Lecture Notes in Economics and Mathematical Systems Founding Editors: M. Beckmann H. P. Kiinzi
Managing Editors: Prof...
50 downloads
804 Views
9MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Lecture Notes in Economics and Mathematical Systems Founding Editors: M. Beckmann H. P. Kiinzi
Managing Editors: Prof. Dr. G. Fandel Fachbereich Wirtschaftswissenschaften Fernuniversitat Hagen Feithstr. 140/AVZ 11, 58084 Hagen, Germany Prof. Dr. W. Trockel Institut fur Mathematische Wirtschaftsforschung (IMW) Universitat Bielefeld Universitatsstr. 25, 336 15 Bielefeld, Germany Editorial Board: A. Basile, A. Drexl, H. Dawid, K. Inderfurth, W. Kursten, U. Schittko
Jaejoon Woo
The Political Economv of Fiscal Policy Public Deficits, Volatility, and Growth
Q - Springer
Author Professor Dr. Jaejoon Woo Department of Economics Kellstadt Graduate School of Business DePaul University 1 East Jackson Boulevard Chicago, IL 60604 USA e-mail: jwool @ depaul.edu
ISSN 0075-8442 ISBN-10 3-540-29640-9 Springer Berlin Heidelberg New York ISBN-13 978-3-540-29640-9 Springer Berlin Heidelberg New York This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media
O Springer-Verlag Berlin Heidelberg 2006 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Camera ready by author Cover design: Erich Kirchner, Heidelberg Printed on acid-free paper
4213153DK
5432 10
To Hyejung
Preface
Political economy is now one of the most active research fields in macroeconomics. My interest in the political economy and many of the ideas that are presented in this monograph were initiated during my PhD economics study at Harvard University. Alberto Alesina introduced me to the political economy literature in his graduate course, "Advanced Macroeconomic Policy." I am indebted to Alberto Alesina, Dani Rodrik, and Aaron Tornell for their insightful discussions and comments. They greatly helped my thinking on the subject. At various stages of writing the chapters (papers), I have also benefited from discussions with or comments from Daron Acemoglu, Gadi Barlevy, John Berdell, Lans Bovenberg, Daniel Cohen, Jorge Braga de Marcedo, Jason Furman, Steinar Holden, Philip Lane, Jongwha Lee, N. Greg Mankiw, Jordan Rappaport, Jeffrey Sachs, Jose Tavares, and Jürgen von Hagen. I also thank seminar/conference participants at Harvard University, State University of New York at Buffalo, ZEI at University of Bonn, the OECD Development Center, DELTA-ENS at Paris, DePaul University, University of Wisconsin at Milwaukee, Loyola University of Chicago, Western Michigan University, University of Calgary, University of Illinois at Chicago, University of Hawaii at Manoa, the European Economic Association Meeting in 2001, the North American Econometric Society Summer Meeting in 2002, the European Econometric Society Meeting in 2003, the International conference of Korea Economic Association and Korea-America Economic Association in 2004, the Southern Economic Association Meeting in 2004, and the Workshop on Fiscal Policy Issues at University of Oslo in 2005 for helpful discussions. A part of this book was written and revised while I was an economist at the Organisation for Economic Cooperation and Development (OECD) in Paris. At the OECD, I also benefited from discussions with colleagues, and learned enormously about the practice in "political economy" of policy debates. I also thank chairman Michael Miller and colleagues at the department for their encouragement and support. I am grateful to editor Katharina Wetzel-Vandai at Springer-Verlag for her encouragement and patience throughout this book
VIII
Preface
project. Mike Carlton and Rebecca Azaola provided excellent research assistance. Sean Moffat provided superb assistance with editing the manuscript. I gratefully acknowledge financial support from the Korea Foundation for Advanced Studies and the Kellstadt Business School at DePaul University for some of the work included here, and from the University Research Council at DePaul University (via Competitive Research Grant, No. 600201) for this book project. I always thank my parents for what I am today. I owe an enormous debt to my wife, Hyejung, who has been a good advisor and critic, and has created a suitable working environment. And to Elizabeth and Alexandra, who always make me laugh.
Chicago October 2005
Jaejoon Woo
Contents
1
Introduction
2
Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization 2.1 Introduction 2.2 Endogenous Growth Model with Optimizing Interest Groups . . 2.3 Endogenous Fiscal Deficit and Volatility: Polarization 2.3.1 Non-cooperative Feedback Nash Equilibrium 2.3.2 Fiscal PoUcy with a Social Planner 2.3.3 Policymakers Faced with Political Uncertainty 2.4 Endogenous Fiscal Deficit and Growth 2.5 Concluding Remarks 2.6 Appendix
3
4
Inflation, Composition of Deflcit Finance, and Social Polarization 3.1 Introduction 3.2 The Economy 3.3 Inflation, Composition of Deficit Finance, and Social Polarization 3.4 Concluding Remarks Social Polarization, Industrialization, and Fiscal Instability 4.1 Introduction 4.2 The Economy 4.2.1 The Pre-Industrialization Regime 4.2.2 The Post-Industrialization Regime 4.2.3 Human Capital Accumulation, Threshold Externality, and Income Distribution 4.2.4 Income Inequality and Polarization 4.3 The Fiscal Policy
1 7 7 13 17 18 23 24 27 31 33 35 35 39 40 45 47 47 51 51 52 54 57 58
X
Contents
4.4 4.5 4.6 4.7 5
6
4.3.1 Polarization and Endogenous Fiscal Deficit 4.3.2 Polarization and Volatility of Fiscal Outcomes 4.3.3 Country Experiences Econometric Evidence Concluding Remarks Appendix A. The Social Planner's Solution Appendix B. Data
Economic, Political, and Institutional Determinants of Public Deficits 5.1 Introduction 5.2 Economic Variables on Public Deficits 5.2.1 The Benchmark Framework 5.3 Fiscal Politics 5.3.1 Political Instability 5.3.2 Government Fragmentation 5.3.3 Political Regime and Electoral Law 5.4 Social Polarization: Income Inequality and Ethnic Divisions . . . 5.5 Institutions 5.6 Robustness and Sensitivity Analysis 5.7 Concluding Remarks 5.8 Data Appendix Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence 6.1 Introduction 6.2 Inequality and Growth: Cross-Country Regression 6.3 Income Inequality and Fiscal Volatility 6.4 Fiscal Volatility and Growth 6.5 Comparison with Sociopolitical Instability and Fertility Decision 6.6 Concluding Remarks 6.7 Data Appendix
58 60 61 64 73 74 75 77 77 80 80 84 84 87 89 92 96 100 103 107 117 117 119 129 135 140 144 146
References
151
Index
161
Introduction
In discussing issues of macroeconomic policy, the traditional approach used to assume a benevolent social planner who maximizes some social welfare function and to derive optimal policies. However, recurrent macroeconomic problems, which often manifested themselves in unsustainable macroeconomic imbalances (domestic and external) observed in many countries in the past decades, are at odds with the traditional optimal policy prescription. Indeed, it is hard to believe that the government is a monolithic entity that seeks to implement policies to achieve a socially optimal outcome. It has been increasingly recognized that the policymakers have their own objective function and agenda, and strategically interact with other current and/or future policymakers and with the pubUc. They may care about remaining in office or represent the interests of particular groups. At the same time, socio-political factors and the associated incentive constraints facing policymakers greatly influence policy decisions. This monograph presents new, theoretical and empirical studies on fiscal and macroeconomic issues, while highlighting social polarization as an essential organizing principle in a political economy approach. One of the most striking macroeconomic developments during the last three decades is the rise and persistence of large fiscal deficits in a number of countries. The rising public debts have been the subject of great concern to both policymakers and researchers since it has often been an important source of macroeconomic instability or even economic crises such as external debt crisis and hyperinflation. Accordingly, fiscal discipline and a strong fiscal position have been recognized as a necessary condition for sustaining economic growth. Many developing countries have embarked on major fiscal reforms in the wake of external debt crises. Also, some industrialized countries-for example, Denmark, Ireland, New Zealand, and Sweden-have adopted institutional changes that strengthened fiscal rules and procedures in order to achieve greater fiscal discipline in the 1980s and 1990s. In euro area, fiscal consolidation efforts culminated in the run up to monetary union as countries needed to comply with the convergence criteria stipulated in the Maastricht Treaty.
2
1 Introduction
Yet many countries suffer from recurrent large fiscal imbalances that often reflect lack of fiscal discipline. Recent financial crises in Brazil in 1999 and Argentina in 2001 were closely related to high budget deficits or unsustainability of fiscal outcomes.^ With the advent of the euro, the deficit bias has reemerged in euro area, though not in all countries. The 3% ceiling for the deficit-to-GDP ratio has been breached by a number of euro area countries (OECD 2005a).^ Some recent study even suggests that such institutional arrangements have induced cosmetic measures such as shifting expenditures to off-budget accounts or engaging in creative accounting, rather than genuine efforts to restrain fiscal spending (Easterly 1999). While it remains an active research topic how stringent fiscal rules and procedures will affect the fiscal performance in a longer time horizon, these developments vividly illustrate the difficulties in achieving fiscal discipline, and raise more fundamental questions such as why some countries have recurrent fiscal problems, while others do not. This question equally applies to developing countries of today. An important puzzle arising from the contrasting macroeconomic experiences across developing regions in the past decades is why certain countries, notably in Latin America and sub-Saharan Africa, have repeatedly adopted unsustainable fiscal policies, while others, notably in East Asia, have maintained sound fiscal policies.^ The unsustainable fiscal policies are often characterized by chronically large deficits and/or high volatility of fiscal outcomes, which are believed to have aggravated macroeconomic instability and contributed to poor growth or even growth collapses. Our research presented in this monograph is motivated by these fiscal developments and related macroeconomic issues. In particular, we advance social polarization thesis that social polarization and degree of social polarization are key factors behind the fiscal policy making process and the adoption of unsustainable policies that have caused macroeconomic problems of fiscal The Real Plan of Brazil was successful in reducing inflation between 1994 and 1998, but not in containing the fiscal deficit. The large fiscal deficit, close to 8% of GDP in 1998, also contributed to a widening of the current account deficit to 4.5% of GDP in the same year (IMF 1999), Argentine crisis in 2001 was largely driven by unsustainable fiscal debt dynamics compounded by the high initial debt level and the rigid exchange rate system. Argentine failed to strengthen its public finance sufficiently when it should have run a surplus during the years of strong growth. Thus, deficits were run throughout the 1990s (Hausmann and Velasco 2002; Allen 2003). ^ Thus, the excessive deficit procedure has been invoked for these countries, but its enforcement ended in a stalemate in November 2003. Since then, the rules have been amended, allowing under certain conditions more time to correct an excessive deficit. The European Council has made a decision in March 2005 that will allow more flexibility in applying the rules. ^ This is true even in the aftermath of the 1997-98 Asian crisis. Also, it is widely accepted that it was not caused by irresponsible fiscal policy but by heavy, private short-term borrowing, which was compounded by weakness of the financial system (see Radelet and Sachs, 1998).
1 Introduction
3
deficits, volatile fiscal outcomes, poor growth and inflation. We contribute to the literature by developing dynamic models of fiscal policy in which social polarization (conflicts of interest among socio-economic groups) plays a central role in generating aforementioned undesirable economic problems, and by presenting supporting evidence as well as comprehensive evaluation of relative importance between our explanation and other existing studies in the literature. (There is now a large political economy literature. See Drazen 2000; Persson and Tabellini 1999a for a literature survey.) Intuitively, a high degree of social polarization of preferences may make it hard for policymakers to agree on ideal government policies because of potential conflicts of interests, and hence may cause a coordination failure among the policymakers. In the presence of polarization of social preferences over public choices, heterogeneous policymakers may have greater incentives to insist on their preferred policies and may end up choosing individually rational but collectively inefficient policies, especially when institutional restraints on policymakers are lacking. We make this intuition more concrete and rigorous in theoretical models and successfully test the implications using cross-country data. Social polarization is one of the oldest ideas found in the political economy literature. Political scientists often distinguish between two major types of social polarization or social conflict: economic cleavage and cultural cleavage (Powell 1982). An important source of economic cleavage is unequal income distribution. The idea that inequality deepens factionalism and dissension in a society is really an old one, perhaps dating back to an ancient Greek philosopher Aristotle. James Madison also noted in 1787 that "the most common and durable source of factions has been the various and unequal distribution of property." Social polarization arising from struggles over the income distribution in turn can be a major impediment to successful economic performance.^ For example, high income inequality or ethnic fragmentation have long been mentioned as an important explanation for recurrent populist fiscal policies and macroeconomic crises in Latin America and sub-Saharan Africa. Conversely, initial low income inequality in East Asia seems conducive to economic growth by promoting stable macroeconomic environment and encouraging human capital accumulation. (See Sachs 1989; Kauffman and Stallings 1991; Birdsall et al. 1995; Rodrik 1996; Engerman and Sokoloff 1997 among others). Recently, empirical growth literature has found that income inequality and ethnic divisions are detrimental to growth (see Perotti 1996; Easterly and Levine 1997; Easterly 2002; Woo 2004 among others). In a celebrated book, The Elusive Quest for Growth, Easterly (2001) bluntly says "One way to summarize the conditions favorable for growth is that progrowth policies are more likely when the two most common forms of social polarization, class confiict and ethnic tensions, are absent." Some of the earlier work on this topic in modern time can be found in Tarantelli (1986) (see Sachs 1989).
