1

Title: “Political Budget Cycles and Fiscal Illusion – a panel data study”

Author: Paulo Reis Mourao; Department of Economics & NIPE (Núcleo de Investigação em Políticas Económicas); University of Minho; Gualtar, 4700 Braga – Portugal. E-mail: paulom@eeg.uminho.pt Abstract: This paper presents the first cross-country evidence of electorally-motivated changes in the political budget cycles in a large range of democracies, including the effects derived of the Fiscal Illusion problem. To date, empirical tests of political business cycles have ignored the important issue of the Fiscal Illusion effects. Results show that, without considering the effects of Fiscal Illusion as a complex process of hiding the real fiscal situation from the political agents (policymakers and voters), election-year government balance shifts downwards and post-election year government surplus shifts upwards. Considering the Fiscal Illusion phenomenon, it was observed that countries with higher values of a Fiscal Illusion Index show worse budget deficits and their budget cycle is more negatively pronounced (more significant negative differences to the average national budget deficit). Evidence further shows that these effects are contingent on the economic development level and on the maturity of the democratic system. Key Words: Political Budget Cycles; Fiscal Illusion; panel data estimation (JEL: H6, O11, D72)

2

1. Introduction Different models of political cycles highlight either the “opportunistic” or the “partisan” incentives of policymakers. In the first group of models, the policymakers maximize their popularity or their probability of being reelected. In the second group, different political parties represent different constituencies, and when in power, they follow policies favourable to their supporting groups. Besides these two groups, the most recent researches focus on Adverse Selection models. These models emphasize the opacity of the governance practices felt by most of voters and strengthened by the impossibility of full knowledge on their own competence by the politicians. Though, the Political Budget Cycles models have been importing more and more elements from a fiscal thematic rediscovered by Buchanan in the 1960s: the Fiscal Illusion problem. Some years later than the Scottish Enlightment, in Italy, Amilcare Puviani (1903) intended to answer to the question “How can a politician best use his powers of the purse to promote his

political projects?” with his work “The Theory of Fiscal Illusion”. Although his bibliography is more extensive, the cited title had become his masterpiece, especially due to the seminal contribution of other Italian author, Mauro Fasiani (1941), and to the great impulse from Buchanan’s (1960 and 1967) quotes. Puviani (1903) introduced the hypothesis of “Fiscal Illusion” as an observable answer to the reported question. In spite of the precedence of Mill (1848) writing, Puviani (1903) was the first author recurring to the terms of “Fiscal Illusion”. With these terms, Puviani (1903) wanted to point the opacity that could be administered by public decision-makers in the imposition of taxes or in public spending management. However, the sense of these terms had changed through the years, the preliminary hypothesis of Puviani (1903) has promoted a large discussion on the methodological field and on the achieved evidence and, curiously, it was registered that other economic subjects (like the Monetary Economics, for instance) imported the terminology to explain different phenomena. The rest of the paper is organized as follows. Section 2 provides a summary of the Fiscal Illusion literature and the consequences of Fiscal Illusion on Political Budget Cycles and its predictions. Section 3 discusses the construction of a Fiscal Illusion Index available to a large range of democracies (the 68 democracies particularly studied by Brender and Drazen (2004), since 1960 to 2006). In Section 4, there appear the specification and the dynamic panel data estimation technique that had been applied to evaluate the political budget cycles of these democracies, now considering as an explanatory variable the Fiscal Illusion Index. Our results showed that Fiscal Illusion is an important issue in the political business cycle thematic. It was observed that countries with higher values of the Fiscal Illusion Index (discussed in the Section 3) show worse budget deficits and their budget cycle is more negatively pronounced (more significant negative differences to the average national budget deficit). The differences among the groups of countries mainly reveal the expected results: the Fiscal Illusion phenomenon is more influent in new democracies and in developing countries, deteriorating government accounts. Section 5 concludes.

3

2. Fiscal Illusion and Political Budget Cycles: the theoretical framework 2.1- Fiscal Illusion: a brief summary This sub-section tries to highlight the deep complexity behind the studies around the Fiscal

Illusion thematic. For those interested in more theoretical developments, Mourao (2006a) is a work that expands these authors and their claims. James Buchanan signed in 1967 the work Public Finance in Democratic Process: Fiscal

Institutions and Individual Choice. In the Chapter 10, the terms The Fiscal Illusion appear as title. He confesses that, at the time, the discussion of Amilcare Puviani’s (1903) main theoretical contribution – the original Illusione Finanziaria – that he has already promoted in Fiscal Theory and Political Economy, edited in 1960, remained the only available summary in English. After Buchanan’s quotes, other authors have recurred to Fiscal Illusion for many purposes and with many different senses, as notices in Mourao (2006a). This sub-section intends to highlight the most prominent of these researches on Fiscal Illusion. According to Puviani’s original idea, the objective of the ruling group becomes that of arranging or organizing the fiscal structure so that the resistance of the dominated class is effectively minimized. Consequently, the rulers ask: “If we desire to minimize taxpayer resistance for any given level of revenues collected, how will it set out to organize the fiscal system?” The answer recurs to both sides of the budget – “illusions” are created through taxes and through public spending programs. The most relevant side is the branch of public revenues. This branch can be subdivided into seven means of introducing fiscal illusion: Obscuration of the individual shares in the opportunity cost of public outlays; utilization of institutions of payments that are planned so as to bind the requirement to a time period or an occurrence which the taxpayer seems likely to consider cheering; charging of explicit fees for nominal services provided upon the occurrence of impressive or pleasant events; levying taxes that will capitalize on sentiments of social fear, making the burden appear less than might otherwise be the case; use of “scare tactics” that have a propensity to make the alternatives to particular tax proposals emerge worse than they are; fragmentation of the total tax weight on an entity into numerous small levies; and opacity of the final incidence of the tax. In his seminal book, Downs (1957) recognized that politicians have little incentive to correct fiscal illusion – their incentive is to spend more on public investment projects that pay off within a four-to-five-year electoral cycle. Galbraith (1958) also accused a not very usual image of the traditional case of Fiscal Illusion. As a consequence, Galbraith (1958) identified that governments would opt for a “sub-optimally low” value of public provision of services. Galbraith (1958) argues that public spending is less than optimal and notes that advertising and marketing are greater in the private sector. In recent times, Twight (1994) and Alesina and Perotti (1996) outline several means by which politicians may make public budgets less transparent, thereby raising the transaction costs of monitoring fiscal conditions for a public subject to fiscal illusion or incomplete information: biased macroeconomic forecasts, biased estimates of the effects of policy changes on

4

budgetary outcomes, strategic use of on- and off-budget expenditures and receipts, manipulation of budgetary baselines, and multiyear budgeting. Von Hagen and Harden (1994) developed a framework in which there is a failure to fully internalise the true economic costs of public expenditure – another kind of fiscal illusion. The narrow interests of individual spending ministers dominate over the collectivists concerns of the Minister of Finances. Consensus is arrived at in cabinet on the basis of the spending ministers, either explicitly or implicitly, backing each other bids resulting in “something for everyone” and thus a sub-optimal overall level of spending. If this framework accurately reflects the actual process of public expenditure determination then intra-governmental institutional reform may be required to redress the situation. The opportunity to expand the assumption of (full) rationality in models of Public Economics promoted a reaction from a diversity of authors synthesized in Wittman (1995). Wittman (1995) does not believe in models assuming homogeneous misinformed electors or consumers. The costs of decision making are either ignored or assumed not to distort choice. When outcomes do not take place with certainty then economists typically assume that individuals maximize expected utility. In this neoclassical framework, anomalies (of the individual perception) are the exception rather than the rule. For instances, in numerous social areas, individuals do not have the “requisite skills”, yet they are able to make the correct decision. Also if voters have specific interests or concerns, they can consult special interest groups for information on the candidates’ positions on the issues in question. Cohen and Percoco (2004) refer that the most recent macroeconomic literature has focused on the effect of public spending contraction and has provided two alternative theories: the theory of asymmetric effects of public spending and the theory of fiscal illusion. Besides Easterly (1999), the impact of downward in public investment in the lack of competitiveness and a consequent worsening of fiscal deficit has also been studied by Calderón, Easterly and Servén

(2003) who develop a theoretical framework aiming to provide an explanation of what is called fiscal illusion. In particular, fiscal adjustment can be thought as an illusion when it reduces the budget deficit but the government net worth remains unaffected. Easterly (2001) shows that, under certain conditions, a government will lower the conventional deficit while leaving its path of net worth unchanged and when required to lower its debt accumulation, the government will lower its asset accumulation or increase its hidden liability accumulation by an equal amount, which follows the structural argument from Easterly (1999). Jensen and Vestergaard (1999) define fiscal illusion as a situation where public decision-makers (namely, the European Union, EU) only incorporate a part of the costs incurred by the constituents (the Member States). Fiscal illusion means that EU does not full take into account the tax costs of the Member States when maximizing the benefit. The degree of the fiscal illusion may also be interpreted as compliance and enforcement costs associated with letting the EU tax the Member States on the basis of individual producers (in their study, the fishermen). Wagner (2001) also recognized that Puviani (1903) gave most of his attention to taxation - it is there where the term Fiscal Illusion precisely obtains its meaning. Consequently, the politician should make taxes become less of a burden than they really are. In his work, Wagner (2001) identifies trade taxes as a good form of taxation due to its bad perceptibility by voters.

5

Searching for the psychological foundations of fiscal illusion, Sanandaji and Wallace (2003) reported the Theory of Mental Accounting. The Theory of Mental Accounting studies the set of cognitive operations used by individuals and households to organize, evaluate and keep track of financial activities. According to this theory, physical money is more valuable than electronic checks and there are evidences for a kind of public hedonic editing – electors actually prefer not to be reminded of the costs of public programs. Therefore, this perspective offers both arguments that the underestimation of tax levels could be beneficial to a hedonist

society but also arguments that support the predictions from the Public Choice thought – tax

illusion can be used to facilitate rent seeking and be harmful to the same society. Some examples of illusions that arise by a nexus between monetary and fiscal factors are provided by Forte (2004): i) fiscal drag due to the automatic increase of real tax rates, in a personal income tax, due to the loss of value of monetary income subject to the progressive rates and of the lump sum deductions from the taxable income; ii) taxation of revenues of capital, in the income tax, at their face value, which normally includes a compensation for the loss of value of the capital invested; iii) taxation of profits due to the fact that depreciations allowances are based on the book value of the assets and this value in most cases is not the actual value but this historical one; iv) the Maastricht rules based in nominal deficits rather than real deficits (that is identified to the formula Index of Consumer Prices*Debt/GDP +

Nominal Deficit) which works for countries with an higher Debt/GDP and a greater propensity to inflation (the obtained results from the imposition of budgetary restrictions of the Maastricht and Amsterdam Pacts are not sufficiently strong to improve the performances of those countries). For Garcia-Alegre and Lopez-Casasnovas (2004), the topic of financial illusion and its relevance in public management has been traditionally discussed in the context of fiscal

illusion and public expenditure growth. They had described the existence of financial illusion

in public accounting since it allows for larger public expenditure increases and managerial slack. Therefore, three strategies can be pursued to manipulate the citizen (or the modelled median voter): i) political actors and bureaucrats may try to show that the tax-price of public sector services appear to be lower than actually is; ii) politicians may find desirable to foster the idea that the median voter is in receipt of larger real income increases as a result of tax/expenditure decisions; iii) politicians and bureaucrats can attempt to alter the preferences of the voters to raise the absolute value of the marginal rate of substitution between public supplied goods and the rest. Finally, for P. Jones (2006), fiscal illusion is asymmetric. Within overall government budgets, domestic programs are very likely to crowd out international programs. This asymmetric fiscal illusion is also evident in questionnaire responses on public expenditure priorities. In some polls (like the mentioned British Social Attitudes Survey), health care and education are invariably considered first or second priority for additional expenditure. Overseas aid has remained at the bottom with defense expenditure just a little higher. While the relative benefits of international programs are underestimated, the relative costs are exaggerated, according to P. Jones (2006). 2.2 - Fiscal Illusion and Political Budget Cycles We can distinguish three generations of political budget cycles models, following Andrikopoulos et al. (2004), Shi and Svensson (2004) or Mink and De Haan (2005).

