Counting on the Environment - World Bankdocuments.worldbank.org/curated/en/... · Counting on the...

E N V I R O N M E N T A L E C O N O M I C S S E R I E S

Counting onthe EnvironmentForest Incomesand the Rural Poor

Paul VedeldArild AngelsenEspen SjaastadGertrude Kobugabe Berg

PAPER NO. 98

June 2004

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Administrator

30026

Papers in this series are not formal publications of the World Bank. They are circulated to encourage thought and discussion. The useand citation of this paper should take this into account. The views expressed are those of the authors and should not be attributed tothe World Bank. Copies are available from the Environment Department of the World Bank by calling 202-473-3641.

Counting onthe EnvironmentForest Incomesand the Rural Poor

Paul VedeldArild AngelsenEspen SjaastadGertrude Kobugabe Berg

THE WORLD BANK ENVIRONMENT DEPARTMENT

June 2004

The International Bank for Reconstructionand Development/THE WORLD BANK1818 H Street, N.W.Washington, D.C. 20433, U.S.A.

Manufactured in the United States of AmericaFirst printing June 2004

Photographs by Paul Vedeld, of Goka Villagers in the Usambara Mountains, Tanzania, 2002.

iiiEnvironmental Economics Series

Contents

FOREWORD viiABSTRACT ixACKNOWLEDGMENTS xiEXECUTIVE SUMMARY xiii

Forest Environmental Income, Income Dependence, Diversification, and Distribution xiiiMethodological Problems xivResearch Recommendations xvPolicy Recommendations xv

Chapter 1Setting the Scene 1

Objectives of This Study 1Problem Formulations 2

Dependence on Income from Forest-Related Environmental Resources 2Best Practices for Empirical Studies of Forest-related Environmental Income 3Outline of Report 4

Chapter 2Forest-Related Environmental Incomes and Poverty 5

Toward a Definition of Environmental Income 5Sources of Environmental Income 5Income Measures 6Location, Production Processes, and Market Chains 7Costs 9The Definitions 10

The Forest-Related Environmental Income and Poverty Link 10Types of Benefits and Stakeholders 10Roles of Forest Income in Rural Livelihoods 12The Poverty–Forest Dependence Connection: A Two-Edged Sword? 17Other Conditioning Factors of Forest Dependence 18

Chapter 3Methodology of Meta-Study 21

Sources of Information 21Selection of Cases 21Description of Sample 21

iv Environment Department Papers

Counting on the Environment — Forest Incomes and the Rural Poor

Representativity and Sources of Bias 23Methodology 24Models and Statistical Analyses 25

Chapter 4Income from Forest Environmental Resources 27

What is the Level of and Variation in Forest Environmental Income? 27Level of Forest Environmental Income 27Relationship between Total Income and Forest Environmental Income 28Key Sources of Forest Environmental Income 28Household Characteristics and Forest Environmental Income 29Contextual Factors and Forest Environmental Income 30A Closer Look at Forest Environmental Income Dependence 33

What is the Distribution of Forest Environmental Income? 35Forest Environmental Income and Income Distribution Between the Cases 35Distribution of Forest Environmental Income Within the Cases 36

How is Forest Environmental Income Used in Household Diversification Strategies? 38Diversification and Main Economic Activities 38Number of Activities and Income Levels 39Cash Forest Environmental Income Strategies 40

Chapter 5Best Practices in Forest Environmental Income Studies 43

Theoretical and Methodological Pluralism 43Differences in the Purpose of the Study 43Types of Studies 44Methodological Approach 44Other Factors 45Best Practices 46

Economic Concepts and Bookkeeping 46Price and Cost Estimates 46External Factors 47Processed Goods 47Environmental Services 48Double Counting 49Stocks, Flows, and Sustainability 50Cash Versus Subsistence Income 51Omitted Variables 52

Representativeness and Data Collection 54Representativeness 54Data Collection 54

Summary on Best Practices 55

Chapter 6Conclusions and Recommendations 59

Key Findings on Forest Environmental Income and Poverty 59Methodological Experiences and Common Pitfalls 62Recommendations on Research and Best Practices 64

vEnvironmental Economics Series

Contents

“Getting the Basics Right” in Individual Case Studies 64Ideas for Future Research on Forest Environmental Income 65

Policy Implications 65Policies to Enhance Forest Environmental Values and Incomes 65Policies That Address Dependence, Distribution, and Diversification 66Policies to Secure the Biodiversity Base 67

Concluding Remarks 67

Appendix AMeasures and Tests Relevant to Forest Environmental Income 69

Basic Measures of Forest Environmental Income and Dependence 69Simple Tests on Role of Forest Environmental Income in Different Households 70A “Forest Environmental Income Kuznets Curve”? 71Income Inequality and Distribution 72

Gini Coefficients 72Kuznets Ratios 73

Diversification 74Forest Environmental Income Fluctuations 74

Appendix BCase Studies 77

Appendix CVariables in the Meta-Analysis 85

Background to the Case Study 85Geographic Characteristics 85Characteristics of Environmental Products 85Characteristics of the Environmental Area 86Socioeconomic Characteristics 86Characteristics of the Market 89Institutional Characteristics 89

NOTES 91

REFERENCES 93

vi Environment Department Papers


BOXES

4.1 Establishment of Mt. Elgon National Park and the Effect on the Access of LocalPeople to Resources 32

5.1 Waiting for Godoy 56

FIGURES

2.1 The relationship between different concepts of environmental income 64.1 The relationship between total and forest environmental income 284.2 Total income and diversification 404.3 Relative forest environmental income and diversification 404.4 Relationship between cash share and total forest environmental income 415.1 Types of forest environmental services included in the studies 48

TABLES

2.1 Importance of different forest benefits to various groups of beneficiaries 112.2 Direct roles of forests in household livelihood strategies 133.1 Distribution of cases by region and forest type 223.2 Household characteristics and external factors in sample 234.1 Average contribution to total and cash income, by sector 274.2 Distribution of forest environmental income, by source 294.3 Forest environmental income and household external factors 304.4 Incomes by Region (number of cases in parentheses) 324.5 Distribution of relative forest environmental income, by case characteristics 344.6 Relative forest environmental income, by Region and forest type (number of cases

in parentheses) 354.7 Forest environmental income for quintiles of cases (standard deviation in parentheses) 364.8 Gini coefficients with and without forest environmental income 374.9 Income distribution within cases and forest environmental income 37

4.10 Importance of various income sources 394.11 Cash forest environmental income share, by household income groups 415.1 Purpose of study, by total and relative forest environmental income 445.2 Type of study by total and relative forest environmental income 455.3 Inclusion of processed goods in valuations of forest environmental income 485.4 Inclusion of sustainability issues, by total forest environmental income 515.5 Perceptions of change in resource supply, by total forest environmental income 515.6 Inclusion of cash and/or subsistence and forest environmental incomes 525.7 Non-inclusion of cash incomes by total forest environmental income 525.8 Labor cost inclusion and total forest income in the sample 535.9 Factors affecting total and relative forest environmental income 56

viiEnvironmental Economics Series

Foreword

It is well-known that poor people are oftendirectly dependent on natural resources thatare not cultivated. But how important, actually,is income from non-cultivated resources? Towhat extent can we measure it? What are wemissing in our poverty assessments? Indeed,what is “environmental income”? These weresome of the questions that inspired this study.

Thanks to the financial support of theNorwegian Government through the TrustFund for Environmentally and SociallySustainable Development, and the WorldBank’s Environment Department, it waspossible to launch the analytical work behindthis report.

The authors are all associated with theAgricultural University of Norway, and itsNoragric Department (Vedeld, Sjaastad, andBerg) or its Department of Economics andResource Management (Angelsen). Let meextend my sincere thanks to the authors fortheir diligent and ambitious efforts to compilea broad empirical data base, and transform thisdiverse raw material into an accessible andsuccinct product.

This report takes us through an intriguingdiscussion of the concept of “environmentalincome” and suggests an operationaldefinition. Through a unique meta-study, it

summarizes no less than 54 empirical studies offorest-related environmental income. Many ofthese studies were not previously easilyaccessible to a general audience. The resultsshould not be carelessly extrapolated to largepopulations, but still tell us a story about justhow important forest-related environmentalincome can be for poor people in many parts ofthe world.

The authors also uncover a methodologicalterrain that is difficult to navigate. If ourunderstanding of environmental income is toimprove, our methods for measuring it must besharpened. This report gives concrete advice onhow to do that. For developing countrygovernments, donors—including the WorldBank—academics, NGOs and everyoneengaged in poverty reduction, there is achallenge here to better capture environmentalincome, and ensure that policies andinterventions will support sustainable use ofthese resources.

Jan Bojö

Adviser

Office of the VP for Environmentallyand Socially Sustainable Development

World Bank

ixEnvironmental Economics Series

Abstract

Environmental income is defined as rent (orvalue added) captured through consumption,barter, or sale of natural capital within the firstlink in a market chain, starting from the pointat which the natural capital is extracted orappropriated.

The present study, which focuses on forestenvironmental income, had two mainobjectives. The first was to investigate theextent to which people in rural areas ofdeveloping countries depend on income fromforest environmental resources, and how thisdependence is conditioned by differentpolitical, economic, ecological, and socio-cultural factors. This is accomplished by ameta-analysis of 54 case studies. The secondobjective was to review research methodologyand make recommendations for “bestpractices” in assessment of forestenvironmental income.

Although there are substantial variations inmethodology and quality of case studies,results indicate that forest environmentalincome represents a significant income sourcewith an average contribution to householdincome of some 22 percent in the populations

sampled. The main sources of forestenvironmental incomes are fuelwood, wildfoods, and fodder for animals. Forestenvironmental income has a strong andsignificant equalizing effect on local incomedistribution. Cash income constitutes abouthalf of total forest environmental income.

The report recommends the development ofresearch protocols, field methods, and simpleanalytical models to analyze the role ofenvironmental income in rural livelihoods.More in-depth studies are needed to unravelthe roles of local heterogeneity and socialdifferentiation. Extended studies that generatetime series data would assist in understandingthe role of environmental income in bothindividual household strategies and in broaderdevelopment strategies.

The omission of forest environmental incomein national statistics and in poverty assessmentsleads to an underestimation of rural incomes,and a lack of appreciation of the value ofenvironment. In areas where environmentalincome is important, this omission may alsolead to flawed policies and interventions.

xiEnvironmental Economics Series

Acknowledgments

This study and publication were financiallysupported by the World Bank and theNorwegian Government through the TrustFund for Environmentally and SociallySustainable Development. We would like tothank Jan Bojö, Adviser, Office of the VP,Environmentally and Socially SustainableDevelopment, World Bank, who has been theTask Manager of this study. He has given ussubstantial scientific and moral support. Inaddition we would like to thank variousreviewers: Sven Wunder and WilliamSunderlin from CIFOR, William Sutton and

Carter Brandon from the World Bank, andparticipants at a World Bank workshop onForest Environmental Income held inSeptember 2003, for very thorough and usefulcomments. Their comments notwithstanding,the authors remain responsible for views andopinions, as well as errors, in this report. Wewould also like to thank consultant LindaStarke for her patient style and languageediting of this report. James Cantrell of theWorld Bank is thanked for desk-topping thefinal manuscript.

xiiiEnvironmental Economics Series

Executive Summary

Understanding the role that environmentalincome plays in poor people’s livelihoods isimportant for at least two reasons. First, it helpspolicymakers design and implement effectivepoverty reduction strategies. Second, the sizeand nature of environmental income hasimplications for issues of conservation andsustainable resource use.

There is no generally agreed definition of“environmental income,” and this paperdiscusses different uses of the term. We settledon the following definition: Environmental incomeis rent (or value added) captured throughconsumption, barter, or sale of natural capital withinthe first link in a market chain, starting from thepoint at which the natural capital is extracted orappropriated. Obviously, the level of income willbe influenced by the degree of processing thatmight take place before any transaction. For theeconomically most important categories of forestenvironmental income (fuelwood, wild food,fodder), however, the modifications of naturalcapital will be negligible.

The present study, which focuses on forestenvironmental income, had two main objectives.The first was to investigate the extent to whichpeople in rural areas of developing countriesdepend on income from forest environmentalresources, and how this dependence isconditioned by different political, economic,ecological, and sociocultural factors. This isaccomplished by a meta-analysis of 54 case

studies. The second objective was to reviewresearch methodology and makerecommendations for “best practices” inassessment of forest environmental income.

Note that the many methodological problemsand weaknesses identified in the case studies inconnection with the second objectivesignificantly reduced the accuracy androbustness of the results arrived at in connectionwith the first objective. Figures andrelationships based on the meta-study of 54cases must therefore be interpreted and usedwith caution.

Forest Environmental Income, IncomeDependence, Diversification, andDistribution

• The mean annual forest environmentalincome was US$678 (adjusted forpurchasing power parity) per household inour sample, which was equivalent to 22percent of total household income.

• The median forest environmental incomewas US$346, equivalent to 19 percent ofmedian total income.

• Although agriculture and off-farm incomehad higher income shares, forest incomerepresented a significant source of income,which is of particular importance tohouseholds living close to the survival line.

Environment Department Papersxiv


• Forest environmental income was ofparticular significance with respect to “gap-filling” and “safety nets”—that is, as anadditional income source in periods of bothpredictable and unpredictable shortfalls inother livelihood sources.

• Wild food and fuelwood were by far thetwo most important forest products. Fodderwas systematically under-reported, but itwas important where reported.

• Cash income constituted about half of totalforest environmental income.

• The share earned in cash declined withhigher forest environmental income. Wealso found a negative relationship betweenthe cash share of forest income and totalincome. Thus, cash forest products were atleast as important for communities with lowforest environmental and low totalhousehold income as for communities thatwere better-off.

• Forest environmental income increased withtotal income, both between the cases(communities) and between householdswithin the communities studied. Forestenvironmental income was thus importantnot only for poor communities. Theelasticity of forest environmental incomewith respect to total income was close tounity.

• Dependence on forest environmentalincome, measured as its share of totalincome, declined with increasing totalincome when analyzed across households.There was no significant relationship acrosscommunities.

• This tendency was even stronger whenlooking at the income distribution between

households: the poor were more dependenton forest environmental income, which hada strong and significant equalizing effect onlocal income distribution.

• Forest income can be seen as part of ruralhouseholds’ diversification strategies. Wefound that high total income was associatedwith less income diversification, indicatingthat higher income was achieved though aprocess of specializing in one or a few high-return activities.

Methodological Problems

The studies reviewed displayed a high degree oftheoretical and methodological pluralism, andthe substantial variability in reporting ofspecific variables and results is partly explainedthrough such pluralism. This variability must,however, also be attributed to methodologicalpitfalls and weaknesses observed in manystudies.

Few studies analyzed the role of or evendistinguished between different types of forestenvironmental income, although this usuallywill be necessary in order to devise appropriatepolicies aimed specifically at the poor.

• Valuation studies reported higher forestenvironmental incomes than studiesfocusing on forest environmental incomedependence and distribution. This may bedue to the inclusion of a wider array ofgoods in the former type of study, selectionof sites with high forest values, or deliberateattempts to inflate the value ofconservation. Our data, however, did notpermit rigorous investigation of theseexplanations.

• Double counting, even triple counting, wascommon.

xvEnvironmental Economics Series

Executive Summary

• Few efforts to define “environmentalincome” were found, and there was littleconsistency in measures used across cases.

• The use of flawed price and cost estimates,such as subsistence good values based onworld markets, was prevalent.

• Many studies omitted important forestenvironmental income sources.Unfortunately, some 72 percent of thestudies ignored fodder, causingunderestimation of income levels.

• Fewer than half the studies consideredwhether current resource use is sustainable.

Research Recommendations

• First, there is a need to “get the basicsright”—to avoid the problems justdescribed—in the individual case studies ofenvironmental income and poverty. Simpleresearch protocols, field methods, andanalytical models are needed to understandthe role of environmental income in rurallivelihoods.

• Second, there is a need for more in-depthstudies to understand the role ofenvironmental income, both in individualhousehold strategies and in broaderdevelopment strategies. Is forestenvironmental income primarily for gap-filling and safety nets, an employment oflast resort, or a pathway out of poverty?Long-term studies with time series datawould help answer these questions. Further,very few of the environmental incomestudies dealt with issues of localheterogeneity and social differentiation.

• Third, there is a need to increase the scopeof such studies through larger research

projects to get comparable data fromdifferent socioeconomic and ecologicalsettings. Our meta-analysis revealed someof the problems of synthesizing case studieswith widely differing methodologies andobjectives. There is a need to generate alarger sample of case studies that havecollected a minimum set of comparabledata. This would permit a much higherlevel of generalization. Moreover, analyzingthe impact of meso- and macro-level factorscan only be done through a meta-analysis.Achieving this requires a concerted effort bylarge players in the research field, such asthe World Bank or the Center forInternational Forestry Research.

Policy Recommendations

A main policy message to governments, donors,and international agencies is that leaving forestenvironmental income out of national statisticsand poverty assessments will lead tounderestimation of rural incomes. This willdepress average and median incomes and alsocause overestimation of the numbers of ruralpoor if such estimates are made from householdincome levels.

In areas where environmental income isimportant, these problems may be serious andcause substantial rural-urban, inter-sectoral,inter-regional, and international distortions,with flawed conclusions and policies as a result.Furthermore, environmental policy in particularmust be based on understanding how theenvironment contributes to the livelihoods ofthe rural poor under changing circumstances.

We also recommend that governments, donors,and international agencies:

• Agree on an operational definition ofenvironmental income that makes income

Environment Department Papersxvi


comparable across cases and acrosslivelihood sources

• Contribute to a worldwide systematiccollection of base-line data in the field

• Include important forest environmentalincome sources in poverty assessments andin Poverty Reduction Strategy Papers,

concentrating on the key resources—food,fuel, and fodder

• Develop sets of policies that secure theresource base, increase the possibilities toenhance the environmental values of forests,and address issues of dependence,distribution, and diversification among, inparticular, poor households and communities.

1Environmental Economics Series

Setting the Scene

More than 1 billion people depend to varying de-

grees on forests for their livelihoods. About 60 mil-

lion indigenous people are almost wholly dependent

on forests. Some 350 million people who live within

or adjacent to dense forests depend on them to a high

degree for subsistence and income. In developing

countries about 1.2 billion people rely on agroforest-

ry farming systems that help to sustain agricultural

productivity and generate income. Worldwide, for-

est industries provide employment for 60 million

people. Some 1 billion people depend on drugs de-

rived from forest plants for their medical needs.

(World Bank, 2001: 15)

Although considerable uncertainty surroundsthe figures often cited regarding the number ofpeople dependent on forests, there is no doubtthat this natural resource plays an importantrole in the livelihoods of vast numbers ofindividuals. Over the last decade or two,numerous studies have documented people’suse and dependence on forest resources. Thesestudies share broad objectives but varysubstantially in focus and approach.

As developing countries elaborate PovertyReduction Strategies based on a diagnosis of thenature and extent of poverty, government,donors, and others must start from a correctassessment of the actual level of poverty. If suchassessments do not capture important sources ofpoor people’s income, such as forests, the basesfor poverty reduction strategies will be flawed.

This study aims to improve knowledge aboutthe level of dependence on forest resourcesamong different groups, as reflected in previousstudies. In addition, the report examines theexisting literature to identify commonmethodological pitfalls and shortcomings, and itrecommends expedient approaches to futurestudies of forest resources, poverty, andlivelihoods.

Objectives of This Study

To consider the accuracy of estimates ofdependence on forests, we looked at existingdata from a selection of studies of forest-relatedenvironmental income for poor people indeveloping countries. This should contribute tothe policy debate on the role of forestenvironmental income as a safety net, a gap-filler, and a contribution to long-term povertyreduction. In analyzing how findings werearrived at in other reports, we also hoped toenhance the manner in which data on forestenvironmental income are gathered and can beincorporated into poverty assessments.

This study thus had two separate objectives:

1. To investigate the extent to which people indeveloping countries depend on incomefrom forest environmental resources andhow this dependence is conditioned bydifferent sociocultural, ecological, economic,and political factors

1

Environment Department Papers2


2. To review and make recommendations on“best practices” for empirical studies ofdependence on forests for environmentalincome, based on both a conceptualdiscussion and a review of the array ofmethodological approaches used in studies.

We chose to focus on environmental incomefrom forests for a number of reasons. From apractical view, the time available for the study(three person-months) forced us to focus on onesubset of environmental income. Further, thereappeared to be more studies on income fromforests than from other sources, and forest-related income forms an important part of ruralincome in many poor regions. This does notmean that environmental income from othersources is less important in individual cases.

This report generally analyzes forest-relatedenvironmental incomes, which we have termedforest environmental incomes throughout thestudy. In Chapter 2, where we explore moregeneral and overall perspectives on costs andbenefits from the environment, we use the moreoverall concept of environmental income.

Problem Formulations

Dependence on Income from Forest-RelatedEnvironmental Resources

In rural household economic researchundertaken 20–30 years ago, a key emphasiswas placed on agricultural production,neglecting what we now know is crucial to suchhouseholds: off-farm incomes through wagework, direct employment such as petty tradeand transport, self-employment, remittances,pensions, and so on. The inclusion of thesesources of income at that time gave a quitedifferent picture of both total and relativeincomes from different sources.

Could the same kind of neglect also be true forforest environmental incomes? What would theinclusion of such incomes in poverty assessmentefforts mean for policy planning andimplementation at large? Forest environmentalincomes are particularly important for poorpeople, presumably because these activities areoften easily accessed, require low levels capitaland purchased inputs, and offer only lowreturns to labor, which is not so attractive forwell-to-do households. These characteristicsthus favor poor households with few otherincome-generating options. The measurement oflevels of dependence for different categories ofhouseholds and under different externalconditions would therefore be important toinvestigate. The share of incomes relative tototal income, how the share enters intosubsistence and cash income strategies, andhow it relates to other income opportunities interms of both scale and vulnerability would beimportant to explore.

People’s diversification strategies for survivaland livelihood involve forest environmentalincome. They have to address current needs,and they also face shocks and other types oftrouble. They furthermore try to developstrategies to accumulate capital and find theirway out of poverty. Forest environmentalincome may constitute part of this process aswell.

Last, it is important to consider if the presentresource use patterns are sustainable. Ideally,people’s use of and dependence on forestenvironmental income should constitute alasting structure, as individuals in principlewould have no objective reasons for depletingthe renewable resource base on which theydepend. This yields a win-win situation in termsof economic growth and the environment. Tothe extent that people are cutting into stocks,


Setting the Scene

depleting resources, or degrading their quality,the long-run effects will threaten bothbiodiversity values and the economic benefitsaccrued from the resource base. An importantquestion is therefore how robust the resource isin the face of different utilization schemes.

These considerations lead to four researchquestions relative to the first objective:

1. How large is the forest environmentalincome?

2. What is the level of dependence on forestenvironmental incomes?

3. What is the distribution of forestenvironmental income between differentgroups of households in the sample?

4. How does the forest environmental incomerelate to strategies for householddiversification?

These issues are analyzed in relation to threedimensions:

1. Forest environmental incomes measured astotal, relative, cash, and subsistence levels

2. The impact of household-level variables onforest environmental incomes: wealth, age,tribe/ethnicity/kinship, educational levels,main occupation, access to land and othernon-forest capitals, and so on

3. The impact of various contextual factors(community or higher-level variables) onforest environmental incomes: averageincome levels, market access andconditions, legal frameworks constrainingaccess and use, administrative andorganizational frameworks, political

frameworks, ecological conditions (such astype of forest or rainfall), and so on.

Best Practices for Empirical Studies ofForest-related Environmental Income

The case studies included in the meta-analysiswere heterogeneous and represented an array ofpractices in terms of the theoretical base andmethodological approach, the scientificbackground of those involved, the purpose ofthe work, the type of analysis, and so on. Theresults of the work were most likely influencedby such factors.

In the analysis of cases and comparison betweenthem, consistency was looked for on a numberof issues, such as income definitions, clear costdefinitions and avoidance of double counting,good differentiation between stocks and flowconcepts of natural resources, and a number ofother possible pitfalls.

Research and development activities have alsoincreasingly become part of the legitimization ofpolitical decisions. To some extent, thedevelopment of epistemic communities isevident, in which politicians, bureaucrats, andresearchers share common scientific andpractical values and norms and depend on eachother in their daily work. It could be that theresults from forest environmental incomestudies are flavored by such issues, althoughthis is difficult to document.

Given all this, we formulated three researchquestions on the second objective on bestpractices:

1. What are the most common overalltheoretical and methodological principlesand approaches used to analyze forestenvironmental incomes?



2. What are common problems in currentpractices analyzing forest environmentalincomes?

3. What are best practices in analyzing forestenvironmental incomes?

Outline of Report

Chapter 2 addresses theoretical and conceptualissues related to environmental incomes and

poverty-environmental income relationships.Chapter 3 briefly outlines the methodologyapplied for the meta-study. Chapter 4 presentsthe results from the meta-study, while Chapter 5deals with common pitfalls found in the studiesand outlines best practices in future work in thefield. Chapter 6 includes some recommenda-tions for future research, a proposal forincluding environmental income in povertyassessments, and some tentative policyimplications.


Forest-Related EnvironmentalIncomes and Poverty

This chapter looks at various theoretical andconceptual issues related to environmentalincomes and at the link between poverty andsuch incomes derived from forest resources.