4
1 Introduction
Surprisingly, however, there are very few systematic theoretical and/or empirical studies on the role of social polarization (of preference among socioeconomic groups) in explaining fiscal policy decisions and its fiscal outcomes. We fill this void in the literature. Importantly, the issue of social polarization is not confined to some developing countries inherited with a history of unequal income distribution or ethnic division. It is also relevant to industrialized countries. Some economists have warned that the on-going globalization process and technological advance can be a potential source of social polarization, creating a new set of class divisions, unless domestic policies and institutions are in place to cushion the negative effects associated with these trends. In an influential study, Rodrik (1997) writes "The on-going globalization process is exposing social fissures between those with the education, skills, and mobility to flourish in an unfettered world market—the apparent "winners"— and those without. These apparent "losers" are increasingly anxious about their standards of living and their precarious place in an integrated world economy. The result is severe tension between the market and broad sectors of society, with governments caught in the middle." He argues that this tension explains the growth of the size of governments in the international integration process as the governments try to preserve social cohesion by fiscal policy tools including social welfare systems.^ Social polarization is becoming a policy concern even in a so-called "miracle economy" like South Korea of today, as greater integration with the world economy and liberalization are generating a new set of class/sector divisions (Lee et al. 2004; Woo, S. 2005). Also, there is historical evidence that international trade and migration had significant consequences for income distribution, which is a source of social polarization. Williamson (1998) documents that globalization accounted for more than half of the rising inequality in rich, labor-scare countries (like United States, Argentina, and Australia) and for a little more than a quarter of the falling inequality in poor, laborabundant countries (like Sweden, Denmark, and Ireland) in the pre-World War I period. Therefore, we believe that the macroeconomic implications of social polarization are an important research subject relevant to industrialized and developing countries alike. ^ Anxieties about globalization are widely spread, although they are often exaggerated and are not always backed up by data. For example, OECD (2005b) recently reports that rising imports, outflows of foreign direct investment (sometimes tied directly to the relocation of production) and inflows of immigrants all contribute to rising job insecurity in OECD countries. The rapid integration into the world trading system of China and India, with their huge pools of low-wage labor, and the recent enlargement of the European Union have fuelled fears of job losses and wage cuts. In the United States, job outsourcing was one of the hotly debated issues during the 2004 presidential election campaigns. In a similar vein, Agell (1999) argues that the globalization of economic activity tended to lead to increased demand for various labor market rigidities.
1 Introduction
5
The plan of the book is as follows. In Chap. 2, we present a model of endogenous fiscal policy in a simple growth framework where social polarization (of preferences among different socio-economic groups) over the ideal composition of government spending plays a central in the fiscal dynamics. In a highly polarized society, a deficit occurs endogenously, fiscal spending path becomes more volatile, output collapses, and economic growth rate is reduced along the transition path to a new lower level of output. One novel feature is that the size of fiscal deficit, the magnitude of fiscal volatility, and the size of reduction in output and growth rate are explicitly shown to be increasing functions of the degree of social polarization. Thereby, we offer a fiscal instability channel that negatively links social polarization and growth, which is an alternative yet distinct explanation for the empirical finding that social polarization is harmful to growth. Moreover, we fully distinguish the incentive to engage in such short-term policies under political uncertainty from that under polarization. Polarization and political uncertainty are shown to be distinct forces that can drive the aforementioned macroeconomic problems. In the next chapter, we extend the framework that is developed in Chap. 2 to study infiationary consequence of the fiscal policy set by polarized policymakers by introducing money and infiation tax into the economy. We derive a positive relationship between inflation and social polarization, which is consistent with empirical findings in the literature. Interestingly, we show that inflation dynamics depend on the money-bond ratio in the deflcit finance and an increase in that ratio is less infiationary for a given degree of polarization. Chapter 4 is motivated by an important puzzle arising from the contrasting macroeconomic experience across developing regions, that is, in sharp contrast to East Asia, much of Latin America and sub-Saharan Africa has often engaged in unsustainable fiscal policies, leading to huge fiscal deficits, external debt crises, or hyperinfiation. We study the theoretical relation among income inequality, industrialization, and social polarization on the one hand, and the relation between the resulting social polarization and fiscal outcomes on the other hand. In a two-sector economy with a manufacturing sector and a traditional sector, the more unequal the initial income distribution, the larger the sectoral income gap during industrialization and the more likely the polarization of sector preferences for different types of government spending. In a highly polarized society, a fiscal deficit results and fiscal outcomes exhibit greater fluctuations over time. We successfully test these two predictions with data on consolidated public sector fiscal balance, and present the first econometric evidence on the positive link between income inequality and undesirable fiscal outcomes in a panel of 90 countries in the period of 1970-90. Countries that have suffered from the greatest fiscal instability (deficits and volatility) are those with highly polarized societies as measured by indicators of income inequality. In Chap. 5, we provide a comprehensive empirical test on a large pool of potential explanatory variables of public deficits in a panel of 57 countries including OECD countries and derive robust conclusions about which of
6
1 Introduction
these variables (or which theories) are important in explaining cross-country variation in fiscal balance. Financial depth, income inequality, assassinations, cabinet size, and centralization of authority in budgetary decisions are found to be significant and robust determinants of public deficits. We also propose a working hypothesis and provide supporting econometric evidence: social polarization is important in explaining differences in fiscal outcomes across countries, yet its effects may be even more pronounced or suppressed, depending on the political and institutional structures through which social polarization is linked to the fiscal policy-making process. Indeed, effects on public deficits of the sociopolitical variables tend to be smaller in countries with better institutional arrangements. Conversely, the sociopolitical polarization has very strong effects on deficits in the presence of poor institutions. In general, there are rich interactions among socio-political and institutional factors. The results are confirmed by extensive robustness tests such as the sensitivity analysis and the robust estimation method. Finally, we empirically investigate the fiscal instability channel that negatively links income inequality to growth in Chap. 6. Although we have derived the theoretical result on the fiscal instability channel in Chap. 2, we have not tested it yet. We present supporting new evidence in a cross section of 93 countries for the period of 1970-2000. There are numerous empirical papers on the relationship between income distribution and growth, yet this is the first study to present such an evidence on this specific fiscal instability channel. As a matter of fact, most studies have focused on the reduced-form growth regression that includes an income distributional measure as an explanatory variable, rather than investigating specific channels through which income distribution affects growth. When tested against other competing theories, the data continue to support our fiscal instability channel. We pay careful attention to robustness and consistency of our results with respect to various issues in running a regression, such as outlier and endogeneity problems.
Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
2.1 Introduction "The fundamental difference between redistributionist and developmentalist governments is social polarization. Societies divided into factions fight over division of the spoils; societies united by a common culture and a strong middle class creates a consensus for growth—growth that includes the poor." from Easterly (2001, p. 256) In recent years, empirical studies on long-term growth have found that social polarization that arises from ethnic divisions or struggles over the income distribution is detrimental to growth. (See Rodrik 1999; Easterly and Levine 1997; Alesina and Rodrik 1994 among others.) Countries with polarized societies, as measured by ethnic fractionalization or income inequality, seem to be more prone to adopt growth-retarding policies—for example, unsustainable fiscal policies that lead to large budget deficits, volatile fiscal outcomes, and growth collapses.^ (See Figs. 2.1 and 2.2.) In such countries, socio-economic groups may have sharp disagreements on ideal government policies, which may cause a coordination failure among policymakers that leads to an adoption of individually rational but collectively inefficient policies. Perhaps social polarization is one of the oldest ideas found in the political economy literature. Yet there are very few (or no) systematic theoretical studies on the role of social polarization (of preference) in collective decisionmaking process and in the development of aforementioned macroeconomic problems. In general, the heterogeneity of preferences is one factor that has Reprinted from European Economic Review, Vol. 49, August, 2005, pp 1451-1477, Woo J, "Social Polarization, Fiscal Instability and Growth", with permission from Elsevier. In a comprehensive study of public sector deficit, Easterly et al. (1994) conclude that large fiscal deficits are largely explained by conscious fiscal policy choices and not by external or by domestic macroeconomic shocks.
8
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
not been well-recognized as critical to the coordination failure in collective action. This chapter provides a systematic analysis of the role of polarization among socio-economic groups in the evolution of fiscal instabilities (large deficits and fiscal volatilities) and their negative effects on the capital accumulation process.^ We build a dynamic game model of fiscal policy in a simple growth framework in which social polarization (more precisely, polarization of preference for types of government spending between socio-economic groups) plays a central role in both generating fiscal instability and growth collapse. Thereby, we emphasize that society's polarization and degree of polarization are key factors underlying policy decisions that are responsible for such undesirable macroeconomic outcomes. On the other hand, social polarization may not only be responsible for a coordination failure but is often thought to be associated with socio-political instability. Empirical growth literature also finds that socio-political instability is harmful to growth. (See Perotti 1996; Alesina et al. 1996.) High levels of socio-political unrest may not only make the downfall of the present government more likely but may dramatically shorten the horizons of politicians. With a shortened expected tenure in office, the government would be more likely to engage in short-term policies at the expense of macroeconomic stability. In this chapter, we fully distinguish the incentive for policymakers to engage in individually rational but socially inferior policies under political uncertainty from that under polarization of preferences. We make a contribution to the literature by clearly bringing out the different roles of polarization and political uncertainty (refiected in the discount factor) in generating fiscal deficits, volatile fiscal outcomes and output collapses, and in reducing economic growth along a transition path to a steady state in a unified framework. Interestingly, social polarization and political uncertainty are shown to compound to produce even worse outcomes in terms of fiscal instability and poor growth. Here we consider an economy in which two heterogeneous policymakers jointly control fiscal policy, but have different objective functions. Each policymaker maximizes her own utility from her public good provision that benefits a specific group (or sector) more than the other. Each represents a different group that may have a different preference for the public goods. The two groups can be thought of as either capitalists and labor workers, manufacturing (formal) and traditional (informal) sectors, right-wing and left-wing parties, urban and rural sectors, or two powerful ethnic groups. When the preferences for types of government spending differ substantially among policymakers (or equivalently among the groups they represent), a fiscal deficit occurs endogenously. This is due to strategic behaviors of the policymakers who have different preferences yet share the government budget. Each ^ Throughout this chapter, the term "fiscal instability" means both a large fiscal deficit and fiscal volatility.
2.1 Introduction
1
J
• Taiwan
•Kore%SfK&apore
• Hong Kong
• Malaysia • Thailand^ Botswana • Indonesia
• Ireland Portugal • Tunisia
I i
•Sweden ^ New Zealand • Switzeriand
^tGree«rf»«BWf^gn^. """^ •Mexico • Morocco # Nigeria #(IMIi|iiHifl »IJIHtfirt'iM \ •Ecuador • Costa Rica ^^^^^-••..„„,^^^^ • Argentina
^
• Venfeiteiaaica
'Wl^^^&A^^ß^^^^^ ^ ™ V B f e n ^ a l ^ p^^ • South Africa ^ cote d'lvoire • Liljeria
1
• Zambia • Sierra Leone 0
|
0.5
SOCPOLA
Fig. 2.1. A scatter plot of average growth rate of real per capita GDP for the period of 1970-98 against a composite index of social polarization (SOCPOLA) that is based on Gini coefficients, ethno-linguistic fractionalization, and institutional quality around 1970. Higher values of SOCPOLA indicate higher social polarization. Data Source: World Bank (2000), Deininger and Squire (1996), and Easterly and Levine (1997). policymaker is aware that whatever government resources she does not exploit may or may not be available for the future provision of her preferred public good, depending on the spending decision of the other policymaker. When policymakers disagree on the composition of government spending, each of them has a greater incentive to overexploit the common government resources and consequently exerts a net negative externality on the other. This prevents them from achieving a socially optimal fiscal outcome. The size of deficit rises with the degree of preference polarization because of the positive relationship between the preference polarization and the incentive to exploit the common resources (polarization effect). Political uncertainty facing policymakers, as reflected in a high discount rate, has an effect similar to that of polarization. Importantly, an economy with a higher degree of polarization will also exhibit greater fluctuations in fiscal spending in response to shocks to the government revenue. The higher the degree of polarization, the more volatile the fiscal path. Relatively heavy discounting of the future events by the policymakers will cause a volatile fiscal spending path too. In the presence of preference polarization (or impatience) among the policymakers, a shock to tax revenue is translated into a more than proportional change in spending. That is, if the degree of polarization is positive or if the policymakers' subjective discount rate is substantially high, then government spending rises (falls)
10
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
• Uganda
• Brazil
• Sierra Leone
• Argentina
•NiaP2imbia
^ Mala>^ ^ ^^~ • Malays«SSlÄtf^^^
I °]
• Italy
C O
i
* ^TßläH
• Belgluoi^
1 J
^ — « C ^ a •Spain tJi^SlI^^ j^Kina^Omited States rk • France • Sweden
• Sri ferteBP^Jca^^^,. '"'•TMorocco
• Egypt^
• C ^ s f i f e ' " ' ^ " *^®P; • South Africa ^ PhiiiDDinl&'^^^ • Costa Riq^Tunisia • Indonesia * Kniiippin^ • Vtnf tateid • B Salvactoeenegal • Colombia • Guatemala
• Norway
0
0.5
SOCPOLA
Fig. 2.2. A scatter plot of a measure of macroeconomic instability over the period of 1970-98 against the social polarization index (SOCPOLA). The index of macroeconomic instability is based on CPI inflation rates, central government deficits, and volatility of real GDP growth (measured by standard deviation), whose higher values indicate greater macroeconomic instabilities. Data Source: The same as in Figure 2.1.
more than proportionally in response to a positive (negative) shock to tax revenue—^this can shed light on the procyclicality of fiscal policies in Latin America that is extensively documented by Gavin and Perotti (1997) (see Woo 2005b for empirical evidence on this prediction). The intuition is similar to the polarization effect. This is due to an interaction between the shock to the government revenue and the policymakers' incentives to exploit it in the presence of the dynamic negative externality operated by the preference polarization or the political uncertainty. The output level and the transition dynamics of economic growth also depend crucially on polarization and political uncertainty. A fiscal deficit arising from polarization or political uncertainty leads to ineflficient capital accumulation in the private sector, and permanently lowers the levels of capital stock and output in the economy. This is because policymakers waste valuable government resources to maximize their utility from producing public goods; hence, they overspend beyond the socially optimal level for a given tax revenue. In the presence of polarization (and/or political uncertainty), the economic growth rate is reduced along a transition path to a new steady state of a lower level of output. The higher the degree of polarization or the subjective discount rate is, the sharper the decline in the growth rate is. In our model the growth collapse is caused by fiscal instability that is ultimately
2.1 Introduction
11
attributed to social polarization (and/or political uncertainty). We characterize the transitional dynamics of growth as a function of polarization and political uncertainty. Our fiscal mechanism is related to a growing literature on fiscal politics (see Alesina and Perotti 1995; Persson and Tabellini 1999a for a literature survey), and particularly to the common pool or pork barrel problem approach. The related papers are Weingast et al. (1981), Chari and Cole (1993a), Tornell and Lane (1998), Hallerberg and von Hagen (1999), and Velasco (1999).^ Under this approach, an excessive spending or deficit (i.e., an overexploitation of a common property) can arise because interest groups that have access to the government resource fail to internalize the full cost of their own appropriation. Each interest group enjoys the full benefit of a specific public spending, while it pays only a fraction 1/n of the cost (i.e., the cost spread through generalized taxation). These studies typically consider n-player symmetric games (although in different contexts), addressing issues such as whether an increase in the number of groups n leads to a worse economic outcome, the possibility of delayed fiscal reforms, or the effect of budgetary institutions on the size of deficits.^ Here we address different issues, however. In general, the heterogeneity of preferences among the players is one factor that has not been well-understood as critical to the coordination failure (i.e., dynamic negative externality) in collective action in a common pool setting. So is the subjective discount factor of the players. In this chapter, a more general model is developed that introduces two new dimensions of preference polarization and political uncertainty into a class of two-player common pool games.^ It is shown that preference polarization and discount factor are critical conditions for the dynamic negative externality to become operative in the common pool context. Not only would this help us to theoretically identify critical conditions leading to overexploitation of the common resources, but also would yield richer, yet distinct, theoretical and empirical implications for fiscal dynamics and the capital accumulation process as shown in the chapter. For example, our result suggests that the common pool problem would be more likely to occur and be more severe in societies with higher degrees of polarization (and/or political The common pool problem refers to a situation in which a productive asset is exploited jointly by economic agents whose noncooperative behavior results in an overexploitation of the asset that is not Pareto optimal. The existence of Pareto inefficient Nash equilibrium in this context was first shown by Levhari and Mirman (1980). See Fudenberg and Tirole (1992) for more details. ^ The relationship between the number of groups n and the common pool problem is addressed by Weingast et al. (1981), Tornell and Lane (1998), and Velasco (1999); the possibility of delayed fiscal reforms by Velasco (1999); and the effect of budgetary institutions on the size of deficits by Hallerberg and von Hagen (1999). ^ Thus, a typical two-player common pool model in the existing literature can be viewed as a special case in our framework.