6

The first generation models emphasize the incumbent’s intention to secure re-election by maximizing its expected vote share at the next election (Nordhaus, 1975). It is assumed that the electorate has backward-looking expectations and evaluates the government on the basis of its past track record. As a result, these models imply that governments, regardless of ideological orientation, adopt expansionary fiscal policies in the late year(s) of their term in office in order to stimulate the economy. The second generation models – adverse selection-type models – emphasize the role of temporary information asymmetries regarding the politicians’ competence level in elucidating electoral cycles in fiscal policy. In these models, signalling is the most important force behind the cycle. The most well-known of these models is due to Rogoff and Sibert (1988), who assume that each political candidate has a competence level (high or low), which is only known to the politician and not to the electorate. Voters want to elect the more competent politician and form rational expectations regarding the incumbent’s type based on the most perceptible fiscal policy outcomes. Before the election, the high-type incumbent will attempt to signal his type (and thereby increase his chances of re-election) by engaging in expansionary fiscal policy, which is less “costly” for him than it is for the low type. This leads to a pre-election increase in the government deficit when a competent politician is in office. Persson and Tabellini (2000) or Shi and Svensson (2002) are examples of the third generation models. As in the adverse selection models, it is assumed that each politician has some competence level which is not full identified by the electorate. Additionally, it is assumed that also the politician himself cannot previously observe this competence level. Therefore, politicians are unsure about how well they will be able to handle future problems. Voters are rational and want to elect the most competent politician since that would imply higher post-election public goods production. Their inference is based on the observable macroeconomic performance of the incumbent government. The key assumption is that the incumbent government can exert a hidden effort, that is, to use a policy instrument unobservable to the public, which is a substitute for competence. The success of this kind of “secret weapons” depends on the outcomes derived from fiscal manipulation (short-term borrowing and revenues mix), the incumbent’s (and the challenger’s) credibility, the democratic maturity, and on the soundness of the campaigns – all issues studied by the Fiscal Illusion literature for a long. Interestingly, we observe that these three schools show a developing thought that is becoming closer to the Fiscal Illusion issues. Therefore, Fiscal Illusion literature is needed to deepen the Political Budget Cycle evidence on democratic societies. As observed by the previous literature, Fiscal Illusion generates a higher number of opportunities for budget manipulation and for changing fiscal aggregates, which by its turn promote less fiscal stability highly associated with a weak economic growth1. Therefore, and according to the Sub-Section 2.1, it is expected that higher values from Fiscal Illusion pattern promote higher budget changes and worse values of the government surplus. According to Persson and Tabellini (2000) or Shi and Svensson (2002), Fiscal budget is asymmetric (there is not a full correction by a budget surplus in the k years after an election that corrects the budget deficits observed in the k years before that election). Therefore, incumbents who hold political offices in countries with higher levels of Fiscal Illusion should face fewer costs of recurring to Fiscal Illusion practices than incumbents who were elected by more conscious

1 See Ndulu and O’Connell (1999) or Barro and Lee (1997).

7

electorates2. As a consequence, the processes of signalling competences by higher differences between visible expenditures and hidden revenues should be greater in the former countries, inducing worse deficits. The budget changes, reflecting the difference between the budget surplus of a random year and the national average in the democratic years, should also evidence more negative values in the countries with more Fiscal Illusion. The reasons are various and they are more notorious in the first years of democratic life: the short democratic experience of the electorate, the coming effects from the transition political period (usually, the democracies substitutes autocratic regimens, with an extraordinary expansion of the public expenditures in the first years accompanied by weak budget control3), the absence of national convergent aims among the political agents, dispersed in many parties, and the special slackness of the pressure from international creditors. The Figure 2.1 suggests these implications (a stylized asymmetric cycle, where deficits area is greater than the surpluses one, with Fiscal Illusion interaction).

Though, we are in position to test these claims, .i.e., to empirically verify if there are significant differences among countries due to their national levels of Fiscal Illusion, if higher levels of Fiscal Illusion distort the stylized pattern of political budget cycles and which direction these distortions take. Next sections (3 and 4) will, respectively, report the steps related to the construction of a Fiscal Illusion Index available to a large range of democracies since 1960 and the obtained empirical results.

2 An ingenious comparison, using our own current experience, recognizes that illusionists have more success with children than with their parents. 3 See Morales and McMahon (1999), for the latin american experiences, or Mourao (2006b) for the case of a young OECD democracy (Portugal).

Figure 2.1 - Fiscal Illusion and the Stylized Political Budget Cycle

pre-

election t (election) post-

election

Budget Surplus

Low Fiscal Illusion High Fiscal Illusion Average (low fiscal illusion) Average (High Fiscal Illusion

Difference to the average (low fiscal illusion) Difference to the average (high fiscal illusion)

8

3. Towards a Fiscal Illusion Index

3.1 - The rationale behind an index for the Fiscal Illusion As observed in the previous sections, the phenomenon of Fiscal Illusion is rather complex. It is complex because nowadays there is a large set of authors who contributed to its study with different senses, it is complex because it is referred to a wide range of economic realities and, finally, its complexity is also derived upon the methodological use that is given to Fiscal

Illusion itself. As Mourao (2006) states, sometimes authors use Fiscal Illusion as an assumption, other researchers employ the terms relating them to hypotheses of solving previous problems and other economists identify Fiscal Illusion to consequences of fiscal manipulation. In these cases, the construction of an Index that combines the many different dimensions of the studied phenomenon is largely suggested. Evidence suggests that studying Indexes of complex fiscal realities is more efficient than analyzing isolated variables (Alesina et al., 1996; Hameed, 2005). Alesina et al. (1996) collected information on the budget institutions of Latin America countries. They classified those countries as a function of the values returned from their Index of Budgetary Institutions and also depending on the presence of budgetary practices of control. Their Index incorporated ten basic dimensions: constitutional constraints, legal requirement for the approval of a macro program, borrowing constraints, authority of minister of finances, amendments by the Congress, consequences of Congress’ rejection of the Budget, opportunity of modifying the Budget after Congress’ approval, opportunity of cutting spending by the Government after Congress’ approval, assumption by the Government of other political Agencies’ debt, and autonomy of these other Agencies to borrow. They concluded that transparent procedures were associated to more fiscal discipline. Hameed (2005) developed indices of fiscal transparency for a broad range of countries based on the IMF's Code of Good Practices on Fiscal Transparency, using data derived from published fiscal transparency modules of the Reports on the Observance of Standards and Codes. The indices cover four clusters of fiscal transparency practices: data assurances, medium-term budgeting, budget execution reporting, and fiscal risk disclosures. Hameed (2005) concluded that more transparent countries are shown to have better credit ratings, better fiscal discipline, and less corruption, after controlling for other socioeconomic variables. Alesina et al. (1996) and Hameed (2005), among others, specially studied the reverse of Fiscal Illusion – the Fiscal Transparency. Consequently, they selected the analyzed dimensions recurring to the nearest ones of the Governance practices fields. As the authors who specifically studied the Fiscal Illusion notice, this phenomenon is not restricted to the ruler agents but it is also verified in the ruled ones, electors and firms. Therefore, a good Index for Fiscal Illusion must contemplate this variety of agents with their behaviour. Nardo et al. (2005) also recognize that indexes, as composite indicators, provide simple comparisons of countries that can be used to illustrate complex and sometimes elusive issues in wide ranging fields. These indicators often seem easier to interpret by the general public

9

than finding a common trend in many separate indicators and have proven useful in benchmarking country performance. Finally, Mourao (2005) stated that working with analytical indexes is better to understand the economic phenomenon instead of its particularized components. Working with indexes also avoids the introduction of redundant variables in econometric models, with the common trouble of loosing degrees of freedom and, finally, it is more suitable to truly approach the involved methodological complexity. 3.2 -The Fiscal Illusion Index In this sub-section, the main steps behind the construction of the Fiscal Illusion Index will be revealed. As Nardo et al. (2005) state, economic or social indexes can send misleading policy messages if they are poorly constructed or misinterpreted. Therefore, it is very relevant to follow prudent steps in order to reach significant aims and to avoid simplistic lectures. Following the previous sections, Table 3.1 provides a basis for the theoretical framework behind the construction of an index related to the phenomenon of Fiscal Illusion. From the original suggestion of Puviani (1903) and his lecture by Buchanan (1960) to some of the most recent authors, like Wagner (2001), Sanandaji and Wallace (2003) or Jones (2006), we clearly identify that the focused dimensions are numerous. Besides the traditional dimensions (Composition of Public Revenues, Money creation, Composition of Public Debt, or Relevance of certain revenue sources), we also find the Governmental discourse manipulation and electorate believes, the Immaturity level of the democracies and the Interaction between interest groups and political behaviour, among others. TABLE 3.1 – Authors and their focus on Fiscal Illusion Authors Purpose of discussing Fiscal Illusion Focused Dimensions

Puviani (1903) Buchanan (1960 and 1967)

To explain the budgetary behaviour of the rulers

Composition of Public Revenues; Money creation; Composition of Public Debt; Relevance of certain revenue sources

Downs (1957) To understand the reproduction of bad governance practices

Political strategies of the ruler group

Galbraith (1958) Downs (1960)

To understand the under-provision of public goods

Public expenditures manipulation

Twight (1994) Alesina and Perotti (1996)

To discuss means by which public budgets are less transparent

Governmental discourse manipulation and electorate believes

Von Hagen and Harden (1994)

To explain sub-optimal overall levels of public spendings

Number of governmental Ministries; Different objectives of governmental agents

Wittman (1995) To cover systematic public decisions Immaturity level of the democracies Easterly (1999 and 2001) Calderón et al. (2003) Cohen and Percoco (2004)

To understand some budgetary practices like the preference for cutting certain kind of less visible investment outlays

Composition of Public Capital outlays

Jensen and Vestergaard (1999)