Toward a Definition of EnvironmentalIncome

How much, and in which ways, doenvironmental incomes contribute to thelivelihoods of the rural poor? To answer thisquestion, a clear operational definition ofenvironmental income must be found. Adefinition is required for internal as well asbroad consistency in use of the term—in otherwords, to ensure that incomes from differentenvironmental goods are measured consistentlywithin a single study and are also comparableacross different studies and locations.

In any definition, three issues are of primeimportance: the possible sources ofenvironmental income, the appropriate incomemeasure, and the level of “remoteness” from thesource—in geographic or economic terms—where environmental income may be captured.

Sources of Environmental Income

With growing concerns about environmentaldegradation, the concept of natural capital hasreceived increasing attention. And this is a goodplace to begin an investigation of theenvironmental income concept in earnest. Could

it be that environmental income is simply theflow derived from the stock of natural capital?

In a very limited sense, the answer is yes.Environmental income can be seen as incomederived from natural resources, sinks, andprocesses created by nature rather than byhumans—from soils, oils, minerals, trees, fish,game, air and water bodies, and climateregulation.

A strict definition of natural capital is “capital inwhose production humans play no part.”Defining environmental income in a similarmanner is, of course, impossible, sincerealization of income requires humaninvolvement. On the one hand, practically allincome involves the consumption or exchangeof goods that ultimately and at least in part arederived from nature. On the other hand, hardlyany goods are consumed or exchanged withoutsome human modification.

There is, however, an important distinctionbetween crops from agricultural fields and logsfrom natural forests. In the latter case it is thenatural capital itself, strictly defined, that atsome point is modified or harvested, but in thecase of crops it is not. And while logs can beseen as “altered trees,” it is much more of astretch to view crops as “altered soil.”

Environmental income can be defined as incomeearned from wild or uncultivated naturalresources. In this case, forest plantations, fish

2



farms, and agricultural fields cannot generateenvironmental income.

A further point of relevance to this studyconcerns forest income. While a naturallyregenerated tree in an agricultural field maygenerate environmental income, it will notgenerate forest income, in contrast to wildgrasses and animals found within a forest. Thiswould be in line with the definition of forestused by the U.N. Food and AgricultureOrganization (FAO), in which, among otherthings, an area is not considered forest if thedominant use is agriculture.1

Income Measures

Three different income measures are implicitlyapplied in the measurement of resource valuesand environmental income: the gross value ofextracted goods, value added, and rent.2 Valueadded equals gross value minus costs of capitalconsumption and intermediate inputs. A furthersubtraction of labor costs and normal profityields the rent. (See Figure 2.1.)

Use of gross value, characterized by the failureto subtract capital consumption andintermediate input costs, cannot be defended.The discussion here will therefore focus onvalue added and rent.

The key to realization of rent from a resource isits appropriation—the act of taking the resourcefrom “the wild” and making it your ownproperty. A good deal of economic analysis(including general equilibrium models) ignoresthis issue, since the starting point is thatproperty rights over all assets are wellestablished.

The existence of rent is often thought of assomething undesired, indicating distortions andinefficiencies. But the presence of rent at theinitial appropriation stage does not implydistortions. In fact, a high resource rent will justbe an indication of a valuable resource.

Appropriation will often occur at the point ofextraction. But an appropriated resource may ofcourse also be held as capital for some timebefore “extraction” converts assets into income.Thus, rents related both to holding and torealizing assets can and often will be capturedthrough the appropriation of natural resources.

An alternative to defining environmentalincome as natural rent is to define it as valueadded, in line with the way “income” isnormally understood in economics and nationalaccounting. Value added can be defined as grossbenefit minus the regeneration costs of factorsexpended during its production. In practical

terms, this means gross benefitnet the costs of capitalconsumption and intermediateinputs. In contrast to rent, whatwould not be subtracted areopportunity costs related tolabor (family or wage labor)and normal profits (return tocapital).

Which measure is preferable?There is, in our view, no

Figure 2.1 The relationship between different concepts ofenvironmental income

Profit

Gross income

Capital costs

(depreciation)

Costs of

intermediate

inputs

Value added

Labor costs

(wages)

Profit

Normal profit

(return to capital)

Rent (super

normal profit)


Forest-Related Environmental Incomes and Poverty

obvious reason to prefer either rent or valueadded. When alternative investment andemployment opportunities do not exist, thesefigures will be identical. In some scenarios, thecosts of capital consumption and intermediateinputs will be trivial, so that even gross valuewill approximate rent. In reality, though,opportunity costs of use of labor and capitalwill almost always be above zero, although theymay also be very small.

The advantage of using rent is partlyconceptual: it measures only income that can beattributed to nature, omitting income from laborthat could earn as much in alternativeemployment. The distinction between marginaland total considerations must, however, beemphasized here. While alternativeemployment opportunities may exist for anygiven individual, local economies might fail toabsorb the labor made available if the entirenatural resource (and thus environmentalincome opportunities) were to disappear.Estimating the “true” opportunity cost of laborunder such counterfactual circumstances isdifficult, to say the least, as demonstrated bynumerous valuation exercises.

Income earned from processing rather thanappropriation and extraction poses particularproblems for value added, since a substantialcomponent of “environmental income” willconsist of returns to skilled labor. In contrast,natural rent will remain identical. We discussthis further in the next section.

Finally, identification of rent will always permita simple calculation of value added—forexample, for purposes of comparison with otherincome sources—while the reverse does notapply.

The advantage of using value added is partlypractical: when the opportunity cost of labor

and capital is difficult to ascertain, as may bethe case in many rural areas of developingcountries, the use of value added will simplifyresearch. In many cases, revealing the actuallabor inputs would also be time-consuming,requiring recording over longer periods of time.

An additional argument for using value addedis the fact that it is in line with normal conceptsof income used elsewhere and thus provides abasis for direct comparisons.

Location, Production Processes, and MarketChains

While it seems intuitively right to regard at leastpart of the value of a fish to the fisher asenvironmental income, it seems equally rightnot to regard the cash generated in a fancyrestaurant in a distant city as alsoenvironmental income, even though the star ofthe meal was that same fish. The key questionis: At what point does environmental incometurn into something else?

Aspects of location could potentially be used todefine the limits of environmental income; suchincome, for instance, may only be earned bylocal people living within a certain distance ofthe resource in question. Alternatively, aspectsof the production process could be incorporatedinto the definition; for example, onlyconsumption or sale of raw materials, notprocessed goods, could count.

Both of these dimensions—location andproduction processes—suffer from problems ofambiguity and imprecision, however. How closedo you have to live to a resource to qualify as “alocal”? Is charcoal a processed good andtherefore excluded from the range of goodscapable of providing environmental income?



A simple case can be made for limiting thedefinition of environmental income to thatwhich is earned only through the first link in apotentially extended market chain. That casebegins with a consideration of natural rent andcompetitive equilibria.

After initial appropriation or extraction, andprovided markets are competitive, no increasesin the value of a good through processing can beattributed to nature; they will appear instead asreturns to invested labor and human-madecapital. Thus in an economy characterized bycompetitive markets, rent can be earned onlythrough the initial appropriation of a resource.If costs are properly deducted and there arecompetitive markets for any already extractedor processed good, environmental income—defined as natural rent—will not rise beyondwhat is earned through initial appropriation orextraction. This is true regardless of the numberof market links and production stages that mustbe negotiated before an end product emerges.

Then the relevant question is: Under whatcircumstances is this not true? Upon itsextraction, a natural resource may be consumedimmediately, sold immediately, or processed ormodified by the extractor. In the first case,consumption, there is no problem. But if it issold immediately, the problem is that rent maybe appropriated at later points in the marketchain, particularly under imperfect marketconditions due to asymmetric information,monopsony (having only one buyer), or pricecontrols. Appropriators may sell extractedresources at a lower price than could have beenobtained with perfect information andcompetitive markets. Thus market distortions insubsequent links in the market chain represent anecessary but not sufficient condition for“natural rent” to be captured by others thaninitial appropriators.

Note, however, that if a definition ofenvironmental income includes “rentattributable to nature” realized through laterlinks in the market chain, there would be nogood reason not to include incomes fromagricultural processing, fancy restaurants, orany productive activity that for some reasonand at some point involved the acquisition ofinputs at prices below the competitive marketequivalent. Market distortions may shift thecapture of natural rent away from extractors,but this rent is then realized precisely because ofthese market distortions, not because of theresource itself.

The problem in the third case, when theresource is processed or modified by theextractor, is that part of the rent captured byappropriators and extractors may beattributable to factors other than “nature.” Intwo otherwise identical settings, anentrepreneur who lives in an area with labormarket failures may capture more rent than onewho does not. And while the monopsony rentearned by carpenters who extract their own rawmaterials would count as environmentalincome, that earned by carpenters whopurchase these materials would not.

How common are such conditions? There is noproblem when markets for processed goods andlabor are competitive, but market failures areprevalent in developing countries, particularlyin rural areas. The problems only appear whenincomes are compared across rather than withincases and study sites. Also, labor wages amongpoor people in comparable rural areas are smallin any case and unlikely to differ much inabsolute terms. What of monopsony rent forprocessors? In most cases, the measurement ofenvironmental income is much easier when theentire first link is considered; that is, it is easier



to measure directly the income accruing tocarpenters or charcoal burners than the incomeattributable only to the natural inputs they use.Only when processing by extractors is common,and monopsony rent is prevalent in one localitybut not another, can a case be made for limitingenvironmental income only to rent earned fromnon-processed goods.

The problem is more serious when value addedis used to measure environmental income. Ifprocessing is common, simple differences in thedivision of labor—some processors extract theirown raw materials while others purchasethem—may generate substantial variation inenvironmental income. And this will apply evenwhen markets are competitive.

Income earned through “the first link in apotentially extended market chain” refers toincome earned by the extractor throughconsumption or exchange of the extracted good.But the first link in the market chain in thiscontext refers to the first link involvingconsumption or transformation of the naturalcapital. Thus multiple sales of the same forest orfishing right, or the acquisition of licenses fortimber harvesting and fishing, do not representlinks in this chain. In some countries, gatereceipts from national park visitations aregathered centrally by government before beingpartially redistributed to communities locatedadjacent to the park. This is environmentalincome. But the wages paid to the tour guide orthe waiters at the park lodge are not.

Competitive markets may provide a theoreticaljustification for limiting the concept ofenvironmental income to rent realized throughthe first link of a market chain only. But theadvantages of such a delineation lie in theresulting linking of environmental income toonly those who have direct access to natural

capital and in the ease of measurement that thisentails.

Costs

Natural resources not only provide income, theycan also generate costs. A typical example iswhen wild animals cause damage to crops orlivestock. Although the damage affectsagricultural or pastoral activities, its source isnatural capital and the costs should properly bededucted from environmental income.

More generally, conflicting interests with respectto the use of natural resources may mean thatone person’s revenue is another person’s cost. Ina typical case, tree cutting by local inhabitantsmay significantly reduce the value of a forestenvironment to visiting tourists or of the wateravailable to downstream users. Should such“costs” also be deducted? No, they should not,because they arise through the absence of thenatural capital rather than its presence. Thesame logic applies to rights and their absence.Costs should not be deducted simply because ofan absence of rights to use natural capital—whether such a lack of rights is due to theestablishment of a national park or theimposition of fishing quotas.

On the other hand, the imposition of duties,such as compulsory fencing, will lead torelevant costs. In all cases, the relevant questionis: What are the costs and benefits attributableto the presence of the natural capital comparedwith a situation where it does not exist?

In this regard, a particular problem arises whenthe presence of natural capital andenvironmental goods deprives locals ofalternative livelihoods. An obvious example ofthis is when protected forest areas precludeconversion to agriculture or pasture. Under



many circumstances, forests may in factrepresent a net loss of wealth for localpopulations. The counterfactual here—anabsence of forest and instead an alternative landuse—will most often be impossible to measurewith any accuracy, although numerousvaluations make the attempt. The problem ofnatural capital as an obstacle rather than as asource of benefits must, however, always bekept in mind.

Finally, any attempt at limiting environmentalincome to that which is earned sustainablywould be meaningless. This is not because ofthe measurement problems involved inseparating sustainable and non-sustainablecomponents of natural resource extractions.Rather, as noted above, the relevant comparisonis with a situation where the natural capital isabsent, not a situation where it is used ormanaged sustainably.3

The Definitions

Based on the above deliberations, a definition ofenvironmental income as rent is as follows:

Environmental income is rent captured throughexchange or consumption of natural capitalwithin the first link in a market chain, startingfrom the point at which the natural capital isextracted or appropriated.

The corresponding definition of environmentalincome as value added is very similar:

Environmental income is the capture of valueadded in exchange or consumption of naturalcapital within the first link in a market chain,starting from the point at which the naturalcapital is extracted or appropriated.

Neither these nor any other definitions will be“perfect” or optimal under all circumstances.

Since the use of the value added definition mayentail substantial variations in income estimatessimply through minor and random differencesin production processes and supply chains, as adefault we recommend the use of the rent definition.

Use of value added may, however, be warrantedunder certain circumstances: when a lack oflabor markets renders the measurement of laborcosts impractical or when production processes,supply chains, and the division of labor inprocessing are largely homogenous across thepopulation being studied.

The Forest-Related Environmental Incomeand Poverty Link

It has long been argued that environmentalincome is relatively more important for thepoor, and—as a corollary—that overuse anddegradation will hurt the poor more than othergroups. A second and more controversialcorollary is that the poor might be hurt more bystrict environmental conservation as well.4

In the case of non-timber forest products(NTFPs), Neumann and Hirsch (2000) note the“overwhelming evidence that the poorestsegments of the societies around the world arethe populations principally engaged in NTFPextraction.” The apparently close associationbetween environmental income and povertyneeds to be considered, however, in terms oftypes of environmental income, absolute orrelative dependence, the relation to householdand contextual factors, and more generally therole of environmental income in livelihoodstrategies. For the purposes of this paper, thefocus in this section is on forest income.

Types of Benefits and Stakeholders

Forests provide a wide range of benefits to poorpeople (see Table 2.1) .



Table 2.1 describes four major types of benefits.First, converting forests (temporarily orpermanently) to agricultural land can be one ofthe most valued benefits locally. Shiftingcultivation represents an intermediate casebetween permanent conversion andconventional forest products, where soilrecuperation and fertilization through biomassburning and decay are essential forest services.Forests also can provide other non-land inputsto agriculture such as arboreal fodder andforage, as well as non-consumptive inputs suchas shade, windbreaks, and contour vegetation.

The case studies we reviewed focused, however,on the more direct forest products—the secondand third categories in the table. NTFPs cover awide range of products from fruits and wildmeat for subsistence uses to commercialproducts such as rubber and para nuts. Thetimber category covers everything from small-scale timber (such as poles and buildingmaterial) to industrial timber. The indirectbenefits of the final category—on-site ecologicalservices (excluding payment for off-siteecological services)—can be important, but littleis known about their magnitude.

Table 2.1 Importance of different forest benefits to various groups of beneficiaries

Source: Angelsen and Wunder (2003), with user groups based on the classification of Byron and Arnold (1999) and with the addition of the�consumers of forest products� category.

Types of economic benefits

User groups

Agricultural land and other inputs in

agricultural production

NTFPs

Timber On-site ecological

services

Forest dwellers Hunters and gatherers

Shifting cultivators

Minor benefit

Main benefit

Main benefit

Important

supplement

Supplementary if

transport access

exists

Variable, but

normally

important

Farmers living adjacent to forests Smallholders

Landless

Major “land

reserve”

Not important

Supplementary

Important

supplement

Supplementary if

transport access

exists

Variable

Commercial users Artisans, traders, small

entrepreneurs

Employees in forest

industries

None

None

Important

Supplementary

Important

Main benefit

None

None

Consumers of forest products Urban poor

None

Some ”pro-poor”

products

Variable

None



Forest dwellers depend directly on a wide rangeof forest products for their existence. For thefarmers living adjacent to forests, who are thefocus of most of the case studies, forests providea major agricultural land reserve. They dependon various forest products such as firewood,construction material, medicinal plants, and soon; forests often provide a food reserve to bedrawn on in periods of crisis or during seasonalfood shortages (before the agricultural harvest).And commercial forest products sometimes alsoprovide an important source of cash. We discussthese uses later.

A third user group consists of those whose mainoccupation is based on forest products, eithersmall-scale operators (such as artisans andcarpenters) or those employed in timberextraction for downstream activities. Under thedefinition of environmental income in theprevious section, most of the income of thesegroups should not be classified as forest income.This, combined with the limited focus given tothese individuals in the studies reviewed, meantthat we did not deal much with thesebeneficiaries. The last group, poor urbanconsumers who depend on forest products suchas firewood and charcoal, is also not dealt withmuch here.

Roles of Forest Income in Rural Livelihoods

Income diversification is a distinguishingfeature of rural livelihood strategies in poorcountries (Ellis 2000: 4). Most households thusmanage a broad portfolio of activities andincome sources. Subsistence and cash incomesfrom “non-cultivated forest-related resources”complement other sources, with a continuumrunning from households that depend almostentirely on these incomes to those that basicallydo not depend on them at all.

The reasons for diversification are several, andtheir relative importance is controversial. Thestandard argument of diversification as a risk-reducing strategy is questioned by Dercon(2000). First, income diversification will tend toreduce expected income; second, becausefluctuations in income sources tend to becovariant, diversification is not a very effectiverisk-reduction means. Thus relatively littleincome smoothing is achieved at the price ofreduced mean incomes. Diversification shouldinstead be understood in terms of theconstraints poor households face. No singleactivity is sufficient, so people have to getinvolved in “whatever is there” to make it addup to a reasonable living. Following this line ofargument, income growth due to the emergenceof one or a few high-return activities shouldresult in less diversification.

Diversification should further be understood interm of the seasonality of different activities.Further, the absence of well-functioning marketsalso necessitates “diversification”: the hightransaction costs (purchase and sales pricewedges) mean farmers try to achieve self-sufficiency in food production, and therebymore diversification. On the other hand, theabsence of labor markets will shut down onepossible avenue of income diversification forhouseholds and can make environmentalincome-generating activities become moreinteresting.

One lesson from this is that environmentalincome must be assessed in light of the role itplays in the total portfolio of activities.Following Cavendish (2003), we mightdistinguish between three different functions offorest income in rural livelihoods:

• Safety nets: Forest products are used toovercome unexpected income shortfalls orcash needs.



• Support of current consumption: Forestproducts are important to maintain thecurrent level of consumption and preventthe household from falling into (deeper)poverty. This role would largely correspondwith the term “coping strategy.”

• A pathway out of poverty: Forest productsprovide a way to increase householdincome sustainably (poverty reduction)either through a “stepping out” strategy(accumulation of capital to move into otheractivities) or a “stepping up” strategy(intensification and specialization inexisting activities) (cf. Dorward andAnderson 2002).

These three roles are of course interlinked, andparticular products can serve the three functionssimultaneously. The distinction is, however,useful to clarify the role environmental incomescan and do play in poverty alleviation. The

functions and descriptions of the three roles areprovided in Table 2.2.

SAFETY NET

The safety net function refers to the role forestscan play during periods of hardship. We choosea rather narrow definition of safety nets andrelate it to the more unpredictable and irregularevents that cause a temporary need for extraincome (either shortfall of other incomes orextraordinary cash needs). Examples of suchemergency events include family illness ordeath, natural disasters, economic crisis, andcivil war.

We do not use this term for the normal seasonalgap-filling functions that forests products oftenplay, in particular before the main agriculturalharvest. A third common usage of forests, whichwe also do not classify as safety net in thispaper, is as a more regular income source for the

Table 2.2 Direct roles of forests in household livelihood strategies

Source: Classification based on Arnold (2001), Kaimowitz (2002), Angelsen and Wunder (2003), and Belcher, Ruiz-Perez, and Achdiawan (2003).

Poverty aspects Function Description

Safety net Insurance Food and cash income in periods of unexpected food and income

shortfall

Gap-filling Regular (seasonal, for example) shortfall of food and income

Regular subsistence uses Fuelwood, wild meat, medicinal plants, and so on

Support

current

consumption

Low-return cash activities A wide range of extractive or “soft management” activities,

normally in economies with low market integration

Diversified forest strategies Forest activities that are maintained in economies with high market

integration

Specialized forest strategies Forest activities that form the majority of the cash income in local

economies with high market integration

Poverty

reduction

Payment for environmental

services

Direct transfers to local communities from off-site beneficiaries



poor who otherwise have few alternatives. Itsaves them from being even poorer.

The role of forests as a safety net depends on thehousehold’s vulnerability—that is, both theprobabilities of being exposed to such eventsand the existence of other safety nets. Manyforest-dependent people have little access tocredit and limited options in terms of formalsector employment. In these cases, forestproducts can be an important “naturalinsurance” against, for example, agriculturalshocks (Pattanayak and Sills, 2001).

Measuring the role of forests as a safety netraises several difficult issues. Using averageannual income (absolute or share of totalhousehold income) will not fully capture thisrole of environmental income. Availability andtiming are key for assessing the safety net role.Ultimately, safety nets help people survive.Further, forest income can be used toaccumulate capital and build up other safetynets (as well as income sources), such as buyingcattle or sending a son to the city to work.

SUPPORT OF CURRENT CONSUMPTION

The second function of forest income in rurallivelihoods is its regular uses in support ofcurrent consumption, but with no or limitedscope of lifting people out of poverty. Threedifferent sets of activities can be distinguishedin this case.

Seasonal Gap-FillingForest products are most extensively used toovercome seasonal food shortfalls—that is,before the main harvest. “The importance offorest product income is usually more in theway it fills gap and complements other incomethan in its absolute magnitude or share ofoverall household income,” note Byron andArnold (1999: 792). Also, on the subsistence-use

side, the authors note that “forest productsseldom provide the staple, bulk items thatpeople eat” (1999: 792).

The gap-filling function is distinguished fromthe safety net function by its regularity andhigher predictability. For example, some forestfruits are harvested each year in the monthsbetween staple harvests. They fill a gap in thesense that they provide a periodic andreasonably predictable contribution to foodsecurity, serving as a “seasonal buffer.”

The gap-filling function raises measurementproblems. Ideally, any valuation should takeinto account that during slack seasons theopportunity costs of labor are lower and thevalue of extra income is higher than in the restof the year.

Regular Subsistence UsesMoving a step further, the category of “regularsubsistence uses” includes forest uses that aredone more or less continuously throughout theyear. Most forest products are consumeddirectly. This is not included in official statisticsand has therefore been referred to as the“hidden harvest.” It is particularly importantwith regards to firewood, fodder, wild food(including meat), and medicinal plants. In someareas with good forest access, the contributioncan be as high as 30–40 percent of total income(cash and subsistence).

FAO estimated that in the mid-1990s some2 billion people used fuelwood (firewood andcharcoal) as their main or sole source ofdomestic energy.5 Although not all of this comesfrom forests, fuelwood is probably the forestproduct with the largest number of direct users.This subject received some attention followingalarming reports in the 1970s about a globalfuelwood crisis, but governments, private firms,and donors have largely ignored fuelwood sincethen. Yet it continues to be important to the bulk



of rural households in developing countries,and an increasing number of them facefuelwood shortages.

A recent literature survey by Arnold and others(2003) shows that fuelwood’s commercial rolecan also be significant. In peri-urban areas ofSub-Saharan Africa, tens of thousands of poorfarmers and small traders supplement theirincomes by selling fuelwood. Sometimes thisactivity even becomes their main source of cash.Notably, this includes also the poorest of thepoor; for instance, many rural landless peopleare among those specializing in fuelwoodproduction.

The wild food used regularly includes literallythousands of different species: fruits, roots andtubers, mushrooms, and so on. These provideimportant vitamins and complement regularstaples such as rice or maize. Wildlife alsoprovides a large share of the calories andprotein of rural households, particularly inCentral and West Africa, the Amazon, andSoutheast Asia. In more than 60 countries,wildlife and fish contribute more than 20percent of the animal protein in rural diets.

In the early 1980s (the most recent dataavailable), the World Health Organizationestimated that 80 percent of people indeveloping countries met their primary healthneeds through traditional medicine, mainlywith some 10,000–20,000 different medicinalplants. The role of these plants, a large share ofwhich comes from forests, has been ignored inthe international debate on health and poverty,probably reflecting western “scientific”approaches to health and ignoring indigenousknowledge and use of these resources.

Low-Return Cash ActivitiesAs noted, rural households are normallyinvolved in several activities, and forest

products play an important role in suchdiversified livelihood strategies. Many of theforest activities (collection of honey, meat, nuts,and so on) are, however, low-return activities.Why? There are two sets of answers to thatquestion. The first one deals with why poorpeople get involved in such activities; this isaddressed in the next sub-section.

The second answer deals with why many (if notmost) forest activities yield low returns. Naturalforests are often economically inferiorproduction environments (Angelsen andWunder 2003). Low per-hectare densities ofcommercially valuable species imply thatextraction tends to be spread over large areas,triggering high costs and low net returns. Thereare also characteristics of the products,especially extractive products, that putproducers at a disadvantage. They arecommonly collected in de facto open accesssituations. There are sharp seasonal and otherfluctuations in supply due to the phenology ofplants, migration patterns of animals, orfluctuating climatic conditions. Many haveheterogeneous quality and are highlyperishable.

Further, forest product trade is oftencharacterized by monopsonies and exploitativemarketing chains. The heterogeneity and qualitydifferences of some products, combined with alack of well-defined standards, encourageintransparencies and manipulations. Perhapsmost important, the remoteness of many forestcommunities reduces their level of informationabout changing end-use prices and theirbargaining power in relation to the few traderswho are commercializing low volumes. Finally,limited growth in the demand for many NTFPs,both over time and with rising income levels,restricts the options for expanding producerincomes. If producers increase supply, the



downward pressure on the price might actuallyreduce their combined income.

