to negotiateagreements and produce goods. The effect of prior rgdppc on rgdppc isexpected to be positive.

investment is the logarithm of the percentage ratio of private and publicinvestment in physical capital to real GDP, lagged one year. It is worth noting

8 The Solow/Islam model works in GDPPC level, not in GDPPC growth rate. Thus, it differs fromthe Barro (1998) type of economic growth model. As shown, moving the lagged dependentvariable in the level model to the left-hand side does not lead to an economic growth model.


that foreign direct investment is part of this variable. These data come fromthe Penn World Table 6.1. The effect of investment on rgdppc is expectedto be positive.

population growth is the rate of population growth of a country,lagged one year. Data come from the Penn World Table 6.1. The effect of arise in population growth on rgdppc is expected to be negative.

education denotes the level of national education, measured by theaverage number of years of education in the overall population. Data comefrom Barro and Lee (2000) and are lagged one year. The effect of this variableon rgdppc is expected to be positive.

instability is set to 1 if a country experiences at least one revolutionor coup d’etat in a given year, and 0 otherwise. The data come from CNTS(2008).9 An occurrence of instability is expected to reduce rgdppc (Feng,2003).

trade is the sum of the values of imports and exports of a country withthe world, divided by the country’s GDP, lagged one year. The data comefrom the Penn World Table 6.1, where the division by GDP controls forthe importance of trade to the country. The effect of a rise in trade ondevelopment is debated. Neoclassical economics expects a positive effect,whereas a neo-Marxist/dependency interpretation expects a negative effect,particularly for LDCs.

country fixed effects and yearly fixed effects are country andyearly specific effects, which, as noted, are called for by economic theory(Mankiw et al., 1992; Islam, 1995). In general, the yearly effects captureeconomic dynamics. The country effects capture structural attributes ofdevelopment such as climate and topography. Since these effects can becorrelated with other independent variables, we might need to model themas fixed effects, an issue to which we will return.

The specification of the democracy equation follows studies in compar-ative politics. The endogenous variable on the right-hand side is rgdppc.The exogenous variables include three types of variables: economic perfor-mance, slow-moving structural variables, and external influences. The eco-nomic performance variables include economic growth rate and inflation.10

Structural variables (e.g., ethnic and religious compositions, party fraction-alization, and institutional qualities) are also sometimes included. These

9 CNTS (2008) defines revolution as an illegal or forced government change, an attempt to generatesuch a change, or an armed rebellion whose aim is independence from the government. A coupd’etat is an extraconstitutional or forced change of the government or its effective control.

10 For the use of inflation and economic growth in democracy models, see Gasiorowski (1995),Londregan and Poole (1996), Feng (2003), and Li and Reuveny (2003).


theoretical factors tend to be stable over time, suggesting that democracymay exhibit path dependence or inertia. To model this inertia, we includethe lagged level of democracy on the right-hand side, which, in addition toits theoretical role, assists in modeling the effects of variables that are notpresent in the model.11 As one study notes, “with such a pervasive controlin place, it is more difficult for spurious effects to be reported” (Burkhartand Lewis-Beck, 1994: 905).12 A third set of control variables involves inter-national forces: trade, and diffusion of democratic ideas and norms.13 Tothese variables, we add a yearly trend, as democracy generally has risen inthe period we study.

prior democracy is the lagged value of democracy. As noted, democ-racy should exhibit inertia. A high level of democracy in the previous periodis likely to promote democracy in the current period. The effect of priordemocracy on democracy is expected to be positive.

year controls for the possibility that democracy has a linear trend in ourperiod. Casual observation suggests that democracy increased after 1945 inmany countries, so we expected the effect of year on democracy to bepositive.

trade is the same variable defined for the development equation. Asdiscussed in Gasiorowski (1995) and Li and Reuveny (2003), the effect ofa rise in trade on democracy is under debate and could theoretically bepositive, negative, or negligible.

diffusion denotes the average democracy score of countries within aregion around each country, lagged one year. The regions are Europe, theMiddle East, Africa, Asia, and America. The diffusion of democratic normsis captured through channels such as economic relations, tourism, andcommunication networks (Starr, 1991; Gasiorowski, 1995; Li and Reuveny,2003). The effect of diffusion on democracy is expected to be positive.

growth is the yearly percent growth rate of real GDP, PPP adjustedand lagged one year, computed using data on real GDP from the WorldDevelopment Indicators (World Bank, 2002). Several studies use negativeeconomic growth as a proxy for economic crises; however, the effect of

11 See, e.g., Burkhart and Lewis-Beck (1994), Muller and Seligson (1994), Muller (1995), and Liand Reuveny (2003).

12 The inclusion of the lagged dependent variable might absorb the variations in the dependentvariable that could be explained by other independent variables, making it harder to findstatistically significant results. Hence, our approach can be described as being conservative.

13 For a study focusing on international forces, see Li and Reuveny (2003). See also Starr (1991)for the effect of diffusion, and Gasiorowksi (1995) for the effect of trade openness.


economic crisis on democracy is debated. Some analysts claim the effect ispositive; others argue the effect is negative (for discussions, see Gasiorowski,1995; Haggard and Kaufman, 1995; and Li and Reuveny, 2003).

inflation is the yearly inflation rate in percent from the consumer priceindex, lagged one year. Data come from the World Development Indicators(World Bank, 2002). As discussed by Haggard and Kaufman (1995) and Liand Reuveny (2003), the effect of a rise in inflation on democracy is alsodebated.

Although there are reasons for including country fixed effects in the devel-opment equation (as noted), no similar reasons exist for including themin the democracy equation. Moreover, because democracy often changesrelatively slowly over time, fixed effects are likely to absorb most of thevariations in democracy, leaving little to be explained by other substantivevariables.


Modeling SimultaneityFor our SEM of democracy and development, we first need to consider theissue of identification. The model is clearly overidentified; the number ofexogenous variables excluded from each of our two equations far exceedsthe required number of one in our case, based on the number of endogenousvariables in a particular equation minus one.

As our model is identified, we can proceed to the estimation. Apply-ing ordinary least squares (OLS) to each equation yields biased estimatesbecause the endogenous variables are correlated with the error terms. Todeal with the simultaneity, one needs to use SEM estimators, which areinstrumental variable (IV) estimators, with one difference. Whereas theinstrument in IV estimation can be any variable that is correlated with theendogenous variable and not correlated with the error term, in SEM esti-mation the instruments are based on all the exogenous variables, which isthe best approach to deal with simultaneity; other methods (e.g., Heckmanselection estimation) are second-best approaches.

Two types of SEM estimators exist: 2SLS and 3SLS. In 2SLS, in stage one,each endogenous variable is regressed on all the exogenous variables in themodel. In stage two, the endogenous independent variables are replaced byfitted values from stage one. In 3SLS, stage one is 2SLS; stage two uses theestimates to compute the system’s variance–covariance matrix. Stage threeuses the matrix in estimating the system from generalized least squares.


Both 2SLS and 3SLS estimators are consistent. The 3SLS estimator ismore efficient than the 2SLS estimator, but it also is more sensitive to anyspecification error than the 2SLS estimator. Moreover, as summarized byKennedy (2005: 165), “Monte Carlo studies have shown it [2SLS] to havesmall sample properties superior on most criteria to all other estimators.They have also shown it to be quite robust . . . to the presence of otherestimating problems such as multicollinearity and specification errors.”Therefore, we employ 2SLS as the main estimator.

While we model simultaneity between development and democracy,some of our control variables may be affected by democracy or develop-ment (e.g., traders may be affected by changes in democracy). Obviously wecannot model all the possible simultaneous effects in one study. Our modelalready is complex by considering the reciprocal effects between democracyand development. The literature typically deals with this issue by lagging thecontrol variables.14 Although this is an imperfect solution, we also employit here.

Modeling Country Effects in the Development EquationAs noted, for theoretical reasons, we include country effects in the devel-opment equation (but not in the democracy equation). This leads to twostatistical issues. First, country effects can be modeled as random or fixed.Random effects are modeled as part of the error term and fixed effectsare set as country dummies. The random-effects estimator is more effi-cient than the fixed-effects estimator, but it is consistent only if the countryeffects are uncorrelated with the independent variables. Consistency of thefixed-effects estimator, or the least squares dummy variable (LSDV) model,is not subject to this requirement. Because our country effects captureattributes such as climate or input endowments, they are likely correlatedwith our independent variables (e.g., investment, population growth). TheHausman test compares the random-effects and fixed-effects estimators.The results corroborate our suspicion, suggesting the fixed-effects esti-mator should be used to model the country effects in the developmentequation.

Second, the presence of the lagged dependent variable, together with thecountry fixed effects, makes the fixed-effects or LSDV estimator inconsistentwhen the asymptotics are considered in the N direction (where N is thenumber of countries). Amemiya (1967) and others have shown that the

14 For this approach, see, e.g., Muller and Seligson (1994) and Li and Reuveny (2003).


LSDV estimator is consistent in the T direction. The size of the inconsistencyin the N direction is about 1/T (Nickell, 1981).15

Beck and Katz (2004) compare the properties of the LSDV estimatorin a dynamic panel setup such as ours (TSCS with a lagged dependentvariable) to the properties of two alternative but consistent estimators ofsuch a model: the Anderson and Hsiao (1982) (AH) estimator and the Kiviet(1995) estimator. The AH estimator removes the fixed effects by using firstdifference and uses the second lag of the dependent variable or the secondlag of the first difference of the dependent variable as an instrument forthe first difference of the lagged dependent variable. The Kiviet estimatoremploys LSDV estimates but corrects them with a formula that plugs invalues from a consistent estimator such as the AH estimator.

Since the finite sample properties of these estimators are unknown, Beckand Katz (2004) studied their bias and root mean square error in MonteCarlo simulations. On the basis of their findings, Beck and Katz recommendthat “the AH estimator should not be used for TSCS [time series crosssection] data. We see little reason, in general, not to prefer LSDV over theKiviet estimator when T is twenty or more” (2004: 14). They add: “Fortypical comparative TSCS data, it does not appear that OLS with fixedeffects and a lagged dependent variable (LSDV) is problematic” (2004: 29).In another Monte Carlo study, Judson and Owen (1999: 13) reach a similarconclusion: “When T > 30, LSDV performs just as well or better than theviable alternatives.” Based on these results and the number of years (35)in our sample period, we employ the LSDV estimator for the developmentequation.16

For robustness check, we also use the estimator of Bruno (2005), whichextends the Kiviet (1995) estimator (that applied to balanced panels) tounbalanced panels. Because the Kiviet algorithm has not yet been extendedto 2SLS, we apply the two-step estimation. We refer to this estimator asthe 2SLS-Kiviet estimator. Performing 2SLS step by step produces biasedstandard errors in the second stage because the procedure uses the predictedvalues for the endogenous variable in computing the standard errors insteadof the required actual values (Greene, 2003). But if the R2 in the first stageis high, say 0.8 and above as in our case, the size of the bias is expected tobe quite small.17 The Kiviet estimator requires one to choose a consistentestimator to provide initial values, and the order of the bias to adjust for.

15 For detailed discussion, see Beck and Katz (2004).16 To be exact, the size of the asymptotic bias is only about 1/35, or 2.8%.17 We verified this point in personal communication with William H. Greene, the author of the

popular Econometric Analysis (e-mails dated April 30 and May 2, 2005, and May 11–13, 2008).


We use the estimator developed by Arellano and Bond (1991) for initialvalues, and 1/T for the bias (see also Bun and Kiviet, 2003).

Heteroskedasticity, Serial Correlation, and NonstationaritySince our sample involves panel data, we need to consider the possibilities ofserial correlation and heteroskedasticity in the error terms of both equations.We deal with these issues in two ways. First, the lagged dependent variablein each equation helps to address serial correlation and mitigate the risksassociated with this potential problem (Beck and Katz, 1995a, 1995b, 2004).Second, we estimate the standard errors of parameter estimates using avariant of the White (1980) estimator of robust standard errors that adjustsfor clustering over countries, which is typically referred to as the Huber–White sandwich estimator. This estimator yields consistent estimates of thecovariance matrix under very general conditions of heteroskedasticity andserial correlation (Wiggins, 1999).

One may also be concerned that, if the variables in the model are non-stationary, the estimates generated by the 2SLS (or 3SLS) may not be valid.As it turns out, this is not the case. As shown mathematically by the impor-tant study of Hsiao (1997) – see also Johnston and DiNardo (1997) – the2SLS estimator is valid when the dependent variables are nonstationaryor co-integrated. “In structural equation applications, what one needs toworry about are the classical issues of identification and estimation, notnon-stationarity” (Hsiao, 1997: 385).

Empirical Findings

In this part of the appendix, we first discuss the results of the exogeneity testand the Hausman test, then explain the computation of the long-run effectsof development and democracy, next present the results for the controlvariables, and finally discuss briefly the results of the alternative estimatorsfrom robustness tests.

Exogeneity TestBefore we evaluate our SEM, we need to establish whether there is anempirical basis for such a model. Burkhart and Lewis-Beck (1994) employthe bivariate Granger causality test. They find that economic developmentGranger-causes democracy, whereas democracy does not Granger-causedevelopment. The pioneering result of Burkhart and Lewis-Beck is impor-tant but does not suffice for our case, which involves a multivariate model.


We therefore employ the Davidson–MacKinnon test of exogeneity forregression estimated via instrumental variables. The null hypothesis is thatan OLS fixed-effects estimator of the same equation would yield consistentestimates. Rejection of the null indicates that the instrumental variablesfixed-effects estimator should be employed (Davidson and MacKinnon,1993; Wooldridge, 2002). In other words, the null hypotheses are thatdemocracy is exogenous in a single-equation model of development andthat development is exogenous in a single-equation model of democracy,respectively, whereas the alternative hypotheses are that the two affect eachother in some direction (according to the theories discussed earlier).

Turning to the results, the test statistic has a p-value of 0.010 for democ-racy in the development equation and a p-value of 0.035 for development inthe democracy equation. Hence, we reject the null hypotheses of exogeneityin both cases: treating development as exogenous in a model of democracyis not supported by the data. Similarly, treating democracy as exogenousin a model of development also is not supported by the data. Ignoring thesimultaneity would lead to inconsistent estimates.

Fixed-Effects–Random-Effects TestAs noted in the main text, the development model calls for the use ofcountry effects, which can be modeled as fixed or random. We conductthe Hausman test, which compares the random-effects and fixed-effectsestimators for the development equation. The test statistic has a p-valueof 0.007, rejecting the null hypothesis that the coefficients from the twoestimators are indistinguishable. This result supports the use of the countryfixed-effects estimator.

Computing Long-Run Effects of Development and Democracyon Each OtherIn our SEM, a rise in economic development in the current period affectsthe level of democracy in the current period and continues to affect the nextperiod of democracy via prior democracy in the model. The long-runeffect of some change in development on democracy accumulates over timeand is computed by the following formula:

long-run change in democracy

= Change in rgdppc

× [rgdppc Coefficient/(1 − prior democracy Coefficient)].


The expression in the square brackets of the formula is given by[0.2127/(1 − 0.9237)], which is 2.787. Thus, when real GDP per capitarises from 416 to 3,994, the long-run change in democracy is 6.30 unitson the 20-point democracy scale, which is based on 2.787×[log(3,994) −log(416)].

How large is the long-run effect of democracy on development? Thelong-term effect of a rise in the level of democracy is given by the followingformula:

long-run percent change in gdp per capita

= 100 × Change in democracy

× [democracy coefficient/(1 − prior rgdppc coefficient)].

The expression in the square brackets of the formula is given by[−0.0006/(1 − 0.9537)], which is −0.013. Thus, the long-run percentchange in gdp per capita due to a change of 16 units of democracy scoreis −20.73%, which is based on [100 × 16 × (−0.013)].

Results for Control VariablesNext, we discuss the results for the control variables from the 2SLS estimator.Beginning with the development equation, the effect of a rise in priorrgdppc on development is statistically significant and positive for bothsamples. Economic development exhibits positive inertia, as expected. Alsoas theoretically expected according to the core Solow framework, the effectof a rise in investment on the level of economic development is positive andsignificant, and the effect of a rise in population on the level of economicdevelopment is negative and statistically significant. We obtain these resultsfor both the all-countries sample and the LDC sample. The effect of a rise intrade on the level of economic development is statistically significant andpositive. Trade openness promotes economic development, supporting theneoclassical position in this particular debate.

The effect of a rise in the level of education on economic developmentis positive, as expected by the extended Solow framework, but is not sta-tistically significant in our case. This result is sometimes reported in aneconomic development model, reflecting the correlation between this vari-able and the first lag of rgdppc. Note, however, that the lagged value ofrgdppc is included in the model due to theoretical reasons called for by theIslam (1995) extension of the Solow model in a dynamic panel data setup.It should also be noted that the coefficient of education for the LDC sampleis considerably smaller than this coefficient for the all-countries sample,


reflecting the fact that the level of education in LDCs is relatively lower ingeneral and, therefore, has a weaker effect on development.

Finally, the effect of a rise of instability from 0 to 1, indicating the pres-ence of coup or revolution or both, on the level of economic developmentis negative and statistically significant in both the all-countries and LDCsamples. Consistent with Feng’s (2003) findings, political instability in theform of coup and revolution reduces the level of economic development.

In the democracy equation, for the sample of all countries, the effectof a rise in prior democracy on democracy is significant and positive,demonstrating the importance of democratic inertia. The effect of a risein year is significant and positive, reflecting the general positive trend ofdemocracy in our sample. The effect of a rise in diffusion of democracyideas on the level of democracy in a country is positive and significant. Asthe number of democratic countries in a country’s geographical region rises,so does the country’s own level of democracy. All these reported effects arerobust in the sample of LDCs only.

As we found in Chapter 2, a rise in trade is found to reduce the level ofdemocracy in a country. One explanation of this result is that trade widensthe social cleavages between winners and losers, leading to more socialunrest and repressive regime. Further discussions are presented in Chapter 2.The effect of a rise in trade also is negative for the LDC sample but isnot statistically significant at standard levels. The effects of a rise in therates of inflation and economic growth on democracy are statisticallyinsignificant for both samples, again as reported by Feng (2003), Li andReuveny (2003), Gasiorowski (1995), and others.

Results from Alternative EstimatorsTables 4.A1 and 4.A2 present the results from the 3SLS and 2SLS-Kivietestimators. The effect of a rise in the level of economic development on thelevel of democracy remains positive and significantly different from zeroacross the two alternative estimators and between the two samples. Likewise,the effect of a rise in the level of democracy on the level of economicdevelopment is negative and significantly different from zero across the twoalternative estimators and between the two samples.


Table 4.A1. Democracy and development, 3SLS


RGDPPC 0.2127∗∗∗ 0.1904∗∗∗

(0.0477) (0.0654)PRIOR DEMOCRACY 0.9236∗∗∗ 0.9148∗∗∗

(0.0067) (0.0081)DIFFUSION 0.2882∗∗∗ 0.3612∗∗∗

(0.0744) (0.0943)INFLATION 0.00002 0.00001

(0.0001) (0.0001)ECONOMIC −0.0073 −0.0076

GROWTH (0.0075) (0.0088)YEAR 0.0194∗∗∗ 0.0284∗∗∗

(0.0039) (0.0051)TRADE −0.0014∗∗ 0.0003∗∗∗ −0.0009 0.0003∗∗∗

(0.0009) (0.0001) (0.0010) (0.0001)PRIOR RGDPPC 0.9543∗∗∗ 0.9562∗∗∗

(0.0052) (0.0061)DEMOCRACY −0.00051∗∗ −0.00047∗

(0.0003) (0.0003)INVESTMENT 0.0016∗∗∗ 0.0019∗∗∗

(0.0002) (0.0003)POPULATION −0.0066∗∗∗ −0.0070∗∗∗

GROWTH (0.0015) (0.0018)EDUCATION 0.0024 0.0023

(0.0022) (0.0031)INSTABILITY −0.0102∗∗∗ −0.0107∗∗∗

(0.0029) (0.0034)Constant −40.3984∗∗∗ 0.3507∗∗∗ −58.3035∗∗∗ 0.0000

(7.6554) (0.0598) (10.1004) (0.0000)Observations 2914 2914 2230 2230



Table 4.A2. Democracy and development, 2SLS-Kiviet


RGDPPC 0.2237∗∗∗ 0.2010∗∗∗

(0.0510) (0.0641)PRIOR DEMOCRACY 0.9233∗∗∗ 0.9148∗∗∗

(0.0095) (0.0113)DIFFUSION 0.2837∗∗∗ 0.3549∗∗∗

(0.0875) (0.1068)INFLATION 0.00002 0.00001

(0.00002) (0.00002)ECONOMIC −0.0050 −0.0053

GROWTH (0.0081) (0.0088)YEAR 0.0189∗∗∗ 0.0277∗∗∗

(0.0040) (0.0055)TRADE −0.0015∗∗ 0.0003∗∗∗ −0.0010 0.0002∗∗∗

(0.0009) (0.0001) (0.0010) (0.00008)PRIOR RGDPPC 0.9793∗∗∗ 0.9825∗∗∗

(0.0049) (0.0063)

DEMOCRACY −0.00054∗∗ −0.00055∗

(0.00029) (0.0003)INVESTMENT 0.00149∗∗∗ 0.0017∗∗∗

(0.00024) (0.00023)POPULATION −0.0062∗∗∗ −0.0066∗∗∗

GROWTH (0.0016) (0.0019)EDUCATION 0.0011 −0.00001

(0.0028) (0.0033)INSTABILITY −0.0102∗∗∗ −0.0108∗∗∗

(0.0029) (0.0039)Constant −39.7159∗∗∗ 0.0000 −57.1632∗∗∗ 0.0000

(8.1152) (0.0000) (10.9474) (0.0000)Observations 2917 2858 2233 2193


PART II

BRINGING IN CONFLICT

FIVE

Democracy and Conflict

INTRODUCTION

Interstate military conflict as an extreme form of political exchange is a cen-tral feature of the international system and has received the enormous atten-tion it deserves from scholars and policymakers. Currently state sovereigntyand anarchy make world peace seem elusive. Chapters 2–4 in Part I modeledthe effects of economic openness on democracy, the effects of democracyand economic openness on income inequality, and the reciprocal effects ofeconomic development and democracy, respectively. These analyses ignoredthe possible impact of democracy and economic openness on interstatepolitical relations. In reality, democracy and economic openness developin and around nation-states such that they inevitably influence politicalrelations. In fact, scholars have argued and sought to demonstrate that bothdemocracy and economic openness affect interstate military conflict, eventhough they debate the nature of their effects.

Part II of our book explicitly tackles the relevance of democracy (Chap-ter 5) and economic openness (Chapter 6) to interstate military conflict.In Chapter 5, we are particularly interested in the relationship betweendemocracy and military conflict. This chapter speaks to two separate bodiesof literature: one on democratic peace and the other on the causes of democ-racy. Both topics have been the subjects of voluminous research, generatingtwo large bodies of literature. Our contribution is not to overthrow butto build on and connect these two separate bodies of literature, highlight-ing and analyzing the previously overlooked reciprocal causal link betweendemocracy and conflict.

The claim that democracies do not fight one another has received wideattention. Most statistical studies investigating this claim share similardesigns, employing a dyadic level of analysis and estimating single-equation

125


models. In these studies, the dependent variable is typically dichotomous,measuring the presence or absence of a militarized interstate dispute (MID).The independent variables include dyadic, joint (dyadic) democracy, andmore or less the same collection of control variables. The vast majority ofstudies report that the probability of military dispute between two coun-tries declines as their joint democracy level increases, although a number ofstudies reject this argument. We note that a much smaller group of studiesargues that democracies are more peaceful in general, not only toward otherdemocracies. This claim has received less empirical support. In this chapter,we focus on the dyadic version of the democratic peace argument.

Side by side with the democratic-peace literature is a growing num-ber of international relations scholars who argue that military conflictaffects democracy. But they debate about the direction of the effect ofconflict on democracy. Conceptually, these studies form a subset of thelarge monadic literature within comparative politics that studies the deter-minants of democratization.

This chapter is motivated by the observation that the democratic-peaceliterature and the literature on the effect of conflict on democracy havebasically ignored the insights offered by each other. In particular, mostinternational relations scholars have treated the effect of joint democracyon conflict as unidirectional, even though conflict has been demonstratedto influence democracy. Once the two separate bodies of literature havebeen linked, one can reasonably argue that conflict and democracy affecteach other simultaneously.

Our goal in this chapter is to address the simultaneity between conflictand democracy within a unified statistical framework. We stress at the out-set that our agenda is not to prove or disprove the existence of democraticpeace. Rather, we believe that democracy and dyadic conflict affect oneanother. Thus, researchers need to model the reciprocal effects explicitly,which is not an easy task; it requires connecting monadic and dyadic levelsof analysis. We develop a novel way to link the two levels of analysis whilemaintaining consistency with both the dyadic democratic-peace literatureand the literature on the monadic determinants of democracy. In our simul-taneous equations model, the endogenous variables include dyadic conflictmeasured by the presence or absence of a MID, the higher level of democ-racy in a dyad, and the lower level of democracy in a dyad. As in studiesby Dixon (1994) and Oneal and Russett (1997), the higher level of democ-racy in a dyad measures political regime dissimilarity and the lower level ofdemocracy reflects dyadic, joint democracy. Our model usefully integrates

Democracy and Conflict 127

the dyadic determinants of dyadic conflict with the monadic determinantsof democracy within one system of simultaneous equations.

Some may view the simultaneity of dyadic conflict and democracy merelyas a methodological nuisance. It is worth noting that, in contrast to thisview, we consider the issue to be both theoretical and methodological. Infact, various scholars share this view. For example, Chan (1997a) arguesin his comprehensive review that the democratic-peace literature needsto take more seriously the possibility of reverse causality from conflict todemocracy. Gleditsch and Ward (2000) also note briefly that conflict oftenaffects democratization. As discussed in detail later, in the first study thatemploys a partial simultaneous setup to examine the reciprocal relationshipbetween conflict and democracy, James et al. (1999) argue that the effectof joint democracy on dyadic conflict is generally insignificant. Their studyhas led to a lively debate, but the conflict–democracy simultaneity issueremains unresolved.1

We believe that further investigation of the relationship between dyadicconflict and democracy within a simultaneous framework is warranted. Asnoted by Alvarez and Glasgow (2000: 163), “if there are theoretical reasons tosuspect endogeneity in the model, it must be modeled. Ignoring endogeneitywill lead to biased estimates.” Of course, the actual effect of simultaneityon the dyadic conflict–democracy nexus could be large or small; but unlessone models the reciprocal interaction between democracy and conflict, onesimply cannot evaluate the effect of ignoring the simultaneity, which is acommon practice in the conflict community.

Our work in this chapter assesses empirically the question of simultaneitybetween democracy and dyadic conflict across a sample of politically rele-vant dyads from 1950 to 1992.2 Our empirical findings can be summarizedas follows. First, dyadic conflict reduces both lower and higher levels ofdemocracy in a dyad. Second, joint democracy reduces the probability ofdyadic MID involvement. When compared with the single-equation esti-mate of Oneal and Russett (1997), the absolute magnitude of the effect inour study is smaller but in relative terms, the effect is similar in size. Theeffect of joint democracy on MID involvement is considerably smaller fornoncontiguous countries than for contiguous countries. Third, the level ofdemocracy in a country is affected by a number of economic and social

1 See Oneal and Russett (2000) and the response of James et al. (2000). For an earlier expositionof the position stated by James et al. (1999, 2000), see Wolfson et al. (1998).

2 Politically relevant dyads are dyads involving at least one major power (United States, UnitedKingdom, France, USSR/Russia, and China) or geographically contiguous countries.


variables that have commonly received attention in the democratizationliterature. Fourth, several control variables in the MID equation that priorstudies have found to be significant in the single-equation framework areshown to lack statistical significance in our simultaneous equation modeldesign. Notably, we find that preference similarity between two countries,based on the United Nations (UN) voting patterns, does not have a signifi-cant effect on dyadic MID involvement.

We organize the chapter as follows. The next section reviews the demo-cratic-peace literature as well as the literature on the effect of conflict ondemocracy, and the section that follows discusses the theoretical expec-tations. We then discuss the few empirical studies that have investigatedsome aspects of the simultaneity between democracy and conflict. The fifthsection of this chapter presents our conceptual model. The sixth sectiondiscusses the statistical model, research design issues, and key findings andis followed by a conclusion section.

TWO BODIES OF LITERATURE ON DEMOCRACYAND MILITARY CONFLICT

This section discusses the two bodies of literature that focus on the effectof democracy on military conflict and the effect of conflict on democracy,respectively.

Effect of Democracy on Conflict

The democratic-peace literature has been reviewed extensively elsewhere.3

In this section, we briefly summarize this literature. Most scholars arguethat democracies do not fight one another and that they are unlikely toengage in militarized disputes with one another.4 The monadic argumentthat democracies are less likely to be involved in MIDs with all countriesalso has been made but is relatively more controversial.5 As noted, we focuson the dyadic version of the argument.

The democratic-peace literature is based on the empirical observa-tion that democracies have almost never fought wars against each other.

3 For example, see Levy (1989), Russett (1993), Gleditsch (1995), Ray (1995), Chan (1997), andRussett and Oneal (2001).

4 For example, see Bremer (1993), Maoz and Russett (1993), Dixon (1994), and Oneal and Russett(1997).

5 Rummel argues that “the more a nation is democratic, the less severe its overall foreign violence”(1997: 5). However, he admits that this argument is controversial.


Although a few cases are said to disprove this empirical regularity, a fewstudies explain why these cases are questionable.6 Given this empirical reg-ularity, the thrust of the literature has focused on explaining why democ-racies do not fight each other, whether this regularity can be attributed todemocracy or other forces, and whether it is simply due to random historicaloccurrence.

The proponents of the democratic-peace proposition offer several expla-nations for their claims. The first, so-called institutional, explanation tracesback to Kant (1795). The decision to wage war faces more checks andbalances in democracies than in autocracies (Russett, 1993). Democraticgovernments confront more constraints by the rule of law. The slowingdown of the decision-making process leaves more time for negotiation totake place and reduces the risk of errors and misunderstandings, both ofwhich reduce the probability of war (Maoz and Russett, 1993). Yet anotherinstitutional explanation concerns the informational role of democraticinstitutions (Fearon, 1994; Schultz, 1998; Bueno de Mesquita et al., 1999).Because democratic leaders are held accountable and compete for reelec-tion, military defeats may lead to electoral defeats by political opponents.When democracies fight, they mobilize their resources and tend to win;therefore, they prefer to mediate disputes and, in the event of war, choosetheir fights carefully to increase the likelihood of victory. Democracies donot fight each other because they are unattractive targets for other democ-racies. So the fact that a democracy mobilizes and escalates a crisis sendsa clear signal regarding its resolve to the other party. Because all democra-cies behave in this manner and their leaders know it, democracies are lessaggressive toward one another.

According to a second so-called normative argument, democratic gov-ernments tend to tolerate their domestic opposition and resolve domes-tic conflicts peacefully. Analogously, they also behave in this way towardother countries (Weart, 1998). Although these effects are monadic, theyare allegedly stronger when democracies interact with one another. Sincethe domestic politics of democracies is transparent, democracies can morereadily trust each other (Maoz and Russett, 1993; Russett, 1993). It followsthat conflicts among democracies are more likely to be mediated by thirdparties than they are to escalate (Dixon, 1994).

6 For a discussion of these cases see, e.g., Doyle (1997). Russett and Oneal (2001) emphasizethat the effect of joint democracy on dyadic conflicts is probabilistic. Even if a few democraticdyads exhibit militarized disputes, these cases do not refute the claim that democratic dyads aregenerally less likely to exhibit militarized disputes than other dyads.


A number of studies reject the democratic-peace proposition. Farber andGowa (1995), for example, argue that the purported negative effect of jointdemocracy on conflict is an artifact of the common interests that were sharedby democracies when facing the former USSR during the Cold War. Layne(1994) argues that the American Civil War is a strong disconfirming case ofthe democratic-peace argument. Polachek (1997) and Polachek and Robst(1998) argue that democracies do not fight each other because of theirextensive trade relations, since conflict reduces trade.7 Lemke and Reed(1996) account for this apparent regularity by arguing that democraciestend to be content with the status quo, whereas autocracies tend to bedissatisfied with the status quo and, therefore, are more prone to conflict.This theory explains both the lack of wars between democracies and thehigher incidence of wars between democracies and autocracies. From adifferent angle, Gartzke (1998) argues that democracies are less likely todisagree and, thus, have no need to fight each other. When one includesa measure of affinity based on the UN General Assembly voting patternsin the typical model, the effect of joint democracy on the probability ofMIDs becomes weak and insignificant.8 Rosato (2003) does not challengethe empirical regularity of democratic peace but argues that extant evidencedoes not substantiate the different causal mechanisms offered by proponentsto rationalize the existence of democratic peace.

Effect of Conflict on Democracy

Whereas most scholars seem to agree that democracies do not fight eachother, the effect of military conflict on democracy is a subject of controversy.According to one view, conflict reduces democracy. This view originatedfrom the writings of de Tocqueville (1835: 650): “All those who seek todestroy the freedom of the democratic nations must know that war is thesurest and the shortest means to accomplish this.” War may not immediatelyresult in a military government, but it will eventually concentrate power inthe hands of the government, leading to despotism and the decline ofdemocracy. Elaborating on the negative effect of conflict on democracy,Lasswell (1977) argues that because some groups may not want to join the

7 For additional discussion of the link between democratic peace and economic relations, seeMousseau and Shi (1999).

8 Another argument that weakens the democratic-peace proposition is offered by Mansfield andSnyder (1996), who argue that young democracies are more prone to engage in war with eachother than are established democracies.


war effort, the state may become increasingly autocratic to reassert power.Moreover, conflict may require resources that the state does not have. Whenpeople do not agree to give up wealth for the war effort, the state may have tobecome oppressive to pay for the war and subsequently force the people todo so.9 Layne (1994) and Thompson (1996) suggest a different mechanism.Building democracy requires peace. States facing external threats need toallocate resources to national defense and maintain a large governmentalapparatus; both factors reduce the level of resources available for democraticconsolidation.

A competing view posits that conflict leads to democracy (Tilly, 1992).Historically, many leaders, when mobilizing their populations, grantedpolitical rights or land in exchange for their support of the war efforts.Such redistribution of wealth enlarges the number of people that are ableto affect governance and, thus, facilitates democratization. Moreover, warsoften provide the impetus for social reforms, including democratization(Higgs, 1988; Porter, 1994; Kasza, 1996). Skocpol et al. (2001) demonstratethis effect in a case study and argue that the U.S. Civil War and U.S. partici-pation in World War I exerted a strong positive effect on the U.S. democraticcivil society.10

A third group of studies, mostly in the field of international relations(e.g., Modelski and Perry, 1991; Bueno de Mesquita et al., 1999; Mitchellet al., 1999), offers a conditional argument for the expected effect of militaryconflict on the level of democracy. When an interstate war ends in a victoryby a democratic country, the winner often imposes democratization onthe loser, replicating its own political system design. In contrast, when thewinner of a war is autocratic, the level of democracy tends to decline in thelosing country.11

Despite competing expectations about the direction of the effect, all threeviews agree that military conflict does influence democracy. The overall ornet effect of conflict on democracy thus appears to warrant empirical inves-tigation. However, these studies do not take into account the converse pos-sibility: democracy may affect conflict behaviors. Again, as noted, ignoringsuch reciprocal effects may lead to erroneous empirical findings.

9 For a similar argument, see Mintz (1985), Almond (1990), Martin (1994), Segal (1994),Midlarsky (1995), and Gates et al. (1996).

10 For other case studies that demonstrate the positive effect of acute conflict on democracy forces,see, e.g., Marwick (1988), Porter (1994), and Reuveny and Prakash (1999).

11 For example, compare the cases of Germany and Japan after World War II, where some Europeancountries were under Nazi or Soviet control.


IMPLICATIONS AND THEORETICAL EXPECTATIONS

We have reviewed two bodies of literature, one on the effects of democracyon conflict and the other on the effects of conflict on democracy. By syn-thesizing the two bodies of literature, we hypothesize that the relationshipbetween dyadic conflict and democracy is reciprocal; in other words, thetwo forces simultaneously affect one another. Again, some may dismiss thesimultaneity as a methodological nuisance; however, we believe this issuehas important theoretical, methodological, and policy implications.

The democratic-peace literature builds on results from single-equationmodels. However, such results rely on the regression of dyadic conflict onjoint democracy – a technique that cannot distinguish whether joint democ-racy affects or is affected by MIDs. In contrast, our simultaneous-equationsframework may yield one of three possible empirical outcomes. First, con-flict influences joint democracy, but joint democracy does not affect conflict.Second, dyadic conflict does not affect joint democracy, but joint democracyinfluences conflict. Third, joint democracy and dyadic conflict affect eachother. We cannot determine the empirical validity of these three distinctoutcomes until we model the democracy–conflict simultaneity explicitly.

That said, we believe the arguments that attribute peace to joint democ-racy are theoretically plausible. Furthermore, we cannot ignore the factthat the democratic-peace argument has received wide empirical support inmany single-equation studies. We therefore state the following hypothesis:

H1: Joint democracy reduces the incidence of MIDs between two countriesin a dyad.

In terms of the effect of conflict on democracy, although the arguments inthe literature are made in the context of a monadic level of analysis, they canbe readily applied to the dyadic level, as dyads are merely pairs of countries.If conflict affects the level of democracy of any of the two countries in adyad, it also could, in principle, affect the joint democracy of that dyad.

Building on our discussion in the previous section, we categorize theeffects of conflict on democracy into four types. The absence of conflictinvolvement by a country enables the provision of resources for the nurtur-ing of democratic institutions. In contrast, the presence of conflict reducesthe level of resources available to support democratization. The preparationfor conflict may increase autocracy as the state coerces the people to par-ticipate in, and contribute resources to, an unpopular war. This channel,however, may also promote democracy: the state may grant political rights


to groups in return for their support of the war. The occurrence of conflictgenerates a sense of crisis, which legitimizes power centralization. However,the state may also grant democratic rights during conflict to win supportfor the war. The postconflict effect comes in two variants. In one variant,conflict is a social shock, stimulating reforms that increase democracy. Asecond variant involves the winners of a war imposing their own politicalregime on the losers.12 In summary, the existing theories in the literature onthe effect of conflict on democracy offer conflicting expectations, implyingthat the overall (net) effect of conflict on democracy is theoretically unclearand leading to the following expectation:

H2: Conflict may raise or reduce the level of democracy.

PREVIOUS EMPIRICAL STUDIES OFCONFLICT–DEMOCRACY SIMULTANEITY

In general, the issue of conflict–democracy simultaneity has not receivedmuch attention in the empirical literature. Before we turn to our analysis,however, we believe it is beneficial to review the few studies that have dealtwith some aspects of this simultaneity. Investigating the reverse causality,Mousseau and Shi (1999) focus on the proposition that democracy declinesas countries prepare for war. However, their study does not find statisti-cal support for this claim. Crescenzi and Enterline (1999), on the otherhand, argue that the relationship between war and democracy is recipro-cal, as reflected by the portion of democratic states in the system and thefrequencies of democratization and war in the system. They found, usingvector autoregression analysis, that these variables do indeed influence oneanother, but the strengths of their effects change across time and space.Rasler and Thompson (2000) argue both that states perceiving high exter-nal threats will be less democratic and that democracies are less likely tohave MIDs with one another. Focusing on nine major powers from 1816 to1992, they estimate two equations separately. In one equation, the monadiclevel of democracy is regressed on the change in democracy, the level ofexternal threat, the change in the level of external threat, a dummy variablemeasuring war participation, and the lagged level of democracy. In a secondequation, a dyadic MID dummy is regressed on a joint democracy dummyand the level of external threat. They conclude that a high level of external

12 Mousseau and Shi (1999) suggest a similar typology involving war’s anterior, concurrent, andposterior effects.


threat reduces democracy and that democratic dyads are less likely to expe-rience MIDs. None of these studies, however, has employed a simultaneousequations model.

James et al. (1999) were the first to study the simultaneous relationshipbetween joint democracy and dyadic conflict. They estimated the followingstructural equations model:

hostility = F (regime, growth, proximity, ally), (5.1)

regime = G (hostility, growth, stability). (5.2)

where hostility is the highest level of dyadic hostility in a year, with fivepossible values (0 = no dispute, 1 = threat of force, 2 = display of force,3 = use of force, and 4 = war); regime [(rH + rL)/(rH − rL +1)] isa measure of joint democracy from Maoz and Russett (1993), where rH

and rL are the higher and lower regime scores in a dyad, respectively; r =c(d − a) is the regime index of a country, where autocracy, democracy, andconcentration of power are represented by a, d, and c, respectively. growthis the average of the growth rates of the two countries, A and B, in thepast three years. proximity captures if countries A and B are contiguousor if they can project force beyond their contiguous neighbors (i.e., UnitedStates, United Kingdom, France, China, and the Soviet Union). ally reflectsthe alliance ties between two countries. stability measures joint regimestability within a dyad.

In their estimation, James et al. (1999) ignored the simultaneity forEquation (5.1) and estimated it as a single-equation model. Equation (5.2)is estimated with a two-stage procedure, taking into account the simultaneityof regime and hostility. The authors find that the effect of joint democracyon the probability of MIDs is generally not significant; the hypothesis thatpeace induces democracy receives stronger support.

Oneal and Russett (2000) criticize James et al.’s (1999) study on severalgrounds. First, they criticize the use of a dated measure of joint democracy.Second, they argue that the multinomial approach to hostility is destinednot to find significant effects for democracy on conflict since MIDs are rareevents – with the five categories of MIDs representing even rarer events.Third, Equation (5.1) omits the capability ratio of the two countries inthe dyad. Oneal and Russett obtained different results for Equation (5.2)than did James et al. when this variable was included in Equation (5.1).Fourth, Equation (5.2) omits variables routinely used in the comparativepolitics literature on the determinants of democracy, such as GDP per capita,inflation, or economic growth (Gasiorowski, 1995). Fifth, Equation (5.2)assumes that dyadic conflict affects joint democracy. However, because it


is a monadic quality, democracy ought to be affected by all the conflicts acountry engages in, not only by dyadic conflict.

James et al. (2000) replied that it is not clear why one empirical measureof joint democracy is superior to others and that the results of Oneal andRussett are sensitive to the measure used. Second, they argue that the aggre-gation of all MID types into one measure can introduce bias. Diplomaticprotest, for example, is an action different from war. Third, they claimthat capability ratio ought to be considered an endogenous variable, notan exogenous variable as Oneal and Russett assume. Moreover, includingcapability ratio in the equation of hostility still produces a weak effect forregime on hostility. Fourth, the sensitivity of the results from Equation(5.2) to the inclusion of a capability ratio in Equation (5.1) could reflect thesensitivity of the model to the practice of temporal aggregation. In the end,James et al. argued that whereas their simultaneous equations model is notmeant to be the last word on the reciprocal relationship between militaryconflict and democracy, studies ignoring this issue do not offer reliableguidance for public policy.

CONCEPTUAL MODEL

In putting together our statistical model in this chapter, we need to specifytwo types of equations: one for dyadic conflict and a second for democracy.The level of democracy of a country is a monadic attribute. Dyadic conflictis obviously a dyadic attribute. In the democratic-peace literature, dyadicconflict is assumed to decline with joint democracy. In our conceptualframework, joint democracy and dyadic conflict depend on each other.But democracy depends not only on dyadic conflict – it also depends onvarious monadic attributes, including the conflict a country has with otherthird-party countries. We include these factors in our model.

Our measure of joint democracy follows the weak link assumption(Dixon, 1994). In this approach, joint democracy is given by the lowerdemocracy score of the two countries in a dyad. This assumption formalizesthe expectation that the likelihood of a dyadic militarized dispute dependson the regime score of the politically less constrained dyad member (i.e., thecountry that is less democratic). Authors of many statistical studies employthe weak-link assumption-based measure of joint democracy, includingDixon (1994), Oneal and Russett (1997), and Russett et al. (1998).

Some studies also use the higher democracy score in a dyad as an inde-pendent variable in their single-equation model of MIDs. In fact, one wouldexpect that the democracy scores of both countries in a dyad would inter-act with conflict; after all, it takes two states to engage in dyadic conflict.


Therefore, if democracy affects conflict, each of the democracy scores in thedyad should affect its likelihood. Consequently, we also include the higherdemocracy score in the dyad in our analysis and treat it as a third endoge-nous variable. As suggested by Oneal and Russett (1997) and Russett et al.(1998), by holding the other regime score in a dyad constant, one mayinterpret the higher democracy score in a dyad as representing the regimedissimilarity between two countries. These authors and others expect con-flict to decline with the lower democracy score in a dyad but to rise with thehigher democracy score in a dyad.

We formally develop the conceptual model in the appendix. Here wepresent our set of estimated equations, Equations (5.3), (5.4), and (5.5),where F indicates functional dependency:

demL = F (ML, midABU, midLRU), (5.3)

demH = F (MH, midABU, midHRU), (5.4)

midABU = F (XAB, demL, demH). (5.5)

In these three equations, demL is the lower one of the two democracyscores (one for each country) in dyad AB, where A and B are the two dyad-member countries; demH is the higher democracy score in dyad AB; midABU

is the unobserved propensity of dyad AB to engage in a militarized dispute;ML indicates the monadic attributes that explain the level of democracy forthe demL country (the one with the lower democracy score in a dyad); andMH indicates the attributes that explain the level of democracy for the demH

country (the one with the higher democracy score).For midLRU in Equation (5.3) and midHRU in Equation (5.4), we first

need to define the concept of politically relevant international environment(PRIE). The PRIE of some country A includes all the countries that sharea border with A, as well as those that are classified as major powers in atime period (Maoz, 2001a). midLRU is the propensity of the low democracycountry in a dyad to engage in conflicts with third parties in its politicallyrelevant international environment, a variable that is unobserved (the vari-able subscript LRU is from the italicized words). Similarly, midHRU is theunobserved propensity of the demH country to engage in military disputeswith third parties in its PRIE.


Similar to the previous chapters, this section first presents our statisticalmodel for the empirical analysis and then discusses several research designissues. Then the section presents the key results from the empirical analysis.


As in the other chapters, the discussion in this section is self-contained anddoes not require any specific statistical expertise. The technical details of thestatistical model, measure construction, and data sources are in the chapterappendix and follow the same order of presentation as in the main text.

Empirical Model

The next three equations state our statistical model and follow the logicin Equations (5.3), (5.4), and (5.5). The statistical model provides a struc-ture for us to guide statistically uninitiated readers through the empiricalexercise. The model distinguishes endogenous and exogenous (or control)variables. The endogenous variables appear on the left-hand side of oneof the equations and the right-hand sides of the other two equations, andexogenous variables appear only on the right-hand sides of the equations.Small capital letters denote variables, and the Greek letters �, � , and �

denote their coefficients, which we will estimate. The Greek letters εt, �t,and �t represent the error terms in the three equations, respectively. Thesubscripts L and H refer to the lower and higher democracy countries in adyad, respectively. The subscripts t and t − 1 denote the time period of avariable, where t denotes the current period and t − 1 denotes the previousperiod (or a lagged variable).

Equation (5.6) operationalizes Equation (5.3) in the conceptual model,which explains the causes of democracy in the lower-democracy country ina dyad:

demLt = �0 + �1midABUt + �2prior demLt−1 + �3yeart−1

+ �4inflationLt−1 + �5inflationLt−1∗yeart−1

+ �6gdppcLt−1 + �7gdppcLt−1∗yeart−1

+ �8growthLt−1 + �9tradeLt−1 + �10diffusionLt−1

+ �11third-party midt−1 + εt. (5.6)

Equation (5.7) operationalizes Equation (5.4) in the conceptual model,which explains the causes of democracy in the higher democracy countryin a dyad:

demLt = �0 + �1midabut + �2prior demLt−1 + �3yeart−1

+ �4inflationLt−1 + �5inflationLt−1∗yeart−1

+ �6gdppcLt−1 + �7gdppcLt−1∗yeart−1 + �8growthLt−1

+ �9tradeLt−1 + �10diffusionLt−1

+ �11third-party midt−1 + �t. (5.7)


Finally, Equation (5.8) operationalizes Equation (5.5) in the conceptualmodel, which explains the causes of dyadic military conflict:

midABUt = �0 + �1demLt + �2demHt + �3contiguityt−1

+ �4alliancet−1 + �5capability ratiot−1

+ �6trade dependencet−1 + �7growtht−1

+ �8affinityt−1 + �t. (5.8)

The endogenous variables in this system are the unobserved propensityfor conflict in dyad AB at time t (midABUt), the lower level of democracyin dyad AB at time t (demLt), and the higher level of democracy in dyadAB at time t (demHt). Recall that the lower democracy score in a dyadrepresents the joint level of democracy shared by the countries in a dyad,and the higher democracy score in a dyad reflects the extent of regimedissimilarity between two countries. Also, we expect that demL reducesmidABU, demH increases midABU, and the effects of midABU on demL anddemH could be negative or positive. In empirical analysis, because we donot observe midABU, we replace it with an observed dichotomous indicator,midAB. It takes on a value of 1 when two countries are engaged in a dyadicmilitarized dispute (MID) and 0 otherwise. The involvement in a MIDimplies the threat of force, the display of force, the use of military force,or war.

Table 5.1 summarizes not only the relationships among the three endoge-nous variables but also the expected effects of the exogenous variables. Theexogenous variables ML and MH in the two democracy equations closelyfollow the model from the democratization literature, as in Chapter 2. Tosimplify the notation, we drop subscripts t and t − 1. In terms of the eco-nomic determinants of democracy, gdppcL and gdppcH indicate the levelsof economic development for the low- and high-democracy countries in thedyad; growthL and growthH measure the yearly growth rates of the twocountries; inflationL and inflationH indicate the annual inflation ratesfor the low- and high-democracy countries; tradeL and tradeH representthe countries’ levels of trade openness, where trade openness is the ratioof total trade over GDP; and year is a linear trend variable. The termsinflationL

∗year and inflationH∗year are interaction terms between

inflation and year in the low- and high-democracy countries, and the termsgdppcL

∗year and gdppcH∗year are interaction terms between year and

GDP per capita in the two countries. These interaction terms capture thetime-varying effects of these factors.


Table 5.1. Expectation of direction of effects in the simultaneous equations

DEML equation DEMH equation MIDAB equationendogenous Sign endogenous Sign endogenous Sign

MIDAB +, − MIDAB +, − DEML −DEMH +

EXOGENOUS EXOGENOUS EXOGENOUSPRIOR

DEMOCRACYL

+ PRIORDEMOCRACYH

+ CONTIGUITY +

YEAR + YEAR + ALLIANCE −INFLATIONL +, − INFLATIONH +, − CAPABILITY

RATIO−

INFLATIONL∗YEAR +, − INFLATIONH

∗YEAR +, − TRADEDEPENDENCE

−

GDPPCL + GDPPCH + GROWTH −GDPPCL

∗YEAR +, − GDPPCH∗YEAR +, − AFFINITY −

GROWTHL +, − GROWTHH +, −TRADEL +, − TRADEH +, −DIFFUSIONL + DIFFUSIONH +THIRD-PARTY

MIDL

+, − THIRD-PARTYMIDH

+, −

In terms of the noneconomic determinants of democracy, prior demL

and prior demH are the one-year lagged democracy scores of the low-and high-democracy countries; diffusionL and diffusionH indicate theaverage democracy scores of the countries in the PRIEs of the low- andhigh-democracy countries in a dyad; and third-party midL and third-party midH are dichotomous variables, which indicate whether the low-and high-democracy countries, respectively, are involved in a militarizeddispute with at least one third-party country in their respective PRIEs.

The exogenous (or control) variables in the MID equation are the usualsuspects in the dyadic studies of democratic peace: contiguity indicateswhether the two states in a dyad are geographically contiguous (or separatedby up to 150 miles of water); alliance indicates whether the two statesin a dyad are members in a military alliance; capability ratio reflectsthe differences between two countries in terms of material and militarycapabilities; trade dependence indicates the level of trade interdependencebetween two countries; growth captures the shared economic growthrate in a dyad; and affinity measures the level of preference similaritybetween two states, based on the congruence of their votes in the UNGeneral Assembly.



To implement our statistical model requires us to address several specialdesign issues. Although we discuss their technical details in the appendix, weoffer a brief overview here for general readers. First, the simultaneous equa-tions model includes equations for two observed continuous endogenousvariables (demL and demH) and one unobserved continuous endogenousvariable (midABU). As is typical in such cases, we replace the unobserved vari-able with an observed conflict indicator. Such models need to satisfy someeconometric requirement so that their estimation is meaningful. In thechapter appendix, we show that this condition, which was developed byHeckman (1978), holds in our case.

Second, in the empirical estimation of the model’s coefficients, we use atime-series cross-sectional data set (also known as panel data). The data areyearly, and the sample pools all the politically relevant dyads from 1950 to1992. Scholars in the literature define politically relevant dyads as pairs ofcountries that are either geographically contiguous or that include at leastone major power (United States, United Kingdom, France, USSR/Russia,and China). The politically relevant dyads are allegedly most likely to expe-rience dyadic political conflict.

Third, the estimation of any simultaneous equations model differs fromthe estimation of a single-equation model in that it needs to take intoaccount the simultaneity relationships between variables or their reciprocaleffects on one another. We estimate our model using a variant of the two-stage least squares method suggested by Maddala (1983), which is designedfor a system of simultaneous equations of the type developed in this chapterand includes both dichotomous and continuous endogenous variables. Weuse the word “variant” since, as we discuss in the appendix and in the nextsubsection, the Maddala method is designed for the case with one dichoto-mous and one continuous variable, but our model includes one dichoto-mous and two continuous variables.

Finally, as noted in previous chapters, the error term in the statisti-cal model needs to satisfy some assumptions for statistical inferences tobe valid. We address the related assumption violations using appropriateeconometric techniques. We also discuss other complications in our case ofthe simultaneous equations model in the appendix.

Empirical Findings

This section presents our key empirical findings for the relationship betweenmilitary interstate conflict and democracy. We first discuss our main results,


followed by results from additional analysis. In the appendix of this chapter,we fully present the findings of the analysis and provide all the technicaldetails for interested readers.

Tables 5.2 presents the empirical results from the three equations of thesimultaneous equations system. General observations on the performanceof our simultaneous equations model suggest that the model performs well.The model’s goodness of fit between the predicted values and the data ishigh. The empirical results for the control variables, which we fully discussin the appendix, are generally consistent with our expectations and previousfindings from the literature.

Starting with the results for the demL equation, the coefficient of midAB isnegative and statistically different from zero. This means that dyadic conflictreduces the lower level of democracy in a dyad. In terms of the size of effect,a dyadic military dispute reduces the lower level of democracy in a dyadby about 0.16 – a decline of about 38% in the average lower democracyvalue in the sample (0.425). Now, because dyadic conflict also affects thelower level of democracy in the next period through its effect on the laggedlower level of democracy in the model, we can compute the effect of MIDinvolvement on the lower level of democracy in the long run, based on theformula described in the appendix. By employing the coefficients of dyadicMID and the lagged lower level of democracy, we find that dyadic MIDinvolvement reduces the lower level of democracy by 2.5 points over thelong run – a 590% decline in the average value of the variable in the sample.Therefore, both the short- and long-run effects of dyadic military disputeon joint democracy are large in size.

In terms of the results for the demH equation, the coefficient of midAB

is negative and statistically different from zero. Thus, dyadic conflict alsoreduces the higher level of democracy in a dyad. With respect to the sizeof effect, a dyadic military dispute reduces the higher level of democracyin a dyad by about 0.13 – a decline of about 1.7% in the average higherdemocracy value in the sample (7.9). Similarly, because dyadic conflict alsoaffects the higher level of democracy in the next period through its effect onthe lagged higher democracy in the model, we compute the effect of MIDinvolvement on the higher level of democracy in the long run. By employingthe coefficients of dyadic MID dispute and the lagged higher democracy, wefind that dyadic MID involvement reduces the higher level of democracy byabout 1.6 points over the long run – a 20% decline in the average value of thevariable in the sample. These short- and long-run effects of dyadic conflicton regime dissimilarity are not as large as their effects on joint democracy.

Based on the results for the MID equation, the effect of demL on thelikelihood of midAB is negative and statistically different from zero. Hence,

Tabl

e5.

2.In

tera

ctio

nsam

ong

join

tdem

ocra

cy,r

egim

edi

ssim

ilari

ty,a

nddy

adic

confl

ict

Equ

atio

nof

DE

ML

Equ

atio

nof

DE

MH

Equ

atio

nof

MID

PR

IOR

DE

MO

CR

AC

YL

0.93

64∗∗

∗P

RIO

RD

EM

OC

RA

CY

H0.

9163

∗∗∗

CO

NT

IGU

ITY

0.78

29∗∗

∗

(0.0

053)

(0.0

088)

(0.1

375)

YE

AR

0.06

19∗∗

∗Y

EA

R0.

0547

∗∗∗

AL

LIA

NC

E−0

.090

7(0

.012

5)(0

.019

3)(0

.138

8)IN

FLA

TIO

NL

−0.0

385∗∗

INFL

AT

ION

H−0

.087

2C

APA

BIL

ITY

RA

TIO

−0.0

285

(0.0

150)

(0.0

647)

(0.0

409)

INFL

AT

ION

L∗ Y

EA

R0.

0000

2∗∗IN

FLA

TIO

NH

∗ YE

AR

0.00

004

TR

AD

ED

EP

EN

DE

NC

E−2

0.38

36(7

.52e

-06)

(0.0

0003

)(1

4.50

72)

GD

PP

CL

13.4

305∗∗

∗G

DP

PC

H12

.175

0∗∗∗

GR

OW

TH

−0.0

287∗∗

(2.9

239)

(4.2

280)

(0.0

116)

GD

PP

CL∗ Y

EA

R−0

.006

6∗∗∗

GD

PP

CH

∗ YE

AR

−0.0

061∗∗

∗D

EM

L−0

.020

7∗∗∗

(0.0

015)

(0.0

021)

(0.0

067)

GR

OW

TH

L−0

.007

1∗∗G

RO

WT

HH

−0.0

053

DE

MH

0.00

27(0

.003

2)(0

.004

6)(0

.010

6)T

RA

DE

L−0

.427

6∗∗∗

TR

AD

EH

−0.1

059

AFF

INIT

Y(0

.084

4)(0

.068

5)D

IFFU

SIO

NL

0.00

91∗∗

∗D

IFFU

SIO

NH

0.00

47∗∗

∗C

onst

ant

−2.1

944∗∗

∗

(0.0

011)

(0.0

010)

(0.1

493)

TH

IRD

-PA

RT

YM

IDL

0.08

45T

HIR

D-P

AR

TY

MID

H0.

0701

∗∗N

1164

4(0

.053

5)(0

.034

8)�

296

.73∗∗

∗

MID

AB

−0.1

597∗∗

∗M

IDA

B−0

.132

7∗∗∗

(0.0

418)

(0.0

310)

Con

stan

t−1

24.9

733∗∗

∗C

onst

ant

−108

.728

6∗∗∗

(24.

7767

)(3

8.06

68)

N11

644

N11

644

Adj

ust

edR

20.

95A

dju

sted

R2

0.94

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗

∗si

gnifi

can

tat

5%;∗

∗∗si

gnifi

can

tat

1%.

142


Table 5.3. Probabilities and relative risks of MID involvement

Probability Relative risk

Scenario 1: Contiguous, nonallied dyadsBaseline: all continuous variables are set at mean values 0.0546Increase in DEML by one standard deviation; other

variables at baseline values0.0402 73.6%

Scenario 2: Noncontiguous, nonallied dyadsBaseline: all continuous variables are set at their mean

values0.0086

Increase in DEML by one standard deviation; othervariables at baseline values

0.0057 66.3%

Note: Based on coefficients in MID equation in Table 5.2.

democratic countries are less likely to engage in MIDs against each other.This result agrees with those reported by many single-equation studies insupport of the democratic peace proposition. The effect of demH on midAB

is not statistically different from zero. Russett et al. (1998) and Oneal andRussett (1997) argue that demH is an important part of the link betweenpolitical regimes and conflict. For example, interpreting demH as politicaldistance, Russett et al. (1998: 457) argue that “greater political distancemakes a dyad more prone to conflict.” The result in Table 5.2 does notsupport this claim, however; the difference is attributable to the fact that wemodel demH–conflict simultaneity, whereas Oneal and Russett (1997) andRussett et al. (1998) do not.

Table 5.3 illustrates the substantive effects in the MID equation bycomputing the probabilities of MID for contiguous and noncontiguouspolitically relevant dyads. Scenario 1 sets all the continuous variables attheir mean values, contiguity at 1 (contiguous countries), and allianceat 0 (nonallied countries). Scenario 2 repeats the experiment but setscontiguity at 0 (noncontiguous countries).

In Scenario 1, the baseline probability of dyadic MID involvement is5.46%. When demL rises by one standard deviation (6.48 units), the prob-ability of midAB drops to 4.02%, a decline of 1.44%. Because conflicts arerare events, it makes more sense to talk about the relative risk between thetwo scenarios (i.e., dividing the new probability by the baseline probabil-ity). The computed relative risk is 73.6%, which means that relative to thebaseline contiguous dyad, the risk of MID involvement is 26.4% lower ifthe dyad’s joint democracy rises by one standard deviation. How do theseeffects compare with those found by Oneal and Russett (1997)? We observe


a smaller absolute effect of joint democracy on the likelihood of dyadic MID(a decline of 1.44% in our results vs. a decline of 2.1% from 7.8% to 5.7%in their results). But the changes in the relative risk are similar between thetwo studies (26.4% vs. 26.9% from Oneal and Russett).

In Scenario 2 (noncontiguous countries), the substantive effect of demL

(joint democracy) on dyadic MID involvement is smaller by an order ofmagnitude. A rise of one standard deviation in demL reduces the probabilityof midAB (dyadic MID involvement) from 0.86% (in the baseline case) to0.57%. The results of this experiment, which are not reported by Onealand Russett (1997), lend support to the argument of James et al. (1999)that the effect of joint democracy on the probability of dyadic MID may besmall. However, the relative risk is 66.3%. Relative to the baseline (0.86%),the reduction of 0.29% amounts to a decrease of 33.7% in the relative riskof MID involvement. That is, relative to the baseline noncontiguous dyad,the risk of MID involvement drops by 33.7% if the dyad’s joint democracyincreases by one standard deviation.

Overall, the empirical results support our theoretical expectation that therelationship between dyadic military conflict and democracy is reciprocal –they provide support for our argument that models that ignore the simul-taneity of conflict and democracy miss something important about thisstructural relationship. Dyadic military conflict reduces both the lower andhigher democracy levels in a dyad, although its short- and long-run effectson the lower democracy level are much stronger. At the same time, the lowerdemocracy level in a dyad reduces the probability of dyadic military dispute,whereas the higher democracy level in a dyad does not.

Additional Analysis

One additional analysis concerns the debate over the impact of affinity,which presumably measures preference similarity based on UN voting pat-terns. We present the additional results in Table 5.A1, based on a simulta-neous equations model with affinity in the MID equation. The purposeis to check the sensitivity of our results in light of the debate between ErikGartzke on one side and Bruce Russett and John Oneal on the other. Thisdebate is not our main focus, but it provides an opportunity to evaluate therobustness of our results. For our purpose, we compare the results betweenTables 5.2 and 5.A1 and find them quite similar across all three equations.The effects of demL and demH and midAB on each other are similar whetheraffinity values are included or not.


As we further explain in the appendix, Maddala developed a method toestimate a simultaneous equations model with one endogenous continuousvariable and one endogenous dichotomous variable. However, his methodonly works in this particular case and we therefore could not use it forour three-equations model. Thus, we conduct another analysis to assess thesensitivity of our results to the Maddala correction of standard errors. Inthis additional analysis, we estimate a model with two endogenous variables,midAB and demL. In other words, we drop the demH equation. We do thisestimation in two ways. In one way, we use the Maddala method per se,including the special features it includes. In the second way, we do notuse the full Maddala method but rather we use its two-stage features andcompute robust standard errors with cluster, the approach we employedearlier for the three-equations models. The results from these two tests arevery similar. We therefore expect that had a Maddala method been availableand applied to our three-equation setup, it would not have changed theinferences reported here.


Our simultaneous equations model of dyadic conflict and democracy ismarkedly distinct from the typical single-equation research designs in thedemocratic-peace literature. In previous studies, midAB was assumed to bethe dependent variable and the simultaneity of conflict and democracy wasignored. We demonstrate that conflict and democracy do in fact influenceone another. Our major innovation is the combining of the dyadic analysisof military conflict with the monadic determinants that explain democracylevels in a dyad. Our approach exploits the insights from two bodies of lit-erature that have remained on two separate courses: the monadic literatureon the causes of democratization and the dyadic democratic-peace litera-ture. The empirical findings support our choice of a simultaneous modelingapproach.

The simultaneous modeling approach in this chapter brings several newthings to the study of democracy and dyadic conflict. First, it shows demo-cratic peace is not an artifact of the effect of conflict on democracy, whichhas been a topic of debate in the literature. Second, the analysis showsthat the pacifying effect of joint democracy is smaller in the absolute sizethan in the work of Oneal and Russett once one considers the democracy–conflict simultaneity. Third, it shows that dyadic conflict reduces bothjoint democracy and dyadic regime dissimilarity, rejecting the view that


conflict promotes democracy. Finally, it shows that the congruence of inter-est between countries does not explain the lack of conflict between democra-cies in the context of a simultaneous equations framework, an issue debatedamong scholars who use single equations for the effect of democracy onconflict.