To discuss the European Union Politics Governmental rent-seeking

Wagner (2001) To explain the public preference for certain types of taxation

Relevance of trade taxes

10

Sanandaji and Wallace (2003)

To understand the opacity of the design of the taxes

Interaction between interest groups and political behaviour

Forte (2004) To discuss the efficacy of Maastricht Rules

Real Public Budget (considering inflation rates and public debt)

Garcia-Alegre and Lopez-Casasnovas (2004)

To discuss the relevance of public reports in the public management

Composition of Public Expenditures

P. Jones (2006) To explain the crowding-out on financing international programs

Electorate preferences on national issues (poor societies prefer ‘economic growth’ as a main issue, also following Maslow, 1970)

After the identification of the theoretical framework, it is the moment to find the equivalent variables and to select data. Table A1 (in Annexes) provides a synthesis of this effort. Finding variables in order to fulfil the requirements of the focused dimensions was a very patient work, combining different data sources and discussing the best way to reduce the count of missing values of the provided databases. The range of twenty-six variables (please confirm all of them in Table A1) was selected considering their analytical soundness, measurability, country coverage, relevance to the phenomenon being measured and relationship to each other. The data are related to 68 democracies, including developing and developed countries. These countries4 were selected using Polity IV filter, following Brender and Drazen (2004) who have chosen only those democracies with positive values from the filter. In the cases of data scarcity of some variables, Shi and Svensson (2002) and Nardo et al. (2005) were followed, when they suggest to substitute the missing values by the national average values of the variables. The main sources were Barro and Lee (2000), Cross-National Time-Series Data Archive (2006), Database of Political Institutions (2004), Government Finance Statistics (2006), International Country Risk Guide (2006), International Financial Statistics (2006), International Labour Organization Statistics (2006), Voter Turnout since 1945 (2002), and World Development Indicators (2006). The web sites http://www.worldpublicopinion.org and http://www.idealist.org also provided data. The panel starts in 1960 and the data goes to 2006. Almost all the variables were used with their provided values or enriched with the suggestions of Shi and Svensson (2002) and Nardo et al. (2005), in this step. The exceptions were the variables built upon Herfindahl Indexes or growth rates. In the former case, the Herfindahl Index of a referred fiscal dimension followed Pommerehne and Schneider (1978) or Becker (1983) and it was defined in the simplest way, as the sum of the squares of the shares of each individual component of that fiscal dimension. In the latter case, the growth rates were computed as the difference of two consecutive observations of the logarithmized variable, yearly observed. To avoid the common critic ‘indicator rich but information poor’, when there is the use of indicators selected in an arbitrary manner with little attention paid to the interrelationships between them, the data were observed through multivariate analysis. When there are arbitrary

4 Argentina, Australia, Austria, Belgium, Bolivia, Brazil, Bulgaria, Canada, Chile, Colombia, Costa Rica, Cyprus, Czech Republic, Denmark, Dominican, Ecuador, El Salvador, Estonia, Fiji, Finland, France, Germany, Greece, Guatemala, Honduras, Hungary, Iceland, India, Ireland, Israel, Italy, Japan, Korea, Lithuania, Luxembourg, Madagascar, Malaysia, Mali, Mauritius, Mexico, Nepal, Netherlands, New Zealand, Nicaragua, Norway, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Romania, Russia, Slovakia, Slovenia, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Trinidad and Tobago, Turkey, United Kingdom, United States of America, Uruguay, and Venezuela.

11

weights given to the indicators in order to constitute an index (usually, all indicators have the same weight), this can lead to indices which overwhelm, confuse and mislead decision-makers and the general public. Although there are some available methods (see Nardo et al , 2005), the chosen method to explain the variance of the observed data through a few linear combinations of the original data was the Multivariate Analysis. Some interesting references on this method are Hair et al. (2005), Kent, Bibby and Mardia (2006), and Johnson and Wichern (2007). Synthesizing, even though there are Q variables, much of the data’s variation can often be accounted for by a small number of variables – principal components, or linear relations of the original data that are uncorrelated. At this point there are still Q principal components, i.e., as many as there are variables. The next step is to select the first, say P<Q principal components (factors) that preserve a “high” amount of the cumulative variance of the original data. Table 3.2 shows that five factors were retained, using the method of principal component factors (for economy of space, the other factors, non-significant, were omitted). These five factors account for more than 80% (84,46% precisely) of the total variation. TABLE 3.2 - Component loadings for Fiscal Illusion variables Factor Eigenvalue Difference Proportion Cumulative

1 12.64945 4.17212 0.4080 0.4080 2 8.47733 6.02921 0.2735 0.6815 3 2.44812 0.88984 0.0790 0.7605 4 1.55827 0.50874 0.0503 0.8107 5 1.04953 0.17508 0.0339 0.8446 Table 3.3 reveals the Rotated Factor Loadings for Fiscal Illusion variables, a powerful suggestion of the weights that will calibrate each variable in the aggregate index5. These results were achieved through the Principal Components Extraction Method with varimax normalised variation. TABLE 3.3 - Rotated Factor Loadings for Fiscal Illusion variables

Variable 1 2 3 4 5 Uniqueness

trd 0.92504 -0.15369 0.12477 0.15922 0.00746 0.07971

icrg -0.43434 0.38172 0.71098 0.00244 -0.02269 0.15962

inv_vot -0.78653 -0.55043 0.08307 0.01592 -0.06039 0.06760

pub_employ 0.18532 0.93288 -0.25430 -0.06257 -0.00565 0.02678

pres_parl -0.52962 -0.16898 -0.34620 0.28901 -0.32231 0.38369

checks -0.62336 -0.01468 0.11139 0.36003 0.36115 0.33875

highedu 0.06366 0.74337 0.51790 0.11634 0.18080 0.12891

gov_confid -0.98835 -0.05771 -0.02640 0.03472 -0.07200 0.01275

higprefer 0.27676 0.89509 -0.31793 -0.00737 -0.00324 0.02107

npopmilli 0.59355 0.77254 -0.18554 -0.04467 0.05065 0.01190

5 See, please, for an explanation of the factor loadings Hair et al. (2005), Kent, Bibby and Mardia (2006), or Johnson and Wichern (2007).

12

mediacs -0.08509 0.52399 0.58645 -0.22323 -0.32325 0.21995

cabin_size -0.69747 -0.25477 0.26530 -0.23952 0.04595 0.31876

money 0.65390 0.50930 0.36216 -0.22573 0.09536 0.12182

shortdebt 0.09845 0.46136 0.26501 0.51703 0.33816 0.32556

pccaptransf 0.80430 -0.28685 0.06437 0.27005 -0.06888 0.18901

pctransfpart -0.16092 0.94149 -0.26535 -0.02251 -0.00277 0.01677

pcgood 0.79165 0.46705 -0.10527 -0.09330 0.05073 0.13279

pcinttrade 0.83732 -0.44819 0.14381 0.18450 -0.03746 0.04189

txherfind 0.84163 0.14843 -0.03391 0.25341 -0.00365 0.20425

pceduc 0.86191 -0.43404 0.17245 0.07932 -0.01953 0.03232

herfdesp -0.84717 0.26357 -0.05646 -0.11338 0.04697 0.19458

ratcurcap -0.78792 0.26162 -0.07167 -0.20566 0.09487 0.25431

pcprofit 0.56727 -0.59957 0.14311 -0.24502 0.20214 0.19734

pcinherita 0.43490 -0.45362 0.15586 -0.42078 0.26486 0.33360

realbud -0.11555 -0.56247 -0.44526 0.17898 0.45313 0.23466

gnidebt 0.22851 0.83302 -0.22043 -0.15045 -0.05071 0.18007 High and moderate loadings (>0.50) indicate how the sub-indicators are related to the principal components. The first factor has high positive coefficients (loadings) with trd (0.93), pccaptransf (0.80), pcgood (0.79), pcinttrade (0.84), txinddir (0.71), txherfind (0.84), pceduc (0.86) and pcprofit (0.57), indicating that Factor 1 may be due to Fiscal Illusion in its strictus

sensu, motivated by fiscal manipulation. Factor 2 is mainly dominated by political-economic variables: pub_employ (0.93), highedu (0.74), higprefer (0.90), npopmilli (0.77), mediacs (0.52), money (0.51), pctransfpart (0.94), txinddir (0.66) and gnidebt (0.83). Factors 3, 4 and 5 are mainly subject to the government’s ability to persuade economic agents and to the budget restrictions. As a normalisation method, it was chosen for each (country-year) observation the percentile rank considering all observations from each variable and the expected effect on Fiscal Illusion by a rise of the variable6. This method allows to express prior units with different measures into normalised (and more likely comparable) variables. In the last step of the production of the Fiscal Illusion Index, it is the moment to deal with the construction of the weights from the matrix of factor loadings after rotation, given that the squares of factor loadings represent the proportion of the total unit variance of the indicator which is explained by the factor. The approach used by Nicoletti, Scarpetta and Boylaud (2000) is that of grouping the sub-indicators with the highest factors loadings in intermediate

composite indicators, whose number is equal to the number of factors. Therefore, each normalised variable with a significant factor loading (greater than 0.7) will have a weight equal to the square of the factor loading divided by the explained variation by the factor. At the end, each intermediate composite indicator will have a weight equal to its proportion of the variance explained by all the factors. In our case, the final value given to each country-year observation is re-scaled, using again the percentile rank and considering all weighted values. Table A.2 shows only two values (1960 and 2006, the first and the last years) of the

6 If the expected effect was negative, then the rank was re-ordered, considering the difference between 1 and the (raw) percentile rank. Otherwise, the rank was not modified.

13

Fiscal Illusion Index for each one of the studied countries. Following the construction steps, higher values of the Index reveal higher patterns of Fiscal Illusion (just remember the factor loadings). The margins of error for the aggregate Fiscal Illusion indicator are non-significant7. This point is illustrated in Figure A.1. In the two panels of Figure A.1, we organize countries in ascending order according to their point estimates of Fiscal Illusion Index in 1960 and in 2006 on the horizontal axis, and on the vertical axis we plot the estimate of the index and the associated 90% confidence intervals. These intervals indicate the range in which it is 90 percent likely that the true score falls. As observed, the ranges are very short, indicating reasonable estimates. Composite indicators, like the Fiscal Illusion Index, often measure concepts that are linked to well-known and measurable phenomena or to other indexes. These links can be used to test the explanatory power of a composite. Simple cross-plots are often the best way to illustrate such links. Figure A.2 illustrates this aspect. There, we can confirm that higher GDP per capita, government transparency and good governance practices are negatively associated to Fiscal Illusion, while a higher international risk is positively associated to Fiscal Illusion. Given these achievements obtained by a cautioned way, we can expect that Fiscal Illusion is not an irrelevant phenomenon in the political and economic dimensions, namely in the political budget cycle of the democracies. Now, it is the moment to go further into a deeper research in order to investigate these suspicions. Next section will evidence the steps associated to this intention. 4. Estimating Political Budget Cycles in Democracies with Fiscal Illusion 4.1 – Data and empirical model The empirical tests presented here draw on a panel of annual data from 68 countries for the period 1960-2006. We basically use the Brender and Drazen (2004) dataset expanded by recurring to the most recent data versions provided by the International Financial Statistics

(2006), World Development Indicators (2006) and the institution ElectionGuide.org. The (log) variables and their sources are shown in Table 4.1. TABLE 4.1 – Variables, sources and identification

Variables Sources Identification

lbalance B&D(2004)+IFS(2006)

Balance, Central Government Surplus [log]

lmbalance Own work

Difference between the years’s balance and the national average balance [log]

lmbelect Own work Difference between the

7 This analysis was supported by SimLab 1.1, Software for Sensitivity and Uncertainty Analysis, commented in Giglioli and Saltelli (2000).