POVERTY REDUCTION

The third function of forest income in rurallivelihoods is poverty reduction. Again, threedifferent sets of activities can be distinguished.

Diversified Forest (cash income) StrategiesAlthough most forest activities yield lowreturns, and therefore are more significant forpoverty prevention than poverty reduction,some forest products—given favorableconditions—can provide a means ofsocioeconomic advancement. The Center forInternational Forestry Research (CIFOR) hasdone a worldwide comparison of 61 cases ofcommercial NTFP production and use. Thecases were then broadly classified, based on twodimensions (Belcher, Ruiz-Perez, andAchdiawan, 2003): the importance of theproduct in local livelihoods (the income fromthe product in percent of total household cashand subsistence income) and the degree ofintegration into the cash economy (thepercentage of total household income earned incash).6

Using a cut-off line of 50 percent for bothdimensions, three categories emerged:7 The firstcategory is the subsistence economy, with lowproduct contribution and low integration in thecash economy. (This is not a subsistenceeconomy in the conventional sense, as all caseshave some cash income from the forest productand possibly also other sources.) This would tosome extent overlap with the category we havediscussed already—low-return cash activities.In terms of geographical distribution, thesubsistence strategies are associated with Sub-Saharan Africa. In general, they have lowertrade volumes and less income, but also

growing populations and market demand thatare increasing pressure on the resource.

The second category is the diversified economy,with low product contribution and highintegration. In this case, forest activities aremaintained even in situations with a highdegree of market integration (more than 50percent of the total income is in cash, and forestcash income is one among several cash sources).This suggests that the forest activities are able tocompete with other cash income-generatingactivities. The diversified economies are foundin Latin America.

The third category is the specialized economy, withhigh product contribution and high integration,as discussed in the next section. The specializedstrategies are associated with Asia.

Specialized Forest (cash income) StrategiesIn the specialized economy, households havefocused on one particular forest product. TheCIFOR study shows that the average income forsuch households is significantly higher (34percent) than the average local income (there isno significant difference in the two othercategories). This indicates that forest productspecialization can be a pathway for povertyreduction.

The specialized economy is further associatedwith a number of favorable conditions: a highvalue-weight ratio for the product, stableproduct market and household involvement inNTFP production, less product adulteration,stable populations of target species, and—interestingly—a lower incidence of customaryrules. It is also associated with more intensivemanagement, but this can either be bydomestication (plantation) or from managedforests. Plantation products are found in only



one-third of the specialized cases, but they areassociated with higher prices and productivity.

Payment for Environmental ServicesForests provide important services to both localon-site forest dwellers (clean drinking water, forexample) and off-site beneficiaries, such asregional users (downstream water benefits, forinstance), national users (urban tourists), andglobal stakeholders (through valuing theexistence of endemic biodiversity). Probably thehighest poverty reduction potential is throughpayments for the off-site benefits enjoyed at theregional, national, or global levels. Currently,most off-site beneficiaries are “free-riders”: theydon’t pay. Hence a poor forest community thatowns a forest has no incentive to take thisservice provision actively into account in itsland use decisions. Thus when land becomesscarce, the service may eventually be lost.

Compensation mechanisms are relevant in atleast four areas: carbon storage andsequestration, biodiversity conservation,hydrological services, and forest-based tourism.We will not go into details for these, but refer tooverviews such as that by Landell-Mills andPorras (2002).

Generally, there is ground for some optimismthat forest-service payments can help reducerural poverty. Not only do they offer anadditional flow of income, the flow will alsotend to be more stable over time than the ones itis designed to replace, such as the fluctuatingprices of timber and cash crops. In addition, theintroduction of payments can also help induceindirect benefits, follow-up investments, andexternal assistance, such as training, improvedcommunity organization, better knowledgeabout forest management, improvedenvironmental quality, better understandingabout urban markets for other products, and so

on. Finally, international markets for ecologicalservices provide a mechanism for long-terminvestment flows from the North to the South.Even when the direct payment recipients arelarger landholders, the transfer can eventuallyalso provide indirect benefits that trickle downto the poor (Angelsen and Wunder 2003).

A key question is to what extent poor peoplewill be able to compete in emerging servicemarkets. Two comparative disadvantages of thepoor are their insecure land tenure and the hightransaction costs generated by smallcommunities, poor infrastructure, and imperfectmarkets. Additional constraints for poor peoplecan arise if there are entry costs into theschemes or if agreements imply long-term,inflexible land use commitments (Pagiola,Bishop, and Landell-Mills 2002: 282–84).

The Poverty–Forest Dependence Connection:A Two-Edged Sword?

Poor households often display “forceddiversification,” with multiple occupations.Activities typically include low-income forest-based actions such as wildlife hunting, charcoalproduction, fish smoking, tobacco curing, andthe gathering of fuelwood, poles, fruits,vegetables, and so on. These prove to be “lowskill and open access activities” with lowreturns. That poor people take them on mayrelate to asset stakes, barriers to entry in certainfactor markets, lack of access to resources, andmore general rights issues. Forest-basedactivities thus attract poor households. “Ease ofaccess to the resource and low skill and capitalthresholds to commercial forest productactivities means that these can be veryimportant in coping strategies of the very poor,”notes Arnold (2001: 3). There is overwhelmingevidence that the poorest households depend



relatively more on these low-return forestactivities, as will be shown in Chapter 5.8

What is the causal link? Are they forest-dependent because they are poor, or poorbecause they are forest-dependent? We wouldargue that the causality runs mainly in the firstdirection: low-return activities—in forestry orother sectors—become an employment of lastresort for poor people.9 Forest dependence is afunction of their poverty: the poor do not haveaccess to more lucrative income opportunities,and low-return activities become the bestpossible use of their labor. For the better-off, onthe other hand, low-return activities are notattractive.10

Based on this, it can be argued that the primeimportance of forest products is with respect toproviding a safety net and a means to maintaincurrent levels of income, not a pathway out ofpoverty. There might, in fact, be a tradeoffbetween these roles. According to Angelsen andWunder: “A key insight for the forestry-povertydiscussion from the literature is the ambiguousrole of NTFPs. The very same characteristicsthat make them important and attractive to thepoor in the first place also limit the potential forfurther income increases” (2003: 25).

The authors also point out that efforts toincrease the value of NTFPs that the poordepend on might be counterproductive: morevaluable resources might be captured by therich, scarcity can cause internal differentiationand development of individualized rights thatexclude the poorest of the poor, higher valuemight lead to overexploitation, anddomestication and synthetic products might bestimulated and replace collection from the wild.

Even if NTFPs do not provide a major way outof poverty, they should not be understood as the

activities that generate poverty. People seekthese activities because they do not have manyother options; as such, the activities areimportant for poor people’s survival andlivelihood. Thus NTFPs should not be seen as apoverty trap, unless “alternative developmentoptions actually exist but . . . policies, donorprojects or other external interventions seek tomaintain people in their low yield forestextraction activities” (Angelsen and Wunder2003: 24).

This does not mean that all forest activities arelow-return ones with poor prospects forsocioeconomic advancement. Some activities aredifferent, and where the right conditions are inplace (resource base, skills, market access, andso on), specialization strategies can makehouseholds better-off (Belcher, Ruiz-Perez, andAchdiawan, 2003). Further, well-to-dohouseholds do diversify by incorporatingenvironmental incomes in their portfolios. Theproducts sought and the level of division oflabor achieved in their pursuit does, however,tend to be much greater. Such environmentalactivities could include semilegal timberextraction and trade, pole production, andintermediary activities for NTFPs.

It may thus be possible to identify povertyprevention strategies as those of less well-to-dohouseholds, where motives related to insurance,gap-filling, and coping strategies prevail. Forthose better-off, the strategies would be seen asvarious types of more and less specialized foreststrategies, enterprise, and industrial activities(Angelsen and Wunder 2003).

Other Conditioning Factors of ForestDependence

A number of factors at household andcommunity level and beyond may have an



impact on the role of environmental incomes forrural households. (Appendix A considers howthese variables, and others, can be applied inmeasuring and explaining variations inenvironmental income levels and dependence.)

HOUSEHOLD-LEVEL FACTORS

Age can affect forest dependence in severalways. Many studies document that youngpeople depend more on forest resources.Collection of valuable forest products can bepart of an accumulation strategy to establish ahousehold. Young households can also beclearing more forest to build up a sufficientamount of cropping or pasture land. Olderhouseholds may have less time and physicalstrength for forest work. At the same time, ourexperience is that some young people considerforest collection old-fashioned. They may alsolack the necessary skills.

Education is expected to be negativelycorrelated with forest dependence, for similarreasons as poverty (income level). Better-educated households have access to a widerrange of income opportunities and wouldtherefore not find it sufficiently rewarding to getinvolved in forest activities. It might also be acultural factor, as with age: Going to the forest isconsidered backward and not for the well-educated.

Ethnicity and place of origin can be importantdeterminants of forest dependence, but not in astraightforward manner. In some contexts,migrants do not have access to forests or theylack the necessary skills, experience, andtradition of forest product collection. In others,they might be restricted from otheropportunities (use of arable land, in particular)and may therefore resort to forest products as“employment of last resort.”

A general finding in rural household surveys isthat female-headed households are poorer thanmale-headed ones. It is worth noting, however,that female heads of household in most culturesare widowed or divorced, or their husbands areworking far away. Thus the adult labor force istypically smaller. In a study from Malawi, forexample, Botha (2003) finds that the averageincome of female-headed households issignificantly lower. In a regression analysis,however, controlling for other householdfactors, the dummy variable for sex ofhousehold head was no longer significant.

COMMUNITY-LEVEL AND CONTEXTUAL FACTORS

Beyond the individual and household level, anumber of factors at the village, district, or evennational level are important determinants offorest dependence. In total, these establish howinterested households are in environmentalincome opportunities and thereby theirdependence on them. Such broader issues mustbe considered for any meaningful descriptionand explanation of adaptations concerningenvironmental income at individual andhousehold level. Unfortunately, many studiesare done in only a few villages and do not havesufficient variation to allow for a useful analysisof these variables.

The availability of forest and environmentalresources will vary substantially, for instance,defining the potential production possibility setfor the household. There is obviously a bigdifference in resource access in montane moistforests in Nepal or Tanzania compared withsuburban areas around Khartoum, Sudan, interms of both types and amounts of products.Furthermore, the ecological variation alsodefines the agroecological conditions that havean impact on income possibilities in agriculture



and therefore the need to resort to forestproducts.

Having good access to markets has ambiguouseffects on forest dependence and is alsocorrelated with other factors such as forestavailability and population density. In general,we would hypothesize that good market accessimplies less forest dependence, as alternativeincome opportunities are better and forestavailability lower. But in some cases theopposite might be true, as exemplified by thespecialization strategy discussed above.Specialization in high-value forest productsrequires good access to markets.

The population-environment nexus and theBoserup versus Malthus debate are importanttopics in the environmental arena. In highpopulation density areas, fewer resources percapita are likely to be available from whichenvironmental incomes can be drawn. Further,high-density areas can be expected to havebetter infrastructure and market access as wellas a greater number of other income-generatingactivities, also pulling in the direction of lowerenvironmental income dependence.

There is an extreme variation in legalframeworks regulating access, withdrawal,maintenance, monitoring, and control rightsover environmental income resources.

Communities close to protected areas are inquite different positions for generatingenvironmental incomes than communities closeto forest plantations or those with access tocommon property resources (CPR) or to openaccess areas. Tenurial rights and degree oflandlessness have an impact on the level ofdependence on forest resources. Because ofpoor people’s particular dependence onenvironmental incomes, the management ofCPRs seems to hold an important role inPoverty Reduction Strategy Papers.11

Countries have different administrative, policy,and legal frameworks that affect access toresources. Concessions rights, the degree oflocal participation, and the amount of landunder different types of protection all haveimplications for levels of and dependence onenvironmental income.

Local social institutions regulate access andpossibilities to generate environmental incomesto some extent. Furthermore, knowledge aboutpossibilities for environmental incomes oftenvaries among local ethnic groups. It is observedthat groups “emanating from the forest” tend tohave a much broader array of products andinterests in the forest than agricultural peopleliving in the lowlands, as can be seen, forexample, in a study from Tanzania (Sjaastadand others 2003).


Methodology of Meta-Study

This chapter looks at various methodologicalchallenges related to the two objectives of thisstudy, including sources of information, theselection of cases, representativity, and the useof models.

Sources of Information

Within the limitations provided by the selectionof cases criteria below, the general approach wasto include studies that could contributequantitatively to our understanding ofdependence on forest-related income indeveloping countries. We selected 28 studiesfrom various international research institutions.Twenty studies were theses written at theAgricultural University of Norway (NLH). Andwe have included three consultancy reports. Wedo not pretend to have made a complete reviewof all quantitative studies done outside NLH.

The studies were found through various Internetsources and databases, including Tropag andRural, ISI Web of Science, CAB Abstracts, Eldis,World Bank, FAO, and the OverseasDevelopment Institute. We have also usedInternet search engines such as Google. And weused the 2,680 Fulltext journals subscribed to byNLH when searching by means of keywordssuch as environmental incomes, forestenvironmental incomes, and dependence onforest environmental resources. Getting access toand overview of potential cases in this way wasthus a first step in the final selection of cases.

Selection of Cases

Generally, we used the following criteria for thecase study selection:

• The case had to include registeredhousehold environmental income andpreferably also incomes from other sources.

• The case had to have registered otheraspects relevant to the households’productive capacity, such as assets, factorsof production, and socioculturalinformation.

• The case had to have some informationabout contextual variables of various types,as described in Chapter 2.

Due to one of the main objectives of this report,we also wanted a mix of different types ofresearch, theses, and consultancy work. Weneeded variations on type of study—economicvaluation studies or different types ofhousehold adaptation studies. We also wanted afocus on studies from Africa, but with somecases from both Latin America and Asia.

These efforts and criteria gave us the sample ofcases included in this study (see Appendix B).

Description of Sample

Our meta-analysis is based on 54 case studiesfrom 17 countries. The cases display great

3



variation in terms of ecological, sociocultural,demographic, political, and economicenvironments. The sample is dominated byAfrican cases: 15 are from East Africa, 18 fromSouthern Africa, 14 from Asia, and 7 from LatinAmerica. This African dominance should bekept in mind when interpreting the results. Thecases are about evenly drawn from wet, semi-wet, and dry forests (see Table 3.1).

Some key household characteristics andcontextual properties of the cases are presentedin Table 3.2. The communities and householdsin our meta-study were generally ruraldwellers, mostly farmers with some access toforests. Most of them were poor in terms ofincome levels; they had low education levelsand hardly any access to capital. Thehouseholds in our cases had on average 5.5years of schooling, a typical feature of ruraldwellers in poor countries. Enrolment rates insecondary and post-secondary education werelow, with 26 percent of households attainingsecondary education and only 4 percentattaining post-secondary education. Loweducation levels typically indicate that thehouseholds are excluded from large segments ofthe labor market.

The average household size for the cases was5.9 people, also a typical level for poor rural

countries. The average annual income for thesehouseholds was US$3,043 (adjusted forpurchasing power parity (PPP)), which gives aper capita income of US$513. This compares tothe average US$675 for the countriesrepresented in the survey (simple average).Together with the fact that most of theenvironmental incomes were not included in theoverall figures, this suggests that communitiesincluded in the samples had average incomeswell below the national averages.

The average land holding was 1.8 hectares.Small plots constrain households in crop andlivestock production and make otheralternatives more attractive, includingexploitation of forest environmental resources.

Livestock numbers were reported to be low,with an average of 4.3 tropical livestock unitsper household but with large variations in thesample. The low number is puzzling, given thatmany rural poor commonly depend heavily onlivestock production. It could indicate thatlivestock represent a significant investmentoutside the reach of many poor households. Ourresults still indicated that livestock contributesignificantly to total household income.12

As a result of having little land and low levelsof capital and education, the sample householdswere more likely to allocate labor to enterprises

Table 3.1 Distribution of cases by region and forest type

Region

Type of forest East Africa South Africa Asia Latin America Total

Wet forests 4 2 7 5 18

Semi wet forests 7 4 3 0 14

Dry forests 4 10 2 0 16

Not specified 0 2 2 2 6

Total 15 18 14 7 54



such as forestry, fisheries, and unskilled wagework. These have small barriers to entry andmay not require much education or high levelsof capital inputs—that is, they have some of the“employment of last resort” characteristicsdiscussed earlier.

The households in our case studies reportedthat they allocate their resources to cropproduction, livestock rearing, forestenvironmental activities, and off-farm work.Off-farm activities included retail trade, wagelabor, brewing, and tailoring. Forestenvironmental activities typically includedanimal hunting; grazing; gold panning;gathering a number of products such asfirewood, thatching grass, wild fruits,vegetables, rattan, wild medicine, insects,timber, and poles; and making baskets, mats,and crafts.

Representativity and Sources of Bias

The meta-analysis should be broadly relevantfor poor rural populations in developingcountries living adjacent to forests who rely on

these for some but not all of their livelihood.Obviously, however, even within this categorythere will still be substantial variations in therelationship to the forest as well as othersocioeconomic variables. Indeed, one purpose ofthe study was to investigate how forest-relatedincome depends on these variables.

For the first objective of the study, the selectionof cases may have an impact on the “levels ofdependence” on forest-related incomes. Theremay be biases relative to continents, countries,or ecological zones. This problem can becontrolled through the use of contextualvariables whenever these are available. Withineach case there may also be substantial biases.

Given the broad sampling of cases, several sortsof bias are possible. Of course, we had nocontrol over the locations that writers chose toconcentrate on, and this may naturally havecaused a geographical bias within a country orregion. There may also have been a tendencyamong researchers to select study areas inwhich NTFPs or other forest goods are ofparticular importance, and this may bias results

Table 3.2 Household characteristics and external factors in sample

Household and external characteristics Total sample Studies included

Age 40.7 years old 14

Size of household 5.9 persons 41

Education level 5.5 years of school 26

Livestock 4.3 tropical livestock units 9

Land size 1.8 hectares 20

Total income US$3,043 48

Cash income US$1,633 18

Gross domestic product of country US$684 51

Elevation 1,219 meters above sea level 30

Precipitation 1,062 millimeters per year 34

Distance to market 19.5 kilometers 13



toward high levels of environmental incomeand dependence. Thus it is not appropriate toclaim that the findings are representative of“rural areas in developing countries.” Inaddition, there were the numerous flawscommitted during field work or analysis withinthe studies themselves, which may bias results;these are discussed in detail in Chapter 5.

The variability is also reflected within the meta-sample in the means arrived at in the separatecases. While the sample averages for annualincome earned from agriculture and off-farmactivities were US$1,499 and US$1,038respectively, several studies reported no suchincome. Although sample average householdsize was 5.9, it ranged from 3.8 to 8.5. Sampleaverage farm size was 1.8 hectares, but thisranged from 0.2 to 3.7 hectares, with a standarddeviation of 0.9.

Given the high variability in sociocultural andsocioeconomic conditions, results from tests thatrely solely on variability across cases rather thanwithin them must be treated with caution.

Another type of bias lies in the selection doneby the case study authors themselves. If thepurpose was to investigate the role of forest-related income, or even to prove thatenvironmental income is important to poorpeople, it is likely to be a selection bias in favorof communities where this is important. Again,the findings are therefore not representative of“rural areas in developing countries.”

For the second objective of this study, theselection of cases could be biased relative to thetotality of studies and the approaches selectedin the “worldwide portfolio of studies.” Thecomposition of studies from different types ofresearch and development environments andthe types of approaches was most likely

somewhat biased in favor of university andresearch work compared with consultancies anddonor-funded assignments. It was hoped thatsome of the typical research biases and faultsare less frequent in the former types of studies,but they are certainly still around and theydeserve attention.

Methodology

The variables used (see Appendix C) wereselected through a perusal of the cases as wellas other meta-studies, in particular the one byCIFOR described in Belcher and Ruiz-Perez(2001). The final set of variables represents, to alarge extent, an amalgamation of those found inthe different cases. As objectives andapproaches vary between studies, the degree towhich each case will supply information on theselected variables differs considerably. In somecases, such as rainfall, it was possible to findmissing information through other means orthrough calculation based on existinginformation. In most cases, however, we wereleft with missing observations for some of thevariables, even for quite crucial information.This was a serious problem in the present study,as it is for most meta-studies of this type.

We also looked at what we have calledcontextual factors. These are beyond theimmediate control of households but mayinfluence levels of total income, environmentalincome, and dependence on the latter.

In order to approach our second objective ofmaking recommendations on “best practices”for empirical studies on this subject, we had toidentify variables that would describe the typeof theoretical and methodological frameworkused in a particular study and that mightinfluence the quality of results and analysis.These issues are discussed in more detail in



Chapter 5, where we present some of the factorswe tentatively expect to be important.

Researchers from different disciplines are doingresearch on the economic importance of naturalresources on people’s livelihoods. We thusfound studies on “dependence onenvironmental income” to be a meeting groundfor a variety of methods, with no consensus onhow to approach estimates and analysis ofenvironmental usage and income. Meta-studiestherefore have to deal with research conductedunder different scientific frameworks. Inaddition to more overarching considerations,there are also a number of methodologicalissues where studies will differ.

Models and Statistical Analyses

Possible measures and procedures for use inanalysis of forest-related income anddependence on such incomes are presented inAppendix A. Most of these, however, are morerelevant to case studies where households arethe unit of analysis, rather than to meta-analyses such as ours, where cases—communities or regions—have been the focus.

A meta-analysis is, at a fundamental level, moreattuned to study differences across communitiesthan differences within them. For an analysis of

the latter to be possible, the cases on which themeta-study is based must provide indicatorsand aggregate variables that permitcomparisons of intra-case variations. Examplesof such indicators—diversification indicatorsand income inequality measures, for instance—are also described in Appendix A.

Some of these variables are of such basicimportance that we would expect them to bepresent in the vast majority of cases.Unfortunately, this is not the case. In addition topointing up fundamental flaws in themethodology applied, however, this may alsocreate problems for meta-analysis of the typeconducted in this study. As the range ofvariables supplied by each case will vary, eachtest will be based on different samples. Thus, forexample, a test of how environmental incomedependence is influenced by overall incomelevels may be based on a different sample thanan analysis of the role of income inequality. Thisis particularly a problem when some casescontain values that represent “outliers” forsome variables. This needs to be borne in mindwhen interpreting the results, and we also keepthis problem in mind in the analysis inChapter 4.

The statistical analyses have been done usingSAS/JPM and Stata.


Income from ForestEnvironmental Resources

This chapter presents the results of the meta-analysis. The first section looks at the magnitudeof forest environmental incomes and thedependence on them, and how these areconditioned by household specific variables(averages for cases) and contextual variables.The second section considers the distribution offorest environmental incomes, both betweencases and between households. The final sectionlooks at the relationship between forestenvironmental income and degree ofdiversification.

What is the Level of and Variation inForest Environmental Income?

This section looks at forest environmentalincome compared with overall income, at thekey sources of the former, and at how income

varies with household characteristics and withfactors beyond the control of the households,such as ecological variations, economic factors,legal status of areas, and different geographicalregions.

Level of Forest Environmental Income

Table 4.1 gives average shares of total incomesderived from different sectors of production.For the cases studied, agriculture (includinglivestock) contributed 37 percent, forestenvironmental activities accounted for 22percent, while 38 percent was derived from off-farm activities, making these the three mainlivelihood strategies.13

The mentioned activities provided householdswith both subsistence and cash income, and the

Table 4.1 Average contribution to total and cash income, by sector

Note: a The figures represent average household income/year for the cases reporting to have the particular income. Figures in parenthesisare percentage contribution to total income for the samples reporting that income. Thus the percentages do not add up to 100. We havenot corrected for the sample size because we do not know if the study excluded income from, for example, off-farm activities because itwas small, zero, or outside the focus of the study.

Total sectoral income Cash sectoral income

Sectors US$-PPP

(%)a Standard deviation

Minimum value

Maximum value N

US$ PPP (%) N

Agriculture (crop and

livestock)

1,499 (37%) 1,792 0 6,455 28 271 (31%) 15

Forest environmental

incomes

678 (22%) 791 1 3,459 51 202 (24%) 17

Off-farm activities 1,038 (38%) 1,424 0 6,129 24 176 (43%) 8

4



contribution to cash income for these activitiesis also shown in Table 4.1. Note that the forestenvironmental income is close to the average ofother rural income sources in terms of itssubsistence-cash ratio (although some care isneeded in comparing these figures—22 versus24 percent—as the sample for the latter is muchsmaller).

The households in our sampled cases didacquire substantial forest environmentalincomes. As seen from Table 4.1, the averagehousehold derived around US$678 per year(PPP adjusted) as forest environmental incomeout of a total average income of US$3,043—around 22 percent. Removing the three caseswith forest environmental income aboveUS$2,000 gives a new mean of US$533, or 19.5percent of total income. And removing all caseswith forest environmental income aboveUS$1,500 gives a mean of US$401, but this wasstill 20 percent of total income.

There were substantial variations in the level offorest environmental income. The maximumlevel was US$3,460 per household and yearwhile the minimum was US$1.30. Thedistribution was quite skewed, with the mean

income (US$678) almost twice that of themedian (US$346), rendering a skewednessmeasure of 1.78. The median share of forestenvironmental income was 18.9 percent, muchcloser to the average, with the cases rangingfrom 0.4 to 60.4 percent.