12
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
instability). By sharp contrast, the existing literature tends to associate the severity of the common pool problem with the number of participants in the collective decision-making process. However, the theoretical relationship between the number of groups and the common pool problem is fragile because it turns out to depend crucially on the assumptions about the shape of utility function (see Kontopoulos and Perotti 1999). Moreover, we go well beyond the issue of budget deficit by analyzing the theoretical linkage of social polarization (and political uncertainty) to fiscal volatility, procyclicality of fiscal spending, and economic growth process. Again, in contrast, the existing models of fiscal deficits do not address the volatility or procyclicality of fiscal outcomes, let alone economic growth. The innovation of this chapter is that the size of fiscal deficit, the magnitude of fiscal volatility, and the decline in output and economic growth rate are explicitly shown to be increasing functions of the degree of social polarization and the subjective discount factor. Social polarization has long been mentioned as an important explanation for populist fiscal policies and poor macroeconomic performance in many developing countries.^ Yet there have been very few systematic theoretical studies on the role of social polarization per se in these problems.^ We fill this void in the literature by demonstrating how social polarization can cause the aforementioned fiscal and growth problems.
As a matter of fact, most of the existing studies tend to focus on either income inequality or ethnic fractionalization (sources of social polarization), rather than social polarization of preferences per se. For example, many economists have argued that unequal income distribution provides an important answer to the questions of why populist fiscal policies appear more often in Latin American countries than other regions. See Rodrik (1996) and Kaufman and Stallings (1991) among others. Even in this literature, however, there are very few theories that explain why unequal income distribution can lead to large deficits and volatile fiscal outcomes! A partial exception is Alesina and Tabellini (1990) who briefly discuss the linkage between polarization and budget deficit in a different mechanism. They show that faced with re-election uncertainty, the incumbent may fail to internalize the costs of additional debt and hence tend to have a deficit bias. Then they extend their argument into the case where two political parties have different preferences. Let alone the different mechanism we employ in our model (i.e., the common pool setting), however, in this type of model it is not the polarization of preference per se but the re-election uncertainty that is the critical condition for an endogenous deficit to arise. (See also Chari and Cole 1993b on this point.) If the government is not faced with the re-election uncertainty, deficit bias would not occur regardless of the opponent's preference. On the other hand, it is surprising that social polarization has largely been ignored in the empirical studies of fiscal deficits. Woo (2003b) presents the first econometric evidence that countries that have suffered greatest fiscal deficits also tend to be those with highly polarized societies as meeisured by indicators of income inequality in a panel of 90 countries over the period of 1970-90.
2.2 Endogenous Growth Model with Optimizing Interest Groups
13
Last but not the least important, we offer an alternative explanation for an important empirical finding that social polarization is harmful to economic growth: a fiscal instability channel. This can be viewed as an alternative fiscal policy channel, but distinct from the redistributive policy channel proposed by Alesina and Rodrik (1994) and Persson and Tabellini (1994). In their models, distributive conflicts within a society lead the government to engage in redistributive policies that may be harmful to economic growth. However, there seems to be a lack of empirical support. What matters for growth in their models is the distortion caused by income tax that accompanies redistributive spending. Perotti (1996) does not find any negative relationship between tax variables and growth. By contrast, our theoretical explanation is highly plausible since fiscal deficits are found to be harmful to growth in numerous empirical growth studies (Fischer 1993, to begin with). And recently Woo (2003a, 2003b) finds strong evidence that social polarization, as measured by income inequality, is robustly and positively associated with fiscal deficits in a comprehensive empirical investigation (see Chaps. 4 and 5). In Chap. 6, we also present evidence both for the positive relation between income inequality and fiscal volatility and for the negative relation between fiscal volatility and growth in a cross section of countries in the 1970-2000 period. The plan of the chapter is as follows. Sect. 2.2 presents a simple endogenous growth model with optimizing interest groups. Sect. 2.3 derives an endogenous government fiscal policy and establishes our main results on the linkage between social polarization and fiscal instability. The social planner's solution is then computed and compared with the non-cooperative feedback Nash equilibrium determined by two heterogeneous policymakers. This is followed by an extension of the model into the case of policymakers facing political uncertainty. Sect. 2.4 analyzes the effect of this endogenous fiscal policy on the capital accumulation and growth in the presence of polarization and political uncertainty. Our conclusions are in Sect. 2.5.
2.2 Endogenous Growth Model with Optimizing Interest Groups We consider an endogenous growth model with no population growth. The economy is populated by a government and a private sector composed of two groups, indexed by i, i = 1,2. These two groups may represent either capitalists and labor workers, manufacturing (formal) and traditional (informal) sectors, urban and rural sectors, two powerful vested interest (ethnic) groups, or right-wing and left-wing parties. Each group consists of a large number of atomistic individuals. The government and the private sector have perfect foresight. The infinitely-lived representative agent in group i seeks to maximize her lifetime utility, which is additively separable:
14
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
J'=
f
[log{ci)-{-Xilogigi) + {l-Xi)log(92)]e-^'dt,
(2.1)
where Ci is private consumption; gi and p2 are two different public goods provided by the government; and p is a subjective discount rate, p > 0. Being small, each member of group i takes gi as given and has the same preference for the two public goods within the group. But these two groups differ in their preferences for the public goods, which is reflected by Aj. We assume that 0 < Aj < 1, for i = l , 2 and A2 < | < Ai. This implies that group 1 prefers gi to ^2 and group 2 prefers p2 to pi. Even though the agent in group i may not like the public good gj, j ^ i as much as gi, it is included in her utility function because of non-exclusiveness of public goods. We also assume that she derives positive utility from the consumption of public good which is not her most favorite one.^ We define ^ = Ai — A2 and interpret it as the degree of difference in their preferences for two public goods. We can think of Ö as a degree of polarization between the two groups. We note that 0 < ^ < 1. While 9 = 1 implies the complete disagreement on the composition of two public goods between two groups, 0 = 0 implies the total agreement in their preferences. We will see the important role played by 6 in the evolution of fiscal deficit and fiscal volatility. Also, capital accumulation and growth process depend crucially on 0, as we will show later. In the economy there are two kinds of real assets: capital, denoted by k, and government bonds, denoted by b. The bonds are assumed to be a perfect substitute for capital and therefore to pay the same rate of interest, r. The dynamic budget constraint of the representative agent in group i is then for Vt > 0 and ao > 0 given, äit = rau - Cit - n,
(2.2)
where an is the asset held by an agent and hence an = ku + ba^ and TI is a lump-sum tax collected by the government from group i.^ We also impose the No-Ponzi-Game (NPG) condition: lim ttite"^* > 0.
(2.3)
As long as marginal utility is positive, the agent will not want to have increasing wealth forever at the rate of r, and that condition will hold as an equality. (See Barro and Sala-i-Martin 1995.) The representative agent in group i maximizes the lifetime utility function (2.1) with respect to c^, subject to equations (2.2) and (2.3). It follows from the first-order conditions for this maximization problem that we get ^ For example, even though an agent may care about the public education expenditures much more than the national defense expenditures, she will also benefit from the national defense. ^ We assume n = r2. The assumption of lump-sum taxation is mainly for simplicity of algebra and can be relaxed without affecting the qualitative implications in the chapter. In what follows, we mean ^ by x.
2.2 Endogenous Growth Model with Optimizing Interest Groups ^ = r - p .
15 (2.4)
Cit
Thus, the optimal consumption path of the agent i is at = coe^^-"'*,
(2.5)
where CQ remains to be determined. (See footnote 12 for the solution.) The budget constraint for the whole private sector is at = rat -ct-T,
Vt > 0,
(2.6)
where at = ait+0L2t, h = kit-{-k2t, h = 6it+&2t5 Q = Cit+C2t, and r = r i + r 2 , for alH > 0 (we normalize as if there is one agent in each group). Now, following Barro (1990), we introduce firms that have the linear production function: (2.7) y = f{k) = Ak, where A > 0 is the constant marginal product of capital. We can think of capital as encompassing human and nonhuman capital.^^ In a competitive equilibrium, the marginal product of capital is equal to the rental price for a unit of capital services. This is the first-order condition for maximization of profit. Therefore, in competitive equilibrium A = r-\-ö,
(2.8)
where 6 is a. constant depreciation rate and r -h (^ is the rental price for a unit of capital services. Thus, consumption at time t is given by^^ at = coe(^-^-^)*.
(2.9)
From the dynamic budget constraint equation (2.6) and the profit maximization condition equation (2.8), we get the following (We suppress the time index, t, when there is no confusion.): k = (A-6)k-c-gi-g2.
(2.10)
Thus, the equilibrium for the private sector is completely described by equations (2.9), (2.10), and (2.3), given government debt, 6, and the government budget constraint. ^° Also see Barro and Sala-i-Martin (1995). ^^ If the marginal productivity of capital is sufficiently large so that A - 5 - p > 0, then consumption grows over time. However, this does not yield unbounded utility because for given gi and p2, f=
/•oo
^0
[log{co) + (A-S-p)t
+ Xilog{gi) + (1 - Xi)log{g2)]e-'''dt < oo.
16
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
Suppose t h a t t h e government budget is balanced at each point in time. T h e n 6 = 0 and bt = bo,yt > 0. Under t h e balanced budget assumption, t h e equilibrium capital stock kt is given by^^
fc^^IZl^
+ ^e^^-P)*
(2.11)
T h e capital accumulation equation specified in equation (2.11) will be useful in computing t h e impact of polarization on capital stock and growth in Sect. 2.4. It is straightforward t o see t h a t t h e asymptotic growth r a t e of capital {k /k) is A — 6 — p. In fact, t h e growth rates of consumption, capital stock, and o u t p u t all asymptotically approach A — 6 — p}^ In other words, lim - = lim - = lim t—^oo k
t—»-oo c
'^=r-p
= A-S-p.
(2.12)
t-*oo y
For now, we assume t h a t t h e agents' subjective discount r a t e is equal t o t h e interest rate, i.e., p = r, so t h a t capital stock and o u t p u t stay constant under a balanced budget {b = 0). Also, note t h a t consumption is constant if p = r. This assumption is only made to serve as a benchmark and t o highlight ^^ Using equation (2.10) and the NFG condition (2.3), we get kt. Under the assumption of a balanced budget, equation (2.10) is k — rk = rbo — r — 2coe^^~^^*. To solve this first-order differential equation, multiply both sides of the equation by the integration factor, e~^*, and integrate it from t to oo. We then have /»oo
oo
/
e-"*(fc - rk)dt = / e-"*(röo - r - 2cQe^''-^^')dt. By applying the NPG condition limt-^oo kte~'^* = 0 to the integration, we can derive equation (2.11). Since y = Ak^ we easily find the output path by using equation (2.11): y = Ak. The initial level of consumption, co, is determined by the following condition. From equation (2.11), T-rbo 2co fco =
h
,
r p and Co is thus determined by the initial level of the capital stock, the lump-sum tax, and initial bond holdings. ^^ Note that in the standard AK growth model, there is no transitional dynamics; Ct, kt, and yt grow at the constant rate of A — S — p. However, in our model where we introduce tax and the government bond, the growth rates of kt and yt are not constant at A — S — p, but only asymptotically approach that rate. This can be checked in equations (2.11) and (2.12). This also implies that changes in polarization affect the growth rate of the economy only along the transition path to the steady state; the same is true for political uncertainty. For more about this, see Sect. 2.4.
2.3 Endogenous Fiscal Deficit and Volatility: Polarization
17
the main points of the chapter. Later, we will discuss the consequences of relaxing this assumption (see Sects. 2.3 and 2.4). Before we move to government fiscal policy, we make an additional assumption on the timing structure. Policymakers are assumed to simultaneously move before the private sector moves; therefore, when the private sector makes a decision, it has information about government fiscal policy that was determined by policymakers and takes the government policy as given.