14

years’s balance and the national average balance of an electoral year [log]

lgdp_pc B&D(2004)+IFS(2006) Per-capita GDP [log]

lgdp_rhp B&D(2004)+Own work

Difference between real GDP and its (country specific) trend, estimated using a Hodrick-Prescott Filter [log]

lpop65 B&D(2004)+WDI (2006)

Fraction of the population over age 65 [log]

lpop1564 B&D(2004)+WDI (2006)

Fraction of the population between ages 15 and 64 [log]

ltrd B&D(2004)+IFS(2006) Ratio of international trade to GDP [log]

elect1l www.electionguide.org

Year before national (parliamentary or presidential) elections [dummy]

elect www.electionguide.org

Year of (parliamentary or presidential) national elections [dummy]

elect1 www.electionguide.org

Year after national (parliamentary or presidential) elections [dummy]

lindexfi Own work Fiscal Illusion Índex [log]

Legend – B&D (2004): Brender and Drazen (2004); IFS (2006): International Financial Statistics (2006); WDI (2006): World Development Indicators (2006) These data include the usual variable “lbalance”, the log of central government surplus, to explore national political budget cycles. But they also include another two variables that will be used as alternative dependent ones: “lmbalance”, the log of the difference between the years’s balance and the national average balance, and “lmbelect”, the log of the difference between the years’s balance and the national average balance of electoral years8. It is expected that these last two variables introduce a suggestion of the amplitude of budget cycles, following similar proposals from Shi and Svensson (2002). As control variables, the database follows those used by Persson and Tabellini (2002) and the subsidiary works, and this research also includes Per-capita GDP, the Difference between real GDP and its (country specific) trend, estimated using a Hodrick-Prescott Filter, the Fractions of the population over age 65 or between ages 15 and 64, and the Ratio of international trade to GDP. The key independent variables are binary election variables. We respectively code “Elect1l”, “Elect” and “Elect1” the Years before, of and after (parliamentary or presidential) elections. Finally, it is also used as explanatory variable “lindexfi”, the logs of the Fiscal Illusion Index, commented in Section 3.

8 As a normalization process of these variables containing negative values, and to keep country-specific differences, the final log values are the logs of an intermediary variable that is equal to each raw variable plus the minimum value considering all the observations from the variable.

15

Table A.3 provides descriptive statistics for the variables included in the analysis. The full data set includes 2227 country-year observations at its maximum length, considering the fiscal variable as dependent ones. For some countries not all years have complete data, resulting in a sample varying from 2182 to 2977 observations, depending on the variable. Because we are interested in studying cross country variations in political budget cycles, we partition the sample into two sub-samples: developing/developed countries and new/old democracies. Following Brender and Drazen (2004), a variable “developed” was created and it takes the value 1 for developed countries and 0 otherwise. Developed countries include the OECD Economies that were members of the organization during the entire sample period, plus Spain, Portugal, Greece and Turkey, examples of “new” democracies. 24 countries in our sample belong to this group and the other 44 are classified as developing countries. Old democracies include the established democracies (that is, all countries which were in a sample of democracies using the POLITY filter, excluding the new democracies. In our sample, 32 countries were considered as “old” democracies and the other 36 as “new” ones. Table A.4 identifies this selection, revealing the associated values from the binary variables “developed” and “old”. Following the theoretical framework and the previous empirical literature on electoral cycles in fiscal policy, a relation between a given fiscal variable y in country i and year t (yi,t) and the electoral cycle can be described as follows:

{ }1,0,1

,,,,,,1 1

,,,,,,

−∈

++++++= −−

= =

−−∑ ∑

k

ELECTFIFIELECTXyy tiiktitiktiti

k

j

m

j

ktikitijjjtijti εµηλδγβ

(Eq. 4.1) where ELECTi,t-k is a dummy election variable that, depending on the value for k, is associated to a pre-electoral year (k=1), to an electoral year (k=0) or to a post-electoral year (k=-1), Xi,t is a vector of m controls, FIi,t is a proxy variable for fiscal illusion conditioning the electoral policy manipulations, µi is a specific country effect, and the term εi,t is a random error that is assumed i.i.d. This specification represents a dynamic panel model, where the dependent variable is a function of its own lagged levels, a set of controls and the electoral timing conditioned by the presence of country-specific levels of fiscal illusion. Using country fixed effects (FE) in an OLS regression with lagged dependent variables introduces a potential estimation bias that is of order 1/T, where T is the length of the panel. (See, for example, Nickell [1981] or Wooldridge [2002].) The bias arises because the initial condition yi,0 is correlated with the country fixed effect µi , so that the lagged dependent variable is correlated with the error term. This problem is thought to be especially severe in micro panel data, where the number of individuals i is large, while T is quite small, often less than half-a-dozen. Since the potential bias of the fixed effects estimator is of order 1/T, the magnitude of the bias in our estimates reported below depends on which sample and fiscal indicator we use. In a panel of all democracies from 1960-2006, the average length of the sample is 26 years in the whole sample, 36 years in the developed countries subsample, and 24 years for less-developed countries subsample. (Remember that some countries do not have data for the entire period.) The average length of the time series in our panel of “old” democracies is longer – 37 years, with few countries having a time series shorter than the

16

maximum. Hence, the bias from using a fixed effects estimator in these regressions is likely to be small. To address the problem that potential bias may be greater in the panels of elections in “new democracies” or “developing counties”, we also present Generalized Method of Moments estimations (GMM), using the Arellano-Bond procedure. In this case, the regressions were provided with robust standard errors, with finite-sample correction for the two-step covariance matrix developed by Windmeijer (2005). GMM results allow to compare FE estimations and to discuss the consistency of the values, as Hsiao (2005) states when he writes “if an estimator is consistent in the fixed-T, large–N case, it will remain consistent if both N and T tend to infinity, irrespective of how they do so (p.296)”. In GMM estimates, and as specification checks, we compute Hansen’s J-statistic for the test for overidentifying restrictions, an extension of the Sargan-test statistic to include a robust error structure. We test for second order serial correlation in the error terms which, if present, would render the estimator inconsistent. 4.2 – Empirical evidence Firstly, the results presented in this section strongly confirm the empirical predictions of political budget cycle theory, though there is a negative effect in the budget surplus in electoral years and a positive effect in the following year. This evidence is reported in the columns from Table 4.2.1, following the conclusions of many authors like Alesina et al.

(1997), Beck (1987), Drazen (2001) or Alt and Lassen (2005). This basic achievement can be deepened considering the substantial differences stated by the previous studies, depending on grouping the countries according to their economic development level or according to the maturity of their democracies. Table 4.2.2 and Table 4.2.3 show the evaluation made of the basic political budget cycle distinguishing the countries. Therefore, and again, the previous expectations are confirmed: “new” democracies exhibit a more significant pattern of political budget cycle than “old” ones. Interestingly and again, the achieved evidence for the dependent variable “lbalance” does not return significant values to the dummy variable related to the pre-electoral year (“elect1l”) but it does return significant values to the electoral year (“elect”), independently of the chosen estimation method (see Table 4.2.2), with the exception of “old” democracies estimated by GMM. Particularly observing the variables “lmbalance” and “lmbelect” in GMM regressions from Tables 4.2.2 and 4.2.3, we confirm that fiscal cycle deviations caused by electoral proximity are less significant in “old” democracies. The results obtained from the partition in developing/developed countries are rather similar to those achieved from the partition in new/old democracies. Developing countries exhibit a budget cycle more affected by electoral moments than developed countries, being the significant values concentrated on the regressions with “lbalance” as dependent variable, indicating that the set of available explanatory variables may fail to capture the dynamics of the other dependent variables “lmbalance” and “lmbelect”. These findings are in line to other previous conclusions, like those expressed by Shi and Svensson (2002), who demonstrated that political budget cycles are much larger in developing countries than in developed countries.

17

According to the previous sections 2 and 3, the Fiscal Illusion phenomenon may substantially distort the expected outcome in the political and economic decisions. Therefore, it is reasonable to expect that different levels of Fiscal Illusion can produce significant changes in the political budget cycle pattern, since the incentives for such manipulation are greater when elections are more meaningful and the incumbent has a positive probability of losing office. This caveat is of particular importance in applications of political budget cycle theory to developing countries or “new” democracies (with higher values of Fiscal Illusion, confirm in Tables A.2 and A.4) where electoral competitiveness is not to be taken for granted and the maturity of the democratic institutions is not yet consolidated. Looking at the raw cross-country data (Table A.3), the average Fiscal Illusion Index score differs between the four subsamples, inducing different political budget cycle reactions depending on the group in which each country is included. As shown in the first two columns from Table 4.2.6, the impact of fiscal illusion index on political budget cycles is significant. Considering all countries, a higher FI value promotes a decreasing in the government surplus: an 1% raise of the associated value induces a reduction of the government surplus (as a proportion of GDP) around 0,12 percentual points (0,09 percentual points according to the FE estimations). This result, following our prior expectations, proves that Fiscal Illusion is a source of larger government budget deficits, as Twight (1994) and Alesina and Perotti (1996) clearly enunciated. Related to the variables “lmbalance” and “lmbelect”, we find again evidence that larger fiscal illusion levels cause more negative deviations from the national balance values to their sample’s average or to their electoral year’s average, suggesting more pronounced cycles. The coefficients of the interaction terms9 (“elect1l_lindexfi”, “elect_lindexfi”, or “elect1_lindexfi”) measure how the electoral effect on the fiscal budget varies among countries with different Fiscal Illusion levels. By an interesting lecture, in the regressions obtained with all countries (Table 4.2.6), there is evidence on this interaction in the dependent variables “lmbalance” and “lmbelect” in pre-electoral years. This fact suggests an important achievement – Fiscal Illusion is (mainly but not exclusively) related to fiscal manipulation by incumbent agents in order to be re-elected, combining the expected results from the original suggestion of Puviani (1903) with those suggested by the proposals of the “Moral Hazard School10” of Rogoff and Sibert (1988), Rogoff (1990) or Shi and Svensson (2002), among others. Some of the coefficients related to the electoral dummies lost their prior statistical significance, suggesting that these binary variables are statistically encompassed when the model incorporates other kind of variables for measuring the cyclical behaviour of the agents and of the institutions. Considering now the differences among old and new democracies (Tables 4.2.7 and 4.2.8), it is observed that the Fiscal Illusion variable has a greater relevance on the latter cases in all the three fiscal variables (“lbalance”, lmbalance” and “lmbelect”), independently of the proximity to elections, functioning as an institutional condition. In the former cases, we mainly observe significant results in the variables related to the amplitude of the cycle, “lmbalance” and “lmbelect”, indicating that the Fiscal Illusion phenomenon exist in the old democracies but it

9 These interaction terms are equal to the products of the dummy related to a pre- , post- or electoral year with the value obtained from the Fiscal Illusion Index. 10 The term “Moral Hazard” was caught from the selection of the approaches on the political budget cycle thematic suggested by Shi and Svensson (2002).