Relationship between Total Income and ForestEnvironmental Income

Forest environmental income is closely andpositively related to total income. We found thatthe forest environmental income (AFI) increasedwith increasing total income (AI). A simpleregression analysis (using a log-log model,which gave the best fit) produced the followingresults (t-values in parentheses):

ln AFI = -1.823 + 0.987 ln AI (-1.68) (6.93)

R2 (adj) = 0.5005; N = 48

Thus the elasticity of forest environmentalincome with respect to total income was close tounity (see also the next section, on distribution).Inclusion of control variables (region, foresttype) did not yield significant coefficients anddid not improve the overall fit of the model

(adjusted R2). Omitting theoutliers for total income did notchange signs or significance. Theplot and regression line are foundin Figure 4.1.

Key Sources of ForestEnvironmental Income

What are the key sources of forestenvironmental incomes? Wildfoods, fuel, fodder, and thatchgrass were the economicallyimportant resources (see Table4.2). As discussed earlier, there are

Figure 4.1 The relationship between total and forestenvironmental income

1,500

1,000

500

0

0

-500

2,000

2,500

3,000

3,500

10,0005,000 15,000

Total household income/year

Ab

solu

te e

nvi

ron

men

tal in

com

e


Income from Forest Environmental Resources

good reasons to believe that there wassubstantial underreporting for a number ofsources, including timber, fodder, andenvironmental services. By adjusting the meanincomes for the number of cases where thedifferent activities were reported, the relativeimportance of wild foods, fuelwood, and “othersources” increased. Our impression from thereview is that many of the studies focused on alimited set of forest environmental goods whileneglecting others. The mean income from aparticular source, averaged out over only thosestudies that reported it, may therefore be a morerealistic measure of its relative importance thanthe sample mean (which assumes a failure toreport the source is a result of its irrelevance).

Source of forest environmental income Forestenvironmental income for cases reporting thatsource (USD PPP-adj.) Share of total sample

Household Characteristics and ForestEnvironmental Income

How do household characteristics (caseaverages) correlate with variations in forestenvironmental income? Gaps in the dataprevented thorough econometric analyses. Wetherefore divided the total sample into threegroups of equal size according to the absolutelevel of forest environmental income—high,medium, and low—and compared them interms of household characteristics. Thecharacteristics included age of head ofhousehold, his or her education level,household livestock units, land size, totalincome, and cash income. Only the latter twowere significantly different for the groups (asdescribed later).

The importance of household characteristicswould be better answered by a household-level

Table 4.2 Distribution of forest environmental income, by source

Notes: a The share of total sample income is calculated by multiplying mean income (based only on cases reporting the source) with thenumber of such cases (N), and dividing by the total number of cases (with breakdown into different sources) (51). This is correct if �notreporting� an income source means that it is zero. b The total sample income is calculated by multiplying mean income (based only oncases reporting the source) with the number of such cases (N), and dividing by the total number of cases.

Source of forest environmental income


income for cases reporting that

source

(USD PPP-adj.)

Share of total

sample incomea

Standard deviation

Maximum value

Minimum value N

Wild foods 286.5 38.3 584.4 2,589.4 2.3 20

Fuelwood 215.5 31.7 283.8 1,036.7 1.4 22

Fodder 123.5 5.8 292.9 786.0 0 7

Timber 28.0 2.3 41.6 155.0 0 12

Grass/Thatch 82.5 5.0 76.7 250.0 0 9

Wild medicine 46.5 3.7 78.1 257.3 0 12

Gold panning 6.2 0.2 10.9 26.7 0 6

Others 128.9 13.0 72.6 970.0 0 15

Totalb 677.9 100.0 791.0 3,460.0 1.3 51



External factors Low forest

env. income Med. forest env. income

High forest env. income

Sample mean Anova test Prob>F N

Mean forest env.

income (USD)

101.6 354.3 1577.8 677.9 0.001 51

Total income (USD) 1,501 3,053 4,905 3,043 0.005 48

Relative forest

environmental income

13.9 17.3 37.7 22.0 0.001 48

Elevation (masl) 1,355 1,205 1,103 1,219 nsa 30

Precipitation (mm) 1,057 1,063 1,073 1,062 nsa 34

Wet forest

Semi-wet forest

Dry forest

6

3

8

4

6

5

6

5

2

18

13

16

nsa 45

Gross domestic product

of country (US$)

459 456 1,147 684 0.0079 51

Distance to market

(km)

1.7 19.0 33.5 19.5 nsb 1 out; 0.003 13

Human-to-forest ratio 173 86 100 123 nsb 17

East Africa

South Africa

Asia

Latin America

9

4

4

0

5

6

5

1

1

5

5

6

14

18

14

7

aLikelihood ratio

prob > chi.sq.

0.008

Pearson > chi.sq.

0.015

51

analysis than a meta-analysis of cases. While thelatter in essence compares averages betweendifferent communities, the former will be able tocapture intra-community variations, which issurely the more important objective whenanalyzing household characteristics.

Contextual Factors and Forest EnvironmentalIncome

Chapter 2 discussed a number of factors beyondthe direct control of households that mayinfluence their willingness and ability to takepart in forest environmental income-generatingactivities. These relate to physical, political, andsocial constraints to resource access.

Again, the sample was divided into low,medium, and high forest environmental incomegroups of equal size. Contextual factorsconsidered included forest availability,ecological conditions, market access, overalleconomic conditions, and demographicconditions (see Table 4.3). The sample data didnot permit analysis of legal frameworks andsociocultural conditions.

Only economic and geographical conditionswere found to be significantly different betweenthe three groups. A particular note must beattached to market distance, where increasingforest environmental income was found toimply greater distance to markets. This result

Table 4.3 Forest environmental income and household external factors

Notes: a The chi.sq. test is questionable, as many cells have fewer than 5 counts. b Substantial variations and outliers.



was, however, only significant if an outlier wasremoved, and was based on only 13observations. The results are discussed in moredetail in the next section.

MARKET ACCESS AND OVERALL ECONOMIC

CONDITIONS

The mean gross domestic product (GDP) valuein the high forest environmental income groupwas substantially higher than for the other twogroups. The causal link may of course run bothways here, but living in a generally vibranteconomy will of course imply benefits related toa number of factors such as infrastructure,marketing, credit, and prices.

Forest environmental income seemed to increasewith distance to markets. The normalexpectation would be that superior marketaccess would improve incomes from forestresources. But other factors correlated withmarket access were apparently more important.A remote location generally means both moreabundance of forest resources and feweralternative income opportunities. Again,however, this result must be interpreted withcaution, given the caveats mentioned earlier.

DEMOGRAPHIC CONDITIONS

We had expected that increasing populationdensity would yield lower forest environmentalincomes per household. However, the numberof people, the area size from which resourcesare drawn, and the combined human-to-forestenvironmental ratio did not show anysignificant variation across the income groups.Three of the studies were done in urban areaswith extremely high population densities, butremoving these did not yield significantdifferences.

FOREST AVAILABILITY AND ECOLOGICAL

CONDITIONS

Low, medium, and high forest environmentalincome was not significantly related to the typeof forest or woodland studied (wet,intermediate, or dry). We also divided thesample into three different groups based onthese types of forest, and we compared overallincome, forest environmental income, and forestenvironmental income dependence between thegroups. Only overall income, however, wassignificantly different, with incomes generallyincreasing with “humidity.” Forestenvironmental income was not significantlydifferent, while the level of dependence wasvirtually identical for the three groups.

Possible explanations for these results are thatwet forest environments may be systematicallyassociated with more severe restrictions on use,that drier forests may be more accessible, andthat major product groups such as fuelwoodand fodder may just as easily be harvested fromdry environments. Note also the probableunderestimation of timber values—which wewould expect to be much greater in humidforests—in many of the sampled cases.

LEGAL AND POLITICAL FRAMEWORKS

The institutional framework is an importantconditioning factor for the role of forestenvironmental income in local communities andalso for the income’s distribution amonghouseholds and communities. Many studiesreveal ways in which rules and regulations mayrestrict or facilitate access to forest resources(see Box 4.1).

Our data did not, however, permit any rigorousanalysis of legal and political frameworks. Interms of both property and managementregimes, only a few cases reported these



contextual variables, and there was too littlevariation within the regimes being reported.

REGION AS A COMPOSITE CONTEXTUAL

VARIABLE

As seen from Table 4.3, chi square tests betweenforest environmental income groups (low,medium, high) and region (East Africa, SouthAfrica, Asia, Latin America) yielded significantresults, with relatively more cases from EastAfrica being found among the low-incomecategory, with cases from Latin America tendingtoward the high-income category, and with thetwo remaining regions evenly distributed acrossforest income groups. These results were

confirmed when cases were grouped accordingto region and both total income and forestenvironmental income were compared (seeTable 4.4).

Moreover, Latin American communities werefound also to be significantly more dependenton forest environmental income. This result maybe due to selection bias in our sample. The LatinAmerican sample was biased toward studies ofindigenous people (by Godoy and collaborators,see chapter 5), while the colonos (forest clearingin-migrants)—a dominant forest-margin-dwelling group—were not represented in thesample.

Box 4.1Establishment of Mt. Elgon National Park and the Effect on

the Access of Local People to Resources

In 1993, the Government of Uganda decided to convert Mt. Elgon Central Forest Reserve to a National Park.This was part of a larger, donor-supported effort to transform seven of the most valuable forest reserves inUganda into national parks. With the transfer of this area from the Ministry of Forests to the Ministry ofTourism and Wildlife, rules and regulations for access to forest-related resources changed, as did the authority,goals, and competence of field officers.

A study by Gosamalang (2003) indicates that the conversion had substantial impact on local people’s access toforest resources. Important economic activities such as livestock grazing, hunting, and timber harvesting werebanned. Gosamalang found that the number of households using the forest decreased from 72 to 30 percent,while the number of households involved in cash income-generating activities declined from 14 to 2 percent.The number of livestock held by local people was reduced by some 50 percent. No initial compensation wasoffered. Collaborative models are now being tried out to reduce the effects of the change.

Table 4.4 Incomes by Region (number of cases in parentheses)

Note: Regions as follows: East Africa�Uganda, Kenya, Tanzania, Sudan, and Ethiopia; Southern Africa�Zimbabwe, Malawi, Zambia, andSouth Africa; Asia�India, Nepal, Sri Lanka, and Lao PDR; Latin America�Peru, Bolivia, Honduras, and Nicaragua.

Variable East Africa South Africa Asia Latin

America Kruskal-Wallis test (p-value)

Total income (US$, PPP) 1,697 (15) 2,010 (12) 4,055 (14) 5,676 (7) 0.0017

Forest environmental income

(US$, PPP)

219 (15) 766 (15) 573 (14) 1,681 (7) 0.0012

Relative forest environmental

Income

16% (15) 25% (12) 18% (14) 35% (7) 0.0168



A Closer Look at Forest Environmental IncomeDependence

Dependence on forest environmental incomecan be measured by the share of total income(see Appendix A) or by relative forestenvironmental income, here denoted RFI. Weranked the cases according to this share anddivided them into three equally sized groups:low RFI, medium RFI, and high RFI.

As reported earlier, the relative forestenvironmental income was 22 percent, while themedian was 19.5 percent. Even if the variationbetween cases was substantial, this findingindicates that there is every reason to takeincomes from the environment seriously inresearch, policymaking, and developmentplanning activities.

The pattern emerging from Table 4.5 is that thelow RFI cases were characterized—notsurprisingly—by low absolute forestenvironmental income, but also by aboveaverage total incomes. The medium RFI groupappeared to be the “most fortunate,” in thatthey had above average forest environmentaland also total income. And a large share of theirincome was in cash. They were also on averagelocated much closer to markets than the othertwo groups. The high RFI group, who derivedsome 42 percent of their total income from forestenvironmental resources, were characterized bymuch lower total income levels than the twoother groups. They also scored lower onhousehold capital indicators such as educationand livestock ownership and were on averagethe group that was furthest from markets. It canalso be noted that the medium and high RFIgroups did not have very different forestenvironmental incomes; it was the total incomethat differed between the two groups.

This pattern could be related to the differentstrategies discussed in Chapter 2. The mediumRFI resembles the diversified forest strategies.Forest environmental activities were maintainedin a situation with high incomes but did notconstitute the major source of income. The highRFI, on the other hand, could moreappropriately be considered belonging to one ormore of the three strategies identified as“supporting current consumption.”

We also estimated various regression models onthe relationships between forest environmentaldependence and total income and otherhousehold level and contextual variables, butwe did not find any statistically significantrelationships. The statistically insignificant linkbetween relative forest environmental incomeand total income should be no surprise. Weearlier found the elasticity of total forestenvironmental income with respect to totalincome to be close to unity, meaning that theelasticity for relative forest environmentalincome should be close to zero.

Among household characteristics reported forthe samples, only education was significant.Communities with low dependence on forestenvironmental income had a significantlyhigher educational level than those withmedium or high dependence. This associationbetween high forest environmental dependenceand low educational levels was also confirmedwhen looking at differences between thehouseholds (not just cases): several of the casestudies found that households with loweducation tend to depend more on forests thanthose with higher education levels (see, forexample, Aryal 2002, Fisher 2002, and Stoian2003).

In terms of contextual factors, also provided inTable 4.5, we found surprisingly few statistically



significant relationships. Concerning foresttypes and RFI, we found no significant patternsor even trends. We would expect thatcommunities close to wet forests with highenvironmental values would tend to dependrelatively more on environmental income, butwe did not see this in our sample. The economicconditions seemed to affect the dependence on

forest environmental incomes. Both the highand low dependence groups were associatedwith low GDPs. Long distance to markets alsoseemed to be associated with high levels ofdependence. There was a tendency of low RFIsbeing associated with East Africa and Asia,whereas Latin America and also SouthernAfrica were associated with higher RFIs.

Table 4.5 Distribution of relative forest environmental income, by case characteristics

Notes: RFI = Relative Forest Environmental Income (forest environmental income as share of total income). a Substantial variations andoutliers.

Characteristics Low RFI Medium RFI High RFI Anova test Prob>F N

Relative forest environmental

income (%)

5 19 42 0.001 48

Forest environmental income

(US$, PPP)

173 743 837 0.008 48

Total income (US$, PPP) 3,339 3,945 1,846 nsa 48

Cash income (US$, PPP) 555 2,861 692 nsa 18

Education level (years) 6.4 4.8 4.4 0.09 25

Elevation (meters above sea

level)

1,430 1,187 990 nsa 28

Precipitation

(millimeters/year)

1,386 858 998 nsa 31

Wet forests

Semi-wet forest

Dry forests

8

4

4

3

4

5

5

4

5

nsa 45

Gross domestic product per

capita. (US$, PPP)

429 692 497 0.008 48

Distance to market

(kilometers)

22 7.9 34.7 nsa ; removed 1 case

and got 0.002

13

Man-to-forest ratio

(people/km2)

227 73 87 nsa 0.1 17

East Africa

South Africa

Asia

Latin America

7

1

8

0

5

5

3

3

3

6

3

4

Likelihood Ratio

prob>chi.sq. 0.02

Pearson>chi.sq.0.05

48

N 16 16 16 48



Population density did not seem to have muchimpact on the levels of dependence, even if wesee that the low dependence group hadsignificantly higher population density, in linewith what we would expect.

In a cross tabulation with region and forest type,for the Latin American cases close to 41 percentof the income was derived from forest resources(see Table 4.6). The forest dependence was alsohigher than average in the Southern Africancases. Furthermore, there were smallerdifferences across forest types than betweenregions.

The dependence on environmental incomes wasgenerally high and it also varied considerablyamong cases. Communities with highdependence tended to have lower overallincomes, they were further from markets, theyhad lower population densities, and LatinAmerica and Southern Africa were over-represented among them. How representativeour average findings are for developingcountries in general is, however, hard to tell.There are reasons to believe that the sites formany of our cases were selected preciselybecause forest environmental goods wereassumed to be important. We take thisdiscussion a step further in Chapter 5.

What is the Distribution of ForestEnvironmental Income?

How does forest environmental dependencechange with total income? We analyzed this attwo levels. First we conducted an inter-community analysis—that is, we compared andanalyzed the cases (communities) to identifydistributional patterns in forest environmentaldependence between them. Then we conductedan intra-community analysis, where weinvestigated the same questions at thehousehold level: What can the studies tell aboutthe variations in dependence between differentgroups of households within each community?

Forest Environmental Income and IncomeDistribution Between the Cases

In Table 4.7, the sample cases have been dividedinto quintiles based on total income. The totalforest environmental income increased as wemove from the bottom to the top quintile. Thereis, however, an interesting pattern emergingwhen it comes to the relative forestenvironmental income: the relationship appearsto be bell-shaped, with the highest share for themiddle-income quintile.

To test the bell-shaped relationship, we applieda quadratic regression model with relative forest

Table 4.6 Relative forest environmental income, by Region and forest type (number ofcases in parentheses)

Forest type

Region Wet Semi-wet Dry

Total (N)

Eastern Africa 18.8 (4) 17.2 (7) 14.4 (4) 16.9 (15)

Southern Africa — 25.7 (2) 25.0 (8) 25.2 (10)

Asia 7.0 (7) 25.0 (3) 26.7 (2) 14.7 (12)

Latin America 40.9 (5) — — 40.9 (5)

Total 20.5 (16) 20.6 (12) 22.2 (14) 21.1 (42)



environmental income as the dependentvariable and total income as the independentvariable. It did not yield any statisticallysignificant results using either the basic model,which included various control variables, or amodel where some of the extremely high-income cases were excluded. Thus theregression analysis could not confirm the bell-shaped relationship illustrated in the table.

But the alternative approach used in Table 4.7still demonstrates that the poorest and richestgroups have a significantly lower dependenceon forest environmental income than theintermediate group.

Distribution of Forest Environmental IncomeWithin the Cases

To look at what the cases tell about thedifferences in forest environmental incomeamong households at different income levels,we focused on three measures—the Ginicoefficient, income quintiles, and the Kuznetsratios (see Appendix A). The central question is:Do poor households within communitiesdepend more on the forest environment as asource of income than wealthier households do?

Only a few of the cases reported on the intra-community distribution related to forest

environmental income. Table 4.8 summarizesthe seven recent studies that reported the Ginicoefficients for household income both with andwithout forest environmental income. Anadditional six studies calculated the Ginicoefficient for one of the income measures.

Forest environmental income reduced incomeinequality between households significantly.Although there were only a few cases, theresults were clear and in line with what othershave observed and with the discussion inChapter 2. On average, the Gini coefficientincreased from 0.41 to 0.51 when forestenvironmental income was excluded from thecalculations. Only in one case (from India) didincome inequality decrease slightly when forestenvironmental income was excluded.

Some of the case studies did carry out moredetailed analyses of dependence on forestenvironmental income across different incomegroups. Table 4.9 shows that absolute forestenvironmental income was highest for therichest income group, but the differences weresmaller than perhaps could have been expected.We need, however, to take into account the factthat the sample is small, particularly for thevery rich and very poor groups. We further seethat forest environmental income sharedecreased with increasing total income, also inline with what we hypothesized.

Table 4.7 Forest environmental income for quintiles of cases (standard deviation in parentheses)a

Note: a The sample consists of 48 cases from 12 different studies; not all studies had information for all quintiles, which affected theestimations of total average income as well as the forest environmental income share.

Income quintiles

Variable Lowest 20% 20–40% 40–60% 60–80% Top 20 % Total

Total income (US$ PPP)

381 (157)

1,195 (263)

2,301 (394)

3,647 (457)

7,568 (3,325)

3,043 (2,984)

Forest env. income (US$ PPP)

90 (67)

305 (218)

654 (471)

876 (787)

997 (974)

584 (678)

Forest env. income share (%)

21.9 (14.6)

24.4 (13.8)

29.0 (20.6)

23.4 (19.5)

11.9 (8.2)

22.0 (16.3)



Income category

Mean forest environmental

income (US$, PPP)

Relative forest environmental income (%)

Standard deviation Minimum Maximum N

Very rich 897.40 16.85 10.25 7.0 28.3 6

Rich 516.00 17.90 14.30 0.5 35.7 12

Average 476.00 20.57 12.72 4.4 39.3 11

Poor 497.40 32.43 20.60 5.0 79.0 12

Very poor 376.30 31.96 14.28 6.6 50.0 7

Table 4.8 Gini coefficients with and without forest environmental income

Table 4.9 Income distribution within cases and forest environmental income

In Appendix A we suggest using theEnvironment Kuznets Ratio (EKR) as a measureof the poverty profile of forest environmentalincome. We can differentiate between the

absolute and relative EKR. The absolute ratio(AEKR) measures the relationship between theaverage incomes of the richest and the poorestsegments. It is thus a measure of the forest

Study Gini for total income

Gini without

forest env. income Change (units)

Aryal (2002) 0.55 0.61 (0.06)

Botha (2003) 0.41 0.54 (0.13)

Fisher (2002) 0.49 0.56 (0.07)

Shaanker et al. (2003b) 0.23 0.43 (0.20)

Shaanker et al. (2003a) 0.56 0.76 (0.20)

Shaanker et al. (2003c) 0.36 0.34 0.02

Stoian (2003) 0.26 0.34 (0.08)

Average 0.41 0.51 (0.10)

Sharma (1999) 0.35

Cavendish (1999a) 0.29

Cavendish (1999b) 0.39

Muderis (1998) 0.36

Nsubuga (1998) 0.25

Nyland (1998) 0.17

Average total 0.36 0.49 (0.13)



environmental income disparity within thecommunities. The relative ratio (REKR)measures the average forest environmentalincome share of these two segments. It is thus ameasure of the forest environmental incomedependence disparity within the communities.

From Table 4.9 we find that the average REKRwas about 0.59, using the six cases that reportedthe income shares for the very rich/very poorratio. Comparing the income shares for therichest and poorest (for which we have moreobservations) yielded a similar value. Thus, thepoorest households were close to twice asdependent on forest environmental income asthe richest.

Our statistical analysis further demonstrated apositive correlation between REKR and forestenvironmental income (both absolute andrelative). In other words, lower forestenvironmental incomes were associated withthe poor having a high dependency on theforest environmental incomes (low REKR). Thisresult is in line with what would be expectedfrom our “employment of last resort” model inChapter 2: low forest environmental incomesindicate that they are low-return activities,making them attractive for only the poorestgroups.

To summarize, for inter-household differencesthere is a relatively clear pro-poor profile forforest environmental incomes. This tendency ismuch weaker for the inter-communitycomparison, where the link between relativeforest environmental and total incomes is weak.The methodological problems involved in aninter-community analysis notwithstanding, ourfindings thus suggest that the pro-poor profileof forest environmental income is importantprimarily within communities rather thanacross communities.

How is Forest Environmental IncomeUsed in Household DiversificationStrategies?

We looked at relationships between levels oftotal and forest environmental householdincomes and households’ diversificationstrategies. Rural households pursue a widerange of economic activities to secure a living.These can have different motivations involvingsafety nets, insurance, consumption support,income maximization, and capitalaccumulation. Diversification is thus animportant feature of rural life, and incomesfrom the forest enter into complex relationshipswith other household activities. In this sectionwe look first at different sources of householdincomes, then we look at the relationshipbetween number of activities and income levels,and finally we look at diversification strategiesto secure cash incomes.

Diversification and Main Economic Activities

Most of the people in households in the selectedcases were agriculturalists with varyinginvolvement in crop production, livestockproduction, forest environment-relatedactivities, and off-farm engagements of differenttypes. Earlier we looked at levels of incomeacquired from different sources and found thatin most cases agriculture was the key incomegenerator, followed by off-farm activities andforest environmental incomes. Income levelsfrom different activities constitute a simplemeasure of diversification.

We constructed a “strategy variable” where, foreach case, we set up a ranked list of activitiesaccording to their economic importance in thecases, with 1 as the most important activity. Themain income strategy was agriculture, followedby off-farm activities (see Table 4.10). Forest



environmental activities were ranked as thethird most important strategy for those caseswhere it was reported. We note that forestenvironmental incomes were roughly at thesame level as off-farm activities and livestock.

Number of Activities and Income Levels

Do household incomes increase withdiversification? To get a measure of the degreeof diversification in terms of number ofactivities, we used a diversification index, DITI(see Appendix A), in which a value of 0represents no diversification (a single source)and a value of 1 represents maximumdiversification (an infinite number of sources ofequal size).

There was substantial variation in the degree ofdiversification among the 22 cases reporting onlivelihood sources, ranging from less than 0.1 toalmost 0.9. The majority of cases (15), however,fell between 0.5 and 0.7.

It is possible, of course, to discuss the directionin which causality between income anddiversification will go. We ran a regression withabsolute income, denoted AI, as theindependent variable, using a quadratic

function. The plot and regression line arepresented in Figure 4.2. The sample of 22 isquite small. Nevertheless, the regression yieldeda function suggesting that diversification isreduced when income levels increase:

DITI = 0.62256 – 0.0001 AI – 8.13E-9 AI2

(5.96) (-0.20) (-0.9)

R2 (adj) = 0.36; N = 22; Prob>F = 0.0052

Related to our discussion in Chapter 2, thisresult may support the claim that diversificationcomes at the expense of higher income or that insome types of diversification rural dwellersmust make do with whatever is available at anygiven time. Higher income, then, seems to beassociated with some specialization in one or afew higher return activities.