2.3 Endogenous Fiscal Deficit and Volatility: Polarization In the previous section, we took the government budget as given and assumed a balanced budget, 6 = 0. Now we consider the endogenous fiscal policy controlled by two policymakers (interchangeably called ministers) who jointly represent the Fiscal Authority (FA) of the government. We assume that the government can transform the consumption goods produced by the private sector into two non-storable public goods, gi and p2. Two ministers indexed by i, i = 1,2, represent the corresponding group, i = 1,2, in the private sector. Minister i provides the public good gi to the private sector, which is financed by government revenues. Each minister, i, derives greater utility from the provision of her favorite public good gi than from the other gj. Since they have different preferences for the two public goods and seek to maximize their own utility, two ministers faced by the common government budget constraint behave strategically in determining the amount of public goods they provide. To describe this endogenous fiscal policy determination process, we consider a diflPerential game between two ministers. Specifically, we explore the set of feedback Nash equilibria, which are subgame perfect and time consistent, in a game-theoretic model in which two ministers can jointly exploit the government net revenues to maximize their utility from providing their favorite public goods.^^ Each minister, i, has the following objective function: V' = / [Xilog{gu) + (1 - Xi)log{g2t)]e-'^'dt, Jo
(2.13)
where 0 < A 2 < | < A i < l and the minister's discount rate is assumed to be equal to the interest rate.^^ Minister i prefers gi to gj, j ^ i and i, j = 1,2,
15
It is well known that the subgame perfect equilibrium is time consistent. For more about the feedback Nash Equilibrium and time-inconsistency problem, see Cohen and Michel (1988). Each minister's utility is assumed not to depend on her consumption, which might bias the policies towards an oversupply of public goods. Yet the assumption that the discount factor is equal to the interest rate delivers a constant consumption path as one can see from equation (2.5). Without loss of generality, we can nor-
18
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
which implies that she puts more weight on her favorite public good, gi, in her utility function. Minister i shares the same weight Xi with her favorite group, so that 0 = Xi — X2 is also the degree of preference polarization between the two ministers. Now we turn to the budget constraint of the government which faces the ministers. The government collects the lump-sum tax of r from the private sector (with normalization of the number of agents in each group to one). Government expenditures can also be financed by issuing bonds at a constant real rate of r. The government budget constraint at each instant is then b = rb-\-gi-\-g2-T,
(2.14)
where b is the stock of national debt.^^ 2.3.1 Non-cooperative Feedback Nash Equilibrium Each minister i chooses her control variable, pj, so as to maximize her utility function (2.13) subject to the government budget constraint equation (2.14) and the NPG condition for every possible choice of the other minister's control variable gj, j 7^ i. Here we employ the feedback Nash equilibrium concept, which allows players to revise their actions through time as the game evolves.-^^ To facilitate the computation of equilibrium in this game, we define government net revenue, Rt, as Rt = r - rbt.
(2.15)
In general, the feedback strategy is a function of time and state; however, very few diflFerential games can be solved in closed form because the firstorder condition for this feedback Nash equilibrium involves a system of partial differential equations. In order to get a closed-form solution, we restrict the strategy set to linear Markov strategies that depend on the current state. We
16
malize the constant consumption to 1. Then log(c) = 0 justifies the specification of the utility form. Note that the No-Ponzi-Game condition relevant for the government budget constraint is lim 6te-"* = 0. t—>oo
17
In differential games, open-loop and feedback Nash equilibria are among the most commonly employed equilibrium concepts. Open-loop strategies are ones for which each player chooses all the values of his control variable for each point in time at the outset of the game. This is relatively easy to solve for, but in general is time inconsistent. On the other hand, feedback strategy consists of a contingency plan that indicates what the best thing to do is for each value of the state variable at each point in time. That is, it allows the player to revise her action at each instant on the basis of the state at that point in time. Hence, the feedback strategy has the property of being subgame perfect.
2.3 Endogenous Fiscal Deficit and Volatility: Polarization
19
then conduct transformation of the variables so that we can construct a game structure in which an open-loop strategy calls for the same rate of public good provision at every point in time as that of the feedback strategy. This is known as "synthesizing the feedback control." ^^ Let us now consider the following linear strategies: 9it=XiRu (2.16) where xi will be endogenously determined as a part of the solution. We assume that the set of strategies is xi ^ [0, oo). Let t/^t = logRt- Then, we can rewrite the minister's objective function (2.13) as /•oo
V^'(Xi,X2)= / [A^/op(xl) + (l-A,)/op(x2) + V't]e-^*rft, Jo
(2.17)
and the budget constraint becomes i^t = r-rxi
-rx2'
(2.18)
Now we solve for the feedback Nash equilibrium, which is also the Markov perfect equilibrium, by maximizing the objective function of minister z, equation (2.17), with respect to Xi subject to equation (2.18). Minister i's Hamiltonian is given by = l>^ilo9{xi) + (1 - K)log{x2) + V'tje"^* + fiit[r - rxi - rx2], (2.19) where Xi is the control variable, fiu the costate variable, and tpt the state variable. The feedback Nash equilibrium to this game is as follows (See Appendix A for the derivation of the following feedback Nash equilibrium.): H'{xuX2,ipt)
Xi = Al, X2 = (1 - A2); and hence p*, = XiRt, git = (1 - \2)Rt- (2.20) Substituting p*^ = x?-^* and Rt = r — rbt into equation (2.14) yields bt = (Al - A2)(r - rb) = 0{T - rb) > 0.
(2.21)
Recall that the parameter 0 £ [0,1] is the degree of the polarization between two ministers. Whenever there are differences in the ministers' preferences for two public goods (i.e., Ö > 0), there occurs an endogenous fiscal deficit, 6 > 0.^^ This result is due to the strategic behaviors of these ministers who ^^ In the differential game literature, it is very common to consider linear strategies due to the aforementioned technical complications. See Fudenberg and Tirole (1992) for more about synthesizing the feedback control. ^^ However, the growth of debt is not explosive. If we solve the differential equation (2.21) for öt, assuming 60 = 0 for simplicity, we obtain ht = ^ — ^e~^^*. Thus, the NPG (No-Ponzi-Game) condition is satisfied: limt-.oo btß-''* = {^-^e-^''*)e-''^ =
20
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
have different preferences, but share the government budget. Each minister is aware that whatever government resources she does not exploit may or may not be available for future provision of her preferred public good, depending on the spending decision of the other minister.^^ Thus, when they disagree on the ideal composition of government spending, each has an incentive to overexploit the common resource today. The polarization of preference leads each policymaker to insist on a higher spending for her favorite sector and to exert (net) negative externality on the other, contributing to bigger overall spending and a larger deficit than the social optimum. Whenever two ministers value public goods with different weights, the negative externality of minister j ' s one-unit provision of QJ on minister i's utility through the state variable h always dominates the positive effect that directly enters minister i's utility function. Moreover, the incentive for each minister to overexploit the government revenues increases with the amount of disagreement between the two ministers (polarization effect). This is because the positive effect of minister j ' s one-unit provision of QJ that directly enters minister i's utility function gets smaller, whereas its negative externality operating through the state variable h gets bigger, as the degree of polarization gets bigger. From the point of view of minister i, therefore, one unit of resource devoted to her opponent's favorite type of spending brings a greater net negative externality, inducing her to spend even more for her favorite item ahead of her opponent. This implies that the size of the current budget deficit is a positive function of the degree of polarization (see Fig. 2.3). Along with this result, the intertemporal budget constraint implies that polarization is positively associated with greater changes in fiscal outcomes over time, such as spending and fiscal balance for a given path of tax revenue. The greater the polarization is, the larger the fiscal spending and current fiscal deficit are. But this only raises the debt level more quickly and reduces available government resources, which forces policymakers to cut tomorrow's spending by more.^^ The intertemporal budget constraint means that larger 0. As t —^ oo, 6 approaches J . This is because the game and the strategies are constructed such that it is in each minister's best interest to spend less so as to satisfy the NPG condition. Note that each minister's spending gi depends only on the state variable, net tax revenue {r — rh). As debt (6) is accumulated, the net tax revenue shrinks, which forces each minister to spend less. As a result, total government spending shrinks asymptotically to zero when there is polarization (^ > 0). This is merely due to the lump-sum tax assumption. By contrast, when 0 = 0, the budget is then balanced, and total government spending equals the lump-sum tax for each time period. ^° It is not the level of tax that causes a deficit in our model. This result holds true for any given level of tax revenue. ^^ It should be noted that our model is not related to the Ricardian equivalence experiment. The Ricardian equivalence proposition implies that the timing of taxes does not matter as long as the present value of taxes is equal to the present
2.3 Endogenous Fiscal Deficit and Volatility: Polarization
21
deficits today must be met by larger surpluses tomorrow, causing an even bigger swing of fiscal policy over time (see Fig. 2.3 again). Importantly, an economy with a higher degree of polarization will exhibit greater fluctuations in fiscal spending in response to shocks to government revenues. We can illustrate this point by using the solution for gi and p2Using equation (2.20), we can write the total government spending (g) at time t as 9t = g^u + g^t = (1 + 0){T - rbt).
(2.22)
For any point in time t, government spending is proportional to the net tax revenue r — rb. Note that the policymaker revises her action at each instant on the basis of the state at that point in time, which should be r — rbt- • Thus, if we take the total differentiation on this equation at an instant t + , then (Qt-\= (1 + Ö) > 1 with equality when 0 = 0. (2.23) dr This yields a striking prediction that government spending rises more than proportionally in response to an increase in tax revenue if the degree of polarization is positive {9 > 0). Whenever there is a positive (negative) shock to the government net revenue, it is translated into a more than proportional increase (decrease) in government spending in the presence of polarization.^^ The absolute size of the change in 'g will be even greater with the size of 0. value of government spending plus the value of the initial government debt. This is because the government spending path is exogenously given, whereas taxes are endogenous. In sharp contrast, government spending is endogenous, while taxes are exogenous in our model. Therefore, government debt acts like net wealth for the private sector. Recall that the level of initial consumption depends on the initial levels of capital stock and government bond holdings, and the lump-sum tax (see Footnote 12). In the Ricardian world, higher bond holdings just mean a higher present value of future taxes, which makes government bonds irrelevant for consumption. In our model, however, higher initial bond holdings reduce the present value of future government spending. ^^ This result is reminiscent of the voracity effect in Tornell and Lane (1999) that interest groups' total appropriation of the economy capital stock rises more than proportionally to the windfall to the capital stock. Although the motive and the issues we address in our chapter are sharply different from theirs, the underlying mechanism for both polarization effect and voracity effect is the negative externality in common pool settings. Aside from other major differences from the existing common pool problem literature, the novel feature of our model is that we not only demonstrate that the polarization of preferences (and the political uncertainty, as will be shown in Sect. 2.3.3) cause the fiscal volatility to arise endogenously but also fully show that the size of fiscal volatility itself is an increasing function of the degree of polarization and the discount factor, which yields a new testable prediction. Moreover, the polarization of preference and the discount factor are shown to be the critical conditions for the dynamic negative externality to become operative in a more general two-player common pool games.
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
22
The intuition behind this result is quite similar to that behind the polarization effect. Recall that the equilibrium Markov strategy in equation (2.20) calls for minister I's spending to be equal to the multiproduct of Ai (or (1A2) for minister 2) and net tax revenue R. For a given shock to tax revenue Ar, minister 1 will claim Ai x A r , while minister 2 will want to increase her favorite spending by (1 — A2) x Ar. Unless Ai = A2 = 1/2, it will result in a more than proportional increase in total spending (Ap = {1 -\- 9) x A r ) . In the case of complete agreement (Ai = A2 = 1/2), the increase in tax revenue will be evenly split between two types of spending so that a balanced budget is maintained. 1
1
1
1
1
1
1
1
1
60
^^^-^"^^\^^,.^--^^^^"^^^^
40
e-
^^ ^^<^^^^
20
CO CO
0
ÜS
theta=0
-
y^^^^^
theta=0.2
__...----—-''''''''''''''^^
theta=0.4
^^,^^--^''^''''^^^^yy
_
-20
E E
theta=0.6/^
0
-
> / /
0
-eo, ^ ^ a = 0 ^ /
-
/
-80 /iheta?^
-ino /
]
1
1
1
\ 25
—
1
1
1
1
Time
Fig. 2.3. Primary surplus is given by r — p(t : Ö, p) = r — (1 + &)(r — rbo)e~^^*. The figure displays time path of primary surplus for various degrees of polarization, 0 (theta), where it is assumed that r = 100, 60 = 0, and r = 3.5%. This relation between polarization and government spending volatility is shown in a diagram when we allow Rt (either T or r) to follow a stochastic process (see Fig. 2.4).^^ As one can see from Fig. 2.4, even for the same size of exogenous shock to revenue Rt, government spending (gt) is much more volatile in an economy with greater polarization. We can relate this result to the procyclicality of fiscal outcomes predominantly observed in Latin American countries, which is documented by Gavin Therefore, the kind of voracity effect cannot occur in our framework, if (i) there is no polarization of preferences, and (ii) players are patient enough. ^^ Fig. 2.4 was drawn under the assumption of a uniform distribution for a random variable, r G [0.035,0.1]. If Tis a random variable instead, we still get qualitatively the same result.
2.3 Endogenous Fiscal Deficit and Volatility: Polarization
23
Fluctuations of Government Spending
Fig. 2.4. Thefigureshows fluctuations in government spending in response to shocks to net tax revenue for different degrees of polarization, 0 (theta), where government spending is given by 'g{t : Ö,p) = (1 + 6){T — rbo)e~^'^*. We have used r = 100 and 6o = 0, and assumed that r is a random variable following a uniform distribution o n r € [0.035,0.1]. and Perotti (1997). For example, during a boom (recession), the fiscal spending rises (falls) more than tax revenue does, causing a deficit (surplus) over this period. This procyclicality can be explained in our framework. Suppose that the tax revenue is no longer a lump-sum tax but an income tax, T = tY, where t is a fixed tax rate and y is a total output of the economy. If Y and T rise during a boom in a country with high polarization, total spending p rises more than proportionally, yielding a procyclical pattern of fiscal spending.^"* 2.3.2 Fiscal Policy with a Social Planner In this section, we establish that a balanced budget is the social optimum. A social planner's objective is to maximize both ministers' welfare with respect to Qi and g2, subject to the government budget constraint equation (2.14). As in Sect. 2.3.1, we confine the strategies, p^, to linear Markov strategies. Then the social planner's problem is to maximize the following objective function W{xi,X2) with respect to xi and X2, subject to equation (2.18):^^ ^^ Alternatively, Talvi and Vegh (1996) argue that procyclical fiscal policy can be optimal if there is greater political pressure for higher government spending with rising output levels. ^^ The social planner's objective function is obtained by adding these two ministers' utility functions.