18

only conditions the amplitude of the budget cycle, not the magnitude of the budget itself (mainly suggested by “lbalance”). Close results were obtained with the countries grouped according to their development level (Tables 4.2.9 and 4.2.10). Government budgets from developed countries show less significant impacts from higher values of the Fiscal Illusion Index, while the public finances of the developing nations are more reactive to their Fiscal Illusion levels, a case that is especially visible in the regressions with “lmbalance” and “lmbelect” as dependent variables. The differences between the two selections (old/new democracies and developed/developing countries) can be explained by the not entirely correspondence between the sets: there are new democracies that are developed countries and old democracies that are developing countries (please confirm Table A.4). This fact also results in the different number of significant coefficients when considering the whole sample or when sub-sampling the observations. Additionally, the GMM estimation, with their robust results on small-T and large-N samples, may promote some substantial differences to the FE estimation (preferred) when T is rather large (as in this case). More enlightening details are provided by Arellano (2003), Brender and Drazen (2004) or Hsiao (2005). Synthesizing the findings, this section showed that Fiscal Illusion is an important issue in the political business cycle thematic. While an old suggestion, primarily suggested in 1903, it is a phenomenon that persists in the democratic countries, influencing their economies, mainly their fiscal aggregates. In this section, it was observed that Fiscal Illusion mainly contributes to explain the differences among countries relatively to the magnitude of their government budget surplus and to explain the amplitude of their budget cycles. It was observed that countries with higher values of the Fiscal Illusion Index (discussed in the Section 3) show worse budget deficits and their budget cycle is more negatively pronounced (more negative differences to the average national budget deficit). The differences among the groups of countries mainly reveal the expected results: the Fiscal Illusion phenomenon is more influent in new democracies and in developing countries, deteriorating their government accounts. 5. Conclusion

The most interesting result of this paper is that the more recent models of political budget cycle gain if they consider the Fiscal Illusion phenomenon. We observe that the present “third generation of models”, like those developed by Persson and Tabellini (2000) or Shi and Svensson (2002), are consistent with the overall pattern of results for the panel of studied countries. Therefore, it was developed an empirical investigation based in these models. These models were augmented to account for for fiscal illusion. Fiscal Illusion appearing in the traditional dimensions, like bad governance practices, fiscal manipulation or opacity on public decisions but also Fiscal Illusion appearing in the media, which constitutes a new signal. The main findings of this paper can be summarized as follows. The phenomenon of Fiscal Illusion is rather complex. It is complex because nowadays there is a large set of authors who contributed to its study with different senses, it is complex because it is referred to a wide range of economic realities and, finally, its complexity is also derived upon the methodological use that is given to Fiscal Illusion itself.

19

After constructing a Fiscal Illusion Index, available for 68 democracies, we have confirmed that higher GDP per capita, government transparency and good governance practices are negatively associated to Fiscal Illusion, while a higher international risk is positively associated to Fiscal Illusion. We also have showed that Fiscal Illusion is a relevant subject in the political business cycle thematic. Despite the fact that it is an old idea, primarily suggested in 1903, Fiscal Illusion is a phenomenon that perseveres in the democratic countries, conditioning their economies, mainly their fiscal aggregates. In the section 4, it was observed that Fiscal Illusion mainly helps to explain the differences among countries relatively to the magnitude of their government budget surplus and to explain the amplitude of their budget cycles. It was observed that countries with higher values of the Fiscal Illusion Index (discussed in the Section 3) show worse budget deficits and their budget cycle is more negatively pronounced (more significant negative differences to the average national budget deficit). The differences among the groups of countries mainly reveal the expected results: the Fiscal Illusion phenomenon is more influent in new democracies and in developing countries, deteriorating government accounts. A fruitful direction for future research would be to consider the determinants of fiscal illusion. One of the limitations of this study was the difficulty in establishing causality due to lack of appropriate instrumental variables. Alt and Lassen (2006) suggest political competition as an instrument for fiscal illusion, based on the observation that political parties prefer more transparency if there is a high probability that they might lose office in the next term. Other factors that may affect fiscal illusion could include level of democracy, existence of coalition governments, colonial history, and legal origins. Another proposed direction of future research is to consider a time-series dimension to the fiscal illusion index. This could provide some evidence on effects of changes in Fiscal Illusion and help pin down the direction of causality.

20

References

Alesina, A. and R. Perotti (1996); “Fiscal Discipline and the Budget Process”; American Economic Review

Papers and Proceedings; 86; 401-407 Alesina, A., Haufman, R., Hommes, R. and E. Stein (1996); “Budget institutions and Fiscal performance in Latin

America”; NBER Working Papers Serie; Working Paper 5586 Alesina, A., Roubini, N. and G. Cohen (1997); Political Cycles and the Macroeconomy; MIT Press; Cambridge,

MA Alt, J. and D. Lassen (2005); “The Political Budget Cycle is Where You Can’t See It: Transparency and Fiscal

Manipulation”; manuscript; Economic Policy Research Unit and Department of Economics; University of Copenhagen

Alt, J. and D. Lassen (2006); “Fiscal Transparency, Political Parties, and Debt In OECD Countries”; European

Economic Review; 50, 6; 1403-1439 Andrikopoulos, A., Loizides, J. and K. Prodromidis (2004); “Fiscal policy and political business cycles in the

EU”; European Journal of Political Economy; 20; 125–152. Arellano, M. (2003); Panel Data Econometrics; Oxford University Press; Oxford Barro, R. and J. Lee (1997); “Schooling Quality in a Cross-section of countries”; NBER Working Papers Series;

Working Paper 6198 Barro, R. and J. Lee (2000); “International Data on Educational Attainment: Updates and Implications”;

manuscript; Harvard University Beck, N. (1987); “Elections and the Fed: Is There a Political Monetary Cycle?”; American Journal of Political

Science; 31;194-216 Becker, G. (1983); “A Theory of Competition Among Pressure Groups for Political Influence”; Quarterly

Journal of Economics; 47; 3; 371-400 Brender, A. and A. Drazen (2004); “Political Budget Cycles in New versus Established Democracies”; NBER

Working Papers Series; Working Paper 10539 Buchanan, J. (1960); Fiscal theory and Political Economy; The University of North Carolina Press; Chapel Hill;

pp. 59-64 Buchanan, J. (1967); Public Finance in democratic process: fiscal institutions and individual choice; The

University of North Carolina Press; Chapel Hill Calderón, C., Easterly, W. and L. Servén (2003); “Infrastructure Compression and Public Sector Solvency in

Latin America”; in Easterly, W. and L. Servén, eds. (2003); The Limits of Stabilization: Infrastructure,

Public Deficits, and Growth in Latin America; Stanford University Press and the World Bank CNTSDA (2006); Cross-National Time Series Data Bank, 1815-1999; Databanks International; Jerusalem Cohen, R. and M. Percoco (2004); “The Fiscal Implications of Infrastructure Development”; Actas da

Conferencia La inversion en infraestructura en América Latina y el Caribe: Perspectivas y problemas

de segunda generación; Washington Downs, A. (1957); An economic theory of democracy; Harper Collins; New York Downs, A. (1960); “Why the Government is “too small” in a democracy”; World Politics; 12; 4; 541-563 DPI (2004); Database of Political Institutions; World Bank; Washington Drazen, A. (2001); “Laying Low during Elections: Political Pressure and Monetary Accommodation”;

Manuscript; University of Maryland Easterly, W. (1999); “When is fiscal adjustment an illusion?”; Policy Research Working Paper Series number

2109; The World Bank Easterly, W. (2001); “Growth Implosions, Debt Explosions, and My Aunt Marilyn: Do Growth Slowdowns

Cause Public Debt Crises?”; World Bank Policy Research Working Paper 2531; The World Bank Fasiani, M. (1941); Principi di Scienza delle Finanze; Vol. I; Torino Forte, F. (2004); “Fiscal and Monetary Illusion and the Maastricht rules”; XVI Conference SIEP – Societá

Italiana di Economia Pubblica; Pavia Galbraith, K. (1958); The affluent society; Houghton Mifflin Company; Boston Garcia-Alegre, E. and G. Lopez-Casasnovas (2004); “The effects of poor financial information systems on the

long term sustainability of local public services - Empirical evidence from the Catalan Municipalities.”; UPF Working Paper 746; Universitat Pompeu Fabra

Giglioli, N. and A. Saltelli (2000); “SimLab 1.1, Software for Sensitivity and Uncertainty Analysis, tool for sound modelling”; mimeo; available from http://arxiv.org/abs/cs/0011031

GFS (2006); Government Financial Statistics; International Monetary Fund; New York Hair, J. et al. (2005); Multivariate Data Analysis (6th Edition); Prentice Hall; New York Hameed, F. (2005); “Fiscal Transparency and Economic Outcomes”; IMF Working Paper; WP/05/225 Hsiao, C. (2005); Analysis of Panel Data; Cambridge University Press; Cambridge

21

ICRG (2006); International Country Risk Guide (2006); The Political Risk Services Group; New York IFS (2006); International Financial Statistics; International Monetary Fund; New York ILO (2006); International Labour Organization Statistics; International Labour Organization; Geneva Jensen, F. and N. Vestergaard (1999); “Regulation of Renewable Resources in Federal Systems – the case of

fishery in EU”; Working Papers 3/99; Department of Environmental and Business Economics; University of Southern Denmark

Johnson, R. and D. Wichern (2007); Applied Multivariate Statistical Analysis (6th Edition); Prentice Hall; New York

Jones, P. (2006); “Taking self-interest into account – a Public Choice analysis of international cooperation”; in Kaul, I. And P. Conceiçao (2006); The New Public Finance; Oxford University Press; Oxford; 304-323

Kent, J., Bibby, J, and K. Mardia (2006); Multivariate Analysis (Probability and Mathematical Statistics; Elsevier; New York

Maslow, A. (1970); Motivation and Personality; 2nd Ed.; Harper & Row; New York Mill, J. (1848); Principles of Political Economy; Oxford University Press; Oxford Mink, M. and J. De Haan (2005); “Has the Stability and Growth Pact impeded Political Budget Cycles in the

European Union?”; CESIFO Working Paper Nr. 1532 Morales, J. and G. McMahon, eds. (1996); Economic Policy and the Transition to Democracy - The Latin

American Experience; Macmillan; New York

Mourao, P. (2005); “A importância do desenvolvimento regional na localização das equipas de futebol profissionais. O caso português 1970-1999”; Revista Portuguesa de Estudos Regionais; vol. 8; 31-45