What is the relationship between diversificationand forest environmental incomes? The testyielded a rather weak but statisticallysignificant relationship, in that the total forestenvironmental income increased with increasingdiversification up to a certain point. Beyond thispoint, diversification decreased again. However,we also ran the test for relative forestenvironmental income and found a similar butmuch more significant relationship (see Figure

4.3). The sample isstill small andinterpretations caneasily becomespeculative.However, we see thatsamples with lowforest incomedependencediversified less thanhouseholds withmedium forestincome dependence.

Table 4.10 Importance of various income sources

Note: a For the N cases where this type of income is reported.

Activity Averagea Standard deviation Minimum Maximum N

Crop 1.9 0.93 1 4 41

Off-farm 2.4 1.08 1 4 27

Forest 2.6 1.31 1 6 38

Livestock 2.8 1.47 1 5 28

Wage work 3.3 1.49 1 5 12

Remittances 3.4 0.89 3 5 5

Others 3.6 0.89 2 4 5



Figure 4.3 Relative forest environmental income and diversification

FEI

DIT

I n

ew

0,4

0,3

0,2

0,1

0,0

0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70

0,5

0,6

0,7

0,8

0,9

DITI = 0.17381 + 3.25347 RFI – 4.57824 RFI2

(3.33) (6.19) (-5.00)

R2 (adj.) = 0.70; N = 22; Prob>F 0.001

It would have been relevant also to investigatethe relationship between cash income (forestenvironmental and total) and diversification,but we did not have sufficient data for ameaningful test of these relationships.

Cash Forest Environmental Income Strategies

Rural households have strategies for totalincome generation and also for cash incomes.Cash is needed both for consumption of goodsand services as well as for various investmentsand inputs in production.

We would generally expect that cash income, aswell as the cash share of income, increases withoverall incomes. Well-to-do households (andcommunities) will typically be more integratedin market economies (the causal link probablyruns both ways). We did observe this in oursample. The purpose here, however, is toinvestigate whether this is true also for absoluteforest environmental cash income (ACFI): doesit increase with total forest environmentalincome?

The cash forest environmental income wasreported in only 17 cases, with an average ofUS$202 per household. For these 17 cases, thetotal forest environmental income was slightly(but not statistically significantly) higher thanthe average for the full sample: US$691

compared with US$678.The distribution wasquite skewed—with alarge number of caseswith no or very lowcash income levels,while one case reportedincomes aboveUS$1,000.

The average cash sharewas 53 percent forforest environmentalincome. Some casesreported 100 percentcash incomes. This wasprobably due to the

Figure 4.2 Total income and diversification

Total household income

DIT

I n

ew

0,4

0,3

0,2

0,1

0,0

0 5,000 10,000 15,000

0,5

0,6

0,7

0,8

0,9



objective of theparticular study (a focuson cash incomes) ratherthan the households inthe sample not engagingin subsistence forestenvironmentalactivities. We alsosuspect that cashincome was reportedonly in cases where itwas important and thatthe cash share of 53percent therefore wasnot representative forthe full sample.

We earlier asked howforest environmentalcash income changes with both forestenvironmental and total income. We consideredhow the forest environmental income cash share(FICS, which is forest environmental cashincome divided by all forest environmentalincome) changes with total forest environmentalincome in the communities.

In Table 4.11 the average FICS is computed forthe three groups of cases, based on level offorest environmental income. The results aresurprisingly clear: communities with low forestenvironmental income had a much higher cashshare than those with medium and high levels:

78 percent compared with 32 percent and 28percent for the latter groups. In absolute terms,however, we did find—quite naturally—that thehigh forest environmental income groups alsohad the highest forest environmental cashincome.

The link between forest environmental incomeand the cash share can also be studied by aregression analysis. A quadratic regressionwhere the FICS is explained by total forestenvironmental income (AFI) yielded the

following (and see Figure 4.4).

Table 4.11 Cash forest environmental income share, by household income groups

Income source and share

Low env. income

(standard deviation)

Medium env. income


High env. income (standard deviation)

Anova test Prob>F N


cash income

65.9 (103.3) 119.8 (146.1)

484.4 (130.7) Rsquare = 0.32

Prob>F 0.063

17


income cash share

0.78 (0.12) 0.32 (0.17)

0.28 (0.16) Rsquare 0.36

Prob>F 0.042

17

Figure 4.4 Relationship between cash share and total forestenvironmental income

Absolute forest environmental income

EIC

S

0,4

0,3

0,2

0,1

0,0

-500,000 1,000,000 2,000,000 3,000,000500,000

0,5

0,6

0,7

0,8

0,9

1,0

1,1



FICS = 0.79416 – 0.00078 AFI + 2.11E-7 AFI2

(6.16) (-2.38) (1.86)

R2 (adj.) = 0.27; N = 17; Prob>F 0.0442

A U-shaped curve gives the best fit, with bothcoefficients significant at the 0.10 level. Thus thecash share appeared to be highest for caseswhere total forest environmental incomes wereeither very high or very low. With medium-levelforest environmental incomes, the cash sharewas low. As can readily be seen from the Figure,the result is sensitive to the outlier with highforest environmental income and high cashshare. Removing this yields forestenvironmental income cash share as a globallydeclining function of total forest environmentalincome.

These results are surprising. It is commonlyassumed that at low levels of forestenvironmental income the activities are mainly

for subsistence, while higher levels areassociated with a high cash share. After all,there are limits to how much wild food a personcan eat and how much firewood can be burned.The figure shows that for most cases the cashshare was low, but there were a few cases withboth high cash shares and low levels of totalforest environmental income.

In short, cash income and market access are alsoimportant for communities at low levels offorest environmental income. In fact, they arerelatively more important for thesecommunities. We also analyzed the relationshipbetween the FICS and total income levels andfound a similar, but weaker, relationship.

We should again note, however, the reservationsthat attend these results, given the smallnumber of cases and potential biases in thesample.


Best Practices in ForestEnvironmental Income Studies

This chapter looks at problems generally foundin research on environmental income and at theparticular problems in the case studies welooked at. It also provides guidelines on bestpractices in each problem area identified.

Theoretical and Methodological Pluralism

Our review of different cases revealed a highdegree of theoretical and methodologicalpluralism. All studies contained sections dealingwith forest environmental income. The studieswere, however, undertaken for differentpractical reasons, for different scientificpurposes, by people with very differentscientific backgrounds and proficiency levels,and with quite different budgets and ambitionlevels.

Differences in the Purpose of the Study

The studies reflected a variety of researchobjectives, and we categorized them alongcertain key purpose categories: valuationstudies, studies of environmental incomedependence, and socioeconomic studiesfocusing more on internal distribution ofincomes within local communities. A studyaiming to explain rapid forest conversion willprimarily focus on comparing per-hectareprofitability of sustainable forest extraction andagriculture and might not look for datadescribing forest environmental income

dependence within the communities. In a studyconsidering the effects of the creation of anational park on a local population, on the otherhand, such data will be crucial. Practically allvaluation studies of non-timber forest productsmeasure the absolute value of forest extractionsby local populations and were thus of interestfor our study at a basic level. But surprisinglyfew studies concerned themselves with levels ofdependence and the internal distribution ofincomes within communities. This made itdifficult to ascertain the relative importance ofthe forest-related environmental incomes.

Differences in focus may, however, also affectestimated forest environmental income levels.While valuation studies often aim to ascertainthe socioeconomic value of forests at thenational or even global level, other studiesmight try to examine the private economicvalue of resources for a community or forindividual households. The decision to includeenvironmental services is discussed furtherlater.

The stated purpose of the case studies wasclassified as overall forest values, dependenceon forest environmental income, distribution offorest environmental income, or a combinationof these three. Table 5.1 shows that most of thecases were socioeconomic studies aboutdependence on forest environmental incomes orvaluation studies; only a few went into detail ondistribution of the forest environmental income.

5



Purposes affect results. We found a ratherstrong correlation between the purpose of thestudy and total forest environmental income:valuation studies systematically reportedalmost three times higher total forestenvironmental incomes than the environmentaldependence studies did. “Distribution” studiestended to report very small numbers. The largedifference is partly explained by the fact thatfive of the seven studies from Latin America,which has a much higher average income thanother regions, were valuation studies. But forSouthern Africa, with five valuation studiesincluded, the cases also yielded estimates offorest environmental income about twice theaverage for all studies from that region.

This finding is interesting. One possibleexplanation is the inclusion of public goods andglobal values in valuation studies, even if weare not able to document this in our study (asdiscussed later). It could also relate to thedifference between potential and actual usevalues. Moreover, valuation studies mightsystematically seek areas with very highbiodiversity and existence values, which seemslike a reasonable explanation. More dubiousattempts to consciously inflate values invaluation studies cannot be discounted,however.

Relative forest environmental income followedthe same pattern as total income, but theexplanatory power of the test was weaker there.

Types of Studies

The particular scientific approach used andeven the proficiency levels of researchers canalso have an impact on results. The sampleincluded research articles, theses works, andconsultancy reports. Differences in the type ofstudy might affect results through associatedvariations in time, funding, ambition levels, andproficiency.

We tested and found significant differencesbetween theses and research studies (see Table5.2). Theses systematically reported lower totalforest environmental incomes than the researchworks.

A number of underlying factors may beaffecting this result: there were more thesisstudies from low-income countries in Africa,they were based on shorter fieldwork, they hadless funds, and they typically focused onhousehold dependence and distribution.

Methodological Approach

The studies reflected different methodologicaland conceptual approaches: household

Table 5.1 Purpose of study, by total and relative forest environmental income

Forest income

Valuation studies


Dependence on env.

income studies

Distribution of env. income

studies Combination of

purposes Anova test

Prob>F N

Total forest

env. income

(US$/househo

ld/year)

1,405 (1,050)

N= 14

525 (505)

N=26

152 (111)

N=5

230 (149)

N=9

R square 0.35

Prob>F 0.001

54

Relative forest

env. income

32% (175)

N=11

21% (136)

N=23

26% (293)

N=5

19% (218)

N=9

R square 0.02

Prob>F 0.02

48


Best Practices in Forest Environmental Income Studies

economic approaches, stakeholder approaches,and broader valuation studies. Even among thestudies with a household economic approach,however, surprisingly few focused explicitly onmore “sophisticated issues” around poverty,local heterogeneity, entitlements, anddistribution. Just seven studies involved the useof Gini-coefficients, and we were only able togenerate Kuznets ratios for 12 studies. Thusonly a small number looked at the communityinternal distribution between groups ofhouseholds for forest environmental incomes.Hardly any studies identified or focused ondifferent forest environmental income functionsand strategies or disaggregate forest income tocapture variations depending on different goodsor categories of goods.

Measures related to forest income dependencewere also mostly very crude, although stilluseful. The most common approach was to lookat the share of forest environmental incomerelative to total income.

Last, few studies linked forest environmentalincomes to broader contextual factors beyondthe household level. Households’ access toresources under different legal, economic,political, and ecological conditions—and howaccess affects the ability and willingness togenerate forest environmental income—wasgenerally neglected.

Other Factors

Researchers from different disciplinesinvestigate the importance of environmentalresources and people’s livelihoods.

An epistemic framework is characterized byparticular common values, norms, and ways toaddress research problems (Knorr-Cetina 1981).Especially in such a young and heterogeneousapplied field as environmental income, it maybe that the more diffuse and intractable problemof researcher bias with respect to achievingsome more or less hidden objectives isparticularly present. Sheil and Wunder (2002)state that some researchers inflate numbers inorder to make their work more interesting forother researchers and for funding agencies.They accuse Peters, Gentry, and Mendelsohn(1989), authors of a well-known valuation studyfrom the Amazon, of producing inflated valuesfor NTFPs. According to Panayotou and Ashton(1992), the study by Peters, Gentry, andMendelsohn sought and found an audience thatwanted NTFP extraction from tropical forests tobe economically competitive. It was notpossible, however, to rigorously investigate thistype of problem in a meta-study like ours.

Even if we were not able to assess this in detail,most of the reviewed cases had a disciplinarycore. Some cases were dominated byeconomists, ecologists, foresters, or

Table 5.2 Type of study by total and relative forest environmental income

Forest income

Thesis

(stand. dev.)

Research

(stand. dev.)

Consultancy

(stand. dev.) Anova test

Prob>F N

Total forest env.

income

(US$/household

/year)

314.7 (288)

N=20

988.2 (937)

N=28

202.7 (67.9)

N=3

R square 0.19

Prob>F 0.006

51

Relative forest

env. income

17% (14)

N=20

28% (17)

N=25

8% (3)

N=3

R square 0.16

Prob>F 0.02

48



anthropologists; other studies were moreinterdisciplinary. It could well be that thedisciplinary focus of a case affected the results.An ecologist, for example, might find or identifymore and higher biodiversity values on accountof his or her knowledge. An anthropologistmight detect more use-values because of aninclination to listen and talk to local people andto observe through participation. Timeunfortunately did not allow for interviews withinvolved researchers to clarify the origin of“dominating researchers” for each case.

Last, the level of research and resource inputcan have a direct impact on results. Long-termstudies with substantial fieldwork inputs mightlead to the discovery and registration of moresources of environmental income, and thushigher income levels. And the precision of dataand possibilities to correct for contextualvariables of different types could also increasewith the resource input.

To sum up, a meta-study like ours must usewhatever data are available in the casesexamined. The wide disparity in the range anddetail of variables reported in these cases doesnot necessarily signify glaring omissions or poorquality. Instead, the disparity may reflectvariation in why, how, by whom, and for whomthe study was carried out.

Best Practices

Most of the problems just described were notprimarily due to flaws in technique ormethodology but instead linked to differencesin the purpose of the study, the specificcompetence of the investigators, or the scopeand methodology applied. Since most of theseproblems were given at the outset, it is difficultto recommend a set of “best practices.”

Some general points can, however, be outlined:

• The findings imply a need for researchers tobe self-critical of and self-reflective abouttheir theoretical approaches andmethodological practices.

• Use the information contained in the datacollected to its fullest extent; environmentalincome measures will generally be based onaggregation of measures for different goods,so the reverse process—disaggregation—should not pose any difficulties. And thedependency, distribution, anddiversification implications for a good suchas timber may be very different from thosefor wildlife or medicinal plants.

• From a research user perspective, it isimportant to note that results might beaffected by the way the TOR is formulated,the objectives defined, the time andresources allocated, and the disciplinarybackground of researchers.

Economic Concepts and Bookkeeping

Price and Cost Estimates

In terms of price and cost estimates, unrealisticassumptions are often made using, for example,national minimum wages or world marketprices as proxies for the absent localequivalents. Estimating the actual opportunitycosts of labor in the presence of substantialmarket distortions can often be difficult andresource-demanding.

The studies usually used local market priceswhen markets existed. When markets areimperfect or missing, problems may arise. Insome cases researchers omitted goods that werenot bartered or traded locally, such as wild



fruits, thatching grass, and wild animals. Thisimplies that forest environmental incomes wereunderestimated.

Other studies used shadow prices forestimation. Godoy, Brokaw, and Wilkie (1995)used the contingent valuation method. Thismight create an upward bias. A few studiesused local prices from the nearest markets,which may give biased results depending on thecharacteristics of the market. Some studies (suchas Cavendish 2000) used substitute prices. Oneproblem with this method is to find closesubstitutes.

In some cases, prices used to value all craftswere the prices paid by tourists, which areusually higher than the local prices.

In general, average prices rather than marginalprices were used; this is understandable, giventhe difficulties generally associated withdetermination of the latter.

BEST PRACTICES:

• For goods that are sold, the actual price andincome obtained by the household shouldbe determined; differences in market accessacross households may be important.

• For goods consumed, the average forestgate price for all households should beused; use of world market or even localmarket prices may lead to significantdistortions.

External Factors

Many external factors affect the possibility andwillingness of actors to acquire forestenvironmental incomes, such as legalframeworks and rights of access for different

groups; political, administrative, and socialinstitutions; and other sociocultural factorsaffecting access. One problem was that manystudies only looked at potential economicvalues, as stated by Sheil and Wunder (2002):“Social obstacles to realization of potentialforest values were generally ignored.”

We found good reasons to stress the accessissue—not only in the difference betweenpotential and realizable economic values, buteven more so in dependence, distribution, anddiversification studies. The often sociallyconstrained access of some groups to particulartypes of forest products is a key reason whysome households but not others can use theforest as a way out of poverty.

BEST PRACTICES:

• Collect and use data on importantexogenous variables related to climate,ecology, economy, law, and otherinstitutions. These may inform generallevels of environmental incomedependence, and their omission fromsubsequent data analysis might, if localvariations are present, lead to significantbias in results and interpretations.

Processed Goods

A more subtle problem concerns the definitionof environmental income in terms of its locationin the market chain. While in one case goodsendlessly processed and exchanged areincluded as long as these processes take placelocally, in others only raw materials areconsidered. This does not necessarily representa problem for the particular study itself, as longas a consistent and conscious choice is made.But this is not always the case.



Table 5.3 indicates that the inclusion ofprocessed goods as forest environmental incomeaffected estimates for such earnings. It yieldedan average of US$1,263 per household per yearcompared with US$434 for studies that did notinclude processed goods. We also found that therelative environmental income was different inthe two sub-samples, with 31 percent for thecases including processed goods compared with19 percent without valued processed goods.Both findings indicate inconsistency in the waysthe term “forest environmental income” wasdefined and used by different authors. Itconstitutes an important problem and calls forcaution when comparing cases.

BEST PRACTICES:

• Use a definition of environmental incomebased on rent realized in the first link of themarket chain.

• If particular circumstances merit the use ofan alternative definition, it is important thatthis alternative definition is clearly statedand applied throughout.

Environmental Services

Many studies neglected the role of forestenvironmental services in local communities in,for example, the form of erosion control andstabilization of water flows. Again, the intendedscale of the study is of relevance here; butresearchers, when investigating the value of aresource at community or household level, mayfail to examine not only spillovers ordownstream effects of environmental servicesbut also the significance of these services tolocal people. The more traditional non-usevalues are of less significance to localcommunities.

Figure 5.1 Types of forest environmental services included in the studies

Table 5.3 Inclusion of processed goods in valuations of forest environmental income

Forest income

Value processed goods (standard

deviation)

Do not value processed goods (standard

deviation)

Anova test

Prob>F N

Total forest env.

income

(US$/househol

d/year)

1,262.6 (1,133)

N=15

434.3 (415)

N=36

R square 0.23

Prob>F 0.0003

51

Relative forest

env. income

31% (12)

N=12

19% (16)

N=36

R square 0.12

Prob>F 0.02

48

14

3

10

5 5 5 5 4

10

0

24

6

8

1012

14

16

Env. service in

general

Carbon se

questrati

on

Rural w

ater r

etention

Urban w

ater s

upply

Soil ero

sion preventio

n

Wind ero

sion preventio

n

Electricity

Other services

Biodiversity



The results presented in Figure 5.1 indicate thatmany researchers neglected the roles that forestenvironmental services play not only at nationaland community levels, but even at individualhousehold levels. Many of these studies werecarried out to find the dependence on forestenvironmental activities and should thereforehave included forest environmental services.

Out of the 54 cases studied, only 26 percentdiscussed some kind of forest environmentalservices, mainly stressing water retention. Fewincluded important services like erosionprotection and clean water. And only one studyattempted to quantify incomes from forestservices (Sjaastad and others 2003).

This supports the argument that reported forestincome estimates often represent a lower boundof actual values, all other factors apart.

BEST PRACTICES:

• Establish, at an early stage, theenvironmental services that are ofimportance to the local population.

• Gauge the degree to which environmentalservice values will vary across households.

• For important services with homogenousvalues across households, broadercommunity-wide estimates are sufficient;acceptable approximations may be availablein the literature.

• For important services with variable valuesacross households, a separate household-level valuation will be necessary.

Double Counting

The double or even triple counting of benefits isa common problem in many studies. This can

occur in numerous ways, including thefollowing:

• The benefits of tourism are added to benefitestimates of biodiversity, where the value oftourism already is partially included.

• Government revenue is added on top ofvaluation estimates (although it is “just” atransfer of income).

• Replacement costs for wells or pumpsneeded to counter lower groundwater levelsare added to afforestation costs.

• Costs of livestock lost to wild animals areoften computed both directly, in terms ofstock value, and as a loss of present andfuture income flows—a failure todistinguish between stocks and flows.

• The value of fodder harvested from theforest may sometimes be included as abenefit in calculations, but not subsequentlysubtracted as an input cost into livestockproduction, escalating the householdincome estimates.

The first two examples are mainly associatedwith more large-scale valuations and thus oflimited relevance to our meta-study. The twolatter problems, though, are common in cases ofthe type considered here.

BEST PRACTICES:

• Focus on end products and goods and workbackwards. For example, loss of cleandrinking water may be countered throughafforestation, drilling of boreholes, sinkingof wells, or municipal plants. Are allmeasures required or just one, and in thelatter case, at what cost?



• Focus on production processes. Does oneparticular forest good enter into theproduction process of another forest benefit,as in the case of fodder and livestockproduction?

Stocks, Flows, and Sustainability

A problem of particular importance forvaluations is the existence of alternativemeasures of the goods contained within forests.Batagoda and others (2000) distinguished fourdifferent measures:

• The value of the stock, or inventory

• The value of maximum sustainable yield, ormaximum potential flows

• The value of maximum potential flows asconstrained by access (legal or physical)

• The value of actual extraction.

Stock values and different measures of flowsmay yield widely disparate figures. Godoy,Lubowski, and Markandya (1993) assert that:“For most purposes, the value of the inventoryis a meaningless concept related neither topresent or to sustainable use.” Stock value maybe of interest in studies of rapid forestconversion and strategies related to suchconversion, but it has no direct bearing on forestincome estimates.

The third category, while rarely applied invaluations, again highlights a wider problem:the necessity of considering law, conventions,and other social institutions and constraintswhen analyzing and explaining resource valuesand their distribution.

The main point, however, is that researchersshould be aware of the different options and

should carefully specify the type of measurethat is targeted. This is often not the case, asboth Batagoda and others (2000) and Godoy,Lubowski, and Markandya (1993) point out.

A related problem concerns sustainability ofextraction. Regrettably, assumptions about thesustainability of actual extraction are oftenmade beforehand, with no subsequentinvestigation of whether these hold. Thus astandard practice is to measure actual extractionand treat this as if the values derived could beperpetuated automatically. This could createproblems, especially when calculating netpresent values of forest incomes.

In many cases, however, there is a genuine lackof knowledge about sustainable yield levels, aswith medicinal plants, bamboo shoots, rats,caterpillars, and butterflies, so the omission mayoften be for a reason.

Finally, the question of the extent to which use,potential or actual, is efficient (in theconventional economic sense of equalizingmarginal costs and benefits) is very rarelyconsidered.

In Chapter 2 we noted that many studies didnot fully take into account whether present useis sustainable. Net present value calculationsassuming that present use can go on “forever”without considering whether present use iscutting into stocks will in general tend tooverrate forest environmental incomeestimations.

Many of the studies that did addresssustainability issues did it based primarily oninterviews with local people and not on detailedfield-level ecological studies, which are moreappropriate.



Variable Total forest env. income (US$/household/year)

Standard deviation N

Anova test

Prob>F

Included sustainability issues 882.9 949.2 29

Neglected sustainability issues 407.7 392.4 22

R square 0.09

Prob>F 0.03

Around half the studies we reviewed did notaddress sustainability issues at all (see Table5.4). We found that studies that tooksustainability into account seemed to havehigher forest environmental incomes than thosethat did not.

By using the interview material, we furthermorefound a strong positive correlation betweenlocal people stating that there was a sufficientsustainable supply of resources and the level offorest environmental income (see Table 5.5).

Again, ideally the studies should addresssustainability through elaborate ecological fieldinvestigations, preferably over time, and alsofollowing the different types of key resources inquestion. Very few studies did this.

These results underline the importance ofconsidering sustainability.

BEST PRACTICES:

• Be clear and consistent in the choice ofmeasures across different forest goods—maximum sustainable yield values, access-constrained values, or values based oncurrent extraction.

• Unless reliable data on the sustainability ofcurrent extractive practices are available,avoid present value estimates based ondiscounted periodical measures; instead,focus on periodical estimates only.

Cash Versus Subsistence Income

Many studies failed to separate cash incomeand consumption, or—even worse—studiescounted only cash income as income. Thisproblem may take many forms, but the mostcommon is the simple omission of consumptionin reported income estimates.

In the cases reviewed, six studies defined totalincome as cash income. This problem wasidentified both in student theses and other typesof work. As can be seen in Table 5.6, studiesthat defined cash income as total income tendedto underestimate both forest environmentalincomes and forest environmental incomedependence.

Table 5.7 shows that only 17 studies (31.5percent) explicitly recorded cash forest incomesat all. Testing for the difference relative to totalforest environmental incomes, however, did not

Table 5.4 Inclusion of sustainability issues, by total forest environmental income

Table 5.5 Perceptions of change in resource supply, by total forest environmental income

Variable Total forest env. income (US$/household/year)

Standard deviation N

Anova test

Prob>F

Insufficient resource supply 511.9 570.7 22

Same as before 2,021.9 747.1 5

Sufficient resource supply 2,117.9 1,897.2 2

R square 0.0.49

Prob>F 0.0001



Variable

Total forest env. income

(US$/household/ year)

Standard deviation

Minimum

Maximum

N

Cash income not included 671.2 723.8 1.3 3459.5 34

Cash income included 691.2 935.3 2.9 3186.5 17

yield statistical significant differences. We foundthe same non-significant trend for relative forestenvironmental income.