Our findings suggest two policy implications that single-equation stud-ies of MID fail to uncover. First, there is a negative relationship betweendyadic conflict and joint democracy. Dyadic conflict reduces joint democ-racy, which makes future conflict more likely and moves the dyad awayfrom peace. Hence, the preservation of peace requires early intervention indisputes before they cause the belligerents to become seriously less demo-cratic. Second, there exists a positive relationship between dyadic peace anddemocracy. Peaceful dyads experience a rise in democracy for both nations.Thus, by brokering peace between adversaries, actors such as major pow-ers or international organizations can help to facilitate the development ofdemocracy, which in turn promotes peace. Although these effects may notbe large, they are statistically significant and could well be important toolsin conflict resolution.

We need to stress that this chapter represents the beginning of a researchprogram rather than a final analytic effort. We believe that the field ofinternational relations could gain insight by repeating previous investiga-tions of the relationship between conflict and democracy while using oursimultaneous framework. In particular, we believe that the issue of the sizeof the effect of joint democracy on the likelihood of MIDs merits furtherinvestigation. For example, studies could use other control variables, otherestimators, or other measures to gain additional insight. However, in ourview, the simultaneity of conflict and democracy is too important to beignored by future research.

SUMMARY AND OUTLOOK

Many statistical studies in international relations investigate the claim thatdemocracies do not fight one another. Virtually all these studies employ asingle-equation design, where the dependent variable measures the presenceor absence of a dyadic MID. A separate group of studies argues that conflictaffects democracy and that its effects could be positive or negative. By andlarge, these two bodies of literature have not incorporated one another’sinsights. We argue that democracy and dyadic conflict affect each othersignificantly and that statistical models that ignore the reciprocal nature ofthese effects may make incorrect inferences.


In this chapter, we developed a simultaneous equations model of democ-racy and dyadic conflict. The model not only connects the monadic literatureon the causes of democratization with the dyadic democratic-peace litera-ture, but it also includes, in an innovative manner, the monadic causes ofdemocracy and the dyadic causes of military conflict in one system of equa-tions. In a sample of all the politically relevant dyads from 1950 to 1992,we find that dyadic military disputes reduce joint democracy, and jointdemocracy reduces the probability of MIDs. The effect of military conflicton joint democracy is large. And the absolute effect of joint democracy onconflict in our study is smaller compared with the single-equation estimatesin the literature, whereas, in relative terms, the effect is similar in size. Theeffect of joint democracy on MID involvement is considerably smaller fornoncontiguous countries than for contiguous ones.

In this chapter, international trade played a role in the analysis as oneof our control variables in both the democracy and the military conflictequations. The effect of trade on military conflict, however, is one of thoseissues that in itself has generated enormous controversy in the fields ofinternational relations and international political economy. Studies havecontinuously debated the merits of trade as a tool to achieve internationalpeace, and the issue is certainly far from reaching closure. Part of thereason for this controversy probably has to do with some previously notedtechnical differences in terms of samples, measures, estimators, and controlvariables.

In the next chapter, we argue that another important substantive sourceof differences, which so far has not received much scholarly attention, hasto do with the composition of trade in terms of both sectoral types andthe directions of trade flows (import and export). Our work focuses onthese very issues in the following order. We first offer a formal model of theeffects of trade, broken down along sectoral type and trade flow directions,on the propensity of nations to initiate conflict against their trading part-ners. The model is then tested empirically using statistical methods. As weshall demonstrate, our work will bring new insight into this relatively olddebate.

APPENDIX

CONCEPTUAL MODEL

The conceptual model focuses on linking the monadic causes of democracyand the dyadic causes of military conflict. Recall that demL is the lower


democracy score between demA and demB in dyad AB, where A and B arethe dyad member countries; demH is the higher democracy score in dyadAB; midABU is the propensity of dyad AB to engage in a militarized dispute –a variable that is unobserved; MA is a vector of monadic attributes affectingthe democracy of A (demA); MB is a similarly defined vector for B (demB);and F denotes a functional dependency.

demA is a function of three components: MA, midABU, and midAJU. Sim-ilarly, demB is a function of MB, midABU, and midBJU. The elements ofthe vectors MA and MB are based on the literature of the determinants ofdemocracy discussed in Chapter 2. The democracy score of country A isaffected by any conflict that country A conducts with any country in itsPRIE (see the text) and not only by conflict with country B.

We divide the PRIE of country A into two parts: country B and otherthird-party countries (denoted J, where J �= B). The propensity of conflictbetween A and B has already been denoted as midABU. The unobservedaggregated propensity for conflict that country A has with all third-partycountries J in its PRIE is denoted by midAJU. Hence, in notation:

demA = F(MA, midABU, midAJU)

demB = F(MB, midABU, midBJU). (5.A1)

Combining the definition of demL and Equation (5.A1), we get

demL = F(MB, midABU, midBJU) if demA> demB;

demL = F(MA, midABU, midAJU) if demA< demB. (5.A2)

Similar to demL, the derivation of demH is given by

demH= F(MA, midABU, midAJU) if demA> demB;

demH= F(MB, midABU, midBJU) if demA< demB. (5.A3)

midABU depends on the democracy scores of both countries in the dyad.Hence,

midABU= F(XAB, demL, demH), (5.A4)

where XAB is a vector of dyadic attributes that affect the likelihood of dyadicMID. The variables included in XAB are those used within the conflict liter-ature. The democratic-peace literature expects that a rise in demL reducesmidABU and that a rise in demH raises midABU.


The full simultaneous equations model is now given by Equations (5.A5),(5.A6), and (5.A7), which were presented in the text:

demL = F(ML, midABU, midLRU), (5.A5)

demH = F(MH, midABU, midHRU), (5.A6)

midABU = F(XAB, demL, demH). (5.A7)


Empirical Model

The elements of the vectors ML and MH in Equations (5.A3) and (5.A4) ofthe conceptual model include the following variables. We provide details ondata sources, expected effects, and measure construction. The discussion ofthe democracy equations follows Chapter 2.

gdppcL and gdppcH are the logged real GDP per capita values for thelow- and high-democracy countries in the dyad, respectively. These data aretaken from the Penn World Table 5.6. As noted in Chapter 2, many studiesuse GDP per capita as an indicator of economic development. We expectthat gdppcL and gdppcH positively influence demL and demH, respectively,an expectation that is in line with the modernization theory – that economicdevelopment breeds democracy.

growthL and growthH are the yearly growth rates of real GDP, com-puted using data from the Penn World Table 5.6, for the low- and high-democracy countries in a dyad, respectively. As noted in Chapter 2, thedemocracy literature registers conflicting expectations regarding the effectof economic growth on democracy. Some studies argue the effect is negative,and other studies claim the effect is positive.

inflationL and inflationH are the yearly inflation rates, based on theconsumer price index, for the low- and high-democracy countries in a dyad,respectively. Several studies use inflation as a proxy for economic crisis, butscholars debate its effect on democracy (see Chapter 2). Data are from theInternational Financial Statistics CD-ROM (2000).

tradeL and tradeH are the trade openness ratios for the low- and high-democracy countries in a dyad, respectively. Trade openness is calculatedas the sum of the values of imports and exports of a country divided byits GDP. Data are from the Penn World Table 5.6. As noted in Chapter 2,the debated effect of trade openness on democracy could be positive ornegative.


year tests whether democracy has a linear trend. In addition, the effects ofinflation and economic development on democracy may change over time.As we performed and discussed in Chapter 2, to capture this possibility weinclude interaction terms between year and inflation, and year and GDPper capita, in the equations for demL and demH, respectively.

prior democracyL and prior democracyH are the prior democracyvalues for the low- and high-democracy countries in a dyad, respectively. Weexpect these variables to influence demL and demH positively. The strategyof including the lagged dependent variable in the model was explained inChapter 2.

diffusionL and diffusionH are the averages of the democracy scores ofthe countries in the PRIEs of the low- and high-democracy countries in adyad, respectively. As in Chapter 2, they capture the diffusion of democraticnorms due to contact-based mechanisms such as economic relations, com-munication networks, and influence by major powers, but in this chapterthey are computed for the PRIEs of the low- and high-democracy countries,not their geographical regions. We expect the effect of diffusion on demL

and demH to be positive.third-party midL and third-party midH are dichotomous variables.

The value of 1 denotes a situation in which the low- or high-democracycountry in a dyad is involved in conflict with at least one third-party countryin its respective PRIE. These variables capture the notion that the democracylevel of country A, for example, may be affected by its conflict with anythird-party country, not just by its conflict with country B. Drawing onthe literature behind Hypothesis H2, we expect that the effect of third-party conflict on democracy could be positive or negative. The third-partydummy variable has been coded based on Zeev Maoz’s (2001b) data onthird-party conflict in the PRIE of each country in a dyad.

Moving to the MID equation, unless otherwise specified we use data fromOneal and Russett (1999a). As noted in the empirical test, the unobservedmidABU variable is replaced by its realization midAB, which is the presenceor absence of a MID in dyad AB. Another possibility is to replace midABU

with a multinomial representation, classifying several types of MIDs, as inJames et al. (1999). However, this method is not employed here becausethe estimation of simultaneous equations models involving continuous andmultinomial variables is not well developed in the literature and because wewish to compare our results to the bulk of the democratic-peace literature,which employs the dichotomous operationalization of MIDs.

contiguity takes the value of 1 when the two states in a dyad are geo-graphically contiguous (or are separated by up to 150 miles of water) and


zero otherwise. Contiguous states are expected to be more likely to experi-ence conflict with one another.

alliance equals 1 when both states within a dyad are members in a com-mon military alliance, and zero otherwise. Alliance partners are expected tobe less likely to experience conflict with one another.

The capability ratio is the natural logarithm of the ratio of the morecapable state’s composite national capabilities score to that of the less capablestate in the dyad. For each country, this score rises as a state commands alarger resource base, comprising total population, urban population, energyconsumption, iron and steel production, military manpower, and militaryexpenditures. Dyads with an imbalance in capabilities are expected to beless likely to experience conflict with one another.

trade dependence measures dyadic trade interdependence. Each statein a dyad has a bilateral-trade-over-GDP ratio, which represents the relativeimportance of trade with a dyadic partner. Following Oneal and Russett, weuse the lower of the two trade dependence values in a dyad to measure thecommon level of dyadic trade dependence shared by the two states. Manyscholars expect that dyadic trade dependence reduces conflict involve-ment, but the issue is debated.

growth is the lower of the two economic growth rates in a dyad com-puted from the rate of change in the real GDP per capita for each state. Thegrowth rates are obtained by using moving averages over a three-year period(substituted by one- or two-year-period averages when required values aremissing). The literature offers conflicting expectations about the directionof this variable’s effect. This variable is computed using data from PennWorld Table 5.6.

affinity measures the level of preference similarity between two statesin a dyad. This variable is based on the similarity of the states’ votingpatterns at the UN General Assembly. Data are from Gartzke (1998). Severalscholars debate whether it is affinity or joint democracy that causes peacein a dyad. Gartzke (1998) argues that preference similarity accounts forthe hypothesized effect of democratic peace and reports that in the samemodel, affinity is statistically significant whereas joint democracy is not. Incontrast, Oneal and Russett (1999b) argue that joint democracy indirectlycontributes to preference similarity. Gartzke (2000) rejects this possibility.


The relationship between some of our exogenous and endogenous variablesmay be simultaneous. Although this possibility applies to virtually all studies


in the literature, we nevertheless believe we ought to discuss the issue. Ofcourse, we cannot model all potential simultaneous effects between ourvariables, because our model is already quite complicated. To deal with thisissue, we lag each of our exogenous variables by a single year, as we did inprevious chapters.

The next issue concerns the Heckman (1978) condition for our model.Heckman shows that a consistency requirement needs to be satisfied whenone estimates a simultaneous equations model in which an unobservedendogenous variable is replaced by an observed dichotomous variable. Ifthis condition is not fulfilled, the estimation is not meaningful because thestructural equations are not consistent with one another (Amemiya, 1978).Heckman considers a model with two equations: one for the observed vari-able, and the other for the unobserved. The dichotomous realization of theunobserved variable is an independent variable in each equation. Supposethe coefficient of the endogenous variable in the equation of the unobservedvariable is � 2, and the coefficients of the dichotomous variable (linked tothe unobserved variable) are �1 and �2 in each equation, respectively. Heck-man’s condition requires �2 = � 2�1. In our case, Heckman’s condition holdsbecause the coefficients of midAB in all equations are zero. That is, the demL,

demH, and midABU equations do not include midAB as an independent vari-able. In addition, � 2 �= 0 in our case, because the coefficients of demL anddemH in the midABU equation are not zero. Hence, we can say that in ourcase the Heckman condition holds.

We estimate our model using a variant of the two-stage least squaresmethod suggested by Maddala (1983) and pooling the politically relevantdyad from 1950 to 1992. As in other studies, we do not deal with the time-series dynamics of democracy and conflict. Dealing with issues of dynamicswould require time-series techniques, which are still not well understood inthe context of MIDs. We defer this issue to future research.

In the first stage of Maddala estimator, each endogenous variable isregressed on all the exogenous variables in the model. The midAB equationis estimated using probit. The equations of demL and demH are estimatedusing ordinary least squares. In the second stage, the endogenous variableson the right-hand side of each equation are replaced by their predictedvalues from the first stage, respectively. In the MID equation, the linearpredictor is used – not the predicted probability.

Maddala (1983) assumes that the error terms are white noise. He employsthe coefficients from the second stage, as we do, but he applies a correctionfor their standard errors. Our model consists of one dichotomous and twocontinuous endogenous variables and we do not assume that the error termsare necessarily white noise. We do not use Maddala’s correction because,


to the best of our knowledge, the literature has not extended it either to amodel with more than two equations or to a case with possible nonsphericalerrors.

Following the suggestions of Guilkey et al. (1992), Bollen et al. (1995),and Alvarez (1997), we use the standard errors computed in the second stageof the procedure to test our hypotheses. However, to assess the sensitivity ofour results to Maddala’s method, we estimate a model with two endogenousvariables (midAB and demL) in two ways: one with the Maddala correctionand the other with robust standard errors (that can control for potentialnonspherical errors).

We need to consider the potential existence of heteroskedasticity andserial correlation in the model’s error terms. One may suggest that weexamine the correlation structure of the error terms in the equations. So farthe literature has not developed a diagnostic test that is appropriate for ourcase. Developing such a test is difficult for at least two reasons. First, oneof the equations in our model has a dichotomous endogenous variable ina time-series cross-sectional design. Second, we have a structural equationsmodel with one dichotomous and two continuous endogenous variables.Conventional tests such as Durbin–Watson are not designed to diagnosesuch models. Hence, future research is called for in this area.

That said, in the presence of heteroskedasticity and serial correlation,a model’s estimated coefficients are consistent, but its standard errors areinefficient and could be biased. To correct for these potential problems,we estimate our model using a variant of the White (1980) estimator ofrobust standard errors in the presence of heteroskedasticity and adjust forclustering over dyads to help account for possible temporal dependence ofthe error terms within each dyad. This estimator yields consistent estimationof the covariance matrix under very general conditions of heteroskedasticityand serial correlation (Wiggins, 1999).

Finally, we demonstrate the importance of the results by examining theeffect size of the key variable. We look at the direct effect of the key variablesand their long-run effects through the lagged endogenous variables in thetwo democracy equations. The long-run impact of dyadic MID on demo-cracy is based on the following formula: [coefficient of MID variable/(1 −coefficient of lagged endogenous democracy variable)].

Empirical Findings

In this part of the appendix, we provide additional details on the sampleand some diagnostics and discuss results for the control variables. Thesample sizes for Models 1 and 2 in the tables shown in the text are 11,644


and 11,022 observations, respectively. These samples are somewhat smallerthan the samples used in some single-equation-based studies of MIDs,which is due to the fact that our model includes more variables than doesthe typical single-equation design and the data on some of these variablesare missing for some years. While this places a potential limitation on ourempirical analysis, we believe that our results are applicable to other samples,because our theoretical argument is not sample-dependent and our sampleis nevertheless large.

Table 5.2 in the main text presents the statistical results for all threeequations. As noted in the text, the model’s goodness of fit is high. Theadjusted R2 for the democracy equations is 0.94 or 0.95, and the �2 modelfit test in the MID equation is statistically significant.

We now discuss the results for the control variables in the demL equa-tion. The effect of prior democracyL on demL is significant and pos-itive, demonstrating the importance of democratic inertia. The effect ofdiffusionL is positive. As the level of democracy in the less democraticcountry’s PRIE rises, so does its own democracy level.

The effect of year is significant and positive, reflecting the tendency ofdemocracy to increase over time in our sample. The effect of inflationL ondemL is significant and negative, and its interaction with year is significantand positive. Hence, a rise in inflation reduces a country’s level of democracy.This effect is not stable over time, however, as is also reported by Gasiorowski(1995) and in Chapter 2. The effect of GDP per capita on demL is significantand positive, whereas its interaction term with year is significant andnegative. This replicates a known result in the literature; that is, democracyrises with economic development. We also find, however, that this effectdeclines over time.

The literature offers contradicting expectations regarding the effect ofeconomic growth on democracy. We find that this effect is significant andnegative. As in Chapter 2 and Gasiorowski (1995), we also find that greatertrade openness reduces the lower democracy score in a dyad. One inter-pretation is that trade widens the social cleavages between winners andlosers, thus reducing democracy. This finding is consistent with the resultin Chapter 2.

Next we turn to the results for the demH equation. As one could expect,the results for the control variables are generally consistent with those fromthe demL equation, which adds justification for our modeling approach.The only exceptions are the effects of inflation, growth, and trade ondemH, which are no longer significant in spite of having the same signs asin the DEML equation. This is not surprising because the high-democracy


countries in a sample of politically relevant dyads tend to be developedcountries, for which the levels of democracy tend to be high and stable. Onthe other hand, the low-democracy countries are typically less developed,exhibiting larger changes in democracy over time.

In our sample, most of the MIDs are purely dyadic. Thus, we oughtto expect that democracy levels would mostly be affected by the variablemidAB. However, despite their relative rarity, third-party disputes also couldaffect democracy. We find that the effects of third-party mid on demH

and (less so) on demL are positive and statistically significant. We canexplain this empirical result by noting that the demH countries in a sam-ple of politically relevant dyads tend to be countries such as the UnitedStates, the United Kingdom, France, Israel, and West Germany. Between1950 and 1992, these countries were at times involved in multiple mili-tarized disputes – a fact that has manifested itself in the positive corre-lation between demH and third-party mid (correlation = 0.136). ThedemL countries in the sample are typically less developed and tend to beinvolved in fewer MIDs with third parties, which is also demonstrated bya positive but weaker correlation between demL and third-party mid(correlation = 0.049).

Finally, we discuss the results for the MID equation in Table 5.2. Theeffect of contiguity on the likelihood of MIDs is statistically significantand positive, as one would expect from the extant literature. The effectof growth on midAB is statistically significant and negative, which alsois consistent with some prior results. The effects of alliance, capabilityratio, and trade dependence, however, are not statistically significantin our results, though they have the same sign as in comparable single-equation studies. We attribute these differences to the fact that our modelaccounts for the simultaneity of conflict and democracy, whereas previousstudies have not.

Additional Analyses

Table 5.A1 reports the results for all three equations based on the additionalanalysis and concerns the debate over the impact of affinity. The three-equation model adds the affinity variable to the MID equation. As noted,these results shed light on a recent debate in the single-equation MIDliterature. Gartzke (1998, 2000) argues that democratic peace is driven bydyadic preference similarity. In his studies, when affinity is included,the effect of joint democracy on the likelihood of MID is not statisticallysignificant. Oneal and Russett criticize Gartzke’s approach and reject his

Tabl

e5.

A1.

Inte

ract

ions

amon

gjo

intd

emoc

racy

,reg

ime

diss

imila

rity

,and

dyad

icco

nflic

t,co

ntro

lling

for

affin

ity

Equ

atio

nof

DE

ML

Equ

atio

nof

DE

MH

Equ

atio

nof

MID

PR

IOR

DE

MO

CR

AC

YL

0.93

46∗∗

∗P

RIO

RD

EM

OC

RA

CY

H0.

9143

∗∗∗

CO

NT

IGU

ITY

0.80

41∗∗

∗

(0.0

056)

(0.0

090)

(0.1

346)

YE

AR

0.06

15∗∗

∗Y

EA

R0.

0602

∗∗∗

AL

LIA

NC

E−0

.054

3(0

.013

1)(0

.020

4)(0

.157

7)IN

FLA

TIO

NL

−0.0

388∗∗

INFL

AT

ION

H−0

.084

1C

APA

BIL

ITY

RA

TIO

−0.0

498

(0.0

150)

(0.0

649)

(0.0

445)

INFL

AT

ION

L∗ Y

EA

R0.

0000

2∗∗IN

FLA

TIO

NH

∗ YE

AR

0.00

004

TR

AD

ED

EP

EN

DE

NC

E−1

7.08

85(7

.56e

-06)

(0.0

0003

)(1

3.71

60)

GD

PP

CL

13.1

769∗∗

∗G

DP

PC

H13

.412

3∗∗∗

GR

OW

TH

−0.0

261∗∗

(3.0

734)

(4.4

671)

(0.0

123)

GD

PP

CL∗ Y

EA

R−0

.006

5∗∗∗

GD

PP

CH

∗ YE

AR

−0.0

067∗∗

∗D

EM

L−0

.018

8∗∗∗

(0.0

016)

(0.0

023)

(0.0

065)

GR

OW

TH

L−0

.006

6∗∗G

RO

WT

HH

−0.0

060

DE

MH

−0.0

018

(0.0

033)

(0.0

048)

(0.0

110)

TR

AD

EL

−0.4

638∗∗

∗T

RA

DE

H−0

.138

4∗A

FFIN

ITY

−0.2

145

(0.0

886)

(0.0

722)

(0.2

127)

DIF

FUSI

ON

L0.

0096

∗∗∗

DIF

FUSI

ON

H0.

0050

∗∗∗

Con

stan

t−2

.069

9∗∗∗

(0.0

012)

(0.0

011)

(0.1

777)

TH

IRD

-PA

RT

YM

IDL

0.09

68∗

TH

IRD

-PA

RT

YM

IDH

0.06

35∗

N11

022

(0.0

557)

(0.0

360)

�2

82.7

5∗∗∗

MID

AB

−0.1

742∗∗

∗M

IDA

B−0

.129

5∗∗∗

(0.0

434)

(0.0

316)

Con

stan

t−1

24.3

320∗∗

∗C

onst

ant

−119

.726

3∗∗∗

(25.

9593

)(4

0.20

63)

N11

022

N11

022

Adj

ust

edR

20.

95A

dju

sted

R2

0.94

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗

∗si

gnifi

can

tat

5%;∗

∗∗si

gnifi

can

tat

1%.

156


findings. Our model facilitates a more accurate assessment of this debate bymodeling the reciprocal relationship between demL and midAB.

We find that the effect of affinity on the likelihood of dyadic MIDinvolvement is negative but statistically insignificant. On the other hand,the effect of joint democracy on the likelihood of MID involvement remainsnegative and significant. Our results thus support the position of Oneal andRussett in this debate. However, it is worth noting that affinity, based onUN voting patterns, may be a weak proxy for dyadic preference similarity.Hence, the question of whether preference similarity affects dyadic MIDinvolvement would benefit from the development of better measures of thisconcept in future research.

Finally, as noted, to assess the sensitivity of our results to the Maddalacorrection of standard errors, we estimate a model with two endogenousvariables (midAB and demL) in two ways: using the Maddala correctionand using robust standard errors. Keshk (2002) provides details on theimplementation of this estimator. The results from these tests are verysimilar, which is consistent with the findings from Monte Carlo simulationsreported by Guilkey et al. (1992) and Alvarez (1997). They find that theMaddala correction has little effect on statistical inferences. We thereforeexpect that had a Maddala correction been available and applied to ourthree-equation setup, it would not have changed the inferences reportedhere.

SIX

Economic Openness and Conflict

INTRODUCTION

In this chapter, we shift our attention to the interaction between militaryconflict and the other focus of our book – economic openness – whichplayed a crucial part in Chapters 2 and 3. Economic openness is measuredin various ways, but in this chapter we emphasize the trade dimension ofeconomic openness, which connects the analytical framework of our bookto an important debate in the related fields of international relations andinternational political economy: the trade and conflict controversy.

The idea that trade promotes peace dates from at least the late eighteenthcentury,1 and has two contemporary explanations: the liberal argumentand the bargaining argument. The antithesis that trade generates conflictalso has a long intellectual history,2 which is often explained from a neo-Marxist view or a neomercantilist derivative of realism. So far, scholarshave mathematically demonstrated the logical consistency of the liberaland bargaining arguments.3 These formal models consider aggregated ortotal trade, ignoring variations of trade across economic sectors and flowdirections (export and import). With a few exceptions, statistical studieshave also used total bilateral trade. Most of these studies find that a risein trade reduces the likelihood of conflict. It is a fair statement that mosttheoretical and empirical works on the effect of trade on conflict ignoreboth the composition of trade across economic sectors and the trade flowdirections in terms of export and import.

We challenge the prevalent approach and provide new insights on theeffect of trade on conflict. Indeed, it does not seem likely that international

1 See, e.g., de Montesquieu (1748), Smith (1776), Kant (1795), and Angell (1912).2 See, e.g., Hobson (1902), Lenin (1916), Hirschman (1945), and Waltz (1970).3 See, e.g., Polachek (1980) and Gartzke et al. (2001).

158

Economic Openness and Conflict 159

exchanges of heterogeneous commodities and services (say, oil, bananas,corn, machinery, fish, and computers, to name just a few) should haveidentical effects on conflict. As one renowned scholar has repeatedly warned,“unduly extensive aggregation of trade data” is likely to produce “biasedestimates of the ability of bilateral trade to diminish conflict” (Polachek,1992: 113; 2000: 10).4 We suggest that the theoretical and empirical focuson total trade results in two aggregation biases: one from aggregation acrosseconomic sectors or goods and the other from aggregation across tradeflow directions (import and export).5 To be sure, some studies do look atthe effects of trade in different sectors on military conflict – but withoutformally providing a microfoundation – and almost all the existing statisticalstudies of this subject except that of Dorussen (2006) have used very limitedsamples. Meanwhile, to the best of our knowledge, the possibility that theeffects of trade on military conflict may differ between that of export andimport has not received attention in the literature.

In this chapter, we offer a theoretical formal model to explain howexport and import flows in specific economic sectors influence the deci-sion to initiate military conflict, a question almost all previous theorieshave ignored. Like virtually any formal model, our model also relies onsimplifying assumptions. As in neoclassical economics, we model trade asperformed by economic agents in the marketplace. As in international rela-tions, we model conflict as a choice made by leaders. We believe that theseassumptions are generally reasonable and allow us to model and anticipatethe individual effects on military dispute initiation of bilateral imports andexports in different sectors.

Simply put, our theory suggests that leaders are less likely to initiateconflicts that cause trade losses. This idea is not new, but our first contri-bution here is to show that this consideration could vary across import,export, and traded sectors. And our theory goes beyond this argument: itpredicts that leaders are more likely to initiate conflicts that increase profitsto their trade. Formalizing these ideas, we show that the costs and the ben-efits of conflict can vary across trade sectors, exports and imports, and thatthey work through the expected effect of conflict on trade prices. Leadersinitiate conflict if they conclude that it will be economically beneficial by

4 Polachek is not alone. Russett and Oneal (2001: 141) also note that “dyadic [total] trade,even when adjusted for the size of the overall economy, is an imperfect indicator of economicinterdependence. For one thing, the composition of trade is not considered.” Surveys of thetrade and conflict literature conclude that researchers should consider disaggregating total trade.See Sayrs (1990: 22), McMillan (1997: 53), and Reuveny (2000: 37).

5 We use the terms “goods,” “economic sectors,” and “sectors” interchangeably.


reducing the price of their import and increasing the price of their export,ceteris paribus. The decision depends on certain good-, export- and import-specific parameters.

Using estimates of these parameters, our model predicts the effects ofchanges in exports and imports in the agriculture/fishery, energy, chemical/mineral, manufactured products, and miscellaneous-consumption sectorson military interstate dispute (MID) initiation. We test these predictionsusing a large-N sample of all the directed-dyads from 1970 to 1997 forwhich data are available. We find that the effects of bilateral trade on MID-initiation vary across sectors and between flow directions. Increases in agri-culture/fishery imports and energy imports reduce the probability of MIDinitiation, and increases in energy exports, manufacturing imports, andmanufacturing exports increase this probability. Yet changes in agriculture/fishery export, chemical/mineral trade, and miscellaneous consumptionproducts trade generally do not affect MID initiation.

Our research, we believe, offers important insights for both academicsand policymakers. As we explain in detail in the concluding section, ouranalysis sheds new light on a long-standing debate in international relationsand offers an opportunity to rethink the logic of how trade affects conflict.Our theory encompasses the liberal argument as a special case and offersan alternative explanation to the bargaining, neo-Marxist, neomercantilistrealist, and classical realist arguments. It also has various important policyimplications.

The remainder of this chapter is organized as follows. The next sectionpresents the motivations of our analysis and discusses previous studies,which is followed by three sections that develop the theoretical model, asection that discusses the model’s implications and formulates hypothe-ses, and a section that presents the empirical model and discusses severalresearch design issues. With these preliminaries out of the way, we turn tothe statistical results of our analysis and the public policy implications ofthese findings. The last section summarizes the analysis and links the dis-cussion to the next chapter. The appendix of this chapter presents technicaldetails pertaining to the design of our analysis and the empirical results.

DISAGGREGATED TRADE IN THE CONFLICT LITERATURE

The contemporary literature on the effect of trade on conflict is too large tofully review here.6 Generally speaking, scholars offer five main theoretical

6 For extensive reviews of the literature, see, e.g., Sayrs (1990), McMillan (1997), Reuveny (2000),Mansfield and Pollins (2003), and Schneider et al. (2003).


explanations. The classical liberal view argues that trade promotes peacebecause military conflict reduces the trade that countries value (e.g., Rose-crance, 1986; Polachek, 1992, 2000; Russett and Oneal, 2001). The bar-gaining approach considers trade a tool that facilitates bargaining oversome contested political issue, expecting that trade reduces the likelihoodof military conflict over the issue (e.g., Morrow, 1999; Gartzke et al., 2001;Crescenzi, 2003). A neo-Marxist view (e.g., dos Santos, 1970; Choucri andNorth, 1975; Ashley, 1980) and a neomercantilist derivative of realism (e.g.,Waltz, 1970, 1979; Gilpin, 1981; Borrus and Zysman, 1992; Barbieri, 2002)envision trade as a tool of influence and exploitation, which can cause con-flict if countries resent it or to peace if they comply. Classical realism expectsthat trade has no systematic effect on military conflict, the causes of whichare political and strategic instead (e.g., Buzan, 1984; Gilpin, 1987; Ripsmanand Blanchard, 1997; Keshk et al., 2004). Most existing empirical evidencesupports the expectation of the first two explanations, that bilateral tradereduces interstate militarized disputes; but evidence to the contrary alsoexists, which the classical liberal view and the bargaining approach cannotexplain.

Attempts to apply some, but not all, of the five explanations to account forthe effects of trade in different sectors on peace have produced conflictingconjectures. One argument expects that trade in sectors in which nationshave a decisive comparative advantage facilitate peace, because such tradebrings more economic benefits that would be lost to conflict. Anotherchannel to peace is said to involve imported goods (e.g., oil, machinery,minerals) that a country considers important, the loss of which hinderseconomic growth. A third channel argues that because imported consumergoods benefit consumers, they support leaders that promote peace withthe exporter.7 Other studies focus on strategic goods, including minerals,chemicals, steel, fuels, and high-tech goods. When strategic goods haveno readily available substitutes, their trade may generate conflict,8 becausecountries may feel threatened by their vulnerability and resort to conflict tomitigate the scarcity. Countries also may be tempted to use trade as a politicaltool to exploit vulnerabilities of and exercise influence on others, causingresentment and conflict.9 Yet trade in strategic goods may also generatefriendliness as one country supports another to protect its trade. A relatedexpectation is that trade in goods with fewer substitutes may generate peace

7 Gasiorowski and Polachek (1982), Arad et al. (1983), Domke (1988), Polachek (1980, 1992).8 See Hirschman (1945), Baldwin (1985), Gasiorowski (1986), Sayrs (1989), and Førland (1991).9 Reuveny (1999a, 1999b) shows how Israel uses its trade with the Palestinians to affect their

actions. For general discussions, see Gilpin (1984), Sen (1984), Borrus and Zysman (1992), andVogel (1992).


because countries tend to cooperate to secure the continuing flows of suchgoods.10

Some scholars debate the usefulness of the concept of strategic goods.Førland (1991: 197) quotes President Eisenhower, saying, “if our opponentneeds something badly, then that something is strategic.” Schelling (1958:500) notes that any good will be strategic to a country if it plans to consume itbut does not produce it. Baldwin (1985: 214–215) explains that the strategicgood can be “[any good] that is needed to pursue a given strategy and that isrelatively inefficient to produce at home.” He warns that it is hard to identifya priori the “strategic quality of a good” since it depends on the extent thatthe good has substitutes and on whether the good is important for a certaincase.

Studies on international competitiveness argue that trade in high-techgoods creates positive externalities for economic growth and the productionof weaponry. International competition over trade shares in these goods canlead to conflict. Moreover, states wishing to excel in economic growth, orstates suspecting that others may translate growth into military power inthe future, will maximize relative gains from trade in high-tech goods. Theattempt of one country to gain more from trade than their trade partnercan, in turn, lead to tension and ultimately conflict.11

Only a few empirical studies attempt to evaluate the effect of disaggre-gated trade on conflict, and none of them, with one exception, focuses onmilitary conflict and their samples include only a few dyads. Bennett et al.(1992) studied the effect of U.S.–Japan trade on political relations from 1948to 1978 and concluded that trade in steel, textiles, and electronics generatesconflict. Park et al. (1976) found that, as oil exports rise, oil-exporting coun-tries become hostile toward their trade partners. Gasiorowski and Polachek(1982) conclude that trade in agriculture and manufactured goods inducescooperation by the Warsaw Pact toward the United States. Polachek andMcDonald (1992: 277) study the effect of disaggregated trade in raw mate-rials and manufactured goods on conflict for 14 OECD countries in 1973.They report that trade reduces conflict but caution that these results are pre-liminary due to limited data coverage. Using a large-N sample, Dorussen(2006) finds that disaggregated trade reduces the likelihood of threats anduse of force but not in all sectors; the effects of nonmanufactured goods,chemicals, metal products, building supplies, and electronics are not statis-tically significant.

10 On the possibility of trade in strategic goods leading to cooperation due to compliance, seeHirschman (1945), Baldwin (1985), Reuveny and Kang (1998), and Reuveny (2001b).

11 This view often traces back to List (1856). See also Tyson (1992) and Mastanduno (1992).


In sum, the literature presents conflicting theoretical conjectures andmixed findings. In the following sections, we attempt to add clarity to thetrade and conflict literature through a model of the effects of trade, disaggre-gated across sectors and flow directions, on the initiation of military conflict.

THEORETICAL MODEL

To study the effects on conflict initiation of bilateral trade broken downacross goods and between flow directions, we need to model the attributesof traded goods such as sensitivities to conflict and price changes, and mar-ket forces such as import demand, export supply, and equilibrium. Statedbroadly, our model considers two countries that trade in various goodswith one another and maintain some political relations. We conceptualizeconflict initiation (or the lack thereof) as a change in bilateral relations.A change in trade affects conflict initiation because conflict is expectedto affect the price of a bilaterally traded good. When the price rises, theexporter’s profit and the importer’s cost rise. A country initiates conflict toincrease export profit or reduce import cost, ceteris paribus. In our model,trade provides the incentive or disincentive for one state to initiate conflicttoward another.

Our logic, attributing the effect of trade on conflict to considerations ofprofit and cost, resembles the classical liberal argument, whose microfoun-dation is best demonstrated in the decision-theoretic model of Polachek(1980, 1992, 2000). We employ the approach of Polachek, but our modeldiffers from his in two major ways. First, Polachek assumes that an exportercountry that initiates conflict against a target will be paid less for its exportto that country, and that an importer country that initiates conflict againsta target will have to pay more for its import from that target. We do notmake such assumptions; rather we explicitly allow import and export pricesto be determined by demand and supply in the market. Second, Polachek’smodel predicts that a rise in trade always brings peace, regardless of its com-position or pattern. Our model predicts that the effect of a rise in bilateraltrade on conflict initiation changes across traded sectors and between tradeflow directions, generating ex ante sector- and flow-specific hypotheses.

Our model includes two countries, A and B. We conceptualize politicalleaders of countries A and B as unitary rational actors whose utility functionsrise with national consumption and military conflict they initiate towardeach other. The conflict that country A sends toward country B is denotedby conflictAB. The unitary actor assumption is a simplification intendedto keep the math of the model tractable. The assumption that leaders’ utility


rises with conflict, ceteris paribus, may seem counterintuitive, but in factit is implicit or explicit in all conflict models. Although conflict is costly,leaders may conclude that its benefits outweigh its costs. Assuming leadersare rational actors, when country A is observed to initiate conflict towardcountry B, it must be the case that country A’s leader believes this actionwill increase his or her expected utility.12

As usual, consumption equals national output plus import of A fromB, importAB, minus the export of A to B, exportAB. When leaders chooseconflict, they take consumption, import, export, and national output asgiven, as in Polachek’s model. Country A’s leader chooses the level of conflictagainst its trade partner, B, to maximize utility, constrained by the need tomaintain some trade balance with country B. The mathematical details ofthe model are in the chapter appendix. Here we offer an accessible, verbaldiscussion of the key ideas.

As the conflict that A initiates against B rises in our model, the prices ofcountry A’s imports from country B and its exports to B change. These pricechanges are determined endogenously inside the model. This expectationdiffers from Polachek’s framework, which assumes that conflict always raisesthe price a country pays for its import and reduces the price it receives forits export. Although this is certainly a plausible possibility to consider, it isnot the only possibility. The reason, as we shall demonstrate, is the fact thatthe prices of traded goods are determined in international markets ratherthan in a deterministic manner through model assumptions.

Economic logic suggests that supply and demand determine the pricesand quantities of traded goods between countries A and B, and this logicis fully embedded in our model. Each traded good involves a supply sideand a demand side. The demand of an importer defines how much of agood to import at different prices, whereas the supply of an exporter defineshow much to export at different prices. As the price of a good rises, thedemand for this good falls and the supply of this good rises. The pointat which supply equals demand determines the equilibrium price of eachtraded good in the dyad.

More specifically, suppose in dyad A–B that A exports to B in somesector i (i.e., B imports from A), and A imports from B in another sector j(i.e., B exports to A). In equilibrium, country A’s demand for good j equalscountry B’s supply of good j to A. If B does not send enough of good j to A,country A’s demand for good j becomes larger than country B’s supply. In

12 Polachek (1980) employs the same leader’s utility function we employ here. For an exposition ofthe unitary actor model in international relations, and the expected utility approach to modelingthe determinants of military conflict, see Bueno de Mesquita (1981).


this case, the price of good j rises, prompting B to increase its supply to A.Sometimes, country B’s supply of good j is larger than country A’s demand.In this case, the price of good j declines due to the lack of demand in A, andB will reduce its supply. In equilibrium, the demand for bilateral import ofgood j equals the supply of bilateral export of that good. The same logic alsoapplies to the export flow in sector i from A to B.

How do trade prices change in response to conflict? This question standsat the center of our model. To answer it, we need to start with the supply anddemand of each traded good. The formal import demand model presentedin the appendix embeds the usual economic expectations that country A’simport demand for a product from B rises with country A’s income and fallswith the good’s price. To this setup, we add a political–economic feature.We assume that conflict reduces country A’s import demand, for which weprovide justifications shortly.

The exporter side also follows the mainstream approach. As is usual ineconomics, the formal model in the appendix assumes that export supplyrises with a good’s price and with the exporter’s income. The supply equationin the appendix includes a similar political–economic feature as in theimport demand equation; that is, country B’s export supply to A falls withdyadic conflict.

Several reasons explain our assumptions that both the import demandand the export supply fall with conflict. Importers and exporters may wantto punish hostile partners by reducing trade. The importer may seek toreduce dependence on a hostile supplier, which may stop selling, whereasthe exporter may seek to reduce dependence on a hostile importer, whichmay stop buying. Additionally, conflict often raises costs for both sides(e.g., higher risk insurance premium, delays in transportation, damages togoods), reducing trade. Furthermore, governments often restrict trade withfoes, seeking to influence their opponents by weakening their economicand military capabilities. Various empirical examples and statistical studiessupport this assumption.13

In equilibrium, bilateral import demand for good j equals its bilateralexport supply. We employ this condition to solve for the equilibrium priceof good j. The solution indicates that the price of good j from B to A riseswith country A’s income and falls with country B’s income, as is usually thecase in bilateral trade models. More important, the direction of the effectof conflict on the price of the imported good depends on the difference

13 For additional discussion of these ideas see, e.g., Pollins (1989), Morrow et al. (1998), Reuveny(2001a, 2001b), and Li and Sacko (2002).


between two parameters: �3 (the effect of conflict on country B’s exportsupply of good j to A) and �3 (the effect of conflict on country A’s importdemand for good j from B). Specifically, if �3 < �3, the import price,denoted as PM, falls with conflict. If �3 > �3, the import price rises withconflict. If �3 = �3, the import price does not change with conflict.

Figure 6.1 illustrates how the equilibrium import price, PM, for goodj is determined at the intersection of the bilateral import demand andexport supply curves. In each panel, the system is initially at equilibriumE1, where country A’s import demand (MD

A1) and country B’s export supply(MS

B1) intersect. The subscript 1 indicates the initial state of affairs, whereconflict has not occurred. Now, as A initiates conflict against B, curve MD

A1

(country A’s import demand) shifts inward to curve MDA2, because country

A demands less from country B. The curve MSB1 (country B’s export supply)

shifts inward to curve MSB2, because country B supplies less to country A.

The subscript 2 indicates the new state of affairs due to conflict. The shifts inthe supply and demand curves due to conflict in Figure 6.1 are determinedby �3 and �3, respectively, which, as noted, represent the strengths of theeffects of conflict on supply and demand. The market adjusts and reachesthe new equilibrium at E2.

A key point from Figure 6.1 is that the new equilibrium price at E2 can behigher than, lower than, or the same as the initial equilibrium price at E1.In panel A, the export supply curve (MS

B) shifts inward less than the importdemand curve (MD

A ) (i.e., �3 < �3), so the price of good j (PM) falls. Inpanel B, MS

B shifts inward more than MDA (that is, �3 > �3), so PM rises. In

panel C, the two inward shifts are equal in size, so PM does not change.What determines the size of the shift of the demand or supply curve?

If export supply is sensitive to conflict, the export’s effect on raising PMincreases with �3. This effect, however, declines if the export supply turnsmore elastic; that is, if the supply curve becomes flatter in Figure 6.1.Similarly, if import demand is sensitive to conflict, the import’s effect onreducing PM rises with �3. And this effect decreases if import demandbecomes more elastic or flatter.

So far, we have discussed country A’s import of good j from B (i.e.,country B’s export to A). We now turn to country A’s export of good ito B (i.e., country B’s import from A). Country B’s demand for countryA’s export and country A’s supply of country B’s import are determinedin manners similar to that discussed earlier for the opposing trade flow.The mathematical details in the chapter appendix demonstrate similar out-comes. In other words, the price changes of this trade flow depend on thesensitivity of the demand of good i to conflict (which we denote as �, whichis comparable in meaning to � in the opposing trade flow discussed earlier)


Figure 6.1. Disaggregated bilateral trade equilibrium.


and the sensitivity of the supply of good i to conflict (which we denote as � ,which is comparable to � in the opposing trade flow discussed earlier). Theshifts of these demand and supply curves for good i also depend on theirprice elasticities (or slopes).

Having identified how prices would respond to conflict, we analyze howthese anticipated price changes influence the effects of trade on conflict ini-tiation. In the chapter appendix, we present the details of the mathematicalderivations. Following are the expectations derived from the model:14

�(conflictAB) = KE

(� i

3 − �i3

) · �(exporti

AB

)i = 1, 2, . . . ; (6.1)

�(conflictAB) = KI

(�

j3 − �

j3

) · �(import j

AB

)j = 1, 2, . . . . (6.2)

According to Equations (6.1) and (6.2), where i and j indicate particulareconomic sectors, changes (denoted by the Greek letter �) in bilateral tradeflows affect the changes in conflict that country A sends toward country B.Specifically, Equation (6.1) indicates that a change in country A’s export ofgood i to country B,�(export i

AB), influences country A’s decision to changethe conflict it initiates against country B, �(conflictAB). The term KE is apositively signed term determined by the math of the model, as shown inthe appendix. Since KE is positive, the sign of the effect of country A’s exporton conflict depends on the relative size between �3 and � 3. Therefore, forany given good exported by A to B, if � 3 > �3, a rise in exportAB (countryA’s export to B) raises the conflict A initiates toward B, conflictAB. If � 3

< �3, a rise in exportAB reduces this conflict. If � 3 = �3, a rise in exportAB

does not affect conflictAB.Equation (6.2) indicates that a change in country A’s import of good j

from B, �(importjAB), affects country A’s decision to send conflict against

B, �(conflictAB). The term KI is a positively signed term determined bythe math of the model, as shown in the appendix. Since KI is positive,the sign of the effect of country A’s import on conflict depends on therelative size between �3 and �3. Therefore, for any given good importedby A from B, if �3 < �3, a rise in importAB (country A’s import from B)raises conflictAB. If �3 > �3, a rise in importAB reduces this conflict. If�3 = �3, a rise in importAB has no effect on conflictAB.

Implications of the Model and Hypotheses

What are the theoretical implications of our model predictions? And howdo we formulate testable hypotheses based on these model predictions? Weaddress these questions in this section. The theoretical model generates two

14 In Equations 6.1 and 6.2, i and j are superscripts denoting good type, not mathematicalexponents.


new implications for the trade and conflict literature. First, the effects ofcountry A’s export to B and country A’s import from B on the conflict Ainitiates against B can differ, because they depend on different coefficients(�3 and �3 for import, and � 3 and �3 for export). Second, the effectsof different traded goods on the conflict that A initiates against B canvary across sectors because �3, �3, � 3, and �3 vary across sectors. Theseimplications differ from the thrust of the literature and provide new insightsinto the effect of trade on conflict. Focusing on total bilateral trade, previousstudies ignore, implicitly or explicitly, the disparities in the effects of differenttrade flows in heterogeneous economic sectors.

What is the logic behind the model? The effects of trade on conflict in ourmodel are driven by the economic costs and benefits one anticipates from theexpected conflict-induced price movements. For example, exporters expectto benefit from conflict if a rise in conflict causes the export price to rise. Thisexport price increase occurs if the bilateral supply falls with conflict (raisingprice) more than the bilateral demand falls (lowering price), but when doesthe bilateral supply of a traded good fall more than its bilateral demand?This situation depends on the extent of substitution of the traded goodby alternative suppliers or buyers. Expecting higher costs due to conflict,exporters may pursue alternative buyers, and importers may seek othersuppliers. Both actions divert trade from foes, but imported and exportedgoods have varying degrees of substitution, ranging from full to none. Ifthe exporter has alternative buyers for its good, it supplies less to its foe andmore to others, which implies an inward shift of the supply curve, raisingthe price of the traded good. Expecting conflict, the importer may also buyfrom other suppliers, which implies an inward shift in the demand curve,reducing the price of the traded good. The trader that has relatively morealternatives is better positioned to benefit from conflict-induced marketchanges, and vice versa.

Figure 6.1 illustrates this logic graphically. Suppose country A has moredifficulty shifting its demand for some good j to other suppliers than countryB has in shifting its supply of good j to alternative buyers. Country A dependsmore on country B as a supplier of good j than B depends on A as an outletfor this good. Panel B of Figure 6.1 represents this scenario. As conflict rises,the price of good j rises, making it more expensive for A to import each unitof good j from B. Hence, A has an incentive to initiate less conflict against Bif its import of good j from B is large, ceteris paribus. This scenario is similarto Polachek’s prediction but it is not the only possible scenario in our model.

Now, suppose country A’s import demand of good j from B is moresensitive to conflict than country B’s export supply of good j to A. Country Ahas less difficulty in locating other suppliers of good j than country B has


in finding alternative buyers for good j. In this scenario, Panel A of Figure6.1 shows that country A’s import demand shifts inward more than countryB’s export supply. As conflict rises, the price of good j falls, making it lessexpensive for A to buy each unit of good j from B. If the amount of import islarge, country A has an economic incentive to initiate more conflict againstB, ceteris paribus. Polachek’s model as well as other previous research fail toconsider this scenario.

Another logical possibility implied by our model, which previous researchalso fails to anticipate, is that country A’s import demand and coun-try B’s export supply of good j could be equally sensitive to conflict. Inthis case, when conflict rises, country B’s export supply and country A’simport demand shift equally in absolute terms. The import price remainsunchanged. No additional economic incentive exists for A to initiate conflictagainst B even if country A’s import from B is large.

Similar to the preceding three scenarios with respect to country A’simport of good j from B, country A’s export of good i can also generatethree distinct effects on the conflict A sends toward B. Likewise, these effectsdepend on the sensitivities of country A’s export supply and country B’simport demand to the conflict.

Finally, it is worth noting that the substitution of traded goods, whichplays an important part in the theoretical model, can occur across countries,within sectors, or both. For example, the United States may substituteRussian wheat with Mexican wheat (across countries), or with Americancorn (within sector), or with Canadian corn (across countries and withinsector). The ease of substitution within sectors may vary. For example, itmay be easier to substitute some agricultural goods (e.g., wheat with rice)but harder to substitute some manufactured or energy goods (e.g., televisionwith cars, oil with nuclear energy).

So far, we have discussed the model’s implications for generic types ofgoods. To test these implications, we need to derive testable hypotheses forspecific traded goods (e.g., agricultural or manufactured goods). Predictingthe effects across specific exported and imported goods requires estimatesof coefficients �3, �3, � 3, and �3. The estimation of �3, �3, � 3, and �3

requires data on bilateral trade quantities and prices, which are not readilyavailable for all goods. Reuveny (2001b) is the only study that has pro-duced these coefficient estimates for five economic sectors: (1) agriculture/fishery, (2) energy, (3) chemical/mineral, (4) manufactured products, and(5) miscellaneous consumption products. Table 6.1 presents the five sec-tors employed by Reuveny, the range of goods in each sector, the impliedsigns of (�3 − �3) and (� 3 − �3) per sector, and the expected effects on

Tabl

e6.

1.Tr

ade

cate

gori

esan

dco

nflic

texp

ecta

tion

s Exp

ecte

def

fect

Exp

ecte

def

fect

Sect

orSe

ctor

con

ten

t�

3−

�3

ofim

port

�3−

� 3of

expo

rt

(1)

Agr

icu

ltu

re/fi

sher

yA

gric

ult

ure

prod

uct

s(f

ruit

s,ve

geta

bles

,cer

eals

,tr

ees,

suga

rs),

fish

,fish

ery

prod

uct

s>

0R

edu

cin

gco

nfl

ict

<0

Red

uci

ng

con

flic

t

(2)

En

ergy

Coa

l,lig

nit

e,ot

her

briq

uet

tes,

prod

uct

sof

cook

ing,

cru

depe

trol

eum

,nat

ura

lgas

,refi

ned

petr

oleu

man

dpr

odu

cts,

elec

tric

pow

er,g

as,

stea

m,r

adio

acti

vem

ater

ials

,ore

s

>0

Red

uci

ng

con

flic

t>

0In

crea

sin

gco

nfl

ict

(3)

Ch

emic

al/m

iner

alFe

rrou

sor

esan

dm

etal

s,n

onfe

rrou

sor

esan

dm

etal

s,n

onra

dioa

ctiv

eor

esan

dm

etal

s,n

onm

etal

licm

iner

als,

chem

ical

s

>0

Red

uci

ng

con

flic

t>

0In

crea

sin

gco

nfl

ict

(4)

Man

ufa

ctu

red

prod

uct

sM

etal

san

dm

iner

als

prod

uct

s,ag

ricu

ltu

rala

nd

indu

stri

alm

ach

iner

y,of

fice

and

data

proc

essi

ng

mac

hin

es,p

reci

sion

inst

rum

ents

,opt

ical

inst

rum

ents

,ele

ctri

calp

rodu

cts,

mot

orve

hic

les,

oth

ertr

ansp

ort

equ

ipm

ent

<0

Incr

easi

ng

con

flic

t>

0In

crea

sin

gco

nfl

ict

(5)

Mis

cella

neo

us

con

sum

ptio

npr

odu

cts

Mea

ts,m

eat

pres

erve

s,m

ilkpr

odu

cts,

food

,be

vera

ges,

toba

cco

prod

uct

s,te

xtile

s,cl

oth

ing

and

acce

ssor

ies,

leat

her

s,le

ath

eran

dsk

ingo

ods,

foot

wea

r,w

oode

npr

odu

cts,

furn

itu

re,

pape

rpr

odu

cts,

prin

tin

gpr

odu

cts,

rubb

ergo

ods,

plas

tic

prod

uct

s,go

ods

not

clas

sifi

ed

<0

Incr

easi

ng

con

flic

t<

0R

edu

cin

gco

nfl

ict

Not

e:�

3=

con

flic

t’sef

fect

onco

un

try

B’s

expo

rtsu

pply

ofgo

odjt

oA

;�3

=co

nfl

ict’s

effe

cton

cou

ntr

yA

’sim

port

dem

and

for

good

jfro

mB

;�3

=co

nfl

ict’s

effe

cton

cou

ntr

yB

’sim

port

dem

and

for

good

ifro

mA

;�3=

con

flic

t’sef

fect

onco

un

try

A’s

expo

rtsu

pply

ofgo

odit

oB

.

171


conflict initiation. Here we employ his estimates for these five specific sectorsand generate the following testable hypotheses.

H1: An increase in country A’s imports from country B in the agriculture/fishery, energy, or chemical/mineral sectors reduces the probabilityof A initiating military conflict against B.

H2: An increase in country A’s imports from country B in the manufac-tured or miscellaneous consumption products sectors increases theprobability of A initiating military conflict against B.

H3: An increase in country A’s exports to country B in the agriculture/fishery or miscellaneous consumption products sectors reduces theprobability of A initiating military conflict against B.

H4: An increase in country A’s exports to country B in the energy,chemical/mineral, or manufactured products sectors increases theprobability of A initiating military conflict against B.


This section first presents our statistical model for the empirical analysis andthen discusses several research design issues, followed by key results from theempirical analysis. As in the other chapters, the discussion is self-containedand does not require any statistical expertise. The technical details of thestatistical model, data sources, and measure construction are in the chapterappendix, which follows the same structure of presentation as this section.

Empirical Model

To test the hypotheses suggested by the theoretical model, we specify andestimate the following statistical model of dyadic conflict initiation:

midt = �0 + �1agriculture importt−1

+ �2agriculture exportt−1

+ �3chemical-mineral importt−1

+ �4chemical-mineral exportt−1

+ �5energy importt−1 + �6energy exportt−1

+ �7manufactured importt−1 + �8manufactured exportt−1

+ �9miscellaneous consumption importt−1

+ �10miscellaneous consumption exportt−1

+ �11gdp initiatort−1 + �12gdp targett−1 + �13contiguityt−1


+ �14distancet−1 + �15target democracyt−1

+ �16initiator democracyt−1 + �17regime dissimilarityt−1

+ �18power balancet−1 + �19initiator capability ratiot−1

+ �20alliancet−1 + �21minor powert−1 + εt. (6.3)

The model specification follows our theoretical expectations to includedyadic conflict, indicators for import and export flows in five hypothesizedeconomic sectors, as well as some control variables identified in the conflictliterature. We denote variables with small capital letters and their coefficientswith Greek notation. Each coefficient indicates the effect of the independentvariable on the dependent variable. The notation εt denotes the unexplainedrandom error in the model. The subscripts t and t − 1 indicate the timeperiod of the variable, where t represents the current time period andt − 1 the previous time period (a lagged variable). The use of lagged variablesis fully discussed in the appendix. To simplify the presentation, we refer tothe variables without their time subscripts.

The phenomenon we seek to explain in the theoretical model,conflictAB, is dyadic conflict. We measure it using the empirical indi-cator mid, which takes on the value of 1 if country A initiates a MID againstcountry B, and zero otherwise. As usual, the event of a MID includes thethreat of military force, the display of force, the use of force, or war. Oursample period of 1970–1997 includes 352 MIDs in our estimation sample.

Our theory indicates that the empirical model ought to include measuresof importAB and exportAB, that is, country A’s import from B in vari-ous economic sectors and country B’s export to A in these sectors. Basedon the hypotheses, we include five import variables and five export vari-ables, all measured in terms of the value of bilateral flow in real terms:agriculture import is the real dollar value of country A’s agriculturalproducts imported from B; chemical-mineral import is the real dollarvalue of imported chemical and mineral products; energy import is thereal dollar value of imported energy products; manufactured import isthe real dollar value of imported manufactured products; and miscella-neous consumption import is the real dollar value of the import of allremaining consumption goods that are not included in the aforementionedgroups. The variables agriculture export, chemical-mineral export,energy export, manufactured export, and miscellaneous consump-tion export are defined similarly for country A’s export to B.

The decision whether to initiate military conflict, of course, does notdepend only on trade variables. We need to control for possible confounding


factors to avoid spurious inferences. In recent years, the conflict literaturehas generally used a similar set of control variables, which we also use: gdpinitiator and gdp target are real GDPs of the conflict initiator and thetarget, respectively; contiguity indicates whether two states in a dyad arecontiguous on land or through up to 150 miles of water; initiator demo-cracy is the level of democracy of the MID initiator; target democracy isthe level of democracy of the target; regime dissimilarity is the absolutevalue of the difference between the democracy levels of the initiator andthe target; and power balance reflects the relative balance in materialscapabilities between the two states in a dyad. The initiator capabilityratio represents the relative importance of the initiator’s capabilities in thedyad, alliance indicates the presence or absence of dyadic military alliance,and, finally, minor power indicates whether both states in a dyad are onlyminor powers.


Several issues guide the design of our research in this chapter. The firstthree design issues concern the measurement of our dependent variable,MID initiation. First, one may argue whether it is appropriate to aggregatethe different types of MID to a single, 1–0 measure. This argument followsmany studies from the literature and enables the comparison of our resultswith previous findings, and we address this issue in an additional analysis.

Second, the theoretical concept of conflict in the model, conflictAB, is acontinuous variable, which leaders can raise or lower, but the mid variablein the statistical model is dichotomous, which seems to make for a relativelycrude measure. This issue, however, does not pose a key problem in our case,because our statistical technique and results inform us of the continuouslikelihood of conflict, not the dichotomous event of conflict or no conflict.

Third, the MID data may have measurement error from identifying whichside of a dyad started a dispute. Since scholars code the MID data retrospec-tively based on historical accounts, it may not be easy to know who startedthe dispute. In spite of this limitation, the mid variable does indicate whichside crosses the threshold into the realm of military force.

Fourth, because our theoretical framework and implied hypotheses antic-ipate the likelihood of conflict initiated by country A against country B, weuse a directed dyad framework. In this framework, the dyad US (A) =>

USSR (B) (the United States initiates conflict against USSR), for example,differs from the dyad USSR (A) =>US (B) (USSR initiates conflict againstthe United States). Our sample includes all those directed dyads for whichtrade and conflict data are available from 1970 to 1997. Fifth, conflict as


the dependent variable may affect some right-hand-side variables; that is,the relationship between some of the right-hand-side variables and conflictmay be reciprocal. For example, conflict may affect dyadic trade or the levelof democracy in each side of a dyad. Ignoring the issue can produce wrongresults, and we deal with this risk by applying the appropriate econometrictechnique.

Sixth, for statistical inferences to be valid, the error term εt needs to satisfysome assumptions. If the error term does not have certain desired properties,our ability to make correct inferences is compromised. We address theseissues with appropriate econometric techniques in the chapter appendix.

Seventh, the right-hand-side variables should not be excessively corre-lated to one another. This issue warrants our attention because sectoral tradeflows could be so highly correlated that they affect statistical inference. Wediscuss the diagnostic test results in appendix.

Finally, we assess the effects of sectoral trade flows on conflict initiationin several ways. We test whether the results support the hypotheses andwe compute the substantive effect of any key variable that has an effectstatistically different from zero. We then test whether the effect of trade inone sector on military conflict initiation differs from that of trade in anothersector, which allows us to evaluate whether a key assumption in previousstudies is appropriate. Finally, we provide information on how well ourmodel predicts the frequency of MIDs within our sample.

Empirical Findings

We present the results of our main analysis in Table 6.2, where the dependentvariable is coded as 1 whenever a country initiates a MID of any type ina year. Column 1 of Table 6.2 reports the results for hypothesis testingand column 2 presents the size of the effect of key variables. Starting withcolumn 1, the coefficient of agriculture import is negative, as expected,and statistically different from zero. The coefficient of agriculture exportis also negative, as expected, but the effect is statistically not different fromzero. Hence, a rise in a country’s agriculture/fishery import reduces thelikelihood of MID initiation toward the exporter, but the effect of a rise in acountry’s agriculture/fishery export is very weak. When a country dependson the import of agriculture/fishery products from its partner more thanits partner relies on the country as an export outlet, the former is less likelyto initiate a conflict against the latter. But a country’s greater dependenceon the other as an export outlet for agriculture/fishery products has littleeffect on its propensity to initiate military conflict.

Table 6.2. Effects of bilateral import and export in five sectors onMID initiation

RelativeAll MIDs risk

Agriculture-fishery import −0.00046 0.78[4.27]∗∗∗

Agriculture-fishery export −0.00006[0.71]

Energy import −0.00004 0.96[1.66]∗

Energy export 0.00004 1.07[2.67]∗∗∗

Chemical-mineral import −0.00004[0.76]

Chemical-mineral export −0.00008[1.19]

Manufactured import 0.00003 1.13[1.89]∗

Manufactured export 0.00003 1.16[2.27]∗∗

Miscellaneous consumption import −0.00005[0.85]

Miscellaneous consumption export 0.00002[0.56]

Initiator capability ratio 0.35938 1.44[2.34]∗∗

Power balance −0.01451[0.08]

Initiator democracy −0.00789 0.84[2.26]∗∗

Target democracy −0.00181[0.50]

Regime dissimilarity 0.00941 1.20[2.90]∗∗∗

Alliance −0.01330[0.18]

Log of distance −0.08084 0.71[4.93]∗∗∗

Contiguity 0.80909[6.15]∗∗∗

Minor power −0.32540[1.89]∗

Initiator GDP 0.06681 1.51[2.83]∗∗∗

Target GDP 0.09974 1.84[4.40]∗∗∗

Observations 213790Pseudo R2 0.31

Note: Z statistics in brackets. ∗ significant at 10%; ∗∗ significant at 5%;∗∗∗ significant at 1%. Constant, peace-year, and spline variables not reported.


The coefficient of energy import is negative, as expected, whereas thecoefficient of energy export is positive, as expected. Both coefficients arestatistically different from zero. A rise in a country’s energy import reducesthe likelihood of MID initiation toward the exporter, but a rise in energyexport increases the likelihood of MID initiation toward the importer. Whena country depends on the energy import from its partner more than itspartner relies on the country as an export outlet, the former is less likely toinitiate a conflict against the latter. Energy dependence matters. In contrast,when a country’s dependence on its partner as a destination for its energyexport is smaller than its partner’s dependence on its supply of energyexport, the country is more likely to initiate a conflict.

The coefficients of both manufactured import and manufacturedexport are positive, as expected, and significantly different from zero. Arise in a country’s import of manufactured products raises the likelihoodof MID initiation toward the exporter, and a rise in a country’s exportof manufactured products raises the likelihood of MID initiation towardthe importer. When a country depends on the import of manufacturedproducts from its partner less than its partner relies on it as an export outlet,the former is more likely to initiate a conflict against the latter. And whena country depends on its partner as an export outlet of its manufacturedproducts less than its partner relies on its supply, the former is more likelyto initiate a conflict against the latter.

The coefficients of chemical-mineral import/export and miscella-neous consumption import/export are not statistically different fromzero, which may be due to aggregations of very different products: chemi-cals and minerals in one case and various consumption goods in the other.The reader may recall that these particular aggregations are dictated bythe availability of data, for which we provide further discussion in theappendix.

At this point, one may ask whether those trade flows whose effects arestatistically different from zero are large or small from the substantive pointof view. Column 2 of Table 6.2 answers this question. As customary in theconflict literature, we address this question by computing the relative riskof MID. Relative risk is particularly useful when we examine the effect sizeof rare events like MID. It is the ratio of the predicted probability of MIDwhen the trade variable of interest increases by one standard deviation aboveits mean over the predicted probability of MID at the baseline. We definethe baseline scenario as one in which all continuous variables are at theirsample means and the dichotomous variables are set to the contiguous,minor power, no-alliance dyad.