24
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization /•OO
W{xuX2)=
/
[(Ai + A 2 ) M x i ) + ( 2 - A i - A 2 ) M X 2 ) + 2^t]e-^*o^t, (2.24)
where it is assumed that the social planner's discount rate equals the interest rate r. The social planner's solution to this optimization problem can be computed in a way similar to each minister's maximization problem. Appendix B shows that the solutions are ^ Al -f A2 , * 2 — Al — A2 Xi = — ^ — and X2 = 2 •
,- --x ^^ ^
It is clear that the ministers' feedback Nash equilibrium is not socially optimal when compared to the above solution, equation (2.25). It is only when two ministers have the same preferences, that is, Ai = A2, that the feedback Nash equilibrium is socially optimal. Since A2 < | < Ai, it is straightforward to see that Xi ^ Xi^nd X2 ^ X2- That is, for a given size of revenue R, the social planner's optimal spending level is always lower than the non-cooperative feedback Nash solution, except when Ai = A2 = | (i.e., no polarization, 0 = 0). Also, the social planner's solution requires that the government budget balance all times. This can be easily checked. Since the first-best allocation of public goods is gl = ^^^^R and g^ = ^~^^~^^i^, gl-\-g^ = R and b = 0, Vt > 0; therefore, each public good is produced depending on preferences, A^, such that the total expenditure on public good provision is the same as the government net revenue. The smaller the polarization between two ministers, the closer to social optimum the decentralized solution. Finally, if they have the same preferences for public goods, their decentralized solution is sociallyoptimal. It is also important to note that the first-best solution is not that each policymaker simply splits the net tax revenue in half (i.e., xl ¥^ \ unless Al = A2 = | ) . Rather, it depends on how policymakers value one type of public good relative to the other. 2.3.3 Policymakers Faced with Political Uncertainty So far, we have assumed that policymakers' subjective discount rate is equal to the interest rate. We relax this assumption on policymakers' time preference and discuss how this affects our previous results. As will be shown below, the qualitative results and their positive implications remain the same. Yet it introduces a separate channel that works toward generating fiscal deficits and volatile fiscal outcomes, which yields richer yet distinct implications for fiscal dynamics and the capital accumulation process. To see this, we allow the possibility of p 7^ r, and specifically introduce the political uncertainty that faces the policymakers. Let us suppose that they face a constant probability of being removed from office per unit time, p > 0. This uncertainty of tenure effectively increases the ministers' subjective discount rate. We can interpret the subjective discount rate p as the sum of
2.3 Endogenous Fiscal Deficit and Volatility: Polarization
25
real interest rate r and the probability of being removed from office p. We may then write the objective function of minister i, equation (2.13), as /•OO
/ [Xilog{git) + {l-Xi)log{g2t)]e-''dt, where p = r-^p. (2.13') Jo Minister i's feedback Nash equilibrium level of public good provision becomes V^=
Xl = Ai^, X2 = (1 - Aa)^; and gt, = ^Rt,
^-^—^Rf
(2.26)
bt = ( A i - A 2 + l - - ) ^ ( r - r 6 ) = ( ^ + l - ! : ) ^ ( r - r 6 ) > 0, if p > ^ .
(2.27)
gl, =
Thus, the government budget flow equation is
For a moment, suppose that there is no polarization between two ministers (9 = 0), but that there is a constant probability of being removed from office (p > 0). It is clear from equation (2.27) that if the ministers are impatient enough (p > r), an endogenous fiscal deficit can occur even in the absence of polarization between them (^ = 0). The more impatient the ministers, the bigger the size of deficit today. The political uncertainty, as reflected in a greater subjective discount rate, means that the eflFective horizon of ministers is shortened.^^ With a shortened expected tenure in office, the ministers fail to fully take account of the costs of additional debt, and hence would have greater incentives to spend more today. As their horizon shortens, the deficit bias grows even larger. It is worthwhile emphasizing that polarization and political uncertainty are two separate forces working toward generating fiscal deficits. For a given level of the time preference p, the deficit is larger when ^ = 1 relative to when ^ = 0. It is also clear that the size of deficit is bigger when the policymakers are more impatient as measured by time preference p for any given degree of polarization 9. Figs. 2.5 and 2.6 illustrate this point. Fig. 2.5 shows the time path of primary surplus in the absence of polarization (^ = 0), whereas Fig. 2.6 displays it in the presence of polarization {9 = 1). Thus, the fact that the government budget is a common property that policymakers can jointly exploit does not automatically causes a fiscal deficit. Fiscal deficit cannot occur without either preference polarization or impatience. In other words, they are the critical conditions for the dynamic negative externality to be operative in a common pool context. ^^ With the constant probability assumption (say, exponential distribution), it can be shown that the expected time until removal from office is - , which can be thought of the effective time horizon of policymakers. As the probability increases, it shortens the horizon. On the other hand, as the probability p goes to zero, the horizon becomes infinite.
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
26
Primary Deficit in the Presence of Political Uncertainty
25
Time
Fig. 2.5. Primary surplus is given by T-g{t: d,p) = r - ^i^^(r-r6o)e~^^"'"^~?^^*. The figure displays time paths of primary surplus in the absence of polarization (6 = 0) for different values of the discount rate p (rho), where is is assumed that r = 100, 6o = 0, and r = 3.5%. Fiscal Deficit In the Presence of Political Uncertainty and Polarization 100
^
0 rtio=0.035 -100
^
rho=0.05/y/
Z
^
-
(0 ^ -200
e-
rho=0.l/
_
CO
ffl -300 Q.
-j
1 ^00 E
0)
(3
-500 ^
-600
-700
/ rtK>=0.15
J H
J
Fig. 2.6. Primary surplus is given by r-5^(t : l9,p) = T-^i±^(T-r6o)e"^^'^^"?^^*. The figure displays time paths of primary surplus in the presence of polarization {0 = 1) for different values of the discount rate p (rho), where is is assumed that r = 100, 6o = 0, and r = 3.5%.
2.4 Endogenous Fiscal Deficit and Growth
27
Similarly, the uncertainty of tenure contributes further to the volatility of fiscal outcomes at times of shocks to revenue. For a given shock to revenue A r at an instant t—, the change in spending is now Agt-\- = (1 + 9)pAr/r, whose size will be more than proportional to the shock if either p > r or 0 > 0 (and p/r is not small enough to make (1 + 0)p/r less than 1!) or both. Note that if /) = r, Agt^/Ar is back to equation (2.23). That is, the uncertainty of tenure only reinforces the polarization effect on the fiscal instability and vice versa. The social planner's solution still requires the budget to be balanced as long as the social planner's discount rate equals the interest rate r. Only when there is neither polarization nor any difference between the ministers' subjective discount rate and the interest rate does the ministers' non-cooperative feedback Nash equilibrium coincide with that of the social planner. If we compute the social planner's solution, assuming that the social planner's discount rate is p (which may or may not be equal to r), then we obtain
Xl = ( ^ l i - ^ and x2 = ^'"V"'^"-
('•'«)
And the fiscal debt accumulation equation under the social planner's solution becomes 6^ = (1 - - ) ^ ( r - rb) > 0, if p > r, (2.29) with equality if p = r. We can easily check that if the social planner's time preference rate is equal to the real interest rate r, a balanced budget remains the socially optimal fiscal outcome.
2.4 Endogenous Fiscal Deficit and G r o w t h We return to capital accumulation and growth in the decentralized economy with the optimizing private sector and endogenous fiscal policy jointly but non-cooperatively determined by two ministers. Here we are interested in how polarization is linked to capital accumulation and the growth process through the fiscal instability channel that was described in the previous section. First, if there is neither polarization nor political uncertainty, the government budget would be balanced. We will then have the same equilibrium condition and capital accumulation path for the economy as we saw in Sect. 2.2. Second, when there is polarization of preference on the composition of the public goods between ministers, a fiscal deficit arises endogenously. We show below that debt accumulation has negative effects on the capital accumulation of the private sector. This, in turn, affects output level and transitional dynamics of economic growth. Indeed, the capital stock and output at each point in time will be permanently lower if a fiscal deficit occurs due to polarization. In the presence of polarization, the economic growth rate is also reduced along
28
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
the transition path to the new steady state with a lower level of output, compared to that in the absence of polarization. Political uncertainty, as reflected in the discount factor of policymakers, also has similar effects on output and the growth process. Later, we characterize the transitional dynamics of growth as a function of polarization and political uncertainty. We first characterize capital accumulation and output in the presence of polarization under the assumption of p = r. It is straightforward to extend our discussion into a more general case such as p 7^ r, as becomes clear later. If we substitute the solution g^ and pj into equation (2.10) and solve the first-order differential equation for k[^, imposing the NPG condition, we obtain j^FD ^ ir-rbo)^.ert * r
^ 2co^ir-p)t ^ (r-rbo)^_e^, p r
2co p
^
^
where the superscript FD stands for fiscal deficit and p = r is assumed as a benchmark. We observe that capital stock is negatively associated with the degree of polarization, and the greater the degree of polarization 9, the smaller the capital stock kt. Also, higher polarization implies a lower level of output y. It is because greater amounts of disagreement lead to larger debt accumulation. This, in turn, reduces the share of output used for capital formation in the economy where there are only two assets: capital and government bonds. In other words, the fiscal authority controlled by two ministers wastes resources and overspends on each minister's favorite public good provision above its socially eflficient level. This causes inefficient capital accumulation and permanently lowers the output level. We can directly show that when the government runs a fiscal deficit, the capital stock is lower than under the balanced budget. A little algebra shows that kfB - fcf P = (^ - ^^0) (1 - e-"-*) > 0, (2.31) with equality if Ö = 0 (i.e., ministers have the identical preference), where the superscript BB stands for a balanced budget and k^^ is the level of capital stock under a balanced budget (see equation (2.11)). In the presence of polarization, kf^ > kf^. The higher the degree of polarization is, the larger the current gap between kf^ and kf^ is (see Fig. 2.7). As t —^ 00, kf^ — kf^ approaches ^^~^ ^K Thus, in a polarized society, both the capital stock and output will be permanently lower while the gap between kf^ and kf^ rises with the degree of polarization.^"^ We can gauge the size of the impact of an increase in polarization on social welfare in terms of consumption, using equations (2.9), (2.11), and (2.30).^^ Note that an increase in polarization 0 from zero to one is associated with a permanent reduction in k at the steady state by the amount of i l z l ^ . 2^ As for output, yf ^ - yf^
= ^^"7^°^(1 - e"^^*) > 0, for Vt > 0, with equality if
^^ I thank an anonymous referee for suggesting this exercise.
2.4 Endogenous Fiscal Deficit and Growth
29
On the other hand, the consumption is given by Ct = p • {kt — ^^-~^) from equations (2.9) and (2.11). A one unit decrease in capital stock Ak amounts to a reduction in consumption by pAk. Therefore, the consumption level at the steady state is permanently reduced by the amount of ^^^"^ "^ when the degree of polarization rises from zero to one. This welfare cost of reduced consumption due to polarization effect can be substantial. According to the average figures of the OECD countries in the period of 1987-2003 from OECD (2004), taxes are 38 percent of GDP, while net debt interest payments are 3 percent of GDP. Assuming p = r, therefore, the size of consumption reduction at the steady state would amount to 35 percent of GDP! 1
thela=1
/
-
•5.1000
- //
1
1
1
1
^"^X^^^^'^^^ / theta=a6
/ /
1
-
y ^
theta=0.3
/
-
1 / theta=0
1
1
1 200 Time
Fig. 2.7. The figure compares capital stock gaps between under a balanced budget and under a fiscal deficit due to various degrees of polarization, 0 (theta): k BB j^FD _ (r-rbp) It is assumed that r = 100, 60 = 0, and r = 3.5%. Note (l-ethat the term "balanced budget" has been abbreviated as "BB," and "fiscal deficit" has been abbreviated as "FD." Transition dynamics of growth depend crucially on the degree of polarization. When there is polarization {0 > 0), the growth rate of capital stock is also lower than that in the absence of polarization {0 = 0) for all finite periods of time. Respectively, the growth rates when ^ = 0 and 9 > 0 are given by
(r-p) (^)l.=o = ^ e - ( - p ) * + l = 0 and
(2.32)
30
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
i..n,
^ e - « « + 0 - . ) * + (,_^) r2co
^
•$rt
r2co
where 6o = 0 is assumed for simplicity, and the last term in each equation above is obtained under the assumption of p = r. It is clear from equations (2.32) and (2.33) that (klk)e^Q > (A:/A;)0>o, for all finite time t > 0. Over time, the growth rate of capital stock in the presence of polarization (9 > 0) converges to the rate of growth under a balanced budget, r — p, whereas the level of k itself is permanently lowered. That is, the asymptotic growth rate of capital is r - p: limt-.oo(|)^^ = r - p (see Fig. 2.8).'^^ Interestingly, however, the relationship between the growth rate and the degree of polarization for the economies with polarization (6 > 0) is not monotonic. A more polarized economy would experience a more dramatic change in its economic growth rate for a given period of time. For example, an economy with a higher degree of polarization 9 would see a sharper collapse of growth rate initially as it runs a larger deficit. However, this same economy would grow more rapidly later as its fiscal balance improves compared to an economy with a lower degree of polarization 6 (again Fig. 2.8). This reflects the fiscal deficit dynamics in relation to polarization as illustrated in Sect. 2.3.1. It is straightforward to see that the above results still hold even when policymakers are impatient enough to discount the future more heavily (i.e., p > r). So far, we implicitly assumed that the policymakers still share the same time preference rate p with the private sector, even when we allowed the case of p j^ r. We can further distinguish them by assuming that policymakers faced with political uncertainty discount the future more heavily than the private sector—^that is, p > p, where the policymakers' discount rate is p = p + p; p is the private agent's discount rate (which may be equal to r or not); and p is the constant probability of policymakers being removed from the office. Even in this case, the result still remains qualitatively the same. The transitional ^^ Since the marginal productivity of capital stock is a constant A (and hence the value of r—p is constant), it can produce perpetual growth without assuming some exogenous technological progress. However, the transitional dynamics of growth as a function of polarization 6 looks similar to that of neoclassical models such as the Solow model. For example, a change in saving rate in the neoclassical model can lead to a permanent effect on the level of output, whereas it does not change the steady-state growth rate. Similarly, a change in the degree of polarization 0 in our model leads to a permanent change in output and capital stock, but not to a permanent change in the steady-state growth rate. Thus, our model can explain why some nations are rich and others are poor, while the differences in growth rates across countries can be explained by appealing to the transition dynamics. From the empirical point of view, this interpretation that differences in growth rates across countries are due to the fact that countries are on different transition paths to their own steady states of output works reasonably well. For seminal empirical papers, see Mankiw et al. (1992), Barro (1991) and Barro and Sala-i-Martin (1992) among others.
2.5 Concluding Remarks
31
Fig. 2.8. The figure depicts the growth rate of capital stock, ( | ) ^ ^
=
'^''^———r < 0, for different degrees of polarization 6 (theta) under the assump+1] tion of r = 100, 6o = 0, 2co = 1, and r = p = 3.5%. dynamics of growth with respect to the probability of losing office p is similar to that with respect to polarization 6, whereas the steady-state growth rate is still r — p. To see this point, note that the government budget flow equation (2.27) becomes 6=(l + ^ - I ) ^ ( r - r 6 ) > 0 , if?>
j ^ ,
(2.34)
and that capital accumulation takes place according to kt =
^Jr-p)t
, jr-rbo)
+
j^_^^t
^
(öp+(i+ö)p)t
(2.35)
We can then clearly see that an increase in political uncertainty as reflected in p leads to a permanent reduction in output and capital stock, but not to a permanent decrease in the steady-state growth rate {r — p).