Omission of cash incomes will not necessarilylead to underestimation; the problem iscommon, for example, in studies where the onlymethodology applied is enumeration of goodsremoved from the forest and the subsequentmultiplication of these with general prices.There may be valid reasons for not recordingcash incomes, but the omission will alsoseverely limit the range of feasible dataanalyses.

BEST PRACTICES:• Valuate cash and subsistence income

separately (see also the section on price andcost estimates and Appendix A).

Omitted Variables

Omitted variables commonly pose a problem.Godoy, Lubowski, and Markandya (1993), forinstance, in reviewing some of the earlyliterature on valuation of NTFPs, identified atendency to examine either flora or fauna butnot both.

They also mentioned the neglect of labor costs.Given that forest environmental income can bedefined in terms of both rent and value added,the “neglect” of labor costs may simplyrepresent the application of a differentdefinition of income. This does not represent aproblem as long as the study explicitly outlinesthe reasons behind the choice and as long as allincomes, if compared, are treated in the sameway.

Table 5.6 Inclusion of cash and/or subsistence and forest environmental incomes

Variable

Forest env. income (US$/household/

year) Standard deviation

Minimum

Maximum

N

Total forest


*Prob>F 0.048

Included both cash

and subsistence as

income

757.3

809.7

1.3

3,459.4

45

Cash as total

income

82.4 71.6 2.9 164.7 6

Relative forest


Prob>F 0.15;ns.

Included both cash

and subsistence as

income

0.23

0.17

0.003

0.60

42

Cash as total

income

0.13 0.10 0.01 0.26 6

Table 5.7 Non-inclusion of cash incomes by total forest environmental income



Other important costs may also be omitted,however, including capital costs, intermediateinput costs, marketing and other transactioncosts, and taxes. There are also potential costs ofliving close to forest resources in terms of weedinfestation, wildfire spillovers, health risks, andpests such as predators preying on livestock orcrops that should be included in acomprehensive assessment of net forest benefits.

In our review of cases, we focused on laborcosts and fodder income.

LABOR COSTS

We found that some studies consistentlyomitted or forgot costs involved in production;56 percent of the cases did not include laborproduction costs while estimating the forestenvironmental income, implying anoverestimation of that income and ofdependence (according to the “economic rent”definition of forest environmental income). Thiswas particularly so if labor costs were taken intoaccount for other household productionactivities (see Table 5.8).

We checked if there was a systematic differencein forest environmental income when laborcosts were included but did not find anysignificant difference. The same applies forrelative incomes. In fact, total incomes weresomewhat higher in studies where labor costswere included. This would, first of all, seem toindicate that labor costs, as expected, are

calculated to be very low in the case studiesconsidered. Second, it may just be that thosecases where labor costs have been includedhave also been more diligent in terms ofidentifying and valuing all potential sources ofenvironmental income.

FODDER INCOME

Most of the households reported havinglivestock of different types, but few of thestudies actually valued fodder. This wouldmean a substantial underestimation of forestenvironmental incomes and dependence, butgenerally not of total household income, as thevalue of fodder for domestic use should enter asa cost in livestock production. In studies wherefodder was included, it competed withfuelwood and wild food in terms of economicimportance, so the omission can be a seriousflaw in forest environmental income estimates.The issue becomes even more problematic forassessment of forest environmental dependenceif fodder costs are not deducted from livestockincomes.

A test of its impact across our cases revealed nosignificant differences, even for relativeincomes. The mean was actually somewhatlower for the households that included fodder.It is hard to explain this finding; one possibilityis that more of the studies in drier forests withlower total forest environmental incomes wouldhave included fodder.

Table 5.8 Labor cost inclusion and total forest income in the sample

Variable Total forest envir. income

(US$/household/year) Standard deviation Minimum Maximum N

Labor costs not

included

653.8 737.8 1.3 3459.5 30

Labor costs

included

712.3 879.1 2.9 3186.5 21



BEST PRACTICES:

• At the pilot stage of an investigation, allfactors that may possibly contribute tohigher or lower environmental incomeshould be mapped out and their relativeimportance assessed.

• If variables are intentionally omitted, thisshould be clearly stated and explained.

• In general, studies should aim to include allthe most important goods, which willgenerally involve fuelwood, wild food,fodder, and timber-related goods.

Representativeness and Data Collection

Representativeness

This study has emphasized the importance orinfluence of contextual variables on forestenvironmental incomes. If individual casespurport to represent province-wide, nationwide,or even regional trends, the cases would have tocontrol for a substantial amount of potentialvariation, such as ecological conditions ofclimate, soils, and vegetation as well aseconomic, legal, sociocultural, political, andadministrative conditions of various types. Thiswould clearly be beyond the scope of moststudies looked into here.

Many cases we looked at did not at all take intoconsideration how representative their samplesare. Some cases were too small to berepresentative of a population’s variability.Nyland (1998), for example, had a sample ofonly 20, implying that it would be incorrect tointerpret his findings in relation to even thelocal population of the study area.

Others, however, took into consideration certainissues such as wealth groups, market access,

and type of management of the forest whileselecting the sample. Fisher’s (2002) sampleselection covered three forest managementtypes and the issue of market access byincluding representative villages. The size ofarea and variation in ecological parameter arealso issues that should be realisticallyrepresented in the studies relative to theobjectives of the study.

The impact of contextual variables is analyzedin Chapter 4. Although we documentsurprisingly few statistically significantrelationships, a key message is that it isimportant to be cautious when concluding andmaking policy recommendations based onfindings from a very limited sample ofhouseholds.

Data Collection

Turner and others (2002), in the context ofvaluations, offer the general observation thatthere is a shortage of studies that consider bothbefore and after conditions with respect toenvironmental change, although such studiesare perhaps the most useful and reliable. Notealso that time series are important for observingwhether forest environmental income riskfunctions primarily relate to safety nets (an exante condition) or gap filling (ex post). There isalso a lack of systematic historical studies thatcould have looked into how people de factoused forest environmental incomes to cope inperiods of natural, social, and economichardships. Finally, extended data collectionmight be necessary simply to catch the seasonaland annual variations that often attendextraction of forest resources.

Only three of our studies contained time seriesdata. Thus, meaningful tests of the implicationsof such data were not possible.



Care must be taken in generalizing about how alack of time series studies could bias forestenvironmental incomes. It seems reasonable,however, to assume that more times serieswould improve the quality of our knowledgeconcerning diverse household strategies in aconstantly changing environment. Times serieswould also allow for more in-depth studies ofpeople’s relationship to potential forestenvironmental incomes, and in that sense theycould possibly also lead to the discovery of newand other sources of forest environmentalincomes that are not detected in short-termsurveys and consultancy work.

An “optimal information gathering strategy” is,in any case, impossible to define, as it willdepend on a variety of factors: purpose of study,social and natural given conditions of the area,economic and social resources at hand, and soon.

A general observation, however, is that relianceon a single method of data collection most oftenwill be insufficient. The two most commonapproaches found in the case studies involvedhousehold questionnaire surveys andenumeration of extracted forest goods on theforest perimeter. The first of these methods willoften be found wanting in terms ofidentification of the range of sources of forestincome and the accuracy of amounts extracted;the second will fail to capture environmentalservices and costs, and also animal grazingwithin the forest, and will be unable todistinguish between cash and subsistenceincome.

In addition to the above two methods, literaturereviews, informal talks with key informants,and more advanced valuation methods may berequired. Again, there is no general, universalmethodology that will work for all cases in allcontexts.

Summary on Best Practices

Much of the “secret” of conducting a properanalysis of forest environmental income and itssignificance—in terms of dependence,distribution, and diversification—for localpopulations lies in the apparently simple task ofnot forgetting several important factors:appropriate price estimates, the entire range ofimportant goods supplied, the risk of doublecounting, focusing on flows and not stocks inincome calculations, and so on. Some of theproblems discussed, and their possible positiveor negative effects on total and relative forestenvironmental income, are summarized in Table5.9, and Box 5.1 provides an example of theimpact of the combined effect on studies.

Information is expensive. There will always be atrade-off between the purpose of a study, thefinancial and human resources available, andthe expected utility of increased information. Aconscious practice should include assessing theimportance of information relative to the costsof acquisition.

For the benefit of the wider research and policycommunity, it is important that the measures offorest environmental income used arecomparable to other types of incomes. Therepresentativeness of the study is equallycrucial.

Given this study’s conclusion regarding theimportance of forest environmental incomes forrural households, leaving such incomes out ofnational statistics and poverty assessments willcontribute to the underestimation of totalincomes among rural dwellers. It could alsoaffect estimates of numbers of rural poor if suchfigures are derived from household incomelevels. For people living close to the povertyline, even small income additions are crucial,and the neglect of something in the range of



one-fifth of household income may have widestatistical and political connotations. The WorldBank, other donors, and the countries withwhich they cooperate thus have compelling

Table 5.9 Factors affecting total and relative forest environmental income

Factor to consider

Anticipated impact on total env. income estimates

Anticipated impact on relative env. income estimates

Purposive sampling of high forest value study

sites

+ +

Lack of resources or proficiency - +/-

Use of world or regional market prices + +/-

Neglect of external factors +/- +/-

Inclusion of all local processed goods + +

Neglect of environmental services - -

Double counting + +

Neglect of sustainability issues + +

Failure to include subsistence income - -

Omitted costs + +

Omitted sources of income - -

Box 5.1Waiting for Godoy

The meta-study includes five case studies undertaken by Godoy and collaborators from Latin America. Theaverage annual forest environmental income in these studies was some US$1,922 per year—three times theaverage for our total sample. The average share of environmental income was 41 percent, or almost twice oursample average. The average total income was around US$5,981 per year, more than twice the average for theoverall sample.

Why were the figures so high? The studies were from Nicaragua, Honduras, and Bolivia, countries with gen-erally high GDPs. They were from tropical rain forest areas, with high precipitation rates (2500 mm per year),and from areas where we would expect the production of forest values to be high. The studies, commendably,encompassed both valuation and livelihoods objectives and involved very close field investigations of differ-ent types of use. The average time spent for the studies was 24 months, well above the sample average. Thecases were from areas where local people have good access to environmental resources. They included a largenumber of environmental goods, such as processed timber. Family labor costs were not included. No explicitassessments were made about the sustainability of present use.

We do, of course, not accuse Godoy and others of consciously inflating figures. On the contrary, the studieswere carefully planned and much effort went into avoiding many of the common pitfalls. They also referred tostudies by others that have produced higher overall figures. The point here is that for a variety of reasons—both naturally given and because of theoretical and methodological choices—the figures were high comparedwith the averages in our sample.

reasons to include forest environmental incomein the surveys on which many important policydecisions are based. In addition to providingmore accurate data for policy use, this would



have the added benefit of testing and refiningmethodologies described here.

We recommend the following:• Develop a set of concepts, tools, procedures,

best practice methodology, andrecommendations on the inclusion of forestenvironmental incomes in povertyassessments.

• Agree on an operational definition of forestenvironmental income that makes it

comparable with other sources of householdincomes.

• Concentrate on a few simple, key resourcessuch as food, fuel, and fodder.

• Contribute to a worldwide systematiccollection of base-line data in the field.

• Include important forest environmentalincomes in poverty assessments and indevelopment programs designed to reducepoverty.


Conclusions andRecommendations

How important is environmental income to therural poor in developing countries? While wehave a number of case studies demonstratingthat it can be quite significant for poor peopleand communities, we still lack a good overallpicture. Knowing more about the significance ofenvironmental income is important for at leasttwo reasons. Obviously, a good description andunderstanding of actual livelihood sources forthe poor is needed in order to design effectivepoverty reduction strategies and targetedpolicies and programs. But knowledge about thesignificance of environmental income is alsoimportant to the environmental conservationdebate. If a substantial share of the poor’sincome is derived from the natural environment,there might be less of a tradeoff betweenpoverty alleviation and conservation objectivesthan some people claim.

Most developing countries experiencedegradation of their natural resource bases;forest areas in particular are vulnerable to bothdecimation and degradation, especially throughconversion of forestland to agriculture, but alsothrough unsustainable use of the existingremaining resources. From an economic point ofview, some of this conversion of natural capitalinto other forms of capital could be efficient andcontribute to overall economic development athousehold, community, and national levels.However, documenting environmental incomecan demonstrate some important costs of thisprocess. If we use some notion of an optimal

level of conversion and cut into stocks ofnatural resources, documenting present usesmust enter into the calculus. A particular policyconcern is that the poorest segments tend todepend more on environmental income.

This report has focused both on studies thatlooked into forest environmental incomes andon typical problems related to such studies. Thischapter sums up key findings in this respect. Wefurthermore make some recommendations forfurther research along with some more-operational policy recommendations that wethink can be drawn from the lessons of thisstudy.

Key Findings on Forest EnvironmentalIncome and Poverty

The meta-study synthesized results from 54 casestudies on forest environmental income. Theserepresented a heterogeneous sample, withhighly varying forest environmental incomesand degrees of forest dependence as thecommon denominator. There were, asdocumented, several gaps and potential biasesin the cases and, therefore, in the meta-analysis.The findings thus need to be interpreted withcare. Caveats notwithstanding, several majorconclusions emerged from our meta-analysis:

• Forest environmental income constituted anaverage of about 22 percent of thehousehold income in our sample. Even

6



though agriculture and off-farm income hadhigher income shares, forest environmentalincome represented a significant source ofincome. In absolute terms, the mean annualforest environmental income was aboutUS$678 (PPP-adjusted) per household in thesample, while the median income wasUS$346, representing about 19 percent oftotal income. This indicates a skeweddistribution, with some cases of very highforest incomes (US$3,460 was the highest).Removing the eight studies with forestenvironmental incomes above US$1,500,forest environmental income was stillaround 20 percent of total income, with theaverage household forest environmentalincome of US$401. Thus the broadconclusion from our limited amount ofcases is that forest environmental incomerepresents about one-fifth to the totalincome of rural households.

• Even if encumbered with substantialuncertainties and variations, the figuressuggest that forest environmental incomescontribute significantly to the economicproduction of goods and services and towelfare levels in these societies. Evencontributions that are relatively “small”may be of utmost importance to familiesliving close to the survival line. Omittingsuch incomes from calculations of nationaleconomic statistics and poverty assessmentswill create biases in the base-line data. Ifsuch data are then used in developmentstrategies and programs and inpolicymaking focusing on livelihoods andpoverty, inefficient resource use may occur.

• There is probably a selection bias in thesense that communities with high forestdependence were selected for study inmany of the cases. This points to the need to

include environmental income estimatesinto poverty and livelihoods surveys. Onthe other hand, the studies focused only ona set of environmental benefits, namelythose from the forest. Moreover, the valuesof forest environmental (both forest andnon-forest) services are difficult to quantify,and almost all studies concentrated onproducts only. This lack of valuation ofbenefits created an estimation bias in theother direction. It implies that the figurescited, all other factors constant, represent alower limit for environmental incomes inthe cases.

• Wild food and fuelwood were by far thetwo most important forest products for thehouseholds in the sample, accounting for anaverage of 70 percent of all forest income.We suspect that some products wereunderreported, such as fodder, which wasreported in only seven cases but still had thethird highest value for these cases. Cuttingtrees for timber got surprisingly low figures.

• Forest income was higher in Latin America,while East Africa had the lowest figures.There was a weak tendency of higherincome from wet forests, but the variationacross regions was stronger than thevariation across forest types.

• Forest environmental income tended toincrease with distance to market—that is,more-remote communities had higher forestenvironmental incomes. This probablyreflects both forest abundance and lack ofother income opportunities. Few of theother contextual factors were found to havea statistically significant impact on forestenvironmental income, including tenureand legal status.

• About half of the forest environmentalincome was earned in cash (only about a


Conclusions and Recommendations

third of the cases distinguished betweencash and subsistence income). A surprisingfinding was that the share earned in cashdeclined with higher forest environmentalincome. We also found a negative butweaker relationship between the cash shareand total income. Thus cash forest productsand market access were at least asimportant for communities with low forestand low total income as they were forbetter-off communities.

• Forest environmental incomes wereparticularly important for poor people.Dependence is measured by relative forestincome (percent of total income). Thesample was divided into three groups: low,medium, and high forest dependence, withrelative forest incomes of 5, 19, and 42percent, respectively. The differencebetween these groups was striking. Thehigh forest dependence group had onaverage only half the total income of thetwo other groups. This group also scoredlower on household capital indicators suchas education and livestock ownership, andthey lived in more-remote locations.

• As expected, there was a strong positiveassociation between forest environmentalincome and total income. Forestenvironmental income was important notonly for poor communities. But in terms offorest dependence (income share), theopposite was true: we found a weak but notstatistically significant trend of decliningforest environmental income share as totalincome increased.

• Only about a quarter of the studiesexplicitly addressed the question of thedistribution of forest income within thecommunities studied. Yet the picture that

emerged was clear: the poor were moredependent on forest income, and forestincome had a strong equalizing effect onlocal income distribution. The pro-poorprofile of forest environmental income wasmuch stronger when looking at inter-household differences than inter-community differences.

• In the seven studies that calculated Ginicoefficients (a measure between 0 and 1 ofthe degree of inequality), the coefficient onaverage increased from 0.41 to 0.51 whenforest income was excluded from thecalculations.

• The forest income share for the poor wasabout twice the share for the richhouseholds in the communities, about 32and 17 percent, respectively. Still, theabsolute level of forest income was higherfor the richest households.

• Forest income can be seen as part of ruralhouseholds’ diversification strategies. Wefound that high total income was associatedwith less income diversification, indicatingthat higher income was achieved though aprocess of specializing in one or a few high-return activities. Interestingly, we found abell-shaped relationship betweendiversification and forest environmentalincome. This result, however, was sensitiveto a few outliers in the data.

• Many studies found that forestenvironmental incomes most typically serveas income supplements and importantsafety net in times of hardship. Only rarelydo they provide a pathway out of poverty.The present study cannot provide anyrigorous analysis of strategies, but ourfindings are consistent with thisobservation.



As stated throughout the report, it is difficult todetermine how representative these findings arefor developing countries in general, for African,Latin American or for Asian countries, forcommunities close to or far away from forests,for rich and poor developing countries, for dryand humid climates, and so on. Caution iscalled for in drawing conclusions and using thefindings.

Methodological Experiences and CommonPitfalls

The studies reviewed displayed a high degree oftheoretical and methodological pluralism. Manyof the case studies were also marred bymethodological and interpretive flaws andweaknesses.

The objectives of the studies differed, and theywere classified as valuation studies, forestenvironmental income dependency studies, orstudies focusing on the distribution of forestenvironmental income. The estimates of thisincome varied systematically in these threecategories, with valuation studies estimatingmuch higher values (US$1,405/household/year) than those dealing with dependence(US$525) or distribution (US$152). One reasoncould be that many of the valuation studies hada social benefit perspective, which implies abroader set of forest benefits. A second reasoncould relate to methodology—using contingentvaluation methods , spending more time in thefield, and so on. Sites for the individual casesmay also have been chosen because high forestincome made them an attractive target for suchstudies. A more speculative explanation is thatthe valuation studies may more often havehidden agendas on generating high biodiversityland values in order to generate arguments forincreased conservation. This phenomenon hasnot been investigated further in this study, but it

is mentioned and discussed by other researchers(such as Sheil and Wunder 2002).

Economic concepts of incomes and costs weretreated differently in the studies, and this hasimplications for the estimates. At the mostfundamental level, the studies differed withrespect to the definition of forest environmentalincome: some meant gross income and othersvalue added (gross income minus costs ofcapital depreciation and intermediate inputs),while some authors used economic rent (valueadded minus opportunity costs of labor andcapital). A related problem concerns cut-offpoints with respect to proximity to resource,processing, and market chains. Furthermore, incases with missing or imperfect markets, theapplication of various types of shadow pricesand opportunity cost principles can beproblematic.

Reviewing the studies, we found the followingpossible sources of problems:

• Labor costs for generating forestenvironmental incomes were not includedin almost 60 percent of the studies,indicating that some concept of value addedhad been used. This could create biases inestimation of forest environmental incomeand in environmental dependency, althoughwe did not find any significant differencesbetween those that included labor costs andthe others.

• Some 70 percent of the studies did notinclude processed goods. This will easilycreate a bias in the estimation of forestenvironmental income. The studies thatincluded processed goods as forestenvironmental income had much higherestimates for forest environmental incomethan the other studies.



• Some studies registered only cash income as“income.” Subsistence forest incomes werenot valued at all in six studies. This leads toan underestimation of the economic valueof forest environmental income and incomedependency. Furthermore, a majority (68percent) of the studies did not distinguishbetween cash and subsistence incomes,something that constrains the analysis of therole of forest environmental income in thehousehold and rural economy.

An important weakness in many of the studieswas related to the range of forest environmentalproducts and services that were included—thatis, the completeness:

• Only 26 percent of the studies includedsome discussion of forest environmentalservices, and only one included a valueestimate for these. This means that most ofthe household studies did not deal withsuch services as water retention and soilerosion control—which clearly may havehousehold-level economic impacts but aredifficult to measure and value.

• Around 72 percent of the studies did notinclude fodder as income or as costs inproduction of livestock, even though animalhusbandry is prevalent in most of the areasunder consideration. This will create aserious bias in the estimate of forestenvironmental incomes and forestenvironmental income dependency.

Few studies seriously analyzed distributionaspects in any detail. Only 13 percent estimatedGini-coefficients. Hardly any study focused onforest environmental income relative to varioustypes of household strategies.

In addition to the handling of economic costsand benefits, we also looked at how the studies

considered the issue of sustainable resource use.More than half of them did not considerwhether the current resource use wassustainable. Some studies calculated net presentvalues, simply assuming that present levels ofexploitation could continue unabated for ever.Most of the studies that addressed the issue didnot carry out detailed field-level ecologicalstudies but based their assessments oninterviews with local people and keyinformants. We found that:

• Studies that included sustainability issueshad, on average, twice the forestenvironmental income of studies of otherstudies.

• In studies where people reportedinsufficient resource supplies from theforest, the forest environmental incomeswere on average one-quarter of the incomesreported in the rest of the studies.

Other important conclusions from the review ofmethodologies include:

• Only 5.5 percent of the studies used timeseries data. Thus few studies were able todeal directly with important issues such asthe role of forests in countering incomefluctuations and the impact of policychanges on forest environmental incomes.

• Many studies tended to overrate therepresentativeness of their findings. Localvariations in ecological, economic, social,and political issues are the rule in casestudies—not the exception. Caution shouldthus be shown when research projects areplanned and launched and in makingprojections or generalizations based on thefindings. The research is often based onsmall samples in confined areas. Theselection criteria for study sites may in



addition unfortunately often reflect resourceconstraints and practical problems in thefield rather than representativeness.

Recommendations on Research and BestPractices

As a point of departure, there is a need to “getthe basics right” in individual case studies onforest environmental income and poverty. Thereis a need for more in-depth studies tounderstand the role of forest environmentalincome in individual household and in broaderdevelopment strategies. There is also a need toincrease the scope of such studies throughlarger research projects that get comparabledata from different socioeconomic andecological settings.

“Getting the Basics Right” in Individual CaseStudies

Some of the recommendations at the end ofChapter 5 are simply conventional goodresearch practices. We will not dwell on thesehere. Also, there are guidelines on how to carryout forest valuation, in particular thosepresented in the edited volume by Campbelland Luckert (2002). There are a number ofcommon pitfalls, omissions (and inclusions),unclear definitions, and myriad models andpractical approaches that are partly compatibleand partly not. Many shortcomings found in thestudies could easily have been avoided. Wehave a few specific recommendations:

• Increased cooperation should be promotedin clarifying definitions and concepts,standardizing ways of measuring forestenvironmental income, and using measuresto assess the role of forest environmentalincome for different social groups

concerning dependence, distribution, anddiversification.

• Although information on different types offorest goods and their uses—by whom, forwhat, and with what rewards—is oftenavailable, it is seldom used analytically.Determining the various ways that differentgroups, in different seasons, benefit fromforest goods is important from policyperspectives related to both povertyreduction and environmental conservation.

• The standard measure of importance offorest environmental income has been eitherabsolute income or income share. But a keyrole of forest income is to fill gaps and serveas a safety net, thus helping people surviveand secure their livelihoods during difficultperiods when other sources fail or areunavailable. We need to develop betterways to assess this dynamic role of forestenvironmental income and other functionsof incomes in the livelihood diversificationstrategies of rural households.

• There is a strong need to develop simplefield methods to assess if present uses ofparticular resources are sustainable and totake this into account when estimatingforest environmental incomes.

Ideas for Future Research on ForestEnvironmental Income

There is a clear need for studies that go beyondthe simple measurement of forest environmentalincome and its contribution to total income fordifferent groups.

• We need to understand better why forestenvironmental income is particularlyimportant for the poor. To what extent is the



income contextually and politicallydetermined, and to what extent is it afunction of some inherent characteristics offorests often being an “employment of lastresort”?

• We need to investigate the extent to whichforest environmental income can provide away out of poverty. This requires paneldata, and very few studies to date havebeen able to address the more dynamicaspects of poverty-forest interactions.

• Many studies do not address how legal,ecological, economic, political, andsociocultural factors outside the control ofhouseholds affect ability and willingness tobecome involved in forest environmentalincome strategies, or how these factorsaffect levels and shares of forestenvironmental income. A more completeand rigorous consideration of controlvariables is necessary.