When country A’s agriculture/fishery import from B rises by one standarddeviation ($187 million) above the mean, country A is 22% less likely toinitiate a MID against B, relative to the baseline level. When country A’senergy import from B rises by $439 million above the mean, it is 4% lesslikely to initiate a MID. When country A’s energy export to B rises by $439million above the mean, country A is 7% more likely to initiate a MID.When country A’s import or export of manufactured products with B risesby $1.5 billion, country A is 13% or 16% more likely to initiate a MID,respectively.

These effects vary across sectors and trade flows, as our theory suggests,but how do they compare in size with those of the nontrade variables?Column 2 also reports the results for the control variables. As in otherchapters, we delegate the discussion of the control variables to the appendix.Here, we provide a summary of the comparison in terms of the size of effect.The substantive effects of increases in energy import and export are smallerthan those for the nontrade variables; the effects of increases in agriculture/fishery import, manufactured imports, and manufactured exports are ofthe same order of magnitude as that for the initiator’s democracy, targetGDP, regime dissimilarity, and distance; and the effects of increases inthe initiator’s GDP and capability ratio are larger than those for the tradevariables. Overall, sectoral trade flows have important effects on conflictinitiation but not the most important effect.

Additional Analysis

We conduct three additional analyses to address several interesting ques-tions. First, previous studies of trade and conflict implicitly assume that thecoefficient for the total bilateral trade variable is the same for each of itscomponent trade flows. One way to test this assumption directly is to testwhether the statistically significant effects of the trade variables in Table 6.2are equal in size across sectors. The results of this analysis in Table 6.A1 inthe appendix show that the effects of increases of trade flows in the statis-tically significant sectors are not equal in size. Hence, aggregating exportsand imports in different sectors masks their differential effects.

Second, one may wonder if our model has good in-sample predictivepower. For each dyad in each year in the sample, we compute the predictedprobability of a dispute based on the model. If the probability is larger thana certain threshold (see the appendix), it suggests that the model expectsa military dispute for that dyad in the year; otherwise, the model predictspeace. By comparing the predictions with actual dyadic events, we compute


the predictive power of the model. As shown in Table 6.A2, the modelpredicts correctly the absence of MID initiation in 91% of the dyad-yearsand the presence of MID initiation in 81% of the cases. Thus, the modelperforms reasonably well in terms of in-sample prediction.

Third, as noted, our dependent variable aggregates the different types ofMIDs, ignoring their variations. We believe a focus on all MIDs also hasmerits. Leaders moving forces to the border or threatening to use force maybe driven by the same considerations as leaders actually using force. It isalso debatable whether the threats to use interballistic missiles or movingmany divisions to the border are less hostile than border skirmishes or evenclashes that cause killings. Nevertheless, to check robustness, we employ twoadditional dependent variables: one involves display of force, use of force,and war; the other includes use of force and war. The results obtained fromusing these two variables are in Table 6.A3 in the appendix. In the modelfor MIDs involving at least the display of force, the results for the tradevariables are identical to those in Model 1. In the model for MIDs involvingthe use of force and war, the results for the trade variables are consistentwith those in Model 1, except that the coefficient of chemical and mineralexport is now negative and significant, rejecting the respective part of H4,and the coefficient of miscellaneous consumption export is negative andsignificant, supporting the respective part of H3.

Comparisons to the Literature and Stylized Examples

Our empirical findings confirm several results reported in previous smaller-scale empirical studies that did not employ the MID measure or our empir-ical model specification. The finding that a rise in energy exports promotesconflict is consistent with the result of Park et al. (1976) for a sampleof oil-producing countries. Our result that a rise in agriculture/fisheryimports reduces conflict initiation agrees with the finding of Gasiorowskiand Polachek (1982) regarding the conflict behavior of the Warsaw Pacttoward the United States. Our finding that a rise in manufactured productsimport or export promotes conflict confirms the Bennett et al. (1992) studyof the U.S.–Japan dyad.

In many ways, the study of Dorussen (2006) is the closest to ours, becauseit also employs a large-N sample and investigates the effect of disaggregatedtrade on the likelihood of MID. In addition, several of our findings are con-sistent with the spirit of Dorussen’s results. For example, chemical/mineraltrade has a weak effect on conflict; some sectoral trade flows reduce theprobability of MID.


Our analysis, however, differs from Dorussen (2006) in other ways.Although his theory is generic like ours, it is not formalized. He arguesthat trade is more pacifying if traded goods have greater opportunity costs.If goods can be easily appropriated, the pacifying effect is weaker, or maybe reversed. Our theory works through changes in trade prices and substi-tution. Dorussen notes the effect of trade may change between export andimport, but he does not predict the effects of export and import in specificsectors. We predict the effects of import and export in different sectorson conflict and test these predictions. Empirically, Dorussen looks at bothMID involvement and onset in a sample of nondirected dyads but excludesminor disputes below the use of force, whereas we study the MID initiationsmeasured in three ways (all MIDs, at least display of force, and at least use offorce) in a sample of directed dyads. Finally, whereas Dorussen finds tradeeither reduces or has no effect on conflict, depending on sector, we find thattrade in some sectors can also increase the likelihood of conflict.

Given the complexity of our model, our exposition could benefit from afew real-world examples, but we must first offer a caveat. Statistical studiessuch as ours typically do not discuss real-world examples, and for goodreasons. Our model predicts a tendency toward conflict or peace due to achange in some trade variable, holding all the other variables in the modelconstant. In the real world, one observes the net effect of all the relevantvariables held at the levels they happen to be, not the partial effects obtainedfrom regression. We cannot, and do not, argue that the presence or absenceof conflict in these examples is only driven by trade. Rather, we employ ourmodel to illustrate the role of sectoral trade in conflict, ceteris paribus.15

Limitations notwithstanding, we may utilize the implications of ourmodel to better understand some stylized examples. Consider first the Iran–U.S. dyad in the period since 1979, which included several U.S.–Iran MIDs.Based on our model, we speculate Iran may intensify conflict with theUnited States to raise the price of its oil exports. In other words, our modelsuggests that Iran, which is likely still selling oil to the United States directlyor indirectly, benefits economically from hostility.

A second example involves Bulgaria, Romania, and Hungary in the 1930s.At that time, their trade flows became highly concentrated on Germany.They imported primarily manufactured goods and coal from Germanyand exported agricultural and light manufactured goods to Germany.16

15 We also need to be careful when using nonlinear models such as probit since employing differentcontrol groups may affect the inferred sizes of effects.

16 On Nazi Germany’s trade with these countries (and others) see, e.g., Hillmann (1940) and Leitz(2004).


Since they were much less able than Germany to substitute these flows, theyrealized that not cooperating with Germany would damage their economies(Hirschman, 1945; Arad et al., 1983). Our model suggests that to pacify thesecountries, Germany could increase coal and agricultural exports and reducemanufactured exports and light manufactured imports.

Finally, consider the U.S.–Japan dyad in the period leading up to Decem-ber 1941. The United States reduced energy export to Japan, hoping tocurb its territorial expansion in Asia, but the policy backfired. Historianssuch as Kennedy (1987) and Keylor (2001) argue that the fall in Japan’senergy import from the United States resulted in economic losses anda growing sense of vulnerability, which contributed to Japan’s hostilitytoward the United States – an explanation consistent with our model pre-diction.


Studies of the effect of trade on conflict typically invoke the notion of totalbilateral trade, implicitly or explicitly assuming the effect does not varyacross economic sectors and between trade flow directions. We challengethese premises. Our theory considers two countries that trade with oneanother and maintain some bilateral relations. The effects of changes inbilateral trade on the decision to initiate conflict vary across economicsectors and between trade flow directions. We employ our theory to identifyex ante the effects on conflict initiation of increases in export and import inthe sectors of agriculture/fishery, energy, manufactured products, chemical/mineral, and miscellaneous consumption products.

In the empirical analysis, we find significant variations in the effects oftrade on MID initiation across economic sectors and between trade flowdirections. As predicted, rises in agriculture/fishery imports and energy im-ports reduce the probability of MID initiation, and rises in energy exports,manufacturing imports, and manufacturing exports increase this probabil-ity. Yet changes in agriculture/fishery export, chemical/mineral trade, andmiscellaneous consumption products trade generally do not affect MIDinitiation. One logical explanation of these insignificant effects is that thesesectors and flows are not important enough to the leader to influence herdecision. But one may also attribute these insignificant results to two empir-ical problems: (1) the limited data for the measurement of the sensitivityparameters of sectoral import and export to conflict and (2) inappropri-ate aggregation of, say, chemicals and minerals into one category and verydifferent consumption goods into one category. Improvements in solving


these two problems should lead to more accurate empirical testing of ourtheory in future research.

Like all regression results, these findings hold when other things do notchange. However, the effects of most of our control variables are also foundto be significant, suggesting that they also affect the likelihood of MIDinitiation. Moreover, although we have estimated the ceteris paribus effectsof changes in all these variables on conflict initiation, only the net effectis observed in reality. Thus, a dyad could be in peace even if some of itssectoral trade flows promote MID, provided that the pacifying forces com-bined outweigh the antagonizing effects together.

How does our model fit with the two primary formal models offered in theliterature to explain the effect of trade on conflict, that is, Polachek’s liberalmodel and Gartzke et al.’s signaling argument? As discussed throughout thechapter, our model subsumes and extends the one of Polachek, predictingthat a rise in trade may promote either peace or conflict and the effectmay vary across goods and imports versus exports, depending on certainparameters. Consider next the signaling argument. In this theory, militaryconflict results from a bargaining failure over some contested issue. Actingunder incomplete information, trading nations are less likely to fight eachother only because they can credibly signal resolve during bargaining byresorting to trade sanctions or trade dissociation (Morrow, 1999; Gartzkeet al., 2001). Within this logic, the classical liberal argument formalized byPolachek does not hold because the size of the benefit from bilateral trade issubsumed into the initiator’s demand and the target’s concession withoutreducing uncertainty–the main cause of bargaining failure and militaryconflict.

One might argue that the same criticism against the classical liberalargument could be leveled against our model. That is, leaders factor expectedeconomic costs into their reservation values when making demands of theiropponents, and they make more concessions to avoid military conflict whenthe expected economic costs are high. We argue that, in fact, our researchposes some important challenges for the signaling argument.

In the signaling argument, the benefit of each actor from economicexchange is modeled by one opportunity cost parameter, which is commonknowledge to both sides (e.g., Gartzke et al., 2001). The situation in ourframework is more complicated because we allow n trade markets for thereare n goods or sectors. As we have shown, each bilateral trade market doesnot necessarily suffer a loss due to conflict; some conflict may actually inducea rise in profit, a condition that is assumed away in the bargaining model.Now, the leader needs to aggregate the costs and benefits across the sectors


in some way to figure out the welfare implications of a conflict decision.To do so, one needs to know who supplies what to whom, who demandswhat from whom, and who are the alternative suppliers and buyers and atwhat prices. The amount of information required is daunting, and much ofthe information is likely to be trade secrets. The leader may have difficultygetting access to all this information; even if she does, she may not be willingto share this financially valuable private information with the other side.

Hence, our analysis of disaggregated bilateral trade creates four majorproblems for the signaling argument. First, the size of the opportunity costis more difficult to pin down in multiple trading markets. Second, the costsand benefits in particular trading markets are not common knowledge suchthat the manipulation of a trading relationship may misinform as to thevalue of a particular trading relationship. Third, if one finds gains in tradefrom conflict that are unknown to the other side, this factor contributesto asymmetric information and uncertainty. Finally, the empirical findingthat some sectoral exports and imports lead to conflict initiation challengesthe theoretical prediction of the signaling argument that, if trade affectsconflict at all, it should reduce the likelihood of military conflict.

Taking a broader perspective, this research advances our theoreticalunderstanding of the effect of trade on conflict initiation. To illustratethis point, let us further compare our theory with the five key theories inthe literature in terms of their expected effect of trade on conflict, policyobjective, policy tool, causal logic, and treatment of markets.

Assuming states maximize material welfare, liberalism argues that lead-ers pursue peace with trade partners because conflict reduces valuable tradegains. The causal logic stresses economic incentives but oversimplifies mar-ket forces, assuming conflict is always costly or conflict lowers export pricefor an exporter and raises import price for an importer. The bargainingapproach assumes leaders manipulate trade to facilitate negotiation over adisputed issue or influence its outcome. Cutting trade ties functions either asa credible signal to demonstrate resolve, enabling states to avoid bargainingfailure and conflict, or as a tool to influence the bargaining outcome, result-ing in less high-level conflict when actors face high exit costs. This approachalso stresses economic ties, but reduces the economics of the marketplaceto one or two parameters that denote the importance of trade to bothsides.

The neo-Marxist and neomercantilist realist arguments postulate thatstates use trade as a policy tool to affect the outcome of a contested issueor to expand their economic gains and political power. The outcome canbe conflict or peace, depending on whether the other side rebels against, or


complies with, the external demand. Economic incentives also play a rolehere, but market forces are not fully modeled – only loosely representedby the differential importance of trade to both sides. The classical realistargument assumes that states maximize power and security, using conflictas a policy tool. In this view, trade should have no systematic effect onconflict, because political relations trump economic relations.

In our theory, states use conflict (or peace) as a policy instrument tomaximize national welfare, which depends on wealth and security. Bilateralimport demand and export supply in different sectors have different sensi-tivities to conflict, reflecting the underlying abilities of states to substitutethese trade flows with others. Markets play an essential role in this logic,informing leaders about the costs and benefits of conflict via the expectedchanges in the bilateral trade prices. The costs and benefits, in turn, motivateor deter conflict initiation.

In general, our theory subsumes the liberal logic as one special case andoffers a distinct explanation alternative to the bargaining/signaling, neo-Marxist, neomercantilist realist, and classical realist arguments. Our theoryalso specifies ex ante when and how export and import flows in specificeconomic sectors influence the likelihood of military conflict initiation, aquestion that has not been systematically addressed before.

We believe the value of our analysis is not merely academic. Our resultsinform policymakers about how to manage bilateral trade flows to improveinterstate relations. Previous studies have focused on the effect of totaltrade on conflict. To a certain extent, finding the net effect of total trade(i.e., aggregated across goods) has limited value for public policy and couldbe misleading, because countries have heterogeneous trade structures, andincreasing trade in some goods could harm political relations. Public policythat seeks peace should promote bilateral trade in peace-promoting goods.Dispute resolution mechanisms should focus on tensions associated withconflict-promoting types of trade. Countries with a history of bilateralconflict that are interested in peace may expand trade in some goods, butnot in others.

SUMMARY AND OUTLOOK

The effect of bilateral trade on interstate military conflict is the subject ofa large and growing body of literature in international relations. Almost alltheoretical and empirical studies focus on total bilateral trade, implicitly orexplicitly assuming that the effect of trade on interstate conflict is constant


across imports and exports in different economic sectors. We challengethe premise of most trade–conflict studies in international relations thatthe effect of trade on interstate conflict does not vary across imports andexports in different economic sectors. We analyze theoretically and empir-ically the effects of bilateral imports and exports in different economicsectors on conflict initiation. Our theory encompasses the liberal explana-tion as a special case and is a distinct alternative to the bargaining/signaling,neo-Marxist, neomercantilist realist, and classical realist arguments. It alsospecifies the different effects of export and import flows, an issue neversystematically addressed before.

Our theoretical model generates ex ante predictions of the effects ofchanges in imports and exports in the agriculture/fishery, energy, chemical/mineral, manufactured products, and miscellaneous consumption prod-ucts sectors on the likelihood of MID initiation. The empirical test employsa sample of directed dyads from 1970 to 1997. We find strong empiricalevidence supporting our expectations for agriculture/fishery, energy, andmanufactured products sectors, and little evidence supporting the expecta-tions for chemical/mineral and miscellaneous consumption products. Thesefindings lead to several conclusions. Rises in agriculture/fishery and energyimports discourage MID initiation. Rises in energy export and trade in man-ufactured products induce MID initiation. Changes in chemical/mineraltrade, agriculture/fishery export, and miscellaneous consumption productstrade generally do not affect MID initiation. These results hold for all MIDsand for MIDs involving at least display of force, and largely so for MIDs in-volving at least the use of force. The significant effects vary across sectorsand trade flows, and their sizes are relatively substantial. The model pre-dicts correctly the large majority of peace and MID events in the sampleand sheds some light on several real-world cases.

In this chapter, we have investigated the effect trade flows disaggregatedacross goods on the propensity of a state to initiate military conflict againsta country with which it also trades. In our analysis of the results and theirimplications, we listed a number of examples referring to a specific conflicttaking place, of course, on Earth. The previous chapters also refer, again ofcourse, to interactions taking place on this very planet. We add the words“of course” naturally and almost in passing; but if so, the reader may argue,why have we not thus far introduced elements pertaining to the physicalenvironment within which these interactions take place?

We think this is a good question that needs to be addressed more oftenin the fields of international relations and international political economy.


In the last part of our book, we take an initial step toward bringing theenvironment into our analysis of the complex transformation taking placein recent decades. In particular, we ask about the effect of the very forcesstudied in the first and second parts of the book – democracy, conflict,and economic openness – on the physical environment. We think thesequestions are important because changes in the physical environment candrastically alter the quality of life on Earth.

APPENDIX

THEORETICAL MODEL

The quantity of trade in good i that country A exports to B is Xi and itsprice is PXi. The volume of trade in good j that country B exports to A isMj and its price is PMj. The number of goods country A exports to B isNX. The number of goods country A imports from B is NM. The utility ofcountry A’s leader is denoted by U (EC, C): EC is economic consumptionof country A; C is the conflict country A initiates against B, defined as apositive number. The signs of the partial derivatives of utility are assumed tobe as in Polachek: UEC > 0 (the marginal utility of economic consumption ispositive, or utility rises with consumption), UEC EC < 0 (the marginal utilityof consumption falls with more consumption), UC > 0 (the marginal utilityof conflict is positive), and UCC < 0 (the marginal utility of conflict fallswith more conflict).

The partial derivative of the price of good i (A exports to B) with respectto conflict is PXi

C (i = 1, 2, . . . , NX). The partial derivative of the priceof good j (A imports from B) with respect to conflict is PM

jC (j = 1,

2, . . . , NM). Their second-order derivatives are PXiCC and PM

jCC. Polachek

assumes certain directions for the effects of conflict on prices (PXiC < 0,

PXiCC > 0, PM

jC > 0, and PM

jCC > 0), but in our model prices and their

derivatives are set endogenously in the marketplace, which is one of our keyinnovations.

Country A chooses the level of conflict (C) against B, taking consumption,import, and export as given, to maximize utility subject to the need tomaintain a zero trade balance. It should be noted using a nonzero tradebalance does not change any of the theoretical results we derive, becausethe math we present next is not sensitive to this particular modeling detail.Turning to the maximization itself, the problem that a decision maker is


assumed to solve is formally stated as follows:

max U(EC, C) by choosing C, subject to 0 =NX∑i=1

Xi · PXi (C)

−NM∑j=1

M j · PM j (C).

(6.1A)

The solution of this problem involves maximizing the followingLagrangian with respect to conflict (C) and the Lagrange multiplier (�).The first-order condition for C gives

UC (EC, C) + �

NX∑i=1

Xi · PXiC (C) −

NM∑j=1

M j · PM jC (C)

= 0. (6.2A)

The effects of changes in country A’s export of good i and import ofgood j on the conflict that A sends toward B are given by computing thecomparative statics of Equation (6.2A) with respect to Xi and Mj:

∂C

∂Xi= −� · PXi

C

UCC + � · ∑NXi=1 Xi · PXi

CC − � · ∑NMj=1 Mi · PMi

CC

i = 1, 2, . . . , NX; (6.3A)

∂C

∂M j= � · PM j

C

UCC + � · ∑NXi=1 Xi · PXi

CC − � · ∑NMj=1 M j · PM j

CC

j = 1, 2, . . . , NM. (6.4A)

Economic logic suggests that supply and demand determine the prices andquantities of traded goods i (i = 1, 2, 3, . . . , NX) and j (j = 1, 2, 3, . . . , NM)between A and B. Each good involves a supply side and a demand side. Thedemand of an importer defines how much of a good to import at differentprices, whereas the supply of an exporter defines how much to export atdifferent prices. As the price of a good rises, the demand for this good fallsand the supply of this good rises. The equilibrium condition defines theprice of each traded good in the dyad.

More specifically, suppose in dyad A–B that country A exports to B insector i (i.e., country B imports from A), and country A imports from B insector j (i.e., country B exports to A). In equilibrium, country A’s demand


for good j equals country B’s supply of good j to A. If B does not sendenough of good j to A, country A’s demand for good j becomes larger thancountry B’s supply. In this case, the price of good j rises, prompting B toincrease the supply to A. Sometimes, country B’s supply of good j is largerthan country A’s demand. In this case, the price of good j declines due to thelack of demand in country A, and country B will reduce the supply of thatgood. In equilibrium, the demand for bilateral import of good j equals thesupply of bilateral export of that good. The same logic also applies to theexport flow in sector i from country A to country B.

How do trade prices change in response to conflict? Demand and supplyare often modeled as linear functions.17 We model country A’s importdemand for good j from B (MD

A ) by

MDA = �0 − �1 P M + �2YA − �3C, (6.5A)

where the superscript j denoting the good is dropped hereafter to simplifynotation. We define all the � parameters in Equation (6.5A) to be positive.YA is country A’s income, and PM is the price country A pays for importsfrom country B. Equation (6.5A) embeds the usual economic expectationsthat country A’s import demand for products from country B rise withA’s income and fall with the good’s price. The effect of conflict on importdemand of country A in Equation (6.5A) is negative (C is defined as positive),a point to which we return shortly.

Country B’s export supply of good j to country A is modeled as follows:

MSB = �1 P M + �2YB − �3C, (6.6A)

where all the � parameters in Equation (6.6A) are positive, YB is the exporterB’s income, and PM and C are defined as in Equation (6.5A). As usual ineconomics, export supply rises with the good’s price and with the exporter’sincome. Equation (6.6A) also indicates that the supply of goods by B to Afalls with conflict (C).

Several reasons exist why import and export fall with conflict in Equa-tions (6.5A) and (6.6A). Importers and exporters may want to punish hostilepartners by reducing trade. The importer may seek to reduce dependenceon a hostile supplier, which may stop selling, whereas the exporter may seekto reduce dependence on a hostile importer, which may stop buying. Addi-tionally, conflict often raises costs to both sides (e.g., higher risk insurance

17 See, e.g., Bond (1985), and Nicholson (2005) for similar formulations that do not includeconflict.


premium, delays in transportation, damages to goods), reducing trade. Fur-thermore, governments often restrict trade with foes, seeking to influencetheir opponents and weaken their economic and military capabilities. Var-ious empirical examples and statistical studies support this assumption.18

In equilibrium, bilateral import demand for good j equals its bilateralexport supply. The equation MD

A = MSB is solved for PM, the equilibrium

price of good j, as follows:

P M = �0

�1 + �1+ �2

�1 + �1YA − �2

�1 + �1YB + �3 − �3

�1 + �1C. (6.7A)

As usual in bilateral trade models, Equation (6.7A) indicates that the priceof good j from country B to country A rises with country A’s income and fallswith country B’s income. The denominator of the expression multiplyingC in Equation (6.7A) is positive. Hence, the effect of C on PM depends on�3 (effect of conflict on country A’s import demand for good j from B)and �3 (effect of conflict on country B’s export supply of good j to A). If�3 < �3, PM falls with conflict (the coefficient of C is negative). If �3 > �3,PM rises with conflict. If �3 = �3, PM does not change with conflict. Figure6.1 illustrates the solution for PM in Equation (6.7A), as discussed in themain text.

Equations (6.5A)–(6.7A) model the determinants of trade prices, leadingto the question, what determines the effect of conflict on trade prices?Suppose A raises hostility toward B by one unit. The effect on PM is givenby the expression in front of C in Equation (6.7A), (�3 − �3)/(�1 + �1). Thechange in PM when C rises depends on the conflict sensitivities of demand(�3) and supply (�3) and on the price slopes of the demand and supplycurves (�1 and �1, respectively). The shift in PM comes from two sources:country B’s export supply [�3/(�1 + �1)] and country A’s import demand[−�3/(�1 + �1)]. When conflict rises, the export’s effect on raising PMincreases with �3, indicating that export supply is sensitive to conflict. Theeffect declines with �1, indicating that export supply is elastic (the supplycurve becomes flatter in Figure 6.1). Similarly, when conflict rises, the effectof import on reducing PM rises with �3, suggesting import demand issensitive to conflict, and falls with �1, suggesting import supply is elastic.

So far, we have discussed country A’s import of good j from B (i.e.,country B’s export to A). We now turn to country A’s export of good ito B (i.e., country B’s import from A). Country B’s demand for countryA’s export and country A’s supply of country B’s import are determined

18 For additional discussion of these ideas, see, e.g., Pollins (1989), Morrow et al. (1998), Reuveny(2001a, 2001b), and Li and Sacko (2002).


in manners similar to those defined by Equations (6.5A) and (6.6A). Thevolume of this flow is denoted as X. We use � for the coefficients in countryB’s import demand in Equation (6.8A) (instead of � for country A’s demandin Equation (6.5A)), and � for the coefficients in country A’s export supplyin Equation (6.9A) (instead of � for country B’s supply in (6.6A)). Thus,country B’s import demand for good i from A is specified as

X DB = �0 − �1 P M + �2YB − �3C (6.8A)

and country A’s export supply of good i to B is specified as

X SA = �1 P M + �2YA − �3C. (6.9A)

As before, all the � and � coefficients are defined to be positive.The condition X D

B = X SA determines the equilibrium price PX for

good i:

PX = � 0

�1 + �1+ �2

�1 + �1YB − �2

�1 + �1YA + �3 − �3

�1 + �1C. (6.10A)

Next we turn to the effects of trade on conflict in Equations (6.3A) and(6.4A). These effects depend on the first and second derivatives of the tradeprices with respect to conflict (C) in Equations (6.7A) and (6.10A):

PXC = �3 − �3

�1 + �1; PXCC = 0; PMC = �3 − �3

�1 + �1; PMCC = 0.

(6.11A)

Substituting these derivatives from Equation (6.11A) into Equations(6.3A) and (6.4A), respectively, and reinstating the goods’ notation in theappropriate places, we get the following solutions:

∂C

∂ Xi=

−� · � i3 −�i

3

� i1 +�i

1

UCCi = 1, 2, . . . , NX; (6.12A)

∂C

∂ M j=

� · �j3 −�

j3

�j1 +�

j1

UCCj = 1, 2, . . . , NM. (6.13A)

Expressions (6.12A) and (6.13A) specify, respectively, the effects ofchanges in country A’s export of good i (i = 1, 2, 3, . . . , NX) and countryA’s import of good j (j = 1, 2, 3, . . . , NM) on its decision to initiate conflictagainst B. The signs of these effects depend on the relative sizes of the �3


and � 3 coefficients for country A’s export to B in Equation (6.12A) and onthe relative sizes of the �3 and �3 coefficients for country A’s import fromB in Equation (6.13A).

Recall that Ucc < 0 in Equations (6.12A) and (6.13A). In Equation(6.12A), if � 3 > �3, a rise in country A’s export to B raises the conflictA initiates toward B. If � 3 < �3, a rise in country A’s export to B reduces thisconflict. If � 3 = �3, a rise in country A’s export to B does not affect countryA’s conflict against B. In Equation (6.13A), if �3 < �3, a rise in country A’simport from B raises the conflict A initiates against B. If �3 > �3, a rise incountry A’s import from B reduces this conflict. If �3 = �3, a rise in countryA’s import from B has no effect on country A’s conflict against B.

Finally, we simplify the presentation of Equations (6.12A) and (6.13A)to obtain Equations (6.1) and (6.2) in the main text. We define KE and KI,which we have also used in the main text of this chapter, as the followingsubexpressions in Equations (6.12A) and (6.13A):

K iE =

−� · � i3 −�i

3

� i1 +�i

1

UCCi = 1, 2, . . . , NX; (6.14A)

K jl = � · (

�j1 + �

j1

)UCC

j = 1, 2, . . . , NM. (6.15A)

With these expression thus defined, we can rewrite ∂C/∂X and ∂C/∂Mfor the partial derivatives in Equations (6.12A) and (6.13A) as their mathe-matical approximation by replacing the partial differential ∂ with the change�, giving the ratios �C/�X and �C/�M instead of the derivatives on theleft-hand side of Equations (6.12A) and (6.13A), which together with Equa-tions (6.14A) and (6.15A) can be rewritten to give Equations (6.1) and (6.2)in the chapter’s main text:

�(conflictAB ) = K E

(� i

3 − �i3

) · �(exportiAB ) i = 1, 2, . . . ; (6.1)

�(conflictAB ) = K I

(�

j3 − �

j3

) · �(import jAB ) j = 1, 2, . . . . (6.2)

IMPLICATIONS OF THE MODEL AND HYPOTHESES

We have seen in the text that the important issue is whether countriesengaged in conflict can substitute the traded goods. This event would dependon the sensitivity of the particular traded good to conflict. Formally, ifcountry A’s import demand over good j from B is less sensitive to conflict


than country B’s export supply of good j to A (�3 < �3), Equation (6.13A)predicts that a rise in country A’s import from B in sector j lowers theconflict A initiates against B. In this case, as conflict rises, country A’simport demand shifts inward less than country B’s export supply, but onlyif A has more difficulty shifting its demand for good j to other suppliersthan B has in shifting its supply of good j to alternative buyers. In otherwords, A depends more on B as a supplier of good j than B depends on A asan outlet for good j. As conflict rises, the equilibrium price of good j, PM,rises, as shown in panel B of Figure 6.1 and in Equation (6.7A), making itmore expensive for A to buy each unit of good j from B. Hence, A has anincentive to initiate less conflict against B if its import of good j from B rises,ceteris paribus. This scenario is similar to the scenario in Polachek’s model(recall that he assumes import price rises with conflict). However, it is notthe only possible scenario in our model.

If country A’s import demand of good j from B is more sensitive toconflict than country B’s export supply of good j to A (�3 > �3), Equation(6.13A) predicts that a rise in country A’s import from B in sector j raisesthe conflict A initiates against B. In this case, country A’s import demandshifts inward more than country B’s export supply, which occurs only if Ahas less difficulty in locating other suppliers of good j than B has in findingalternative buyers for good j. As conflict rises, the equilibrium price of goodj, PM, falls, as in panel A of Figure 6.1 and in Equation (6.7A), making itless expensive for A to buy each unit of good j from B. Hence, A has aneconomic incentive to initiate more conflict against B, ceteris paribus.

Yet another possibility implied by our model occurs when country A’simport demand and country B’s export supply of good j are equally sensitiveto conflict (�3 = �3). In this case, a rise in country A’s import from B insector j does not affect the conflict initiated by A against B. When conflictrises and country B’s export supply and country A’s import demand shiftequally in absolute terms, PM does not change. As a result, country A’sincentive to initiate conflict against B does not change when its import ofgood j from B rises.

The reasoning behind the effect of country A’s export on the conflictA initiates toward B is similar. In Equation (6.12A), if country A’s exportsupply of good i to B is less sensitive to conflict than country B’s importdemand for good i from A (� 3 < �3), a rise in country A’s export to B insector i reduces the conflict A initiates against B. In this case, with conflict,country A’s supply of good i to B shifts inward less than country B’s demandfor good i from A. Country A depends more on B as a buyer of good i than


country B depends on A as a supplier. As a result, the price of good i, PX,falls, as shown in Equation (6.10A). When PX falls with conflict, A gets lessfor each unit of good i exported to B. Hence, A has an incentive to reduceconflict if its export of good i to B rises.19

If country A’s export supply of good i to B is more sensitive to conflictthan country B’s import demand for good i from A (� 3 > �3), a rise incountry A’s export to B in sector i raises the conflict that A sends against B.In this case, as conflict rises, country A’s supply of good i to B shifts inwardmore than country B’s demand for good i from A. Country A depends lesson B as an outlet than B depends on A as a supplier. The price of good i, PX,rises, as shown in Equation (6.10A). Country A gets more for each unit ofgood i exported to B. Hence, A has an incentive to raise conflict if its exportof good i to B rises, ceteris paribus.

If country A’s export supply and country B’s import demand of good iare equally sensitive to conflict (� 3 = �3), a rise in country A’s export to B ingood i does not affect the conflict A initiates against B. In this case, countryA’s export supply and country B’s import demand shift inward equally inabsolute terms in response to a rise in conflict, so the price of good i doesnot change with conflict. As a result, there is no incentive for A to increaseor decrease the conflict it initiates toward B.

Turning to our hypotheses, to generate specific hypotheses on the effectsof different types of traded goods on conflict, one has to know the values of�3, �3, � 3, and �3. The reader may recall that these four parameters are thecoefficients of conflict in the bilateral import demand and export supplyequations of each good. As with all demand and supply equations, theseparameters are not readily observable and have to be estimated empirically.In principle, we could estimate the parameters ourselves, using data onbilateral trade prices and quantities for dyadic trade flows broken downalong economic sectors, but these data are rarely available. Alternatively, wecould use existing estimates of these parameters in the literature. We choosethe latter approach, which offers an initial test of our theoretical model.

Only a few trade studies estimate bilateral import demand and exportsupply using bilateral trade prices and quantities. They use data (Italianer,1986) collected for the European Commission, but they do not includeconflict. To our knowledge, only Reuveny (2001b) includes conflict in esti-mating sectoral bilateral import demand and export supply equations. Heemploys Italianer’s data for five sectors: (1) agriculture/fishery, (2) energy,

19 This scenario is implied by Polachek’s model, but again it is not the only possible scenario.


(3) chemical/mineral, (4) manufactured products, and (5) miscellaneousconsumption products. For each good traded in any dyad, he estimates asystem of six equations with six endogenous variables: export of A to B,import of B from A, export of B to A, import of A from B, conflict from Ato B, and conflict from B to A. The coefficients of conflict in the demandand supply equations provide our �3, �3, � 3, and �3. We use the averageestimates across dyads to compute (�3 − �3) and (� 3 − �3) for each of Ital-ianer’s five sectors, which Equations (6.12A) and (6.13A) employ to predicteffects of the sectoral exports and imports, respectively, on conflict.20

Table 6.1 in the main text presents the five sectors, the signs of (�3 − �3)and (� 3 − �3) per sector, and the expected effects of country A’s importfrom B and country A’s export to B in each sector on the conflict A initiatesagainst B. We employ the signs of these expected effects to generate the fourhypotheses listed in the main text.


Empirical Model

Our theory indicates the empirical model ought to include bilateral sec-toral imports and exports, and national incomes of both states in thedyad. We include import to A from B and export from A to B in thefive Italianer (1986) sectors, denoted as agriculture import, agricul-ture export, chemical-mineral import, chemical-mineral export,energy import, energy export, manufactured import, manufac-tured export, miscellaneous consumption import, and miscella-neous consumption export. The raw data are purchased from the WorldTrade Flows database, sponsored by the University of California, Davis,and the National Bureau of Economic Research (Feenstra, 2000). They areprovided in current U.S. dollars across 10 Standard International TradeClassification (SITC) one-digit sectors and include all the countries thathave data. We convert these data into millions of 1995 constant U.S. dollarsand regroup them to conform to the Italianer (1986) sectors.

The one-digit SITC sectors we use are denoted by numbers from 0 to9 and their nature is as follows: (0) agriculture and fishery; (1) beveragesand tobacco; (2) minerals and inedible crude materials; (3) energy, fuels,

20 Reuveny’s work has two limitations from our perspective. It employs events data to measureall types of conflict, not just MIDs, and its sample includes a few major powers in 1963–1994.Reuveny’s estimates are not perfect (as in all studies), but the implied issue is empirical, nottheoretical.


and lubricants; (4) animals, vegetable oils, fats, and waxes; (5) chemicals;(6) basic manufactured products; (7) manufactured machinery and trans-port products; (8) miscellaneous consumption products; and (9) goodsnot classified by kind. The five Italianer sectors and the way they match theSITC sectors listed in parentheses are agriculture-fishery (SITC 0), chemical-mineral (SITC 2, 5), energy (SITC 3), manufactured products (SITC 6, 7),and miscellaneous consumption products (SITC 1, 4, 9, 8). We do not usethe SITC sectors directly because we do not have the required supply anddemand coefficient estimates to form ex ante expectations for each sector.The trade data record only nonzero values. So for dyads covered by thetrade database, trade values are set to zero for years showing no values, perexchange with Robert Feenstra.

Moving to the control variables, gdp initiator and gdp target are realGDPs of the conflict initiator and the target, respectively, in millions of 1995constant U.S. dollars. The data are from the World Bank (2002). We do notuse the trade-to-GDP ratio measure because it may misrepresent the effectof trade. For example, if trade does not change and GDP falls, the ratio rises.Suppose we find in this case that a rise in the ratio reduces conflict. Oneoften interprets this to imply a rise in trade reduces conflict, but in our caseit is the fall in GDP that reduces conflict, not a rise in trade. This potentialproblem is discussed by Mansfield and Pevehouse (2000) and Keshk et al.(2004). These studies include trade and GDP separately in the model, apractice we follow. By including the GDP, we also control the size of theeconomy, similar to what the trade-to-GDP ratio does.

The remaining controls are standard in studies of directed dyads (e.g.,Bennett and Stam, 2000a, 2004; Hegre, 2004). The variable contiguity isset to 1 when two states in a dyad are contiguous on land or are separatedby up to 150 miles of water and zero otherwise. Contiguous states are morelikely to fight one another. The distance is logged distance between thecapitals of two states in a dyad. Some scholars (e.g., Russett and Oneal, 2001;Dorussen, 2006) argue that a rise in distance decreases the probability ofMID initiation, whereas others (e.g., Ray, 2003) argue against including thisvariable. For completeness, we include both variables in our analysis.

We construct three democracy-related variables to capture the effectsof domestic political regimes. These variables are based on the POLITY2variable from the POLITY IV data set, ranging from −10 to 10 (−10 =high autocracy; 10 = high democracy): initiator democracy is the levelof democracy of the MID initiator, target democracy is the level ofdemocracy of the target, and regime dissimilarity is the absolute valueof the difference between their democracy levels. The initiator’s democracy


variable tests whether democratic countries are more likely to initiate MIDs.the target’s democracy variable tests whether democracies are more likely tobe MID targets, and the difference variable measures the effect of politicaldissimilarity on MID initiation.

Following Bennett and Stam (2000b, 2004), we include two capabilityvariables for each dyad, which are computed based on the COW compositeindex of national capabilities. power balance is the ratio of the largerstate’s capabilities divided by the total of the initiator’s and target’s capabil-ities, ranging between 0.5 (equality) and 1 (dominance). The ratio tests ifMID initiation is less likely for dyads of relatively more balanced capabili-ties. initiator capability ratio is the ratio of the initiator’s capabilitiesover total dyad capabilities and tests whether a rise in the initiator’s capabil-ities encourages MID initiation. alliance is a dummy variable indicatingthe presence (1) or absence (0) of dyadic defense pacts, neutrality pacts,or ententes. minor power is coded 1 if both states in a dyad are minorpowers and zero otherwise. In the EUGene computer program, which isused to generate these variables, the major powers are China, France, theUnited States, the United Kingdom, and USSR/Russia (Bennet and Stam,2000b).


Several issues guide the design of our research in this chapter. The firstissue is that our dependent variable aggregates the different types of MIDs,ignoring their variations. We have addressed this issue in the additionalanalysis in the main text.

A second issue is that conflict rises or falls in the theoretical model, butour dependent variable admits only values of 0 and 1. This issue applies toall studies using MID data and no one argues that conflict is really a 1–0process. The logit/probit estimation, however, predicts the probability ofa dispute, which is a continuous variable (like conflict in our theoreticalmodel), preserving the model’s spirit.

A third issue is that although many studies employ MID initiation (e.g.,Bennett and Stam, 2000a, 2004; Reiter and Stam, 2003; Hegre, 2004; Lai andSlater, 2006), the MID data code initiators as the side that first takes militaryaction. This practice ignores actions falling below the MID threshold, maynot represent who started the quarrel, ignores that for actual conflict tooccur the target is important, and overlooks the possibility that when oneside moves, the other may essentially ignore it. That said, we believe thatthese limitations are not crucial here; because our theory focuses on the


decision of one side to initiate military action, the MID initiation measureis appropriate for our purpose.21

Fourth, since the dependent variable is dichotomous, we use the probitestimator. Our sample includes 213,790 observations but, as usual, only avery small number of directed dyad-years involve MIDs (352).

Fifth, conflict as the dependent variable may affect some right-hand-side variables, including trade and democracy, causing simultaneity bias.This possibility applies to almost all statistical studies that model conflictas a function of various variables. Deviating from this norm, Keshk et al.(2004) model the simultaneous relationship between aggregate trade andMID. This method, however, is not feasible here. In our case, the systemof equations would have to include 11 endogenous variables (5 imports, 5exports, and 1 conflict variable), but the Maddala–Amemiya method usedby Keshk et al. (2004) only works for a system of two endogenous variables(one continuous and one dichotomous). Therefore, we employ a suboptimalsolution for addressing the simultaneity issue; we lag the right-hand-sidevariables, a strategy that is typically applied in the literature.

Sixth, given the pooled nature of our data, the error term of our modelmay be subject to serial correlation, heteroskedasticity, and temporal depen-dence within dyads. To control for heteroskedasticity and serial correlation,we estimate robust standard errors clustered over dyads. We also includethe peace-year counter (peace_years) and three cubic spline variables(splines) to control for the duration dependence of peace within dyads, assuggested by Beck, Katz, and Tucker (1998).

Seventh, sectoral trade flows may be correlated with one another, reflect-ing economic interrelationships, and may also affect other independentvariables, which may lead to multicollinearity. We investigate this possibil-ity by inspecting bivariate correlations and variance inflation factors (VIFs).VIF = 1/(1 − R2

I ), where R2I is R2 from regressing the ith independent vari-

able on all the other independent variables. If VIF > 10, multicollinearityis a problem. In this case, estimated coefficients and their variances remainunbiased, but the variances may be large. If the results are significant evennow, they are robust since the test is demanding (Achen, 1982; Gujarati,2002; Kennedy, 2005).

Finally, we assess the effects of trade variables on conflict by testinghypotheses on the signs of effects, computing the sizes of statistically sig-nificant effects, testing hypotheses on the equality of significant effects, and

21 Whereas most quantitative conflict studies use the MID data, some studies use events data.Several studies discuss advantages and disadvantages of events data versus MID data (e.g.,Reuveny, 2002).


employing the model for in-sample prediction. Because our sample is large,we employ the relatively more conservative two-tailed tests for the statisti-cal significance of the effects of our variables, even though our hypothesesare directional and one-tailed tests are appropriate. We employ significancetests because we analyze a sample, not the entire population. Even if oursample included all the countries in the 1970–1997 period, this would nothave been the population since we would not have data for 2000, 1960, andso on, and we only have one collected data set.

In evaluating the size of the effect, we compute the relative risk for eachsignificant variable, which is the probability of an event in a treatment groupdivided by its probability in a control group (e.g., Gartzke et al., 2001; Russettand Oneal, 2001; Dorussen, 2006). Specifically, we compute the probabilityof MID initiation when a variable rises by one standard deviation aboveits sample mean, divided by the probability of MID initiation when thisvariable is set at its sample mean, ceteris paribus. Strictly speaking, theinterpretation is as follows: Assume there are 100 MIDs and the relative riskrises 5% when variable X rises. If we had many samples for a model whenX rises, the average number of predicted MIDs across these samples wouldbe 105. The one-standard-deviation rise is widely used.

We also test whether the statistically significant effects of our trade vari-ables are equal in size (the insignificant effects are statistically indistinguish-able from zero), which is an important test of the practice of using total tradedata aggregated across sectors and flow directions and which is common inthe literature. This practice is equivalent to assuming the same coefficientapplies to imports and exports in different sectors or that all these tradeflows have identical effects on conflict.

We also evaluate the results with in-sample predictions. This evaluationcomputes the predicted probability of a MID for each directed dyad-yearand compares it to a threshold, which is typically set to the proportion ofMID initiation in the sample. The sample proportion of MID initiations is areasonable cutoff because an uninformed observer can always use this valueas a guide to provide an educated guess of the probability of MID initiation.If the predicted probability is larger than the threshold, the model is said tohave predicted a MID, and vice versa. The proportion of correctly classifiedcases indicates the model’s predictive power.

Empirical Findings

Table 6.2 in the main text presents the estimation results. The model has apseudo R2 above 0.3, indicating good explanatory power. In terms of thecontrol variables, the results largely resemble previous findings. The effect of


a rise in initiator capability ratio on the likelihood of MID initiation ispositive and significant at the 5% level. A state is more likely to initiate MIDswhen its share of the dyad’s total capability increases. The effect of a risein power balance is insignificant. Conflict is not less likely between stateswith relatively equal capabilities. The coefficient of alliance is insignificant,suggesting that the presence of a military alliance has little effect on MIDinitiation. The effect of a rise in initiator democracy is negative andsignificant at the 5% level, the effect of a rise in target democracy isinsignificant, and the effect of a rise in regime dissimilarity is positiveand significant at the 1% level. More democratic countries are less likelyto initiate MIDs, but not more likely to be targeted. Countries are morelikely to initiate conflict against those whose regimes differ more from theirown.

The effects of contiguity and distance are significant at the 1% leveland are positive and negative, respectively. Contiguous states are relativelymore likely to initiate MIDs against each other, whereas a rise in distancebetween countries reduces the likelihood of MID initiation. The effect ofminor power is negative and significant at the 10% level. Minor powerdyads are less likely to experience MIDs. The effects of initiator gdp andtarget gdp are positive and significant at the 1% level. Larger economiesare more likely to be initiators and targets of MIDs.

The results for the trade variables are discussed in the text. Here we listagain the obtained sign of the effect and note its statistical significancelevel. The coefficient of agriculture import is negative as expected andsignificant at the 1% level. The coefficient of agriculture export is neg-ative as expected but not significant. The coefficient of energy importis negative as expected and significant at the 10% level. The coefficientof energy export is positive as expected and significant at the 1% level.The coefficients of manufactured import and manufactured exportare both positive as expected and significant at the 10% and 5% levels,respectively. The coefficients of chemical-mineral import/export andmiscellaneous consumption import/export are not significant.

These effects vary across sectors and trade flows, as our theory sug-gests, but how do they compare with those of the nontrade variables? InTable 6.2, relative to the control group, the risk of MID initiation due to aone-standard-deviation rise above the sample mean is 44% higher for theinitiator capability ratio, 16% lower for initiator democracy, 20% higher forregime dissimilarity, 29% lower for log distance, 51% higher for initiatorGDP, and 29% higher for target GDP. Hence, the comparison in terms ofsubstantive effects indicates that trade has an important effect on conflictinitiation, but not the most important effect.


Table 6.A1. Equality tests on significant positive and negative effects ofsectoral trade flows

Agriculture/fishery import Energy import

Energy export 21.39∗∗ 9.30∗∗

Manufactured import 22.34∗∗ 4.68∗

Manufactured export 21.15∗∗ 5.35∗

Note: ∗ significant at 5% level; ∗∗ significant at 1% level.

We examine the diagnostic tests for multicollinearity. Even though thelargest correlation between any two independent variables is quite high (0.87for manufactured goods export and chemical/mineral export), all the indi-vidual VIFs are smaller than the threshold value of 10. In all cases, the corre-lation between any two export flows is the same as that between the importflows of the same sectors, which is one check of the validity of the data. TheVIF diagnostics indicate that multicollinearity is not a serious concern.

Additional Analysis

Table 6.A1 presents results for the coefficient equality tests. The first rowdenotes the two flows (agricultural import and energy import) that reducethe probability of MID initiation. The first column denotes the three flows(energy export, manufactured import, and manufactured export) thatencourage MID initiation. The results show that all the pairwise equalitytests are statistically significant at conventional significance levels fortwo-tailed tests. Hence, the effects of these sectors are not equal in size,suggesting (again) that aggregating exports and imports in different sectorsis inappropriate.

Table 6.A2 presents results for the in-sample prediction of MID initia-tion for Model 1. The cutoff probability is 0.0016 (the sample average

Table 6.A2. In-sample prediction of MID initiation

Predicted MID

Actual MID 0 1 Total

0 193,647 19,791 213,4381 67 285 352

Total 193,714 20,076 213,790

Note: Model predicts MID Initiation = 0 correctly in 91%of the cases; model predicts MID Initiation = 1 correctly in81% of the cases.


Table 6.A3. Effects of bilateral sectoral flows on display of force and use of force

(1) Display of force+ (2) Use of force+Agriculture-fishery import −0.00046 −0.00035

[4.25]∗∗∗ [3.31]∗∗∗

Agriculture-fishery export −0.00007 −0.00007[0.74] [0.74]

Energy import −0.00004 −0.00005[1.76]∗ [1.83]∗

Energy export 0.00004 0.00004[2.61]∗∗∗ [2.26]∗∗

Chemical-mineral import −0.00004 −0.00001[0.69] [0.28]

Chemical-mineral export −0.00008 −0.00012[1.26] [1.80]∗

Manufactured import 0.00003 0.00003[1.94]∗ [2.05]∗∗

Manufactured export 0.00003 0.00005[2.32]∗∗ [3.14]∗∗∗

Miscellaneous consumption import −0.00006 −0.00003[0.88] [0.75]

Miscellaneous consumption export 0.00002 −0.00012[0.51] [1.82]∗

Initiator capability ratio 0.36393 0.49638[2.35]∗∗ [2.87]∗∗∗

Power balance 0.04679 0.08097[0.26] [0.40]

Initiator democracy −0.00708 −0.00618[1.93]∗ [1.52]

Target democracy −0.00019 0.00356[0.05] [0.83]

Regime dissimilarity 0.01147 0.00898[3.40]∗∗∗ [2.31]∗∗

Alliance 0.00617 0.00714[0.08] [0.08]

Log of distance −0.07961 −0.07474[4.85]∗∗∗ [4.13]∗∗∗

Contiguity 0.82614 0.81936[6.33]∗∗∗ [5.63]∗∗∗

Minor power −0.34438 −0.39713[2.01]∗∗ [1.91]∗

Initiator GDP 0.06672 0.03584[2.84]∗∗∗ [1.41]

Target GDP 0.10066 0.11342[4.32]∗∗∗ [4.37]∗∗∗

Observations 213790 213790

Note: Z statistics in brackets. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at 1%.Constant, peace-year, and spline variables not reported.


probability of MID initiations). Model 1 correctly predicts the absence ofMID initiation in 91% of the dyad-years and the presence of MID initia-tion in 81% of the cases. Thus, the model performs reasonably well in termsof in-sample prediction.

Table 6.A3 reports the model results based on alternative indicators ofdyadic conflict. For both models, it is clear that the results closely resemblethose obtained for the model based on all MIDs. In the model for MIDsinvolving at least the display of force, the statistical results for these controlvariables are essentially identical to those based on all MIDs. For the modelfor MIDs involving the use of force and war, the results for the controlvariables are similar to those based on all MIDs, except that the effects ofinitiator democracy and initiator gdp are not significant.

PART III

BRINGING IN THE ENVIRONMENT

SEVEN

Democracy and the Environment

INTRODUCTION

In this chapter, we begin the third and last part of this book. So far, theinquiries we have conducted could in principle have taken place anywhere,be it on planet Earth or somewhere else in space, as long as there existpeople who interact with one another, markets that enable people to attachmonetary values to goods and services, and governments that decide to useforce against other countries. The global political economy and the interac-tions it entails, however, invariably occur within the confines of the physicalenvironment, the biosphere of planet Earth. Consequently, the political andeconomic forces we focus on in this book – democracy, economic openness,and military conflict, and the complex transformations they entail – caninfluence this very physical environment within which we all live.

The questions we seek to answer in the third part of the book concernwhether and how these effects on the environment occur. Is a rise in thelevel of democracy good or bad for the environment? How does a rise in thelevel of international trade affect the environment? What are the implica-tions of being involved in military conflict for a country’s physical envi-ronment? Chapters 7, 8, and 9 in the third part of the book address theseimportant questions.

Does a rise in the level of democracy of a country reduce or increaseits environmental degradation? The democracy–environment link has cap-tivated policy makers and the media with conflicting arguments and evi-dence, for which we are not taking sides. For example, former U.S. VicePresident Al Gore has asserted that greater political and civil freedoms pro-mote environmental quality (Gore, 1992). Geographer and media figureJared Diamond (2005b) noted that the Scandinavian democracies have thebest environmental record in the world. The experience of Central and

205


Eastern Europe is also quite telling (Forest Watch Indonesia and GlobalForest Watch, 2002). Under the Communist system, the environment seri-ously deteriorated; the arrival of democracy to the region resulted in muchbetter environmental policies. On the other hand, Diamond argues thatautocracy also can benefit the environment because not only can it legis-late and implement laws by decree but it also may possess a long-run viewdue to its potentially long-lasting reign. In his recent bestseller Collapse:How Societies Choose to Fail or Survive, Diamond (2005a) cites a numberof autocrats that protected forests, including seventeenth-century Japan’sTokugawa Shogunate and the Dominican Republic’s Rafael Trujillo andJoaquin Balaguer. Diamond (2005b) also points to contemporary China,which phased out leaded gasoline over a period of only one year and endedlogging at once by decree, demonstrating the beneficial impact of autocracyfor the environment. Hence, Diamond claims that one cannot generalizethe effect of democracy or autocracy on the environment.

Scholars also debate the effect of democracy on environmental degrada-tion, both theoretically and empirically. Some theorists claim that democ-racy reduces environmental degradation; others argue that democracy doesnot reduce environmental degradation and may even harm the environ-ment. Despite the contentious nature of this issue, systematic empirical evi-dence is relatively scant and mixed, particularly in terms of large-N statisti-cal analysis. The various theoretical mechanisms on the effect of democracyon environmental degradation, though often conflicting, are all plausiblegiven their theoretical assumptions. In other words, it is not possible toadjudicate the competing claims based solely on the theoretical arguments.Because these causal mechanisms may operate at the same time, it is impor-tant to study their overall or net effect empirically, which is the focus of ouranalysis in this chapter.

The dependent variable in our empirical/statistical model, environmentaldegradation, is measured with five salient types of human-induced degra-dation: carbon dioxide (CO2) emissions, nitrogen oxide (NOx) emissions,land degradation, forest depletion in terms of the rate of deforestationand the share of forested area, and organic pollution in water. For eachtype of degradation, we estimate several regression-based statistical modelsemploying various measures of democracy or autocracy.

Our empirical analysis differs from previous studies in a number ofways. First, previous empirical analyses studied the effect of democracy onthe signing of international environmental agreements, resource scarcityand access to environmental amenities, and human actions that directlyharm the environment. We focus on human activities that directly affect

Democracy and the Environment 207

environmental quality. Although it is important to study resource scarcity,access to environmental amenities, and commitment to environmental pro-tection, we believe it is also important to study human actions that directlydamage the environment. After all, the best way to protect the environmentis to minimize the damage to the environment in the first place.

Second, our sample size is generally larger than that of previous studies.Depending on data availability for the environmental indicators, the num-ber of countries included in the empirical sample varies from 105 for theland degradation analysis to 143 for the CO2 emissions per capita analysis.A larger sample size allows us to have more variations in the data to facilitatestatistical inference and gives us more confidence in the generalizability ofour findings.

Third, the empirical results for the effect of democracy on the environ-ment are consistent across all the aforementioned types of human-induceddegradation, except for deforestation: a rise in democracy reduces environ-mental degradation and improves environmental performance. The effectson forest depletion are mixed. Our cross-sectional analysis for forests findsthat democracies tend to have relatively larger shares of forested area com-posing their total land area, but they also have higher rates of deforestationrelative to nondemocracies. The substantive effect of democracy on theenvironment is considerable, but it varies in size across different aspects ofenvironmental degradation. We also find that democracy has nonmono-tonic effects that vary across the environmental indicators.

The remainder of this chapter proceeds as follows. The next section dis-cusses the theoretical channels linking democracy to environmental degra-dation. The section that follows presents our empirical model in terms ofthe variables, data, measures, and expected effects. The fourth section of thischapter discusses several technical design issues pertaining to the empiri-cal analysis, followed by a section presenting and discussing our empiricalresults. Finally, we summarize our key findings.

EFFECT OF DEMOCRACY ON THE ENVIRONMENT

This section discusses the two opposing views regarding the effect of demo-cracy on the environment and the associated empirical evidence. The debatein the literature turns on institutional attributes of political regimes affectingenvironmental problems: the role of public opinion in policymaking, inter-est groups’ aggregation and representation, state autonomy, social move-ment mobilization, and the flow of information. Scholars take positions inthe debate by emphasizing some of these regime characteristics.


Democracy Improves Environmental Quality

A rise in the level of democracy in a country improves its environmentalquality. Scholars holding this view have offered five different supportingarguments. According to the first argument, political rights and freedom ofinformation promote the cause of environmental interest groups, which inturn raise public awareness and encourage environmental legislation (see,e.g., Schultz and Crockett, 1990; Payne, 1995). This effect works throughenvironmental groups and public opinion at large. Information on envi-ronmental issues flows more freely, and political rights are more numerousand better protected in a democracy than in an autocracy. Environmentalgroups, therefore, are often more successful at informing people and orga-nizing them to act on environmental problems in a democracy than in anautocracy. Whereas the elite in an autocracy may be more educated than thepublic (as education tends to rise with income), the autocratic regime cen-sors information flows, and its decision making is more autonomous thanthat of a democratic government. Environmental degradation may not bereported by the media to the people. In contrast, because democracy allowsfor free media, environmental problems are more likely to be reported inthe news. People in a democracy, therefore, are more likely to be informedabout the environment than are members of the elite in an autocracy. Betterinformed actors, in turn, are more likely to act on environmental problems,raising environmental quality.

A second argument is that democracies are more responsive to the envi-ronmental needs of the public than are autocracies (Kotov and Nikitina,1995). This argument works through electoral accountability and the abilityof groups to mobilize socially, achieve political representation, and influencepublic policymaking. Democracies hold regular and free elections, whichcan bring to power new parties, including those friendly to the environ-ment (e.g., the Green Party in Germany). In an autocracy, the distributionof political power is concentrated, reducing the likelihood that environmen-talists will come to power. Thus, environmentalists stand a greater chanceof affecting policymaking in a democracy than they do in an autocracy.Of course, this logic implies that people can also freely elect extreme anti-environmental parties. Casual observation, however, suggests that in realitysuch situations do not occur frequently.

A third argument focuses on institutional and ideational features ofdemocracy. According to this argument, democracies are more likely tocomply with environmental agreements because they respect the rule of


law, which in turn raises environmental quality (Weiss and Jacobsen, 1999).Also, democracies respect economic freedom and, therefore, have marketeconomies (Berge, 1994). The market, in turn, promotes environmentalquality. Furthermore, because democracies respect human life more thanautocracies, they are more responsive to life-threatening environmentaldegradation. And to the extent that democracies engage in fewer wars, theyshould have a higher level of environmental quality because wars destroythe environment (Gleditsch and Sverdlop, 2003). Finally, famines promoteenvironmental degradation because they divert attention away from long-run environmental concerns. Famines tend not to occur in democraciesbecause democratic governments are more responsive to the needs of thepeople (Sen, 1994). Hence, environmental degradation will be higher inautocracies than in democracies.

A fourth argument expects that the elite in an autocracy will be lesspro-environment than the masses or the public at large in a democracy(Congleton, 1992). The logic of this argument relies on environmental reg-ulation that curtails pollution and waste. With prevailing technologies andmaterials, environmental regulation lowers production and consumption,which, in turn, imposes a higher cost on the elite in an autocracy than on themasses in a democracy, because the ruling elite in an autocracy hold a muchlarger share of national income than most people in a democracy and aretherefore relatively less pro-environment than the masses in a democracy.

A fifth argument begins with the observation that because environmentaldegradation typically develops slowly, the discount rate and time horizonof the government have important impacts on environmental regulation(Congleton, 1992). The masses in a democratic country should have less atstake over regime change than the elite in an autocracy. In an autocracy, theelite are tightly linked to the leader. If the leader loses power, the elite maysuffer heavy losses or even lose their lives. Facing this possibility, the elitemay wish to prevent regime change by force, and to this end more resourcesare allocated to oppression. The elite may also think that the change isinevitable, becoming hedonic. Both actions raise the discount rate andreduce the time horizon of the autocratic government. As a result, theruling elite in an autocracy will ignore the environmental damage expectedin the future. If they invest more today to suppress real or potential rebels,they allocate resources away from environmental issues. If they consumemore today, they ignore environmental degradation that takes a long timeto rectify or current activities that will cause damage in the future. In bothcases, environmental quality will decline.


Democracy May Worsen Environmental Degradation

Democracy does not reduce environmental degradation and may evenincrease it. Scholars have provided several explanations for this position.First, Hardin (1968) warns about the impending hazards of uncheckednatural resource exploitation and environmental mismanagement by self-interested individuals and groups. When private property rights of naturalresources are not well defined, as is often the case with “the commons”(e.g., clean air, oceans, forests), free individuals or interest groups tend tooverexploit such resources and ignore the damage their economic actionsinflict on the environment. Gleditsch and Sverdlop (2003: 70) note thatHardin’s “Tragedy of the Commons” does not encourage confidence in theeffect of economic and political freedom on environmental quality.

Second, Paehlke (1996: 28) argues that “the great danger for both demo-cracy and the environment is that, while economy and environment are nowglobal in character, democracy functions on only national and local deci-sion levels.” Thus, global environmental problems may not necessarily beattended to in a timely manner. Heilbronner (1974) argues that global pop-ulation growth threatens global environmental quality. Being autonomousdecision makers, autocracies can curtail human reproduction, but democ-racies are held accountable by the public and therefore respect citizens’rights, including those involving human procreation.

Third, Dryzek (1987) notes that democracies tend to be market econo-mies, where business interest groups have considerable clout. His argumenthighlights the asymmetric influence of profit-oriented corporate interestsin capitalist democracies. Dryzek (1987: 121) lists countries in which demo-cracy is systematically skewed in favor of corporate interests, while “envi-ronmental groups have a hard time getting a foot in the door.” Corporateinterests, in turn, seek to maximize profit, not necessarily to better environ-mental quality. Thus, democratic leaders accountable to business intereststhat support their coming to power may not necessarily value environmentalquality. “Polyarchy,” Dryzek argues, “will normally yield to the imperativesof the market, if not always to the interests of large corporations. . . . Attheir corporatist worst, polyarchies degenerate into caricatures of the ideal,with some dire consequences for ecological rationality” (Dryzek, 1987:125).1

1 It is possible to argue that Dryzek’s argument is not about democracy but rather about the effectsof wealth in a democracy. Our reading of Dryzek differs somewhat: his argument focuses on theability of business to affect government policies through the democratic process and, hence, isabout democracy.


Fourth, Midlarsky (1998) argues that democracies often experience pub-lic policy inaction where environmental degradation is concerned. Demo-cratic leaders have the tendency to please competing interest groups to winas many votes as possible. “Corporations and environmental groups canfight each other to a standstill, leaving a decision making vacuum instead ofa direct impact of democracy on the environment. As a result of budget con-straints, democracies may not be responsive to environmental imperativesbut to more pressing issues of the economic subsistence of major portionsof the voting public” (Midlarsky, 1998: 159). In addition, democratic gov-ernments may be reluctant to alleviate environmental degradation becausesome groups are expected to benefit (or lose) from environmental policiesmore than others (Midlarsky, 1998: 159).

Previous Empirical Studies

Extant empirical studies may be categorized according to the nature of theirdependent variables. One set of studies examines government commitmentto environmental quality in terms of signing international agreements thatprotect the environment (e.g., Congleton, 1992; Neumayer, 2002). A secondset of studies investigates resource scarcity and access to environmentalamenities such as safe water or sanitation (e.g., Shafik, 1994; Torras andBoyce, 1998). A third set of studies explores human activities harmful tothe environment, such as greenhouse gas emissions (e.g., Midlarsky, 1998;Gleditsch and Sverdlop, 2003).

Focusing on international agreements assumes (implicitly or explicitly)that they promote environmental quality. These agreements, however, alsoreflect international negotiations and bargaining. The end product maynot necessarily address the specific problems of any particular country. Attimes, these agreements also include “cheap talk” that is intended to appeaseenvironmentalists. Resource scarcity and access to environmental amenitiescan reflect nonenvironmental and structural conditions concerning wealthand resource endowment, which are not directly related to environmentalquality.

The reader may recall that we seek to study the effect of democracy onhuman actions that directly degrade the environment. Before we proceed,a word of caution is necessary. Although the distinction between signingagreements and human actions that hurt the environment is clear, thedistinction between conditions of resource scarcity and actions that causeenvironmental degradation may be blurred. For example, land degrada-tion can suggest soil pollution by humans, but it may also indicate the


scarcity of productive land that forces humans to overexploit land alreadyin use.