2.5 Concluding R e m a r k s This chapter has presented a dynamic model of fiscal policy in a simple growth framework where social polarization of preferences among socio-economic groups plays a central role in the evolution of fiscal instability and growth collapse. One key feature of the chapter is that the size of fiscal deficit, the
32
2 Fiscal Deficit, Fiscal Volatility and Growth: Social Polarization
magnitude of fiscal volatility, and the size of reduction in output and growth rate are explicitly shown to be increasing functions of the degree of social polarization and the degree of political uncertainty (reflected in policymakers' subjective discount factor). Broadly consistent with recent empirical studies on fiscal instability and growth, our results can particularly contribute to explaining why many Latin American countries in the past decades have suflFered from chronic fiscal deficits, volatile fiscal outcomes, procyclicality of fiscal policies, and disappointingly poor growth. Our theory suggests that all of these problems can be ultimately attributed to the polarization within the government or among socio-economic groups in a society. Although social polarization might be among the oldest ideas in the political economy literature, we first theoretically demonstrate how social polarization can cause the aforementioned fiscal and growth problems. Furthermore, we fully distinguish the incentive for policymakers to engage in short-term policies under political uncertainty from that under social polarization of preferences. Not only do we show that these two are separate and distinct forces driving fiscal dynamics and growth process but also that these are critical to the coordination failure (that is, the negative dynamic externality) in a more general two-player common pool game. On the other hand, recent studies on budgetary institutions have presented evidence that stringent budgetary procedures and rules can directly influence fiscal outcomes (see Alesina and Perotti 1996 for a literature survey).^^ However, our theory suggests that social polarization may lie in depth behind the fiscal problems. So whether institutional arrangements (including budgetary institutions) can be made to mitigate the negative fiscal eflfects from social polarization remains to be an important question. In this regard. Woo (2003a) reports encouraging econometric evidence that the eflPects on public sector deficits of the social polarization tend to be smaller in countries with better (budgetary) institutional arrangements. Conversely, the social polarization has very strong effects on deficits in the presence of poor institutions. This also has important policy implications. For example, the fiscal decentralization that many countries have recently undergone may produce diflferent results depending on the underlying social polarization and institutional factors. In a highly polarized country, decentralizing power to local governments may only increase tension among central governments, regions and different groups, and can threaten macroeconomic stability unless it is checked by proper institutional restraints. This has been observed in some countries such as the Balkans, Indonesia, and Brazil in recent years.^^ ^° See von Hagen (1992), Hallerberg and von Hagen (1999), and Alesina et al. (1999) among others. ^^ Rodden et al. (2003) provide deatiled country case studies on various issues related to fiscal decentralization for eleven nations. In many countries, the fiscal decentralization often led to difficulty in maintaining fiscal discipline.
2.6 Appendix
33
2.6 A p p e n d i x A. Derivation of the decentralized feedback Nash equilibrium. Minister I's Hamiltonian is H\xuX2,i^t)
= [Ai/opxi + (l-Ai)/opx2+i/'t]e-"*+/iit[r-rxi-rx2]. (2.36)
The first-order conditions are Hl^ = —e-^* - fiir = 0; Hl = e"^* = - / i i ;
lim fi{t) = 0
(TVC).
(2.37) Using the transversality condition (TVC), we solve the first-order differential equation and we can get —rt flit) = -—. (2.38) Thus, the solution is Xi = Ai. Similarly, we can solve the optimization problem of minister 2, getting X2 = (1 — A2). Note that the first-order condition for the control variable, xii and the solution of fi{t) do not contain the state variable ip. This makes the open-loop and feedback strategies coincide. B. Derivation of the solution of social planner By using a social planner's objective function and the government budget constraint, we set up the following Hamiltonian function: = [(Ai+A2)/opxi+(2-Ai-A2)/opX2+2i/'t]e"''*+/xt[r-rxi-rx2]. (2.39) The first-order conditions are H{xi,X2,ipt)
Al + A2 _j.t ^ jj. 2 — Al — A2 _j.+ „ ,^ .^v e "^^ - /j,r = 0; (2.40) e '^'^ - fir = 0; H^^ = .-rt H^ = 2e-^' = - / i ; and lim fi{t) = 0 (TVC).
H^^ =
t—^00
Using the transversality condition (TVC), the first-order differential equation is solved to get ß{t) =
.
(2.41)
Thus, the solution is ^ Al -f- A2 , * 2 — Al — A2 Xi = —1^— and X2 = k -•
/r^ ^«s (2-42)
Inflation, Composition of Deficit Finance, and Social Polarization
3.1 Introduction "Milton Friedman's famous statement that inflation is always and everywhere a monetary phenomenon is correct. However, governments do not print money at a rapid rate out of a clear blue sky. They generally print money to cover their budget deficit. Rapid money growth is conceivable without an underlying fiscal imbalance, but it is unlikely. Thus, rapid inflation is almost always a fiscal phenomenon." from Fischer and Easterly (1990, pp. 138-39)
The previous chapter has developed a political economy model of fiscal policy in which social polarization plays a crucial role in driving macroeconomic outcomes such as fiscal deficits, fiscal volatility and poor growth. Now we extend the model to study inflationary consequence of the fiscal policy set by polarized policymakers by introducing money and inflation tax into the economy. The relation between fiscal deficits and inflation has long been a subject of economic research. As the above quote from Fischer and Easterly (1990) suggests, the fiscal consequences of the debt crisis seem to be the initiating disturbances in many countries that have suffered extreme inflation in the past decades (Dornbusch et al. 1990; Fisher et al. 2002). More recently, Catao and Terrones (2003) reaflfirm the positive association between deficits and inflation among high-inflation and developing country groups in a panel of 107 countries over the 1960-2001 period, although not among low-inflation advanced economies.^ The inflationary consequences of fiscal deficits are often more severe in developing countries because of the lack of central bank independence from political pressures to extend cheap credits or help to finance ^ Gosh et al. (2002) also report a positive and significant effect of fiscal deficit on inflation in a panel of 147 countries over the period 1970-99. Alfaro (2005) finds similar evidence in a cross-section of 130 for the period 1973-98, although the result becomes less significant in the annual panel data for the same period.
36
3 Inflation, Composition of Deficit Finance, and Social Polarization
budget deficits, and the lack of sufläciently developed domestic capital markets that can buy newly issued governments debt (see Chap. 5 and Caballero and Krishnamurthy 2004 for evidence on fiscal deficits and financial market development). In this chapter, however, we emphasize the role of social polarization as a fundamental force that may drive the relation between fiscal deficits and inflation. We derive a result that if there is a high degree of polarization and a large deficit arises as a result, there will be high steady-state inflation once the government turns to money financing. Moreover, the higher the degree of polarization, the higher the steady state inflation rate. Based on Israel's high inflation experience in the mid 1980s, Patinkin (1993) provides a supportive observation for what our model attempts to capture. He regards high inflation as resulting from the prisoners' dilemma in the coalition government. Out of fear that her position, relative to that of other minister may be worsened, each minister insists on her increased nominal budget. The overall budget, whose planned expenditures exceed its expected revenues, then generates inflation. As we saw in the previous chapter, when there is greater disagreements on the ideal composition of government spending and therefore high degree of polarization, each policymaker will have a greater incentive to insist on greater spending for her favorite type of spending program or public good, leading to overspending that causes inflation via monetary financing.^ Fig. 3.1 shows the scatter plot of inflation tax over the period of 1970-2000 against income inequality measured in 1970s (as a proxy for social polarization).^ A positive correlation is quite discernible. Indeed, several empirical studies have reported evidence of a positive correlation between income inequality and inflation over the period of 1965-1990 (Sachs 1989; Beetsma and Van Der Ploeg 1996; Al-Marhubi 1997). Desai et al. (2005) extend the sample period through 2000 and find a significant positive relation between the two variables in countries with democratic political systems in a panel of 120 countries. Also, there is a unconditional, positive and significant relation between inequality and inflation in the OECD country sample.^ Aizenman (1990) studies the consequences of inflationary finance when the inflation tax is determined by several policymakers, each of whom can print money via the central bank. He suggests that the failure of the central bank to impose the needed discipline with friction among the various policymakers may be responsible for the recent inflationary episode (Yugoslavia and Russia in the early 1990s). Cukierman et al. (1992) find that the level of inflation in a cross-section of countries is related to measures of political instability. Inflation tax is measured by 7r/(l4-7r), where TT is the inflation rate that is obtained from World Bank (2005). The income inequality indicator AGINI70 is the average of all available Gini coefficients in the 1970s in Deininger and Squire (1996) data set. Beetsma and Van Der Ploeg (1996) develop a theory of inflation that arises from distributional conflicts, which is similar to the reasoning in Alesina and Rodrik (1994). A democratic society in which the wealth distribution is unequal elects
3.1 Introduction
37
Argentin
.998283 IsraelUganda
Uruguay
funcey Venezuel
Costa Ri
ace
Dominica Kenya
Egypt Paldstan
South Af I
Sri Lanic
India United K Funiand
Norway Denmaric
Malawi
Jamaft^^'"^'^
Trinidad New Zeal
Barbaaos Fiji
Austraii traii Iwedei Sweden
Senegal
Thailand
Fr^niWglade
Hong Kon Belgium
.768751
Switzerl Germany
28.1
Singapor
AGINI70
65.38
Fig. 3.1. Inflation tax and income inequality. Interestingly, however, we show that inflation dynamics depend on the money-bond ratio in the deficit finance and an increase in that ratio is less inflationary for a given degree of polarization. That is, more reliance on moneyfinancing relative to bond-financing can be less inflationary. This paradoxical result is consistent with earlier studies by Turnovsky (1978) and Liviatan (1983). However, these studies derive the result for an exogenously given government spending. More important, our explanation is very diflFerent from theirs. The reason for our result is that as each policymaker tries to finance her expenditure through inflation tax, this increases the rate of inflation which, in turn, reduces the private agent's real money demand, tax base. Therefore, they would not be able to spend as much as they would with pure debt-finance. As a result, money-financing provides more discipline to these polarized ministers. Therefore, it follows that the money-financing can less inflationary if the government relies more on inflation tax relative to bond-financing from the beginning. In other words, bond-financing of fiscal deficit may not cause inflation initially. Yet as government debt accumulates as a result, it further
political parties that are likely to represent the interests of poor people. It is in the interests of their clientele of the resulting governments to attempt to levy inflation taxes in order to erode the real value of debt service and redistribute from the rich to the poor. Consequently, inequality and high levels of nominal government debt result in inflation in more democratic countries due to populist pressures for redistribution.
38
3 Inflation, Composition of Deficit Finance, and Social Polarization
raises the potential inflation that would prevail when the government finally has to turn to seigniorage.^ More recently, Tornell and Velasco (2000) make a similar reasoning in a sharply diflFerent context in which they compare fixed exchange rate with flexible exchange rate regime for their fiscal implications. Conventional wisdom holds that fixed rates provide more fiscal discipline than do flexible rates. The claim that flxed rates induce more discipline stresses that sustained adoption of lax fiscal policies must eventually lead to an exhaustion of reserves and thus to a politically costly collapse of the peg. Hence, under fixed rates irresponsible fiscal behavior today leads to punishment tomorrow. Under flexible rates irresponsible fiscal behavior has costs as well. The diflPerence is in the intertemporal distribution of these costs: flexible rates allow the eflPects of unsound fiscal policies to manifest themselves immediately through movements in the exchange rate (and inflation). Hence, under flexible rates irresponsible fiscal behavior today leads to punishment today. They show that if fiscal authorities are impatient, flexible rates by forcing the costs to be paid up-front provide more fiscal discipline and higher welfare for the representative private agent. Despite the diflFerent mechanism at work, the implication of their result is similar to ours in that monetary financing (under the flexible exchange rate regime) may provide more fiscal discipline (than does fixed rate regime under which the budget deficit cannot be financed by monetary financing as long as the government is credibly committed to the fixed rate). The plan of the chapter is as follows. Sect. 3.2 describes the economy with money, and derives money demand of the private sector. Sect. 3.3 derives our main result on inflation, fiscal deficits, and social polarization. Sect. 3.4 concludes.
This is reminiscent of the argument of Sargent and Wallace (1981) that bondfinancing may be more inflationary for a given deficit, although the reason is very different. They show that when the government deficit is held constant, a temporary tightening monetary policy may eventually lead to higher inflation. This is because tight current monetary policy implies greater bond issuance and growing debt service to finance the given deficit. Related to this, Blanchard (2004) makes an interesting argument regarding the inflationary effects of inflation targeting in an open economy setting. If high inflation triggers a tight monetary policy stance under an inflation targeting that results in an increase in the real interest, it may lead to a real depreciation and to a further increase in inflation. It is because the increase in the real interest rate increases the probability of default on the government debt, which may cause a real depreciation. In this case, fiscal policy, not monetary policy, is the right instrument to decrease inflation. He argues that this is the case of Brazil in 2002 and 2003.
3.2 The Economy
39
3.2 T h e Economy The economy is populated by a government and a private sector composed of two groups, indexed by i, i = 1,2, as in the previous chapter. Each group consists of a large number of atomistic individuals. The government and the private sector have perfect foresight. Now we incorporate money into the lifetime utility by considering a Sidrauski type utility function as follows: [log{ci) + ^logirrii) + Xilogigi) + (1 - Xi)log{g2)]e-'''dt,
(3.1)
^0
where ci is private consumption; P the price level; m = yr the real money balance; and 7 > 0. We assume that the private discount rate p is equal to the real interest rate r, and focus on the inflationary consequences of deficit finance solely from social polarization (that is, there is no political uncertainty either because p = r and hence p = 0).^ As previously, gi and p2 ^ire two different public goods provided by the government. Being small, each member of group i takes gi as given and has the same preference for the two public goods within the group. Yet these two groups may differ in their preferences for the public goods as reflected in A^ G [0,1]. We continue to assume that A2 < | < Ai, which implies that group 1 prefers gi to p2 and vice versa. An agent can hold her wealth in the form of capital k, government bonds 6, and real money balance m. The bonds are assumed to be a perfect substitute for capital and therefore to pay the same rate of interest r. The flow budget constraint of the agent of group i is now for Vt > 0 and ÜQ > 0 äit = rau - Cit - (r + 7rt)mit - n,
(3.2)
where an is the asset held by an agent; an = kn -h hn + rriit ; and r^ is a lump-sum tax collected by the government from group i. The utility maximization subject to the above budget constraint and NPG condition yields the solution Ct = c, and rrit = 7(r + 7rt)~\
(3.3)
where TTt is an inflation rate. Consumption is constant since the private discount rate is assumed to be equal to the real interest rate r. Before we move to government fiscal policy, we make an additional assumption on the timing structure. Following Cohen and Michel (1988), we assume that the fiscal authority moves before the individual at each point of ^ Since the discount rate is assumed to be equal to the interest rate, then consumption will be constant over time as one can recall from a private agent's life-time utility maximization problem in the previous chapter. This makes it easier to focus on the inflationary consequences of endogenous fiscal policies, which is the main purpose of this chapter. This separability assumption will render the behaviors of the financial sector completely independent of the production sector.