• Very few forest environmental incomestudies deal with issues of localheterogeneity and social differentiation.· A more rigorous focus on differences inuser profiles is necessary.

• Although there is a growing literature onforest environmental income, far fewerstudies focus on other natural resources.More studies on other types ofenvironmental incomes, in both rural andurban settings, are needed.

Policy Implications

Policies to Enhance Forest EnvironmentalValues and Incomes

We found that forest environmental incomeconstitutes an important share of total

household income in the cases studied. In mostnational accounts and even povertyassessments, these values are often omitted orunderreported. We strongly recommendidentifying ways to include environmentalvalues in these. Focusing on the key resources offood, fuel, and fodder would facilitate suchinclusion.

Policies to enhance the environmental income ofthe poor must be pursued along several lines.First, it is important to design and implementpolicies to secure and enhance the resource baseon which these incomes are drawn.

Second, improved systems for poor people’saccess to and control of the resource base areneeded. Forest environmental resources aretypically not privately owned but are undervarying degrees of state or communalownership. At the national level, the planningand management authority of forestenvironmental resources generally rests withthree often competing departments: agriculture,forestry, and nature conservation. Thesedepartments have different mandates andresponsibilities as well as different structures,legal frameworks, and management practicesregarding local people.

At the local level, ownership of forestenvironmental resources is often contested. Thestatus can range from a total lack of access and aban on resource withdrawal in many nationalparks and conservation areas to forest reservesand communal lands with a myriad of formaland informal usufruct rights.

Elements of a policy to enhance forestenvironmental values and incomes could thus:

• Improve formal systems for mapping andregistration of environmental values



• Clarify structures, rights, andresponsibilities between involved publicbodies at national, regional, and local levels

• Establish systems of user agreements withparticipatory monitoring and controlsystems to secure and enhance the naturalresource base.

Policies That Address Dependence,Distribution, and Diversification

Policy interventions should not only concentrateon maximizing aggregate values from theenvironmental resources but should also begiven a poverty focus.

The studies on which this report is based tell usclearly that these incomes are important forpoor people and that deprivation of access andwithdrawal of resources have serious impactson rural livelihoods—both in terms ofconsumption and in cash income-generatingstrategies. It still seems unlikely that incomesfrom the environment in most cases can be theprincipal solution for poverty reduction withinrural development efforts. Forest environmentalincomes primarily serve as a necessarysupplement and as an important safety net intimes of household hardships or when society atlarge goes through general crises (war, drought,economic recessions, and so on). Only in a fewcases do they provide, on their own, thepathway out of poverty.

This does not mean that forest environmentalincome is unimportant on the poverty agenda.But it must be understood within the overalllivelihood strategies of households andcommunities. People prepare for and reducepotential effects of crises in advance through“risk management.” They also have to handlethe effects of crises through risk coping

strategies. Forest environmental incomes formpart of such complex processes.

In many of the studies forest income wasderived from areas under of protection, whichoften meant that part of the income was illegallyacquired The number and size of protectedareas in the world is likely to increase in thefuture following international agreements andconventions (such as the Convention onBiological Diversity). Conflicts over access anduse rights are sure to escalate. In many Africancountries, donors and international forestenvironmental NGOs still promote theconversion of forest reserves that local peoplehave traditionally had access to into nationalparks and other types of legal status wherebiodiversity control is stricter and where localpeople are more often given no access.

Many countries are now developingcommunity-based forest management systems.There is a substantial devolution of powers tolower levels of governance and even to localcommunities to manage the forest resources.Decentralization may, however, easily result inlocal elites capturing the lion’s share of thebenefits.

A policy to address dependence, distribution,and diversification could:

• Secure direct but controlled access andwithdrawal rights to and duties regardingcrucial forest environmental resources forpoor people within local communitiesthrough negotiated resource use agreementswith other actors

• Secure dynamic and flexible agreementsthat can cater for both ecological variationsand variations in social and economicconditions over time



• Develop local capacity and competence toharvest and process products and toincrease local value added

• Improve the functioning of forest and non-forest product markets and reduce the roleof inefficient intermediaries.

Policies to Secure the Biodiversity Base

The old “fortress approach” to conservation,including the alienation of local people fromnatural resources, has not worked and is nolonger politically acceptable. Policies must havea broader focus than just conservation. Themore recent social forestry approaches, jointforest management, and other participatoryapproaches also have problems, however:biodiversity is often less than adequatelyprotected, and local benefits, in particular forthe poor, are less than expected (see, forexample, Hulme and Murphree 2001, Barrowand Murphree 2001, Vedeld 2002).

Conserving biodiversity resources will in manycases be difficult to reconcile with localparticipation. We need a better approach thatcombines the ambitions of biodiversityconservation with rural development, valuegeneration, and distribution concerns. There areno blueprint models for this.

A revised policy could:

• Develop simple operational field methodsto assess if current uses of particularresources are sustainable

• Secure the biodiversity resource basethrough zoning and multiple usearrangements, using local institutions andknowledge

• Develop transparent and reciprocal systemsto secure congruence between provision andappropriation of natural resources

• Secure harmonization of external publicbodies of management, monitoring, andcontrol.

Concluding Remarks

The forest resources from which environmentalincomes are drawn are under tremendouspressure in most of the developing world.Converting forest to agricultural land can beeconomically sensible and may in many cases beseen as part of a reasonable developmentprocess. But with the growing scarcity of forestresources, remaining forestlands increase invalue, and the additional land converted toagriculture is typically less productive foragricultural purposes. The conversion processeswould nevertheless typically continue, partlybecause they often entail a transfer ofownership from state or communal to privateproperty and partly because many of thebenefits from standing forests are local or globalpublic goods in contrast to private agriculturalproducts.

In this larger picture of forest incomes anddevelopment, we need a better database onforest environmental values and incomes.Reaching some consensus on best researchpractices is a common goal for both researchersand policymakers. Heterogeneity of studyobjectives and methodologies and thesubstantial variations in results unfortunatelyyield a field riddled with uncertainties. In asituation where political decisions increasinglyare based on—or legitimized by—researchfindings, research works increasingly becomeinvolved in politics.



The assessment of the economic importance offorest resource use involves individuals withconflicting interests, such as conservationists,timber traders, medicine plant merchants, andagriculturalists. This has strong politicalimplications—a fact well acknowledged byresearchers themselves, by bureaucrats, and by

politicians. Different stakeholders will findsupporters within various parts of the researchcommunity. In this context we believe thattransparency, integrated and unified researchmethods, and increased awareness about thestrategic dimensions of research are important.


Appendix A —Measures and Tests Relevant toForest Environmental Income

Basic Measures of Forest EnvironmentalIncome and Dependence

Several basic measures are relevant to theanalyses of forest environmental income and itslink to poverty. The first and most obvious,using simple syntax, is

AI = absolute total income

which would involve a household’s income, inboth subsistence (direct consumption) and cashform, from all available sources. A furthermeasure is

ACI = absolute cash income

which would involve cash income from allavailable sources. A final measure related toabsolute income from all sources would then be

ASI = absolute subsistence income

where AI = ACI + ASI. The two latter can also bemeasured in terms of their relative rather thanabsolute contribution to total income. That is,

RCI = relative cash income = ACI/AI

and

RSI = relative subsistence income = ASI/AI

Each of the above has a counterpart for forestenvironmental income. The simplest is

AFI = absolute forest environmental income

which is forest environmental income, in bothconsumption and cash form, from allenvironmental income sources and products.Furthermore,

ACFI = absolute cash forest environmental in-come

which is the cash component of AFI and

ASFI = absolute subsistence forest environmen-tal income

which is the consumption component of AFI.Again, these two latter measures can beexpressed in terms of their relative rather thanabsolute contribution to absolute forestenvironmental income, so that

FICS = forest environmental income cash share= ACFI/AFI

and

FISS = forest environmental income subsistenceshare = ASFI/AFI

Just as important as the two latter, however, aremeasures relating environmental income toincome from all sources. First,

RFI = relative forest environmental income =AFI/AI

which measures absolute forest environmentalincome as a share of absolute income from allsources and is thus a simple but important



measure of reliance, or dependence, on forestenvironmental income. The equivalent for cashincome is

RCFI = relative cash forest environmental in-come = ACFI/ACI

which measures cash forest environmentalincome as a share of cash income from allsources, while

RSFI = relative subsistence forest environmen-tal income = ASFI/ASI

which measures subsistence forestenvironmental income as a share of subsistenceincome from all sources.

Simple Tests on Role of ForestEnvironmental Income in DifferentHouseholds

If the above measures are available, a number ofsimple tests and procedures can be carried outthat shed light on the role of environmentalincome in poverty alleviation. In the following,unless otherwise indicated, it is assumed thattests are carried out with households as the unitof observation.

Forest environmental income will, of course,depend on numerous endogenous andexogenous variables, as discussed in Chapter 2.In general, therefore, all available explanatoryvariables of possible relevance should beincluded in the regression equations. Forsimplicity, however, we here specify the keyrelationship only and leave out the controlvariables. These relationships generally wouldbe non-linear, which should be considered in themodel specification.

1. Research question: (How) does total incomelevel influence forest environmental incomelevel?

This is the simplest test, asking basicallywhether total income contributes to higherforest environmental incomes or not:

AFI = f(AI) (1)

There is, however, a potential endogeneityproblem in this model, since forestenvironmental income will tend to contribute tohigher overall income. In many cases, in fact,the reverse specification may be the relevantone.

2. Research question: Does total income levelinfluence the dependence on forestenvironmental income?

This question looks at how dependence varieswith overall income. Using RFI as a proxy fordependence, the model appears as

RFI = f(AI) (2)

This model may also directly relate to questionsregarding forest environmental Kuznets curves,discussed later.

3. Research question: Does cash/consumptionrealized from forest environmental incomeincrease with absolute forest environmentalincome?

These tests simply look at whether forestenvironmental cash income or forestenvironmental subsistence income increase withtotal forest environmental income. The test mayreveal whether increases in forestenvironmental income primarily contribute tocash income or consumption, with implicationsfor the role of forest environmental income as asafety net, as support for consumption, or as apathway out of poverty.

FICS = f(AFI) (3a)

FISS = f(AFI) (3b)


Measures and Tests Relevant to Forest Environmental Income

Note, however, that both of these tests maysuffer from the same endogeneity problem asthat which potentially afflicts (1).

4. Research question: Do “poor people” dependmore on the environment as a source of cashthan the wealthy do?

The following specifications are relevant:

ACFI = f(AI) (4a)

RCFI = f(AI) (4b)

The first of these looks at absolute cash forestenvironmental income as a function of absolutetotal income, while in the second the dependentvariable is cash forest environmental income asa share of cash income from all sources.

In specifications where absolute total incomeenters as an independent variable and whereproblems of endogeneity crop up, a wealthindicator—based on households’ aggregatepossession of productive and exchangableassets—may usefully be substituted. Wealthindicators have the advantage over absoluteincome of being a more reliable proxy preciselyfor wealth, which is usually the more interestingexplanatory variable, and are at the same timemuch less sensitive to the endogeneityproblems.

Each of these questions, and the attendantspecifications, have dealt with forestenvironmental income in general. It may,however, be useful to consider groups ofenvironmental goods—or even individualgoods—separately. We might, for example, wishto look at dependence on timber, NTFPs, andenvironmental services—and at how these varywith wealth—separately, since such an analysiswill reveal more accurately the nature ofdependence and the sustainability of

environmental use for different social groups.Or, following up on the discussion in Chapter 2,we could test the hypothesis that certain low-return forest activities are an “employment oflast resort” for the poor, while the rich are ableto monopolize high-return activities.

Thus, for example, model (2) could be changedinto:

RFIi = f(AI) (5)

where i denotes the specific good or group ofgoods under consideration.

A “Forest Environmental Income KuznetsCurve”?

The manner in which forest environmentalincome, and dependence on it, varies withincome or wealth may be neither strictlyincreasing or decreasing, and more flexiblemodels may be necessary. One example of thiscould be a slightly reinterpreted environmentalKuznets Curve, where instead of looking atpollution or resource degradation we look atforest environmental income levels anddependence and postulate U-shaped relationsbetween these and total income.

The logic underlying the forest environmentalincome Kuznets curve (FIKC) is that there issome trajectory of environmental degradationas the income of societies increases, which couldthen also be applied to environmental income. Ifthe unit of analysis is households, it couldsimilarly be argued that there is an FIKC-liketrajectory that households pass through asincome increases (cf. Bulte and van Soest 2001).

The following model could be useful to test thisrelationship:

AFI = a(AI) + b(AI)2 + u (6)



which can be transformed into a linearrelationship by dividing both sides of theequation with AI:

AFI/AI = a + b(AI) + u/AI (7)

which in turn equals:

RFI = a + b(AI) + u/AI (8)

In investigations of dependence, RFI may besubstituted for AFI in equations (6) and (7).

Note that (8), apart from the error term, isequivalent to (2). Ideally, given the weightederror term, the estimators should here betransformed through the standard generalizedleast squares procedure.

The assumption here is that degradation isproportional to use or extraction, which in turnis proportional to forest environmental income.While the latter may seem quite unproblematic,the former assumption neglects sustainabilitythresholds.

The model can (and should) also be used toexamine how income from and dependence onspecific environmental goods or groups ofgoods vary with income and wealth levelswithin communities:

AFIi = a(AI) + b(AI)2 + u (9)

where AFIi is income from environmental good i(or group i). And, again, a wealth indicator mayprofitably be substituted for absolute totalincome.

Income Inequality and Distribution

Many of the regression models suggested,including the FIKC, aim at finding out howenvironmental dependence differs amongincome groups. This sub-section suggests othercomplementary measures of income inequality

and how they can be used in the analysis ofenvironmental dependence.

Gini Coefficients

The Gini coefficient can be calculated as

µ2

1 1

2n

AIAI

G

n

i

n

jji

AI

∑∑= =

−=

where n is the sample size and m is the sampleaverage. So the Gini coefficient for incomeinequality is simply the relative mean differencebetween all possible income pairs i and j in thesample. For small samples, the expressionshould be multiplied with n/(n – 1) to providean unbiased estimator.

If we construct a new variable for “absolutenon-environmental income”—that is, absoluteincome from all sources other than theenvironment—such that

ANI = AI – AFI

then we can calculate a Gini coefficient forabsolute incomes excluding forestenvironmental income:

µ2

1 1

2n

ANIANI

G

n

i

n

jji

ANI

∑∑= =

−=

A comparison of these two Gini coefficients willreveal whether, and to what extent, forestenvironmental incomes contribute to reducinginequality. For example, Aryal (2002) finds thatthe Gini coefficient in his study area ofBudongo, Uganda, increases from an alreadyrelatively high level of 0.55 to 0.61 when forestincome is excluded.



Kuznets Ratios

Comparisons of forest environmental incomeand dependence between the rich and the poorare of interest in general, regardless of thespecific nature of the relationship. The Kuznetsratio (not to be confused with Kuznets curves) isthe ratio between the average income of therichest x percent with the poorest y percent.(‘How many times richer are the richest xpercent compared with the poorest y percent?’)We extend the definitions of the Kuznets ratioby distinguishing between the absolute andrelative ones, and we also apply them to forestenvironmental income.

In the following, values for the richest andpoorest 20 percent will be used, since separationinto quintiles is common, but other percentagesmay be used (typically some have used thebottom 40 percent).

The “absolute Kuznets ratio,” in terms of totalincome, can be defined as:

%)20(

%)20(

⋅⋅⋅⋅=

poorestAImean

wealthiestAImeanAKR

The corresponding “absolute environmentalKuznets ratio” can then be defined as:

%)20(

%)20(

⋅⋅⋅⋅=

poorestAFImean

wealthiestAFImeanAEKR

Relative Kuznets ratios specify the share ofmean absolute income attributable to differentsources. The “relative environmental Kuznetsratio” can thus be defined as:

%)20(

%)20(

⋅⋅⋅⋅=

poorestRFImean

wealthiestRFImeanREKR

While AEKR measures the ratio of absoluteforest environmental income among the wealthyto that of the poor, the REKR expresses the ratioof dependence among the wealthy to that of thepoor. Thus it is entirely plausible, as Cavendish(2003) hypothesizes, that while the wealthy willhave absolute forest environmental incomesexceeding those of the poor (AEKR > 1), thepoor will be more dependent on forestenvironmental incomes (REKR < 1). This wasindeed observed by Aryal (2002) and Botha(2003) in Uganda and Malawi, respectively.

Kuznets ratios can also be used as explanatoryvariables when analyzing forest environmentalincome in meta-studies. Relevant tests could, forexample, be:

mean AFI = f(AKR) (10)

or

mean RFI = f(AKR) (11)

These would, respectively, analyze absoluteforest environmental income and dependenceon forest environmental income as a function oftotal income inequality.

If data permit, meta-studies might also usefullyemploy pairwise t-tests or ANOVAs to analyzeabsolute forest environmental income andincome dependency as a function of incomegroup (such as wealthiest and poorest 20percents) across cases.

More generally, the Kuznets ratios can also beapplied to income from and dependence onspecific environmental goods or groups ofgoods, for example:

%)20(

%)20(

⋅⋅⋅⋅

=poorestAFImean

wealthiestAFImeanAEKR

i

ii



where AFIi is forest environmental income fromgood i (or group i), or

%)20(

%)20(

⋅⋅⋅⋅

=poorestRFImean

wealthiestRFImeanREKR

i

ii

where RFIi is forest environmental income fromgood i (or group i).

Diversification

Diversification of income sources is central toanalysis of the role of forest environmentalincome in rural livelihoods. The simplestindicator would be a digit indicating thenumber of different income sources in ahousehold.

Two relevant diversification indexes, based onthe same logic as the Simpson Index for fieldfragmentation, would be:

Diversification index, total income =

DITI = 1 – å(Ii/AI)2

where Ii is income (consumption + cash) fromsource i.

Diversification index, cash income =

DICI = 1 – å(CIi/ACI)2

where CIi is cash income from source i.

These indexes, like the simple digits, may sufferfrom problems related to the specific manner inwhich incomes are classified. For example,should forest environmental income constituteone category, or should it be divided intoincome from timber, NTFPs, and environmentalservices? The specific purpose of the analysis,and experimentation, may reveal the mostexpedient classification. But the comparison of

cases presumes that a similar level ofaggregation has been used in defining theincome categories.

Diversification indexes can be used to test anumber of hypotheses, such as:

Total income increases with total incomediversification:

AI = f(DITI) (12)

Cash income increases with cash incomediversification:

ACI = f(DICI) (13)

Forest environmental income is important forincome diversification:

DITI = f(AFI) (14)

Forest environmental cash income is importantfor cash income diversification:

DICI = f(ACFI) (15)

The implied causality in the above tests can, ofcourse, be discussed. Tests incorporatingexogenous variables are, again, recommended.

Forest Environmental Income Fluctuations

Fluctuations in forest environmental income areof general interest, since they may reveal theextent to which the resource functions as a“bank account”—to be tapped only when othersources of income decline or fail—or representsa stable source of income for consumption orsale, which in turn may illuminate questionsrelated to the types of risk (idiosyncratic orcovariant) faced by households, the copingstrategies they use, and the role of forestenvironmental income in poverty reduction oravoidance.



To investigate this, time series are necessary,and these are unfortunately but understandablyrare. When time series data—and in particularpanel data—are available, however, variancesand standard deviations in forest environmentalincome (also from different goods, and as cashor consumption) for each household over timemay be computed. These, in turn, can becompared across different income, wealth, orsocial groupings.

Panel data may also permit proceduresexplaining forest environmental incomefluctuations, for example:

AFIt+1 – AFIt = f(ANIt+1

– ANIt, DITIt+1– DITIt) (16)

where t is time period (1…..n). This wouldattempt to explain fluctuations in absolute forestenvironmental income as a function offluctuations in non-forest environmental incomeand income diversification.