The empirical literature pertaining to our focus on human actions thatdirectly damage the environment is relatively small. In a sample of 118countries for the year 1989, Congleton (1992) found that democracies havehigher levels of methane and chlorofluorocarbon emissions per capita thanautocracies. Midlarsky (1998) reports several findings. A rise in the level ofdemocracy increases CO2 emissions per capita in a sample of 98 countriesin the year 1990, soil erosion by water in a sample of 97 countries for the1980s, and the percentage of annual deforestation between 1981 and 1990in a sample of 77 countries. The level of biodiversity, as measured by thepercentage of protected land area, rises with democracy for a sample of 100countries in 1993. But democracy does not affect fresh water availability in asample of 97 countries in 1990 and the level of soil erosion by chemicals in asample of 97 countries during the 1980s. Barrett and Graddy (2000) foundthat a rise in democracy lowers per capita sulfur dioxide (SO2) emissions ina pooled sample of countries for about 33 years and particulate emissionsin a sample of 27 countries for about 38 years. The effect of democracy onwater pollution is not statistically significant in a sample of 45 countries forabout 29 years. Torras and Boyce (1998) found that the level of democracy,measured by the 1995 values of the Freedom House indicators of politicalrights and civil liberties, reduced air pollution (SO2, smoke, particulateemissions) in pooled samples of 19–42 countries from 1977 to 1991 anddecreased water pollution (dissolved oxygen, fecal coliform, access to safewater, and access to sanitation) from 1977 to 1991 in pooled samples of about58 countries. These results, however, do not appear to be robust acrosssamples. Based on a sample of the 148–185 sites in 24 countries acrossthree different periods (1979–1982, 1983–1986, and 1987–1990), Scruggs(1998) found that democracy does not affect water pollution and particulateemissions, but it reduces SO2 emissions. Gleditsch and Sverdlop (2003)reported that democracy reduced CO2 emissions per capita for a sample of108 countries in 1990.

We believe that it is safe to conclude that extant empirical evidence in theliterature on human actions that directly degrade the environment is mixed.It is also apparent that some studies employed relatively small samples interms of the country, year coverage, or both. Previous studies also differed interms of empirical model specification. For example, many empirical anal-yses did not control for the environmental Kuznets curve effect (see laterdiscussion), and none of the studies cited earlier controlled for the pos-sible effect of military conflict on the environment (except for Midlarsky,1998) or the effect of trade openness. In addition, these studies focused on


different aspects of environmental degradation and their number of indi-cators ranged between one and seven. Some of the indicators used measureenvironmental amenities, as opposed to our focus on human actions thatdamage the environment. In short, room exists for additional systematicempirical analyses that investigate multiple indicators of environmentaldegradation in samples that are as large as possible while controlling for theKuznets curve effect, trade openness, and military conflict.


This section presents our empirical model for evaluating the effect of democ-racy on environmental degradation, related research design issues, andthe statistical findings. Because the competing causal mechanisms of howdemocracy affects the environment may operate simultaneously, the pur-pose of our empirical analysis is to assess the net effect of these forces,finding out overall if democracy is good or bad for the environment. As inthe other chapters, this discussion is self-contained and does not require sta-tistical expertise. The technical details of the statistical model, data sources,and measure construction are in the chapter appendix, following the samestructure of presentation as in the main text.

Empirical Model

To assess empirically the effect of democracy on the environment, we specifyand estimate a statistical model of environmental degradation. As in otherchapters, we denote variables with small capital letters, and their coefficientswith Greek notations. Each coefficient indicates the effect of the indepen-dent variable on the dependent variable. The notations �y and �c denotevectors of coefficients for year and country fixed-effects variables, respec-tively. The notation εt denotes the unexplained random error in the model.The variable subscripts t and t − 1 indicate the time period of the variable,where t represents the current period and t − 1 the previous time period(a lagged variable). To simplify the presentation, we refer to the variableswithout time subscripts t or t − 1.

environmentt = �0 + B1regime typet−1

+ �2lagged environmentt−1 + �3real gdppct−1

+ �4real gdppc squaredt−1 + �5trade opennesst−1

+ �6population densityt−1 + �7wart−1

+ �yyear fixed effectst

+ �ccountry fixed effectst + εt. (7.1)


The dependent variable, environment, is a multidimensional conceptof environmental degradation. We choose to focus on five specific humanactivities that directly harm the environment: greenhouse gas emissions,air pollution, water pollution, forest depletion, and land degradation. Weanalyze one representative, salient indicator for each area, but two indicatorsfor forest depletion. Hence, we have six dependent variable measures.

Specifically, the first dependent variable focuses on CO2 emissionsbecause this gas constitutes about 60% of greenhouse emissions. The secondindicator is NOx, because this gas is a major health hazard and an impor-tant source of smog in urban areas. The third indicator relates to organicpollution in water. The fourth and fifth indicators are the rate of defor-estation and the forested area share out of total country area, respectively.The sixth indicator concerns land degradation, which seriously underminesagricultural activities.

We measure the key independent variable, regime type, in two differentways: continuously and dichotomously. The continuous or interval measureof regime type is the widely used indicator of the level of democracy fromthe POLITY IV data employed in previous chapters. In the dichotomousmeasurement approach, the variable democracy dummy is set to 1 if acountry’s polity score is greater than or equal to 6, and zero otherwise (i.e.,anocracies and autocracies); the variable autocracy dummy is set to 1 ifa country’s polity score is smaller than or equal to −6, and zero otherwise(i.e., anocracies and democracies).

The literature on environmental degradation motivates the specificationof the control variables in the model. The variable real gdppc is the grossdomestic product (GDP) of a country expressed in real terms, per capita;real gdppc squared is the square of the real GDP per capita variable. Thetwo variables together help capture the nonlinear effect of development onthe environment, a phenomenon referred to as the environmental Kuznetscurve, which is discussed at length in the appendix. The trade opennessvariable is the share of a country’s total trade over its GDP, indicating theimportance of trade for the national economy; population density is thepopulation size of a country divided by its total area; and war indicatesthe presence or absence of a country’s involvement in an interstate orintrastate war in a year.

For the CO2 emissions per capita and water pollution models, wehave pooled time-series cross-sectional data. This type of data structurerequires additional control variables: lagged environment is the valueof the dependent variable from the previous year (for which the ratio-nale is discussed in the appendix), country fixed effects are country


dummy variables that capture country-specific heterogeneity, and yearfixed effects are year dummy variables that measure year-specific effects.These additional control variables do not appear in the other models becausewe only have cross-sectional data for those other environmental variables.


This section discusses several design issues that are special to the empir-ical analysis in this chapter. Again, we refer the interested reader to theappendix of this chapter for further details. The first design issue concernsthe sample information and whether we conduct a cross-sectional analysisor a cross-sectional time-series (pooled) analysis for each environmentalindicator. Data availability determines this choice. For the CO2 emissionsand water pollution, we have time-series cross-sectional data and conductpooled analyses. The sample of CO2 emissions per capita includes 143countries from 1961 to 1997, and the sample of organic pollutants in waterincludes 112 countries from 1980 to 1998. For NOx, deforestation rate,forested land area, and land degradation, we have cross-sectional data andconduct cross-sectional analyses. The NOx sample includes 118 countriesin 1990. The rate-of-deforestation sample covers 134 countries over twodecades (1980s and 1990s). The share-of-forested-area sample also covers134 countries but for three different years (1980, 1990, and 2000). The land-degradation sample includes 105 countries in the 1980s. It is worth notingthat our sample coverages are larger than previous empirical studies on thetopic.

The second design issue concerns the possibility of endogeneity for ourright-hand-side variables. In the appendix, we provide possible examplesfrom our context. We address this possibility by lagging all the right-hand-side variables.

Third, slow-changing national structural variables (e.g., climate, educa-tion) and global or local biophysical attributes (e.g., atmospheric integrity,existing damage) may also affect environmental degradation. Hence, envi-ronmental degradation is likely to exhibit inertia. We model this tendencywith the variable lagged environment when we have time-series cross-sectional data (for CO2 emissions and organic water pollution).

Fourth, we employ appropriate econometric techniques to deal with pos-sible assumption violations associated with the error term of the statisticalmodel. Failure to do so may result in incorrect inferences regarding the effectof democracy on the environment. We also test for evidence of excessivecorrelation among the independent variables.


Finally, to give some substantive meaning to our findings, we discussthe size of democracy’s effect on the environment. We discuss the sizeof the estimated coefficient of democracy, provided that the coefficient isstatistically different from zero. If the coefficient is not statistically differentfrom zero, it is not necessary to discuss the size of the effect. In models thatinclude a lagged dependent variable, we also compute the long-run impactof democracy.

Empirical Findings

In discussing our findings, we first focus on the main analysis using thecontinuous measure of regime type, and then we discuss the results usingthe dichotomous measures of regime type. Findings for the control variablesand additional technical details can be found in the appendix.

Table 7.1 reports the results for the net effect of the level of democracy onenvironmental degradation. The columns of this table present the resultsfor CO2 emissions per capita, NOx emissions per capita, organic pollutantsin water, the rate of deforestation, the share of total land area that is forested,and the share of degraded land, respectively.

Across the six columns in Table 7.1, the net effect of the level of democracyon six different environmental indicators is always statistically different fromzero, consistently in the direction of reducing environmental degradation,except for the deforestation rate. Relative to their less-democratic counter-parts, more democratic countries produce fewer CO2 emissions per capita,fewer NOx emissions per capita, fewer organic pollutants in water, and lessland degradation. The effects of democracy on forest depletion, however,are mixed: democracies experience higher deforestation rates (recall thata negative value for the rate of deforestation implies faster deforestationwhereas a positive value indicates faster afforestation) but have larger sharesof forested land. It is worth noting that the effects of democracy on CO2

emissions and the level of organic pollution in water are significantly dif-ferent from zero in spite of the taxing country and year dummies in themodels.

One may wonder how to reconcile the conflicting findings of the twoindicators of forest depletion. On the one hand, a rise in democracy is asso-ciated with faster deforestation. On the other hand, the share of the forestedland area rises with the democracy level. Since the forested land area mea-sure is largely cross-sectional and the deforestation rate variable capturesthe temporal change from decade to decade in our sample, the results sug-gest that although democracies tend to have relatively more forests, they

Tabl

e7.

1.E

ffec

tofl

evel

ofde

moc

racy

onen

viro

nmen

tald

egra

dati

on

Org

anic

wat

erA

nn

ual

Fore

star

eaD

egra

ded

area

CO

2N

OX

pollu

tan

tsde

fore

stat

ion

shar

eof

shar

eof

per

capi

tap

erca

pita

per

km3

rate

lan

dar

eala

nd

area

LEV

EL

OF

DE

MO

CR

AC

Y−0

.002

1∗∗−0

.019

8∗−0

.003

3∗−0

.118

7∗∗∗

0.08

07∗∗

∗−0

.028

1∗

(0.0

011)

(0.0

140)

(0.0

021)

(0.0

362)

(0.0

236)

(0.0

219)

TR

AD

EO

PE

NN

ESS

−0.0

006

0.00

43∗∗

0.00

050.

0144

∗∗−0

.003

0−0

.366

1∗

(0.0

011)

(0.0

025)

(0.0

004)

(0.0

070)

(0.0

035)

(0.2

692)

WA

R−0

.053

9∗∗∗

0.11

020.

0063

0.48

32∗

−0.3

983∗∗

0.39

17∗

(0.0

189)

(0.1

891)

(0.0

282)

(0.3

418)

(0.1

749)

(0.3

004)

RE

AL

GD

PP

C0.

0000

9∗∗∗

0.00

014∗∗

∗0.

0000

10.

0004

∗∗∗

−7.2

033e

-05

6.37

22∗∗

∗

(0.0

0002

)(0

.000

04)

(0.0

0001

)(0

.000

1)(6

.877

5e-0

5)(2

.347

0)R

EA

LG

DP

PC

SQU

AR

ED

−2.8

0e-0

9∗∗∗

−2.3

9e-0

9∗−4

.83e

-10∗

−1.0

9e-0

8∗∗∗

9.96

e-10

−0.3

84∗∗

∗

(5.3

3e-1

0)(1

.53e

-09)

(3.1

3e-1

0)(3

.57e

-09)

(2.4

9e-0

9)(0

.1.4

5)P

OP

ULA

TIO

ND

EN

SIT

Y0.

0007

∗∗∗

−0.0

005∗∗

0.00

01∗∗

∗−0

.001

1∗∗−0

.000

050.

2854

∗∗∗

(0.0

002)

(0.0

002)

(0.0

000)

(0.0

006)

(0.0

002)

(0.0

966)

LAG

GE

DE

NV

IRO

NM

EN

T0.

8658

∗∗∗

0.78

95∗∗

∗

(0.0

228)

(0.0

832)

Con

stan

t0.

1247

∗1.

9067

∗∗∗

1.40

91∗∗

∗−2

.840

2∗∗∗

3.40

46∗∗

∗−2

2.69

86∗∗

(0.0

643)

(0.1

815)

(0.5

286)

(0.5

745)

(0.2

879)

(9.4

663)

Obs

erva

tion

s38

3310

813

4420

425

510

5R

20.

990.

410.

990.

280.

130.

20

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nta

t10%

;∗∗

sign

ifica

nta

t5%

;∗∗∗

sign

ifica

nta

t1%

.Coe

ffici

ents

for

year

and

cou

ntr

ydu

mm

yva

riab

les

not

show

nfo

rm

odel

sof

CO

2an

dor

gan

icw

ater

pol

luta

nts

.

217


also tend to deplete them faster. At the theoretical level, democracy mayincrease or decrease environmental degradation. Empirically, deforestationturns out to be the only aspect of environmental degradation on which theeffect of democracy is positive.

How large are the substantive effects of democracy on these six environ-mental indicators? In this computation, we raise the democracy measureone standard deviation above its mean in each sample while we consistentlyhold other continuous variables constant at their sample means and the warvariable at zero. As discussed in the appendix, because of the differences inresearch design and data availability, we can show both the immediate andthe long-run effects of democracy (accumulated via the lagged dependentvariable) for CO2 emissions and water pollution. For the other indicators,we are only able to show the immediate effect of democracy.

Based on Table 7.1, a one-standard-deviation increase in the level ofdemocracy (7.6) above its sample mean (0.84) causes CO2 emissions percapita to fall by 0.47%. The immediate effect of democracy on CO2 emis-sions per capita appears small. The effect that cumulates over time throughthe lagged dependent variable is 0.12 metric tons per capita, which is about7.5 times of the size of the immediate effect. This amounts to a declineof about 4% in CO2 emissions. It should be noted that these numberspertain to per capita carbon emissions. Hence, if its democracy level risesone standard deviation above its sample mean, a country such as China,whose population is about 1.3 billion people, will see a decline of 156million metric tons per year in the long run – a large drop that can makea difference for the global environment, considering that in 1998 the totalcarbon emissions of Asia and the Pacific region (2.5 billion people) were2,167 million metric tons per year (UNEP, 2003: 215).

In the second column of Table 7.1, the direct effect of democracy on NOx

emissions per capita is much larger. An in-sample one-standard-deviationincrease in democracy (7.60) above its sample mean (1.3) causes NOx

emissions per capita to decline by 14%. This effect is larger than the long-run effect of democracy on CO2 discussed earlier, which in and of itself isnot small.

In the third column, an in-sample, one-standard-deviation increase indemocracy (7.26) above its sample mean (3.44) causes the level of organicpollutants in water to decline by about 2.4%. The long-run net effect ofa one-standard-deviation increase in democracy is much larger, reaching11%.

In the fourth column, an in-sample, one-standard-deviation increase indemocracy (6.6) above its sample mean (1.25) causes the average annual


deforestation rate to rise by about 271%. This obviously is a very largeeffect. In the fifth column, an in-sample one-standard-deviation increase indemocracy (6.9) above its mean (1.86) raises the share of forested land in acountry by 75%, which also is a very large effect.

Finally, in the sixth column, a one-standard-deviation rise in democracy(7.8) above its mean (0.2) causes the share of severely and very severelydegraded land out of total land area to decline by 20%, and is also a relativelylarge effect.

Additional Analyses

Seeking to further probe the robustness of our results, we conduct severaladditional analyses, based on (1) the dichotomous measure of democracy,(2) the dichotomous measure of autocracy, and (3) a composite environ-mental indicator that aggregates a number of environmental attributes. Forthese three analyses, we discuss the results for the key regime-type variablesand discuss the control variables in the appendix.

Tables 7.2 and 7.3 present results for the dichotomous measures of democ-racy and autocracy, respectively. Although these measures lose informationcontained in the level variable, they provide a sharper contrast of the dif-ference in effect between democracy and nondemocracy (a group includ-ing anocracies and autocracies), or between autocracy and nonautocracy(a group including anocracies and democracies). We therefore view thedichotomous measures of democracy and autocracy as complementary to,rather than substitutes for, the level of democracy measure.

In Table 7.2, we find that the net effect of the transition to democracyis not consistently significant across the five dimensions of environmentaldegradation. Democratic regimes do not appear to be different from non-democratic regimes in terms of CO2 emissions per capita, NOx emissionsper capita, or the level of organic pollutants in water. But they do havesignificantly lower levels of land degradation, larger forested land area, andhigher deforestation rates compared with nondemocratic countries.

In contrast, Table 7.3 shows that relative to nonautocratic countries,autocratic regimes experience higher CO2 emissions per capita, higher NOx

emissions per capita, higher levels of organic pollutants in water, smallerforested areas, and lower deforestation rates. They also tend to exhibit moreland degradation, an effect that is statistically somewhat weaker.

Taken together, Tables 7.1, 7.2, and 7.3 demonstrate the patterns in theeffects of political regime type on different dimensions of environmen-tal degradation. As Table 7.1 shows, the continuous democracy measure

Tabl

e7.

2.E

ffec

tofd

emoc

racy

onen

viro

nmen

tald

egra

dati

on

Org

anic

wat

erA

nn

ual

Fore

star

eaD

egra

ded

area

CO

2N

Ox

pollu

tan

tsde

fore

stat

ion

shar

eof

shar

eof

per

capi

tap

erca

pita

per

km3

rate

lan

dar

eala

nd

area

DE

MO

CR

AC

YD

UM

MY

0.01

19−0

.165

1−0

.023

5−0

.883

2∗∗0.

8884

∗∗∗

−0.6

729∗∗

(0.0

153)

(0.1

785)

(0.0

282)

(0.4

409)

(0.2

907)

(0.3

475)

TR

AD

EO

PE

NN

ESS

−0.0

006

0.00

44∗∗

0.00

050.

0146

∗∗−0

.002

8−0

.378

5∗

(0.0

011)

(0.0

025)

(0.0

004)

(0.0

074)

(0.0

037)

(0.2

672)

WA

R−0

.054

1∗∗∗

1.03

e-01

0.00

690.

4943

∗−0

.382

7∗∗0.

3272

(0.0

189)

(2.0

2e-0

1)(0

.027

5)(0

.359

5)(0

.184

5)(0

.299

4)R

EA

LG

DP

PC

8.37

e-05

∗∗∗

1.17

e-04

∗∗∗

7.10

e-06

0.00

04∗∗

∗−5

.05e

-05

6.12

17∗∗

∗

(1.7

8e-0

5)(3

.72e

-05)

(8.9

2e-0

6)(0

.000

1)(6

.75e

-05)

(2.3

207)

RE

AL

GD

PP

CSQ

UA

RE

D−2

.71e

-09∗∗

∗−1

.64e

-09

−4.4

7e-1

0∗−8

.96e

-09∗∗

∗3.

78e-

10−3

.65e

-01∗∗

∗

(5.3

7e-1

0)(1

.46e

-09)

(3.0

9e-1

0)(3

.68e

-09)

(2.4

9e-0

9)(1

.43e

-01)

PO

PU

LAT

ION

DE

NSI

TY

0.00

072∗∗

∗−0

.000

5∗∗0.

0001

1∗∗∗

−0.0

010∗

−0.0

001

0.27

94∗∗

∗

(0.0

0020

)(0

.000

2)(0

.000

04)

(0.0

006)

(0.0

002)

(0.0

971)

LAG

GE

DE

NV

IRO

NM

EN

T0.

8660

∗∗∗

0.79

07∗∗

∗

(0.0

228)

(0.0

832)

Con

stan

t0.

1251

∗2.

0430

∗∗∗

1.39

18∗∗

∗−2

.450

0∗∗∗

3.08

09∗∗

∗−2

1.61

05∗∗

(0.0

650)

(0.1

765)

(0.5

254)

(0.5

281)

(0.2

658)

(9.4

136)

Obs

erva

tion

s38

3310

813

4420

425

510

5R

20.

990.

400.

990.

220.

090.

21

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nta

t10%

;∗∗

sign

ifica

nta

t5%

;∗∗∗

sign

ifica

nta

t1%

.Coe

ffici

ents

for

year

and

cou

ntr

ydu

mm

yva

riab

les

not

show

nfo

rm

odel

sof

CO

2an

dor

gan

icw

ater

pol

luta

nts

.

220

Tabl

e7.

3.E

ffec

tofa

utoc

racy

onen

viro

nmen

tald

egra

dati

on

Org

anic

wat

erA

nn

ual

Fore

star

eaD

egra

ded

area

CO

2N

Ox

pollu

tan

tsde

fore

stat

ion

shar

eof

shar

eof

per

capi

tap

erca

pita

per

km3

rate

lan

dar

eala

nd

area

AU

TO

CR

AC

YD

UM

MY

0.03

63∗∗

∗0.

4211

∗∗∗

0.04

98∗∗

1.54

45∗∗

∗−0

.623

7∗∗0.

3654

†

(0.0

138)

(0.1

727)

(0.0

296)

(0.5

144)

(0.3

113)

(0.2

894)

TR

AD

EO

PE

NN

ESS

−0.0

006

0.00

43∗∗

0.00

050.

0147

∗∗−0

.003

2−0

.380

2∗

(0.0

011)

(0.0

025)

(0.0

004)

(0.0

069)

(0.0

035)

(0.2

736)

WA

R−0

.054

4∗∗∗

9.09

e-02

0.00

700.

6424

∗∗−0

.466

3∗∗∗

0.42

69∗

(0.0

188)

(1.8

1e-0

1)(0

.028

4)(0

.360

7)(0

.197

2)(0

.309

1)R

EA

LG

DP

PC

8.62

e-05

∗∗∗

1.58

e-04

∗∗∗

6.41

e-06

0.00

04∗∗

∗−2

.27e

-06

6.67

09∗∗

∗

(1.7

7e-0

5)(3

.59e

-05)

(9.0

2e-0

6)(0

.000

1)(6

.03e

-05)

(2.2

913)

RE

AL

GD

PP

CSQ

UA

RE

D−2

.80e

-09∗∗

∗−3

.11e

-09∗∗

−4.5

7e-1

0∗−9

.50e

-09∗∗

∗−5

.19e

-10

−4.0

5e-0

1∗∗∗

(5.3

1e-1

0)(1

.48e

-09)

(3.1

2e-1

0)(3

.31e

-09)

(2.3

6e-0

9)(1

.40e

-01)

PO

PU

LAT

ION

DE

NSI

TY

0.00

071∗∗

∗−0

.000

5∗∗0.

0001

2∗∗∗

−0.0

008∗

−0.0

002

0.28

91∗∗

∗

(0.0

0020

)(0

.000

2)(0

.000

04)

(0.0

006)

(0.0

002)

(0.0

967)

LAG

GE

DE

NV

IRO

NM

EN

T0.

8659

∗∗∗

0.79

32∗∗

∗

(0.0

228)

(0.0

840)

Con

stan

t0.

1102

∗1.

6855

∗∗∗

1.35

78∗∗

−3.1

578∗∗

∗3.

4179

∗∗∗

−23.

8747

∗∗

(0.0

640)

(0.1

867)

(0.5

390)

(0.6

532)

(0.3

434)

(9.2

644)

Obs

erva

tion

s38

3310

813

4420

425

510

5R

20.

990.

431.

000.

270.

060.

19

Not

e:St

anda

rder

rors

inpa

ren

thes

es.†

sign

ifica

nt

at10

.5%

;∗si

gnifi

can

tat

10%

;∗∗

sign

ifica

nt

at5%

;∗∗∗

sign

ifica

nt

at1%

.Coe

ffici

ents

for

year

and

cou

ntr

ydu

mm

yva

riab

les

not

show

nfo

rm

odel

sof

CO

2an

dor

gan

icw

ater

pol

luta

nts

.

221


exhibits significant effects on the five dimensions of environmental degra-dation. Tables 7.2 and 7.3, however, indicate that the effect of democracyis not monotonic across various segments of the continuous democracymeasure. For CO2, NOx, and organic pollution in water, the significanteffects of the continuous measures are to some extent driven by the differ-ence between autocratic and nonautocratic regimes. For land degradation,the effect of the continuous democracy measure is driven by the differencebetween democracy and nondemocracy. For forest depletion (deforestationand forested area), the effects of democracy appear to be monotonic, as theyare significant regardless of how regime type is measured.

Finally, although it is useful to study the effect of democracy on differ-ent dimensions of environmental degradation separately, one may wonderhow democracy affects some composite environmental indicator. Recently,the Environmental Performance Measurement Project (EPMP) (2002) hasembarked on a large-scale effort to construct composite environmentalmeasures of several types, which we describe in the appendix. It is worthreiterating that these composites lump together different environmentalindicators, just as gross domestic product does for the economy. But theenvironmental story is much more complex than the economic story. Forone thing, whereas money value provides a common metric to aggregateeconomic activities, environmental forces do not have a similar, readilyavailable common metric (i.e., they are each measured in specific physical,chemical, or geographical units). Thus, one should interpret the compositemeasure-based results with caution.

With this in mind, we employ two core composite contributors to theEnvironmental Sustainability Index (ESI), put together by the EPMP. Theenvironmental systems quality composite aggregates measures in the areas ofair quality, water quantity, water quality, biodiversity, and terrestrial degra-dation. The reducing environmental stresses composite aggregates measuresin the areas of reducing air pollution, reducing water stress, reducing ecosys-tem stress, and reducing waste, consumption, and population pressures onthe environment. The data come from EPMP (2002) for the year 2002.

Table 7.4 includes six columns. The first three columns focus on theenvironmental systems quality composite. The next three columns concernthe reducing environmental stress composite. In each case, we report resultsusing the three measures of political regime type, level of democracy, andthe two dichotomous indicators of democracy and autocracy.

Limitations notwithstanding, the results in Table 7.4 are in the spirit ofthose reported in Tables 7.1, 7.2, and 7.3. Table 7.4 shows that the effects ofthe level of democracy on both the environmental systems quality composite

Tabl

e7.

4.E

ffec

tofp

olit

ical

regi

me

type

onen

viro

nmen

talc

ompo

site

s

Red

uci

ng

Red

uci

ng

Red

uci

ng

Env

iron

men

tal

Env

iron

men

tal

Env

iron

men

tal

envi

ron

men

tal

envi

ron

men

tal

envi

ron

men

tal

syst

ems

qual

ity

syst

ems

qual

ity

syst

ems

qual

ity

stre

ssst

ress

stre

ss

LEV

EL

OF

DE

MO

CR

AC

Y0.

1474

0.15

30(0

.275

3)(0

.151

8)D

EM

OC

RA

CY

DU

MM

Y2.

5670

3.10

22∗∗

(2.9

863)

(1.8

368)

AU

TO

CR

AC

YD

UM

MY

−6.8

081∗∗

−2.4

238

(3.4

276)

(2.6

298)

TR

AD

EO

PE

NN

ESS

−0.0

226

−0.0

212

−0.0

286

−0.0

304

−0.0

287

−0.0

327

(0.0

324)

(0.0

324)

(0.0

325)

(0.0

265)

(0.0

259)

(0.0

269)

WA

R1.

0001

0.98

330.

6373

0.78

100.

8037

0.48

03(4

.554

6)(4

.460

3)(4

.480

5)(2

.725

8)(2

.639

4)(2

.804

1)R

EA

LG

DP

PC

0.00

040.

0003

0.00

04−0

.000

0−0

.000

20.

0001

(0.0

006)

(0.0

006)

(0.0

005)

(0.0

004)

(0.0

004)

(0.0

004)

RE

AL

GD

PP

CSQ

UA

RE

D−3

.23e

-10

2.54

e-09

−2.6

3e-0

9−3

.04e

-08∗∗

−2.6

5e-0

8∗∗−3

.29e

-08∗∗

∗

(1.9

3e-0

8)(2

.02e

-08)

(1.8

1e-0

8)(1

.44e

-08)

(1.4

4e-0

8)(1

.39e

-08)

PO

PU

LAT

ION

DE

NSI

TY

−0.0

511∗∗

∗−0

.051

8∗∗∗

−0.0

507∗∗

∗−0

.026

0∗−0

.026

9∗−0

.025

4∗

(0.0

208)

(0.0

205)

(0.0

203)

(0.0

174)

(0.0

171)

(0.0

172)

Con

stan

t53

.976

8∗∗∗

53.3

760∗∗

∗55

.290

9∗∗∗

61.0

057∗∗

∗60

.256

1∗∗∗

61.5

675∗∗

∗

(2.8

579)

(2.8

392)

(2.9

870)

(2.5

402)

(2.5

593)

(2.6

550)

Obs

erva

tion

s10

810

810

810

810

810

8R

20.

280.

280.

290.

450.

460.

45

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗

∗si

gnifi

can

tat

5%;∗

∗∗si

gnifi

can

tat

1%.

223


and the reducing environmental stress composite are positive, as in Table7.1, but they are weak or not statistically different from zero. A transitionfrom nondemocracy to democracy raises environmental system quality, butthe effect is insignificant. On the other hand, this transition raises the reduc-ing environmental stress composite, and the effect is statistically significant.Hence, democratic transition is good for reducing the environmental stresscomposite. A transition from nonautocracy to autocracy reduces environ-mental systems quality, and the effect is statistically significant – implying,once again, that democracy is good for the environment. This transitionalso lowers the reducing environmental stress composite, but this particulareffect is weak.


Our research in this chapter shows that scholars debate the nature of theeffect of democracy on the environment. Existing evidence is mixed andrelatively scant. Our research in this chapter contributes to the democracy–environment literature by empirically testing the net effect of democracyon environmental degradation. The empirical scope of our data analysis interms of sample size and the number of indicators is generally wider thanin previous studies. We employ a wide array of important types of humanactivities that degrade the environment: CO2 emissions, NOx emissions,organic pollution in water, forest depletion in terms of the rate of deforesta-tion and the share of forested area over land area, land degradation, and twocomposite environmental measures. For democracy, we employ a continu-ous measure of the level of democracy or autocracy and two dichotomousmeasures of democracy and autocracy.

The results presented in this chapter have several important implicationsfor public policy. Democracy is generally good for the environment and, assuch, policies that expand democracy could reduce human activities thatlead to environmental degradation. But this conclusion should be quali-fied in terms of which regime transitions are likely to cause what type ofimprovement. A transition from nondemocracy to democracy (across the+6 threshold on the level-of-democracy scale) does not significantly influ-ence CO2 emissions per capita, NOx emissions per capita, or organic pollu-tants in water, but it lowers land degradation and yet raises the deforestationrates. In contrast, a transition from an autocracy to a nonautocracy (acrossthe −6 threshold on the level-of-democracy scale) reduces CO2 emissionsper capita, NOx emissions per capita, organic pollutants in water, and,somewhat less so, land degradation, but it also raises the deforestation rate.


These findings suggest that moving upward along the level-of-democracyscale always leads to faster deforestation but not always to a significant envi-ronmental improvement in terms of the other four measures; the effectdepends on where a country is along the level-of-democracy scale. In fact,one could expect greater improvements of the environment when a transi-tion occurs across the −6 threshold than when a transition occurs across the+6 threshold. In other words, increasing the democracy score of a highlyautocratic country above the −6 threshold delivers more bang for the buckthan increasing the democracy score of a less autocratic country above the+6 threshold.

The findings concerning the size of effect also have important implica-tions. Whereas, in all cases, a rise in democracy produces a noticeable effecton environmental degradation, the effect of the level of democracy on envi-ronmental degradation varies across environmental dimensions. We couldexpect considerable effects of political regime type in terms of deforesta-tion rate, forested area, NOx emissions per capita, and land degradation.Furthermore, timing is also of the essence here; one should not expect tosee the effects of a rise in democracy kick in full-force very quickly. Theimmediate (annual) effects of a rise in democracy on organic pollutants inwater and CO2 emissions per capita appear to be small, but the cumulativebeneficial effects of this rise in democracy are much larger over time. Thatsaid, these two effects are still smaller than the effects of democracy onNOx, deforestation rate, forested area, and land degradation. Hence, a risein democracy reduces some types of environmental degradation more thanother types and results in faster deforestation rate.

Our results in this chapter also suggest that democratization could indi-rectly promote environmental degradation through its effect on nationalincome. This effect is subtle and works through the environmental Kuznetscurve discussed in the appendix of this chapter. Across the five aspects ofenvironmental degradation, we find evidence supporting the existence ofan environmental Kuznets curve for CO2 emissions per capita, NOx emis-sions per capita, and the level of land degradation. Hence, when incomeper capita is low, a rise in income per capita causes more degradation;once past a certain threshold, a rise in income per capita reduces degrada-tion. Although existing evidence on the effect of democracy on economicgrowth is inconclusive – to the extent that a rise in democracy promoteseconomic growth – the environmental Kuznets curve effects we find suggestthat democracy could indirectly cause more environmental degradation forthe aforementioned indicators at the initial stage of development but couldhelp to reduce the degradation with further development.


Finally, the reader may recall the examples with which we opened thischapter, referring to the works and evaluations of former U.S. Vice PresidentAl Gore and the popular writer Jared Diamond. By and large, the insightof Al Gore that democracy benefits the environment is generally more onthe mark than that of Jared Diamond. The latter suggests in one placethat autocracy has its advantages when it comes to environmental qualityand in another place argues that the effect of political regime type on theenvironment cannot be generalized.

In light of our research, a caveat is in place for the arguments of both fig-ures. Our analysis reveals the danger of making general statements regardingcomplex transformation using stylized facts – which is not to suggest that thestylized observations invoked by Al Gore and Jared Diamond are incorrect.But when it comes to public policy formulation, the more important issue iswhether the anecdotal evidence translates into generalizable patterns. Ourpoint is that stylized facts do not justify general claims.

Al Gore is correct in observing that democracy is good for the envi-ronment, but he fails to identify that a rise in democracy leads to fasterdeforestation and that the effects of democracy vary in size across environ-mental indicators. These failures should not be taken lightly. For one thing,continued faster deforestation in democracy implies that Al Gore’s expec-tation that a rise in democracy would resolve our environmental problemsmay turn out to be a false prediction in the long run. Faster deforestationmay elicit a rise in carbon dioxide emissions (because diminishing forestsabsorb less emissions), a decline in water resources due to global warm-ing, and intensified land degradation and higher water pollution due toa decline in water resources. Furthermore, failing to recognize the vary-ing effects of democracy, policymakers will not realize that the beneficialeffects of democracy on the environment will be small and slow in terms ofreducing water pollution and CO2 emissions.

At the same time, Jared Diamond’s insights for deforestation receivesupport in our large-N sample: democracy induces faster deforestation. Butour statistical findings suggest that overall Jared Diamond is “more wrong”than Al Gore. Whereas Diamond seems to signal that autocracy may not bea bad thing for the environment, his conclusion based on a few cases anddeforestation per se does not hold for all environmental indicators, acrossmany countries and over many years.

Diamond’s second argument, that the effect of democracy on the envi-ronment cannot be generalized and that both democracy and autocracycan end up damaging the environment, is also wrong. Our results for car-bon dioxide, nitrate dioxide, land degradation, pollution in water, and the


aggregate environmental indicators suggest that a rise in democracy reducesenvironmental damage. Hence, if all the countries in the world turned intodemocracies tomorrow, the expectation is that, all else being equal, carbondioxide emissions per capita, for example, should decline as a result.

This example, which assumes ceteris paribus conditions, however, alsodemonstrates the danger of employing the type of stylized analysis of AlGore and Jared Diamond. Both thinkers failed to include the channelfrom democracy to the environment that works through the environmentalKuznets curve. This approach is typical of relying on stylized observations.Stylized analysis can neither sort out nor estimate the effects of variouscausal determinants. The statistical approach not only estimates the ceterisparibus effects of a large number of causal factors but also provides anopportunity of forecasting the near future using the estimated coefficients.

SUMMARY AND OUTLOOK

This chapter evaluated the controversial effects of democracy on the envi-ronment. We found that a higher level of democracy leads to fewer CO2

emissions per capita, fewer NOx emissions per capita, less organic pollutionin water, and less land degradation. Compared with nondemocratic coun-tries, democracies have larger forested areas, but, going against the spirit ofthe preceding findings, they deplete their forests relatively faster.

We also find that the effect of democracy on the environment is discontin-uous along the continuous scale of democracy. Relative to nondemocraticcountries, democratic regimes do not appear to be different in terms ofCO2 emissions, NOx emissions, and pollution in water, but they have lowerrates of land degradation and yet higher deforestation rates. In contrast,relative to nonautocracies, the autocratic regime experiences more CO2

emissions, greater NOx emissions, higher pollution in water, somewhat lessland degradation, and yet lower deforestation rates. Hence, whereas theeffects of democracy on the deforestation rate and the forested area aremonotonic along the democracy scale, the difference between autocracyand nonautocracy significantly influences CO2 emissions, NOx emissions,and organic pollution in water, and the difference between democracy andnondemocracy affects land degradation.

So far in Part III of our book, we have focused on the effect of democracyon human activities that hurt the environment, treating economic open-ness as a control variable. In the next chapter, we shift the focus of analysisfrom the level of democracy of a country to its level of economic open-ness and investigate its effect on the environment. Economic openness, as


we saw in Chapter 3, is a multidimensional concept that involves variousaspects, including trade, foreign direct investments, portfolio investments,and so on. One aspect of economic openness that has attracted the mostattention is the effect of international trade flows on the environment. Theeffect of trade on the environment has been controversial and unsettled.The subject also has important policy and international implications sincetrade is the only aspect of economic globalization that has led to the mostinternational collaboration and the most aggressive development of inter-national institutions – the formation of the General Agreement on Tradeand Tariffs, followed by its successor regime, the World Trade Organization(WTO). Whether or not the WTO should consider environmental issuesin the global trade talks and in the adjudication of trade disputes is highlydebated. It is against the background of this debate that we turn our atten-tion to the effect of international trade flows on the environment in the nextchapter.

APPENDIX


Empirical Model

The dependent variable environment is a multifaceted concept. We focuson greenhouse gas emissions, air pollution, water pollution, forest depletion,and land degradation. We analyze separately one representative, salientindicator for each area, with the exception of forest depletion, for whichwe use two measures – the rate of deforestation and the share of totalland area that is forested. Although these indicators are not exhaustive, theycollectively provide a comprehensive picture of environmental degradation.They also have relatively more comprehensive data coverage.

We conduct six different empirical tests. Each test employs several mea-sures of democracy to be discussed later. The first empirical test focuses oncarbon dioxide (CO2) emissions because this gas constitutes about 60% ofall greenhouse emissions, generated by energy-related activities and sourcesincluding industry, burning of solid fuels (e.g., coal), liquid fuels (e.g.,petroleum), gaseous fuels (e.g., natural gas), gas flaring (burning of gasreleased in petroleum extraction), cement manufacturing, and bunker fuels(stored fuels). We use per capita CO2 emissions to adjust for country-sizedifferences. Data are from the World Development Indicators (World Bank,2002).


The second test analyzes nitrogen oxide (NOx) because this gas is a majorhealth hazard and an important source of smog in urban areas. The NOx

emissions are generated primarily from fossil fuel combustion in motorvehicle engines, and also from other processes including biofuel combustion,oil and gas production, solvent use in industry and other sectors, and wasteburning. We employ the logged NOx emissions per capita. Data come fromthe GEO Data Portal (UNEP, 2006) maintained by the United NationsEnvironment Programme.

The third test examines a major form of water pollution. The level oforganic pollution in water is measured in terms of the amount of oxygenin kilograms, which bacteria living in the water consumes when breakingdown organic matter (typically denoted as biochemical oxygen demand, orBOD). This standard measure is comparable across countries. Our measureis the logged ratio of BOD to the amount of internal renewable waterresources in a country. Data on organic pollution in water come from theWorld Development Indicators (World Bank, 2002), and data on internalrenewable water resources in a country come from the World ResourcesInstitute (2001).

The fourth and fifth tests focus on forest depletion because forests areimportant drivers of ecosystem health. Moreover, deforestation has beenidentified as one of the primary agents of climate change (IPCC, 2007a,2007b). Forest depletion is measured here in two complementary ways.One measure is the rate of deforestation, where a positive value of thisvariable indicates a rise in forested area over time (afforestation) and a neg-ative value implies a fall in forested area over time (deforestation). Availabledata measure deforestation in terms of permanent manmade conversion ofnatural forests into other uses (e.g., mining, ranching, agriculture). Areaslogged with the intent of regeneration and areas degraded by acid rain andforest fires are not included. Data on average annual deforestation ratesper decade are collected from the World Resources Institute (1999) andthe State of the World’s Forest Report (2001). A second measure of forestdepletion is a country’s share of total land area that is forested. Althoughthe deforestation rate allows us to evaluate the effect of democracy on thechange in forested area, this change does not necessarily reflect the size offorested area in a country.2 Hence, we also assess the effect of democracy onthe size of the forested area. To that end, we employ the logged percentageshare of the forested area in total land area. Data come from the Food andAgriculture Organization (FAO, 2000, 2002, 2006).

2 For example, a country that has destroyed a lot of forest area before the sample period may havea low deforestation rate in the sample period but may have only a small forested area.


The sixth test looks at severe land degradation, which greatly harms agri-cultural activities. Data come from the FAO (2000, 2002, 2006). As discussedin FAO (2000), the only available uniform global source of human-inducedland degradation data comes from this report. These data measure landdegradation from water erosion (e.g., loss of topsoil, deformation, sedi-mentation), wind erosion (e.g., loss of topsoil, deformation, overblowing),chemical deterioration (e.g., loss of nutrients, salinization, acidification),and physical deterioration (e.g., compaction, crusting, lowered water table).Data are from the late 1980s and are reported as the share of degraded landout of total land at five levels: not degraded, light degradation (reducedagricultural suitability), moderate degradation (greatly reduced productiv-ity), severe degradation (largely destroyed, unreclaimable at farm level), andvery severe degradation (fully destroyed biotic functions, unreclaimable).Our indicator is the logged share of severely and very severely degraded landout of total land area.

The key independent variable is political regime type. We employ severalmeasures for this concept: one continuous, level of democracy, and theother two dichotomous, democracy dummy and autocracy dummy.The continuous or interval measure of regime type is the widely usedcomposite indicator of the level of democracy, ranging from −10 to +10,from the POLITY data set that we used in the previous chapters. Despitethe popularity of this measure of democracy, one may question whetherthe POLITY score needs to be treated as interval or dichotomous. One mayalso question whether the effect is constant across a range of values alongthe scale. For example, the effect when the democracy score rises from−10 to −5 may not be the same as the effect when the score rises from0 to +5.

In light of these tensions over the measurement of democracy, some schol-ars (e.g., Dixon, 1994; Fearon and Laitin, 2003) often employ dichotomousmeasures of democracy and autocracy that are coded based on a continu-ous indicator. In these studies, a country is often defined as a democracy ifthe continuous measure of democracy is greater than or equal to 6 and asan autocracy if its score is smaller than or equal to −6. The dichotomousvariables democracy and autocracy are coded 1 if a country is democraticor autocratic and zero otherwise.

The chosen threshold of 6 is, of course, to some extent arbitrary inthe sense that 7 or 5 could also be used (but probably not −7 or −1,etc.). Nevertheless, the threshold of 6 has been used in many studies. Tomaintain compatibility with this norm, we use the threshold level of 6. Theconsiderations guiding our choice of the threshold level of −6 are similar.


The control variables we include are those used in a number of empiricalstudies in environmental economics and environmental politics. The firsttwo variables are real income per capita, real gdppc, and its squared term,real gdppc squared. A large literature argues that economic growth hascompeting effects on environmental quality. A larger economy generatesmore output and, therefore, more pollution and waste. Some types of tech-nological progress, which are associated with growth, also may damage theenvironment (e.g., greenhouse gases). The effect is typically referred to as thescale effect. As income per capita rises above some threshold, the importanceof environmental quality for people is said to rise, and they begin to employcleaner production techniques and fewer natural resources, thereby increas-ing investment in environmental regulation. This behavior is referred to asthe income effect. The combined operation of the scale and the income effectgenerates an inverted U-shaped figure when environmental degradation isplotted against income per capita.

The inverted U shape is known in environmental economics and environ-mental politics as the environmental Kuznets curve (EKC).3 Its empiricalexistence, however, is debated, an issue to which we later return. Relevantto our analysis, one may also frame the Kuznets curve debate in terms ofwhether it is an economic effect (environmental quality as a luxury good thatis affordable at higher per capita incomes) or a political effect (the emergingmiddle class as a byproduct of industrialization asserts itself politically onissues of air and water quality). Since the economic effect would occur inde-pendent of the level of democracy but the political effect arguably wouldnot, it is important to include in the Kuznets curve analyses both demo-cracy and income per capita on the right-hand side of the statistical model,as we do.

Although the empirical existence of the Kuznets curve is debated, moststudies allow for it in statistical models by including income per capita andits squared term. If the EKC exists, the coefficient of income per capitashould be positive, the coefficient of income per capita squared should benegative, and the coefficient of income per capita should be larger than theabsolute value of the coefficient of income per capita squared. GDP percapita, in purchasing power parity–adjusted, constant 1996 internationaldollars, is from the Penn World Table 6.1 (Heston et al., 2002).4

3 The name EKC is used in the literature, because the original Kuznets curve hypothesizes theexistence of an inverted U curve for income inequality as a function of income per capita. Forreview of these arguments, see Dinda (2004) and Panayotou (2000b).

4 An alternative indicator of technology and knowledge is GDP per worker. But GDP per capitaand GDP per worker are correlated very highly (correlation coefficient = 0.98).


The third control variable is trade openness, which is a country’s totaltrade over its GDP. The trade and environment literature argues that tradecan affect the environment in two broadly defined ways. In one way, thepattern of domestic production and consumption, and the methods of pro-duction, change under trade openness as countries follow their comparativeadvantages and/or adopt certain more efficient, cleaner (or not) technolo-gies to produce for other countries. For example, a country that trades envi-ronmentally clean goods would see its environmental quality rise, and viceversa. This channel also involves changes to environmental regulation, assome international trade treaties may require regulatory changes at home. Inaddition, trade may affect the environment by promoting economic growthand, hence, altering people’s behaviors over time. Although the combinedempirical effect of trade on the environment is debated, it needs to beincluded in the model.5 We employ a popular measure of the importance oftrade openness to a national economy, which is the sum of national exportsand imports divided by GDP. Data are from the Penn World Table 6.1.

The fourth control variable is population density (population dividedby land area).6 The effect of population density on environmental degrada-tion may change across indicators. A rise in population density is expectedto generate more CO2 emissions as a larger population consumes and pro-duces more. But it may generate fewer NOx emissions as denser areas tendto use more public transportation and fewer cars (the primary generator ofthis type of emissions). Water pollution is expected to rise with populationdensity. As more people engage in consumption and production, organicwater pollution should rise. Many densely populated nations tend to bemore urbanized and depend less on the environment for livelihood (e.g.,consider Western European countries such as the Netherlands or Belgium).As such, they may clear fewer forests. Greater population implies morepressure to use agricultural land for food and industry and, therefore, moreland degradation. Data are from the World Development Indicators (WorldBank, 2002).

Finally, as discussed by Reuveny (2002), for example, military conflictalso can affect the environment. Military conflict, however, may generatecompeting effects on environmental degradation. The variable war is coded1 if a country is involved in an interstate or intrastate war in a given period,and 0 otherwise. According to Gleditsch et al. (2002), from which we obtainthe conflict data, a militarized interstate or intrastate conflict is defined as awar if it involves at least 1,000 battle-related deaths per year.

5 For a review of these arguments, see, e.g., OECD (1994b), Pugel (2003), and Harris (2006).6 For examples of studies employing this indicator, see Panayotou (2000b).



First, the choice between cross-sectional analyses or cross-sectional time-series (pooled) analyses is dictated by data availability. As noted in themain text, for the CO2 emissions and water pollution, we have time-seriescross-sectional data. For NOx, deforestation rate, forested land area, andland degradation, we have cross-sectional data. Thus, we conduct pooledanalyses for the first two indicators and cross-sectional analyses for the lastsix indicators.

Given the various data structures in these indicators, we ensure that thetime periods of the right-hand-side variables match the time period of thedependent variable observations. The sample design for other indicators areclearly discussed in the main text, but the sample design for the forest deple-tion variables requires additional clarification. For the rate of deforestation,the sample covers 134 countries. For each country, the data include twoaverage annual deforestation rates, one rate for the 1980s and the other forthe 1990s. The annual rates during a decade are computed based on the totalforest areas for 1980, 1990, and 2000. The right-hand-side variables takeon their decade-average values to represent their values during the decade,capturing the cross-sectional patterns. For the share of total land area that isforested in a country, we have three data points for each country (1980, 1990,2000). Our data source does not specify when exactly a measurement wastaken for a country; we assume that these possible measurement problemsare absorbed into the error term.

The second design issue concerns the possibility of endogeneity for ourright-hand-side variables. Our empirical framework treats level of demo-cracy, democracy dummy, autocracy dummy, real gdppc, real gdppcsquared, population density, war, and trade openness as exogenousvariables. One may argue that environmental degradation can affect thesevariables. For example, land degradation may reduce the exports froman agrarian economy. Or environmental degradation may lead to con-flict, a hotly debated issue in the literature. To mitigate the potential riskof simultaneity, the right-hand-side variables are lagged one year, as isdone in many studies (e.g., Oneal and Russett, 1999a, 1999b, 1999c; Li andReuveny, 2003). This popular practice is sufficient for addressing the possi-bility of endogeneity under the weak exogeneity assumption (Wooldridge,2002).

Third, national structural variables (e.g., climate, education) and globalor local biophysical attributes (e.g., atmospheric integrity, existing damage)may also affect environmental degradation. Because these factors tend tochange slowly, environmental degradation is likely to exhibit inertia, which is


modeled empirically through the inclusion of the lagged dependent variableon the right-hand side for the pooled analysis.

We further guard against the potential risk of missing structural variablesin our pooled analyses by using the two-way fixed-effects estimator. Thisestimator includes separate intercepts for each country and each year. Inaddition, the lagged dependent variable and the fixed-effects country andyear dummies help to control for the fact that different types of environ-mental degradation may have different causal determinants. Unfortunately,due to data availability, we could only do so for CO2 emissions and organicwater pollution, not for NOx emissions, deforestation, forested area, andland degradation.

Including the lagged dependent variable, lagged environment, andusing the two-way fixed-effects estimator, however, do not come withoutcost. It is well known that they soak up the variations in the dependent vari-able that could otherwise be explained by other right-hand-side variables,which should make it harder for us to find statistically significant results.Hence, this approach to empirical modeling can be considered conservative.As a caveat, this approach suggests that statistical results and inferences forCO2 emissions and organic water pollution are likely to be more reliablethan results for the other indicators with fewer data. Future research withmore data for the other indicators is certainly in order.

Fourth, we consider the risks of heteroskedasticity and serial correlation.When error terms are not spherical, the estimated regression coefficientsare consistent, but their standard errors are inefficient and biased. To dealwith the risk of heteroskedasticity, we estimate Huber–White robust stan-dard errors (White, 1980). Both the year dummies and, as Beck and Katz(1995a, 1995b) suggest, the lagged dependent variable capture the temporaldynamics in the pooled data, controlling for possible serial correlation.

Fifth, we need to consider the issue of multicollinearity. This potentialproblem is a cause of concern when the effects of key variables are statis-tically insignificant in models of good fit (high R2). In this case, statisticalinsignificance may be an artifact of multicollinearity that increases the stan-dard errors of the coefficient estimates. We assess the extent of this possibleproblem using the variance inflation factor (VIF) diagnostic.

Sixth is the issue of which statistical significance level to use when oneinterprets the results. In evaluating the effect of democracy on the envi-ronment, we have discussed two types of theories. One type expects thatdemocracy will promote environmental quality, whereas the other typeexpects the opposite. As noted, it is not possible to reject either set of the-ories based a priori on theoretical grounds. Given their assumptions, both


types of theories appear correct concerning the expected sign of the neteffect. In each model, we therefore test the sign of this net effect of democ-racy on the environment against the null hypothesis of no effect – or ratherthe two competing effects are equal in size – by employing a one-tailedtest in reporting the results. Many other studies have used this approach(see, e.g., Oneal and Russett, 1999a; Li and Reuveny, 2003; Reuveny and Li,2003).

In the interpretation of the results, we employ the 10% significance levelin addition to the 1% and 5% levels. We take this approach because someof our samples are relatively small due to limits on data availability. Ourapproach is justifiable also for our pooled models since we include fixedeffects and the lagged depended variables, which soak up the variations inthe dependent variable.

Finally, we need to measure the size of democracy’s effect on the envi-ronment. We discuss the size of the estimated coefficient of democracy inour models, provided that the coefficient is statistically significant. If thecoefficient is statistically insignificant, the size of effect is so small that itis statistically zero. For the size of change in the independent variable (inour case democracy), we follow other studies by using its one-standard-deviation change in the sample. We proceed by first computing a base valuefor the dependent variable, holding all the continuous right-hand-side vari-ables at their sample mean and setting the dichotomous war variable atzero, and then computing a new value for the dependent variable when thecontinuous democracy variable is increased by one standard deviationin the sample. The two values are then compared to each other, and thedifference is expressed in terms of percent change.

However, this practice does not tell the full story when the lagged depen-dent variable is included on the right-hand side in the pooled analysisbecause it only captures the immediate impact of democracy; it does nottake into account the fact that the impact of democracy on the environmentfrom previous periods is absorbed into the effect of the lagged environmen-tal degradation variable, which is also on the right-hand side. Democracyaffects the current environment via its direct effect and continues to affectthe environment in the next period via the lagged environmental degrada-tion. These effects accumulate over time. The long-run impact of a changein democracy produces the following change in environmental quality:[coefficient of democracy/(1 − coefficient of lagged environmental degra-dation)] × (change in democracy). We compute the long-run effect in thepooled analyses of CO2 and water pollution, because the approach is notapplicable to cross-sectional data.


Empirical Findings

Having discussed in the main text the effect of democracy in Table 7.1,we now focus on the control variables. The variable trade openness hasno effect on CO2 emissions per capita, on organic water pollution, or onforested area. But the effect of trade openness on NOx emissions percapita is positive and statistically significant. The effect of this variable isnegative and statistically significant on land degradation, but positive andsignificant on the deforestation rate. Trade is not necessarily a boon for theenvironment, but it appears to reduce the deforestation rate.

The effect of war on CO2 emissions per capita is statistically significantand negative, reflecting the net effect of competing forces. CO2 emissionsmay rise during wars from greater production and operation of weaponry.But they may also fall during wars due to the reduced normal economicactivity that generates emissions as the labor force is drafted and/or partsof the economy are destroyed. This logic applies to all our environmentalindicators. For CO2, the emission-reducing effect of war is larger thanthe increasing effect. The net effects of war on NOx emissions per capitaand organic pollution in water are statistically insignificant. In these cases,the competing effects of war are of about the same size. In contrast, warsignificantly reduces the rate of deforestation and raises land degradation.Compared with countries at peace, countries involved in war in our cross-national analysis for forest depletion have smaller shares of forested area.

The effects of the lagged CO2 emissions per capita and the lagged organicpollution in water on the contemporary values of these indicators, respec-tively, are both positive and statistically significant. Hence, these indicatorsexhibit inertia.

As noted, population density may also exhibit competing effects on theenvironment. We find that the net effects of population density on CO2

per capita, organic pollution in water, and land degradation are statisticallysignificant and positive. In these cases, a larger population density leads tomore environmental damage. In contrast, the statistically significant effectof population density on NOx emissions per capita is negative. As forforest depletion, more densely populated countries exhibit significantlyhigher deforestation rates, but the effect of population density on the shareof forested area is not statistically significant in our sample.

Our findings for the signs and significance levels of real gdppc and realgdppc squared suggest that CO2 emissions per capita, NOx emissions percapita, and land degradation exhibit an EKC (for land degradation, incomeis logged). Stated in real 1996 international dollars, the turning points are


$16,071 for CO2; $25,951 for NOx; and $4,012 for land degradation. Organicwater pollution and the share of forested land do not exhibit a Kuznets curve.For the rate of deforestation, the coefficient of GDP per capita is positive andsignificant, and the coefficient of GDP per capita squared is negative andsignificant. Hence, deforestation declines as GDP per capita rises and thenrises as GDP per capita rises above some level, which indicates a nonlineareffect that does not follow the shape of the EKC.

Our results for the environmental Kuznets effect are within the rangeof results reported in the extensive literature on this topic. The locationof the turning point and whether the Kuznets curve exists are empiricalissues. Results vary among studies and depend on model specification, data,indicators, and estimators. For example, Cole et al. (1997) found turningpoints of $25,100 for CO2 and $15,100 for NOx; Moomaw and Unruh (1997)found $18,333 for CO2; and de Bruyn et al. (1998) did not find Kuznetscurves for CO2 or NOx. Many studies did not find a Kuznets curve forwater pollution (Panayotou, 2000b). For forests, Shafik and Bandyopadhyay(1992) and Barbier (2001) did not find a Kuznets curve. We did not find anystudies of the Kuznets curve for land degradation. Investigating the sourcesfor differences in the Kuznets curve literature is a very large task and isbeyond the scope of this chapter.

It is apparent that our results support both economic and political effectsdiscussed earlier. However, the income variables largely capture the eco-nomic effect, because the political effect (via the middle class) is controlledfor by the democracy variable. To the extent that Seymour Lipset is correctabout the modernization thesis of democracy, part of democracy’s effect onthe Kuznets curve also traces to economic development, which is present inthe model.

Finally, we should discuss whether multicollinearity affects our resultsin Table 7.1. Multicollinearity becomes a concern when the diagnostic VIFstatistic exceeds the threshold of 10. For the CO2 model in Table 7.1, theaverage VIF is 5.7. Hence, multicollinearity is not a concern. For the NOx

model in Table 7.1, the average VIF is 8.7, suggesting that multicollinearitymay be a concern. Using the matrix of variance decomposition, we find thatthe multicollinearity is caused by the high correlation between real incomeper capita and its squared term. The VIFs for other variables are smaller than2.5, indicating that multicollinearity for them is not a concern. For the modelinvolving organic pollution in water, the average VIF is 4.8, suggesting thatmulticollinearity is not a concern. For the deforestation model, the averageVIF is 4.9, also suggesting that multicollinearity is not a concern. For theland degradation model, the average VIF is 95; hence, multicollinearity is


a concern. However, using the matrix of variance decomposition, we findthat the source of the collinearity is the correlation between real income percapita and its squared term. The VIFs for other variables are all smaller than2, indicating that multicollinearity is not a concern.

Additional Analyses

Tables 7.2 and 7.3 present the results for the dichotomous measures ofdemocracy and autocracy, respectively; these results are fully discussed inthe text. The additional analysis section in the text also discusses the Envi-ronmental Performance Measurement Project’s (EPMP’s) (2002) compositeenvironmental measures.

The project aggregates environmental indicators by computing the aver-age of their z-scores in their respective distribution. The z-score transfor-mation subtracts the variable’s sample mean from its value in the sampleand divides the result by its standard deviation. For details, see Appendix Ain http://www.yale.edu/esi/.

As noted in the main text, we employ two core composite contributorsto the ESI, put together by the EPMP: the Environmental Systems Qualitycomposite and the Reducing Environmental Stresses composite. The ESIaggregates five core composites, of which we use two.7 In addition, EPMPhas released data for 2005. We do not use these data because we do not havedata on other variables for 2005.

The results are reported in Table 7.4. Among the control variables, onlythe effect of population density is consistently significant across the sixcolumns; its sign is negative. This result is not surprising given that thecontrol variables often have different effects on the various dimensions ofenvironmental degradation, as shown in Tables 7.1–7.3 and in the literature,which should caution us against being overly confident about the resultsbased on any composite environmental indicator; the environment, unlikethe economy, cannot be easily aggregated because its various componentslack a single metric.

7 The other three core contributors involve economic, social, and political indicators in areas suchas science and technology, capacity for public debate, private sector responsiveness to environ-mental issues, and governance. Given our focus on human actions that harm the environment,we do not employ these core areas or the ESI.

EIGHT

Economic Openness and the Environment

INTRODUCTION

In Chapter 7, we studied the effect of democracy in a country on vari-ous types of environmental degradation, controlling for international trademerely as a secondary force. This chapter studies how international tradeand its interaction with democratic governance influence the terrestrialenvironment. Specifically we examine land degradation and deforestation,two indicators that reflect the health of the terrestrial environment of a coun-try. These two dimensions of the terrestrial environment have an importantimpact on food production and climate change – two issues that have posedserious challenges to the contemporary world.

Land degradation significantly constrains a country’s ability to providefood for its people and animals, including animals to be consumed. A risein land degradation often reduces the land’s productivity and, therefore,food yields. Because of growing populations and land degradation, manyless developed countries (LDCs) find that their people have to confrontdeclining food security. The problem is intensifying as rising oil prices haveinduced food crises in many poor countries.

Deforestation has important environmental implications, including glo-bal warming: it speeds up land erosion, causes fresh water supplies to decline,reduces the level of biodiversity, raises the levels of sedimentation in riversand lakes, intensifies sand and dust storms, increases the occurrence ofmudslides, and deteriorates air quality. In fact, the Intergovernmental Panelon Climate Change has listed deforestation as a key driver of climate change.Shrinking forests fail to absorb carbon dioxide emissions, the accumulationand trapping of which is recognized as the most significant cause of globalwarming.

239


Many commodities that are traded internationally depend heavily onthe endowment of land and forests in countries. Consequently, interna-tional trade produces important effects on the terrestrial environment. Butthe effect of international trade on the environment is controversial. Inone view, trade reduces environmental degradation and promotes environ-mental quality; in another view, trade harms the environment, increasingenvironmental degradation. The controversy in and of itself suggests thatfurther analysis is warranted and also justifies our focus on one main aspectof globalization to keep our analysis tractable. Meanwhile, the democracy–environment relationship is also debated, as we have demonstrated in thepreceding chapter. In general, the democracy–environment debate andthe trade–environment controversy have produced separate literatures thatremain largely on separate courses.

We argue that the effects of trade and democracy on environmentaldegradation need to be studied together. Whereas both trade and democracyhave controversial effects on the environment, the literature also argues thatinternational trade flows influence democracy, as we will see later. Empiricalstudies that focus on the effect of only trade or democracy on environmentaldegradation, excluding the other, produce incorrect statistical inferences byattributing the effect of one factor to the other. Furthermore, internationaltrade and democratic governance may well interact with each other toamplify or mitigate the other’s impact on the environment – a possibilitythat has not been contemplated in the literature or in our previous chapter.

In this chapter, we are interested in the interactions among trade, demo-cracy, geography, and the environment. We discuss the various mechanismsthrough which trade and democracy influence the environment and howtrade and democracy relate to each other. In our comprehensive empiricalanalysis, we address various possibilities: trade and democracy may interactwith one another, and their effects may vary across development levels,between democracies and autocracies, for countries with low and highlevels of trade, and across geographical regions. A country in Africa, forexample, may be exposed to conditions different from those in the UnitedStates and may exhibit different rates of development and urbanization.

Our key empirical findings suggest that a rise in trade openness reducesthe rate of deforestation in autocracy but increases the rate of deforesta-tion in democracy. These effects are similar between LDCs and developedcountries (DCs). A rise in trade openness reduces land degradation, but theeffect is not robust and does not depend on regime type. A rise in democracyincreases deforestation and reduces land degradation, but these effects are

Economic Openness and the Environment 241

weaker in LDCs than in DCs. In addition, open-economy democracies expe-rience faster deforestation than closed-economy democracies. The effect ofdemocracy on land degradation does not depend on trade openness.

THEORETICAL AND EMPIRICAL LITERATURES

This section summarizes the arguments on the effect of democracy on theenvironment presented in the preceding chapter, discusses the theories of theeffect of trade on environmental degradation, highlights the link betweentrade and democracy, and goes deeper into the environmental Kuznets curvementioned in the Chapter 7 appendix. Although our primary interest lieswith trade and its interaction with democracy, the various topics we surveywill assist us in putting together the statistical model. The relevant empiricalliterature pertaining to our primary interest is also summarized.

Theories on Trade and Environment

International trade influences environmental degradation both directly andindirectly. The direct causal mechanisms involve microlevel economic activi-ties that affect the environment, regardless of whether the economy is grow-ing or not. The indirect causal mechanisms concern the positive effect of in-ternational trade on the economic growth of a country, which in turn affectsthe environment.1 The positive effect of international trade on economicgrowth is established in many studies and will be taken here as a given.2

International trade can directly increase or decrease the level of environ-mental degradation in a country. These direct effects occur as internationaltrade changes the profitability of producing some products at home andshapes the patterns of domestic and foreign consumption and the alloca-tion of resources across borders.

The structural effect of trade on the environment involves changes in thepatterns of domestic production, consumption, investment due to trade,and in the location of agricultural activities. The structural effect of tradecan, for example, reduce the production of chemical-intensive crops, whichpromotes environmental quality. However, this effect also could promote

1 For reviews of the literature on the effect of trade on the environment, see, e.g., the discussionsin OECD (1994b), Panayotou (2000b), Pugel (2003), and Harris (2006). Our discussion buildson and synthesizes these sources.