Mourao, P. (2006a); “The Economics of Illusion – a discussion based in Fiscal Illusion”; Journal of Public

Finance and Public Choice; 25; 1 (Forthcoming) Mourao, P. (2006b); Determinantes da Despesa Pública em Portugal; Caixotim; Porto Nardo, M. et al. (2005), "Handbook on Constructing Composite Indicators: Methodology and User Guide";

OECD Statistics Working Papers, 2005/3;OECD Publishing Ndulu, B. and S. O’Connell (1999); “Governance and Growth in Sub-Saharan Africa”; Journal of Economic

Perspective; 13(1); 41-46 Nickell, S. (1981); “Biases in Dynamic Models with Fixed Effects”; Econometrica; 49; 1417-1426 Nicoletti, G., Scarpetta, S. and O. Boylaud (2000); “Summary indicators of product market regulation with an

extension to employment protection legislation”; OECD, Economics department working papers 226 Nordhaus, W. (1975); “The political business cycle”; Review of Economic Studies; 42; 169-190 Persson, T. and G. Tabellini (2000); Political Economics: Explaining Economic Policy; MIT Press; Cambridge,

MA Persson, T. and G. Tabellini (2002); “Do Electoral Cycles Differ Across Political Systems?”; manuscript, IGIER

and Bocconi University Pommerehne, W. and F. Schneider (1978); “Fiscal Illusion Political Institutions and local spending“; Kyklos; 31;

381-408 Puviani, A. (1903); Teoria della illusione finanziaria; Sandron; Palermo Rogoff, K. (1990); “Equilibrium Political Budget Cycles”; American Economic Review ;80; 21-36 Rogoff, K. and A. Sibert (1988); “Elections and Macroeconomic Policy Cycles”; Review of Economic Studies;

55;1-16. Sanandaji, T. and B. Wallace (2003); “Fiscal Illusion and fiscal obfuscation: an empirical study of tax perception

in Sweden”; M. Sc. Thesis in Economics; Stockolm School of Economics; Stockolm Shi, M. and J. Svensson (2002); “Political budget cycles in developed and developing countries”; manuscript;

Institute for International Economic Studies; Stockholm University Shi, M. and J. Svensson (2004); “Political Budget Cycles: A Review of Recent Developments”; mimeo Twight, C. (1994); “Political Transaction-Cost Manipulation: An Integrating Approach”; Journal of Theoretical

Politics; 6; 189 – 216 Von Hagen, J. and I. Harden (1994); “Budget Processes and Commitment to Fiscal Discipline”; European

Economic Review; 39; 771-779 VTs1945 (2002); Voter Turnout since 1945: a global report; International Institute for Democracy and Electoral

Assistance (eds: Rafael López Pintor and Maria Gratschew); Stockolm Wagner, R. (2001); “From the Politics of Illusion to the High Cost of Regulation”; Of public interest; 3 (8) WDI (2006); World Development Indicators; Word Bank; Washington Windmeijer, F. (2005); “A finite sample correction for the variance of linear efficient two-step GMM

estimators”; Journal of Econometrics; 126; 25– 51 Wittman, D. (1995); The myth of democratic failure: why political institutions are efficient; The University of

Chicago Press; Chicago Wooldridge, J (2002); Econometric Analysis of Cross Section and Panel Data; MIT Press; Cambridge, MA

22

Cited Electronic Sites: http://www.electionguide.org/ http://www.idealist.org/ http://www.worldpublicopinion.org/

23

ANNEXES

Table A.1 – Dimensions of Fiscal Illusion, related variables and databases Focused

Dimensions Variable,

[short

denomination]

Expected

effect on

national

Fiscal

Illusion (by a

rising of the

variable)

Source Notes on

missing

values

Herfindahl Index of Public Revenues, [txherfind]

- GFS a)

Percentage of taxes on goods and services in total taxes revenues, [pcgood]

+ GFS a)

Composition of Public Revenues

Ratio between indirect and direct taxes revenues, [ratcurcap]

+ GFS a)

Money creation Growth rate of M2 aggregate, [money]

+ IFS

Public Debt Percentage of Public Debt in the Gross National Income, [gnidebt]

+ WDI

Composition of Public Debt

Percentage of short-term public debt in the national public debt, [shortdebt]

- WDI

Percentage of taxes on transfers, on inheritances and gifts in total taxes revenues, [pcinherita]

+ GFS a) Relevance of certain revenue sources

Percentage of taxes on corporate profits in total taxes revenues, [pcprofit]

+ GFS a)

Political strategies of the ruler group

Average value of radio receptors, tv sets and newspapers per capita, [mediacs]

- CNTSDA

Public expenditures manipulation

Percentage of capital and current transfers in the total expenditures, [pctransfpart]

+ GFS a)

Percentage of education expenditures in the total expenditures, [pceduc]

- GFS a)

Percentage of higher school complete in the total population, [highedu]

- Barro and Lee (2000) b)

Governmental discourse manipulation and electorate believes

Number of governmental checks and balances, [checks]

- DPI c)

Size of cabinets, [cabin_size]

+ CNTSDA Different objectives of governmental agents

Parliamentary power in the Democracy, [pres_parl]

- DPI c)

Immaturity of the democracies

Percentage of invalid votes in parliamentary elections, [inv_vot]

+ VTS1945 c)

24

Composition of Public Capital outlays

Percentage of expenditures on capital transfers in the total expenditures, [pccaptransf]

+ GFS a)

Government confidence (in public polls), [gov_confid]

+ http://www.worldpublicopinion.org c) Governmental rent-seeking

International country risk, [icrg]

- ICRG a)

Openness of the economy, [trd]

- IFS Relevance of trade taxes

Percentage of trade taxes in total taxes revenues, [pcinttrade]

+ GFS a)

Number of nonprofit organizations per million of people, [NPOpmilli]

- http://www.idealist.org Interaction between interest groups and political behaviour

Number of public employees, [pub_employ]

+ ILO

Real Public Budget

Real Public Budget, according to Forte (2004), [realbud]

+ IFS d)

Composition of Public Expenditures

Herfindahl Index of Public Expenditures, [herfdesp]

- GFS a)

Electorate preferences on national issues

Percentage of answers stating ‘economic growth’ as the most important national issue, [higprefer]

+ http://www.worldpublicopinion.org c)

Legend – CNTSDA: Cross-National Time-Series Data Archive (2006); DPI: Database of Political Institutions (2004); GFS: Government Finance Statistics (2006); ICRG: International Country Risk Guide (2006); IFS: International Financial Statistics (2006); ILO: International Labour Organization Statistics (2006); VTS1945: Voter Turnout since 1945 (2002); WDI: World Development Indicators (2006). Notes – a) Due to the scarcity of data in some of the variables provided by the databases, the missing values were substituted by the national average values of the pointed variable, following previous proceedings of Shi and Svensson (2002) or Nardo et al. (2005). b) Barro and Lee (2000) database provides data in each five years since 1960; therefore, the interstitial missing values were substituted by the value from the previous provided year. c) The missing values were substituted by the national average value using the available data. d) According to Forte (2004), the Real Public Budget is equal to Index of Consumer Prices*Debt/GDP + Nominal Deficit.

25

Table A.2 – Fiscal Illusion (FI) Index, 1960 and 2006

Country year FI

ARGENTINA 1960 0,841

ARGENTINA 2006 0,401

AUSTRALIA 1960 0,496

AUSTRALIA 2006 0,27

AUSTRIA 1960 0,481

AUSTRIA 2006 0,088

BELGIUM 1960 0,754

BELGIUM 2006 0,25

BOLIVIA 1960 0,776

BOLIVIA 2006 0,564

BRAZIL 1960 0,551

BRAZIL 2006 0,366

BULGARIA 1960 0,477

BULGARIA 2006 0,283

CANADA 1960 0,439

CANADA 2006 0,116

CHILE 1960 0,848

CHILE 2006 0,327

COLOMBIA 1960 0,875

COLOMBIA 2006 0,598

COSTA_RICA 1960 0,532

COSTA_RICA 2006 0,152

CYPRUS 1960 0,77

CYPRUS 2006 0,474

CZECH_REP 1960 a)

CZECH_REP 2006 a)

DENMARK 1960 0,339

DENMARK 2006 0,121

DOMINICAN 1960 0,845

DOMINICAN 2006 0,311

ECUADOR 1960 0,6

ECUADOR 2006 0,349

EL_SALVADOR 1960 0,979

EL_SALVADOR 2006 0,569

ESTONIA 1960 a)

ESTONIA 2006 a)

FIJI 1960 0,523

FIJI 2006 0,395

FINLAND 1960 0,438

FINLAND 2006 0,07

FRANCE 1960 0,897

FRANCE 2006 0,433

GERMANY 1960 0,496

GERMANY 2006 0,117

GREECE 1960 0,912

GREECE 2006 0,447

GUATEMALA 1960 0,963

GUATEMALA 2006 0,799

HONDURAS 1960 0,913

HONDURAS 2006 0,859

HUNGARY 1960 0,756

HUNGARY 2006 0,551

ICELAND 1960 0,441

ICELAND 2006 0,147

INDIA 1960 0,928

INDIA 2006 0,833

IRELAND 1960 0,58

IRELAND 2006 0,105

ISRAEL 1960 0,492

ISRAEL 2006 0,306

ITALY 1960 0,811

ITALY 2006 0,312

JAPAN 1960 0,619

JAPAN 2006 0,346

KOREA 1960 0,878

KOREA 2006 0,426

LITHUANIA 1960 a)

LITHUANIA 2006 a)

LUXEMBOURG 1960 0,17

LUXEMBOURG 2006 0,177

MADAGASCAR 1960 0,906

MADAGASCAR 2006 0,849

MALAYSIA 1960 0,678

MALAYSIA 2006 0,48

MALI 1960 0,929

MALI 2006 0,943

MAURITIUS 1960 0,817

MAURITIUS 2006 0,649

MEXICO 1960 0,797

MEXICO 2006 0,255

NEPAL 1960 0,55

NEPAL 2006 0,696

NETHERLANDS 1960 0,368

NETHERLANDS 2006 0,062

NEW ZEALAND 1960 0,227

NEW ZEALAND 2006 0,022

NICARAGUA 1960 0,886

NICARAGUA 2006 0,594

NORWAY 1960 0,551

NORWAY 2006 0,145

PAKISTAN 1960 0,967

PAKISTAN 2006 0,935

PANAMA 1960 0,698

PANAMA 2006 0,518

PAPUA 1960 0,588

PAPUA 2006 0,646

PARAGUAY 1960 0,99

PARAGUAY 2006 0,712

PERU 1960 0,935

PERU 2006 0,768

PHILIPINES 1960 0,889

PHILIPINES 2006 0,685

POLAND 1960 0,986

POLAND 2006 0,767

PORTUGAL 1960 0,791

PORTUGAL 2006 0,276

ROMANIA 1960 a)

ROMANIA 2006 a)

RUSSIA 1960 a)

RUSSIA 2006 a)

SLOVAKIA 1960 a)

SLOVAKIA 2006 a)

SLOVENIA 1960 a)

SLOVENIA 2006 a)

SOUTH_AFRICA 1960 0,627

SOUTH_AFRICA 2006 0,396

SPAIN 1960 0,617

SPAIN 2006 0,09

SRI_LANKA 1960 0,992

SRI_LANKA 2006 0,927

SWEDEN 1960 0,37

SWEDEN 2006 0,069

SWITZERLAND 1960 0,242

SWITZERLAND 2006 0,102

TRINIDAD 1960 0,162

TRINIDAD 2006 0,066

TURKEY 1960 0,838

TURKEY 2006 0,401

UK 1960 0,633

UK 2006 0,232

US 1960 0,381

US 2006 0,219

URUGUAY 1960 0,933

URUGUAY 2006 0,76

VENEZUELA 1960 0,796

VENEZUELA 2006 0,487

Notes – a) Values not available in the pointed dates due to data scarcity.