40
3 Inflation, Composition of Deficit Finance, and Social Polarization
time and is able to precommit for an instant of time. Noting that the private consumption is independent of each minister's public good provision decisions, and taking into account the money demand of the private agent, each minister chooses her own public good provision level, g*, which is the Markov perfect equilibrium. The fiscal authority moves first, announcing its gi and p2 obtained by solving the optimization problem subject to the budget constraint and to the response of the private agent, Ct and rut. The agent takes the government spending as given and solves for Ct and rrit so that she can maximize her life-time utility subject to her budget constraint.^
3.3 Inflation, Composition of Deficit Finance, and Social Polarization Each minister i has the following objective function: /•CX)
V' = /
[Xilog{git) + (1 - Xi)log{g2t)]e-'^*dt,
(3.4)
where 0 < A 2 < | < A i < l ; and the minister's discount rate is assumed to be equal to the interest rate r. Minister i prefers gi to gj, j ^ i and i, j = 1,2, which puts more weight on her favorite public good, pi, in her objective function. As in the previous chapter, we define ^ = Ai ~ A2, and interpret it as a degree of polarization between ministers (equivalently between two groups). Each minister would maximize her objective function subject to the relevant government budget constraint and the private agent's real money demand function (3.3). We assume that money does not enter each minister's utility function because it is only used to finance her public good provision, gi. Thus, the money supply is indirectly determined by the ministers' spending levels, gi. Once the government finances its spending by seigniorage as well as debtfinancing, the relevant government budget constraint becomes b-\-m = gi-\-g2-hrb
— T — 7rm,
(3.5)
where irm is the inflation tax. Using the money demand equation (3.3), we can get irm = 7 — rm. Note that as inflation rate goes up, the inflation tax increases. But there is a maximum of inflation tax revenues that can be collected by printing money because money demand depends negatively on the nominal interest rate.^ ^ Here we do not consider the distinct role of the central bank explicitly by assuming that the fiscal authority controls the central bank for simplicity of analysis. There is no distinction between the central bank and the fiscal authority. ^ To see this, we compute the first and second derivatives of 7rm with respect to TT as follows:
3.3 Inflation, Composition of Deficit Finance, and Social Polarization
41
Since we are primarily interested in inflationary consequences of deficit finance and in the relationship among inflation, composition of deficit finance, and degree of polarization, we consider the following experiment. The fiscal authority begins by using both debt-finance and money-finance at a given ratio of money and bonds (^) and then turns to pure money financing at time T in the future, either because there is a maximum level of debt service or because the fiscal authority has no other way than inflation tax to continue to finance its expenditures.^ The maximum level of debt service is determined by the lump-sum tax r in our model.^^ Let us start with the simplest case of ^ = 0. That is, the government initially uses pure bond-financing before switching to pure money-financing. Also, let us assume, for simplicity, that the money supply is constant at the level 7r~^ before the switch in order to guarantee money market equilibrium. This switch takes place at time T when the fiscal authority changes its financing method so as to freeze the debt at its then current level bx, and hence 6 = 0 from time T. Thus for alH < T government debt is increasing, money balances are constant, and so is private consumption. After T the government debt is not increasing, and money balances are decreasing as the government begins to finance its expenditures by printing money and generating inflation. The Mt = 0 for t < T ; 6t = 0 for t > T ; and the government budget constraints are thus
dTTt
(r-f-7rt)2
but the second derivative is negative. That is, dTT?
(r + 7r03
'
^ There is another incentive to turn to inflationary finance. When the government has issued a large debt, the government might want to reduce the burden of debt by inflating away the debt. This incentive is studied by Missale and Blanchard (1994). They derive the maximum maturity of debt in which this incentive to inflate the debt burden is restrained by the reputational consideration. 10 Temporarily, the government might not be able to finance its spending by printing money, so the government relies on pure bond-financing. For example, the central bank often pursues an independent monetary policy by not accommodating the fiscal policy. Empirically, however, most independent monetary policy regimes are of short duration in practice. See Klein and Marion (1994). Dornbusch et al. (1990) and Drazen and Helpman (1990) also study the inflationary finance and the government fiscal deficits. Specifically, Drazen and Helpman (1990) discuss stabilization policies and inflation, and study how expectations of a policy switch affect economic dynamics before the switch, whose timing or mix among expenditure cuts, tax increase, or increases in money growth rates may be uncertain.
42
3 Inflation, Composition of Deficit Finance, and Social Polarization b = 9i-\-g2-^rb-r, for t < T
(3.6)
m = pi + p2 + rbr — r — Trm, for t>T. We can solve for the endogenous g* by maximizing the minister's utility subject to equation (3.6) and to the response of private agent, equation (3.3). The feedback Nash equilibrium gi for t > T is p^ = Ai(7rm+r—rör) and P2 = (1 — A2)(7rm + r — rbr)- Substituting this solution into the second equation of the government budget constraint equation (3.6) yields the following equation: m = 6>(7rm + r - rbr), for t > T.
(3.7)
We can see that the change in real money balance will be greater as the degree of polarization increases. Note that the dynamics of m around its steady state is unstable (i.e., the coefficient of m in equation (3.7) is On > 0). The unique convergent perfect foresight path includes an immediate jump to the new steady state level: TT^H- = TT*. Furthermore, the higher the degree of polarization 0, the higher the steady state inflation rate that is reached after the switch, which we show below. Since m = 0 in the steady state, by using equation (3.7) and m = 7(r + 7r)~^ in equation (3.3), we can get the steady sate inflation rate TT*: ^'"'^ =rbT-T. r-t-TT
(3.8)
From the above steady state inflation, dn* _ (r -f TT*)^ > 0. dbr 7
(3.9)
The steady state inflation rate is higher if the level of debt at T is higher. Also, an additional bond issuing at T before the switch accelerates the steady state inflation that would prevail after the switch. One can show that the government debt bx at time T should be equal to (assuming 6o = 0)^^ (3.10) br = - - -e-'^^. r r Combining equations (3.8) and (3.10) and then diflPerentiating TT with respect to 0 yields (3.11) - ^ = (r^ + TrrT) > 0. du
The steady state TT* will be higher if the degree of polarization within the government becomes larger. ^^ The government debt path under the non-cooperative Nash Feedback equilibrium is exactly identical to that in Chap. 2 since for t < T the fiscal authority uses the pure bond financing.
3.3 Inflation, Composition of Deficit Finance, and Social Polarization
43
Next, we consider a more general case that the fiscal authority uses both debt-finance and money-finance at a given ratio ^ of money and bonds (i.e. ^ = :^) at the beginning, and turns to pure money financing at time T. Each minister would determine her public good provision level gi, by solving the maximization problem subject to the government budget constraint and the money demand function of the private sector. Let d = b-\-m. Using the expression for d, we can rewrite the government budget constraint as follows: d = gi+g2 + rd — T — {r-^7r)m
(3.12)
= gi-^g2+rd-T-j. The feedback Nash equilibrium, p^ and pj? is Pi = Ai(7 -h r - rd), and P2 = (1 " ^2){7 + r - rd).
(3.13)
To facilitate the interpretation of the relationship among polarization, fiscal sending and inflation rate, note that d = {b -\- m) = {1 -\- Qb where ^ _ m rpj^^ feedback Nash equilibrium, p^ and pj? is equivalent to g^ = Al(7 4- r — r ( l + f)6) and g^ = (1 — A2)(7 + r — r ( l + ^)6). One can quickly notice that each minister's public good provision g* lessens as the moneybond ratio (f) increases. This can be explained intuitively. Since the private agent's money demand is inversely related to the inflation rate, there is a limit on (real) revenues that can be collected by inflation tax for any given lumpsum tax and money-bond ratio in the deficit finance. As ^ becomes larger, the portion of revenue orginated from inflation tax from which the polarized ministers can extract decreases and their spending declines as well. The steady state inflation after the switch takes place at T can be obtained as follows. Since 6 = 0 at T and m = 7(r-|-7rt)~^, the steady state inflation rate TT* is (recall that the government budget is c? = ^(7 + r — rd) from equations (3.12) and (3.13)) ^.^Jib-_rl_ (3.14) 7 -h r — ro Note that the inflation rate will be higher if the debt at T when the switch takes place is higher. Using ^ = y and the derived g*, the government budget constraint can be written as 6 = ^ ( 7 + T - r ( l + 0^>)-
(3.15)
Since d = (1 + 0^^ d={l-\- ^k We go further by computing the br using equation (3.15) and the NPG condition, U _ (7 + -r) ,^ .^r^ (1-e-^^^). (3.16)
r(l + 0
44
3 Inflation, Composition of Deficit Finance, and Social Polarization
It is clear from equation (3.16) that a higher money-bond ratio in the deficit finance or a lower polarization leads to lower debt. Substituting equation (3.16) into equation (3.14) yields (7 + r ) ( l - e - ^ ' - ^ ) - r r ( l + 0 The steady state inflation rate depends on several variables such as polarization, money-bond ratio, and switch time T. First, we note that there is a monotonic positive relationship between the degree of polarization and inflation rate. To see this, we compute the diflPerentiation of TT* with respect to<9: dn^ -^
(rTe-'rT^(:^ + r-rT){l + 0 v^^[nil ^^^r){^ + e-erT)2 > 0, V ^ 6 [0,1].
('^^9^ (3.18)
Note that in general 0 < r < 1 and hence 7 + r — r r > 0. As in the first case, either the higher the degree of polarization or the higher the resulting debt at time T, the higher the steady state inflation rate after switching to monetizing the debt service. If the government had a higher money-bond ratio at the beginning but turns to money finance later, however, then this economy will have a lower steady state inflation rate despite the polarization within the government. In other words, there is a monotonic relationship between ^ and inflation rate for any given 6, dTT* d^
-(7-hT-rr)(l-e-^^^) < 0, V e > 0, (7 + r)(e + e-^-^)2
(3.19)
where 7 + r — r r > 0. The above result suggests that money-financing can be less inflationary. This result is consistent with earlier studies by Turnovsky (1978) and Liviatan (1983), although they derive the result for an exogenously given government spending.^^ Thus, our explanation is very diflPerent. In our model, it is because as ^ increases, each minister's spending decreases. In a sense, money-financing provides discipline to these polarized ministers. When she tries to finance her expenditure by printing more money, this increases the inflation rate that in turn reduces the private agent's real money demand (i.e. tax base). With a money market equilibrium condition and the resulting limit on the inflation tax revenue, she cannot spend as much as she would with pure bond-financing. Therefore, it follows that the money finance is less inflationary if the government relies more on inflation tax relative to bondfinance from the beginning. In other words, bond-finance of fiscal deficit may not cause infiation initially. Yet as debt accumulates, this also increases the potential inflation that would prevail when the government finally turns to seigniorage. ^^ Calvo (1985) and Liviatan (1988) also study the implications of changing the money-bond ratio in deficit financing for an exogenously given government spending.
3.4 Concluding Remarks
45
3.4 Concluding Remarks We have examined the inflationary consequence of fiscal deficit in relation to social polarization, and derived a positive relation between social polarization and inflation: the higher the degree of polarization, the higher the steady state inflation. This prediction is consistent with empirical evidence on inequality and inflation found in the existing studies, given that income inequality has long been mentioned to be an important source of social polarization in the political science literature. Interestingly enough, however, an increase in the money-bond ratio in the deficit financing turns out be less inflationary for any given degree of polarization because the money financing may provide more fiscal discipline to these polarized policymakers.
Social Polarization, Industrialization, and Fiscal Instability
4.1 Introduction This chapter is motivated by an important puzzle t h a t arises from t h e contrasting macroeconomic experience across developing regions in recent decades. In sharp contrast t o East Asia, much of Latin America and subSaharan Africa has often engaged in unsustainable fiscal policies, leading t o huge fiscal deficits, external debt crises, or hyperinflation (see Table 4.1).^ For instance, Korea's largest public deficit in t h e 1980s was 4.3 percent of G D P in 1982 against Mexico's 15.4 percent of G D P in 1982 or Zambia's 28.5 percent of G D P in 1986. Despite t h e substantial progress made on fiscal reform since t h e late 1980s, many developing countries still suffer from chronically recurrent large deficits. Brazil's recent financial crisis is closely related t o its high budget deficit, which is 8.4 percent of G D P as of January in 1999. (See Newsweek of J a n . 25, 1999.) Moreover, past fiscal policies in Latin America and sub-Saharan Africa are characterized by high volatility of fiscal outcomes such as fiscal balance and its components, compared t o East Asia (see Table 4.1).2 Reprinted from Journal of Development Economics, Vol. 72, October, 2003, pp 223-252, Woo J, "Social Polarization, Industrialization, Fiscal Instability: Theory and Evidence", with permission from Elsevier. Even in the wake of the recent Asian crisis, most economists agree that it was not caused by irresponsible macroeconomic policy but by heavy, private shortterm borrowing, which was compounded by weakness of the financial system (see Radelet and Sachs 1998, for example). East Asian countries hit by the financial crisis in 1997 showed somewhat declining fiscal surpluses or emerging deficits in 1997-98, yet the size of deficits is still modest. It ranges from -1.37 percent of GDP in Korea, -0.87 percent of GDP in Thailand, -0.67 percent of GDP in Indonesia, to 2.99 percent of GDP in Malaysia in 1997, which are the most recent figures available from World Development Indicators of World Bank (2000). Closely related to this volatility is the procyclicality of fiscal policy in Latin America, which has been extensively documented by Gavin and Perotti (1997).