2 For recent reviews of the body of literature on international trade and economic growth, seePugel (2003) and Salvatore (2004).


activities that hurt the environment, such as increased drainage of wetlandsto satisfy the demands for greater production for a larger internationalmarket.

The composition effect of trade can damage the environment by changingthe composition of the goods produced at home as countries follow theircomparative advantages when they open their economies for internationaltrade. For example, consider a country that produces two goods: a timber-intensive good by harvesting forests and a labor-intensive good that doesnot affect the environment. According to the neoclassical trade theory, if thecountry is highly forested and abundant in timber, it will have comparativeadvantage in the timber-intensive good. Therefore, international trade inour example increases the production of the timber good and promotesdeforestation, which degrades the environment. If the country is labor-abundant, on the other hand, a rise in international trade increases theproduction of the labor-intensive good and reduces the production of thetimber good, which slows down deforestation.

The technology effect of trade can reduce environmental degradationby shifting domestic production to methods that are less environmentallydamaging due to the requirements brought about by international trade.For example, a rise in the foreign demand for goods produced with farm-ing methods that use fewer fertilizers may benefit the environment. Trademay also create more contacts and increase the diffusion of environmen-tally cleaner, new technologies and innovations. However, trade could alsospread the use of “dirtier technologies,” whose use damages the environ-ment. Such technologies could be cheaper to employ and make exportersmore competitive in international markets while remaining legal within thebounds of existing domestic environmental laws. The result would be a fallin environmental quality.

The regulatory effect of trade is expected to encourage the pro-environment policies. Some trade agreements require countries to limitenvironmental damage. Trade openness also may lead to more stringentenvironmental regulation across the board if an influential country thatsets the pace for others is pro-environmental. The approach of this countrycould influence other countries that are interested in selling their productsin the former’s market. Yet this effect also may work in the opposite direc-tion: the influential country may not be pro-environmental in some or allareas. In addition, parochial trade interests may push for a relaxation ofexisting regulation to shift to cheaper production methods that may dam-age the environment but strengthen international competitiveness. Because


other countries may follow this course of action, the result could be a raceto the bottom, increasing environmental damages.

The existing multilateral trade accords are virtually all sector-specific.The Convention on International Trade on Endangered Species (CITES),for example, has regulated trade in endangered species since 1973 (e.g.,Frankel, 2003). The Montreal Protocol on Substances that Deplete theOzone has regulated and phased out the production and trade of most chlo-rofluorocarbons (e.g., Brack, 1996; Parson, 2002). Wider setups are eitherbilateral or regional. Regional and bilateral trade agreements (e.g., the Euro-pean Union, Australia–New Zealand Closer Economic Relationship, NorthAmerican Free Trade Agreement) have required members to harmonizeenvironmental standards and regulate environmental damage due to trade-related activities (e.g., Hufbauer et al., 2000; Crutsinger, 2001; Frankel,2003). Environmental activists have tried to bring their concerns into theWorld Trade Organization, but thus far have failed. The United NationsEnvironmental Programme, which is currently the only truly across-the-board multilateral environmental organization, thus far has not focused ontrade (e.g., Frankel, 2003; Harris, 2006).

Arguments on Democracy and Environment

In Chapter 7 we demonstrated the debate over the effect of democracy on theenvironment; here we summarize the arguments. Beginning with the viewthat democracy may not benefit the environment, the market argument pos-tulates that various businesses such as corporations, investors, banks, andtheir subcontractors normally seek to maximize profits, not environmentalquality. Because these business interests are influential, democratic govern-ments may not necessarily pursue environmental quality. If environmentalconcerns jive with business concerns, the environment will be preserved;otherwise business considerations may trump environmental concerns.

The policy inaction argument considers the inner workings of democraticgovernments, which confront checks and balances and respect the power ofpublic opinion of competing voices. Democratic governments may exhibitpolicy inaction when it comes to the environment and its preservation,because deliberations on environmental issues often are not able to convergeon an outcome acceptable to all parties involved.

The global-commons thesis observes that democracy is a national phe-nomenon, whose effects stop at the border. As such, democracy may not alle-viate global environmental problems, particularly when natural resources


do not have well-developed property rights or can be freely accessed bya collective. Actors left on their own may seek to maximize their interest,ignoring the damage their actions may inflict on the environment. Becauseall actors may behave in this manner, environmental degradation will rise.

The population growth channel argues that population growth in a coun-try tends to increase environmental degradation, all other things beingequal. Because democratic governments respect free choice, including thefreedom of humans to procreate, the spread of democracy may enable ahigher global population growth rate, indirectly promoting environmentaldegradation.

Now, the view that democracy benefits the environment also invokesmultiple arguments. The responsiveness argument is that democracies aregenerally more responsive to the needs of their citizenries than are autoc-racies because of electoral accountability and political competition. Theseneeds also include environmental preferences. If the public comes to valuethe environment, democratic governments are more likely than autocraticones to translate the public preference into pro-environment policies.

The rule of law argument observes that democratic regimes respect therule of law, whereas autocracies often cater to the needs to their support-ing elite, regardless of the law, and are subject to the whims of the auto-crat, clientelism, and extreme rent-seeking. Thus, democratic governmentsare more likely to comply with international and domestic environmentalagreements, regulations, and laws than autocratic leaders.

The freedom-of-information channel suggests that greater freedom forinformation flow characterizes democracies and their market economiescompared to autocracies (with or without market economies). This charac-teristic facilitates the pursuit of environmental groups, raises public aware-ness of environmental problems and possible solutions, and encouragesenvironmental legislation that curtails degradation.

The famines and human life argument suggests that famines lead to envi-ronmental degradation by diverting attention away from environmentalconcerns and pushing people to exploit natural resources excessively forsurvival. Environmental degradation itself can put lives at risk. Becausedemocracies respect human life more than do autocracies, they would movemore swiftly to alleviate famines and life-threatening environmental condi-tions. As a result, their environmental qualities should be higher than thatin autocracies.

The war channel reasons that countries with democratic regimes mayengage in fewer international wars and experience fewer civil wars thancountries with autocratic regimes. Meanwhile, interstate and intrastate wars


often damage the environment. To the extent that both tendencies hold,democracies should have better environmental quality than autocracies.

Environmental Kuznets Curve Effect

In the appendix of Chapter 7, we discussed the basics of the environmen-tal Kuznets curve (EKC) theory. The implication of this argument is thatinternational trade flows may affect the environment through their effecton the rate of economic growth in a country. The argument that tradeis the engine of growth can be traced to British economists Adam Smithand David Ricardo. Since then this argument has become one of the hall-marks of economics. In this argument, nations specialize in or focus onproducing the goods of their comparative advantage, ending up with a sur-plus of these goods. They then export some of this production to othercountries and import in return the goods that other countries can producerelatively more efficiently. Since, in this arrangement, each partner to thetrade uses its factors of production most efficiently, all partners end upwith more output and higher efficiency. Trade then, as noted, is the engineof economic growth, but what is the effect of economic growth on theenvironment?

The economic growth effect of trade on the environment falls under EKCtheory. International trade promotes economic growth, which in turnaffects the level of environmental degradation in a country. The theoryargues that up to some threshold of income per capita, environmentaldamages rise as the economy grows in scale. As income per capita growsabove this level, environmental damages decline. The plot of environmentaldegradation as a function of income per capita is the EKC, which takes theshape of an inverted U. “EKC” is given by analogy with the original Kuznetscurve, which plots income inequality as a function of income per capita andalso takes the shape of an inverted U.3

What drives the inverted U shape of the EKC? EKC theory attributes theinverted U shape to two economic forces. They are known as the scale effectand the income effect, which affect the environment differently. Beginningwith the scale effect, as the economy grows, income per capita grows; butas the economy produces and consumes more goods and services, it alsorequires more natural resources of various types. With the current typeof energy sources, technologies and methods of production, and means

3 For recent reviews and discussions of the principles of the EKC see, e.g., Panayotou (2000a) andDinda (2004).


of transportation, higher levels of production and consumption simplygenerate more pollution, waste, and other forms of environmental degra-dation. The impact of the scale effect on the level of environmental degra-dation in a country is positive (i.e., degradation grows with income percapita).

However, this is not the only postulated effect of economic growth. Asincome per capita rises, human preferences are expected to shift towardconsuming goods and products that generate less environmental damagein consumption, as well as goods whose production techniques generateless environmental damage. This gives rise to the income effect of economicgrowth. Richer people (with higher income per capita) also are argubly morewilling to pay for cleaner goods produced with cleaner technologies and thatgenerate less pollution. The same people are also more able to and arguablymore willing to pay for and cover the consequences of more environmentalprotection. Therefore, the sign of the income effect of economic growth isexpected to be negative.

Putting the income and scale effects to work together, the EKC argumenthypothesizes that as income per capita rises above some threshold – the levelof which is unknown based on theoretical reasoning, but can be estimatedempirically – the income effect of economic growth comes to dominate thescale effect of growth. The combined operation of the two forces gener-ate the inverted U shape of the EKC. Because trade – both domestic andinternational exchanges of goods and services – is said to be the engine ofeconomic growth, it follows that we should be able to observe an EKC inthe presence of international trade flows.

Theoretical Link between Trade and Democracy

This chapter investigates the effect of trade and democracy on environmen-tal degradation, but international trade flows and the level of democracyare theoretically related. Political scientists acknowledge that internationaltrade flows are determined by economic forces, but they argue that politicalfactors also can be important determinants of trade, particularly in deter-mining who trades with whom, in what, and how much. Scholars have longbeen aware that international trade can be used by one country as a toolof foreign policy and statecraft to influence another (see, e.g., Hirschman,1980; Hufbauer, 2005). Stating this argument succinctly, Diaz-Alejandro(1975) writes: “which markets are allowed to operate and how, which areencouraged and which are repressed – these are political decisions, bothnationally and internationally” (p. 214).


If international trade flows can be used to exert power at the serviceof meeting some foreign policy goal, who will trade with whom? Grieco(1988), Gowa (1994), and Gowa and Mansfield (1993), among others,expect states to avoid trade with those they consider their actual or potentialadversaries. In such situations, the concern for relative gains – who gainsmore from trade – may reduce trade flows to trickles, as the side that gainsless worries that the side that gains more may translate its larger economicbenefits into greater military power. During the Cold War, for example, theUnited States regulated its international trade relations with the Soviet blocbased on this logic, and today the argument suggests that the United Statesshould regulate its international trade with China based on similar princi-ples. Indeed, bilateral cooperation and conflict have significantly influencedthe level of bilateral trade flows (Pollins, 1989; Li and Sacko, 2002). Today,because democracies are less likely to engage in wars against each other, aswe discussed and analyzed extensively in Chapter 5, democracies should feelmore secure in their bilateral trade regardless of who gains more. Conse-quently, they should trade more with each other (e.g., Morrow et al., 1998;Russett and Oneal, 2001). Moreover, because democracy tends to respectthe rule of law and private property, businesses should face less risk andhigher economic returns when they trade and invest in democracies (e.g.,Olson, 1993; Clauge et al., 1996; Li and Resnick, 2003; Souva et al., 2008).

At the same time, scholars have argued that international trade opennessitself can affect the level of democracy in a country. In fact, Chapter 3focuses explicitly on this relationship. The theoretical picture we paintedin Chapter 3 is one of a controversy, where scholars debate the effect ofeconomic openness on democracy. On one side, some observers and writersexpect that a rise in international trade should raise the level of democracyin a country because, for example, trade promotes economic growth andtechnological diffusion. The expanding middle class that benefits fromtrade is more educated. Its members push for democracy, which servesthem best with its transparency and preference for the rule of law andstable property rights. On the other side, scholars expect that a rise ininternational trade should have a negative effect on democracy by creatingwinners and losers in a country, intensifying income inequality, weakeningthe state’s ability to govern autonomously when facing external economicforces, and pushing the governments to serve particular economic interestsinstead of the public interest.4

4 See the sources cited in Chapter 3, including Bhagwati (1994), Im (1996), Schmitter (1996),Martin and Schuman (1997), Diamond (1999), and Held et al. (1999).


Summary of Theoretical Forces Affecting the Environment

The causal mechanisms through which trade and democracy influence theenvironment are many and complex. Tables 8.1A–C list these causal mech-anisms and note the sign of each expected effect on the environment. Table8.1A focuses on the effects of international trade, Table 8.1B emphasizes theeffect of democracy, and Table 8.1C clarifies the effect of economic growthon the environment (i.e., the EKC effect). For each mechanism, a negativesign implies less environmental quality or higher environmental degrada-tion, whereas a positive sign indicates increasing environmental quality orreduced environmental degradation.

Empirical Literatures

Thus far we have discussed the arguments behind four distinct but relatedrelationships: trade and environment, democracy and environment, growthand environment, and trade and democracy. These arguments have receivedvarious degrees of empirical attention in the literature. Here we discuss theempirical studies related to three of these four topics and refer the readerto the preceding chapter for the empirical literature on democracy and theenvironment.

Beginning with trade and the environment, the number of studies thatstatistically investigate the effect of trade on environmental degradation ina large sample is relatively small. On the one hand, some evidence existsthat trade may have some small, beneficial effect on the environment. Lucaset al. (1992) studied the toxic intensity of output produced in 80 countriesfrom 1960–1988. They concluded that the growth rate of this toxic inten-sity was lower for economies open for trade that grow rapidly, relative tocomparatively closed economies. Grossman and Krueger (1993) employedcity data from 1977, 1982, and 1988 for emissions of sulfur dioxide (SO2;for 42 countries), fine smoke (for 19 countries), and suspended particulatematter (SPM; for 29 countries). They found that trade openness (mea-sured as ratio of export plus import to gross domestic product [GDP])reduced SO2 emissions. The effect of trade openness on smoke and SPM isnot significant. Suri and Chapman (1998) studied the effect of the ratio ofimport over GDP on energy consumption per capita (used as a proxy forair pollution) for 33 industrialized countries in the period 1971–1990. Theyreported that the effect is negative. Antweiler et al. (2001) employed citydata of SO2 for 40 nations in the 1970s and 1980s. They found that tradeliberalization reduced SO2, but the effect was small. De Soysa and Neumayer


Table 8.1A. Summary of causalmechanisms: Trade and the environment

Expected effect onenvironmental

Channel quality

Structural Positive or negativeComposition Positive or negativeTechnology Positive or negativeRegulatory Positive or negative

Table 8.1B. Summary of causal mechanisms:Democracy and the environment


Channel quality

Market NegativePolicy inaction NegativeGlobal commons NegativePopulation growth NegativeResponsiveness PositiveFreedom of information PositiveRule of law PositiveFamines PositiveHuman life PositiveWar Positive

Table 8.1C. Summary of causal mechanisms:Environmental Kuznets curve


Channel quality

Scale NegativeIncome Positive

(2005) showed that during the 1980–1999 period, trade, foreign directinvestment, and economic freedom increased sustainable development asmeasured by the genuine savings rate at which investment in the total stockof manufactured, human, and natural capital exceeds its depreciation.


On the other hand, some evidence exists that trade may harm the envi-ronment. Barbier (2001) studied the effect of trade on agricultural landexpansion for tropical countries in 1961–1994. He found that agriculturalexport promoted agricultural land expansion, concluding that trade inten-sifies pressure on the environment. Dean (2002) employed data on Chinesewater pollution in 1987–1995. She found that trade liberalization, whichis measured from reductions in black market premiums, promotes waterpollution. The mixed results and relatively small number of studies suggest,as summarized by Panayotou (2000a), that room exists for more empiricalanalysis of the effect of trade on environmental degradation.

Next, the body of empirical literature that investigates the presence ofthe EKC is considerably larger than the literature on the effects of tradeand democracy on the environment. Since the seminal study of Gross-man and Krueger (1993), many studies have attempted to verify or rejectthe presence of the EKC. For recent extensive reviews of these studies, seePanayotou (2000b) and Dinda (2004). The empirical studies use data fromvarious sources for national cross sections, panel data, and pooled data, de-pending largely on the availability of data. Most studies focus on air pollu-tion (particularly SO2, but also carbon monoxide and CO2, nitrous oxides,and SPM). Some studies investigate water quality (organic and nonorganicpollution), a few studies investigate deforestation, and, as far as we know,virtually no study investigates land degradation.

In general, some studies have found that the EKC exists for local airpollutants, primarily SO2, and considerably less so for water quality andother types of air pollution. Global direct indicators such as CO2 andmunicipal waste, and indirect indicators such as traffic volume and energyconsumption, generally do not exhibit an inverted U-shaped pattern. TheEKC results for deforestation are mixed and debated (e.g., Bhattarai andHamming, 2001; Bulte and van Soest, 2001). Finally, EKC studies typi-cally exclude trade, democracy, or both from their model specificationsdespite the empirical evidence that demonstrates the link between tradeand democracy.5

As for the effects of trade and democracy on each other, several empiricalstudies find that democracy promotes trade, whereas the sign of the effectof trade on democracy is debated. In any case, these studies taken togethersuggest that trade and democracy affect each other, and we should there-fore include both in our empirical model. For studies of the effect of trade

5 For other reviews and assessments of the empirical literature on EKC, see Stern (1998), Harbaughet al. (2002), Pugel (2003), and Harris (2006).


openness on the level of democracy in a country, we refer the reader toChapter 3, where we show that trade openness decreases the level of democ-racy in a country.

With respect to the effect of democracy on trade, Morrow et al. (1998)found that joint democracy promotes trade among great powers, andRemmer (1998) found that democracies in South America signed moretrade agreements with each other. Dixon and Moon (1993) reported thatthe United States exports more to democracies than to nondemocracies,while Bliss and Russett (1998) and Russett and Oneal (2001) found thatthis result holds for a larger sample. In contrast, Verdier (1998) concludedthat democracy only promotes trade between DCs, whereas Mansfieldand Bronson (1997) and Gowa and Mansfield (1993) reported that theeffect of democracy on trade was not statistically significant. Souva et al.(2008) showed that joint democracy increases bilateral trade, but the effectlargely works through the effect of democracy on property rights protection.


Similar to the previous chapters, this section first presents our statisticalmodel for the empirical analysis and then discusses several research designissues. Next, the section presents the key results from the empirical analysis.As in the other chapters, the discussion in this section is self-contained anddoes not require any specific statistical expertise. The technical details of thestatistical model, measure construction, and data sources are in the chapterappendix, following the same order of presentation as in the main text. Wehave seen that trade may affect the environment through several channelsthat operate in parallel. Hence, as in the previous chapter, we do not seek toinvestigate the relative importance of each of these channels, but rather westudy their net effect.

Empirical Model

To assess statistically the effect of trade and democracy on the terrestrialenvironment, we specify and estimate the following statistical model. Asin other chapters, we denote variables with small capital letters, and theircoefficients with Greek notations. Each coefficient indicates the effect ofthe independent variable on the dependent variable – the phenomenon weseek to explain. The notation εt denotes the random error not explained bythe statistical model. The variable subscript t indicates the current period.To simplify the presentation, we refer to the variables without the time


subscript t. The specification of the model enables democracy, trade, andtheir interaction to affect the environment:

terrestrial environmentt

= �0 + �1democracyt + �2trade opennesst

+ �3democracyt · trade opennesst + �4gdppct

+ �5gdppc squaredt + �6population densityt + εt. (8.1)

The dependent variable, terrestrial environment, is measured in twoways: the level of land degradation and the rate of deforestation in a coun-try. Among the key independent variables, democracy is dichotomous,taking the value of 1 if a country is democratic and zero if a country isautocratic; trade openness is the share of a country’s total trade in itsnational economy; democracy × trade openness represents the interac-tion between democracy and trade openness, testing whether democracyand trade openness mitigate or amplify each other’s effect on the envi-ronment. For the control variables in the model, gdppc is the real GDPper capita of a country, gdppc squared is the square of the gdppc, andpopulation is the population density of a country.


To implement our statistical model, we need to address several specialdesign issues. Although we discuss the technical details in the appendix, weoffer a brief general overview here. First, whereas the theoretical argumentssuggest that multiple causal mechanisms may work at the same time betweendemocracy and trade on the one hand and environmental degradation onthe other, our statistical model estimates the net effect of these mechanisms.Even though it is useful to understand the role of each mechanism, assessingthe net effect of democracy or trade on the environment is in our viewmore important in terms of both resolving the debates among scholarsand grasping the consequences of policies that intend to expand trade ordemocracy.

Second, our statistical model is distinct from those in extant studies.Previous empirical studies typically did not include both democracy andtrade in one model. One could gain insight by studying the two forcestogether, because they influence not only the environment but also eachother. Furthermore, we also include the interaction between democracyand trade in our model. Thus, in addressing environmental degradation,economically open democracies may differ qualitatively from economically


open autocracies, and open-economy democracies may differ from closed-economy democracies. In democracies, leaders are held accountable to theelectorate and their political supporters and may alienate pro-environmentinterests. In autocracies, the ruling elite that have adopted an open economicpolicy may be less responsive to trade-related environmental concerns,because they do not answer to the public.

Third, the rate-of-deforestation sample covers 134 countries over twodecades (1980s and 1990s). The land degradation sample includes 105 coun-tries during the 1980s. We conduct our analysis in two samples: one for all thecountries and the other for LDCs only. As in Chapter 2, we identify LDCs asthose countries that are not OECD members. We are interested in identifyingwhether the patterns that exist in the all-countries sample also stay robustin LDCs alone. For the all-countries sample, both democracy and tradeexhibit more variations because both developed and developing countriesare included. These variations decline considerably in the LDC-only sam-ple. Also, stylized observation suggests that many LDCs are relatively moredependent on trade than many DCs. The Kuznets hypothesis and stylizedfacts lead us to expect that LDCs should have weaker institutions to monitorthe status of the terrestrial environment and promote its preservation. Wereport the results for the LDC sample in the additional analysis.

Fourth, as environmental conditions vary dramatically across conti-nents, we control for regional heterogeneity by including regional indi-cators: ASIA, SOUTH AMERICA, EUROPE, MIDDLE & AFRICA, SUB-SAHARAN AFRICA, and OCEANIA. They are set to 1 when a country isin Asia, South America, North Africa and the Middle East, Sub-SaharanAfrica, Europe, and Oceania, respectively. North America is used as the ref-erence category, so all other regions in the model are compared with NorthAmerica in their environmental outcomes (i.e., more than or less than theoutcome of being in North America). We report the results for regionalindicators in the additional analysis.

Finally, as usual, we need to ensure that the model’s error term behavesas assumed. We address the related technical issues in the appendix.

Empirical Findings

We first present the results for the all-countries sample and then discussthe results of several additional analyses. Findings for the control vari-ables and additional technical details are in the chapter appendix. Table 8.2presents the results for both deforestation and land degradation. For eachindicator, we report the results of two models: one model tests the separate


Table 8.2. Effects of democracy and trade openness on deforestation and land degradation

Deforestation Land degradation

(1) (2) (1) (2)All countries All countries All countries All countries

DEMOCRACY −0.9327∗∗ −0.8698∗∗ −0.6834∗∗ −0.7539∗∗

(0.4415) (0.4205) (0.3427) (0.3707)TRADE OPENNESS 0.0130∗∗ 0.0189∗∗ −0.0068∗ −0.0025

(0.0070) (0.0094) (0.0051) (0.0058)DEMOCRACY∗OPENNESS −0.0206∗∗ −0.0094

(0.0109) (0.0111)GDPPC 0.0004∗∗∗ 0.0004∗∗∗ 6.1067∗∗∗ 5.8017∗∗

(0.0001) (0.0001) (2.3417) (2.4452)GDPPC SQUARED −8.55e-09∗∗ −9.04e-09∗∗ −0.3652∗∗∗ −0.3476∗∗

(3.58e-09) (3.58e-09) (0.1440) (0.1501)POPULATION DENSITY −0.0009∗ −0.0013∗∗ 0.2798∗∗∗ 0.3082∗∗∗

(0.0006) (0.0007) (0.0969) (0.1076)Constant −1.2740∗∗∗ −1.2346∗∗∗ −22.9935∗∗ −21.7429∗∗

(0.3464) (0.3507) (9.4540) (9.8598)Observations 204 204 105 105R2 0.21 0.23 0.20 0.21

Note: Standard errors in parentheses. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at 1%.

effects of democracy and trade, and the other adds the democracy–tradeinteraction.

Starting with deforestation, the coefficient of democracy in Model 1 isnegative and statistically different from zero. Hence, the democratic countryexperiences faster deforestation than the nondemocratic country, becausea negative value of the deforestation variable indicates deforestation and apositive value indicates afforestation, which is consistent with our findingin the preceding chapter. Now, the coefficient of trade openness in Model1 is positive and statistically different from zero. Economies that are moreopen to international trade actually experience slower deforestation.

The interpretation of Model 2 for the all-countries sample is more com-plicated; details are explained in the appendix. Holding trade openness atits mean (we center this variable around its mean), the effect of democracyon deforestation remains negative and significant. Hence, democracy stillexperiences faster deforestation in a country with average trade opennessthan does autocracy. Now the coefficient of democracy × trade opennessin Model 2 is negative and significantly different from zero, suggesting thatin democracy, rising trade openness should lead to even faster deforestation.


That is, trade openness amplifies the deforestation-increasing tendency indemocracy.

In terms of an autocracy (when democracy = 0), its deforestation ratewould be slower than in a democracy, and trade openness does not increasethe rate of deforestation in an autocracy. In fact, because the coefficientof trade openness is positive and significantly different from zero, risingtrade openness in an autocracy increases its rate of afforestation or, in otherwords, reduces its rate of deforestation.

On average, democracy has faster deforestation than autocracy, and tradeopenness speeds up deforestation in a democracy but reduces it in anautocracy.

Next we turn to the results for the all-countries sample for land degra-dation. In Model 1 for land degradation, the coefficient of democracyis negative and significantly different from zero. Hence, democracy has alower level of land degradation than autocracy. The coefficient of tradeopenness in Model 1 is negative and significantly different from zero. Arise in trade openness reduces the level of land degradation in a country.

Is there an interactive effect between democracy and trade for land degra-dation? Model 2 for land degradation answers this question. When tradeopenness is held at its sample mean level, democracy still has a lower level ofland degradation than autocracy. But the coefficient of democracy×tradeopenness is statistically not different from zero, suggesting that the demo-cratic country does not have varying levels of land degradation when itslevel of trade openness is raised. In addition, the coefficient of trade open-ness is not statistically different from zero. Taken together, trade opennessdoes not affect land degradation in a country, and its effect does not dependon regime type.

Additional Analyses

We conduct two types of additional analysis. First, we add regional indicatorsto the models of Table 8.2 for the all-countries sample. Second, we reesti-mate all models for the LDC sample. The results for deforestation and landdegradation are reported in Tables 8.A1 and 8.A2, respectively. Model detailsare introduced in the chapter appendix. Here we provide a summary of thekey results.

With deforestation in Table 8.A1, once we control for regional hetero-geneity, democracy no longer has a faster deforestation rate than autocracy,but rising trade openness in democracy still leads to significantly higherdeforestation rates. As we look at the countries around the world today,


newly democratizing countries also tend to embrace greater trade open-ness. We likely expect to see faster deforestation in these countries. As fortrade openness, the results are similar to those presented in Table 8.2 evenafter we control for regional differences. Open-economy autocracies expe-rience smaller deforestation rates than closed-economy autocracies, butopen-economy democracies witness higher deforestation rates than closed-economy democracies. Controlling for regional differences does not fun-damentally change the results, despite the regional indicators that suggestregional differences do exist (see the appendix).

For the LDC sample in Table 8.A1, the results for democracy are gener-ally similar to those for the all-countries sample. Democracy is associatedwith faster deforestation than autocracy in the first two models for LDCs,but the effect of democracy vanishes once we control for regional differ-ences. But democracies with more trade openness continue to experiencehigher deforestation rates than those with less trade openness, a result thatremains robust in the LDC sample. The effect of trade openness remainsthe same in the LDC sample as in the all-countries sample. Trade opennessreduces deforestation in autocracies but raises it in democracies.

Table 8.A2 reports the results of six models for land degradation. Demo-cracies have lower levels of land degradation than autocracies for the all-country sample and when regional differences are controlled for. In themodels for LDCs without the regional indicators, democracies still experi-ence less land degradation, but the effect is not significantly different fromzero when controlling for regional differences. Also as in Table 8.2, theeffect of democracy on land degradation does not depend on the level oftrade openness, regardless of whether it is the all-countries sample plus theregional indicators or across the LDC models.

The effect of trade openness on land degradation is not significantlydifferent from zero when we control for regional differences for the all-countries sample or across all the models for the LDC sample. Changesin trade openness generally do not affect land degradation in LDCs andthe significant effect of trade openness in Model 1 for the all-countriessample in Table 8.2 seems to be driven by DCs or without controlling forregional heterogeneity. This outcome is intuitive if we recall that trade flowsin DCs tend to be intra-industry, which generally do not degrade land. Incontrast, LDCs tend to export primary commodities, minerals, and minedresources, the production of which tends to degrade land. The effect oftrade openness on land degradation in three out of four models for theLDCs is indeed positive. A rise in trade might increase land degradation,but the effect is not large, probably due to the relatively smaller size of trade


flows in LDCs to begin with. Finally, the effect of trade openness does notappear to be contingent on the nature of a country’s political regime type.


In this chapter, we study the controversial effect of trade openness on theterrestrial environment and further analyze whether the effect of trade onthe environment differs between democracy and autocracy. Our findingsdemonstrate the complex interactions among trade, democracy, and theenvironment. A rise in trade openness reduces deforestation in autocracyand increases deforestation in democracy, and the effect is similar for LDCsand DCs. A rise in trade openness reduces land degradation, but the effectis not robust and does not depend on regime type. Now, a rise in democ-racy increases deforestation and reduces land degradation, but these effectsare weaker in LDCs than in DCs. In addition, the effect of democracy ondeforestation is stronger when trade openness is high. The effect of democ-racy on land degradation does not depend on trade openness.

What are the broader implications for theory and public policy? One thingcomes across clearly; that is, the effects of both trade and democracy on theterrestrial environment are not clear-cut. It appears that trade producessome beneficial effect on the quality of land, but the effect is generallyweak. And although trade reduces deforestation, the beneficial effect tendsto occur in autocracies, not in democracies. In fact, democracy increasesdeforestation. In contrast, democracy seems to reduce the level of landdegradation.

Our results suggest that the spread of democracy has a mixed effect onthe terrestrial environment: it is good for land degradation but bad fordeforestation. The evidence for land degradation is consistent with thenotion that democratic governments have to care for the physical envi-ronment in which their citizens reside and that institutional checks andbalances in democracy empower a broad range of public interests, includ-ing pro-environment groups. The evidence for deforestation, in contrast,is consistent with the notion that democratic regimes may ignore negativeeffects on the environment if strong constituencies can gain from it. In thecase of deforestation, these interests often include logging firms, farms gain-ing cleared land, urbanized areas facing population pressures alleviated byclearing forested land, and landless people seeking to gain land by clearingforests.

At the same time, the effect of international trade on the terrestrialenvironment is mixed. Trade appears to produce some beneficial effect on


the quality of land, but the effect is generally weak. And trade leads toopposite effects on deforestation between democracy and autocracy. Theopening up of the Philippines or Brazil for trade led to increased log-ging, which in the case of the Philippines led to large-scale deforestationof hillsides. Hence, economic reforms may produce unintended, malignantconsequences on the environment.

The public policy problem is potentially serious for LDCs, because theyface pressures from within and outside, which push them to both democra-tize and liberalize their economies. Thus, DCs push LDCs to democratize,and international organizations run by DCs push LDCs to liberalize theireconomies and remove their trade barriers. Facing these pressures, LDCsare neither well equipped in terms of environmental regulatory expertise,nor do they have the capabilities and incentives to deal with the negativeexternalities of these policies. The practical effect for the environment maynot be beneficial, which underscores the need for a careful and fully moni-tored transition.

SUMMARY AND OUTLOOK

In this chapter, we studied theoretically and empirically the effect of inter-national trade openness on the quality of the terrestrial environment, inthe context of countries with either democratic or autocratic regimes. Ourgoal is to better understand the interactions among trade, democracy, andthe environment by linking together literature and debates that have largelybeen separate from each other.

We find that a rise in trade openness reduces deforestation in an autocracyand increases deforestation in a democracy, and the effect is similar forLDCs and DCs. A rise in trade openness reduces land degradation, butthe effect is not robust and does not depend on regime type. A rise indemocracy increases deforestation and reduces land degradation, but theseeffects are weaker in LDCs than in DCs. In addition, the effect of democracyon deforestation is stronger when trade openness is high. The effect ofdemocracy on land degradation does not depend on trade openness.

In this and previous chapters, we have studied the effects of democracyand trade on the environment. In these analyses, an indicator of militaryconflict served as one of the control variables. The next chapter shifts theanalytical focus to conflict. The role of the environment in international andintrastate violent conflict has received ample scrutiny in the literature. Butthe effect of violent conflict on the environment itself has received muchless attention and has not been studied statistically in a large-N sample.


The channel from violent conflict to the environment stands at the centerof the next chapter, which also provides the last empirical analysis in thisbook. We focus particularly on two aspects of the environment: the levelof CO2 emissions and the rate of deforestation. These two environmentalaspects are particularly interesting to study because they play a major role inclimate change. In our next chapter, we turn to something interesting abouthow violent interstate and intrastate conflicts influence human activitiesthat cause climate change.

APPENDIX


Empirical Model

The data sources and measure construction for deforestation, land degra-dation, and the level of democracy are discussed in the appendix ofChapter 7. The democracy variable used in the statistical model is coded 1if a country’s POLITY score is larger than 6 and is coded 0 otherwise. Wemeasure the effect of trade on the environment by the sum of the values ofexport and import of a country divided by its GDP. Data are from the PennWorld Table 6.1 (Heston et al., 2002). We also construct an interactive termbetween democracy and openness to test their conditional effects on thedependent variables. We center the trade variable for ease of interpretation.

We control for the indirect effect of trade on the environment throughits effect on income per capita and the EKC hypothesis. To that effect, weinclude (on the right-hand side) the GDP per capita of a country and thesquare of the GDP per capita. If the EKC holds, the coefficient of GDP percapita should be positive, and the coefficient of GDP per capita squaredshould be negative. Data on real GDP per capita expressed in purchasingpower parity–adjusted, constant 1996 international dollars come from PennWorld Table 6.1 (Heston et al., 2002).

Population pressures affect environmental degradation. To capture thiseffect we use population density (total population divided by land area).Previous studies (see, e.g., Panayotou, 2000b) have shown that populationdensity generally increases the level of environmental degradation. Thesedata come from the World Development Indicators (World Bank, 2002).

Terrestrial environmental degradation may vary across continents. Acountry in Africa, for example, may be exposed to conditions different fromthose in the United States and may exhibit different rates of development


and urbanization (although this variation is controlled for by GDP percapita). To control for regional variations, we include regional dummyvariables in the additional analysis models, set to 1 when a country is inAsia, South America, North Africa and the Middle East, the rest of Africa,Europe, and Oceania, respectively, with North America being the referencecategory. It is worth noting that these dummies are not theoretical and oftentend to absorb much of the variations in the dependent variable that canbe otherwise attributed to substantive variables. Hence, results from themodels including these dummies need to be interpreted with caution.


One empirical design issue we need to consider is the risk of heteroskedasticerror variance. We address this risk by estimating Huber–White robuststandard errors, which are consistent in the presence of heteroskedasticity(White, 1980). Serial correlation is not a concern in our particular case,because we only have cross-sectional data, not time-series data, for bothenvironmental indicators.

For the interpretation of the interactive term democracy × trade open-ness, recall that trade openness is centered around its mean. The coef-ficient of democracy measures the effect of a rise in democracy from 0to 1, holding trade openness at its mean (for which the interaction termdemocracy × trade openness equals 0 as, again, trade openness iscentered around its mean).

Next, note that the coefficient of trade openness measures the effectof raising trade openness by one unit in autocracies (when democracyand the interaction term democracy × trade openness are both equal tozero). Hence, the coefficient of the interaction term democracy × tradeopenness has two interpretations: it indicates the additional effect in ademocracy of raising trade openness (for in an autocracy the interactionterm equals zero) or it indicates the additional effect of raising democracyfrom 0 to 1 when trade openness is not at its mean (making the interactionterm democracy × trade openness nonzero).

Empirical Findings

For the control variables in Table 8.2, in the all-countries sample, the co-efficient of GDP per capita is positive and statistically significant, and thecoefficient of GDP per capita squared is negative and significant. The rate


of deforestation falls (becomes less negative) as GDP per capita rises andthen rises (becomes more negative) as GDP per capita rises above somelevel. Hence, we do not find evidence supporting the EKC hypothesis in thefull sample. The nonlinear effect found in the all-countries sample largelydisappears in the LDC sample.

Next, the effect of population density on deforestation is negative and sta-tistically significant for both samples. High population density is associatedwith more deforestation, as such countries tend to exhibit larger pressures toclear forest from needing more agricultural and urban areas, ceteris paribus.The regional dummy variables suggest that the rate of deforestation is rela-tively more intense in North America (the United States, Canada, Mexico,and Central America, the reference group), than in other regions.

For the control variables in the land degradation models, the EKC isstatistically significant for the all-countries sample, but not for the LDCsample. As before, this result is to be expected, because the DCs, whichtend to exhibit the EKC in our results, are not included in the LDC sample.A rise in population density leads to more land degradation in both theall-countries sample and the LDC sample, which makes intuitive sense. Amore dense country is expected to exert a larger pressure on land due tomore intense agriculture and waste dumping, ceteris paribus. Finally, theregional dummy variables in this set of statistical models suggest that theland in the continents of North America is statistically significantly moredegraded than land in Asia, South America, Africa, the Middle East andNorth Africa, and Oceania and is as degraded as the land found in Europe.

Additional Analyses

We conduct two types of additional analyses. First, we add regional dichoto-mous variables to the models estimated using the all-countries sample. Sec-ond, we estimate the four models (basic, basic + interaction term, basic +regional dummy variables, and basic + interaction term + regional dummyvariables) for a sample of LDCs.

Beginning with deforestation, in Models 1 and 2 for the all-countriessample in Table 8.A1, the results are similar to those in Table 8.2, exceptthat the coefficient of democracy, while negative as before, is statisticallyinsignificant. The coefficients of trade openness and democracy×tradeopenness are similar to those obtained for the all-countries sample in Table8.2. Hence, the addition of the regional dummy variables does not changethe results, despite these dummies being statistically significant.

Tabl

e8.

A1.

Eff

ects

ofde

moc

racy

and

trad

eop

enne

sson

defo

rest

atio

n

(1)

All

(2)

All

cou

ntr

ies

cou

ntr

ies

(1)

LDC

(2)

LDC

(3)

LDC

(4)

LDC

DE

MO

CR

AC

Y−0

.273

2−0

.213

1−0

.618

5∗−0

.520

7∗−0

.240

5−0

.151

7(0

.309

5)(0

.289

6)(0

.381

0)(0

.351

2)(0

.302

5)(0

.281

0)T

RA

DE

OP

EN

NE

SS0.

0119

∗∗0.

0177

∗∗0.

0123

∗∗0.

0189

∗∗0.

0107

∗0.

0170

∗∗(0

.007

0)(0

.009

1)(0

.007

1)(0

.009

3)(0

.007

5)(0

.009

1)D

EM

OC

RA

CY

∗ OP

EN

NE

SS−0

.023

4∗∗

−0.0

262∗

∗−0

.031

0∗∗∗

(0.0

109)

(0.0

112)

(0.0

114)

GD

PP

C0.

0003

∗∗∗

0.00

04∗∗

∗0.

0001

0.00

010.

0002

0∗0.

0002

1∗(0

.000

1)(0

.000

1)(0

.000

1)(0

.000

1)(0

.000

15)

(0.0

0015

)G

DP

PC

SQU

AR

ED

−8.2

7e-0

9∗∗

−9.4

8e-0

9∗∗∗

9.31

e-09

9.52

e-09

3.15

e-09

2.73

e-09

(3.4

8e-0

9)(3

.41e

-09)

(9.7

7e-0

9)(1

.01e

-08)

(9.9

9e-0

9)(1

.02e

-08)

PO

PU

LAT

ION

DE

NSI

TY

−0.0

010∗

−0.0

013∗

∗−0

.001

1∗∗

−0.0

015∗

∗−0

.001

0∗∗

−0.0

014∗

∗(0

.000

6)(0

.000

7)(0

.000

6)(0

.000

7)(0

.000

6)(0

.000

7)A

sia

1.64

08∗∗

∗1.

4892

∗∗∗

1.62

51∗∗

∗1.

3964

∗∗∗

(0.4

275)

(0.4

427)

(0.4

332)

(0.4

586)

Sou

thA

mer

ica

0.76

92∗∗

0.50

050.

8168

∗∗0.

4135

(0.4

550)

(0.4

790)

(0.4

579)

(0.4

952)

Eu

rop

e1.

0592

∗∗∗

1.29

31∗∗

∗1.

4495

∗∗∗

1.83

72∗∗

∗(0

.442

0)(0

.423

1)(0

.476

2)(0

.453

7)A

fric

a1.

4341

∗∗∗

1.42

23∗∗

∗1.

3477

∗∗∗

1.31

59∗∗

∗(0

.355

1)(0

.365

8)(0

.379

3)(0

.399

7)M

iddl

eE

ast

and

Nor

thA

fric

a3.

5943

∗∗∗

3.54

03∗∗

∗3.

1479

∗∗∗

2.97

72∗∗

∗(0

.806

8)(0

.809

6)(0

.848

0)(0

.835

0)O

cean

ia0.

8443

∗0.

7290

∗1.

3831

∗∗∗

1.35

70∗∗

(0.4

617)

(0.5

539)

(0.4

691)

(0.5

898)

Con

stan

t−2

.770

3∗∗∗

−2.6

785∗

∗∗−0

.959

8∗∗

−0.9

026∗

∗−2

.509

9∗∗∗

−2.3

765∗

∗∗(0

.426

7)(0

.438

1)(0

.398

7)(0

.404

1)(0

.558

0)(0

.571

9)O

bser

vati

ons

204

204

187

187

187

187

R2

0.31

0.33

0.24

0.27

0.30

0.34

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗∗

sign

ifica

nt

at5%

;∗∗∗

sign

ifica

nt

at1%

.

262


For the LDC sample in Table 8.A1, in Models 1 and 3, the results fordemocracy are generally similar to the results for the all-countries sample.The effect of a rise in democracy from 0 to 1 on deforestation is negativeand statistically significant in Model 1 and negative but not significant whenincluding regional dummies in Model 3. Hence, the effect of democratiza-tion is weaker in size in the LDC sample than in the all-countries sample,reflecting the fewer democracies among the LDCs. In Models 3 and 4, withthe regional dummy variables, the results from democracy are again similarto those from all-countries sample, but they are not statistically significant.The coefficients of the interaction term democracy × trade openness inModels 2 and 4 for the LDCs are similar to the results for the all-countriessample.

The effects of rises in trade openness on deforestation in Models 1and 3 for the LDC sample are positive and significant, as we obtained forthe all-countries sample. The effect is similar in size across the samples,suggesting that it largely originates from the LDCs. The results for Models2 and 4 (with the interaction terms) for the LDCs are also similar to theresults from the all-countries sample.

Table 8.A2 reports the results for land degradation. The effect of a risein democracy from 0 to 1 on the level of land degradation is negative andsignificant in Models 1 and 2 for the all-countries sample when the regionaldummies are included and for Models 1 and 2 from the LDC sample whenthe regional dummies are not included. The effect of a rise in democracyfrom 0 to 1 is also negative when the regional dummies are included for theLDC sample in Models 3 and 4, but it is not statistically significant.

The effect of a rise in the level of trade openness on land degradation isnot statistically significant when regional dummies are included in Models1 and 2 for the all-countries sample and not statistically significant in anyof the models estimated for the LDC sample. Hence, change in tradeopenness generally does not affect land degradation in LDCs and theresult in Model 1 of Table 8.2 for the all-countries sample seems to be drivenby DCs.

As for the democracy × trade openness interaction term in the addi-tional analyses for land degradation, Model 4 for the all-countries samplewith regional dummies shows that a rise in democracy from 0 to 1 signifi-cantly reduces land degradation when trade openness is held at its mean,but a further rise in trade openness above its mean does not add to thiseffect (because the interaction term is not significant). Similar results holdfor LDCs, but the negative effect of a rise in democracy from 0 to 1 isnot significant when the regional dummies are added in Model 4. For the

Tabl

e8.

A2.

Eff

ects

ofde

moc

racy

and

trad

eop

enne

sson

land

degr

adat

ion

(1)

All

(2)

All

cou

ntr

ies

cou

ntr

ies

(1)

LDC

(2)

LDC

(3)

LDC

(4)

LDC

DE

MO

CR

AC

Y−0

.783

0∗∗

−0.8

557∗

∗∗−0

.580

7∗−0

.682

4∗∗

−0.3

595

−0.3

829

(0.3

417)

(0.3

584)

(0.3

549)

(0.4

095)

(0.3

732)

(0.4

230)

TR

AD

EO

PE

NN

ESS

−0.0

033

0.00

12−0

.003

30.

0004

0.00

250.

0032

(0.0

053)

(0.0

065)

(0.0

053)

(0.0

057)

(0.0

052)

(0.0

062)

DE

MO

CR

AC

Y∗ O

PE

NN

ESS

−0.0

105

−0.0

118

−0.0

027

(0.0

104)

(0.0

117)

(0.0

100)

GD

PP

C4.

7759

∗4.

7158

∗1.

4095

1.27

570.

7217

0.76

65(3

.014

5)(3

.070

2)(3

.558

3)(3

.686

8)(4

.771

6)(4

.799

3)G

DP

PC

SQU

AR

ED

−0.3

069∗

∗−0

.307

0∗∗

−0.0

591

−0.0

526

−0.0

419

−0.0

452

(0.1

819)

(0.1

842)

(0.2

217)

(0.2

295)

(0.2

946)

(0.2

960)

PO

PU

LAT

ION

DE

NSI

TY

0.14

900.

1775

∗0.

4358

∗∗∗

0.46

09∗∗

∗0.

3033

∗∗∗

0.31

01∗∗

∗(0

.123

6)(0

.130

8)(0

.104

6)(0

.107

8)(0

.120

3)(0

.120

7)A

sia

−0.8

948∗

∗−0

.975

8∗∗∗

−0.7

592∗

∗∗−0

.778

4∗∗∗

(0.4

064)

(0.4

002)

(0.2

603)

(0.2

688)

Sou

thA

mer

ica

−0.5

313∗

−0.6

179∗

∗−0

.737

7∗∗∗

−0.7

578∗

∗∗(0

.365

7)(0

.359

9)(0

.250

8)(0

.249

7)E

uro

pe

−0.1

790

−0.0

769

−0.4

155

−0.3

938

(0.4

980)

(0.4

565)

(0.4

172)

(0.4

182)

Afr

ica

−1.6

576∗

∗∗−1

.718

1∗∗∗

−1.5

231∗

∗∗−1

.526

0∗∗∗

(0.5

754)

(0.5

868)

(0.5

260)

(0.5

324)

Mid

dle

Eas

tan

dN

orth

Afr

ica

−1.4

416∗

∗∗−1

.433

8∗∗∗

−1.0

329∗

∗∗−1

.038

9∗∗∗

(0.4

879)

(0.4

528)

(0.3

600)

(0.3

660)

Oce

ania

−2.1

813∗

∗∗−2

.082

0∗∗∗

−3.8

804∗

∗∗−3

.819

0∗∗∗

(0.8

356)

(0.8

152)

(0.4

677)

(0.5

646)

Con

stan

t−1

4.54

44−1

4.07

56−5

.667

3−5

.072

50.

2205

0.06

59(1

2.58

96)

(12.

8749

)(1

4.13

14)

(14.

6453

)(1

9.42

82)

(19.

5502

)O

bser

vati

ons

105

105

8484

8484

R2

0.31

0.32

0.25

0.26

0.40

0.40

Not

e:St

anda

rder

rors

inpa

ren

thes

es.∗

sign

ifica

nt

at10

%;∗∗

sign

ifica

nt

at5%

;∗∗∗

sign

ifica

nt

at1%

.

264


LDC sample, the interaction term is not statistically significant across theboard, suggesting that the effect of a rise in democracy does not becomestronger when trade openness is above the mean, and the effect of tradeopenness does not depend on whether democracy is 1 or 0 (or on regimetype).

NINE

Conflict and the Environment

INTRODUCTION

This chapter studies the effects of war on the environment in an innovativeway and as part of the story told in this book. Thus far we have studiedthe effects of economic openness and democracy on conflict, inequality,development, and the environment, and how these forces interact. Theconcern of analysts and policy makers is growing, including those in thePentagon, that the environment will play a role in conflict as climate changeprogresses. The effect of conflict on the environment, in the context ofglobalization, thus becomes an important complex transformation to study.

The Intergovernmental Panel on Climate Change (IPCC) expects thatclimate change will intensify during this century, causing growing envi-ronmental degradation (IPCC, 2007a, 2007b). Consequently, we may seemore wars as countries vie for degraded resources or quarrel over whoshould cover the costs of ameliorating or reducing environmental damages.Although the link from climate change to war is still only a possibility,many scholars have observed that environmental forces have already causedconflicts. In contrast, the effect of war on the environment has receivedlittle scholarly attention. Part of the reason for this neglect is perhaps thetemptation to assume that wars always destroy the environment. One mayrecall the pollution caused by burning oil wells during the 1991 Gulf Waror the destruction of forests during the Vietnam War. We argue that it isnot the only possibility; war may benefit the environment, albeit perversely,by disrupting environmentally harmful economic activities. And it followsthat war may have no net effect on the environment, for its positive andnegative effects could cancel each other. If all these effects are theoreticallypossible, the net effect of war on the environment is an empirical issue.

266

Conflict and the Environment 267

What are the effects of war on the domestic environment of a countrytaking part in the war? Does war intensify or slow down environmentaldegradation? Perhaps war does not have any net effect on the environment.Perhaps the effects of war on the environment vary across environmentalforces. Virtually all extant empirical studies of these questions either citeanecdotal evidence or employ in-depth case studies focusing on one or afew wars and types of environmental degradation. Overall, extant empiricalfindings paint an ambiguous picture, suggesting that war can either damagethe environment or reduce environmental degradation. Naturally, one istempted to ask what the general tendency is across many countries andwars. These research questions need to be addressed statistically in a large-N sample of countries and wars over time. This important task has thus farnot been performed, which motivates this chapter.

We begin our analysis by theorizing on the channels leading from warto the environment. This theoretical analysis leads us to expect that thenet effect of wars on the environment may depend on the type of environ-mental degradation and on whether wars are fought abroad or at home.Our statistical models employ a large-N sample of countries and wars overtime. The dependent variables are two salient aspects of environmentaldegradation: CO2 emissions per capita and the rate of deforestation. Theseenvironmental indicators received much attention in recent years and forgood reasons. CO2 emission is the most important contributor to the globalstock of greenhouse gases in the atmosphere, which drives climate changeon the source side. Deforestation weakens nature’s ability to reduce theglobal stock of carbon emissions, which is the most important contributorto climate change on the sink side. The theoretical mechanisms discussedin the next section by which war can affect the environment are applicablefor other environmental forces (e.g., water, land) but, given the scope of thecurrent empirical analysis, we believe they are better evaluated in separatestudies.

In studying the effects of war on the environment, we obviously do notseek to investigate if warfare could be used as a tool to reduce environmentaldegradation; rather we seek to learn about how the world works, which canhave important implications. For example, to the extent that warfare leadsto faster deforestation, the effect of localized wars can in fact be much largerthan realized thus far – also generating effects on the global climate system.Moreover, if environmental degradation causes warfare, as many scholarsargue, a finding that warfare degrades the environment would suggest thepresence of a negative feedback that can cause even more warfare andenvironmental degradation.


To summarize our primary empirical findings, war reduces the level ofCO2 emissions in a country regardless of the location of the conflict. Warfought at home increases the rate of deforestation at home, whereas warfought abroad reduces this rate. These results hold in a number of sensitivityanalyses that employ different empirical model specifications, samples, andmeasures. When evaluated in the context of climate change, these findingsimply that wars produce competing effects on climate change, promotingand alleviating environmental degradation that affects climate change.

The remainder of the chapter proceeds as follows. The next section theo-rizes the effects of warfare on environmental degradation within countries.The section that follows presents an overview of the empirical researchdesign and statistical model, followed by a discussion of the key empiricalresults. The last two sections discuss implications for global environmentalpolicy and summarize the chapter, respectively. Technical details and resultsfrom additional analyses are provided in the appendix.

THEORIZING THE EFFECTS OF WAR ON THE ENVIRONMENT

What are the effects of war on the environment of a country? Although itis tempting to assume that war always destroys the environment, there aretwo other logical possibilities: war benefits the environment, or war doesnot affect the environment. This section argues that all three possibilitiesare theoretically plausible.

War may damage the environment through direct and indirect channels.Beginning with indirect channels, the movement of armies to the battlefield,for example, may destroy fields and vegetation and degrade land. Bombingcampaigns may cause unintended fires and destroy grazing land, forests, andcropland and kill animals. Fighting may also discharge wastes that polluterivers and lakes. War may also affect the composition and scope of produc-tion, which in turn affects the environment. For example, the economy mayshift to produce more weapons and fewer consumption goods, which, inturn, may increase pollution and waste. Finally, Dasgupta (1995) reasonsthat war may weaken social norms that support environmental protectionof the commons, increasing environmental degradation.

Other than indirect effects caused by outright combat, peacetime mili-tary preparations for war can also indirectly damage the environment. Forexample, the need to sustain armies may intensify water extraction and live-stock consumption, and munition storage may release toxic materials. Asdiscussed in Westing’s (1990) edited volume, nuclear and chemical weaponsproduction and storage may damage ecosystems. Additionally, as noted by


Singer and Keating (1999), standing armies may dump materials such asshells and chemicals into the environment. The buildup of army campsmay destroy arable land and vegetation and increase logging, and trainingarmies may degrade arable land and cause pollution. Another indirect chan-nel involves war refugees. As observed by Allan (1987) for Afghani refugees,the temporary settlement of many refugees in hastily made-up camps maycreate mountains of waste and destroy the surrounding vegetation andforestry.

Moving to the direct effects, war can destroy the environment. Thewarring parties may destroy the environments of each other as part of theirstrategies to win the war. For example, armies may set forests on fire ordump defoliates on them to deny the enemy timber or hiding places, as wellas set oil wells on fire; destroy fresh water resources, crops, grazing fields,and vegetation; and kill domesticated animals to prevent their use by theenemy. Armies may also flood large areas by destroying dams or openingfloodgates or destroy other natural resources to slow down or preventenemy movements. Alternatively, the warring parties may intensify theexploitation of natural resources to sell them and finance their war efforts.

A number of stylized facts support these expectations. McNeill (2001)lists wars that destroyed forests, including the French war in Morocco in the1920s, World War II, the Greek Civil War in the late 1940s, the British war inMalaya in the 1950s, the Vietnam War, and the Russian war in Afghanistanin the 1980s. Articles in The Economist (1995, 2003) report that warringparties in Myanmar, Angola, and Sierra Leone caused intensified timberlogging in forests to finance their war efforts. Deacon (1994) noted thatwars in less developed countries (LDCs) eroded norms of preservation,increasing the rate of deforestation. Finally, during the 1991 Gulf War theIraqis set oil wells on fire and spilled oil into the Persian Gulf, generatingpollution and destroying natural habitats.

Sometimes, however, wars may also benefit the environment. If the wareffort reduces the scope of ordinary economic activities that destroy theenvironment at home, and the damaging effects of war are not too high,the net effect of war could be beneficial. For example, fewer fishing boatsmay go to sea during wars because fishermen are enlisted in the armyor seek to avoid maritime combat locations. As a result, fish stocks maygrow. A second example involves changes in industrial activities. Wars maydestroy industrial plants and transportation infrastructures. As productiondeclines, pollution and waste discharges also decline. Furthermore, warsmay also alter the composition of production and consumption at home.For example, more fuel may be shipped to the front, creating fuel shortages


at home. Consequently, ordinary economic consumption and productionmay decline, reducing pollution and waste. Or, as more people are enlistedin the army, economic activities in sectors using them in peacetime mayshrink, including timber logging, fishing, and industrial production. As aresult, stocks of forests and fish may increase, and industrial pollution maydecline.

Turner (1990) and McNeill (2001) provide a number of examples of warsthat benefited some environmental resources. For example, World War IIhelped fish stocks increase in the Atlantic Ocean because normal fishingdecreased. In German-occupied Europe and Japan, industrial emissions fellduring World War II because of coal shortages and the damages that resultedfrom the Allies’ bombing campaigns. Iraqi land mines kept people fromgoing into the Kuwaiti dessert, allowing the stocks of formerly overexploitedplants and animals to revive.

Thus far we have discussed positive and negative effects of war on theenvironment. However, two reasons exist for why war may not affect theenvironment. First, the intensity or nature of the fighting may not suffice todamage the environment. For example, surgical air strikes may not destroymuch of the economy and, therefore, may not largely affect emissions.Second, the negative and positive effects of wars on the environment maycancel each other. One could probably break down the positive and negativeeffects along sectors and forces, but this task is currently not feasible due todata unavailability. That said, one ought to note that theoretical ambiguityand competing effects are quite common in environmental studies, includ-ing the effects of democracy and trade on the environment as indicated inChapters 7 and 8, as well as the effects of the environment on war. In suchsituations, the relative sizes of competing forces are not known theoreti-cally, but their net effect is estimable. As in these and other social scienceareas, we can and should study the net effect of war on the environmentempirically.1

Before we turn to our research design, we need to consider three addi-tional theoretical issues. First, our discussion suggests that the effects of waron the domestic environment may vary across environmental features. Itfollows that our empirical analysis ought to examine various environmen-tal aspects, and our statistical results may also differ across these disparate

1 Other areas that involve theoretical ambiguity and competing forces include the effect of tradeon armed conflict, the effect of globalization on democracy, and the effect of economic growthon inequality.


dimensions. In other words, wars may benefit some environmental aspectsand harm others.

A second issue concerns the locations of wars. Often conflicts are cat-egorized into two types: interstate and intrastate. Intrastate conflicts havebeen on the rise; interstate conflicts have declined but not disappeared. Inthe context of the environmental impact of war, the interstate–intrastatedistinction is less useful than the location where a conflict actually occurs:at home or abroad. It is tempting to argue that only wars fought on a coun-try’s own territory should matter for its environment; however, this is notthe only possibility. Wars fought abroad may change the relative balanceof sectors in the national economy and alter the scope and composition ofproduction and consumption at home. For example, fuel may be sent to thefront abroad while rations are imposed at home. People may be sent to fightabroad, reducing production in some domestic sectors. Production in othersectors may grow as war materials are sent abroad. All these war-inducedeconomic changes can significantly affect the state of the environment athome.

Finally, the effect of war on the environment may also differ betweendeveloped countries (DCs) and LDCs. In the case of CO2 emissions, weexpect to see a weaker effect of war for LDCs than for DCs. Consider,for example, two economies X and Y that obtain 90% and 10% of theiroutputs, respectively, from emission-generating sectors (e.g., industry, util-ities). Because Y has fewer emission-generating activities, the net effect ofwar on Y’s emissions should be weaker than that on X’s emissions. On theother hand, in the case of deforestation, we expect to see a stronger effectof war for LDCs than for DCs. Because LDCs are more likely to be agrarianeconomies and their output depends more on forests, the effects of war onthe rate of deforestation should be larger in LDCs than in DCs. On a moreapplied level, in the time period for which our data are available (recentdecades), LDCs generally have experienced more interstate and intrastatewars than DCs, further suggesting the need to distinguish LDCs in theempirical analysis.

To summarize, wars produce competing theoretical effects on the domes-tic environments of the countries fighting them. These effects may changedepending on the environmental features at issue, the locations of warfare,and the development levels of the countries. However, it is worth stressingthat despite this complexity, it is feasible and desirable to estimate the neteffect of wars on particular environmental features, an issue to which weturn next.



This section illustrates our empirical model, discusses related researchdesign issues, and presents the main findings. As in the previous chapters,this discussion is self-contained and does not require statistical expertise.The technical details of the statistical model, data sources, and measureconstruction are shown in the chapter appendix, following the same orderof presentation as in the main text.

Empirical Model

To assess the effect of war on the environment, we specify and estimate thefollowing statistical model of the environment. We denote variables withsmall capital letters and their coefficients with Greek notations. The notation�c denotes a vector of coefficients for country fixed-effects variables. Eachcoefficient indicates the effect of the independent variable on the dependentvariable, the phenomenon we seek to explain. The notation εt denotes therandom error not explained by the statistical model. The variable subscriptst and t − 1 indicate the time period of the variable, where t represents thecurrent period and t − 1 the previous time period (a lagged variable).To simplify the presentation, we refer to the variables without their timesubscripts t or t − 1. The model specification is based on the aforementionedtheoretical discussions and the literature on democratization. It providesa structure for us to guide the statistically uninitiated readers through theempirical exercise.

environmentt = �0 + �1wart−1 + �2democracyt−1

+ �3real gdppct−1 + �4real gdppc squaredt−1

+ �5trade opennesst−1 + �6population densityt−1

+ �7lagged environmentt−1 + �8year

+ �ccountry fixed effects + εt. (9.1)

The dependent variable, environment, represents two different aspectsof environmental degradation: CO2 emissions per capita and the rate ofdeforestation. These two forces are important, if not the most important,aspects of global environmental degradation today, and they will becomeincreasingly so in the coming decades. CO2 emissions are a major causeof climate change because they contribute to the greenhouse effect. Defor-estation reduces the availability of fresh water because forests reduce losses


due to runoff and evaporation and absorb much of the world’s rainwa-ter into underground aquifers. Deforestation also promotes land degrada-tion because forests keep topsoil in place and prevent land erosion. More-over, deforestation influences the global climate system by reducing nature’sability to absorb and break down CO2 emissions. Hence, both indicatorscorrelate with climate change, albeit in distinct manners. The CO2 emis-sions indicator reflects the outcome of human activities that produce emis-sions, whereas the rate-of-deforestation indicator measures the outcomeof human activities that absorb CO2 emissions. The stock of emissions atany given point in time depends on both types of human activities (Harris,2006; IPCC, 2007a, 2007b). The two indicators give us an opportunity toinvestigate how warfare relates to different processes that influence climatechange.

The key independent variable, war, is measured in several ways. The firstindicator, war at home or abroad, reflects the presence or absence of acountry’s involvement in at least one interstate, intrastate, or international-ized violent conflict that is identified as a war in our time period (yearly forCO2 emissions per capita, and decade for the rate of deforestation). Thisaggregate indicator does not distinguish wars in terms of their geographicallocations. The reader may recall that our theory allocates importance towhether a war is fought at home or abroad. We therefore also create twomore war-related indicators: war abroad reflects a country’s involvementin any war fought abroad and war at home represents a country’s involve-ment in any war fought at home (including a civil war). These indicatorsallow us to test whether the location of warfare matters in environmentaloutcomes. All these indicators of war denote armed conflict situations thatcause at least 25 battle deaths. We describe the details about these variablesin the chapter appendix.

Like the other models in this book, this statistical model also controlsfor other forces that influence the environment. The empirical literatureexplaining environmental degradation motivated our inclusion of the con-trol variables: real gdppc is the gross domestic product (GDP) of a countryexpressed in real terms, per capita; real gdppc squared is the square of thereal GDP per capita variable. As discussed at length in the previous chap-ter, these two variables together represent the environmental Kuznets curveeffect. democracy is the level of democracy in a country, measured (as inthe other chapters) over a continuum ranging from complete democracy tocomplete autocracy; trade openness is the share of a country’s total tradein its national economy; and population density is the population of acountry divided by the area of its territory.


Because the model of CO2 emissions per capita uses pooled time-seriescross-sectional data, it requires some additional control variables to ensurethe validity of statistical inference: year is a yearly counter that modelsthe linear trend in the emissions, lagged environment represents theCO2 emissions from the previous year, and country fixed effects arecountry dummies that capture heterogeneity across different countries.These additional variables are not necessary for the deforestation model,because its data are largely cross-sectional.


In designing our empirical analysis, we also need to consider several tech-nical issues. We provide an overview of these issues here and delegate thedetails to the appendix. The first issue is whether we should conduct a pooledtime-series cross-sectional analysis or a cross-sectional analysis. The sam-ple of CO2 emissions per capita includes 143 countries from 1961 to 1997.The deforestation sample covers 134 countries over two decades (1980s and1990s). Given the availability of data, we employed the pooled design for theCO2 model, and the cross-sectional design for deforestation. In the cross-sectional design, the variables, except for the dichotomous war variables,take on decade-average values, and in the pooled design they take on yearlyvalues.

The second issue concerns the properties of the error terms and thepossibility that the right-hand-side variables are excessively correlated. Weapply appropriate econometric techniques to address any issue related to theerror term and provide some diagnostic test for the possible high correlationamong independent variables.

Third, we model the possibility of path dependence for carbon dioxideby including the lagged level of CO2 emissions per capita as one of theindependent variables, which also helps to control for the effect of anypossibly omitted variable. For deforestation the lagged dependent variableis not included, given the cross-sectional nature of the sample.