26

Table A.3 - Descriptive statistics

Grouped variables

lbalance lmbalance lmbelect lgdp_pc lgdp_rhp lpop65 lpop1564 ltrd elect1l elect elect1 elect1l_lindexfi

elect_lindexfi

elect1_lindexfi

lindexfi

N 2227 2182 2182 2694 2563 2977 2976 2650 2996 2996 2996 2996 2996 2996 2996

max 3,921 3,808 3,862 10,821 2,623 2,996 4,278 5,437 1,000 1,000 1,000 0,000 0,000 0,000 -0,373

min -7,470 -6,949 -7,083 5,613 -3,288 0,693 3,871 2,078 0,000 0,000 0,000 -1,345 -1,310 -1,358 -1,362

mean 3,259 3,167 3,218 8,375 1,228 1,913 4,102 3,961 0,193 0,196 0,190 -0,140 -0,142 -0,139 -0,719

sd Total 0,274 0,254 0,256 1,080 0,629 0,617 0,099 0,584 0,395 0,397 0,393 0,294 0,296 0,294 0,180

N 975 949 949 1318 1228 1586 1586 1301 1594 1594 1594 1594 1594 1594 1594

max 3,646 3,603 3,716 10,171 2,440 2,865 4,278 5,153 1,000 1,000 1,000 0,000 0,000 0,000 -0,397

min -7,470 -6,949 -7,083 5,613 -3,288 0,833 3,871 2,078 0,000 0,000 0,000 -1,345 -1,310 -1,358 -1,362

mean 3,247 3,164 3,218 7,910 1,214 1,742 4,078 3,836 0,137 0,140 0,136 -0,092 -0,094 -0,092 -0,695

sd old=0 0,370 0,352 0,357 0,967 0,736 0,582 0,108 0,562 0,344 0,347 0,342 0,240 0,242 0,241 0,203

N 1252 1233 1233 1376 1335 1391 1390 1349 1402 1402 1402 1402 1402 1402 1402

max 3,921 3,808 3,862 10,821 2,623 2,996 4,247 5,437 1,000 1,000 1,000 0,000 0,000 0,000 -0,373

min 2,430 2,318 2,410 5,964 -2,576 0,693 3,893 2,079 0,000 0,000 0,000 -1,132 -1,115 -1,116 -1,132

mean 3,269 3,169 3,219 8,820 1,240 2,109 4,130 4,081 0,257 0,259 0,252 -0,194 -0,196 -0,192 -0,747

sd old=1 0,165 0,139 0,135 0,991 0,511 0,596 0,080 0,580 0,437 0,438 0,435 0,337 0,340 0,338 0,144

N 1284 1251 1251 1664 1549 1939 1939 1633 1948 1948 1948 1948 1948 1948 1948

max 3,921 3,808 3,862 10,057 2,623 2,833 4,278 5,437 1,000 1,000 1,000 0,000 0,000 0,000 -0,373

min -7,470 -6,949 -7,083 5,613 -3,288 0,693 3,871 2,079 0,000 0,000 0,000 -1,345 -1,310 -1,358 -1,362

mean 3,250 3,164 3,217 7,887 1,268 1,622 4,067 3,954 0,153 0,156 0,149 -0,101 -0,103 -0,099 -0,683

sd developed=0 0,335 0,316 0,320 0,963 0,714 0,542 0,101 0,578 0,360 0,363 0,357 0,246 0,248 0,246 0,193

N 943 931 931 1030 1014 1038 1037 1017 1048 1048 1048 1048 1048 1048 1048

max 3,763 3,688 3,727 10,821 2,524 2,996 4,247 5,421 1,000 1,000 1,000 0,000 0,000 0,000 -0,493

min 2,255 2,318 2,460 7,279 -0,227 1,262 3,986 2,078 0,000 0,000 0,000 -1,132 -1,115 -1,116 -1,132

mean 3,272 3,170 3,220 9,163 1,167 2,457 4,169 3,973 0,268 0,270 0,266 -0,212 -0,214 -0,212 -0,786

sd developed=1 0,157 0,132 0,126 0,738 0,464 0,294 0,048 0,593 0,443 0,444 0,442 0,356 0,358 0,357 0,128

27

Table A.4 - Countries and values from variables

"Old" and "Developed"

Country Old Developed

ARGENTINA 0 0

AUSTRALIA 1 1

AUSTRIA 1 1

BELGIUM 1 1

BOLIVIA 0 0

BRAZIL 0 0

BULGARIA 0 0

CANADA 1 1

CHILE 0 0

COLOMBIA 1 0

COSTA_RICA 1 0

CYPRUS 1 0

CZECH_REP 0 0

DENMARK 1 1

DOMINICAN 0 0

ECUADOR 0 0

EL_SALVADOR 0 0

ESTONIA 0 0

FIJI 0 0

FINLAND 1 1

FRANCE 1 1

GERMANY 1 1

GREECE 0 1

GUATEMALA 0 0

HONDURAS 0 0

HUNGARY 0 0

ICELAND 1 1

INDIA 1 0

IRELAND 1 1

ISRAEL 1 0

ITALY 1 1

JAPAN 1 1

KOREA 0 0

LITHUANIA 0 0

LUXEMBOURG 1 1

MADAGASCAR 0 0

MALAYSIA 1 0

MALI 0 0

MAURITIUS 1 0

MEXICO 0 0

NEPAL 0 0

NETHERLANDS 1 1

NZ 1 1

NICARAGUA 0 0

NORWAY 1 1

PAKISTAN 0 0

PANAMA 0 0

PAPUA 1 0

PARAGUAY 0 0

PERU 0 0

PHILIPINES 0 0

POLAND 0 0

PORTUGAL 0 1

ROMANIA 0 0

RUSSIA 0 0

SLOVAKIA 0 0

SLOVENIA 0 0

SOUTH_AFRICA 1 0

SPAIN 0 1

SRI_LANKA 1 0

SWEDEN 1 1

SWITZERLAND 1 1

TRINIDAD 1 0

TURKEY 0 1

UK 1 1

US 1 1

URUGUAY 0 0

VENEZUELA 1 0

Note: Coefficient of correlation between “Old” and “Developed” = 0,537.

28

Table 4.2.1 - All Democracies (no Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,004 d -0,001 d -0,006 d -0,005 d -0,005 d -0,004 d

Adjusted R2 0,66 0,45 0,49

F-Statistics* 320,14 268,34 271,68

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,15 0,18 0,16

N (countries) 68 68 68 68 68 68

N (observations) 1966 1860 1978 1817 1978 1817

Elect -0,015 a -0,014 a -0,005 d -0,001 d -0,003 d -0,001 d

Adjusted R2 0,67 0,45 0,49

F-Statistics* 121,60 268,23 271,47

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,30 0,19 0,24

N (countries) 68 68 68 68 68 68

N (observations) 1966 1860 1978 1817 1978 1817

Elect1 0,016 a 0,017 a -0,010 b -0,013 a -0,008 c -0,012 a

Adjusted R2 0,66 0,45 0,49

F-Statistics* 323,73 269,06 272,29

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,19 0,17 0,15

N (countries) 68 68 68 68 68 68

N (observations) 1966 1860 1978 1817 1978 1817

Notes: The covariates include one lag of the dependent variable, the log of per-capita GDP, the ratio of international trade to GDP, the fraction of the population over age 65, the fraction of the population between ages 15 and 64, and the log difference between real GDP and its (country specific) trend, estimated using a Hodrick-Prescott filter. d: significant at an higher level than 10%; c: significant at the 10% level; b: significant at the 5% level; a: significant at the 1% level *:F-test is an F test of overall significance **p-values of the Hansen’s J-statistic that is a test for overidentifying restrictions, an extension of the Sargan-test statistic to include a robust error structure ***p-values of the Serial corr. Test that is a test for second-order serial correlation in the first-difference residuals , asymptotically distributed as N(0,1) under the null of no serial correlation. .

29

Table 4.2.2- New Democracies (no Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,005 d -0,001 d -0,014 d -0,013 d -0,012 d -0,008 d

Adjusted R2 0,50 0,25 0,36

F-Statistics* 79,69 59,51 65,44

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,13 0,33 0,12

N (countries) 36 36 36 36 36 36

N (observations) 847 762 824 739 824 739

Elect -0,025 a -0,023 a -0,010 d -0,007 d -0,007 d -0,006 d

Adjusted R2 0,51 0,25 0,35

F-Statistics* 82,04 59,23 65,08

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,10 0,25 0,11

N (countries) 36 36 36 36 36 36

N (observations) 847 762 824 739 824 739

Elect1 0,028 a 0,017 d -0,022 b -0,017 c -0,014 b -0,016 c

Adjusted R2 0,50 0,26 0,36

F-Statistics* 82,74 60,35 66,18

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,11 0,12 0,11

N (countries) 36 36 36 36 36 36

N (observations) 847 762 824 739 824 739

Notes: see Table 4.2.1

Table 4.2.3 - Old Democracies (no Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,002 d -0,001 d 0,000 d 0,001 d 0,000 d 0,001 d

Adjusted R2 0,73 0,58 0,57

F-Statistics* 258,02 234,54 220,85

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,84 0,90 0,83

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Elect -0,005 d -0,009 b -0,002 d -0,002 d 0,000 d -0,001 d

Adjusted R2 0,73 0,58 0,57

F-Statistics* 258,25 234,58 220,85

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,83 0,91 0,83

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Elect1 0,008 d 0,010 d -0,005 d -0,009 c -0,00d -0,009 c

Adjusted R2 0,73 0,58 0,57

F-Statistics* 258,78 234,74 220,99

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,84 0,85 0,92

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Notes: see Table 4.2.1

30

Table 4.2.4 - Developing countries (no Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,010 d -0,007 d -0,013 c -0,009 d -0,012 d -0,009 d