�

��

� ��

��

��

��

!��"��#�&��

��'��

��*�� :�<�"� '�?&��:��&�:@!�

��J��KKQ�� @�&�"�J��<"�X��J��&�<��

'��Y��"�Z��

��&��[��

?�� "��\��&��

��[��]��Z��

�KK^*�KKQ� ]��Z�� '�?&��:��:@!�

��"�_�� ]��J��`��?��

_��Y��

��*�� Y�� '�?&��{�<��

|&��&��?�&��"�?&��&��

:@!�

��"��KK}�� ]��"��@�[��&��?��""��"��

!��"��|&��'�~��

��<�@�[��&��?��

|��<��

�KK�*�K}K� |��<�� '�?&��:��&�:@!�

_�"�&��

��KK^��

|~��&��:��]��

J��&��#�&"��

��"��'��"��#��

�KK�� #�� Q��^��}*

��Q�

_��

��KK��

��J�"�&�:��]��

J��&��"�|&��

�� ?��@��

{�<��&�"��"�|&��&��&�"��

��&�"��Y��[��

J��?��@��

�KK�*�KK�� ?��@�� <��

|�[��"�|&��&��

J�"�&�?��|&��

|�[��J��

?��|��?��<��

:�[��KK��

_��

��KK��

��J�"�&�:��]��

J��&��"�|&��

��'��?��@��

�KK�*�KK�� ?��@�� <��

|�[��"�|&��&��

J�"�&�?��|&��

|�[��J��

?��|��?��<��

:�[��KK��

_��

��KK�&��

��J�"�&�:��]��

J��&��"�|&��

��<��?��

@��

�KK�*�KK�� ?��@�� <��

|�[��"�|&��&��

J�"�&�?��|&��

|�[��J��

?��|��?��<��

:�[��KK��

Appendix B —Case Studies



��

� ��

_��:��_��^�� @��J��J�[��]��

?��"��[��

'�"��

�� '�"�� '�?&��{�<��

|&��&��?�&��"�?&��&��

:@!�

��<��""�_��'��

�"��'��@�&��

'��'��

'��[��#��

?&��KKK��

@�&�"�@�[�"�|&��&��"��

?�[��"��&��""��

��""��

�KK�� !��?<��<��

�� "��^�:��

#��"�#��

|�[��{�[�"�<��

��<��""�_�'��

�"��'��@�&��

'��'��

'��[��#��

?&��KKK��

@�&�"�@�[�"�|&��&��"��

?�[��"��&��""��

��'��Z��""��

��

�KK�� !��?<��<��

�� "��^�:��

#��"�#��

|�[��{�[�"�<��

��<��""�_��'��?��

��

�Z��'��

@�&��'��

'��

��

�� !� �� "�� # ��&�'��

�KKK*�� _��<��"��#��

��#��

��<��""�_��'��

'��@�&��#��

?&��KK��

@�&�"�@�[�"��"��?�[��

��&��?��

��

�KK^� �� K*��

��<��""�_��'��

'��@�&��#��

?&��KK��

@�&�"�@�[�"��"��?�[��

��&��?��'��"��

��

�KK^� �� K*��

��[��

��KKK��

|�<��&�"��"��J�[��

|�[��"��<��&��

��"�!��"��KK^*�KKQ�

�KK^*�KKQ� �� <�<��?��KK��

#�<��"��""��@��

��[��

��KKK��

|�<��&�"��"��J�[��

|�[��"��<��&��

��"�!��"��KK�*�KK��

�KK�*�KK�� <�<��?��KK��

#�<��"��""��@��

��"�[��]��

_��@��

_��

��]��

��

��J��J��<"��{�<��

_��[��|�[��

Y��"��&��Z��

J��J�&��?��:��"�

��[��J��

�KK�� J�� {��<�<��

:|Y{��&��

�


Case Studies

�

�

��

� ��

{�[��{��

?��&�"��|��

��

��"��:��]��J��&��

#��&��#��<"��_��

J�[��@�[�"��

|�[��?��&��

�� ?��

��&��

?�<��J�<��<��

��|��&��

��

��

|��'��

!�"��

��K}K��

?��"��:��:��&�?��"��

��?�&��"��|&��&��?��

��"�]��@��]��

��

�K}K� �� '�?&��:��&�:@!�

|��'��

!�"��

��K}K��

?��"��:��:��&�?��"��

��?�&��"��|&��&��?��

��"�]��@��]��

��

�K}K� �� '�?&��:��&�:@!�

��'�� !��"��"��

{�<��&��"�'�"��

�KKK*�� '�"�� Y�<��"��<�<��

��:��

_��{��

��"��KK��

]��|��&��#�&��|~��&��

��:��]��]��<�&�"��

J��&��'��"�!<��

J��"��?��

#��:�&��

�KK�� :�&�� #��^��

��!��

�[��

{��@��<�Z��

|��_��]��

!��&��

@��|��J��Z�

��&��

��Z�

��

@�&�"��&��"�_��

�KKK*��

�KKK*�� _�"�[�� # �� Q��^K�*

Q�K�

��!��

�[��

{��@��<�Z��

|��_��]��

!��&��

@��|��J��Z�

��&��

��Z�

��

@�&�"��&��"�_��

�KKK*��

�KKK*�� _�"�[�� # �� Q��^K�*

Q�K�



�

�

��

� ��

��!��

�[��

{��@��<�Z��

|��_��]��

!��&��

@��|��J��Z�

��&��

��Z�

��&��

@�&�"��&��"�_��

�KK�*�KK��

�KK�*�KK�� !�� # �� Q��^K�*

Q�K�

��!��

�[��

{��@��<�Z��

|��_��]��

!��&��

@��|��J��Z�

��&��

��Z�

��

@�&�"��&��"�_��

�KK�*�KK��

�KK�*�KK�� !�� # �� Q��^K�*

Q�K�

��"��{��

��^��

��@��"�?��'��|"��

��[��Y��#�<�&��

@�&�"�J��<"�X��@�[�"��

��^� Y�� '�?&��:��&�:@!�

��"��!��'��

{��?��

J��

��

��KK^��

��"��:��]��J��&��

��|&��J��<��"�

�� &�"��:��"�

��"��?��@��

��?��<<��&��

�KK^� ?��@�� Q��^��*

�}��

!��?��

?��<��@��

�&�� ?��

_��KK��

|~��&��:��]��

J��&��#��]��_�"��

��!�""��_�""��

]��<"��"�"��_�]��|~��_"�&��

#��]��"�#�&��

�KK^� #�� ^��Q^*

��

�

�

!��?��

?��<��@��

�&�� ?��

_��KK��

|~��&��:��]��

J��&��#��]��_�"��

��!�""��_�""��

]��<"��"�"��_�]��*�#��_"�&��

#��]��"�#�&��

�KK^� #�� ^��Q^*

��


Case Studies

�

�

��

� ��

��#��KKK�� ]��:��]��

��J��&��!��"��#�&��

��?��'��{��&��

�KKK� Y�� _?&��&��&�"��

��'��

Y��[��

'��

��?��&��?��

�� <��Y��

��&��<�{��&��:�<�"��

��*�� :�<�"� '�?&��:��&�:@!�

'��

��KK��|&��&��"��:��"�J��

|��"��Z��

��_��Z��

]��Z��

�KK�*�KK�� ]��Z�� '�?&��:��&�:@!�

'��'��

��KK}��

��&��{�<��[��?�&��

|&��&��J��"�

!��"��]��

#��J��"��'��"��

��""��|��<��

�KK}*�KKK� |��<�� '�?&��:��&�:@!�

:��?��KKQ�� ?�&��|&��&��|&�"��&�"�?��

'��[��'��

��Z��]��Z��

�KKQ� ]��Z�� '�?&��:��&�:@!�

:��:��

��KK}��

?�&��"�|&��&��J�"��&�"��&��

��?��"��&��Y��

��'��

�KK�*�KK}� Y�� '�?&��:��&�:@!�

:"��]��KK}�� @�&�"�J��<"��Y��@��:��"�

��&��

�

�KK�*�KK}� @��J{�� &��{�<��

|&��&��?�&��"�?&��&��

:@!�

?��{��KK^��J��<"��?��

��"�_��:��"�J��:�<�"�

�KK�*�KK^� :�<�"� '�?&��:��&�:@!�

?�"��!��K}K��|&�"��&�"��?�&��|&��&�"�

��#�&��?��

J��&��[��?��""�?&�"��

'�&��Z��

{��&��?��]��Z�Z��

��

�K}K� ?�� '�?&��:��&�:@!�



�

�

��

� ��

?��Y��

:��?��

��

��'��

:��[��

��{��

��\��

��&��

��

��Z��"�:��J��"��

��_��

��<<��

��^��

@�[�"��|&�"��&�"��

��:]�J�{�<��&��

��"��{�<��&��|&�"��&�"�

��"��'��?��&��

]��!��

|&�"��&�"�?��?��#��

�_�"��?��]��<"��"�"��

?��&��

�� #�� #��"��&��

��"�@�[�"��

_��[��_��

'��*�^�

?��Y��

:��?��

��

��'��

:��[��

��{��

��\��

��&��

��

��Z��"�:��J��"��

��_��

��<<��

��^��

@�[�"��|&�"��&�"��

��:]�J�{�<��&��

��"��{�<��&��|&�"��&�"�

��"��'��?��&��

]��!��

|&�"��&�"�?��?��#��'�"��

'��!�""��

�� #�� #��"��&��

��"�@�[�"��

_��[��_��

'��*�^�

?��Y��

:��?��

��

��'��

:��[��

��{��

��\��

��&��

��

��Z��"�:��J��"��

��_��

��<<��

��^&��

@�[�"��|&�"��&�"��

��:]�J�{�<��&��

��"��{�<��&��|&�"��&�"�

��"��'��?��&��

]��!��

|&�"��&�"�?��?��#��\�[�

��:��"�J��

�� #�� #��"��&��

��"�@�[�"��

_��[��_��

'��*�^�


Case Studies

�

�

��

� ��

?��&�"��?��

?��&��"��]��

:��"�[��

_��&��

_�""��&��KKK��

!��"��"��"��

?��"��"�@�[�"��`�@�&�"�

@�[�"��"��"��&��

��!��?��&��

�KKK� ?��

��&��

J��]��

?��"��'��_��

��?&��&�Y��

��"��

J��?��&��

?��&�"��?��

?��&�"��]��

:��"�[��

_��&��

_�""��&��KKK��

!��"��"��"��

?��"��"�@�[�"��`�@�&�"�

@�[�"��"��"��&��

�� \��?��&��

�KKK� ?��

��&��

J��]��

?��"��'��_��

��?&��&�Y��

��"��

J��?��&��

?��&�"��?��

?��&��"��]��

:��"�[��

_��&��

_�""��&��KKK&��

!��"��"��"��

?��"��"�@�[�"��`�@�&�"�

@�[�"��"��"��&��

��'��?��&��

�KKK� ?��

��&��

J��]��

?��"��'��_��

��?&��&�Y��

��"��

J��?��&��

?��:��]��

��KKK��

��#��#�<�&��

?�&��|&��&�?��J��<"��

:�<�"�

�KKK� :�<�"� '�?&��:��&�:@!�

?��]��

��&�"��:��&�"��"��&��[��

��#��"��]��

!��"��?�&��

�KKK*�� '�?&��:��&�:@!�

?��{�� ]��#�<�&��J��&�<��

'��J��<"��@�[�"��

�<��'<��{��&��"�

��

��*�� '�?&��:��&�:@!�

?\��|��J��

��"��

'��

:��?��

��

��'��"��^��

|&��&��"��&��

��{��&��]��Z��

�� ]��Z�� :��&��<��':�]�

]��Z��

?\��|��J��

��"��

'��

:��?��

��

��'��"��^��

��"��&��

�"��\��{��&��]��Z��

�� ]��Z�� :��&��<��':�]�

]��Z��



��

� ��

?\��|��J��

��"��

'��

:��?��

��

��'��"��^��

��"��&��

]��{��&��]��Z��

�� ]��Z�� :��&��<��':�]�

]��Z��

?��{��^�� '��_��]��"��"�

��J��Y��@�[�"��<��

?��:��]��

J��&��_�"�[��Z��

�KK}� _�"�[�� #��"��&��

��"�@�[�"��

_��[��_��

'��*�^�

��J��

��KKK��

|[�"��?��"��

{��@��Y��?��Y��

�[��_��#��&��

�KK�� ?��@�� '�?&��:��&�:@!�

��:��_��

��J��

��J��KK��

�

��{�<��?��[�[�"�?��

]��"��#�<"�&��

��'��J��

#��

�KK�*�KK�� #�� _��<��"��

_��]��"��

�

�


Appendix C —Variables in the Meta-Analysis

Background to the Case Study

1. Year of data

2. Purpose of original study (academic research,rural development, conservation report, orconsulting report)

3. Duration of the study in months

4. Objective of the study

5. Sample size in the study

6. Methodology used (household survey, othermethod or both)

7. Locality of the study area – Province

8. District

9. Study site

10. Country

Geographic Characteristics

11. Size of the study area (km2)

12. Number of people living in the study area

13. Population density (persons per km2)

14. Human population growth trend - Has thehuman population in the study area increased,remained stable, or decreased over the past 10years (including changes due to migrations)?

15. Mean elevation of the study area (m)

16. Average precipitation of the study area(mm/year)

17. Soil type

18. Forest type

19. Major land uses in the study area inpercentages by the following categories:

20. Rainfed crop production (%)

21. Irrigated crop production (%)

22. Non-cultivated land (%)

23. Grassland (%)

24. Pasture (savannahs/woodlands that areused for grazing) (%)

25. Swidden fallow (%)

26. Forest (%)

27. Settled area (%)

28. Marshy/swampy area (%)

29. Other (%)

Characteristics of Environmental Products

30. Use of the environmental products:



31. Wild animal (own consumption, sale, orboth)

32. Wild fruits (own consumption, sale, or both)

33. Wild vegetables (own consumption, sale, orboth)

34. Wild insects (own consumption, sale, orboth)

35. Other wild foods (mushroom, birds, mice,honey, spices) (own consumption, sale, or both)

36. Wild medicine (own consumption, sale, orboth)

37. Timber (own use, sale, or both)

38. Timber (used as a production input)

39. Firewood (own use, sale, or both)

40. Thatching grass (own use, sale, or both)

41. Woven goods (sleeping mats, baskets,brooms, hats) (own use, sale, or both)

42. Leaf litter (own use, sale, or both)

43. Pottery clays (own use, sale, or both)

44. Termite mounds (own use, sale, or both)

45. Livestock fodder (own use, sale, or both)

46. Ornamental/aesthetic/fashion (own use,sale, or both)

47. Tourism (own use, sale, or both)

48. Charcoal (own use, sale, or both)

49. Resin (own use, sale, or both)

50. Rattan (own use, sale, or both)

51. Fish (own use, sale, or both)

52. Gold (own use, sale, or both)

Characteristics of the Environmental Area

53. Type of land tenure system of the resourcearea:

54. Private land

55. State land

56. Communal land

57. Open access

58. Type of management of the environmentalresource:

59. Strict nature reserve

60. National park

61. Habitat management area

62. Protected landscape

63. Managed resource protected area

Socioeconomic Characteristics

64. Socioeconomic characteristics:

65. Average household size

66. Average number of people involved inenvironmental production per producerhousehold

67. Average age of the household head

68. Average number of years in school

69. percentage of households with no education


Variables in the Meta-Analysis

70. percentage of households with primaryeducation

71. percentage of households with secondaryeducation

72. percentage of households with post-secondary education

73. Average annual household income (US$)

74. Average annual cash income (US$)

75. Share of cash income (%)

76. Income share according to sectors

77. Crop production

78. Livestock

79. Wage labor

80. Off-farm activities

81. Remittances

82. Forest activities

83. Fishing

84. Other

85. percentage income that is earned in cashaccording to sources as follows:

86. Crop income (%)

87. Livestock income (%)

88. Off-farm (%)

89. Forests (%)

90. Average annual forest environmental income(US$)

91. Average forest environmental income bysources:

92. Average timber income (US$)

93. Average fodder income (US$)


95. Average firewood income (US$)

96. Average gold panning income (US$)

97. Average wild food income (US$)

98. Average wild medicine income (US$)

99. Average thatching grass income (US$)

100. Other income (US$)

101. Average annual cash forest environmentalincome (US$):

102. Average timber income (US$)



105. Average firewood income (US$)

106. Average gold panning income (US$)

107. Average wild food income (US$)

108. Average wild medicine income (US$)

109. Average thatching grass income (US$)

110. Other income (US$)

111. Relative forest environmental income

112. Relative cash forest environmental income



113. Does the study indicate costs incurred inproduction?

114. Costs incurred in crop production (labor,seeds, fertilizers, draft…)

115. Costs incurred in livestock production(labor, fodder, cattle purchases...)

116. Costs incurred in forest production (labor,machinery, transport.)

117. Domestic expenses

118. Other expenses

119. Total expenses

120. Net income (US$)

121. Has fodder been valued?

122. Gini coefficient without forest income

123. Gini coefficient with forest income

124. Degree of forest dependency (percentage oftotal households dependent on forests):

125. Not dependent ( percentage of totalhouseholds with 0% share of forestenvironmental income)

126. Weakly dependent ( percentage of totalhouseholds with <=20% share of forestenvironmental income)

127. Dependent ( percentage of total householdswith between 20 and 40% of forestenvironmental income)

128. Strongly dependent ( percentage of totalhouseholds with more than 40% of forestenvironmental income)

129. percentage of total households involved atall in forest environmental income activities

130. Local labor rate - Average daily wage forlabor (US$)

131. Proportion (%) of households involved inforest environmental production

132. Proportion (%) of households involved inprocessing of environmental goods

133. Proportion (%) of households involved inmarketing of environmental goods

134. Average household income of producerhouseholds per year (US$)

135. Proportion of men having forest income (%)

136. Proportion of women having forest income(%)

137. Average land holding (ha)

138. Average livestock holding (TLU)

139. Classification of household according towealth:

140. Very rich (%)

141. Rich (%)

142. Average (%)

143. Poor (%)

144. Very poor (%)

145. Degree of forest dependency according todifferent income groups (role of forest productsin total household production)

146. Cash forest share according to differentwealth groups


Variables in the Meta-Analysis

147. Livelihood strategies adopted, ranking:

148. Crop production

149. Livestock

150. Wage labor

151. Off farm activities

152. Remittance

153. Industrial worker

154. Forest activities

155. Transfers

156. Fishing

157. Other

158. What is the percentage of households withformal employment?

Characteristics of the Market

159. Are there any organizations securingtransport, marketing, and sales of produce:private, cooperative, state, and projects?

160. Age of organization

161. Is the organizational structure efficient?

Institutional Characteristics

162. Customary rules governing forest/productuse:

163. Does the community respect customarylaws

164. Are the rules effectively enforced? (Yes orNo)

165. Are there government regulations affectingresource exploitation?

166. If yes, what is the effect of regulations onexploitation of the resource? (positive wouldmean resource exploitation is tending towardsustainable)

167. What is the effect of governmentregulations on promoting equitable access to theresource? ((positive means equitability ispromoted)

168. Is there government investment to support,encourage, or develop environmentalproduction?

169. If yes, what is the effect of investment onexploitation of the resource? (positive wouldmean resource exploitation is tending towardsustainable)

170. What is the effect of investment onpromoting equitable access to the resource?(positive means equitability is promoted)

171. Do households have recognized legal rightto harvest the product for trade?

172. Do households have recognized legal rightto change the land use to another productionsystem?

173. Have the legal rights of households toharvest the environmental products forcommercial purposes changed in the past 10years?

174. Are households generally aware of thenature of their legal rights to harvest theproduct for commercial purposes?

175. Have there been any official claims byhouseholds to increase land/resource rightsover the last 10 years?



176. Are state laws and traditional rulesconflicting regarding exploitation ofenvironmental products?

177. What is the walking distance in km fromthe environment production area to the market?

178. What is the mode of travel?

179. Does the study indicate any externalsupport?

180. Has external support from donors or NGOsbeen targeted to:

181. Raw material producers

182. Traders

183. Processing/manufacturing industry

184. Retail/export industry

185. What is the main source of externalsupport? (local/national NGO, internationalNGO, international donors, national privatesector, or international private sector)

186. Does the study take into considerationenvironmental services?:

187. Does the study account for carbonsequestration? (Yes or No)

188. Does the study account for water retention?

189. Does the study account for soil erosionprevention?

190. Does the study account for electricity?

191. Does the study account for wind erosion?

192. Does the study account for biodiversity?

193. Does the study include other resourcesother than forestry?:

194. Does the study include fish resources?

195. Does the study include grasslands?

196. Does the study include freshwaterwetlands?

197. Does the study include coastal wetlandsand mangroves?

198. Does the study include freshwater bodies?

199. Does the study include salt-water bodies?

200. Does the study include others?

201. Does the study include the value ofprocessed goods, e.g., woodcarvings?

202. Does the study include subsistence incomein estimation of total income?

203. Does the study consider sustainability?(local perceptions regarding trends in resourcesupply)


Notes

1. The FAO definition of a forest includes aminimum area size of 0.5 ha, a minimumtree canopy cover of 10 percent, and thedominant use not being agriculture orurban; cf. FAO (2001).

2. A fourth possible measure is profit, whichequals value added minus labor costs. Theredoes not, however, seem to be any goodreason to omit labor costs while at the sametime retaining normal profit. Also, of course,in rural areas of poor countries, normalprofit will often be insignificant.

3. Environmental services provided by forestsare numerous, including carbonsequestration, water flow stabilization,erosion control, and biodiversity protection.Vast populations in faraway locations mayin some way benefit from forest resourceswithout ever realizing it. The focus here,however, is on the rural poor.

4. The first part of this section draws onAngelsen and Wunder (2003).

5. See Kaimowitz (2002) for sources and fullreferences for the figures provided in thissub-section.

6. The study used a much larger number ofdimensions to categorize the cases (Belcherand Ruiz-Perez 2001): geographic setting;characteristics of the product; productionsystem; ecological implications of

production; socioeconomic characteristics ofraw material production area; processingindustry and trade; institutionalcharacteristics of producers andgovernment policies; and externalinterventions.

7. The distribution between these categorieswas 18, 28, and 15, although this should notbe given much weight, as the cases were notselected randomly.

8. See Neumann and Hirsh (2000), Kaimowitz(2002), and Angelsen and Wunder (2003)and the references therein.

9. The model is elaborated in Angelsen andWunder (2003: 22).

10. Wealthy households also realize substantialincome from the forest but will more oftenengage in high-return activities that requirecapital (animals, machinery) that isunavailable to the poorest.

11. In brief, “Poverty Reduction StrategyPapers (PRSPs) describe a country’smacroeconomic, structural and socialpolicies and programs to promote growthand reduce poverty, as well as associatedexternal financing needs. PRSPs areprepared by governments through aparticipatory process involving civil societyand development partners, including theWorld Bank and the International Monetary



Fund (IMF).” (See www.worldbank.org/poverty/strategies.)

12. Only 9 studies presented data on livestockownership, while 28 cases presented dataon livestock income. The results thus mostlikely underestimate livestock ownership.The omission could partly also be a result ofdifferent objectives of the cases. Quite anumber of studies were furthermoreconcerned about livelihood strategies and

presented data on incomes rather thanownership.

13. In comparing incomes across countries, weused the purchasing power parity (PPP)conversions to account for price leveldifferences. We used the ratio betweencountries’ PPP GDP and nominal GDPfound in World Bank Development reportsfor the year of the study.


References

Angelsen, Arild, and Sven Wunder. 2003.Exploring The Forest–Poverty Link: KeyConcepts, Issues and Research Implications.Occasional Paper # 40. Bogor, Indonesia:Center for International Forestry Research.

Arnold, J. E. Michael. 2001. Forestry, Poverty andAid. Occasional Paper # 33. Bogor,Indonesia: Center for International ForestryResearch.

Arnold, Michael, Gunnar Kohlin, ReidarPersson, and Gillian Shepard. 2003. FuelwoodRevisited: What Has Changed over the LastDecade? Occasional Paper # 39. Bogor,Indonesia: Center for International ForestryResearch.

Aryal, B. 2002. “Are Trees for the Poor? A Studyfrom Budongo Forest, Uganda.” M.Sc.Thesis, Department of Economics and SocialScience. Agricultural University of Norway.

Barrow, E. and M. Murphree. 2001. “CommunityConservation. From Concept to Practice.” InHulme and Anderson. African Wildlife andLivelihoods. The Promise and Performance ofCommunity Conservation. London: JamesCurry Publishers.

Batagoda, B. M. S., R. K. Turner, R. Tinch, K.Brown. 2000. Towards Policy RelevantEcosystem Services and Natural Capital Values:Rainforest Non-Timber Products. CSERGEWorking Paper. GEC 2000-06.

Belcher, Brian, and Manuel Ruiz-Perez. 2001. AnInternational Comparison of Cases of ForestProduct Development: Overview, Descriptionand Data Requirements. Working Paper # 23.Bogor, Indonesia: Center for InternationalForestry Research.

Belcher, Brian, Manuel Ruiz-Perez, andRamadhani Achdiawan, 2003. “GlobalPatterns and Trends in Non-Timber ForestProduct Use.” International Conference onRural Livelihoods, Forest and Biodiversity.Bonn, Germany, 20–23 May 2003.

Botha, N. B. 2003. “Looking for a Path Out ofPoverty: The Study of Chimaliro ForestryReserve, Malawi.” M.Sc. thesis, Departmentof Economics and Social Science.Agricultural University of Norway.

Bulte, E. H., and D. P. van Soest. 2001.“Environmental Degradation in DevelopingCountries: Explaining the EnvironmentalKuznets Curve.” Journal of DevelopmentEconomics 65(1):225–36.

Byron, Neil, and Michael Arnold. 1999. “WhatFutures for the People of the TropicalForests?” World Development 27 (5):789–805.

Campbell, Bruce M., and Martin K. Luckert,eds. 2002. Uncovering the Hidden Harvest.Valuation Methods for Woodlands and ForestResources. London: Earthscan.



Cavendish, W. 2000. “Empirical Regularities inthe Poverty-Environment Relationship ofRural Households: Evidence fromZimbabwe.” World Development 28(11):1979–2003.

Cavendish, Will. 2003. “How Do ForestsSupport, Insure and Improve theLivelihoods of the Rural Poor. A ResearchNote.” Bogor, Indonesia: Center forInternational Forestry Research.

Dercon, Stefan. 2000. Income Risk, CopingStrategies and Safety Nets. WPS # 2000: 26.Oxford, UK: Center for the Study of AfricanEconomies.

Dorward, A., and S. Anderson. 2002.“Understanding Small Stock as LivelihoodAssets: Indicators for FacilitatingTechnology Development andDissemination.” Unpublished Report onReview and Planning Workshop 12–14August 2002. Imperial College at Wye, U.K.

Ellis, Frank. 2000. Rural Livelihoods and Diversityin Developing Countries. Oxford: OxfordUniversity Press.

FAO (Food and Agriculture Organization). 2001.FRA 2000. On Definitions of Forests and ForestChange. FRA Working Paper # 33. Rome.

Fisher, M. 2002. “Household Welfare and ForestDependence in Rural Malawi.”Unpublished Working Paper.

Godoy, R., N. Brokaw, and D. Wilkie. 1995. “TheEffect of Income on the Extracting of NonTimber Forest Products: Model, Hypothesis,and Preliminary Findings from the SumuIndians of Nicaragua.” Human Ecology 23(1).

Godoy, R. A., R. Lubowski, and A. Markandya.1993. “A Method for the EconomicValuation of Non Timber Forest Products.”Economic Botany 47:220–33.

Gosamalang, D. 2003. “Changing Legal Statusof Mt. Elgon Forest Reserve—Impacts forLocal People’s Livelihoods.” M.Sc. thesis,Noragric. Agricultural University ofNorway.

Hulme, D., and M. Murphree. 2001. AfricanWildlife and Livelihoods. The Promise andPerformance of Community Conservation.London: James Curry Publ.

Kaimowitz, David. 2002. Forest and RuralLivelihoods in Developing Countries. Bogor,Indonesia: Center for International ForestryResearch.

Knorr-Cetina, K. 1981. The Manufacture ofKnowledge. An Essay on the Constructivist andContextual Nature of Science. Oxford:Pergamon Press.

Landell-Mills, Natasha, and I. Porras. 2002.Silver Bullet of Fools’ Gold: A Global Review ofMarkets for Forest Environmental Services andTheir Impact on the Poor. London:International Institute for Environment andDevelopment.

Neumann, R. P., and E. Hirsch. 2000.Commercialisation of Non-Timber ForestProducts: Review and Analysis of Research.Bogor, Indonesia, and Rome: Center forInternational Forestry Research and Foodand Agriculture Organization.

Nyland, T. K. 1998. “Local People’s Use of Landand Natural Resources, Laos.” CandidateAgric. thesis, Department of Economics and


References

Social Science. Agricultural University ofNorway.

Pagiola, S., J. Bishop, and N. Landell-Mills.2002. Selling Forest Environmental Services.Market-Based Mechanisms for Conservationand Development. London: Earthscan.

Panayatou, Theodore, and Peter S. Ashton. 1992.Not by Timber Alone. Economics and Ecologyfor Sustaining Tropical Forests. Washington,D.C.: Island Press.

Pattanayak, Subhrendu K. and Erin O Sills.2001. “Do Tropical Forests Provide NaturalInsurance? The Microeconomics of Non-Timber Forest Product Collection in theBrazilian Amazon.” Land Economics 77(4):595–612.

Peters, C. M., A. H. Gentry, and R. O.Mendelsohn. 1989. “Valuation of anAmazon Rainforest.” Nature 339:655–56.

Sheil, Doug, and Sven Wunder. 2002. “TheValue of Tropical Forests to LocalCommunities. Complications, Caveats, andCautions.” Conservation Ecology 6 (2).

Sjaastad, E., Y. Ngaga, S. Chamshama, K.Magnussen, G. Monela, and P. Vedeld. 2003.“Resource Economic Analysis of CatchmentForests in Tanzania. A Report to MNRT.Tanzania.” Noragric. 2003.

Stoian, D. 2003. “Making the Best of TwoWorlds: Rural and Peri-Urban LivelihoodOptions Sustained by Non-Timber ForestProducts from the Bolivian Amazon.”International Conference on RuralLivelihoods, Forest and Biodiversity. Bonn,Germany, 20–23 May 2003.

Turner, K., J. Paavola, P. Cooper, S. Farber, V.Jessamy, and S. Georgiou. 2002. ValuingNature: Lessons Learned and Future ResearchDirections. CSERGE Working Paper. EDM2002-05.

Vedeld, P. 2002. The Process of Institution-buildingto Facilitate Local Biodiversity Management.Noragric Working Paper No. 26,Agricultural University of Norway.

World Bank. 2001. “A Revised Forest Strategyfor the World Bank Group” (draft). 30 July2001. Washington D.C.

Counting on the Environment - World Bankdocuments.worldbank.org/curated/en/... · Counting on the...

Documents

Transcript of Counting on the Environment - World Bankdocuments.worldbank.org/curated/en/... · Counting on the...