Fourth, we lag the right-hand-side variables in the CO2 models for tworeasons. The effects of the independent variables on the environment maynot be immediate and take some time to materialize. In addition, the right-hand-side variables may themselves be affected by the environment.

Finally, following our argument that the effect of war on environmentaldegradation may differ between DCs and LDCs, we estimate our modelsfor both a sample of all countries and a sample of LDCs only.


Table 9.1. War and CO2 emissions per capita

All countries LDCs

(C1) (C2) (C3) (C4)

WAR AT HOME −0.0375∗∗∗ −0.0173OR ABROAD (0.016) (0.016)

WAR AT HOME −0.027∗ −0.021(0.017) (0.017)

WAR ABROAD −0.057∗∗ 0.004(0.028) (0.020)

DEMOCRACY −0.0016∗∗ −0.0016∗∗ −0.0028∗∗∗ −0.0028∗∗∗

(0.001) (0.001) (0.001) (0.001)REAL GDPPC 0.00008∗∗∗ 0.00009∗∗∗ 0.00008∗∗∗ 0.00008∗∗∗

(0.00002) (0.00001) (0.00003) (0.00003)REAL GDPPC SQUARED −2.75e-09∗∗∗ −2.76e-09∗∗∗ −2.00e-09∗ −2.02e-09∗

(5.70e-10) (5.70e-10) (1.36e-09) (1.36e-09)TRADE OPENNESS −0.0009 −0.0009 −0.0002 −0.0002

(0.001) (0.001) (0.001) (0.001)POPULATION DENSITY 0.0007∗∗∗ 0.0007∗∗∗ 0.0006∗∗ 0.0006∗∗

(0.0003) (0.0003) (0.0003) (0.0003)LAGGED CO2 0.866∗∗∗ 0.866∗∗∗ 0.850∗∗∗ 0.850∗∗∗

(0.023) (0.023) (0.035) (0.035)YEAR 0.001 0.001 0.001 0.001

(0.001) (0.001) (0.001) (0.001)Constant −1.794 −1.597 −1.474 −1.516

(2.13) (2.15) (2.30) (2.30)Observations 3830 3830 3053 3053R2 0.99 0.99 0.97 0.97


Empirical Findings

This section presents our empirical findings. We first present the mainresults and then discuss results for several additional analyses. Findings forthe control variables and additional technical details are presented in theappendix of this chapter.

Table 9.1 presents the results for CO2 emissions per capita. The diagnos-tics used to ascertain the econometric performance of the model performwell, as discussed in the appendix. Models C1 and C2 are based on thesample of all the countries, while Models C3 and C4 focus on the sample ofLDCs. The key independent variable in Models C1 and C3 is war at homeor abroad, which does not distinguish between wars fought abroad and


at home. Models C2 and C4 distinguish between wars according to theirlocations.

The coefficient of the variable war at home or abroad in Model C1is negative and significantly different from zero. War involvement reducesCO2 emissions per capita, regardless of its location. The effects of warabroad and war at home in Model C2 are also negative and significantlydifferent from zero. In the full sample, both war at home and war abroadreduce CO2 emissions per capita. What do these results mean? Recall thatthe effect of war on the environment may occur through changing output,resource allocation and sectoral breakdown, and destroying the economy.After controlling for the output effect (i.e., through real GDP per capita inthe model), the net effect of war on the environment is negative.

How large are these effects? We compute the sizes of the effects of waron CO2 emissions per capita in Models C1 and C2. We assume that thewar variable changes from 0 to 1, and all the other variables in the modelare held at their sample mean values. An involvement in war at home orabroad reduces CO2 emissions per capita by 1.14%, an involvement in warat home reduces CO2 emissions per capita by 0.76%, and an involvementin war abroad reduces CO2 emissions per capita by 1.61%.

These substantive effects represent statistically significant outcomes butseem relatively small in size. At the same time, as in other models in thebook that include lagged dependent variables on the right-hand side, theseimmediate effects do not tell the whole story of the effect of war on CO2

emissions per capita – war affects the current level of CO2 emissions percapita via its direct effect, as computed earlier, and continues to affect theCO2 emissions per capita in the next period via its effect on the laggedemissions, which accumulates over time. With some additional computa-tion, we find that in the long run, an involvement in war at home orabroad reduces CO2 emissions per capita by 8.5%. Similarly, an involve-ment in war at home reduces CO2 emissions per capita by 5.67%, and aninvolvement in war abroad reduces CO2 emissions per capita by 12.01%.These long-run effects are not only substantial but also considerably largerthan the short-run effects.

We have argued that the effect in LDCs may differ from that observed ina sample of all countries. Turning to Models C3 and C4 for the LDCs, wefind that an involvement in war at home or abroad does not affect CO2

emissions per capita. We get the same results for the effects of involvementsin war at home and war abroad. Recalling our theoretical discussion,one may interpret this result in two ways. First, it is possible to argue thatthe competing effects of war on the environment cancel out in the case


Table 9.2. War and deforestation

All countries LDCs

(D1) (D2) (D3) (D4)

WAR AT HOME OR −0.075 −0.188ABROAD (0.326) (0.347)

WAR AT HOME −0.509∗∗ −0.696∗∗∗

(0.281) (0.281)WAR ABROAD 0.840∗∗ 1.065∗∗∗

(0.414) (0.472)DEMOCRACY −0.121∗∗∗ −0.122∗∗∗ −0.099∗∗∗ −0.097∗∗∗

(0.038) (0.037) (0.036) (0.034)REAL GDPPC 0.0004∗∗∗ 0.0004∗∗∗ 0.0002∗ 0.0002∗

(0.0001) (0.0001) (0.0002) (0.0001)REAL GDPPC SQUARED −1.03e-08∗∗∗ −1.02504e-08∗∗∗ 3.83e-09 5.16519e-09

(3.75e-09) (3.58e-09) (9.43e-09) (7.50e-09)TRADE OPENNESS 0.013∗∗ 0.012∗∗ 0.012∗ 0.011∗

(0.007) (0.007) (0.007) (0.007)POPULATION DENSITY −0.0010∗ −0.0009∗ −0.0011∗∗ −0.0010∗

(0.0006) (0.0006) (0.0006) (0.0006)Constant −2.524∗∗∗ −2.328∗∗∗ −2.057∗∗∗ −1.791∗∗∗

(0.657) (0.549) (0.713) (0.579)Observations 204 204 187 187R2 0.249 0.272 0.253 0.292


of LDCs, resulting in a statistically insignificant result. Alternatively, aninterpretation we prefer, the effect of involvement in war at home orabroad, war abroad, or war at home on CO2 emissions per capita inLDCs is so small as to be statistically insignificant. This interpretation is inline with the observation that the extent of CO2-promoting activities percapita in LDCs is very small. As such, involvement in wars simply does notaffect CO2-promoting activities much, driving a statistically insignificantoutcome.

Table 9.2 reports the results for deforestation. As in the case of CO2

emissions per capita, Models D1 and D2 are based on the sample of allcountries, whereas Models D3 and D4 are based on the sample of LDCs.Models D1 and D3 do not distinguish the locations of wars (home orabroad), whereas Models D2 and D4 make this distinction.

In Model D1, the effect of war at home or abroad – war involvementregardless of its location – on deforestation is not statistically different fromzero. But in Model D2, the coefficient of war at home is negative andsignificantly different from zero, whereas the coefficient of war abroad is


positive and significantly different from zero. These results suggest that warinvolvement and the location of the war matter for deforestation. Involve-ment in wars fought at home increases the rate of deforestation, whereasinvolvement in wars fought abroad reduces it.

Our results suggest that wars fought at home destroy forests and/orintensify timber logging and forest clearing at a faster-than-usual speed.Consequently, the rate of deforestation at home increases. In contrast, warsfought abroad alleviate these pressures. As a result, wars fought abroadreduce the rate of deforestation at home. It is plausible that wars foughtabroad divert labor and financial resources from timber logging and forestclearing. In addition, wars fought abroad inflict less damage on forests athome than do wars fought at home.

The opposite effects of wars fought at home and abroad on deforestationalso help to explain the lack of effect of war at home or abroad. Thecompeting effects cancel out when the war variable does not distinguish thelocations of wars. Once wars are separated according to their geographicallocations, their effects become statistically significant.

For the LDCs sample, Models D3 and D4 yield results consistent withthose from the sample of all countries. Wars do not affect deforestation in theLDCs if we do not distinguish their locations. Once we separate wars basedon their locations, their effects start to emerge in patterns consistent withthose from the full sample. Wars abroad slow down deforestation whereaswars at home lead to faster deforestation. In addition, as expected, thecoefficients of the war variables in Model D4 for the LDCs are considerablylarger in size than those in Model D2 for the sample of all countries.

We now compute the sizes of the effects of the statistically significantwar variables, implementing the same procedure as in the CO2 case (i.e.,change a war variable from 0 to 1 while holding all the other variables attheir mean and computing the change in the effect in percent). Based onModel D2 (the sample of all countries) in Table 9.2, we find that war athome increases the rate of deforestation rate by 179.8%, and war abroadreduces the rate of deforestation by 296.8%. Wars fought abroad or at homeproduce very large effects on the rate of deforestation at home, albeit inopposite directions.

Additional Analyses

We have presented a relatively large number of results on the effect ofwarfare on CO2 emissions and deforestation. To establish the robustness ofour results, we conduct two additional analyses: (1) we include a cubic real


GDP per capita term for both the CO2 and the deforestation models and (2)we exclude the lagged dependent variable (LDV) from the CO2 models. Theresults of these analyses are fully presented and discussed in the appendix.

To summarize, we first reestimate our statistical models while assumingthat the dependence of CO2 emissions per capita on real GDP per capita iscubic, not quadratic. The results for the effects of war at home or abroad,war at home, and war abroad are similar to those reported in Table 9.1in terms of their signs, significance, and sizes, indicating the robustness ofthe CO2 models. The coefficients of some of the GDP per capita terms differfrom those in Table 9.1, as could be expected, and they are interpreted in theappendix. In general, the dependence of CO2 emissions per capita on realGDP per capita may be cubic, but the effects of war remain robust, whichis the important issue here.

We then include a cubic real GDP per capita term for the rate of defor-estation. The coefficients for the war and the control variables are similar tothose in Table 9.2. The effects of war on deforestation also remain robust.

Finally, we estimate the models of CO2 emissions per capita presented inTable 9.1 without the LDV. The results, presented in the appendix, are againin line with Table 9.1 in terms of signs and significance, with one exception.In the new analysis, the effect of war at home on CO2 emissions per capitais negative and significantly different from zero in the LDC sample, as inthe full sample. The effects of war on CO2 emissions per capita are nowlarger in size than those in the presence of the LDV. We offer an explanationfor this outcome in the appendix. Hence, the results in Table 9.1 are not anartifact of the presence of the LDV.


Our theoretical discussion suggests that war may slow down or speed upenvironmental degradation of a country participating in the war. The overalleffect may change on the basis of the location of the war (at home orabroad), the particular environmental indicator inspected, and the levelof the country’s economic development. Our empirical analysis, focusingon the net effect of war on CO2 emissions and deforestation, employs astatistical model for many countries and years, which to our knowledge isthe first of its type in the conflict–environment literature.

Summarizing our finding, a country’s involvement in interstate orintrastate warfare reduces its CO2 emissions per capita, but the effect isvery weak in LDCs. Wars fought at home increase the rate of deforestationin a country, whereas wars fought abroad slow deforestation. The effects of


war on the rate of deforestation (either positive or negative) are larger inLDCs than in DCs.

These findings have implications for climate change because our envi-ronmental indicators are important contributors to this phenomenon. Ingeneral, some of the effects of climate change have already been observedtoday, including rising sea levels, retreating glaciers, melting ice poles, andmore frequent and intense extreme weather occurrences such as stormsand droughts. Although exact numbers are still unknown, climate changeis expected to be the largest source of environmental degradation in thecoming decades, assuming business as usual (IPCC, 2007a, 2007b).

How will people respond to these expected changes? In principle, theycould try to mitigate the expected effects, adapt to the changes by alteringpractices, or do nothing and accept the decline in their quality of life.Although adaptation and mitigation may seem reasonable measures, theyface formidable obstacles. For example, adapting to the sea-level rise bybuilding coastal defenses may cost tens of billions of dollars and take manyyears to complete (Allen et al., 1998; IPCC, 2001).

Currently, a number of U.S. and European policymakers appear to bepreparing for the possibility that climate change will increasingly induceviolent conflict (New York Times, February 29, 2004). As Schwartz andRandall (2003: 22) write in a report commissioned by the U.S. Departmentof Defense, “Disruption and conflict will be endemic features of life.” Theissue is not without debate, but many scholars agree that growing environ-mental degradation and natural resource problems already make violentconflict significantly more likely.

The possibility of climate change–induced warfare recasts our findingsin a new light. Will the increased incidence of warfare accelerate climatechange, or slow it down? Since climate change is an evolving force whoseattributes are not yet fully manifested and understood, answering thesequestions requires prediction. In social sciences, prediction typically is basedon historical data. Caution is needed when employing this method, but theworking assumption is that, when analyses can explain historical patterns,they give us useful information about the future.

We believe that our findings help us gain insight on our question. Ourfindings suggest that wars can slow down the pace of climate change, but,of course, we do not argue that warfare is a good policy instrument forcombating climate change. On the contrary, we believe that there are farbetter, more peaceful ways for reducing CO2 emissions and deforestation,including developing alternative sources of energy, appropriate institutions,technological measures, and perhaps reducing the growth rate of production


and consumption while fossil fuel remains our main source of energy.However, our analysis demonstrates that although war is generally believedto be damaging, it also can perversely benefit the environment.

SUMMARY AND OUTLOOK

The role of environmental forces in violent intrastate and interstate warfarehas received much scholarly attention in recent studies, but the reverse causalarrow from warfare to the environment has been largely overlooked. In thischapter, we study the effects of war on the domestic physical environment.Does war have any effect on the domestic environment of a country takingpart in the war? Does it intensify or slow down environmental degradation?Do the effects of war on the environment vary across environmental forces?Only a few empirical studies have investigated the effects of war on thephysical environment, and virtually none of these studies has employedstatistical methods for a large sample of countries and years.

This chapter develops a statistical model of the effects of war on twoimportant types of environmental degradation: CO2 emissions and defor-estation. We find that war reduces CO2 emissions at home regardless ofwhether the war is fought at home or abroad. War fought at home increasesthe rate of deforestation at home, whereas war fought abroad reduces thisrate. These findings have implications for understanding the sources ofclimate change as well as the environmental consequences of warfare.

This chapter contains the last theoretical and empirical analysis we con-duct in this book. Throughout Chapters 2–9, we examined different typesof interactions among democratic governance, economic openness, incomeinequality, national economic development, involvement and initiation ofinterstate and intrastate military conflicts, and various indicators of thequality or degradation of the physical environment within which all thesesocioeconomic and political interactions take place. Guided by various theo-retical arguments, debates, and empirical studies on a wide range of topics,we have demonstrated that these forces interact with one another alongvarious channels over time.

The reader may recall that we started our analytical theoretical and empir-ical journey in Chapter 1 with a stylized graphical model of the various inter-actions among the economic, political, social, and environmental forces westudy in this book. This graphical model offers merely a scheme of howour inquiry would unfold, providing little specifics as to the nature of thevarious interactions in our conceptual framework. In the last chapter ofthe book, we revisit this graphical model. In Chapter 10, we flesh out the


graphical model with the various central findings we obtained in the variouschapters. The revised graphical model offers a relatively complete view ofthe complex transformation the global system is undergoing in our times. Italso uncovers for us the kinds of trade-offs and mutually reinforcing effectsthe various important processes generate for us.

APPENDIX


Empirical Model

The dependent variable, environment, is measured in two ways: carbondioxide (CO2) emissions per capita and the rate of deforestation. The indi-cator of CO2 emissions per capita is expressed in metric tons per capita in acountry in a year, adjusting for cross-national differences in population. Thisindicator captures emissions generated by various human activities, includ-ing industrial production, burning of fossil fuels, burning of gas releasedin petroleum extraction, cement manufacturing, and gases released fromstored fuels. The data come from the World Development Indicators (WorldBank, 2002).

The indicator of deforestation measures the rate at which forests decreaseor increase in a country due to activities such as building settlements,mining, ranching, farming, and reforestation. For this variable, a positivevalue indicates a rise in forested area over time (afforestation) and a negativevalue implies a fall in forested area over time (deforestation). Areas harvestedwith the intent of natural regeneration and areas degraded by gatheringwood for fuel, by acid rain, or by natural fires are not included. Data onaverage annual deforestation rates per decade per country are collectedfrom the World Resources Institute (1999) and the State of the World’sForest Report (2001).

The key independent variable, war, is also measured in several ways. Thevariable war at home or abroad is dichotomous, which is set to 1 if acountry participated in at least one interstate, intrastate, or international-ized violent conflict that is identified as a war in our time period. For theempirical analysis, we need to decide how to identify a violent conflict asa war, which requires some threshold. Many studies employ the thresholdof 1,000 battle deaths for that purpose. For our purpose the threshold of1,000 battle deaths may be too high, because it results in classifying many


periods involving violent conflicts in recent decades as peaceful, thus omit-ting relevant information. Therefore, we decided to employ a lower thresh-old in identifying periods of warfare in this study.

We use data compiled by the Department of Peace and Conflict Researchat the University of Uppsala, Sweden and the International Peace ResearchInstitute, Oslo, Norway. In these data, one can identify an armed conflict asone causing at least 25 battle deaths and involving the use of armed forcebetween two parties, where at least one party is the government of a state(Gleditsch et al., 2002; Strand et al., 2005); we employ this approach.

To evaluate the effects of wars according to whether they are fought athome or abroad, we create two additional dichotomous variables based onthe Conflict Sites data set compiled by Raleigh and Gleditsch (2005). Thedichotomous variable war abroad is set to 1 when a war is fought abroad,and 0 otherwise. The dichotomous variable war at home is set to 1 whena war is fought at home (including a civil war), and 0 otherwise.

The empirical literature explaining environmental degradation moti-vates our inclusion of the following control variables: real gdppc, realgdppc squared, democracy, trade openness, population density,and, in some models, real gdppc cubed, and lagged environment (e.g.,Panayotou, 2000a; Li and Reuveny, 2006; Dinda, 2004).

real gdppc denotes real GDP per capita. As discussed in Chapter 7, theliterature suggests that environmental degradation follows an inverted Uwith respect to real gdppc, known as the EKC. The empirical validity ofthe EKC, we recall from previous chapters, is debated. To model the EKC,we include real gdppc and its squared term, real gdppc squared, in themodel. If the EKC exists, the coefficient of real GDP per capita should bepositive whereas the coefficient of real GDP per capita squared is negative,with the former coefficient larger in size than the absolute value of thelatter coefficient. The data, expressed in constant 1996 international dollarsadjusted for purchasing power parity, come from the Penn World Table 6.1(Heston et al., 2002).

The level of national democracy, democracy, is measured using POLITYIV data (Marshall and Jaggers, 2006). The construction of this variablefollows the one in previous chapters. The measure ranges between −10(most autocratic) and 10 (most democratic). The effect of democracy onthe environment was discussed in Chapter 7.

The level of trade openness of a country, trade openness, is the sumof exports and imports of a country divided by its GDP, reflecting theimportance of trade in the country’s economy. The effect of trade opennesson the environment was discussed in Chapter 8.


A rise in population density is expected to increase CO2 emissions percapita because it leads countries to consume and produce disproportionallymore (e.g., China, India, the United States, Western Europe). Densely pop-ulated nations may require more land for agriculture and industry, raisingthe rate of deforestation. Yet these countries may also be more urbanizedand depend less on the environment for livelihood, thereby clearing fewerforests (e.g., Western Europe). Templeton and Scherr (1999) argue thatdensely populated countries may strengthen property rights or privatizeforests, reducing deforestation, and Panayotou (2000a) concludes that arise in population density creates competing theoretical effects on forests.Data for population density come from the World Development Indicators(World Bank, 2002).

For the model of CO2 emissions per capita, we also include the first lagof the emission variable (lagged environmentt−1 in Equation (9.1)). Therole of the lagged dependent variable is similar to its role in the other modelsin this book: it addresses path dependence in CO2 emissions, amelioratesthe undesired effects of nonspherical regression errors, and accounts forpossible omitted variables.

Finally, for the model of CO2 emissions per capita, we also include yearand country fixed effects; year is a year counter that captures thepossibility of trending in the CO2 emissions data, as discussed in the nextsection, and country fixed effects are dichotomous country dummyvariables. Each variable is set to 1 for all the years for some country, andis set to zero otherwise. These variables capture nonchanging structuralvariables of a country that may affect CO2 emissions per capita, as discussednext.


We consider several design issues. The first issue is whether we shouldconduct a pooled time-series cross-sectional analysis or a purely cross-sectional analysis. Although the pooled design is superior in capturingboth temporal and cross-sectional patterns, our choice is dictated by dataavailability. Our sample for the CO2 models covers 134 countries with dataavailable annually from 1961 to 1997, whereas the sample for the rate ofdeforestation covers 134 countries with data available for two decades (1980sand 1990s). Thus, the unit of analysis in the test is the country-year for CO2

emissions per capita and the country-decade for the rate of deforestation.For deforestation, the right-hand-side variables, except for the dichotomous


war variables, take on their decade-average values to capture the cross-sectional patterns.

The second issue concerns the presence of possible heteroskedastic errorvariance. As in previous chapters, we deal with this issue by estimatingHuber–White robust standard errors for both models of the CO2 emissionsper capita and models of the rate of deforestation.

Third, we consider the possibility of path dependence for CO2, for we havepooled data for this variable. For deforestation, as noted, we have a cross-sectional sample, which makes path dependence a nonissue on technicalgrounds. CO2 emissions per capita are expected to exhibit path dependencebecause the CO2-generating industrial facilities, energy equipment, andtransportation devices depreciate slowly. In the prevailing technologicalparadigm, even when old units are replaced, new units generate emissions.Thus, a high level of CO2 emissions per capita in one period is likely to befollowed by high emissions in the next period. As discussed in Chapter 2, ifnot modeled explicitly, path dependence in the dependent variable is likelyto cause serial correlation in the error term. A more refined modeling ofthis inertia is outside the scope of this chapter. We return to this issue inChapter 10.

The fourth issue concerns the possibility that our models omit someglobal and national structural variables that might affect the environment,subjecting our results to an omitted variable bias. The inclusion of theLDV accounts for some of the omitted variable bias. We further guardagainst this possibility in the CO2 models by using the country fixed-effects estimator and the year variable. The country fixed effects controlfor excluded country-specific characteristics (e.g., climate, regional atmo-spheric integrity, institutional setup, geography); year accounts for thepossibility that CO2 emissions rise over time due to economic growth ora declining ability of the atmosphere to break up emissions over time (asthe stock of emissions rises). Although these extra features make it harderfor us to find statistically significant effects, we prefer to err on the sideof caution. For deforestation, because our sample includes only two crosssections, we employ ordinary least squares without having to apply theseadditional controls.

Fifth, we lag the right-hand-side variables in the CO2 models for tworeasons. The effects of our covariates on the environment may take sometime to materialize. Also, whereas our design treats the right-hand-side vari-ables as exogenous, they may be affected by environmental degradation. Forexample, CO2 emissions may correlate with environmental laws that affect


the real GDP per capita, and environmental degradation, as noted, maycause violent conflict. As we explained in previous chapters, many scholarsoften lag the right-hand-side variables to control for possible simultaneity,and we do so as well.

Sixth, we need to consider the risks of multicollinearity and, for CO2,nonstationarity. As noted, when the effects of key right-hand-side variablesare insignificant and the model R2 value is high, the insignificance canreflect multicollinearity, which inflates the standard errors of the coefficientestimates. Nonstationary dependent variables and error terms may causespurious results. We employ the variance inflation factor (VIF) statisticto diagnose the problem of multicollinearity, and we conduct the Levin–Linard and Im–Pesaran–Shin panel unit root tests to diagnose the problemof nonstationarity.

Empirical Findings

We first evaluate if the model suffers from the presence of multicollinearityand if the error term has good properties. As noted, multicollinearity is acause of concern when effects of key variables are statistically insignificantbut the model goodness of fit is very high. The average VIF is very high(28.7), but it is caused by the country fixed-effects dummy variables, whichis to be anticipated. The individual VIFs for the variables are below 3.5,much lower than the conventional threshold of 10, with the exceptions ofreal GDP per capita and its squared term. This result is expected and in factcannot be helped if one wants to model the EKC. The results obtained fromLevin–Linard and Im–Pesaran–Shin panel unit root tests indicate that theissue of nonstationary is not a cause of concern in our case.

In Table 9.1, the coefficients of war at home or abroad in Model C1and of war abroad and war at home in Model C2 are negative andsignificant. These results are discussed in the main text. Note that the effectis statistically significant even in the presence of the empirically taxingcountry fixed effects. One may wonder why war reduces CO2 emissions percapita if real GDP per capita is included in the model. Recall that the effectof war may work through changing output, affecting resource allocationand sectoral breakdown, and destroying the economy. Controlling for theoutput effect, we find that the net effect of war due to the second and thirdchannels is negative. A sectoral analysis seems desirable but is outside thescope of this chapter. We will return to this point.

For the sizes of the effects of war in Models C1 and C2 we report changesrelative to a base case. We compute the size of the effect as the percent


change in CO2 emissions per capita when the war variable changes from0 to 1, with all the other variables in the model set at their sample meanvalues. On the basis of all estimated coefficients for the base case, we findthat an involvement in a war reduces emissions by 1.14%, an involvementin war fought at home reduces emissions by 0.76%, and an involvement inwars fought abroad reduces emissions by 1.61%. On the basis of only thestatistically significant coefficients for the base case (i.e., setting insignificantcoefficients to zero), we find larger effects: an involvement in a war reducesthe CO2 emissions per capita by 2.77%, an involvement in a war foughtat home reduces emissions by 1.68%, and an involvement in a war foughtabroad reduces emissions by 3.56%.

As discussed in the main text, these effects do not tell the whole story.In Chapter 2, we computed long-term impact via the LDV. Using thatapproach, the long-run impact of war on CO2 emissions per capita is givenby [coefficient of war/(1 – coefficient of lagged emissions)] × change inwar, where the change in war is 1 (a change from 0 to 1). This computationindicates that the long-run effect on CO2 emissions per capita is 7.46 timeslarger than the short-run effect. In the long run, involvement in a war reducesemissions by 8.5%, an involvement in war fought at home reduces emissionsby 5.67%, and an involvement in wars fought abroad reduces emissions by12.01%. If we only use the significant coefficients in the computation, weget 20.66% reduction in emissions for any war, 12.53% for a war fought athome, and 26.56% for a war fought abroad.

In Models C3 and C4, war at home or abroad, war at home, andwar abroad do not affect CO2 emissions per capita for the LDC sample.These results are discussed in the main text.

The results for the control variables in Table 9.1 are the same across all fourmodels. These results are also consistent with theoretical expectations andthe empirical results in previous studies of the determinants of CO2 emis-sions per capita. A rise in democracy reduces CO2 emissions per capita.The effect of the LDV is positive, indicating that CO2 emissions exhibitinertia. Rises in trade openness and year have no effect on CO2 emis-sions. A rise in population density raises CO2 emissions per capita. Thecoefficients of real gdppc and real gdppc squared indicate the existenceof an EKC that turns down at 14,545 constant 1996 international dollars forthe all-countries sample, and at 20,000 constant 1996 international dollarsfor the LDC sample. These turning points are in the range of the resultsreported in the literature. For example, Cole et al. (1997) reported a turningpoint of $25,100 for CO2 emissions per capita, and Moomaw and Unruh(1997) found a turning point of $18,333 for these emissions.


As for the deforestation models in Table 9.2, the R2 values are in the range0.25–0.29, which indicates a reasonably good model fit for a cross-sectionaldesign such as ours. This R2 value is smaller than that for CO2 because thedeforestation model is cross-sectional due to data availability and does notinclude country fixed effects, year, and lagged environmentt−1. Theaverage VIF is 6.3 and individual VIFs are much below the threshold valueof 10. Hence, multicollinearity is not a concern in this model.

In Model D1, the effect of war at home or abroad on deforestation isstatistically insignificant. In Model D2, the coefficient of war at home isnegative and significant, whereas the coefficient of war abroad is positiveand significant. We have discussed the interpretations in the main text. Forthe LDCs, Models D3 and D4 yield results that are similar to those for thefull sample. As expected, the war coefficients appear considerably larger insize in Models D3 and D4 compared with those values in Models D1 and D2.To compute the sizes of effects for the war variables, we implement the sameprocedure as in the CO2 case and we discuss the results in the main text.

The effects of the control variables fall within the range of possibilitiesdiscussed in the literature. A rise in democracy raises the rate of deforesta-tion for all four models in Table 9.2. For the all-countries sample, the rateof deforestation exhibits a significant nonlinear pattern due to real GDPper capita, but not an EKC. The rate of deforestation falls as real gdppcrises; it then grows as real gdppc rises above a threshold. For the LDCsample, we also do not find an EKC for deforestation. Our results supportthose studies that did not find the EKC (e.g., Shafik and Bandyopadhyay,1992; Barbier, 2001) and disagree with those that found the EKC effect(Bhattarai and Hammig, 2001). A rise in trade openness reduces the rateof deforestation, whereas a rise in population density increases the rateof deforestation.

Additional Analyses

We first reestimate our models while assuming that the dependence of envi-ronmental degradation on real GDP per capita is cubic, not quadratic (ashypothesized by the EKC theory). Table 9.A1 presents results for CO2 emis-sions per capita (Models C5 and C6) and deforestation (Models D5 and D6).For CO2 emissions per capita, the results for the effects of war and controlvariables are similar to those in Table 9.1 in terms of signs, significance, andsizes, indicating the robustness of our CO2 models. The coefficients of someof the real GDP per capita terms differ from Table 9.1, as could be expected.


Table 9.A1. War, cubic GDP per capita, CO2 emissions per capita, and deforestation

All countries LDCs All countries LDCs

(C5) (C6) (D5) (D6)

WAR AT HOME −0.027∗ −0.021 −0.532∗∗ −0.633∗∗

(0.017) (0.017) (0.278) (0.291)WAR ABROAD −0.051∗∗ 0.005 0.798∗∗ 1.022∗∗∗

(0.028) (0.020) (0.412) (0.452)DEMOCRACY −0.0015∗ −0.0027∗∗∗ −0.117∗∗∗ −0.095∗∗∗

(0.001) (0.001) (0.035) (0.033)REAL GDPPC 0.00012∗∗∗ 0.00010∗∗∗ 0.00018 −0.00008

(0.00002) (0.00004) (0.0001) (0.0002)REAL GDPPC SQUARED −5.82e-09∗∗∗ −4.87e-09 1.417e-08 4.524e-08

(1.57e-09) (5.25e-09) (1.45e-08) (3.61e-08)REAL GDPPC CUBED 7.85e-14∗∗ 9.69e-14 −6.8145e-13∗∗ −1.3530e-12

(4.16e-14) (2.04e-13) (4.07e-13) (1.25e-12)TRADE OPENNESS −0.0010 −0.0002 0.012∗∗ 0.011∗

(0.001) (0.001) (0.0067) (0.007)POPULATION DENSITY 0.001∗∗∗ 0.001∗∗ −0.001∗ −0.001∗∗

(0.0003) (0.0003) (0.0006) (0.0006)LAGGED CO2 0.866∗∗∗ 0.850∗∗∗

(0.023) (0.034)YEAR 0.001 0.001

(0.001) (0.001)Constant −1.163 −1.345 −1.975∗∗∗ −1.478∗∗∗

(2.129) (2.214) (0.574) (0.489)Observations 3830 3053 204 187R2 0.99 0.99 0.277 0.294


For the all-countries sample (Model C5), the coefficient of real GDP percapita is positive and statistically significant and the coefficient of squaredreal GDP per capita is negative and statistically significant, as in Table 9.1.For the LDCs sample (Model C6), the coefficient of squared real GDP percapita is also negative, but insignificant. The coefficient of the cubic realGDP per capita term is positive in both samples but is significant only forthe all-countries sample.

Model C5 shows that the dependence of CO2 emissions per capita on realGDP per capita may be cubic for the all-countries sample. The emissionsincrease with real GDP per capita up to a local maximum attained at $14,481,after which they decline with real GDP per capita to a local minimum


Table 9.A2. War and CO2 emissions per capita, omitting the lagged dependent variable

All countries LDCs

(C7) (C8) (D7) (D8)

WAR AT HOME OR −0.124∗∗∗ −0.047ABROAD (0.035) (0.028)

WAR AT HOME −0.084∗∗ −0.060∗∗

(0.036) (0.031)WAR ABROAD −0.161∗∗ −0.043

(0.064) (0.041)DEMOCRACY −0.0102∗∗∗ −0.0103∗∗∗ −0.0166∗∗∗ −0.0165∗∗∗

(0.003) (0.003) (0.0028) (0.0028)REAL GDPPC 0.0008∗∗∗ 0.0008∗∗∗ 0.00059∗∗∗ 0.0006∗∗∗

(0.00003) (0.00003) (0.00005) (0.00005)REAL GDPPC SQUARED −1.89e-08∗∗∗ −1.89e-08∗∗∗ −6.55e-09∗∗ −6.64e-09∗∗

(1.13e-09) (1.13e-09) (2.62e-09) (2.63e-09)TRADE OPENNESS −0.0032∗∗ −0.0031∗ −0.0003 −0.0003

(0.002) (0.0016) (0.0017) (0.0017)POPULATION DENSITY 0.0033∗∗∗ 0.0033∗∗∗ 0.0019∗∗∗ 0.0019∗∗∗

(0.0004) (0.0004) (0.0006) (0.0006)YEAR 0.005∗∗ 0.004∗∗ 0.007∗∗∗ 0.007∗∗∗

(0.002) (0.002) (0.002) (0.002)Constant −9.426∗∗ −8.779∗∗ −13.91∗∗ −14.07∗∗∗

(3.79) (3.81) (4.07) (4.08)Observations 3838 3838 3061 3061R2 0.964 0.964 0.951 0.951


attained at $35,481, and then they rise again with real GDP per capita adinfinitum. Model C6 suggests that the cubic relationship does not hold forthe CO2 emissions per capita in the LDC sample.

Models D5 and D6 present the results from including a cubic real GDP percapita term for deforestation. The coefficients of war and control variablesare similar to those in Table 9.2, indicating (again) the robustness of ourmodels. The coefficients of the real GDP per capita variables change, whichis expected. The coefficients of real GDP per capita and real GDP per capitasquared are not statistically significant for either sample. The coefficient ofthe cubic term is only statistically significant for the all-countries sample.Models D5 and D6 indicate that the dependence of the rate of deforestationon real GDP per capita does not follow a cubic relationship.

Finally, Table 9.A2 presents results from estimating the models of CO2

emissions per capita in Table 9.1 without the LDV. We find that the effects


of the war variables agree with those in Table 9.1 in terms of their signs andsignificance, except that now the effect of war fought at home is statisticallysignificant for the LDC sample. The effects of war are now larger in size thanthe effects in the presence of the LDV. This result is expected because theLDV absorbs some of the variations to be explained by other variables in themodel; when it is excluded, this variance is allocated to other variables onthe right-hand side. The results for the effects of the control variables alsoagree with those in Table 9.2, except that now a rise in year has a significantand positive effect across the board.

TEN

Conclusion

We are approaching the end of the intellectual journey we mapped out atthe beginning of our book. Looking back, we have addressed a numberof important questions, crossed various disciplinary boundaries, engagednumerous academic and policymaking controversies, and uncovered theo-retical and empirical linkages among various literatures and topics, some ofwhich have previously received little attention. The preceding eight chaptershave delved into a wide range of issues that pertain to the causes or effects ofdemocratic governance, economic openness, income inequality, economicdevelopment, military violence, and environmental degradation. In eachchapter, we have focused on specific parts of the big picture, holding otherthings in the system constant, and have generated new knowledge, whichwe believe should be very useful to both academics and policy makers.

In this chapter, our goal is to bring together our key findings and drawan elaborate big picture of some of the complex transformations faced inour world today. This approach forces us to ask broader questions and crosschapter boundaries. The insights we thus generate go beyond our findingsfrom the individual chapters. As we relax the ceteris paribus assumptionunderlying each chapter’s analysis and connect the dots across the differ-ent chapters, we see that more processes interact and more disciplinaryboundaries break down, producing new policy trade-offs and solutions.

In this chapter, we seek to accomplish three tasks. Our first task is to takestock of key findings from the various analyses we have conducted. Insteadof simply listing these findings, we fit them into the big picture we employedin Chapter 1 to describe our conceptual framework. This approach providesa visual display of the connections of the key findings in Chapters 2–9.

Our second task in this chapter is to offer caveats regarding our analysesand suggest general directions for future research. Taking a broad view,this manuscript is based on a certain epistemology, research methodology,

292

Conclusion 293

and theoretical thinking. Our investigation is essentially part of an infiniteprocess of pursuing new knowledge. Thus, instead of treating our findingsas the end point, we view our research as part of a larger, perpetual humanendeavor to acquire broader and deeper understanding of sociopolitical–economic interactions between people and between a human being andnature. In this spirit, we highlight possible themes for future research.

Our third task in this chapter is to integrate and synthesize our keyfindings and distill broad policy implications that transcend those identifiedin Chapters 2–9. Our goal is to reveal new linkages, trade-offs, and solutions.This undertaking is only possible if we place the findings from the differentchapters within a larger, unified conceptual framework.

TAKING STOCK OF KEY FINDINGS

In Chapter 1, we developed Figure 1.1 as an intellectual device to moti-vate and organize the social scientific inquiry that was about to unfold inthis book. In taking stock of the key findings from our many analyses, wereturn to this useful device but flesh it out to create Figure 10.1, whichdocuments our findings and illustrates their interconnectedness. In doingso, Figure 10.1 distinguishes among different types of economic opennessto the world economy – trade, foreign direct investment (FDI), and portfo-lio investments – and different types of environmental degradation – CO2

emissions, NOx emissions, land degradation, water pollution, and defor-estation. Figure 10.1 also marks the sign of each causal arrow to indicate thata particular cause–effect relationship is statistically different from no effectand its direction. The figure also takes note of the statistically significantnonlinear and reciprocal effects identified by our analyses. In short, whereasFigure 1.1 provided us with an analytical roadmap at the beginning of ourlong intellectual journey, Figure 10.1 showcases the social scientific insightswe have collected along our way.

Let us first take a look at the role of six forces included in Figure 10.1. Fiveof these forces were dependent variables in different chapters (i.e., we studiedtheir determinants). Democracy was a dependent variable in Chapters 2, 4,and 5; income inequality in Chapter 3; economic development in Chapter 4;military conflict in Chapters 5 and 6; different types of environmentaldegradation were the dependent variables in Chapters 7, 8, and 9; andeconomic openness was one of the key independent variables in Chapters 2,3, 4, 6, and 8.

Within this context, we can group our key findings according to thedependent variables studied, integrate results obtained from the different

Inve

rted U

for C

O 2, NO x, L

and D

egra

tion

Econ

omic

D

evel

opm

ent

Inco

me

Ineq

ualit

y

Envi

ronm

enta

lD

egra

datio

nM

ilita

ry

Con

flict

Econom

icO

pen

nes

sD

emocr

acy

– on

CO2, N

O x, wat

er, la

nd+

on de

fore

statio

n an

d risi

ng in

open

ec

onom

y dem

ocra

cies

– fo

r tra

de+

for F

DIx

for p

ortfo

lio

+ or

–, d

epen

ding

on tr

ade s

ecto

r

– fo

r tra

de, p

ortfo

lio+

for F

DI

++

+ +

Inve

rted U

for L

DCs

– –

– – –

on CO

2 by

all w

ars

+ on

defo

resta

tion

by w

ar at

hom

e–

on de

fore

statio

n by

war

abro

ad

– on

land

– on

defo

resta

tion

in au

tocra

cy

+ on

fore

st in

demo

cracy

Trade

+–

+

+

Figu

re10

.1.

Gra

phic

alla

you

tof

key

fin

din

gs.

294

Conclusion 295

parts of the book, and point out the chapter in which they are based. Indoing so, we focus on the key determinants, not on all the variables includedin each of our models. As we shall see, our key findings suggest a number ofuneasy trade-offs for policymakers who seek to meet multiple objectives.

Determinants of Democracy

Figure 10.1 demonstrates that the level of democracy in a country is affectedby a number of forces that act in different directions. FDI, economic devel-opment, and the diffusion of democratic ideas promote democracy. A risein FDI flows into a country increases its level of democracy, but this pos-itive effect declines over time and approaches zero (Chapter 2). A rise ineconomic development increases the level of democracy, supporting themodernization thesis (Chapters 2 and 4). The flow of democratic ideas intoa country continuously increases its level of democracy (Chapters 2 and 4).

Democracy, however, faces challenges posed by the forces of trade, port-folio investment, income inequality, and military conflict, as indicated inFigure 10.1. Increasing trade and portfolio investments reduce the level ofdemocracy in a country. The negative effect of portfolio investments ondemocracy grows over time (i.e., further reduces the level of democracy ina country as we progress with time; see Chapter 2). Countries embroiledin interstate military disputes can expect to see their levels of democracydecrease (Chapter 6). As the level of income inequality in a country rises,its level of democracy declines (Chapter 2).

Finally, the level of democracy in a country exhibits path dependence; thecurrent level of democracy is positively affected by its own previous value.In other words, democracy is a relatively slow-changing force, suggestingthat attempts to change its level “overnight” may not be successful due toentrenched interests seeking to sustain the political status quo (Chapters 2,4, and 5).

Determinants of Income Inequality

Like democracy, the level of income inequality in a country is influenced bya number of factors (Chapter 3). As shown in Figure 10.1, a rise in the level oftrade openness of a country reduces its income inequality. Income inequalityalso falls as the level of democracy in a country rises, suggesting a positiverole for democracy if the goal is to increase egalitarianism. In contrast, a risein FDI flows into a country raises its income inequality, implying an uneasytrade-off between attracting FDI and achieving greater income equality.


The positive effect of FDI on inequality, however, is attenuated in moredemocratic countries.

The effect of economic development on inequality is complex; it followsa nonlinear, inverted U-shaped curve. As a country develops economicallyand its income per capita rises, the level of its income inequality rises untilits income per capita reaches a fairly high threshold of about $6,500. Afterthis income per capita threshold is crossed, further increases in the level ofeconomic development in a country reduce its level of income inequality.

Like democracy, income inequality also exhibits path dependence, wherethe current value of inequality is positively associated with its previous value.This tendency reflects two primary forces. Rich and powerful actors tend toreject public policies that redistribute wealth from the rich to the poor andtend to support policies that increase their own share of the national pie.In addition, poor people face economic, educational, social, and politicalobstacles when they try to rise out of poverty; these forces create incomeinequality inertia, working against abrupt changes.

Determinants of Economic Development

As shown in Figure 10.1, the level of economic development in a countryis influenced by competing forces, including political regime type, tradeopenness, and path dependence (Chapter 4).

Starting with negative influences, the level of economic developmentdeclines as the level of democracy in a country rises. This finding suggeststhat democracy is not optimal for promoting economic development, whichchallenges the belief of many policymakers. In contrast, the level of economicdevelopment rises with trade openness. This finding supports the claimof neoclassical economics that trade is the engine of economic growth.However, as we shall see in the next subsection, this positive effect alsoentails a cost.

Like democracy and income inequality, development also exhibits inertiaor persistence over time. Economic development changes slowly over timedue to two types of determinants that tend to change slowly over time. Onetype of determinant includes class structure and institutional characteristicssuch as the rule of law and the extent and enforcement of private propertyrights. A second type includes determinants that essentially cannot be dras-tically changed unless some extraordinary events such as natural disastersor conquests by other countries occur (e.g., geography, climate, resourceendowments). Thus, the finding that economic development exhibits inertiasuggests that poor and rich nations are likely to maintain their current iden-tities well into the future.

Conclusion 297

Determinants of Interstate Military Conflict

Figure 10.1 demonstrates that interstate military conflict is subject to com-peting influences from democracy (Chapter 5) and trade (Chapter 6). A risein the level of democracy shared by the two countries in a dyad reduces theprobability of dyadic conflict, supporting the so-called democratic-peacehypothesis. With this result in mind, two things are worth noting. First,the finding is dyadic, not monadic. Although the United States andDenmark, for example, are both democracies, the former clearly engagesin more military conflicts than the latter. Yet our findings suggest that theUnited States and Denmark are less likely to engage in military conflict thanthe United States and Iran. Second, this finding is probabilistic, not deter-ministic, indicating that countries are less likely to exhibit military disputeswhen their level of joint democracy rises, not that they will never engage inmilitary interstate disputes.

Chapter 6 suggests that the effect of bilateral trade on interstate mili-tary conflict is much more complex than has been realized thus far in theliterature. Figure 10.1 shows that the effect of increases in bilateral tradeon the likelihood of dyadic military conflict varies across bilateral imports,exports, and economic sectors. This finding refutes general statements,which are often made in the literature in international relations, that “tradecauses peace” or “trade causes conflict.” We find that rises in the flows ofagriculture/fishery and energy imports reduce the probability of militaryconflict initiation in a dyad, whereas rises in the flows of energy export andmanufactured products increase the probability of conflict.

Military conflict, like our other dependent variables, exhibits path depen-dence. Dyads that have been at peace more years are less likely to exhibitmilitary conflict in the present. This finding suggests that countries maydevelop trust in one another following some long period of peace and maythen learn to resolve their conflicts without resorting to military force.

Determinants of Environmental Degradation

Analyses of the determinants of environmental degradation have pro-duced our most complex findings, as Figure 10.1 demonstrates. Democracy,trade, military conflict, and economic development produce varying effectson environmental degradation, depending on the type of environmentaldegradation studied. As we shall see, this finding implies that the policytrade-offs surrounding environmental issues are also going to be complex.

Specifically, a rise in democracy reduces the levels of CO2 and NOx

emissions per capita in a country, the share of organic pollution in water,


and the share of land degraded in a country. However, a rise in democracyalso speeds up deforestation in the country (Chapter 7). If one expectsthe spread of democracy to benefit the environment, it is important totake into account the trade-off between deforestation and the other aspectsof environmental degradation, because forests play an important role insustaining the environmental integrity of the biosphere.

A rise in trade openness generates even more complicated effects on envi-ronmental degradation than democracy, depending on the type of degrada-tion and the country’s political regime. A rise in trade reduces deforestationin autocracies and land degradation in both democracies and autocracies.However, a rise in trade leads to faster deforestation in democracies andcauses higher NOx emissions per capita in both democracies and auto-cracies (Chapters 7 and 8).

Involvement in interstate and civil wars, regardless of whether the inter-state war is fought at home or abroad, lowers CO2 emissions per capita andintensifies land degradation. A war fought at home raises the rate of defor-estation at home, whereas a war fought abroad reduces this rate (Chapter 9).Of course, we do not argue that war should be used as a policy instrumentfor promoting environmental quality in a country, but the fact that war mayperversely slow these types of environmental degradation demonstrates thecomplexity of our empirical findings.

The effect of an increase in the level of economic development on environ-mental degradation is often nonlinear. CO2 and NOx emissions per capitaand land degradation exhibit a nonlinear, inverted U-shaped curve thatrises with income per capita until reaching a turning point and then falls.The turning points are about $18,000 for CO2, $25,000 for NOx, and $4,000for land degradation. Because most LDCs are far below these income levels,our findings imply that environmental degradation will continue to risefor years to come as these countries continue to develop. Finally, CO2 emis-sions per capita and water pollution exhibit inertia, suggesting that thesetypes of environmental degradation do not decline abruptly (Chapter 7).

Caveats and Future Research

We have uncovered a number of important and interesting interconnectionsamong forces of broad interest, but the quality of our analysis invariablydepends on the conditions related to our current epistemology, researchmethodology, data availability, and theoretical thinking. Therefore, we re-gard the journey we have undertaken in this book as part of an ongoingresearch enterprise.

Conclusion 299

In our view, our research can be fruitfully extended in three directions.First, one could expand the empirical coverage of our samples, constructalternative empirical indicators, collect new data, and apply new estima-tion techniques. Our empirical analyses may benefit from getting data formore years and countries, even though our theoretical arguments and find-ings ought not to be sample-dependent. For example, one might collectmore data and redo our environmental analyses that employed a cross-sectional design. It is also worthwhile to study other indicators of glob-alization such as migration flows and information technology and flows.Future research on income inequality may benefit from employing alterna-tive development indicators, such as the Human Development Indicator,or alternative inequality indicators such as access to medical services andfood. Further research on trade and conflict could benefit from a morerefined sectoral trade breakdown, as well as from better bilateral tradeprice and quantity data used for estimating parameters in our theoreticalmodel. Finally, more refined modeling of democracy and conflict dynamicsin a context of a simultaneous equations model could also be beneficial,which requires techniques that are currently unavailable in the case of ourmodel.

Second, because many of the relationships we study involve multiplecausal mechanisms that operate in parallel, future research may seek toassess the validity of each specific mechanism. For example, whereas wemodeled the net effect of economic openness on democracy, future researchmay attempt to calibrate the relative importance of the channels drivingthis net effect. The same idea applies to the net effect of democracy on theenvironment, the net impact of trade on the environment, and the net effectsof democracy and development on each other. These assessments shouldhelp us evaluate the strength of the competing arguments by facilitating theaccumulation of theoretical and empirical knowledge.

Third, we have connected many topics that so far have been largely stud-ied separately, but we have not considered all the possible interconnections.For example, we did not analyze the impact of FDI on the environment.We treated economic openness as exogenous, but extant research suggeststhat it is affected by political forces such as democratic governance, mil-itary conflict, economic forces such as development, and environmentalforces such as resource endowments. Future research may also go beyondthe topics we covered in this book. For example, one may study the impactof labor mobility across countries on income inequality. Finally, new link-ages may emerge from unpacking aggregate concepts. For example, onecould study whether different types of democracies, such as presidential or


parliamentary systems, or different categories of autocracies, perform dif-ferently in promoting environmental quality.

We believe future research in these three general directions could pro-duce interesting and useful insights for both academics and policymakers.More important, the suggested research extensions reflect the spirit run-ning throughout our book. That is, to understand the impact of complextransformations in our world today, we have to cross disciplinary bound-aries, integrate previously ignored connections, and create a relatively morerealistic, albeit more complex, bigger picture. As existing scholarly pur-suits become more and more specialized, the lessons they generate tendto become increasingly narrow, because scholars tend to view their ownrelationship of interest as the first-order effect and to brush aside otherinterconnections as negligible second-order effects. The danger we all facein taking this approach is that we fail to see the possibility that a usefullesson based on a narrow context may imply costly trade-offs in a broadercontext.

We believe our efforts in this book help reveal some of the inherent ten-sions, complexities, and interconnectedness that operate among our forces.These interactions, illustrated in Figure 10.1, involve direct, indirect, com-peting, and nonlinear interconnections; they demonstrate various tensionsand trade-offs among competing objectives such as economic develop-ment versus environmental quality, economic development versus equi-table income distribution, prodevelopment economic liberalization versusprodemocracy political liberalization, and pro-openness economic liberal-ization versus environmental quality. Further demonstrating and explainingthese tensions, complexities, and interconnectedness is our final task in thisbook.

TENSIONS, COMPLEXITIES, AND INTERCONNECTEDNESS

The preceding two sections summarized our key findings and suggestedfuture research. This section integrates our findings from a public policy per-spective. The reader may recall that the last section of each chapter discussedimplications for public policy, under the assumption that policymakerssought to achieve only one or two goals. For example, in Chapter 2 weassumed that policymakers seek to promote democracy, and in Chapter 4we assumed they merely seek to promote both democracy and economicdevelopment. In reality, policymakers often seek to pursue multiple goalsat the same time. In the parlance of Figure 10.1, we assume that these con-current goals include raising the level of democracy, encouraging economic

Conclusion 301

development, promoting a more equitable income distribution, reducingvarious types of environmental degradation, and strengthening nationalsecurity.

In general, the competing effects included in Figure 10.1 suggest thatpolicymakers who seek to meet several goals at the same time will experiencepolicy tensions and have to confront trade-offs due to the interconnected-ness among various forces in the international system of polity, economy,and environment. A cursory look at our system indicates that as the numberof goals increases, a discussion of the policy implications often becomes verycomplicated. For our purpose, we can illustrate the gist of this argument byconsidering several relatively simple examples in which policymakers seek topursue only two goals at a time. However, before we discuss these examplesis a word of caution: As we will see, even these seemingly simple examplesgenerate a fair amount of complexity, which cannot be helped. In otherwords, complexity is perhaps the most pronounced feature of the interna-tional system of polity, economy, and environment studied in this book.

DEMOCRACY AND EQUITY

Figure 10.1 shows that increasing democracy reduces income inequality,and reducing income inequality increases democracy. Because the two goalssupport each other, there seems to be no tension, but democracy and equityalso interact indirectly with other forces in our system. Increasing democ-racy reduces economic development. A rise in economic development, inturn, increases income inequality if countries are located to the left of theturning point of the inverted U-shaped curve linking development to in-equality.

Consider next the effect of economic openness on the relationship be-tween democracy and income inequality. A rise in trade openness reducesincome inequality and democracy. However, a rise in FDI openness increasesboth these forces. Thus, increasing both trade openness and FDI open-ness at the same time presents tension. Increasing portfolio investmentopenness seems benign, because reducing it promotes democracy withoutraising inequality. However, portfolio investment can finance the accumu-lation of physical capital, facilitating economic development. Thus, reduc-ing portfolio investment, which promotes democracy, may reduce eco-nomic development. Reducing economic development, in turn, underminesdemocracy and raises income inequality if the country is located to the leftof the turning point in the inverted U-shaped curve linking developmentto inequality.


DEMOCRACY AND DEVELOPMENT

We find that a rise in development promotes democracy, whereas a rise indemocracy reduces development. Hence, leaders seeking development mayhave to delay democratization, at least for a while. Promoting development,however, can also conflict with the goal of increasing democracy throughthe channel of income inequality. If a country is located to the left of theturning point on the inverted U-shaped curve that links development andinequality, a rise in development will raise income inequality, which, inturn, will reduce democracy.

Consider now the effect of economic openness on this relationship. Neo-classical economists have long hailed free trade as an engine of development.Our results support this argument, but we also find that a rise in tradereduces democracy, our other goal in this example. Thus, again, we face anunintended policy trade-off.

Next, recall our findings that FDI liberalization promotes democracy,whereas a rise in democracy reduces development. This combination sug-gests that FDI liberalization indirectly impedes our ability to meet the goalsof democracy and development at the same time. Moreover, although notstudied here (and itself a subject of debate), some studies argue that FDIpromotes economic development. The rise in development, in turn, raisesdemocracy, but the rise in democracy may reduce development – suggesting,once again, a policy trade-off.

DEVELOPMENT AND EQUITY

The goals of development and equity are also at tension. Leaders seekingequity and development face a dilemma: inequality rises with developmentuntil an income per capita of about $6,500 is reached and then falls withfurther increases in income per capita. Because the per capita income ofLDCs is typically much smaller than $6,500, their income inequality isexpected to rise as these countries develop.

Democracy further complicates the relationship. As inequality rises,democracy falls. This decline may, in turn, increase development. How-ever, a rise in development will increase inequality if income per capita isless than $6,500. Once income per capita grows greater than $6,500, a risein development reduces inequality. As inequality falls, democracy rises, butpolicymakers are still not entirely out of the woods. As democracy rises,the level of development falls, which may reduce income per capita belowthe $6,500 threshold and may return the nation back to the income range

Conclusion 303

in which development increases inequality, which in turn reduces the levelof democracy.

Turning to openness in Figure 10.1, a rise in trade increases developmentand reduces inequality, supporting both goals. However, trade liberaliza-tion can indirectly increase income inequality through its positive effect ondevelopment. The effect of development on inequality, as noted, is non-linear, increasing inequality until an income per capita of about $6,500 isattained. Thus, development will increase inequality in most LDCs (as inChina, for example). Moreover, if trade liberalization promotes a rise in FDI,which is often observed to be the case, the net effect of the increases in thelevels of trade and FDI openness on inequality becomes unclear, because arise in FDI increases inequality and a rise in trade reduces it.

ENVIRONMENTAL QUALITY AND DEMOCRACY

In this example, we consider leaders who pursue democracy and environ-mental quality. Figure 10.1 suggests that a rise in democracy reduces alltypes of environmental degradation that we studied, other than the rate ofdeforestation. Let us assume, for the moment, that leaders ignore deforesta-tion, coveting more the goals of democracy and environmental quality inthose other areas. Our analysis suggests that this approach is likely to fail.Deforestation, we know, promotes climate change (forests absorb CO2 emis-sions), damages fresh water supply (forests absorb rainwater), and degradeslands (forests prevent land erosion). These damages may progress slowly fora while, but they do accumulate and eventually grow large. Thus, ignoringthe detrimental effect of democracy on deforestation is to postpone the needto deal with the issue. This strategy, however, carries the risk of generatingnonreversible outcomes such as degrading land to the point of becomingpermanently unsuitable for agriculture.

To further complicate things, we find that growing trade opennessincreases deforestation in a democracy, but reduces it an autocracy. Thus,one might be tempted to delay democratization in countries facing severedeforestation, but our results suggest that this strategy would increase othertypes of environmental degradations (because a rise in democracy reducesall the other types of environmental degradation we studied).

ENVIRONMENTAL QUALITY AND DEVELOPMENT

Countries often pursue both development and environmental quality. How-ever, the two goals often conflict with each other. As shown in Figure 10.1,


the effect of development on environmental degradation follows an invertedU-shaped curve. As a country develops, environmental degradation firstrises and only later falls. We find the turning-point income levels for thedifferent environmental attributes that we studied are relatively high andare much larger than the incomes of most, if not all, of the LDCs, wherethe large majority of the world’s population resides. Thus, given the currenttechnological paradigm, policymakers need to decide which goal is moreimportant – promoting development or improving environmental qual-ity. If they choose development first, they will likely see their environmentdegrade for years.

Next, we find that trade liberalization promotes economic development.However, this may not benefit the environment. The effect of a rise intrade on environmental degradation depends on the environmental aspectexamined and the political regime type. In democracies, trade liberalizationintensifies deforestation, which degrades the environment in the long runby reducing fresh water supplies, increasing land erosion, and increasingCO2 stocks, which accelerate global warming and climate change.

DEMOCRACY AND NATIONAL SECURITY

The link between democracy and national security is also not simple. Ifleaders choose to use military force (assuming leaders are rational), wecan assume they deem it to be beneficial for national security. Under thisreading, Figure 10.1 suggests tension in attaining the goals of promotingboth democracy and national security.

We find that a rise in joint democracy reduces the likelihood of militaryconflict, whereas military conflict reduces democracy on both sides of adyad. With this in mind, consider the scenario in which leaders seek topromote democracy while pursuing national security by resorting to force.The goals of national security and democracy will clash as conflict reducesdemocracy. This decline, in turn, may reduce the level of joint democracy,making conflict more likely and further undermining democracy.

TRADE AND SECURITY

One may also consider the decision to liberalize trade to bring about peace,but this approach may fail as well. Our results suggest that leaders shouldscrutinize the composition and pattern of their trade with potential foes,because a rise in some sectoral trade flows makes military conflict morelikely. If peace is a desirable component of national security, bilateral trade

Conclusion 305

in some economic sectors, with some countries, may need to be regulatedand even reduced. Meanwhile, as trade openness rises, democracy declines.This process may cause the level of joint democracy to decline and makemilitary conflict more likely.

ENVIRONMENTAL QUALITY AND NATIONAL SECURITY

Let us now consider the interaction between environmental quality andnational security. In this example, we assume that leaders seek to promoteboth national security and environmental quality. The pursuit of these goalscan generate policy tension. Depending on where the war is fought and theparticular type of environmental degradation considered, war may harmor benefit a country’s environment. Military conflict, however, can alsoharm or benefit the environment indirectly through its effect on democ-racy. Military conflict, we recall, reduces democracy, which, in turn, affectsthe environment. Thus, environmental quality and national security maybecome competing goals, suggesting a potential policy trade-off.

DEMOCRACY AND ECONOMIC OPENNESS REVISITED

Finally, consider tensions that arise from the desire to pursue goals per-taining to the two forces that stand at the center of our book: democraticgovernance and economic openness. In recent decades, many policymakershave sought to promote both economic and political liberalization, assum-ing the two activities improve national welfare that is broadly defined torise with increases in development, equality, environmental quality, andnational security.

Figure 10.1 suggests the risks associated with promoting democracyand economic openness at the same time. Democracy is good for theenvironment, except for deforestation. Democracy also favors equitableincome distribution. These two components increase national welfare.However, a rise in democracy also reduces development – another com-ponent of social welfare. Moreover, if countries are on the right side ofthe turning points on the inverted U-shaped curve linking development toinequality and on the inverted U-shaped curve linking development to envi-ronmental degradation, a fall in development will increase income inequal-ity and environmental degradation, both of which reduce national welfare.

Promoting economic openness as a goal also has competing effects onwelfare. Part of the problem is due to the fact that different componentsof economic openness (e.g., export, import, sectoral trade, FDI, portfolio


investments) exert competing effects on income equality, national security,and environmental quality. Moreover, economic openness affects develop-ment (the fourth component of welfare) both positively, as in neoclassicaleconomics, and negatively, working through its negative effect on democ-racy. The development component, in turn, has implications for both theequality and environmental quality components of welfare.

In all, pursuing both economic liberalization and political liberalizationmay not necessarily increase a multidimensional metric of national welfare.The overall costs of pursuing the two goals together may not outweigh theirbenefits.

Casual observations suggest that political and economic liberalizationshave benefited DCs, but many LDCs have not been as fortunate. In fact, somecountries have given up pursuing at least one of these goals. For example,during the Cold War, a number of countries rejected both goals, thoughtheir experiment failed. China has chosen to promote development but todelay democratization. Russia has considerably scaled back its democracyin recent years and has concentrated on economic development. Malaysiahas decided to reduce its economic openness by imposing capital flowrestrictions. South Korea and Taiwan developed quickly but did so behindhigh trade and currency barriers and a fair amount of autocracy.

Taken together, these stylized facts, as well as some of the two-goal exam-ples discussed earlier, suggest that a one-size-fits-all policy of simultaneouslypromoting political and economic liberalizations, which is often widelyadvocated, may not always be optimal. One should note, however, that wedo not intend to suggest that the goals of promoting democracy and eco-nomic openness themselves are not worthy of pursuing. Rather, we arguethat the inherent complexities and tensions we demonstrated in this bookmay produce some unintended dismal outcome for the simultaneous pur-suit of political and economic liberalization.

FINAL WORDS

This book has demonstrated the growing complexity of formulating publicpolicies to pursue multiple economic, social, political, and environmentalgoals in an interconnected system. However, the overall picture is even morecomplicated than the one we painted. Because all countries face these verychoices, and they may not all value the various conflicting goals in the sameway, policy coordination among countries adds yet another layer of com-plexity – one we have not considered in this book. The collective-action

Conclusion 307

problem entrenched in the anarchy of the international system makes thenational-level policymaking endeavor all the more demanding and chal-lenging, requiring continuous adjustments in the process of meeting one’sown goals.

Leaders are often hard pressed to come up with solutions, and they oftenclaim to have found solutions. However, the truth of the matter is there reallyis no silver bullet. Leaders interested in simultaneously pursuing severalgoals in the system depicted by Figure 10.1 essentially have two options.First, they may have to prioritize and make some tough choices – decidingto pursue goal X at the expense of sacrificing goal Y. Second, in pursuing aparticular policy goal, leaders may have to adopt certain policy measures tominimize the unintended side effects of their effort. For example, when theyopen up the economy to FDI, leaders may want to compensate firms andpeople hurt by the foreign entry to reduce income inequality that rises withFDI. Similarly, if they remove barriers to financial capital flows, they mayhave to strengthen regulation to prevent excessive speculation that leads tofinancial crises, social strife, state repression, and a decline in democracy.

Taking a broader view, one may wonder whether the Westphalian statesystem is up to the task of formulating effective public policies in an inter-connected world. Even if policymakers are well informed about all thecompeting considerations and trade-offs, they may not necessarily be ableto achieve a first best solution, for they may each prioritize conflicting goalsdifferently. This observation suggests that as the world system becomesmore and more interconnected, the pressures for coordinated global gover-nance or even for forming some sort of a world government may grow; butwe believe that this outcome is unlikely anytime soon. In the end, currentpolicymakers have to increasingly confront the daunting challenges andtrade-offs we have demonstrated and those others we have not analyzedhere, which their predecessors only two or three decades ago did not haveto grapple with.

We have arrived at the end of our analytical journey. We hope our readershave benefited from the many messages offered by this book. But if we hadto choose only one message we would like our readers to remember aboutour book, we think it would have to be the following. A tendency existsamong many social scientists to emphasize specialization in a certain disci-pline and to strive for parsimonious theories to explain reality. As we havedemonstrated in this book, however, a disciplinary focus leads to partial,and often erroneous, understandings of how the world works because itignores potentially important processes emphasized by other disciplines.