Adjusted R2 0,58 0,33 0,39

F-Statistics* 100,12 77,05 78,53

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,21 0,60 0,10

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 984 1088 1030

Elect -0,015 b -0,008 d -0,012 d -0,009 d -0,009 d -0,007 d

Adjusted R2 0,58 0,33 0,38

F-Statistics* 100,76 76,97 78,34

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,24 0,13 0,11

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 984 1088 1030

Elect1 0,023 a 0,018 b -0,011 d -0,006 d -0,008 d -0,005 d

Adjusted R2 0,58 0,33 0,39

F-Statistics* 102,14 76,84 78,29

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,24 0,11 0,10

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 984 1088 1030

Notes: see Table 4.2.1

Table 4.2.5- Developed countries (no Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l 0,004 d 0,003 d 0,002 d 0,005 d 0,002 d 0,005 d

Adjusted R2 0,77 0,68 0,68

F-Statistics* 308,81 290,64 288,68

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,96 0,62 0,62

N (countries) 24 24 24 24 24 24

N (observations) 903 845 890 833 890 833

Elect -0,010 c -0,013 b 0,001 d 0,000 d 0,002 d 0,001 d

Adjusted R2 0,77 0,68 0,68

F-Statistics* 310,25 290,60 288,69

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,98 0,65 0,66

N (countries) 24 24 24 24 24 24

N (observations) 903 845 890 833 890 833

Elect1 0,009 c 0,011 c -0,010 c -0,013 c -0,009 c -0,013 c

Adjusted R2 0,77 0,68 0,68

F-Statistics* 309,98 292,16 290,08

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,98 0,67 0,68

N (countries) 24 24 24 24 24 24

N (observations) 903 845 890 833 890 833

Notes: see Table 4.2.1

31

Table 4.2.6 - All Democracies (Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,014 d 0,001 d -0,046 c -0,024 c -0,041 c -0,023 c

Elect1l_lindexfi -0,014 d 0,002 d -0,055 c -0,031 c -0,050 c -0,089 c

lindexfi -0,081 b -0,120 c -0,140 a -0,258 a -0,122 a -0,246 a

Adjusted R2 0,66 0,44 0,47

F-Statistics* 250,05 212,35 214,50

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,21 0,17 0,15

N (countries) 68 68 68 68 68 68

N (observations) 2022 1860 1978 1891 1978 1817

Elect 0,030 d -0,038 c -0,011 d 0,018 d -0,012 d 0,017 d

Elect_lindexfi -0,024 d -0,033 d -0,009 d 0,026 d -0,012 d 0,024 d

lindexfi -0,077 b -0,112 d -0,154 a -0,285 a -0,0134 a -0,273 a

Adjusted R2 0,66 0,44 0,47

F-Statistics* 251,78 211,65 213,78

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,17 0,16 0,14

N (countries) 68 68 68 68 68 68

N (observations) 2022 1860 1978 1891 1978 1817

Elect1 0,054 b 0,056 c -0,028 d -0,044 c -0,021 d -0,039 c

Elect1_lindexfi 0,053 c 0,055 d -0,024 d -0,042 d -0,017 d -0,037 d

lindexfi -0,098 a -0,135 c -0,015 a -0,264 a -0,134 a -0,255 a

Adjusted R2 0,66 0,44 0,47

F-Statistics* 253,58 212,47 214,50

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,18 0,16 0,14

N (countries) 68 68 68 68 68 68

N (observations) 2022 1860 1978 1891 1978 1817

Notes: see Table 4.2.1,

32

Table 4.2.7 - New Democracies (Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,019 d -0,025 d -0,049 d -0,025 d -0,044 d -0,022 d

Elect1l_lindexfi -0,021 d -0,035 d -0,053 d -0,023 d -0,048 d -0,023 d

lindexfi -0,126 c 0,036 d -0,240 a -0,283 b -0,224 a -0,240 a

Adjusted R2 0,49 0,21 0,29

F-Statistics* 62,62 48,49 53,20

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,31 0,18 0,14

N (countries) 36 36 36 36 36 36

N (observations) 847 798 824 776 824 776

Elect -0,038 d -0,041 d -0,013 d 0,014 d -0,012 d 0,000 d

Elect_lindexfi -0,019 d -0,027 d -0,005 d 0,030 d -0,007 d 0,011 d

lindexfi -0,126 c 0,032 d -0,250 a -0,294 a -0,232 a -0,245 b

Adjusted R2 0,50 0,20 0,29

F-Statistics* 64,47 48,12 52,77

Hansen-test** 0,99 0,86 0,99

2nd order test*** 0,25 0,18 0,17

N (countries) 36 36 36 36 36 36

N (observations) 847 798 824 776 824 776

Elect1 0,024 d -0,023 d -0,048 d -0,029 d -0,042 d -0,042 d

Elect1_lindexfi -0,005 d -0,055 d -0,036 d -0,019 d -0,031 d -0,035 d

lindexfi -0,115 c -0,012 d -0,254 a -0,299 a -0,237 a -0,248 b

Adjusted R2 0,50 0,21 0,29

F-Statistics* 64,81 49,31 53,93

Hansen-test** 0,99 0,73 0,99

2nd order test*** 0,28 0,17 0,13

N (countries) 36 36 36 36 36 36

N (observations) 847 798 824 776 824 776

Notes: see Table 4.2.1

Table 4.2.8 - Old Democracies (Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,002 d 0,008 d -0,032 d -0,044 d -0,033 d -0,041 d

Elect1l_lindexfi 0,000 d 0,013 d -0,043 d -0,059 d -0,043 d -0,055 d

lindexfi -0,050 d -0,082 d -0,151 a -0,203 b -0,130 b -0,188 b

Adjusted R2 0,73 0,58 0,56

F-Statistics* 200,76 185,09 173,92

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,84 0,82 0,75

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Elect -0,003 d -0,018 d -0,003 d 0,007 d -0,007 d 0,005 d

Elect_lindexfi 0,002 d -0,013 d -0,001 d 0,011 d -0,009 d 0,008 d

lindexfi -0,050 d -0,088 d -0,163 a -0,225 b -0,141 a -0,207 b

Adjusted R2 0,61 0,58 0,56

F-Statistics* 200,85 184,83 173,63

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,83 0,55 0,50

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Elect1 0,061 b 0,056 d 0,002 d -0,024 d 0,006 d -0,022 d

Elect1_lindexfi 0,070 c 0,059 d 0,008 d -0,020 d 0,013 d -0,018 d

lindexfi -0,074 d -0,095 d -0,166 a -0,216 b -0,147 a -0,200 b

Adjusted R2 0,74 0,58 0,56

F-Statistics* 202,23 184,96 173,74

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,71 0,56 0,51

N (countries) 32 32 32 32 32 32

N (observations) 1175 1098 1154 1078 1154 1078

Notes: see Table 4.2.1

33

Table 4.2.9 - Developing countries (Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l -0,034 d -0,014 d -0,039 d -0,013 d -0,036 d -0,013 d

Elect1l_lindexfi -0,037 d -0,005 d -0,040 d -0,008 d -0,039 d -0,009 d

lindexfi -0,106 c -0,072 d -0,325 a -0,398 a -0,311 a -0,385 b

Adjusted R2 0,57 0,26 0,30

F-Statistics* 78,46 63,98 65,27

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,27 0,12 0,11

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 1030 1088 1030

Elect -0,031 d -0,029 d -0,034 d -0,005 d -0,037 d -0,008 d

Elect_lindexfi -0,024 d -0,031 d -0,034 d 0,004 d -0,044 d -0,003 d

lindexfi -0,109 c -0,071 d -0,328 a -0,405 a -0,312 a -0,390 a

Adjusted R2 0,57 0,26 0,30

F-Statistics* 78,91 63,94 65,19

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,24 0,14 0,12

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 1030 1088 1030

Elect1 0,028 d 0,002 d -0,041 d -0,036 d -0,036 d -0,030 d

Elect1_lindexfi 0,009 d -0,024 d -0,042 d -0,038 d -0,038 d -0,031 d

lindexfi -0,104 c -0,060 d -0,332 a -0,405 b -0,318 a -0,392 b

Adjusted R2 0,58 0,26 0,29

F-Statistics* 79,82 64,06 65,28

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,24 0,12 0,11

N (countries) 44 44 44 44 44 44

N (observations) 1119 1060 1088 1030 1088 1030

Notes: see Table 4.2.1

Table 4.2.10 - Developed countries (Fiscal Illusion as covariate)

1960-2006

lbalance lmbalance lmbelect

FE GMM FE GMM FE GMM

Elect1l 0,069 b 0,093 d -0,036 d -0,029 d -0,033 d -0,024 d

Elect1l_lindexfi 0,082 c 0,111 d -0,047 d -0,040 d -0,045 d -0,034 d

lindexfi -0,115 b -0,177 b -0,109 c -0,119 d -0,079 d -0,114 d

Adjusted R2 0,78 0,69 0,69

F-Statistics* 242,25 227,82 225,59

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,96 0,81 0,85

N (countries) 24 24 24 24 24 24

N (observations) 903 874 890 861 890 861

Elect -0,027 d -0,013 d 0,063 c 0,079 d 0,058 d 0,075 d

Elect_lindexfi -0,022 d -0,002 d 0,077 c 0,095 d 0,070 d 0,090 d

lindexfi -0,085 d -0,145 c -0,146 a -0,160 b -0,113 b -0,151 b

Adjusted R2 0,78 0,69 0,69

F-Statistics* 242,13 228,44 226,15

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,78 0,93 0,98

N (countries) 24 24 24 24 24 24

N (observations) 903 874 890 861 890 861

Elect1 0,066 c 0,014 d -0,034 d -0,032 d -0,025 d -0,030 d

Elect1_lindexfi 0,072 c 0,013 d -0,030 d -0,026 d -0,021 d -0,024 d

lindexfi -0,117 b -0,153 b -0,112 b -0,120 d -0,085 d -0,113 d

Adjusted R2 0,78 0,69 0,69

F-Statistics* 242,94 228,74 226,34

Hansen-test** 0,99 0,99 0,99

2nd order test*** 0,76 0,84 0,89

N (countries) 24 24 24 24 24 24

N (observations) 903 874 890 861 890 861

Notes: see Table 4.2.1

Figure A.1: Margins of Error for Fiscal Illusion Index, 1960 and 2006

Fiscal Illusion Index, 1960

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

percentile rank

Fiscal Illusion Index, 2006

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

percentile rank

Figure A.2 – Links between Fiscal Illusion Index values and GDP per capita, Country risk, Corruption Perception and Governance Indicators (year: 2000)

GDP per capita and Fiscal Illuson

R2 = 0,5598

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 2 4 6 8 10 12

GDP per capita (log scale)

Fis

cal

Illu

sio

n

Country risk and Fiscal IllusionR2 = 0,402

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 0,5 1 1,5 2 2,5

Country risk (source: ICRG)

Fis

cal

Illu

sio

n

Corruption Perception and Fiscal Illusion

R2 = 0,328

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 2 4 6 8 10 12

Corruption Perception Index (source: Transparency International)

Fis

cal

Illu

sio

n

Governance Indicators and Fiscal Illusion

R2 = 0,5187

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

-20 -15 -10 -5 0 5 10 15 20 25 30

Average of Aggregate Governance Indicators (source: World Bank)

Fis

cal

Illu

sio

n