As the global system of polity, economy, and environment continues tobecome increasingly interconnected, the demands on national leaders andtheir advisors will only increase, requiring them to cross disciplinary bound-aries in order to understand and address the policy trade-offs of even morecomplex and interrelated processes than the ones studied here. This is thetime of Complex Transformations.

References

Achen, Christopher. 1982. Interpreting and Using Regression. Beverly Hills, CA: Sage.Ahluwalia, Montek S., Nicholas G. Carter, and Hollis B. Chinaware. 1979. “Growth and

Poverty in Developing Countries.” Journal of Development Economics 6(3): 299–341.Aitken, Brian, Ann Harrison, and Robert E. Lipsey. 1996. “Wages and Foreign Owner-

ship: A Comparative Study of Mexico, Venezuela, and the United States.” Journal ofInternational Economics 40(3): 345–371.

Alderson, Arthur S., and Francois Nielsen. 1999. “Income Inequality, Development, andDependence: A Reconsideration.” American Sociological Review 64(4): 606–631.

Alesina, Alberto, and Dani Rodrik. 1994. “Distributive Politics and Economic Growth.”The Quarterly Journal of Economics 109(2): 465–490.

Allan, Nigel J. R. 1987. “Impact of Afghan Refugees on the Vegetation Resources ofPakistan’s Hindukush-Himalaya.” Mountain Research and Development 7(3): 200–204.

Allen, Leslie, Stephen L. Harris, Catherine Herbert Howell, K. M. Kostyal, and RickSammon. 1998. The Wonder of the World. Washington, DC: National GeographicSociety.

Almond, Gabriel Abraham. 1990. A Discipline Divided: Schools and Sects in PoliticalScience. Beverly Hills, CA: Sage.

Alvarez, R. Michael. 1997. Endogeneity in Choice Models. Division of the Humanitiesand Social Sciences. California Institute of Technology. http://www.hss.caltech.edu/rma/home.html (June 3, 2000).

Alvarez, R. Michael, and Garrett Glasgow. 2000. “Two-Stage Estimation of NonrecursiveChoice Models.” Political Analysis 8(2): 147–165.

Amemiya, Takeshi. 1967. “The Estimation of Balestra-Nerlove Models.” TechnicalReport No. 4, Institute for Mathematical Studies in the Social Sciences, StanfordUniversity.

Amemiya, Takeshi. 1978. “The Estimation of a Simultaneous Equation GeneralizedProbit Model.” Econometrica 46(5): 1193–1205.

Amin, Samir. 1996. “The Challenge of Globalization.” Review of International PoliticalEconomy 3(2): 216–259.

Anderson, T. W, and Cheng Hsiao. 1982. “Formulation and Estimation of DynamicModels Using Panel Data.” Journal of Econometrics 18(1): 47–82.

309

310 References

Angell, Norman. 1912. The Great Illusion: A Study of the Relation of Military Power inNations to Their Economic and Social Advantages. London: Heinemann.

Antweiler, Werner, Brian R. Copeland, and M. Scott Taylor. 2001. “Is Free Trade Goodfor the Environment?” American Economic Review 91(4): 877–908.

Arad, Ruth W., Seev Hirsch, and Alfred Tovias. 1983. The Economics of Peacemaking: AFocus on the Egyptian-Israeli Situation. New York: St. Martin’s.

Arellano, Manuel, and Stephen Roy Bond. 1991. “Some Tests of Specification for PanelData: Monte Carlo Evidence and an Application to Employment Equations.” Reviewof Economic Studies 58(2): 277–297.

Armijo, Leslie Elliott. 1998. “Mixed Blessing: Foreign Capital Flows and Democracyin Emerging Markets.” Presented at the Annual Meeting of International StudiesAssociation, Minneapolis, MN, March.

Ashley, Richard. 1980. The Political Economy of War and Peace: The Sino-Soviet-AmericanTriangle and the Modern Security Problematique. London: Frances Pinter.

Baldwin, David. 1985. Economic Statecraft. Princeton, NJ: Princeton UniversityPress.

Baran, Paul A. 1973. The Political Economy of Growth. Harmondsworth, UK: Penguin.Barbier, Edward B. 2001. “The Economics of Tropical Deforestation and Land Use: An

Introduction to the Special Issue.” Land Economics 77: 155–171.Barbieri, Katherine. 2002. The Liberal Illusion: Does Trade Promote Peace? Ann Arbor,

MI: University of Michigan Press.Barnet, Richard J., and John Cavanagh. 1994. Global Dreams: Imperial Corporations and

the New World Order. New York: Simon & Schuster.Barrett, Scott, and Kathryn Graddy. 2000. “Freedom, Growth, and the Environment.”

Environment and Development Economics 5: 433–456.Barro, Robert J. 1996. “Democracy and Growth.” Journal of Democracy 1: 1–27.Barro, Robert J. 1998. Determinants of Economic Growth: A Cross-Country Empirical

Study. Cambridge, MA: MIT Press.Barro, Robert J. 2000. “Recent Developments in Endogenous Growth Theory.” In The

Determinants of Economic Growth, edited by M. S. Oosterbaan, Thijs De Buyter VanStevenick, and N. Van Der Windt, 13–38. Boston: Kluwer Academic.

Barro, Robert J., and Jong-Wha Lee. 2000. “International Data on Educational Attain-ment: Updates and Implications.” Manuscript, Harvard University.

Batra, Geeta, and Hong Tan. 1997. Malaysia: Enterprise Training, Technology and Pro-ductivity. Washington, DC: World Bank.

Beck, Nathaniel, and Jonathan N. Katz. 1995a. “Nuisance vs. Substance: Specifying andEstimating Time-Series-Cross-Section Models.” Political Analysis 6: 1–34.

Beck, Nathaniel, and Jonathan N. Katz. 1995b. “What to Do (and Not to Do)with Time-Series Cross-Section Data.” American Political Science Review 89: 634–647.

Beck, Nathaniel, and Jonathan N. Katz. 2004. “Time Series-Cross Section Issues:Dynamics.” Manuscript, Department of Political Science, New York University,and Division of the Humanities and Social Sciences, California Institute of Tech-nology.

Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. “Taking Time Seriously:Time-Series-Cross-Section Analysis with a Binary Dependent Variable.” AmericanJournal of Political Science 42: 1260–1288.

References 311

Beck, Ulrich. 1996. “Kapitalismus Ohne Arbeit.” Der Spiegel 20: 140–146.Bennett, D. Scott, and Allan Stam. 2000a. “Research Design and Estimator Choices in the

Analysis of Interstate Dyads: When Decisions Matter.” Journal of Conflict Resolution44(October): 653–685.

Bennett, D. Scott, and Allan Stam. 2000b. “EUGene: A Conceptual Manual.” Interna-tional Interactions 26: 179–204.

Bennett, D. Scott, and Allan Stam. 2004. The Behavioral Origins of War. Ann Arbor, MI:University of Michigan Press.

Bennett, Marlin R., Alan Rosenblatt, and Jianxin Wang. 1992. “The Visible Hand: TheUnited States, Japan, and the Management of Trade Disputes.” International Interac-tions 17: 191–213.

Berge, Erling. 1994. “Democracy and Human Rights: Conditions for SustainableResource Utilization.” In Who Pays the Price? The Sociocultural Context of Environ-mental Crisis, edited by B. R. Johnson, 187–193. Covelo, CA: Island Press.

Bhagwati, Jagdish. 1994. “Globalization, Sovereignty and Democracy.” In Democracy’sVictory and Crisis: Nobel Symposium, edited by A. Hadenius, 263–281. Cambridge,UK: Cambridge University Press.

Bhagwati, Jagdish. 2000. The Wind of the Hundred Days: How Washington MismanagedGlobalization. Cambridge, MA: MIT Press.

Bhattarai, Madhusudan, and Michael Hammig. 2001. “Institutions and the Environ-mental Kuznets Curve for Deforestation: A Cross-Country Analysis for Latin America,Africa and Asia.” World Development 29(6): 995–1010.

Bienen, Henry. 1971. The Military Modernization. Chicago: University of Chicago Press.Birdsall, Nancy. 1998. “Life is Unfair: Inequality in the World.” Foreign Policy 112(Sum-

mer): 76–83.Blanchard, Olivier. 2000. Macroeconomics. Upper Saddle River, NJ: Prentice Hall.Bliss, Harry, and Bruce Russett. 1998. “Democracy Trading Partners: The Liberal Con-

nection, 1962–1989.” Journal of Politics 60: 1126–1147.Blomstrom, Magnus, and Ari Kokko. 1996. “The Impact of Foreign Investment on Host

Countries: A Review of the Empirical Evidence.” Policy Research Paper 1745, WorldBank, Washington, DC.

Blomstrom, Magnus, Robert E. Lipsey, and Mario Zejan. 1992. “What Explains Devel-oping Country Growth.” NBER Paper 4132, National Bureau of Economic Research,Cambridge, MA.

Boix, Carles. 1998. Political Parties, Growth and Equality: Conservative and Social Demo-cratic Economic Strategies in the World Economy. Cambridge, UK: Cambridge Univer-sity Press.

Boix, Carles, and Susan C. Stokes. 2003. “Endogenous Democratization.” World Politics55: 517–549.

Boli, John, and George M. Thomas. 1999. “INGOs and the Organization of WorldCulture.” In Constructing World Culture: International Nongovernmental Organiza-tions Since 1875, edited by J. Boli and G. M. Thomas, 13–50. Stanford, CA: StanfordUniversity Press.

Bollen, Kenneth A. 1979. “Political Democracy and the Timing of Development.” Amer-ican Sociological Review 44: 572–587.

Bollen, Kenneth A. 1983. “World System Position, Dependency, and Democracy: TheCross-National Evidence.” American Sociological Review 48: 468–479.

312 References

Bollen, Kenneth A. 1991. “Political Democracy: Conceptual and Measurement Traps.”In On Measuring Democracy, edited by Alex Inkeles, 3–20. New Brunswick, NJ: Trans-action.

Bollen, Kenneth A., David K. Guilkey, and Thomas A. Mroz. 1995. “Binary Out-comes and Endogenous Explanatory Variables: Tests and Solutions with an Appli-cation to the Demand for Contraceptive Use in Tunisia.” Demography 32(1):111–131.

Bollen, Kenneth A., and Robert W. Jackman. 1985. “Political Democracy and the SizeDistribution of Income.” American Sociological Review 50(4): 438–457.

Bond, Marian E. 1985. “Export Demand and Supply for Groups of Non-Oil DevelopingCountries.” International Monetary Stuff Papers 32: 56–77.

Borenszstein, Eduardo, Jose DeGregorio, and Jong-Wha Lee. 1994. “How Does ForeignDirect Investment Affect Economic Growth?” IMF Paper WP/94/110, InternationalMonetary Fund, Washington, DC.

Borjas, George J. 1987. “Immigrants, Minorities and Labour Market Competition.” In-dustrial and Labour Relations Review 40(3): 382–392.

Borjas, George J. 1990. Friends or Strangers: The Impact of Immigration on the US Economy.New York: Basic Books.

Borjas, George J. 1993. “The Impact of Immigrants on Employment Opportunities ofNatives.” In The Changing Course of International Migration, 191–198. Paris: Organi-sation for Economic Cooperation and Development.

Borner, Silvio, Aymo Brunetti, and Beatrice Weder. 1995. Political Credibility and Eco-nomic Development. New York: Macmillan.

Bornschier, Volker, and Christopher Chase-Dunn. 1985. Transnational Corporations andUnderdevelopment. New York: Praeger.

Boron, Atilio. 1998. “Faulty Democracies? A Reflection on the Capitalist ‘Fault Lines’ inLatin America.” In Fault Lines of Democracy in Post-Transition Latin America, editedby Felipe Aguero and Jeffrey Stark, 41–66. Miami: North-South Centre Press at theUniversity of Miami.

Borrus, Michael, and John Zysman. 1992.“Industrial Competitiveness and AmericanNational Security.” In The Highest Stakes: The Economic Foundations of the NextSecurity System, edited by Wayne Sandholtz, Michael Borrus, John Zysman, KenConca, Jay Stowsky, Steven Vogel, and Steve Weber, 7–54. New York: Oxford UniversityPress.

Brack, Duncan. 1996. International Trade and the Montreal Protocol. London: RoyalInstitute of International Affairs and Earthscan Publications.

Bremer, Stuart A. 1993. “Democracy and Militarized Interstate Conflict, 1816–1965.”International Interactions 18: 231–249.

Brunn, Stanley D., and Thomas R. Leinback. 1991. Collapsing Space and Time: GeographicAspects of Communications and Information. London: HarperCollins Academic.

Bruno, Giovanni S. F. 2005. “Approximating the Bias of the LSDV Estimator for DynamicUnbalanced Panel Data Models.” Economics Letters 87(3): 361–366.

Bryan, Lowell, and Diana Farrell. 1996. Market Unbound. New York: John Wiley.Bueno de Mesquita, Bruce. 1981. The War Trap. New Haven, CT: Yale University Press.Bueno de Mesquita, Bruce, and George W. Downs. 2005. “Development and Democ-

racy.” Foreign Affairs 84: 77–86.

References 313

Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith.1999. “An Institutional Explanation of the Democratic Peace.” American PoliticalScience Review 93: 920–941.

Bulte, Erwin H., and Daan P. Van Soest. 2001. “Environmental Degradation in Develop-ing Countries: Households and the (Reverse) Environmental Kuznets Curve.” Journalof Development Economics 65: 225–235.

Bun, Maurice, and Jan F. Kiviet. 2003. “On the Diminishing Returns of Higher OrderTerms in Asymptotic Expansions of Bias.” Economics Letters 79: 145–152.

Burkhart, Ross E., and Michael S. Lewis-Beck. 1994. “Comparative Democracy: TheEconomic Development Thesis.” American Political Science Review 88: 903–910.

Bush, George W. 2001. Remarks by the President to the Council of the Americas.Washington, D.C.: The White House. May 7. http://www.whitehouse.gov/news/releases/2001/05/20010507–6.html.

Bussmann, Margit, Indra de Soysa, and John R. Oneal. 2005. “The Effect of Globalizationon National Income Inequality.” Comparative Sociology 4 (3–4): 285–312.

Buzan, Barry. 1984. “Economic Structure and International Security: The Limit of theLiberal Case,” International Organization 20: 5–41.

Cameron, Rondo. 1997. A Concise Economic History of the World: From Paleolithic Timesto the Present. New York: Oxford University Press.

Cammack, Paul. 1998. “Globalization and the Death of Liberal Democracy.” EuropeanReview 6: 249–263.

Chan, Steve. 1997a. “In Search of Democratic Peace: Problems and Promise.” MershonInternational Studies Review 41(1): 59–91.

Chan, Steve. 1997b. “Democracy and Inequality: Tracking Welfare Spending in Singa-pore, Taiwan, and South Korea.” In Inequality, Democracy and Economic Develop-ment, edited by Manus I. Midlarski, 227–243. Cambridge, UK: Cambridge UniversityPress.

Chan, Sylvia. 2002. Liberalism, Democracy and Development. Cambridge, UK: CambridgeUniversity Press.

Chirot, Daniel. 1977. Social Change in the Twentieth Century. New York: Harcourt BraceJovanovich.

Choucri, Nazli, and Robert North. 1975. Nations in Conflict: National Growth andInternational Violence. New York: Freeman.

Clauge, Christopher, Philip Keefer, Stephen Knack, and Mancur Olson. 1996. “Propertyand Contract Rights in Autocracies and Democracies.” Journal of Economic Growth1(2): 243–276.

Clinton, William J. 1996. Between Hope and History. New York: Random House.CNTS. 2008. Cross-National Time Series Data Archive, 1815–2003. Binghamton, NY:

Databanks International. http://www.databanksinternational.com/.Coe, David T., Elhanan Helpman, and Alexander W. Hoffmaister. 1994. “North-South

Research and Development Spillover.” IMF Paper WP/94/144. International Mone-tary Fund, Washington, DC.

Cohen, Patricia. 2007. “Does Capitalism Lead to Democracy, and How? ” InternationalHerald Tribune, June 13, London.

Cohen, Yousseff. 1985. “The Impact of Bureaucratic-Authoritarian Rule on EconomicGrowth.” Comparative Political Studies 18: 123–136.

314 References

Colaresi, Michael, and William R. Thompson. 2003. “The Economic Development-Democratization Relationship: Does the Outside World Matter?” Comparative Politi-cal Studies 36: 381–403.

Cole, Matthew A., Anthony J. Rayner, and John M. Bates. 1997. “The EnvironmentalKuznets Curve: An Empirical Analysis.” Environment and Development Economics 2:401–416.

Congleton, Roger D. 1992. “Political Institutions and Pollution Control.” Review ofEconomics and Statistics 74: 412–421.

Cox, Robert W. 1996. “Globalization, Multilateralism, and Democracy.” In Approachesto World Order, edited by Robert W. Cox and Timothy J. Sinclair, 524–536. New York:Cambridge University Press.

Cox, Robert W. 1997. “Democracy in Hard Times: Economic Globalization and theLimits to Liberal Democracy.” In The Transformation of Democracy? Globalization andTerritorial Democracy, edited by Anthony McGrew, 49–72. Malden, MA: Blackwell.

Crescenzi, Mark J. 2003. “Economic Exit, Interdependence, and Conflict: An EmpiricalAnalysis.” Journal of Politics 65(3): 809–832.

Crescenzi, Mark J., and Andrew J. Enterline. 1999. “Ripples from the Waves? A Systemic,Time Series Analysis of Democracy, Democratization and Interstate War.” Journal ofPeace Research 36(1): 75–94.

Crutsinger, Martin. 2001. “Environmental Review OK-ed.” Associated Press, April 21.Dahl, Robert A. 1989. Democracy and Its Critics. New Haven, CT: Yale University Press.Dahl, Robert A. 1994. “From Immigrants to Citizens: A New Yet Old Challenge to

Democracies.” Paper presented at the International Conference on the 21st Centuryand Democracy, Seoul, South Korea.

Dailami, Mansoor. 2000. “Financial Openness, Democracy, and Redistributive Policy.”World Bank Policy Research Paper #2372, World Bank, Washington, DC.

Dasgupta, Partha. 1993. An Inquiry into Well Being and Destitution. Oxford: ClarendonPress.

Dasgupta, Partha S. 1995. “Population Poverty and the Local Environment.” ScientificAmerican 272(2): 40–46.

Databanks International. 2008. Cross-National Time Series Data Archive, 1815–2003.Binghamton, NY: Databanks International. http://www.databanks.sitehosting.net/.

Davidson, Russell, and James G. MacKinnon. 1993. Estimation and Inference in Econo-metrics. New York: Oxford University Press.

de Bruyn, S. M., J. C. J. M. Van Den Bergh, and J. B. Opschoor. 1998. “Economic Growthand Emissions: Reconsidering the Empirical Basis of Environmental Kuznets Curves.”Ecological Economics 25: 161–175.

de Montesquieu, Baron [1748] 1900. The Spirit of Laws. New York: Collier.de Soysa, Indra, and Eric Neumayer. 2005. “False Prophet, or Genuine Savior? Assess-

ing the Effects of Economic Openness on Sustainable Development, 1980–1999.”International Organization 59(3): 731–772.

de Tocqueville, A. [1835] 1969. Democracy in America, translated by G. Lawrence.Vol. 1–2, edited by J. P. Mayer. Garden City, NY: Anchor Books.

Deacon, Robert T. 1994. “Deforestation and the Rule of Law in a Cross Section ofCountries.” Land Economics 70: 414–430.

Dean, Judy. 2002. “Does Trade Liberalization Harm the Environment? A New Test.”Canadian Journal of Economics 35: 819–842.

References 315

Deininger, Klaus, and Lyn Squire. 1996. “A New Dataset Measuring Income Inequality.”World Bank Economic Review 10: 565–591.

Deininger, Klaus, and Lyn Squire. 1998. “New Ways of Looking at Old Issues: Inequalityand Growth.” Journal of Development Economics 57(2): 259–288.

Diamond, Jared. 2005a. Collapse: How Societies Choose to Fail and Survive. London:Penguin.

Diamond, Jared. 2005b. “Tokugawa Shoguns vs. Consumer Democracy.” New PerspectiveQuarterly 22: 1–6.

Diamond, Larry. 1992a. “Economic Development and Democracy Reconsidered.” InReexamining Democracy, edited by Gary Marks and Larry Diamond, 93–139. BeverlyHills, CA: Sage.

Diamond, Larry. 1992b. “Economic Development and Democracy Revisited.” AmericanBehavioral Scientist 35: 450–499.

Diamond, Larry. 1995. “Democracy and Economic Reform: Tensions, Compatibilities,and Strategies for Reconciliation.” In Economic Transition in Eastern Europe andRussia: Realities of Reform, edited by Edward P. Lazear, 107–158. Stanford, CA: HooverInstitution Press.

Diamond, Larry. 1999. Developing Democracy: Toward Consolidation. Baltimore: JohnsHopkins University Press.

Diamond, Larry, Juan J. Linz, and Seymour M. Lipset. 1987. “Building and SustainingDemocratic Government in Developing Countries: Some Tentative Findings.” WorldAffairs 150: 5–19.

Diamond, Larry, and Marc F. Plattner, eds. 1994. “Economic Reform and Democracy:Special Edition.” Journal of Democracy 5.

Diaz-Alejandro, Carlos. 1975. “North-South Relations: The Economic Component.”International Organization 29: 213–241.

Dinda, Soumyananda. 2004. “Environmental Kuznets Curve Hypothesis: A Survey.”Ecological Economics 49: 431–455.

Dixon, William J. 1994. “Democracy and the Peaceful Settlement of International Con-flict.” American Political Science Review 88(1): 14–32.

Dixon, William J., and Terry Boswell. 1996. “Dependency, Disarticulation, and Denom-inator Effects: Another Look at Foreign Capital Penetration.” American Journal ofSociology 102(2): 543–562.

Dixon, William, and Bruce Moon. 1993. “Political Similarity and American ForeignTrade Partners.” Political Research Quarterly 46: 5–25.

Dollar, David, and Aart Kraay. 2000. “Growth is Good for the Poor.” DevelopmentResearch Group Papers, World Bank, Washington, DC.

Domke, William K. 1988. War and the Changing Global System. New Haven, CT: YaleUniversity Press.

Dorussen, Han. 2006. “Heterogenous Trade Interests and Conflict: What You TradeMatters.” Journal of Conflict Resolution 50(1): 87–107.

dos Santos, Theotonio. 1970. “The Structure of Dependence,” American Economic Review80: 1151–1161.

Doyle, Michael. 1997. Ways of War and Peace: Realism, Liberalism and Socialism. NewYork: Norton.

Drake, Paul W. 1998. “The International Causes of Democratization, 1974–1990.” In TheOrigins of Liberty: Political and Economic Liberalization in the Modern World, edited

316 References

by Paul W. Drake and Mathew D. McCubbins, 70–91. Princeton: Princeton UniversityPress.

Drucker, Peter F. 1994. “The Age of Social Transformation.” Atlantic Monthly November,53–80.

Dryzek, John S. 1987. Rational Ecology: Environment and Political Economy. Oxford:Blackwell.

Easterly, William. 1999. “Life During Growth.” Journal of Economic Growth 4(3): 239–275.

The Economist. 1994. “Democracy and Growth: Why Voting is Good For You. ” August27.

The Economist. 1995. “Southeast Asian Wars and Timber.” June 17: 35.The Economist. 2003. “The Global Menace of Local Strife,” May 24: 23–25.Environmental Performance Measurement Project. 2002. Environmental Sustainability

Index. http://sedac.ciesin.columbia.edu/es/esi/ESI2002˙21MAR02tot.pdf.FAO (Food and Agriculture Organization). 2000. World Soil Resource Report: Land

Resource Potential and Constraints at Regional and Country Levels. Land andWater Development Division, Food and Agriculture Organization, United Nations,Rome.

FAO. 2002. TERRASTAT. Land and Plant Nutrition Management Service, Food and Agri-culture Organization, United Nations, Rome. http://www.fao.org/ag/agl/agll/terrastat(April 5, 2009).

FAO. 2006. TERRASTAT. Land and Plant Nutrition Management Service. Food andAgriculture Organization, United Nations, Rome. http://www.fao.org/ag/agl/agll/index.stm (May 3, 2006).

Farber, Henry S., and Joanne Gowa. 1995. “Polities and Peace.” International Security23: 123–146.

Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of InternationalDisputes.” American Political Science Review 88(September): 577–594.

Fearon, James, and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” Amer-ican Political Science Review 97: 75–90.

Feenstra, Robert C. 2000. “World Trade Flows.” Center for International Data, Instituteof Governmental Affairs, University of California, Davis, and National Bureau ofEconomic Research, Cambridge, MA.

Feenstra, Robert C., and Hanson Gordon. 1995. “Foreign Direct Investment and RelativeWages: Evidence from Mexico’s Maquiladoras.” NBER Paper 5122. Cambridge, MA:National Bureau of Economic Research.

Feng, Yi. 1997. “Democracy, Political Stability and Economic Growth.” British Journalof Political Science 27: 391–418.

Feng, Yi. 2003. Democracy, Governance, and Economic Performance: Theory and Evidence.Cambridge, MA: MIT Press.

Feng, Yi, and Paul J. Zak. 1999. “The Determinants of Democratic Transitions.” Journalof Conflict Resolution 43(2): 162–177.

Ferreira, Francisco. 1999. Inequality and Economic Performance: A Brief Overview toTheories of Growth and Distribution. Washington, D.C.: World Bank. http://www.worldbank.org/poverty/inequal/index.htm (November 7, 2000).

Findlay, Ronald. 1990. “The New Political Economy: Its Explanatory Power for theLDCs.” Economics and Politics 2: 193–221.

References 317

Flandreau, Mar. 2007. “Do Democracy and Globalization Go Hand in Hand?The Bayesian Heresy.” http://www.bayesianheresy.blogspot.com/2007/01/does-democracy-and-globalization-go.html (January 3, 2007).

Forest Watch Indonesia and Global Forest Watch. 2002. The State of the Forest: Indonesia.Bogot, Indonesia, and Washington DC: Forest Watch Indonesia and Global ForestWatch.

Førland, Tor Egil. 1991. “Economic Warfare and Strategic Goods: A Conceptual Frame-work for Analyzing COCOM.” Journal of Peace Research 28: 191–204.

Frankel, Jeffrey A. 2003. “The Environment and Globalization.” NBER Working Paper10090. Cambridge, MA: National Bureau of Economic Research.

Freedom House. 1999. Freedom in the World. New York: Freedom House.Frieden, Jeffry A., and Ronard Rogowski. 1996. “The Impact of the International Econ-

omy on National Policies: An Analytical Overview.” In Internationalization andDomestic Politics, edited by Robert O. Keohane and Helen V. Milner, 25–47. Cam-bridge, UK: Cambridge University Press.

Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago Press.Fruhling, Hugo. 1998. “Judicial Reform and Democratization in Latin America.” In Fault

Lines of Democracy in Post-Transition Latin America, edited by Felipe Aguero andJeffrey Stark, 237–262. Miami: North South Centre Press at the University ofMiami.

Fukuyama, Francis. 1992. The End of History and the Last Man. New York: Free Press.Garrett, Geoffrey. 1999. “Global Markets and National Politics: Collision Course or

Virtuous Circle?” International Organization 52(4): 787–824.Gartzke, Erik. 1998. “Kant We All Just Get Along?: Motive, Opportunity, and the Origins

of the Democratic Peace.” American Journal of Political Science 42: 1–27.Gartzke, Erik. 2000. “Preferences and the Democratic Peace.” International Studies Quar-

terly 44: 191–210.Gartzke, Erik, and Quan Li. 2003a. “Measure for Measure: Concept Operationalization

and the Trade Interdependence-Conflict Debate.” Journal of Peace Research 40: 555–573.

Gartzke, Erik, and Quan Li. 2003b. “All is Well that Ends Well: A Reply to Oneal, Barbieriand Peters,” Journal of Peace Research 40: 727–732.

Gartzke, Erik, and Quan Li. 2005. “Mistaken Identity: A Reply to Hegre.” ConflictManagement and Peace Science 22: 225–234.

Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. “Investing in the Peace: Eco-nomic Interdependence and International Conflict.” International Organization 55:339–438.

Gasiorowski, Mark J. 1986. “Economic Interdependence and International Conflict:Some Cross National Evidence.” International Studies Quarterly 30: 23–38.

Gasiorowski, Mark J. 1995. “Economic Crisis and Political Regime Change: An EventHistory Analysis.” American Political Science Review 89: 882–897.

Gasiorowski, Mark J. 1997. “Political Regimes and Industrial Wages: A Cross-NationalAnalysis.” In Inequality, Democracy and Economic Development, edited by ManusMidlarski, 244–267. Cambridge, UK: Cambridge University Press.

Gasiorowski, Mark J., and Solomon W. Polachek. 1982. “Conflict and Interdependence:East-West Trade and Linkages in the Era of Detente.” Journal of Conflict Resolution26: 709–729.

318 References

Gates, Scott. 2002. “Empirically Assessing the Causes of Civil War.” Paper presented atthe Annual Meeting of the International Studies Association, New Orleans, March24–27.

Gates, Scott, Torbjørn L. Knutsen, and Jonathon Moses. 1996. “Democracy and Peace:A More Skeptical View.” Journal of Peace Research 33: 1–10.

Germain, Randall. 1997. The International Organization of Credit. Cambridge, UK:Cambridge University Press.

Gill, Stephen. 1995. “Globalization, Market Civilization, and Disciplinary Neoliberal-ism.” Journal of International Studies 24: 399–423.

Gilpin, Robert. 1981. War and Change in World Politics. New York: Cambridge UniversityPress.

Gilpin, Robert. 1984. “Structural Constraints on Economic Leverage: Market-Type Sys-tems.” In Strategic Dimension of Economic Behavior, edited by Gordon H. McCormickand Richard E. Bissel, 105–128. New York: Praeger.

Gilpin, Robert. 1987. The Political Economy of International Relations. Princeton, NJ:Princeton University Press.

Gleditsch, Kristian S., and Michael D. Ward. 2000. “War and Peace in Space and Time:The Role of Democratization.” International Studies Quarterly 44: 1–29.

Gleditsch, Nils Petter. 1995. “Geography, Democracy and Peace.” International Security20: 297–323.

Gleditsch, Nils Petter. 1998. “Armed Conflict and the Environment: A Critique of theLiterature.” Journal of Peace Research 35: 381–400.

Gleditsch, Nils Petter, and Bjorn O. Sverdlop. 2003. “Democracy and the Environment.”In Human Security and the Environment: International Comparisons, edited by EdwardPaper and Michael Redclift, 45–70. London: Elgar.

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, andHavard Strand. 2002. “Armed Conflict 1946–2001: A New Dataset.” Journal of PeaceResearch 39: 615–637.

Gonick, Lev S., and Robert M. Rosh. 1988. “The Structural Constraints of the WorldEconomy on National Political Development.” Comparative Political Studies 21: 171–199.

Goodell, Grace. 1985. “The Importance of Political Participation for Sustained CapitalistDevelopment.” Archives Europeennes de Sociologie 26: 93–127.

Goodell, Grace, and John P. Powelson. 1982. “The Democratic Prerequisites of Develop-ment.” In Freedom in the World: Political Rights and Civil Liberties, edited by RaymondGastil, 167–176. New York: Freedom House.

Goodin, Robert E. 1979. “The Development-Rights Trade-off: Some Unwarranted Eco-nomic and Political Assumptions.” Universal Human Rights 1: 31–42.

Gore, Albert. 1992. Earth in the Balance: Ecology and the Human Spirit. New York:Houghton Mifflin.

Gowa, Joanne. 1994. Allies, Adversaries and International Trade. Princeton, NJ: PrincetonUniversity Press.

Gowa, Joanne, and Edward D. Mansfield. 1993. “Power Politics and International Trade.”American Political Science Review 87: 408–420.

Graham, Edward, and Erika Wada. 2000. “Foreign Direct Investment in Mexico.” WorldEconomy 20(6), 777–797.

Gray, John. 1996. After Social Democracy. London: Demos.

References 319

Green, William H. 2007. Econometric Analysis. Upper Saddle River, NJ: Prentice Hall.Grieco, Joseph M. 1988. “Realist Theory and the Problem of International Cooperation:

Analysis with Amended Prisoner Dilemma Model.” Journal of Politics 50: 600–624.Griffiths, William E., R. Carter Hill, and George G. Judge. 1993. Learning and Practicing

Econometrics. New York: Wiley.Grossman, Gene, and Alan B. Krueger. 1993. “Environmental Impacts of a North Amer-

ican Free Trade Agreement.” In The Mexico-U.S. Free Trade Agreement, edited by PeterM. Garber, 13–56. Cambridge, MA: MIT Press.

Guilkey, David K., Thomas A. Mroz, and David Taylor. 1992. “Estimation and Testing inSimultaneous Equations Models with Discrete Outcomes Using Cross Section Data.”Unpublished manuscript, University of North Carolina, Chapel Hill.

Gujarati, Damodar N. 2002. Basic Econometrics. New York: McGraw-Hill/Irwin.Gurr, Ted Robert, and Keith Jaggers. 1999. Polity98 Project: Regime Characteristics

1800–1998. http://www.systemicpeace.org/polity/polity4.htm (April 5, 2009).Haddad, Mona, and Ann Harrison. 1993. “Are There Dynamic Externalities from Foreign

Direct Investment?” In Market Structure, and Industrial Performance, edited by ClaudioR. Frischtak and Richard S. Newfarmer, 160–272. London: Routledge.

Haggard, Stephan. 1990. Pathways from Periphery: The Politics of Growth in the NewlyIndustrialized Countries. New York: Cornell University Press.

Haggard, Stephan, and Robert R. Kaufman. 1995. The Political Economy of DemocraticTransitions. Princeton: Princeton University Press.

Harbaugh, William T., Arik Levinson, and David Molloy Wilson. 2002. “ReexaminingEmpirical Evidence for an Environmental Kuznets Curve.” NBER Working Paper7711, Cambridge, MA: National Bureau of Economic Research.

Hardin, Garrett. 1968. “The Tragedy of the Commons.” Science 162: 1243–1248.Harris, Jonathan M. 2006. Environmental and Natural Resource Economics: A Contem-

porary Approach. Boston: Houghton Mifflin.Hatzius, Jan. 1997. “Foreign Direct Investment, Capital Formation and Labor Costs:

Evidence from Britain and Germany.” London School of Economics, Paper 336.London: Center for Economic Performance, London School of Economics.

Hayek, Friedrich. 1944. The Road to Serfdom. Chicago: University of Chicago Press.Heckman, James J. 1978. “Dummy Endogenous Variables in Simultaneous Equation

System.” Econometrica 46: 931–959.Heerink, Nico. 1994. Population Growth, Distribution, and Economic Development: The-

ory, Methodology and Empirical Results. Berlin: Springer-Verlag.Hegre, Havard. 2004. “Size Asymmetry, Trade, and Militarized Conflict.” Journal of

Conflict Resolution 48: 403–429.Hegre, Havard. 2005. “Identifying How Trade Matters in Empirical Studies of Interstate

Conflict.” Conflict Management and Peace Science 22(3): 217–224.Heilbronner, Robert L. 1974. An Inquiry into the Human Prospect. New York: Norton.Held, David. 1991. “Democracy, the Nation State and the Global System.” In Political

Theory Today, edited by David Held, 197–235. Stanford, CA: Stanford UniversityPress.

Held, David. 1992. “Democracy: From City States to a Cosmopolitan Order?” PoliticalStudies 40: 13–40.

Held, David, and Anthony McGrew, eds. 2007. Globalization Theory: Approaches andControversies. Cambridge, UK: Polity Press.

320 References

Held, David, Anthony McGrew, David Goldblatt, and Jonathan Perraton. 1999. GlobalTransformations: Politics, Economics, Culture. Cambridge, UK: Polity Press.

Held, David, Anthony McGrew, David Goldblatt, and Jonathan Perraton. 2009.“Researching Globalization,” Global Transformations. http://www.polity.co.uk/global/research.asp (March 29, 2009).

Helliwell, John F. 1994. “Empirical Linkages Between Democracy and EconomicGrowth.” British Journal of Political Science 24: 225–248.

Heston, Alan, Robert Summers, and Bettina Aten. 2002. Penn World Table Version 6.1,Center for International Comparisons at the University of Pennsylvania (CICUP),October.

Hewlett, Sylvia Ann. 1979. “Human Rights and Economic Realities: Tradeoffs in Histor-ical Perspective.” Political Science Quarterly 94: 453–473.

Hewlett, Sylvia Ann. 1980. The Cruel Dilemmas of Development: Twentieth Century Brazil.New York: Basic Books.

Higgins, Matthew, and Jeffrey G. Williamson. 1999. “Explaining Inequality the WorldRound: Cohort Size, Kuznets Curves, and Trade Openness.” NBER Paper 7224. Cam-bridge, MA: National Bureau of Economic Research.

Higgs, Robert. 1988. Crisis and Leviathan: Critical Episodes in the Growth of AmericanGovernment. New York: Oxford University Press.

Hillmann, H. C. 1940. “Analysis of Germany’s Foreign Trade and the War,” Economica7: 66–88.

Hirschman, Albert. [1945] 1980. National Power and the Structure of Foreign Trade.Berkeley, CA: University of California Press.

Hirst, Paul. 1997. “The Global Economy: Myth and Realities.” International Affairs 73:409–425.

Hirst, Paul, and Grahame Thompson. 1996. Globalization in Question: The InternationalEconomy and the Possibilities of Governance. Cambridge, UK: Polity Press.

Hobson, John A. 1902[1948]. Imperialism, A Study. London: Allen and Unwin.Hsiao, Cheng. 1997. “Statistical Properties of the Two Stage Least Squares Estimator

Under Co-integration.” Review of Economic Studies 64: 385–398.Huber, Evelyne, Dietrich Rueschemeyer, and John D. Stephens. 1993. “The Impact of

Economic Development on Democracy.” Journal of Economic Perspective 7: 71–85.Hufbauer, Gary. 2005. Economic Sanctions Reconsidered. Washington, DC: Institute for

International Economics.Hufbauer, Gary, Daniel Esty, Diana Orejas, Luis Rubio, and Jeffrey Schott. 2000. “NAFTA

and the Environment: Seven Years Later.” Policy Analysis in International Economics,No. 61, Institute for International Economics, Washington, DC.

Hughes, Barry. 2000. Continuity and Change in World Politics: Competing Perspectives.Upper Saddle River, NJ: Prentice Hall.

Human Development Report. 1999. United Nations Development Program. New York:Oxford University Press.

Human Development Report. 2000. United Nations Development Program. New York:Oxford University Press.

Huntington, Samuel. 1968. Political Order in Changing Societies. New Haven, CT: YaleUniversity Press.

Huntington, Samuel. 1984. “Will More Countries Become Democratic?” Political ScienceQuarterly 99: 193–218.

References 321

Huntington, Samuel. 1991. The Third Wave: Democratization in the Late TwentiethCentury. Norman, OK: University of Oklahoma Press.

Ichino, Andrea. 1993. “The Economic Impact of Immigration on the Host Country.”In Migration Policies in Europe and the United States, edited by Giacomo Luciani,145–162. Dordrecht, Netherlands: Kluwer.

Im, Hyug Baeg. 1987. “The Rise of Bureaucratic Authoritarianism in South Korea.”World Politics 38: 231–257.

Im, Hyug Baeg. 1996. “Globalization and Democratization: Boon Companions orStrange Bedfellows?” Australian Journal of International Affairs 50: 279–291.

Inter-American Development Bank. 1998. Facing Up to Inequality in Latin America.Washington, DC: Inter-American Development Bank.

International Financial Statistics CD-Rom. 2000. Washington, DC: International Mon-etary Fund.

IPCC. (Intergovernmental Panel on Climate Change). 2001. Special Report on TheRegional Impacts of Climate Change: An Assessment of Vulnerability, Intergovernmen-tal Panel on Climate Change. United Nations Environmental Programme and WorldMeteorological Organization. http://www.ipcc.ch/ (April 5, 2009).

IPCC. 2007a. Climate Change 2007: The Physical Science Basis, Summary for Policymak-ers, Intergovernmental Panel on Climate Change. United Nations Environmental Pro-gramme and World Meteorological Organization. http://www.ipcc.ch/ (April 5, 2009).

IPCC. 2007b. Climate Change 2007: Climate Change Impacts, Adaptation and Vul-nerability, Summary for Policymakers, Intergovernmental Panel on Climate Change.United Nations Environmental Programme and World Meteorological Organization.http://www.ipcc.ch/ (April 5, 2009).

Islam, Nazrul. 1995. “Growth Empirics: A Panel Data Approach.” Quarterly Journal ofEconomics 110: 1127–1170.

Italianer, Alexander. 1986. Theory and Practice of International Trade Linkage Models.Boston: Kluwer.

Jackman, Robert W. 1993. Power Without Force: The Political Capacity of Nation States.Ann Arbor, MI: University of Michigan Press.

Jaggers, Keith, and Ted Robert Gurr. 1995. “Tracking Democracy’s Third Wave with thePolity III Data.” Journal of Peace Research 32: 469–483.

James, P., E. Solberg, and M. Wolfson. 1999. “An Identified Systemic Model of theDemocracy-Peace Nexus.” Defense and Peace Economics 10: 1–37.

James, P., E. Solberg, and M. Wolfson. 2000. “Democracy and Peace: A Reply to Onealand Russett.” Defense and Peace Economics 11: 215–229.

Jenkins, Rhys. 1996. “Theoretical Perspectives on the Transnational Corporation.” InInternational Political Economy: State-Market Relations in the Changing Global Order,edited by C. R. Goddard, J. T. Passe-Smith, and J. G. Conklin, 439–459. Boulder, CO:Lynne Rienner.

Johnston, Jack, and John DiNardo. 1997. Econometric Methods. New York: McGraw-HillCompanies.

Jones, R. J. Barry. 1995. Globalization and Interdependence in the International PoliticalEconomy. London: Frances Pinter.

Judge, George G., R. Carter Hill, William E. Griffiths, Helmut Lutkepohl, and Tsoung-Chao Lee. 1988. Introduction to the Theory and Practice of Econometrics. New York:Wiley.

322 References

Judson, Katherine A., and Anne L. Owen. 1999. “Estimating Dynamic Panel Models: AGuide for Macroeconomists.” Economic Letters 65: 9–15.

Justman, Moshe, and Mark Gradstein. 1999. “The Industrial Revolution, Political Transi-tion, and the Decline in Inequality in 19th Century Britain.” Explorations in EconomicHistory 36(2): 109–127.

Kant, Immanuel. [1795] 1970. “Perpetual Peace: A Philosophical Sketch.” In Kant’sPolitical Writings, edited by Hans Reiss. Cambridge, UK: Cambridge University Press.

Kant, Immanuel. [1795] 1983. Perpetual Peace and Other Essays on Politics, History, andMorals. Indianapolis, IN: Hackett.

Kasza, Gregory J. 1996. “War and Comparative Politics.” Comparative Politics 35: 355–373.

Keck, Margaret E., and Kathryn Sikkink. 1998. Activists Beyond Borders: Advocacy Net-works in International Politics. Ithaca, NY: Cornell University Press.

Kennedy, Paul. 1987. The Rise and Fall of the Great Powers: Economic Change and MilitaryConflict from 1500 to 2000. New York: Random House.

Kennedy, Peter. 2005. A Guide to Econometrics. Cambridge, MA: MIT Press.Keshk, Omar M. G. 2002. Stata Program for Estimating Simultaneous Equations Model

Using the Madalla Procedure. CDSIMEQ. Columbus, OH: Ohio State University.Keshk, O. M. G., B. M. Pollins and R. Reuveny (2004), “Trade Still Follows the Flag:

The Primacy of Politics in a Simultaneous Model of Interdependence and ArmedConflict,” Journal of Politics, 66:1155–79.

Keylor, William. 2001. The Twentieth Century World: An International History. Oxford:Oxford University Press.

King, Dwight Y. 1981.“Regime Type and Performance.” Comparative Political Studies 13:477–504.

King, Gary. 2001. “Proper Nouns and Methodological Propriety: Pooling Dyads inInternational Relations Data.” International Organization 55(2): 497–507.

King, Gary, and Langche Zeng. 2001. “Explaining Rare Events in International Rela-tions.” International Organization 55: 693–715.

Kiviet, Jan F. 1995. “On Bias, Inconsistency, and Efficiency of Various Estimators inDynamic Panel Models.” Journal of Econometrics 68: 53–78.

Kiviet, Jan F. 1999. “Expectation of Expansions for Estimators in a Dynamic PanelData Model; Some Results for Weakly Exogenous Regressors.” In Analysis of PanelData and Limited Dependent Variables, edited by Cheng Hsiao, M. Hashem Pesaran,Kajal Lahiri, and Lung Fei Lee, 199–295. Cambridge, UK: Cambridge UniversityPress.

Klein, Michael, Carl Aaron, and Bita Hadjimichael. 2001. “Foreign Direct Investmentand Poverty Reduction.” Fiscal Policy Papers 2613. Washington, DC: World Bank.

Kotov, Vladimir, and Elena Nikitina. 1995. “Russia and International EnvironmentalCooperation.” In Green Globe Yearbook of International Cooperation on Environmentand Development, edited by H. O. Bergesen and G. Parmann, 17–27. Oxford: OxfordUniversity Press.

Kummell, Gerhard. 1998. “Democratization in Eastern Europe. The Interaction of Inter-nal and External Factors: An Attempt at Systematization.” East European Quarterly32: 243–267.

Kurzer, Paulette. 1993. Business and Banking: Political Change and Economic Integrationin Western Europe. Ithaca, NY: Cornell University Press.

References 323

Kurzman, Charles, Regina Werum, and Ross Burkhart. 2002. “Democracy Effect on Eco-nomic Growth: A Pooled Time-Series Analysis, 1951–1980.” Studies in ComparativeInternational Development 37: 3–33.

Kuznets, Simon. 1955. “Economic Growth and Income Inequality.” American EconomicReview 45(1): 1–28.

Lai, Brian, and Dan Slater. 2006. “Institutions of the Offensive: Domestic Sources ofDispute Initiation in Authoritarian Regimes, 1950–1992,” American Journal of PoliticalScience 50: 113–126.

Lal, Deepak. 1983. The Poverty of Development Economics. London: Institute of EconomicAffairs.

Lall, Sanjaya. 1974. “Less Developed Countries and Private Foreign Direct Investment:A Review Article.” World Development 2(1): 41–48.

Lall, Sanjaya. 1985. Multinationals, Technology and Exports: Selected Papers. New York:St. Martin’s.

Lappe, Frances, Joseph Collins, Peter Rosset, and Luis Esparza. 1998. World Hunger:Twelve Myths. 2nd Edition. New York: Grove Press.

Lasswell, Harold D. 1977. On Political Sociology. Chicago: Chicago University Press.Lawrence, Robert. 1994. “Multinationals, Trade and Labor.” NBER Paper No. 4836,

National Bureau of Economic Research, Cambridge, MA.Layne, Christopher. 1994. “Kant or Can’t: The Myth of the Democratic Peace.” Interna-

tional Security 19: 5–49.Leitz, Christian. 2004. Nazi Foreign Policy, 1933–1941: The Road to Global War. Oxford:

Routledge.Lemke, Douglas, and William Reed. 1996. “Regime Types and Status Quo Evaluations:

Power Transition Theory and The Democratic Peace.” International Interactions 22:143–164.

Lenin, Vladimir I. [1911] 1993. “Writings on the Commune.” In The Civil War in France:The Paris Commune, 91–132, New York: International Publishers.

Lenin, Vladimir I. 1916. “Imperialism: A Special Stage of Capitalism.” In InternationalPolitical Economy: Perspective on Global Power and Wealth, edited by Jeffry Friedenand David A. Lake, 110–119. New York: St. Martin’s.

Lenski, Gerhard. 1966. Power in Privilege: A Theory of Social Stratification. New York:McGraw-Hill.

Levy, Jack S. 1989. “Domestic Politics and War.” In The Origin and Prevention of MajorWars, edited by Robert I. Rotberg and Theodore K. Rabb, 88–108. Cambridge, UK:Cambridge University Press.

Lewis, Peter. M. 1994. “Economic Statism, Private Capital and the Dilemmas of Accu-mulation in Nigeria.” World Development 22(3): 437–451.

Li, Quan, and Adam Resnick. 2003. “Reversal of Fortunes: Democracy, Property Rightsand Foreign Direct Investment Inflows in Developing Countries.” International Orga-nization 57(1): 1–37.

Li, Quan, and Rafael Reuveny. 2003. “Economic Globalization and Democracy: AnEmpirical Analysis.” British Journal of Political Science 33: 29–54.

Li, Quan, and David Sacko. 2002. “The (IR)Relevance of Interstate Militarized Disputesto International Trade.” International Studies Quarterly 46: 11–44.

Liang, Kung-Yee, and Scott L. Zeger. 1986. “Longitudinal Data Analysis Using General-ized Linear Models.” Biometrika 73: 13–22.

324 References

Lijphart, Arend. 1977. Democracy in Plural Societies: A Comparative Exploration. NewHaven, CT: Yale University Press.

Lin, Justin Y., and Jeffrey B. Nugent. 1995. “Institutions and Economic Development.”In Handbook of Development Economics, edited by Jere Behrman and T. N. Srinivasan,2301–2370. Amsterdam: Elsevier.

Lindblom, Charles E. 1977. Politics and Markets: The World’s Political-Economic Systems.New York: Basic Books.

Lindert, Peter H. 1989. “Modern Fiscal Redistribution: A Preliminary Essay.” Agricul-tural History Paper Series, No. 55. University of California, Davis, CA.

Lindert, Peter H. 1994. “The Rise of Social Spending.” Explorations in Economic History31(1): 1–37.

Linz, Juan J. 1978. The Breakdown of Democratic Regimes: Crisis, Breakdown, and Re-equilibrium. Baltimore/London: Johns Hopkins University Press.

Linz, Juan J. 1994.“Presidential or Parliamentary Democracy: Does It Make a Differ-ence?” In The Failure of Presidential Democracy, edited by Juan J. Linz and ArturoValenzuela, 3–87. Baltimore: Johns Hopkins University Press.

Lipset, Seymour Martin. 1959. “Some Social Requisites of Democracy: Economic Devel-opment and Political Legitimacy.” American Political Science Review 53: 69–105.

Lipset, Seymour Martin. 1981. Political Man: The Social Bases of Politics. Baltimore: JohnsHopkins University Press.

Lipset, Seymour Martin. 1994. “The Social Requisites of Democracy Revisited.” AmericanSociological Review 59: 2–13.

Lipset, Seymour Martin, Kyoung-Ryung Seong, and John Charles Torres. 1993. “AComparative Analysis of the Social Requisites of Democracy.” International SocialScience Journal 16: 155–175.

Londregan, John B., and Keith T. Poole. 1996. “Does High Income Promote Democracy?”World Politics 49: 1–30.

Longworth, Richard C. 1998. Global Squeeze: The Coming Crisis for First World Nations.Chicago: Contemporary Books.

Lucas, Robert E. B., David Wheeler, and Hermamala Hettige. 1992. “Economic Develop-ment, Environmental Regulation and the International Migration of Toxic IndustrialPollution: 1960–1988.” In International Trade and the Environment, edited by PatrickLow, 67–86. Washington, DC: World Bank.

MacDonald, Martha. 1991. “Post-Fordism and the Flexibility Debate.” Studies in PoliticalEconomy 36: 177–201.

Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics.Cambridge, UK: Cambridge University Press.

Mainwaring, Scott. 1993. “Presidential, Multipartisan, and Democracy: The DifficultCombination.” Comparative Political Studies 26: 198–228.

Mankiw, N. Gregory, David Romer, and David D. Weil. 1992. “A Contribution to theEmpirical of Economic Growth.” Quarterly Journal of Economics 107: 69–102.

Mansfield, Edward D., and Rachel Bronson. 1997. “Alliances, Preferential Trading Agree-ments, and International Trade Patterns.” American Political Science Review 87: 624–638.

Mansfield, Edward D., and Jon C. Pevehouse. 2000. “Trade Blocs, Trade Flows, andInternational Conflict.” International Organization 54(4): 775–808.

Mansfield, Edward D., and Brian M. Pollins. 2003. “Interdependence and Conflict:An Introduction.” In Economic Interdependence and International Conflict: New

References 325

Perspectives on an Enduring Debate, edited by E. D. Mansfield and B. M. Pollins,1–30. Ann Arbor: University of Michigan Press.

Mansfield, Edward, and Jack Snyder. 1995. “Democratization and the Danger of War.”International Security 20: 5–38.

Mansfield, Edward, and Jack Snyder. 1996. “The Effects of Democratization on War.”International Security 21: 196–207.

Maoz, Zeev. 2001a. “Democratic Networks: Connecting National Dyadic and SystemicLevels of Analysis in the Study of Democracy and War.” In War in a Changing World,edited by Zeev Maoz and Azar Gat, 143–182. Ann Arbor: University of MichiganPress.

Maoz, Zeev. 2001b. Data Set DYMID Version 1.1. http://spirit.tau.ac.il/∼zeevmoaz.Maoz, Zeev, and Bruce Russett. 1993. “Normative and Structural Causes of Democratic

Peace: 1946–1986.” American Political Science Review 87: 624–638.Markusen, James, and Anthony Venables. 1999. “Foreign Direct Investment as a Catalyst

for Industrial Development.” European Economic Review 43(2): 335–356.Marquand, David. 1997. “The Great Reckoning.” Prospect, July.Marshall, Monty G., and Keith Jaggers. 2000. Polity IV Project: Political Regime

Characteristics and Transitions, 1800–1999. Center for International Developmentand Conflict Management, University of Maryland. http://www.cidcm.umd.edu/inscr/polity/index.htm#data.

Martin, Hans-Peter, and Harold Schumann. 1997. The Global Trap: Globalization andthe Assault on Prosperity and Democracy. London: Zed Books.

Martin, Michael L. 1994. “National Security and Democracy: The Dilemma from aFrench Perspective.” Armed Forces and Society 20: 395–421.

Marwick, Arthur. 1988. “Introduction.” In Total War and Social Change, edited by ArthurMarwick, x–xxi. London: Macmillan.

Marx, Karl. [1867] 1990. Capital Volume 1: A Critique of Political Economy. London:Penguin Classics.

Marx, Karl. [1871] 1993. “The Civil War in France.” In The Civil War in France: TheParis Commune, 23–90, New York: International Publishers.

Matanduno, Michael. 1992. Economic Containment of CoCom and the Politics of East-WestTrade. Ithaca, NY: Cornell University Press.

Mazumdar, Dipak, and Ata Mazaheri. 2002. Wages and Employment in Africa. Aldershot,Hampshire, England: Ashgate.

McMillan, S. M. 1997. “Interdependence and Conflict.” Mershon International StudiesReview 41: 33–58.

McNeill, John Robert. 2001. Something New Under the Sun: An Environmental History ofthe Twenty-Century World. London: W. W. Norton.

Midlarsky, Manus I. 1995. “Environmental Influences on Democracy: Aridity, War-fare and a Reversal of the Causal Arrow.” Journal of Conflict Resolution 39(2): 224–262.

Midlarsky, Manus I. 1998. “Democracy and the Environment: An Empirical Assess-ment.” Journal of Peace Research 35: 341–361.

Midlarsky, Manus I. 2002. “Review of Democracy and Development: Political Institu-tions and Well-Being in the World, 1950–1990, by Adam Przeworski et al., 2000.”Journal of Politics 64: 670–672.

Mill, Stuart J. [1860] 1951. “Considerations on Representative Government.” In Utilita-rianism, Liberty, and Representative Government, edited by H. B. Acton. London: Dent.

326 References

Milner, Helen V., and Robert O. Keohane. 1996. “Internationalization and DomesticPolitics.” In Internationalization and Domestic Politics, edited by Robert O. Keohaneand Helen V. Milner, 3–24. Cambridge, UK: Cambridge University Press.

Minford, P., J. Riley, and E. Nowell. 1997. “Trade, Technology and Labor Markets in theWorld Economy.” Journal of Development Studies 34(2): 1–34.

Mintz, Alex. 1985. “The Military Industrial Complex: American Concepts and IsraeliRealities.” Journal of Conflict Resolution 29(4): 623–639.

Mitchell, Sara McLaughlin, Scott Gates, and Havard Hegre. 1999. “Evolution inDemocracy-War Dynamics.” Journal of Conflict Resolution 43(6): 440–461.

Modelski, George, and Gardner Perry III. 1991. “Democratization in Long Perspective.”Technological Forecasting and Social Change 39: 23–34.

Moomaw, William R., and Gregory C. Unruh. 1997. “Are Environmental Kuznets CurveMisleading Us? The Case of CO2 Emissions.” Environment and Development Eco-nomics 2: 451–463.

Moon, Bruce E. 1991. The Political Economy of Basic Human Needs. Ithaca, NY: CornellUniversity Press.

Moran, Jonathan. 1996. “Contradictions Between Economic Liberalization and Democ-ratization: The Case of South Korea.” Democratization 3: 459–490.

Morrow, James D. 1997. “When Do Relative Gains Impede Trade?” Journal of ConflictResolution 41: 12–37.

Morrow, James D. 1999. “How Could Trade Affect Conflict?” Journal of Peace Research36(4): 481–489.

Morrow, James, Randolph Siverson, and Tressa Tabares. 1998. “The Political Deter-minants of International Trade: The Major Powers 1907–1990.” American PoliticalScience Review 92(3): 649–661.

Mousseau, Michael, and Yuhang Shi. 1999. “A Test for Reverse Causality in the Demo-cratic Peace Relationship.” Journal of Peace Research 36(6): 639–663.

Muller, Edward N. 1988. “Democracy, Economic Development and Income Inequality.”American Sociological Review 53: 50–68.

Muller, Edward N. 1995. “Economic Determinants of Democracy.” American SociologicalReview 60: 966–982.

Muller, Edward N., and Mitchell A. Seligson. 1994. “Civic Culture and Democracy: TheQuestion of Causal Relationship.” American Political Science Review 88(3): 635–652.

Muller, Ronald. 1979. “The Multinational Corporation and the Underdevelopment ofthe Third World.” In The Political Economy of Development and Underdevelopment,edited by Charles K. Wilber, 151–178. New York: Random House.

Nafziger, Wayne. 1997. The Economics of Developing Countries. Upper Saddle River, NJ:Prentice Hall.

Nafziger, Wayne E. 2006. Economic Development. New York: Cambridge University Press.Neumayer, Eric. 2002. “Do Democracies Exhibit Stronger International Environmental

Commitment? A Cross Sectional Analysis.” Journal of Peace Research 39: 139–164.New York Times. 2004. “The Sky is Falling! Say Hollywood and, Yes, the Pentagon.” New

York Times, February 29.Nicholson, Walter. 2005. Microeconomic Theory: Basic Theory and Extensions. Fort Worth,

TX: The Dryden Press.Nickell, Stephen. 1981. “Biases in Dynamic Models with Fixed Effects.” Econometrica

49: 1417–1426.

References 327

Nordlinger, Eric. 1970. “Soldiers in Mufti: The Political Impact of Military Rule uponEconomic and Social Change in Non-Western States.” American Political ScienceReview 64: 1131–1148.

Normand, Charles. 1996. “Health Insurance: A Solution to the Financing Gap?” InMarketing Education and Health in Developing Countries, edited by C. Colclough.,205–221. Oxford: Clarendon Press.

North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance.Cambridge, UK: Cambridge University Press.

North, Douglass C., and Barry Weingast. 1989. “Constitutuions and Commitment:Evolution of Institutions Governing Collective Choice in 17th-Century England.”Journal of Economic History 49(4): 803–832.

O’Donnell, Guillermo. 1973. Modernization and Bureaucratic-Authoritarianism: Studiesin South American Politics. Berkeley: Institute of International Affairs, University ofCalifornia.

O’Donnel, Guillermo. 1979. Modernization and Bureaucratic-Authoritarianism. Berke-ley: University of California Press.

O’Donnell, Guillermo. 1993. “The Browning of Latin America.” New Perspectives Quar-terly 10: 50–53.

O’Donnell, Guillermo. 1994. “Delegative Democracy.” Journal of Democracy 5: 55–69.O’Donnell, Guillermo, Philippe Schmitter, and Laurence Whitehead. 1986. Transitions

from Authoritarian Rule. Baltimore: Johns Hopkins University Press.OECD (Organization for Economic Cooperation and Development). 1994a. The Per-

formance of Foreign Affiliates in OECD Countries. Paris: Organization for EconomicCooperation and Development.

OECD. 1994b. The Environmental Effects of Trade. Paris: Organization for EconomicDevelopment and Cooperation.

Olson, Mancur. 1982. The Rise and Decline of Nations. New Haven, CT: Yale UniversityPress.

Olson, Mancur. 1991. “Autocracy, Democracy and Prosperity.” In Strategy and Choice,edited by Richard J. Zeckhauser, 131–157. Cambridge: MIT Press.

Olson, Mancur. 1993.“Dictatorship, Democracy, and Development.” American PoliticalScience Review 87: 567–576.

Oneal, John R., and Bruce Russett. 1997. “The Classical Liberals Were Right: Democracy,Interdependence and Conflict, 1950–1985.” International Studies Quarterly 41: 267–294.

Oneal, John R., and Bruce Russett. 1999a. “Assessing the Liberal Peace with AlternativeSpecifications: Trade Still Reduces Conflict.” Journal of Peace Research 36: 423–42.

Oneal, John R., and Bruce Russett. 1999b. “Is the Liberal Peace Just an Artifact of ColdWar Interests? Assessing Recent Critiques.” International Interactions 25: 213–241.

Oneal, John R., and Bruce Russett. 1999c. “The Kantian Peace: The Pacific Benefits ofDemocracy, Interdependence, and International Organizations, 1885–1992.” WorldPolitics 52: 1–37.

Oneal, John R., and Bruce Russett. 2000. “Comment: Why ‘An Identified Systemic Modelof The Democratic-Peace Nexus’ Does Not Persuade.” Defense and Peace Economics11: 197–214.

O’Rourke, Kevin. 2001. “Globalization and Inequality: Historical Trends.” NBER PaperNo. 8339, National Bureau of Economic Research, Cambridge, MA.

328 References

Paehlke, Robert. 1996. “Environmental Challenges to Democratic Practice.” In Democ-racy and the Environment: Problems and Prospects, edited by William Lafferty andJames Meadowcroft, 18–38. Cheltenham, England: Edward Elgar.

Panayotou, Theodore. 2000a. “Globalization and Environment.” CID Paper 53, Centerfor International Development and Harvard University, Cambridge, MA.

Panayotou, Theodore. 2000b. “Economic Growth and the Environment.” CID Paper 56,Center for International Development and Harvard University, Cambridge, MA.

Park, Tong, Farid Abolfathi, and Michael Ward. 1976. “Resource Nationalism in For-eign Policy Behavior of Oil Exporting Countries.” International Interactions 2: 247–262.

Parson, Edward. 2002. Protecting the Ozone Layer: Science, Strategy, and Negotiation inthe Shaping of a Global Environmental Regime. Oxford: Oxford University Press.

Payne, Rodger A. 1995. “Freedom and the Environment.” Journal of Democracy 6: 41–55.

Penn World Tables 5.6a. Center for International Economic Comparisons, University ofPennsylvania, Philadelphia, PA. http://pwt.econ.upenn.edu/ (April 5, 2009).

Penn World Tables 6.1. Center for International Economic Comparisons, University ofPennsylvania, Philadelphia, PA. http://pwt.econ.upenn.edu/ (April 5, 2009).

Pindyck, Robert, and Daniel Rubenfeld. 1991. Econometric Models and Economic Fore-casts. New York: McGraw-Hill.

Platner, Marc. 1993. “The Democratic Moment.” In The Global Resurgence of Demo-cracy, edited by Larry Diamond and Marc Plattner, 31–49. Baltimore: Johns HopkinsUniversity Press.

Polachek, Solomon W. 1980. “Conflict and Trade.” Journal of Conflict Resolution 24:55–78.

Polachek, Solomon W. 1992. “Conflict and Trade: An Economic Approach to PoliticalInternational Interactions.” In Economics of Arms Reduction and the Peace Process,edited by Walter Isard and Charles Anderton, 89–120. Amsterdam: Elsevier.

Polachek, Solomon W. 1997. “Why Democracies Cooperate More and Fight Less: TheRelationship Between International Trade and Cooperation.” Review of InternationalEconomics 5: 295–309.

Polachek, Solomon W. 2000. “Trade-Based Interactions: An Interdisciplinary Perspec-tive.” Presidential Address, Peace Science Society. Presented at Yale University, NewHaven, CT, October 28.

Polachek, Solomon W., and Judith A. McDonald. 1992. “Strategic Trade and the Incentivefor Cooperation.” In Disarmament, Economic Conversion, and Management of Peace,edited by Manas Chatterji and Linda Forcey, 273–284. Westport: Praeger.

Polachek, Solomon W., and John Robst. 1998. “Cooperation and Conflict AmongDemocracies: Why Do Democracies Cooperate More and Fight Less?” In The Politi-cal Economy of War and Peace, edited by Murray Wolfson, 127–154. Boston: KluwerAcademic.

Pollins, Brain M. 1989. “Does Trade Still Follow the Flag?” American Political ScienceReview 83(2): 465–480.

Porter, Bruce D. 1994. War and the Rise of the State: The Military Foundations of ModernPolitics. New York: Free Press.

Potter, David, David Goldblat, Margaret Kiloh, and Paul Lewis, eds. 1997. Democratiza-tion. Cambridge, UK: Polity Press.

References 329

Pridham, Goeffery. 1994. “Encouraging Democracy.” In The International Dimension ofDemocratization in Eastern Europe, edited by Goeffrey Pridham, Eric Herring, andGeorge Sanford. New York: St. Martin’s.

Przeworski, Adam. 1985. Capitalism and Social Democracy. Cambridge, UK: CambridgeUniversity Press.

Przeworski, Adam. 1991. Democracy and the Market: Political and Economic Reforms inEastern Europe and Latin America. Cambridge, UK: Cambridge University Press.

Przeworski, Adam, and Fernando Limongi. 1997. “Modernization: Theories and Facts.”World Politics 49: 155–183.

Przeworski, Adam, Michael E. Alvarez, Jose A. Cheibub, and Fernando Limongi. 1996.“What Makes Democracies Endure?” Journal of Democracy 7: 39–55.

Przeworski, Adam, Michael E. Alvarez, Jose A. Cheibub, and Fernando Limongi. 2000.Democracy and Development: Political Institutions and Well-Being in the World, 1950–1990. New York: Cambridge University Press.

Przeworski, Adam, and Fernando Limongi. 2003. “Political Regimes and EconomicGrowth.” In Development and Underdevelopment: The Political Economy of GlobalInequality, edited by Mitchel A.. Seligson and John T. Passe-Smith, 435–448. Boulder,CO: Lynne Rienner.

Pugel, Thomas. 2003. International Economics. New York: McGraw-Hill/Irwin.Putnam, Robert D., Robert Leonardi, and Raffaella Y. Nanetti. 1993. Making Demo-

cracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton UniversityPress.

Pye, Lucian. 1966. Aspects of Political Development. Boston: Little Brown.Quinn, Dennis. 1997. “The Correlates of Change in International Financial Regulation.”

American Political Science Review 91(3): 531–551.Raleigh, Clionadh, and Nils Petter Gleditsch. 2005. War Sites 1946–2003. Centre for the

Study of Civil War, PRIO.Rao, Vaman. 1985. “Democracy and Economic Development.” Studies in Comparative

and International Development 19: 67–81.Rasler, Karen, and William R. Thompson. 2000. “Democratic Peace and the Reversed

Causal Arrow Hypothesis: Sequentially Relating Threat, Democratic Dyads, andReduced Conflict Propensities.” Presented at Midwest Political Science Association,Chicago, April.

Ray, James Lee. 1995. Democracy and International Conflict: An Evaluation of the Demo-cratic Peace Proposition. Columbia: University of South Carolina Press.

Ray, James Lee. 2003. “Explaining Interstate Conflict and War: What Should be Con-trolled For?” Conflict Management and Peace Science 20: 1–32.

Reiter, Dan, and Allan C. Stam. 2003. “Identifying the Culprit: Democracy, Dictatorship,and Dispute Initiation.” American Political Science Review 97: 333–337.

Remmer, Karen. 1998. “Does Democracy Promote Interstate Cooperation? Lessons fromthe Mercosur Region.” International Studies Quarterly 42: 25–51.

Reuveny, Rafael. 1999a. “The Political Economy of Israeli-Palestinian Interdependence.”Policy Studies Journal 27: 643–664.

Reuveny, Rafael. 1999b. “Israeli-Palestinian Economic Interdependence Reconsidered.”Policy Studies Journal 27: 668–672.

Reuveny, Rafael. 2000. “The Trade and Conflict Debate: A Survey of Theory, Evidenceand Future Research.” Peace Economics, Peace Science and Public Policy 6: 23–9.

330 References

Reuveny, Rafael. 2001a. “Bilateral Import, Export, and Conflict/Cooperation Simultane-ity.” International Studies Quarterly 45: 131–158.

Reuveny, Rafael. 2001b.“Disaggregated Trade and Conflict: Exploring Propositions.”International Politics 38: 401–428.

Reuveny, Rafael. 2002. “Economic Growth, Environmental Scarcity and Conflict.” GlobalEnvironmental Politics 2: 83–110.

Reuveny, Rafael, and Heejoon Kang. 1998. “Bilateral Trade and Political Conflict/Cooperation: Do Goods Matter?” Journal of Peace Research 35: 581–602.

Reuveny, Rafael, and Quan Li. 2003. “Economic Openness, Democracy and IncomeInequality: An Empirical Analysis.” Comparative Political Studies 36: 575–601.

Reuveny, Rafael, and Aseem Prakash. 1999. “The Afghanistan War and the Breakdownof the Soviet Union.” Review of International Studies 25: 693–708.

Reuveny, Rafael, and William R. Thompson. 2004. Growth, Trade, and Systemic Leader-ship. Ann Arbor: University of Michigan Press.

Ripsman, Norrin M., and Jewan-Marc F. Blanchard. 1997. “Commercial LiberalismUnder Fire: Evidence from 1914 and 1936.” Security Studies 6: 4–50.

Risse, Thomas, and Kathryn Sikkink. 1999. “The Socialization of International HumanRights Norms into Domestic Practices: Introduction.” In The Power of Human Rights:International Norms and Domestic Change, edited by Thomas Risse, Stephen C. Ropp,and Kathryn Sikkink, 1–38. Cambridge, UK: Cambridge University Press.

Robbins, Donald. 1996. “Evidence of Trade and Wages in the Developing World.” OECDDevelopment Center Technical Papers, No. 119. Paris: Organization for EconomicCooperation and Development.

Roberts, Bryan. 1996. “The Social Context of Citizenship in Latin America.” InternationalJournal of Urban and Regional Research 20(1): 38–65.

Robertson, Roland. 1992. Globalization: Social Theory and Global Culture. London: Sage.Rodrik, Dani. 1997. Has Globalization Gone Too Far? Washington, DC: Institute for

International Economics.Rodrik, Dani. 1998. “Symposium on Globalization in Perspective: An Introduction.”

Journal of Economic Perspectives 12(4): 3–8.Rosato, Sebastian. 2003. “The Flawed Logic of Democratic Peace Theory.” American

Political Science Review 97(November): 585–602.Rosecrance, Richard. 1986. The Rise of the Trading State: Commerce and Conquest in the

Modern World. New York: Basic Books.Rudra, Nita. 2005. “Globalization and the Strengthening of Democracy in the Develop-

ing World.” American Journal of Political Science 49(4): 704–730.Rueschemeyer, Dietrich. 1991. “Different Methods – Contradictory Results? Research

on Development and Democracy.” International Journal of Comparative Sociology 32:9–38.

Rueschemeyer, Dietrich, and Peter Evans. 1985. “The State and Economic Transforma-tion: Towards an Analysis of the Conditions Underlying Effective Intervention.” InBringing the State Back In, edited by Peter Evans, Dietrich Rueschemeyer, and ThedaSkocpol, 69–92. Cambridge, UK: Cambridge University Press.

Rueschemeyer, Dietrich, Evelyne Huber Stephens, and John Stephens. 1992. CapitalistDevelopment and Democracy. Chicago: Chicago University Press.

Rummel, Rudolph J. 1997. Power Kills: Democracy as a Method of Nonviolence. London:Transaction Publishers.

References 331

Russett, Bruce. 1993. Grasping the Democratic Peace. Princeton: Princeton UniversityPress.

Russett, Bruce, and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdepen-dence, and International Organizations. New York: Norton.

Russett, Bruce, John R. Oneal, and David R. Davis. 1998. “The Third Leg of the KantianTripod for Peace: International Organizations and Militarized Disputes, 1950–1985.”International Organization 52(3): 441–467.

Salvatore, Dominick. 1998. International Economics. Upper Saddle River, NJ: PrenticeHall.

Salvatore, Dominick. 2004. International Economics. New York: John Wiley.Sassen, Saskia. 1996. Losing Control: Sovereignty in an Age of Globalization. New York:

Columbia University Press.Sayrs, Lois. 1989. “Trade and Conflict Revisited: Do Politics Matter?” International

Interaction 15: 155–175.Sayrs, Lois. 1990. “Expected Utility and Peace Science: An Assessment of Trade and

Conflict.” Conflict Management and Peace Science 11: 17–44.Schamis, Hector E. 1991. “Reconceptualizing Latin American Authoritarianism in

the 1970s: From Bureaucratic-Authoritarianism to Neoconservatism.” ComparativePolitics 23(2): 201–220.

Scharpf, Fritz. 1991. Crisis and Choice in European Social Democracy. New York: CornellUniversity Press.

Schelling, Thomas C. 1958. International Economics. Boston: Allyn and Bacon.Schmitter, Philippe C. 1996. “The Influence of the International Context upon the

Choice of National Institutions and Policies in Neo-Democracies.” In The Interna-tional Dimensions of Democratization: Europe and the Americas, edited by LaurenceWhitehead, 26–54. Oxford: Oxford University Press.

Schneider, Gerald, Katherine Barbieri, and Nils Petter Gleditsch. 2003. “Does Global-ization Contribute to Peace? A Critical Survey of the Literature.” In Globalizationand Armed Conflict, edited by Gerald Schneider, Katherine Barbieri, and Nils PetterGleditsch, 3–30. Lanham, MD: Rowman & Littlefield Publishers.

Schultz, Cynthia B., and Tamara R. Crockett. 1990. “Economic Development, Democra-tization, and Environmental Protection in Eastern Europe.” Boston College Environ-mental Affairs Law Review 18: 53–84.

Schultz, Kenneth A. 1998. “Domestic Opposition and Signaling in International Crises.”American Political Science Review 92(December): 829–840.

Schumpeter, Joseph A. 1942. Capitalism, Socialism and Democracy. New York: Harper &Row.

Schumpeter, Joseph A. 1950. Capitalism, Socialism and Democracy. New York: Harper &Row.

Schwartz, P., and D. Randall. 2003. An Abrupt Climate Change Scenario and Its Impli-cations for United States National Security. Report Commissioned by the United StateGovernment Department of Defense. http://www.gbn.com/ArticleDisplayServlet.srv?aid=26231.

Scruggs, Lyle A. 1998. “Political and Economic Inequality and the Environment.”Ecological Economics 26: 259–275.

Segal, David R. 1994. “National Security and Democracy in the United States.” ArmedForces and Society 20: 375–393.

332 References

Self, Peter. 1993. Government by the Market?: The Politics of Public Choice. London:Macmillan.

Seligson, Mitchell A., and Edward N. Muller. 1987. “Democratic Stability and EconomicCrisis: Costa Rica, 1978–1983.” International Studies Quarterly 321: 301–326.

Sen, Amartya. 1994. “Liberty and Poverty: Political Rights and Economics.” New Republic210(10): 31–37.

Sen, Gautam. 1984. The Military Origins of Industrialization and International TradeRivalry. New York: St. Martin’s.

Shafik, Nemat. 1994. “Economic Development and the Environmental Quality: AnEconometric Analysis.” Oxford Economic Papers 46: 757–773.

Shafik, Nemat, and Sushenjit Bandyopadhyay. 1992. “Economic Growth and Environ-mental Quality: Time Series and Cross Country Evidence.” Background Paper for theWorld Development Report, WPS0904. Washington, DC: World Bank.

Sheth, D. L. 1995. “Democracy and Globalization in India: Post-Cold War Discourse.”Annals of the American Academy of Political and Social Science 540: 24–39.

Siegle, Joseph T., Michael M. Weinstein, and Morton H. Halperin. 2004. “Why Demo-cracies Excel?” Foreign Affairs 83(September/October): 57–71.

Simpson, Miles. 1990. “Political Rights and Income Inequality: A Cross-National Test.”American Sociological Review 55(5): 682–693.

Singer, David, and John Keating. 1999. “Military Preparedness, Weapon Systems andthe Biosphere: A Preliminary Impact Statement.” New Political Science 21: 325–343.

Sirowy, Larry, and Alex Inkeles. 1991. “The Effects of Democracy on Economic Growthand Inequality: A Review.” In On Measuring Democracy, edited by Alex Inkeles, 234.New Brunswick, NJ: Transaction.

Skocpol, Theda, Ziad Munson, Andrew Karch, and Bayliss Camp. 2001. “Patriotic Part-nerships: Why Great Wars Nourished American Civic Voluntarism.” In Shaped byWar and Trade: International Influences on American Political Development, edited byIra Katznelson and Martin Shefter, 134–180. Princeton: Princeton University Press.

Slaughter, Matthew J., and Phillip Swagel. 1997. Does Globalization Lower Wages andExport Jobs? Washington, DC: International Monetary Fund.

Smith, Adam. 1776. The Wealth of Nations. New York: The Modern Library.Solow, Robert M. 1956. “A Contribution to the Theory of Economic Growth.” Quarterly

Journal of Economics 70: 65–94.Soros, George. 1997. “The Capitalist Threat.” Atlantic Monthly 279(2): 45–58.Souva, Mark, Dale Smith, and Shaun Rowan. 2008. “Promoting International Trade:

The Role of Market Protecting Institutions.” Journal of Politics 70(2): 383–392.Stark, Jeffrey. 1998. “Globalization and Democracy in Latin America.” In Fault Lines of

Democracy in Post Transition Latin America, edited by Felipe Aguero and Jeffrey Stark,67–98. Miami: North-South Centre.

Starr, Harvey. 1991. “Democratic Dominoes: Diffusion Approaches to the Spread ofDemocracy in the International System.” Journal of Conflict Resolution 35(2): 356–381.

State of the World’s Forest Report. 2001. Food and Agriculture Organization, Rome,Italy.

Stern, David I. 1998. “Progress on the Environmental Kuznets Curve?” Environment andDevelopment Economics 3: 173–196.

References 333

Stiglitz, Joseph. 1998. “Towards a New Paradigm for Development: Strategies, Policies, andProcesses.” Prebisch Lecture at UNCTAD, Geneva.

Strand, Havard, Joachim Carlsen, Nils Petter Gleditsch, Havard Hegre, ChristinOrmhaug, and Lars Wilhelmsen. 2005. Armed Conflict Dataset Codebook: Version3–2005. Oslo (PRIO): International Peace Research Institute.

Strange, Susan. 1996. The Retreat of the State: The Diffusion of Power in the WorldEconomy. Cambridge, UK: Cambridge University Press.

Stolper, Wolfgang F., and Paul Samuelson. 1941. “Protection and Real Wages.” Reviewof Economic Studies 9(1): 58–73.

Sullivan, Gerard. 1983. “Uneven Development and National Income Inequality in ThirdWorld Countries: A Cross-National Study of the Effects of External Economic Depen-dence.” Sociological Perspectives 26(2): 201–231.

Suri, Vivek, and Duane Chapman. 1998. “Economic Growth, Trade and Energy: Impli-cations for the Environmental Kuznets Curve.” Ecological Economics 25: 195–208.

Sylwester, Kevin. 2005. “Foreign Direct Investment, Growth and Income Inequalityin Less Developed Countries.” International Review of Applied Economics 19(3):289–300.

Tarkowski, Jacek. 1989. “Old and New Patterns of Corruption in Poland and USSR.”Telos 80: 51–62.

Templeton, Scott R., and Sara J. Scherr. 1999. “Effects of Demographic and RelatedMicroeconomic Change on Land Quality in Hills and Mountains of DevelopingCountries.” World Development 27(6): 903–918.

Thompson, William R. 1996. “Democracy and Peace: Putting the Cart Before the Horse?”International Organization 50(1): 141–174.

Tilly, Charles. 1992. Coercion, Capital and European States, AD, 990–1990. Cambridge,UK: Blackwell.

Torras, Mariano, and James K. Boyce. 1998. “Income, Inequality, and Pollution: AReassessment of the Environmental Kuznets Curve.” Ecological Economics 25: 147–160.

Trent, John E. 1994. “Democracy in Danger.” Presented at the 46th World Congress ofInternational Political Science Association, Berlin.

Tsai, Pan-Long. 1995. “Foreign Direct Investment and Income Inequality: Further Evi-dence.” World Development 23(3): 469–483.

Turner, B. L. 1990. The Earth as Transformed by Human Action. New York: CambridgeUniversity Press.

UNEP (United Nations Environment Programme). 2003. Global Environmental Outlook:Past, Present and Future Perspectives. Nairobi and London: Earthscan Publications.

UNEP. 2006. GEO Data Portal. http: //geodata.grid.unep.ch/ (May 3, 2006).U.S. Department of State. 2007. How Political Democracy and Economic Development Are

Mutually Reinforcing, Paula J. Dobriansky, Under Secretary for Democracy and GlobalAffairs, Remarks at a Freedom House Symposium on Democracy and Development,Washington, DC, January 19.

Van Hanen, Tatu. 1990. The Process of Democratization: A Comparative Study of 147States, 1980–1988. New York: Russak.

Verdier, Daniel. 1998. “Democratic Convergence and Free Trade.” International StudiesQuarterly 42: 1–24.

334 References

Vernon, Raymond. 1971. Sovereignty at Bay. New York: Basic Books.Vogel, Steven. 1992. “The Power Behind Spin-Ons: The Military Implications of Japan’s

Commercial Technology.” In The Highest Stakes: The Economic Foundations of theNext Security System, edited by Wayne Sandholtz, Michael Borrus, John Zysman,Ken Conca, Jay Stowsky, Steven Vogel, and Steve Weber, 55–80. New York: OxfordUniversity Press.

Wade, Robert. 1996.“Globalization and Its Limits.” In National Diversity and Global Ca-pitalism, edited by Susanne Berger and Ronald Core, 60–88. New York: Cornell Uni-versity Press.

Wallerstein, Immanuel. 1974. The Modern World System. New York: Academic Press.Waltz, Kenneth N. 1970. “The Myth of Interdependence.” In The International Corpo-

ration, edited by Charles P. Kindleberger, 205–223. Cambridge, MA: MIT Press.Waltz, Kenneth N. 1979. Theory of International Politics. Reading, MA: Addison-Wesley.Weart, Spencer R. 1998. Never at War: Why Democracies Will Not Fight One Another.

New Haven, CT: Yale University Press.Weiss, Edith Brown, and Harold K. Jacobsen. 1999. “Getting Countries to Comply with

International Agreements.” Environment 41: 16–23.Weitzman, Martin L. 1993. “Capitalism and Democracy: A Summing Up of the Argu-

ments.” In Markets and Democracy: Participation, Accountability and Efficiency, editedby Samuel Bowles, Herbert Gintis, and Bo Gustafsson, 314–335. Cambridge, UK:Cambridge University Press.

Westing, Arthur P. 1990. Environmental Hazards of War: Releasing Dangerous Forces inan Industrialized World. Newbury Park, CA: Sage.

White, Halbert. 1980. “A Heteroskedasticity-Consistent Covariance Matrix Estimatorand a Direct Test for Heteroskedasticity.” Econometrica 48: 817–838.

Whitehead, Laurence. 1986. “International Aspects of Democratization.” In Transitionsfrom Authoritarian Rule, edited by Guillermo O’Donnell, Philippe Schmitter, andLaurence Whitehead, 3–46. Baltimore: John Hopkins University Press.

Whitehead, Laurence. 1993. “Introduction: Some Insights from Western Social Theory.”World Development 21: 1253–1260.

Whitehead, Laurence. 1996. “Three International Dimensions of Democratization.”In The International Dimensions of Democratization, edited by Laurence Whitehead,3–24. Oxford: Oxford University Press.

Wiggins, Vince. 1999. Comparing XTGLS with Regress Cluster ( ). STATA Corporation.http://www.stata.com/support/faqs/stat/xtgls rob.html.

Williamson, Jeffrey G. 1997. “Globalization and Inequality, Past and Present.” WorldBank Research Observer 12(2): 117–135.

Wolf, Martin. 2000. “The Big Lie of Global Inequality.” Financial Times, February 9,p. 25.

Wolfson, Murray, Patrick James, and Eric Solberg. 1998. “In a World of Cannibals Every-one Votes for War: Democracy and Peace Reconsidered.” In The Political Economy ofWar and Peace, edited by Murray Wolfson, 154–176. Boston: Kluwer Academic.

Wood, Adrian. 1994. North-South Trade, Employment, and Income Inequality: ChangingFutures in a Skill-Driven World. New York: Oxford University Press.

Wooddall, Pam. 1995. “The World Economy: Financial Markets Survey.” The Economist,October 17, pp. 1–37.

References 335

Wooldridge, Jeffrey. 2002. Introductory Econometrics – A Modern Approach. Cincinnati,OH: South-Western College Publishing.

World Bank. 2002. World Development Indicators. Washington, DC: World Bank.World Bank Poverty Net. 2000. World Bank, Washington, DC [Online]. http://

www.worldbank.org/poverty/inequal.World Resources Institute. 1999. World Resources. New York: Oxford University Press.World Resources Institute. 2001. World Resources. New York: Oxford University Press.Zacher, Mark W., and Richard A. Matthew. 1995. “Liberal International Theory: Com-

mon Threads, Divergence Strands.” In Controversies in International Relation Theory:Realism and the Neoliberal Challenge, edited by Charles W. Kegley, Jr., 107–150. NewYork: St. Martin’s.

Author Index

Aaron, C., 68Abolfathi, F., 162, 179Achen, C., 197Ahluwalia, M. S., 72Aitken, B., 68n12Alderson, A. S., 68n13Alesina, A., 65n5Allan, N. J., 269Allen, L., 280Almond, G. A., 131n9Alvarez, J. A., 52Alvarez, M. E., 28, 91, 93, 97n3, 99n5, 109n6Alvarez, R. M., 91, 127, 153, 157Amemiya, T., 114, 152, 197Amin, S., 31, 33n17Anderson, T. W., 115Angell, N., 13, 158n1Antweiler, W., 248Arad, R. W., 161n7, 181Arellano, M., 116Armijo, L. E., 34, 34n20Ashley, R., 161Aten, B., 231, 259, 283

Baldwin, D., 161n8, 162, 162n10Bandyopadhyay, S., 237, 288Baran, P. A., 67Barbier, E. B., 237, 250, 288Barbieri, K., 160, 161Barnet, R. J., 67Barrett, S., 212Barro, R. J., 56, 89, 94, 95, 110n8, 111Bates, J. M., 237, 287Batra, G., 68n11Beck, N., 50, 59, 82, 87n28, 115, 115n15,

116, 197, 234

Beck, U., 31, 32n12Bennett, D. S., 195–196Bennett, M. R., 162, 179Berge, E., 209Bhagwati, J., 28, 28n4, 29n5, 247n4Bhattarai, M., 250, 288Bienen, H., 96Birdsall, N., 65n4, 67, 72n15Blanchard, J. F., 66, 161Bliss, H., 251Blomstrom, M., 68, 68n11Boehmer, C., 161, 182, 198Boix, C., 64, 97n3, 99n5, 109n6Boli, J., 28, 29n7, 30n9, 56Bollen, K. A., 27n3, 31, 37n21, 49n25,

58n27, 64n3, 65, 82, 153Bond, M., 188n16Bond, S. R., 116Borenszstein, E., 68Borjas, G. J., 70n14Borner, S., 97Bornschier, V., 68n13Boron, A., 31, 33n16Borrus, M., 161, 161n9Boswell, T., 68n13Boyce, J. K., 211, 212Brack, D., 243Bremer, S. A., 128n4Bronson, R., 251Brunetti, A., 97Brunn, S. D., 28, 29n7Bruno, G. S., 115Bryan, L., 31Bueno de Mesquita, B., 89, 93, 129, 131,

164n12Bulte, E. H., 250

337

338 Author Index

Bun, M., 116Burkhart, R. E., 26n2, 27n3, 37n21, 49,

49n25, 56, 58, 61, 82, 91, 97, 97n2, 112,112n11, 116

Bush, G. W., 24Bussmann, M., 68Buzan, B., 161

Cameron, R., 24n1Cammack, P., 31, 31n11, 32n13Camp, B. SeeCarlsen, J., 283Carter, N. G., 72Cavanagh, J., 67Chan, S., 63–65, 65n6, 127, 128n3Chan, Sylvia, 90, 90n1Chapman, D., 248Chase-Dunn, C., 68n13Cheibub, J. A., 28, 91, 93, 97n3, 99n5, 109n6Chinaware, H. B., 72Chirot, D., 95Choucri, N., 161Clauge, C., 95Clinton, W. J., 24Coe, D. T., 68n11Cohen, Y., 94, 95Colaresi, M., 97n2Cole, M. A., 237, 287Collins, J., 64n3Congleton, R. D., 209, 211, 212Copeland, B. R., 248Cox, R. W., 31, 31n11, 32n12, 33n16, 33n17Crescenzi, M. J., 133, 161Crockett, T. R., 208Cross-National Time Series Data Archive,

111, 111n9Crutsinger, M., 243

Dahl, R. A., 26n2, 31, 32n14, 92Dailami, M., 29n5, 30n10, 32n12Dasgupta, P., 73, 268Davidson, R., 117Davis, D. R., 135, 136, 143de Bruyn, S. M., 237de Soysa, I., 68, 248de Tocqueville, A., 92Deacon, R. T., 269Dean, J., 250DeGregorio, J., 68Deininger, K., 56, 65, 71, 72, 73n16, 80, 81,

81n19, 83n23, 85

Diamond, J., 205, 206, 226Diamond, L., 24n1, 28, 29n6, 29n7, 31,

31n11, 32n13, 49n24, 56, 58, 63, 92,109n6, 247n4

Diaz-Alejandro, C., 246DiNardo, J., 116Dinda, S., 231n3, 245n3, 250, 283Dixon, W. J., 68n13, 126, 128n4, 129, 135,

230, 251Dollar, D., 68Domke, W. K., 161n7Dorussen, H., 159, 162, 179, 180, 195,

198dos Santos, T., 161Downs, G. W., 89, 93Doyle, M., 129n6Drake, P. W., 26, 28, 29n6, 48, 63n2Drucker, P. F., 31, 32n12Dryzek, J. S., 210, 210n1

Easterly, W., 71, 80, 81, 81n19, 85Enterline, A. J., 133Environmental Performance Measurement

Project, 222, 238Eriksson, M., 232, 283Esparza, L., 64n3Esty, D., 243Evans, P., 28, 29n6

Farber, H. S., 130Farrell, D., 31Fearon, J. D., 129, 230Feenstra, R. C., 68n12, 194, 195Feng, Y., 18, 26n2, 37n21, 56, 60, 91, 95, 97,

109n6, 110, 111, 111n10, 119Ferreira, F., 63Findlay, R., 95Flandreau, M., 24Food and Agriculture Organization, 230Forest Watch Indonesia, 206Forland, T. E., 161n8, 162Frankel, J. A., 243Freedom House, 53–59, 61, 212Frieden, J. A., 34, 34n19, 34n20Friedman, M., 89Fruhling, 28, 30n9Fukuyama, F., 89

Garrett, G., 34, 34n19Gartzke, E., 130, 144, 151, 155, 161, 182,

198

Author Index 339

Gasiorowski, M. J., 26, 27n3, 37, 37n21, 48,49n24, 49n26, 53, 65, 65n5, 65n6, 97n2,109n6, 111n10, 112, 113, 119, 134, 154,161n7, 161n8, 162, 179

Gates, S., 131, 131n9Germain, R., 69Gill, S., 31, 31n11, 33n15, 33n17Gilpin, R., 161, 161n9Glasgow, G., 91, 127Gleditsch, K. S., 127Gleditsch, N. P., 128n3, 160n6, 209–212,

232, 283Global Forest Watch, 206Goldblat, D., 1, 2, 17, 23, 24n1, 62, 67, 69–70Gonick, L. S., 97n2Goodell, G., 95Goodin, R. E., 95, 96Gordon, H., 68n12Gore, A., 205, 226Gowa, J., 130, 247, 251Graddy, K., 212Gradstein, M., 65Graham, E., 68n12Gray, J., 31, 31n11Greene, W. H., 59, 115n17Grieco, J. M., 247Griffiths, W. E., 49Grossman, G., 248, 250Guilkey, D. K., 153, 157Gujarati, D. N., 197Gurr, T. R., 47, 81

Haddad, M., 68n11Hadjimichael, B., 68Haggard, S., 26, 31, 32n13, 34, 34n20, 48,

93, 94, 113Halperin, M. H., 11, 90Hammig, M., 250, 288Harbaugh, W. T., 250n5Hardin, G., 210Harris, J. M., 232n5, 241n1, 243, 250n5, 273Harris, S. L., 280Harrison, A., 68n11, 68n12Hatzius, J., 67Hayek, F., 89Heckman, J. J., 140, 152Heerink, N., 73Hegre, H., 131, 195, 196, 283Heilbronner, R. L., 210Held, D., 1, 2, 17, 23, 24n1, 28, 28n4, 31,

31n11, 62, 63n2, 67, 69–70, 247n4

Helliwell, J. F., 26, 37n21, 91, 97Helpman, E., 68n11Heston, A., 231, 259, 283Hettige, H., 248Hewlett, S. A., 94, 95Higgins, M., 71, 72, 80Higgs, R., 131Hill, C., 49Hillmann, H. C., 180n15Hirsch, S., 161, 181Hirschman, A., 13, 158n2, 161n8, 162n10,

162n11, 181, 246Hirst, P., 34, 34n18Hoffmaister, A. W., 68n11Howell, C. H., 280Hsiao, C., 115, 116Huber, E., 92Hufbauer, G., 243, 246Hughes, B., 62Human Development Report, 64n3, 67, 69Huntington, S., 24n1, 26, 26n2, 28, 30n10,

48, 89, 93

Ichino, A., 70n14Im, H. B., 28, 28n4, 31, 32n13, 32n14,

33n16, 33n17, 63n2, 93Inkeles, A., 64n3, 65n7Inter-American Development Bank, 66Intergovernmental Panel on Climate

Change, 229, 266, 273, 280International Financial Statistics CD-Rom,

149International Herald Tribune, 89, 90Islam, N., 110, 111, 118Italianer, A., 193, 194

Jackman, R. W., 64n3, 65, 94Jacobsen, H. K., 209Jaggers, K., 47, 81, 109, 109n7, 283James, P., 127, 127n1, 134–135, 144, 150Jenkins, R., 67Johnston, J., 116Jones, R. J. B., 31, 31n11, 34, 34n18Judge, G. G., 49Judson, K. A., 115Justman, M., 65

Kang, H., 162n10, 162n11Kant, I., 12, 13, 28, 29n5, 30n10, 129,

158n1Karch, A., 131

340 Author Index

Kasza, G. J., 131Katz, J. N., 50, 59, 82, 87n28, 115, 115n15,

116, 197, 234Kaufman, R. R., 26, 31, 32n13, 34, 34n20,

48, 93, 113Keating, J., 269Keck, M. E., 28, 29n7, 30n9, 60Keefer, P., 95Kennedy, P., 114, 181, 197Keohane, R. O., 34, 34n20Keshk, O. M., 157, 161, 195, 197Keylor, W., 181Kiloh, M., 24n1King, D. Y., 96King, G., 59, 85n27Kiviet, J. F., 115–116Klein, M., 68Knack, S., 95Knutsen, T. L., 131n9Kokko, A., 68n11Kostyal, K. M., 280Kotov, V., 208Kraay, A., 68Krueger, A. B., 248, 250Kummell, G., 28, 29n7, 31, 33n17Kurzer, P., 34, 68Kurzman, C., 97Kuznets, S., 68, 72, 84

Lai, B., 196Laitin, D. D., 230Lal, D., 94Lall, S., 67Lappe, F., 64n3Lasswell, H. D., 130Lawrence, R., 68Layne, C., 130, 131Lee, J., 56, 68, 111Lee, T., 49Leinback, T. R., 28, 29n7Leitz, C., 180n15Lemke, D., 130Lenin, V. I., 13, 89, 158n2Lenski, G., 65Leonardi, R., 92Levinson, A., 250n5Levy, J. S., 128n3Lewis, P., 24n1, 73Lewis-Beck, M. S., 26n2, 27n3, 37n21, 49,

49n25, 56, 58, 61, 82, 91, 97n2, 112,112n11, 116

Li, Q., 50, 63n2, 95, 97n2, 109n6, n7,111n10, 112, 112n11, 112n13, 113,114n14, 119, 161, 165n13, 182, 189n17,198, 233, 235, 247, 283

Liang, K., 59Lijphart, A., 26Limongi, F., 27, 28, 49n24, 52, 90, 91, 91n1,

93, 97n3, 99n5, 109n6Lin, J. Y., 24Lindblom, C. E., 31, 31n11Lindert, P. H., 65, 65n5Linz, J. J., 26, 92, 93Lipset, S. M., 26n2, 28, 28n4, 37n21, 65n5,

92, 93, 96, 97n2, 102, 237Lipsey, R. E., 68, 68n12Londregan, J. B., 26n2, 47, 47n22, 51, 82,

97n2, 109n7, 111n10Longworth, R. C., 31, 32n12, 34, 34n20Lucas, R., 248Lutkepohl, H., 49

MacDonald, M., 31, 32n13MacKinnon, J. G., 117Maddala, G. S., 140, 145, 152, 153, 157, 197Mainwaring, S., 26Mankiw, N. G., 110–111Mansfield, E., 47, 47n22, 48, 82, 130n8,

160n6, 195, 247, 251Maoz, Z., 128n4, 129, 134, 136, 150Markusen, J., 68n11Marquand, D., 31, 32n12Marshall, M. G., 109, 109n7, 283Martin, H., 31, 32n12, 33n15, 247n4Martin, M. L., 131n9Marwick, A., 131n10Marx, K., 89Matthew, R. A., 63, 63n1Mazaheri, A., 68n12Mazumdar, D., 68n12McDonald, J. A., 162McGrew, A., 1, 2, 17, 23, 62, 67, 69–70McMillan, S. M., 159n4, 160n6McNeill, J. R., 269, 270Midlarsky, M. I., 90, 91n1, 131n9, 211, 212Mill, J. S., 89Milner, H. V., 34, 34n20Minford, P., 66n9Mintz, A., 131n9Mitchell, S. M., 131Modelski, G., 131Montesquieu, Baron de, 13, 158n1

Author Index 341

Moomaw, W. R., 237, 287Moon, B., 65, 251Moon, B. E., 65n5Moran, J., 31, 32n12Morrow, J., 50, 129, 161, 165n13, 182,

189n17, 247, 251Moses, J., 131n9Mousseau, M., 130n7, 133, 133n12Mroz, T. A., 153, 157Muller, E. N., 26, 26n2, 28, 28n4, 31, 32n12,

37n21, 49n25, 50, 65, 65n5, 82, 85, 93,112n11, 114n14

Muller, R., 67Munson, Z., 131

Nafziger, W., 67, 69, 80n18, 96Nanetti, R. Y., 92Neumayer, E., 211, 248Nicholson, W., 188n16Nickell, S., 115Nielson, F., 68n13Nikitina, E., 208Nordlinger, E., 96Normand, C., 69North, D., 89, 95North, R., 161Nowell, E., 66n9Nugent, J. B., 24

O’Donnell, G., 26, 31, 94, 32n13, 33n16Olson, M., 38, 89, 94–95, 247Oneal, J. R., 18, 28, 29n5, 47, 47n22, 48,

50, 68, 126, 127, 127n1, 128n3, 128n4,129n6, 134–136, 143–145, 150, 151, 155,157, 159n4, 161, 195, 198, 233, 235, 247,251

Opschoor, J. B., 237Orejas, D., 243Organization of Economic Cooperation and

Development, 68n11, 232n5, 241n1Ormhaug, C., 283O’Rourke, K., 70Owen, A. L., 115

Paehlke, R., 210Panayoto, T., 231n3, 232n6, 241n1, 245n3,

250, 259, 283Park, T., 162, 179Parson, E., 243Payne, R. A., 208Penn World Tables, 109–111, 149, 151

Perraton, J., 1, 2, 17, 23, 62, 67, 69–70Perry III, Gardner, 131Pindyck, R., 81n22Plattner, M., 28, 28n4, 92Polachek, S. W., 130, 158n3, 159, 161,

161n7, 162–164, 164n12, 169, 170, 179,186, 192, 193n18

Pollins, B. M., 160n6, 161, 165n13, 189n17,195, 197, 247

Poole, K. T., 26n2, 47, 47n22, 51, 82, 97n2,109n7, 111n10

Porter, B. D., 131, 131n10Potter, D., 24n1Powelson, J. P., 95Prakash, A., 131n10Pridham, G., 26Przeworski, A., 27, 27n3, 28, 30n10, 31,

33n17, 49n24, 52, 90, 91, 91n1, 93, 95,97n3, 99n5, 109n6

Pugel, T., 232n5, 241n1, 241n2, 250n5Putnam, R. D., 92Pye, L., 95

Quinn, D., 69

Raleigh, C., 283Randall, D., 280Rao, V., 94Rasler, K., 133Ray, J. L., 128n3, 195Rayner, A. J., 237, 287Reed, W., 130Reiter, D., 196Remmer, K., 251Resnick, A., 95, 247Reuveny, R., 18, 50, 63n2, 97n2, 109n6,

109n7, 111n10, 112, 112n11, 112n13, 113,114n14, 119, 131n10, 159n4, 160n6, 161,161n9, 162n10, 165n13, 170, 189n17, 193,194n19, 195, 197, 197n20, 232, 233, 235,283

Riley, J., 66n9Ripsman, N. M., 161Risse, T., 28, 29n7, 30n9, 60Robbins, D., 67Roberts, B., 28, 30n8, 30n9Robertson, R., 31, 32n14Robst, J., 130Rodrik, D., 31, 32n12, 62, 65, 65n5, 65n6,

66n9, 67Rogowski, R., 34, 34n19, 34n20

342 Author Index

Romer, D., 110–111Rosato, S., 130Rosecrance, R., 161Rosenblatt, A., 162, 179Rosh, R. M., 97n2Rosset, P., 64n3Rowan, S., 247, 251Rubenfeld, D., 81n22Rubio, L., 243Rudra, N., 29n6Rueschemeyer, D., 27, 28, 29n6, 49n24, 92Rummel, R. J., 128n5Russett, B., 18, 28, 29n5, 47, 47n22, 48, 50,

126, 127, 127n1, 128n3, 128n4, 129,129n6, 134–136, 143–145, 150–151, 155,157, 159n4, 161, 195, 198, 233, 235, 247,251

Sacko, D., 165n13, 189n17, 247Salvatore, D., 67, 241n2Sammon, R., 280Samuelson, P., 66Sassen, S., 31, 33n16Sayrs, L., 159n4, 160n6, 161n8Schamis, H. E., 65n4Scharpf, F., 34, 34n18Schelling, T. C., 162Schmitter, P., 26, 28, 29n5, 29n7, 31, 31n11,

247n4Schneider, G., 160n6Schott, J., 243Schultz, C. B., 208Schultz, K., 129Schumann, H., 31, 32n12, 33n15, 247n4Schumpeter, J. A., 28, 28n4, 89Schwartz, P., 280Scruggs, L. A., 212Segal, D. R., 131n9Self, P., 28, 30n8Seligson, M. A., 37n21, 49n25, 50, 82, 93,

112n11, 114n14Sen, A., 209Sen, G., 161n9Seong, K., 92Shafik, N., 211, 237, 288Sheth, D. L., 28, 30n8Shi, Y., 130n7, 133, 133n12Siegle, J. T., 11, 90Sikkink, K., 28, 29n7, 30n9, 60Simpson, M., 65Singer, D., 269

Sirowy, L., 64n3, 65n7Siverson, R., 50, 129, 165n13, 189n17, 247,

251Skocpol, T., 131Slater, D., 196Slaughter, M. J., 66n9Smith, A., 129, 158n1Smith, D., 247, 251Snyder, J., 47, 47n22, 48, 82, 130n8Solberg, E., 127, 127n1, 134–135, 144,

150Sollenberg, M., 232, 283Solow, R. M., 110, 110n8, 118Soros, G., 62Souva, M., 247, 251Squire, L., 56, 65, 71, 72, 73n16, 80, 81,

81n19, 83n23, 85Stam, A. C., 195–196Stark, J., 28, 30n9Starr, H., 27n3, 28, 30n10, 48, 112, 112n13Stephens, E. H., 27Stephens, J., 27, 92Stern, D. I., 250n5Stiglitz, J., 68Stokes, S. C., 97n3, 99n5, 109n6Stolper, W. F., 66, 66n8Strand, H., 232, 283Strange, S., 69Sullivan, G., 68n13Summers, R., 231, 259, 283Suri, V., 248Sverdlop, B. O., 209–212Swagel, P., 66n9Sylwester, K., 68

Tabares, T., 50, 165n13, 189n17, 247,251

Tan, H., 68n11Tarkowski, J., 31, 33n17Taylor, D., 153, 157Taylor, M. S., 248Templeton, S. R., 284The Economist, 89, 269Thomas, G. M., 28, 29n7, 30n9, 56Thompson, G., 34, 34n18Thompson, W. R., 18, 97n2, 131, 133Tilly, C., 131Torras, M., 211, 212Torres, J. C., 92Tovias, A., 161, 181Trent, J. E., 31, 32n13

Author Index 343

Tsai, P., 68n13Tucker, R., 197Turner, B. L., 270

United Nations Environment Programme,218

Unruh, G. C., 237, 287US Department of State, 90

van den Bergh, J. C. J. M., 237Van Hanen, T., 28, 29n7van Soest, D. P., 250Venables, A., 68n11Verdier, D., 251Vernon, R., 34, 68Vogel, S., 161n9

Wada, E., 68n12Wade, R., 34, 34n18Wallensteen, P., 232, 283Wallerstein, I., 24n1, 31, 33n17Waltz, K. N., 158n2, 161Wang, J., 162, 179Ward, M., 127, 162, 179Weart, S. R., 129Weder, B., 97Weil, D. D., 110–111

Weinstein, M. M., 11, 90Weiss, E. B., 209Weitzman, M. L., 28, 28n4Werum, R., 97Westing, A. P., 268Wheeler, D., 248White, H., 50, 116, 153, 234, 260Whitehead, L., 26, 28, 30n10, 31, 33n16,

63n2Wiggins, V., 50, 82, 116, 153Wilhelmsen, L., 283Williamson, J. G., 70n14, 71, 72, 80Wilson, D. M., 250n5Wolf, M., 62Wolfson, M., 127, 127n1, 134–135, 144,

150Wood, A., 66Wooddall, P., 69Wooldridge, J., 117, 233World Bank, 62, 64, 195World Resources Institute, 229, 282

Zacher, M. W., 63, 63n1Zak, P. J., 26n2, 56, 60Zeger, S. L., 59Zejan, M., 68Zysman, J., 161, 161n9

Subject Index

2SLS. See two-stage least squares2SLS Kiviet. See two-stage least squares

Kiviet estimator

3SLS. See three-stage least squares

absence of conflict, 132, 180Afghanistan, 269Africa, 56, 112, 253AH estimator, 115Asia, 48, 60, 112, 181, 218, 253, 260–262autocracy, 1, 11, 12, 16, 25, 29, 36, 47, 82,

90, 93–96, 98, 103, 104, 109, 129–130,132, 134, 195, 206, 208–210, 212, 214,219, 222–227, 230, 238, 240, 244–245,253, 255–258, 260, 273, 298, 300, 303,306

bilateral trade, 14, 158, 160, 161, 163, 165,168–170, 178, 181–184, 189, 193, 243,247, 251, 297, 299, 304

California, 23capital, 10, 28–30, 33, 36, 37, 42, 44, 45, 47,

62–70, 72, 76, 77, 79, 93, 95, 96, 100,110, 249, 301, 306, 307

ceteris paribus, 42, 52, 163, 164, 169, 170,180, 182, 192, 193, 198, 227, 261,292

China, 25, 46, 93, 96, 98, 106, 127, 134,140, 196, 206, 218, 247, 284, 303,306

classical economics, 10climate change, 16, 229, 239, 259, 266–268,

272, 280–281, 303, 304clustering, 49, 59, 82, 116, 145, 153, 197

CO2, 15, 206–207, 212, 214–228, 232–238,259, 267, 271–282, 284–290, 293,297–298, 303

communication, 1, 48, 92, 112, 150comparative advantage, 161, 232, 242, 245composite environmental indicator, 219,

222, 238composition effect, 242control variable, 7, 36, 39, 43, 46–48, 52, 53,

56, 74, 80, 87, 104, 108, 112, 114, 116,118, 126, 128, 141, 146–147, 153–154,173, 174, 178, 182, 195, 198, 202,214–216, 219, 231, 236, 238, 252, 253,258, 260, 261, 273–275, 279, 287–291

country fixed effects, 59, 77, 87, 101, 113,114, 117, 272, 286

country fixed effects estimator, 117, 285covariance matrix, 50, 113, 116, 153COW composite index, 196

datacross section, 25, 49, 53, 110, 115, 140,

153, 214–215, 274, 284, 285, 288panel, 250time series, 49, 53, 56, 101, 105, 115, 140,

152, 214, 215, 233, 260, 274, 284time series cross section, 101, 115

DCs. See developed countriesdeforestation, 15–17, 206, 207, 212,

214–222, 224–229, 233, 234, 236–237,239, 240, 242, 250, 252–259, 261, 263,267–269, 271–274, 277–282, 284–285,287–290, 293, 298, 303–305

demand, 14, 28, 30, 64, 92, 163–166,169–170, 182, 184, 187–193, 195, 229,242

344

Subject Index 345

democracy, 1–6, 8–16, 18–20, 23–65, 70–81,89–119, 125–157, 174, 175, 178, 186,195, 197, 199, 205–238, 243–266, 270,273, 283, 292–300

democratization, 29, 35, 37, 39, 53, 79, 93,99, 104–107, 126–128, 131–133, 138,145, 147, 225, 263, 272, 302, 303,306

dependent variable, 6, 7, 12, 26, 35, 47, 49,50, 53, 56, 71, 80–82, 85–87, 90, 97, 99,100, 108, 115, 126, 145, 146, 173–175,179, 196, 197, 206, 213, 214, 228,233–235, 251, 252, 260, 272, 282, 284,285, 291, 293

developed countries, 16, 32–34, 38, 41, 44,64, 66–70, 73–76, 78, 79, 83–85, 96, 98,101, 240, 241, 253, 256–258, 261, 263,271, 274, 280, 306

display of force, 134, 138, 173, 179, 180, 185,202

domestic conflict, 129dummy variable, 56, 59, 60, 85, 133, 150,

196, 215, 260, 261, 263dyadic, 3, 12–14, 18, 125–128, 132–141,

143–148, 151, 153, 155, 157, 159, 165,172–175, 178, 193, 196, 202, 297

economic growth effect, 245economic openness, 1–6, 8–14, 18, 19, 23,

24, 30–34, 36, 44–46, 60, 62–66, 70–74,77–82, 85, 87, 100, 107, 125, 158, 186,205, 227, 247, 266, 281, 292, 293, 299,301, 302, 305–306

education, 1, 11, 28, 43, 56, 60, 66, 72,73, 92, 100, 110, 111, 118, 208, 215,233

EKC. See environmental Kuznets curveemission, 15, 16, 206, 207, 211–212,

214–219, 224–229, 232–234, 236, 239,248, 259, 267, 268, 270–291, 293, 297,298, 303

endogeneity, 102, 127, 215, 233endogenous variable, 97, 99, 100, 109, 111,

113, 115, 126, 135–138, 140, 145,151–153, 157, 194, 197

environmental degradation, 2, 9, 15, 16,205–225, 228, 231–235, 238–242,244–246, 248–250, 252, 259, 266–268,272–274, 279–281, 283, 285, 288, 292,293, 297, 298, 303–305

environmental economics, 6, 18, 231

environmental Kuznets curve, 19, 68, 77,212–214, 225, 227, 231, 236, 237, 241,245–246, 248, 249, 253, 273, 283

Environmental Performance MeasurementProject, 222, 238

Environmental Sustainability Index, 222,238

Environmental Systems Quality composite,222, 238

EPMP. See Environmental PerformanceMeasurement Project

equality test, 88, 200error, 51, 60, 68error term, 8, 38, 49, 59, 73, 81, 87, 102, 113,

114, 140, 175, 197, 215, 233, 253, 274,285, 286

ESI. See Environmental Sustainability Indexestimator, 43, 46, 49, 53, 59–60, 77, 79,

85–87, 91, 101, 113–117, 153, 157, 197,234, 237, 285

Europe, 3, 44, 48, 60, 65, 112, 206, 253,260–262, 270, 284

exogenous variable, 97, 99, 100, 110, 111,113, 135, 137, 138, 152, 233

export, 14, 29, 47, 66, 111, 147, 149,158–160, 162–173, 175–181, 183–194,197, 198, 200, 232, 233, 245, 248, 250,251, 256, 259, 283, 297, 305

externalities, 162, 258

factors of production, 30, 45, 66, 95, 245famines and human life argument, 244fixed effects, 101, 110, 111, 114–115, 213,

235, 286fixed effects estimator, 59, 77, 85–87, 114,

117, 234Florida, 23foreign direct investment, 10, 11, 19, 25, 28,

36, 37, 40, 44, 47, 62, 66, 72, 79, 81,111, 228, 249, 299

foreign financial capital, 66, 72, 79formal model, 6, 7, 17, 147, 159, 165,

182France, 24, 46, 127, 134, 140, 155, 196Freedom House, 53, 56–59, 61, 212freedom of information channel, 244

GEE. See general estimating equationGeneral Agreement on Trade and Tariffs,

228general estimating equation, 59, 60

346 Subject Index

Gini coefficient, 10, 56, 64, 65, 71, 72, 80,81, 85

global warming, 226, 239, 304global-commons thesis, 243globalization, 1, 4, 8, 18–20, 23–46, 50–53,

62, 76, 79, 228, 240, 266, 299Greece, 77, 96, 269gross domestic product, 37, 47–49, 56, 68,

72, 81, 82, 85, 87, 96, 97, 100, 102–104,108, 110–112, 118, 134, 138, 149–151,154, 178, 195, 199, 202, 214, 231, 232,237, 248, 252, 259–261, 273, 276, 279,283, 286, 288–290

Gulf War, 266, 269

heteroscedasticity, 49, 59, 60, 82, 116, 153,197, 285

homoscedasticity, 59Huber-White robust standard errors, 59, 82,

234, 260, 285hypothesis, 7, 14, 60, 70, 72, 74–76, 82, 92,

117, 132, 134, 150, 153, 160, 163, 168,170–175, 193, 194, 197, 198, 253, 259,261, 297

null, 50, 59, 117, 235hypothesis testing, 60

IMF. See International Monetary FundIm-Pesaran-Shin panel unit root test,

286import, 14, 47, 66, 111, 147, 149, 158–160,

163–173, 175–194, 197, 198, 200, 232,245, 248, 259, 283, 297, 305

income effect, 231, 245, 246income equality, 63, 78, 80, 295, 306income inequality, 2, 6, 10, 18, 28, 32, 40,

43, 46, 56, 60, 62–87, 95, 125, 245, 247,281, 292–296, 299–303, 307

independent variable, 7, 12, 35, 48, 49, 71,87, 90, 99, 135, 152, 173, 197, 213, 214,230, 235, 251, 272, 273, 275, 282

inertia, 37, 49, 52, 84, 110, 112, 118, 119,154, 215, 233, 236, 285, 287, 296,298

inflation, 18, 26, 37, 48, 52, 53, 100, 111,113, 134, 138, 149, 150, 154

international dollars, 72, 82, 102, 231, 236,259, 283, 287

international economics, 18international labor mobility, 66International Monetary Fund, 51, 78

international organization, 18, 32, 78, 79,146, 258

international political economy, 6, 18, 147,158, 185

international trade, 13, 15, 66, 78, 228, 232,239–242, 245–248, 254, 257, 258

inverted U, 231, 245, 246, 250, 283, 296,298, 301, 302, 304

Iraq, 269, 270

joint democracy, 12, 13, 125–127, 130,132–135, 141–147, 151, 251, 297, 304,305

Kuznets curve, 72, 74, 83, 84, 87, 245

lagged, 233, 235, 236, 274, 276, 283, 284lagged dependent variable, 49–51, 53, 82,

87, 101, 110, 114–116, 150, 216, 218,234, 235, 274, 276, 279, 284, 287, 290,291

land degradation, 15, 16, 206, 207, 211,214–216, 219–228, 230, 232–234,236–237, 239, 240, 250, 252, 253,255–259, 261, 263, 273, 293,298

LDCs. See less developed countriesLDV. See lagged dependent variableleast squares dummy variable, 114, 115less developed countries, 16, 32, 33, 38, 39,

41–42, 44–46, 53, 58, 59, 64, 66–70,73–76, 78–79, 83–85, 95, 96, 101, 102,105, 106, 111, 118–119, 239–241, 253,256–258, 261, 263, 265, 270–271,274–280, 287–291, 298, 302–304,306

Levin-Linard, 286long run change in democracy, 103, 104,

117, 118, 235Lorenz curve, 80LSDV. See least squares dummy variable

Maddala, 152correction, 145, 152, 153, 157estimator, 152method, 140, 145, 153

major powers, 133, 136, 146, 150, 196market argument, 243mean, 7, 75, 82, 143, 177, 178, 218, 219, 254,

260, 263, 278Middle East, 112

Subject Index 347

military conflict, 2, 4, 6, 9, 12–15, 18, 19,107, 125, 126, 128, 130–131, 135, 138,145, 147, 158–164, 172, 175, 182–185,205, 212, 232, 258, 293, 295, 297, 299,304, 305

military interstate dispute, 12, 125–128, 133,135, 138, 139, 141–148, 150, 152–157,160, 172–182, 185, 195–202

model variablesAFFINITY, 138, 139, 144, 151, 155, 157AFRICA, 56, 253AGRICULTURE EXPORT, 173, 175, 194,

199AGRICULTURE IMPORT, 173, 175, 194,

199ALLIANCE, 138, 139, 143, 151, 155, 173,

174, 196, 199ALLY, 134ASIA, 56, 253AUTOCRACY DUMMY, 214, 230, 233CAPABILITY RATIO, 138, 139, 151, 155,

173, 174CHEMICAL-MINERAL EXPORT, 173,

194CHEMICAL-MINERAL IMPORT, 173,

177, 194, 199CONFLICT, 163, 168, 173, 174CONTIGUITY, 138, 139, 143, 150, 155,

173, 174, 195, 199COUNTRY FIXED EFFECTS, 99, 100,

111, 214, 272, 274, 284, 285, 288DEMH, 136, 138–141, 143–145, 147–150,

152, 154, 155DEML, 136–145, 147–150, 152–157DEMOCRACY, 36, 40, 41, 47, 71, 72,

74–77, 81–88, 99, 100, 104, 107–110,112–113, 117–119, 137, 173, 252,254–256, 260–265, 272, 273, 283, 287,288

DEMOCRACY DUMMY, 214, 230, 233DEPENDENCE, 139, 151, 155DEVELOPMENT, 37, 48, 52DIFFUSION, 25, 27, 30, 36, 41–43, 48,

53, 56, 58, 60, 92, 99, 100, 112, 119,137, 139, 150, 154

DISTANCE, 173, 195, 199ECONOMIC DEVELOPMENT, 52, 53,

56, 60ECONOMIC GROWTH, 52, 53, 99, 100,

119EDUCATION, 56, 60, 99, 100, 118

ENERGY EXPORT, 173, 177, 194, 199ENERGY IMPORT, 173, 177, 194, 199ENVIRONMENT, 214, 228, 272, 274,

282, 283EUROPE, 56, 253EXPORT, 164, 168, 173, 177, 199FDI, 25, 33, 36, 37, 40–47, 50, 51, 53, 56,

58, 62, 64, 67–68, 71–79, 83–87, 249,295, 296, 301–303, 305, 307

GDP INITIATOR, 173, 174, 195GDP TARGET, 173, 174GDPPC, 71, 72, 82, 83, 99, 100, 108, 109,

137, 138, 149, 252, 283, 287, 288GDPPC2, 71, 72, 83, 252GINI, 56GINI coefficient, 46, 60GROWTH, 37, 112, 134, 137–139, 149HOSTILITY, 134, 135IMPORT, 164, 168, 173INCOME INEQUALITY, 64, 71, 295INFLATION, 37, 48, 52, 53, 99, 100, 113,

119, 137, 138, 149, 154INGOs, 29, 30, 56, 60INITIATOR CAPABILITY RATIO, 196,

199INITIATOR DEMOCRACY, 174, 195,

199, 202INITIATOR GDP, 199, 202INSTABILITY, 99, 100, 111, 119INVESTMENT, 99, 100, 110, 111, 118LAGGED ENVIRONMENT, 214, 215,

234, 272LEVEL OF DEMOCRACY, 214, 230, 233LONG RUN CHANGE IN

DEMOCRACY, 117, 118LONG RUN PERCENT CHANGE IN

GDPPC, 118MANUFACTURED EXPORT, 173, 177,

194, 199MANUFACTURED IMPORT, 173, 177,

194, 199MID, 136, 137, 141, 173MIDAB, 155MIDDLE EAST, 56, 253MINOR POWER, 173, 174, 196, 199MISCELLANEOUS CONSUMPTION

EXPORT, 173, 194MISCELLANEOUS CONSUMPTION

IMPORT, 173, 177, 194, 199MISCELLANEOUS EXPORT, 173OCEANIA, 253

348 Subject Index

model variables (cont.)PAST INEQUALITY, 71, 72, 77PERIPHERY, 56, 61POLITY, 109POPULATION, 100, 252POPULATION DENSITY, 214, 232, 236,

238, 252, 272, 273, 283, 284, 287, 288POPULATION GROWTH, 99PORTFOLIO, 10, 25, 36, 37, 40–43,

46–48, 50, 53, 56, 58, 71, 72, 75, 77, 78,81, 83, 87, 293, 301, 305

POWER BALANCE, 174, 196, 199PPP, 112PRIOR DEMOCRACY, 37, 52, 53, 58, 99,

100, 102, 112, 117, 119, 137PRIOR RGDPPC, 99, 100, 104, 118PROXIMITY, 134REAL GDPPC, 214, 231, 233, 236, 272,

273REAL GDPPC SQUARED, 214, 231, 233,

236, 272, 273REGIME, 134, 135REGIME DISSIMILARITY, 174, 195, 199REGIME TYPE, 214, 230RGDPPC, 99, 102, 104, 111, 117, 118SEMIPERIPHERY, 56, 60, 61SOUTH AMERICA, 253STABILITY, 134TARGET DEMOCRACY, 174, 195, 199TARGET GDP, 199TERRESTRIAL ENVIRONMENT, 252THIRD-PARTY MID, 137, 139, 150TRADE, 36, 39–43, 47, 56, 58, 71–72,

74–77, 99–100, 111, 112, 118, 119,137–139, 149, 151, 154, 155, 255, 256,260, 261, 283, 287, 288

TRADE DEPENDENCE, 138TRADE OPENNESS, 214, 232, 233, 236,

252, 254–256, 263, 265, 272, 273WAR, 214, 218, 232, 233, 236, 272, 273,

275, 277, 282, 283, 286–288WAR ABROAD, 273, 276–279, 287WAR AT HOME, 273, 276–279, 287WAR AT HOME OR ABROAD, 276–279YEAR, 37, 42, 52, 53, 99, 100, 112, 119,

137, 138, 150, 154, 272, 274, 284, 285,287, 288, 291

YEAR FIXED EFFECTS, 214, 215YEARLY FIXED EFFECTS, 99, 100, 111

monadic, 3, 13, 126, 128, 129, 132, 133, 135,136, 145, 147, 297

Monte Carlo Simulations, 115Montesquieu, Baron de, 13multicollinearity, 114, 197, 200, 234, 237,

286, 288

neoclassical economic theory, 105neoclassical economics, 10, 63, 111, 118,

159, 296, 306neoclassical trade theory, 242North America, 3, 253, 260, 261NOx, 15, 206, 214–216, 218, 219, 222,

224–225, 227, 229, 232–234, 236–237

occurrence of conflict, 133Oceania, 253, 260, 261OECD. See Organization for Economic

Development and CooperationOLS. See ordinary least squaresordinary least squares, 59, 81, 82, 85, 115,

117, 152, 285Organization for Economic Development

and Cooperation, 38, 51–53, 73, 76, 77,87, 88, 101, 162, 253

panel corrected standard errors, 59–60,87

panel data, 59, 116, 118, 140percent change in DEMOCRACY, 42policy inaction argument, 243politically relevant dyads, 127, 140, 143,

147, 155POLITY, 230, 259POLITY III, 47, 48, 56–59, 61, 81POLITY IV, 109, 195, 214, 283polity score, 214pollution, 1, 15, 206, 209, 211, 212,

214–218, 222–229, 231–237, 246,248–250, 266, 268–270, 293, 297,298

population growth, 244Portugal, 96post conflict effect, 133PPP. See purchasing power paritypreparation for conflict, 132profit, 33, 67, 159, 163, 182, 210purchasing power parity, 48, 72, 82, 100,

102, 109, 112, 231, 259, 283

R2, 83, 197, 198, 234random error, 35, 71, 173, 213, 251,

272

Subject Index 349

random-effects estimator, 59, 85, 114real gross domestic product per capita,

96–100, 102–104, 107–111, 117–118Reducing Environmental Stresses

composite, 222, 238regime dissimilarity, 126, 136, 138, 141, 145,

178, 199regime type, 16, 18, 47, 91, 214, 216, 219,

222, 225, 226, 230, 240, 255, 257, 258,265, 296, 304

regression, 35, 81, 82, 85, 109, 117, 132, 133,180, 182, 206, 234, 284

regulatory effect, 242relative risk, 143, 144, 177, 198responsiveness argument, 244robust standard errors, 49, 59, 82, 85, 116,

145, 153, 197, 234, 260, 285root mean square error, 115rule of law argument, 244Russia, 25, 46, 58, 94, 96, 106, 127, 140, 196,

269, 306USSR, 130, 174

sample mean, 42, 102, 177, 198, 199, 218,235, 238, 255, 276, 287

scale effect, 231, 245, 246SEM. See simultaneous equations modelsensitivity analysis, 43, 62serial correlation, 49, 60, 82, 116, 153, 197,

234, 285simultaneity, 50, 91, 97–98, 101, 107, 113,

114, 117, 126–128, 132–135, 140,143–146, 155, 197, 233, 286

simultaneous equation, 50, 127, 140,141

simultaneous equations model, 11, 13, 91,97–99, 101, 102, 107, 113, 116, 117,126, 128, 134, 135, 140, 144–147, 149,150, 152, 299

SITC. See Standard International TradeClassification

Smith, Adam, 13, 245South America, 48, 251, 253, 260–262South Korea, 96Spain, 77, 96standard deviation, 42, 43, 75, 83, 143, 144,

177, 178, 198, 218, 235, 238standard errors, 49, 59, 82, 83, 115, 116,

145, 152, 153, 157, 234, 286Standard International Trade Classification,

194, 195

structural effect, 241supply, 14, 163–170, 177, 184, 187–193, 195,

303sustainable development, 15, 249

Taiwan, 96technology effect, 242Texas, 23theory, 96, 107, 110, 111, 149, 159, 173, 178,

180–184, 194, 196, 199, 245, 257, 273,288

three-stage least squares, 101, 104, 113, 114,116, 119

trade, 1, 5, 10, 13–16, 18, 19, 24, 25, 28, 32,33, 36, 38, 39, 44–47, 50, 51, 53, 56, 62,64, 66, 70, 73, 76–79, 81, 83, 87, 95,100, 111, 112, 118, 119, 130, 138, 139,147, 149, 151, 154, 158–195, 197–199,205, 212, 214, 228, 232, 236, 239–243,245–259, 263, 270, 273, 283, 293–299,301–306

trade barriers, 258trade flows, 10, 14, 147, 169, 178, 179, 181,

182, 185, 193, 197, 198, 228, 246, 247,256, 304

trade liberalization, 25, 44, 248, 250, 303,304

trade openness, 10, 16, 36, 46, 47, 53, 77, 78,83, 138, 149, 213, 232, 240, 248, 251,252, 254–258, 295–296, 298, 301, 303,305

TSCS. See dataTurkey, 96turning point, 236, 237, 287, 298, 301, 302,

305two-stage least squares, 101, 114, 116,

118two-stage least squares Kiviet estimator,

101, 104, 115, 119

UK, 127, 134, 140, 155, 196Britain, 25, 269

United Nations, 128United Nations Conference on Trade and

Development, 51United Nations Environment

Programme, 229, 243United Nations General Assembly, 139,

151use of force, 134, 138, 162, 173, 179, 180,

185, 202

350 Subject Index

variance, 49, 58, 59, 61, 85, 113, 197, 237,260, 285, 286, 291

Variance Inflation Factor, 197, 200, 234,237, 286, 288

VIF. See Variance Inflation Factor

war, 12, 16, 61, 62, 92, 128–135, 138, 173,179, 202, 214, 232, 235, 236, 279–283,286–291, 298, 305

war channel, 244waste, 209, 222, 229, 231, 246, 250, 261, 268,

269

World Bank, 47, 62, 64, 78, 195World Development Indicators, 47, 81,

112, 113, 228, 229, 232, 259, 282,284

World Resources Institute, 229,282

World Trade Flows, 194World Trade Organization, 78, 228,

243WTO. See World Trade Organization

year fixed effects estimator, 59

Democracy and Economic Openness in an Interconnected System: Complex transformations

Documents

Transcript of Democracy and Economic Openness in an Interconnected System: Complex transformations