48
4 Social Polarization, Industrialization, and Fiscal Instability
Table 4 . 1 . Comparison of Developing Regions: Fiscal Balance, Volatility of Fiscal Outcomes, and Income Distribution East Asia Latin America Sub-Saharan Africa Fiscal Balance (% of G D P ) : 1970 - 90 Central Government Balance" General Government Balance^ Public Sector Balance*^
-2.34 -2.12 -3.60
-3.80 -4.52 -4.89
-4.68 -5.21 -6.73
Volatility of Fiscal Outcomes: 1970 - 90' Central Government Balance General Government Balance Public Sector Balance Government Expenditures Tax Revenue Expenditure Components Wages and Salaries Purchases of Goods and Services Subsidies and Transfers
2.40 2.75 2.53 2.91 1.80
3.47 4.14 3.78 4.15 2.41
3.53 3.62 4.43 4.69 2.62
3,35 5.53 4.09
5.00 7.71 8.81
5.19 8.20 4.10
Income Distribution around 1970 Gini Coefficient for Income^ Gini Coefficient for Land Ownership-^
38.99 45.20
50.83 81.19
49.45 47.19
Sources: "" World Bank (1999). ^ IMF (various issues). ^ Easterly et al. (1994). '^ Averages of country-specific standard deviations are reported. All but expenditure components are measured as a percentage of GDP, and the components of expenditure are me8isured as a percentage of government expenditure. Data on the central government expenditure components are from World Bank (1999). ^ Deininger and Squire (1996). f Taylor and Hudson (1972).
Why have certain countries, notably in Latin America and sub-Saharan Africa, repeatedly adopted unsustainable fiscal policies in recent decades, while East Asian countries have maintained stable fiscal policies? We argue that social polarization and degree of polarization are key to understanding diflferences in fiscal outcomes across countries. The unsustainable fiscal paths can best be viewed as politico-economic equilibria that arise from social conAlthough we focus on volatility rather than procyclicality, we can shed some light on this issue.
4.1 Introduction
49
flicts of interest over government spending, created by interaction between initial income distribution and industrialization. In an economy with a manufacturing sector and a traditional sector, the more unequal the initial income distribution, the larger the sectoral income gap during industrialization and the more likely the polarization of sector preferences for different types of government spending. In a highly polarized society, each policymaker has a greater incentive to insist on higher spending for her preferred sector.^ A large fiscal deficit results, and the fiscal spending path becomes more volatile. Income inequality has long been mentioned as a major source of social polarization. Many economists have argued that class and sectoral divisions due to unequal income distribution provide an important answer to the question of why populist fiscal policies appear more often in Latin American countries than other regions. (See Rodrik 1996; Kauffman and Stallings 1991; Sachs 1989; Berg and Sachs 1988.) Yet there are very few theories that explain why unequal income distribution can lead to large deficits and volatile fiscal outcomes. On the other hand, a growing literature on the political economy of fiscal deficits emphasizes the importance of government fragmentation and political instability in understanding fiscal deficits.'* While this chapter shares the same spirit in emphasizing the political economy aspects of fiscal policies, it is also related to the tragedy of the commons literature in that the endogenous fiscal policy is jointly determined by heterogeneous policymakers who share common government resources.^ This chapter, however, explores the theoretical linkage between initial income distribution and social polarization along the industrialization process. It then highlights the role of polarization in the development of fiscal instability, deriving new directly testable implications. Moreover, we address both fiscal deficits and volatilities in a single framework, whereas the existing models of fiscal deficits do not deal with the volatility of fiscal outcomes. In our model, industrialization is viewed as the adoption of modern manufacturing technology that combines skill and capital stock (Goldin and Katz 1998). Industrialization provides an incentive to accumulate human capital and enjoy a higher wage. But not everyone can invest in human capital because education is costly. The agents who mainly benefit from the industrialization are those who are already rich enough to invest in human capital. Assuming the human capital externality with a threshold property, the tranOne of the important findings of a comprehensive study of public sector deficits by Easterly et al. (1994) is that largefiscaldeficits are largely explained by conscious fiscal policy choices and not by external or by domestic macroeconomic shocks. For example, see Alesina and Drazen (1991) and Alesina and Tabellini (1990). Also, papers somewhat related in the growth literature are Alesina and Rodrik (1994) and Persson and Tabellini (1994), who suggest that there may be a tendency of the majority voting to favor large redistributive spending in a democratic country with an unequal income distribution. See Velasco (1999), Hallerberg and von Hagen (1999), Tornell and Lane (1998), and Weingast et al. (1981) among others.
50
4 Social Polarization, Industrialization, and Fiscal Instability
sition dynamics can be dramatically different depending on the initial income distribution.^ If industrialization starts with a large fraction of the population participating, it will experience a narrowing of wage/income inequality as observed in East Asia, contrary to historical experience and contemporary evidence of the Kuznets curve in other regions J By contrast, when only a small fraction of the population is able to participate in industrialization, it leads to a permanent segmentation between the manufacturing and traditional sectors that yields a large sectoral wage gap.^ If a wide wage gap between the two sectors results, the polarization of preference—conflicts of interest over government spending—is sharper. The polarization of preference leads a policymaker representing each sector to spend more for her favorite sector, collectively contributing to bigger overall spending. Given that two policymakers share the government budget, whatever government resources a policymaker does not exploit may or may not be available depending on the other's behavior. When policymakers disagree on the desirable composition of government spending, each of them has a greater incentive to overexploit the common resources, exerting negative externality on the other, which prevents them from achieving the social optimum. Interestingly, a more polarized society also suffers from greater fluctuations in its fiscal outcomes. The higher the polarization, the bigger the fiscal spending and the larger the current fiscal deficit (polarization effect). But this raises the debt level more quickly and reduces government resources, forcing policymakers to cut tomorrow's spending by more. This polarization effect implies that a shock to tax revenue is translated into a more than proportional change in spending. Moreover, the higher is the degree of polarization, the greater is the change in spending in response to a shock to the tax revenue. We test these predictions on fiscal deficits and volatility by running panel regressions, based on a panel data set covering 90 countries over the period of 1970-90. Here we utilize recently assembled quality data on public sector deficits (Easterly et al. 1994) and income inequality (Deininger and Squire 1996). As our theory suggests, income inequality as a proxy of polarization is found to be statistically significant and robust in the regressions of fiscal deficit and volatility. Countries that have suffered from the greatest fiscal instability ^ Human capital externality as a "trickle down" process has been extensively studied in growth literature. See Azariadis and Drazen (1990) and Perotti (1993) among others. ^ Kuznets's inverted-U hypothesis has been questioned by recent empirical studies such as Birdsall et al. (1995) and Fields and Jakubson (1994). The East Asian experience has contributed to this argument (see World Bank 1993). ^ Large wage differentials found in Latin America and sub-Saharan African countries reflect unequal distribution of schooling, which is in turn related to income distribution. Consequently, much of these countries' inequality is associated with large wage differentials, and education is a key channel through which income inequality is perpetuated. See Inter-American Development Bank (1998) for a comprehensive study on income inequality in Latin America.
4.2 The Economy
51
are those with highly polarized economic societies as measured by indicators of income inequality.^ To our best knowledge, this is the first econometric test that reports the significance of income inequality in explaining different fiscal outcomes across countries.^^ Sect. 4.2 describes an economy under a post-industrialization regime and discusses the dynamics of income distribution in the process of industrialization. Sect. 4.3 derives an endogenous fiscal policy and establishes the main results, followed by country experiences that support our conclusions. Sect. 4.4 presents econometric evidence, and Sect. 4.5 concludes. Appendix derives the social planner's optimal fiscal policy. A list of our sample countries and a table of pairwise correlation of consolidated public sector balance with selected variables are provided in the data appendices.
4.2 T h e Economy We consider a two-sector small open economy with a government {fiscal authority) and a continuum of atomistic individuals endowed with perfect foresight. The economy is populated by overlapping generations of individuals who live two periods: young and old. The agents are equally able, but they are endowed with different amounts of initial wealth. For simplicity, we assume that the economy's two sectors produce essentially the same product (i.e., perfect substitutes). The price is normalized to one, p = 1. The population is constant and its size is normalized to one, L = 1. 4.2.1 The Pre-Industrialization Regime The pre-industrialization regime is characterized by linear production technology that uses only labor in both the manufacturing and traditional sectors. Production is described by Xf = X f = J L *
(4.1)
where Xf and X{^ are outputs in the manufacturing and traditional sectors at time t, respectively. The marginal productivity of labor is J > 0 in both sectors. With this technology, the unskilled labor is assumed to be as productive as the skilled labor. The skilled labor is more productive than the unskilled only when combined with capital stock. Thus, there is no incentive for an agent to get an education and become skilled because there is no wage differential {w^ = w^ = 6). ^ It is interesting to note that Malaysia, which had run the largest fiscal deficit and exhibited the greatest volatility of fiscal outcomes among East Asian countries in the period of 1970-90, also had the most unequal income distribution in this group. ^° A notable partial exception is Berg and Sachs (1988), who found that countries with high income inequality had a significantly greater likelihood of having rescheduled the external debts than countries with low income inequality.
52
4 Social Polarization, Industrialization, and Fiscal Instability
4.2.2 The Post-Industrialization Regime The post-industrialization economy consists of the modern manufacturing sector using skilled labor and capital and the traditional sector using unskilled labor only. Respectively, the production technologies are: X f = F ( L f , i ^ , ) , and
(4.2)
X f =:((5 + 0(Lf))Lf,
(4.3)
where F(-) is a constant return-to-scale (CRS) function that exhibits diminishing marginal productivity; L^ and L^ are skilled and unskilled labor, respectively (of course, L^ + L^ — L); and K is capital stock and depreciates completely in one period. To capture the benefit of industrialization for the rest of the economy, we assume that there is a positive human capital externality, 0(L'^), which spills over to the traditional sector, enhancing its labor productivity. The externality term (t>{L^) is a monotonically increasing function of L*^ such that (pLS > 0 and 0(0) = 0. It is assumed that FLS{L^, Kt) > S + (t){L^) for 0 < L^ < L so that the skilled worker is always more productive than the unskilled. It is also assumed that investment in human capital is made one period in advance and that both labor markets are in perfect competition. The government has free access to international capital markets and can issue bonds at the constant world interest rate, r € (0,1). It is assumed that there is no default risk of government debt. While individuals can lend any amount at this rate, they cannot borrow against future labor earnings due to a severe moral hazard problem. Since the number of skilled workers is known one period in advance, the capital stock installed each period is set to ensure that FK{Lf,Kt) = r + l. (4.4) Thus, the skilled worker earns the wage w^ = FL{L^,Kt) in the absence of government subsidy to the manufacturing sector, and w^ is a constant for given r and technology because F(-) is a CRS function. Note that the unskilled worker is paid w^ = MP^j = 5 -\- (j){L^). Each individual can work as unskilled labor in both periods or invest in human capital when young and work as skilled when old. Each supplies one unit of labor inelastically in each working period. Following Galor and Zeira (1993), we assume the indivisibility of human capital. To become a skilled worker, one has to invest /i > 0. Since no individual can have access to the capital market, a poor individual with e* < /i is unable to invest in human capital. The distribution function H{e) of the endowment is defined over the support e* G [e,e] and L = J^ dH{e). Let fit = Jh^^i^t) ^^ ^^^ fraction of the population that is rich and 1 — fit = f^ dH{et) the poor. The young do not consume. Each individual cares about her offspring and leaves a nonnegative bequest if she can afford to do so. The utility function of an individual i is given by
4.2 The Economy
53 (4.5)
U' = u{cudt),
where u{') is a concave function, satisfying the Inada condition; Ct and dt are the second period consumption and the bequest. There are two different types of government spending that affect the individual's budget constraint: the subsidy to the manufacturing sector, gt, and the government spending for the traditional sector, ft. As illustrated later, the government spending programs that may disproportionately benefit certain groups can take many different forms. For simplicity, we call any government spending that benefits the skilled workers the subsidy to the manufacturing sector, g}^ By using g in this way, we try to capture the fact that in many developing countries fiscal transfers to rent-seeking groups (such as industrial powers or public enterprises) take the form of government inputs, such as electricity, gas, soft credit, and raw materials that are provided at subsidized prices. Similarly, the spending / represents any kind of government program that favors workers in the traditional sector such as subsidized food, gas, fertilizer, rural electrification, public-works programs or provision of public employment (for example, see Gillis et al. 1996). These gt and ft are endogenously determined by two policymakers representing different sectors, and individuals take them as given. Since the policymakers share the government budget but seek to maximize their objective functions, they behave strategically in determining the spending level of each item. In a noncooperative dynamic game, policymaker S chooses {gt,gt+i} to maximize her own objective function subject to the government budget constraint for any possible {ft, ft+i} policymaker N chooses. Similarly, we define policymaker AT's constrained optimization problem: 1
Max
J' = Y,ßno^H9t+j)
+ (1 - cx')v{ft+j)], i = S,N.
(4.6)
j=o
subject to h+j - bt+j-i = rbt+j-i + gt+j + ft+j -T,
j = 0,1,
(4.7)
where 0 < ^5 < 1 is the subjective discount rate of the policymakers; b is the government debt; T is the lump-sum tax revenue; and v(-) is a concave function satisfying the Inada condition. We assume that 0 < a^ < | < o^*^ < 1 so that policymaker 5, who represents the manufacturing sector, prefers g to / and the other policymaker N, who represents the traditional sector, prefers / to g. We then define the degree of polarization, 9 = a^ — a^ e [0,1], where ^ = 0 means complete agreement on the composition of spending, and 9 = 1 complete disagreement. ^^ The profit of manufacturing sector is given by7r = X — (1 + r)K — FL{L^ , K) • L^ -\- g > 0 with equality if ^ = 0. The manufacturing sector profit (essentially subsidized by the government) is assumed to be equally distributed to the skilled workers in the sector so that they receive w^ = FL{L^,K) + g/L^.
54
4 Social Polarization, Industrialization, and Fiscal Instability
4.2.3 Human Capital Accumulation, Threshold Externality, and Income Distribution Now consider an optimization problem of an agent with initial endowment e. Only an agent with e > h can invest in human capital and become a skilled worker because no individual has access to the capital market. The optimization problem of an agent with e is, for given g and / : max U{e) = (1 - 7) log Ct + 7 log dt
(4.8)
{ct, dt}
s.t.
ct-hdt<{l
+ r){e-h)-hwf-T
Ct-\-dt<{l-\-r){e-hw^_i)
(4.9)
= y^, iie>h,
+ w^-^f-T
= y^,
iU
where 0 < 7 < 1. For algebraic simplicity, we assume a Cobb-Douglas utility function for u{c,d) and the lump-sum tax, T.^^ The bequest in period t + 1 and hence the endowment of a worker's offspring in period t-\-l are: et+i = dt+i{et) = 7{(1 + r){et - h)-\- wf^^ -T}= or
jy^, if e^ > /i, (4.10)
= dt+i{et) = 7{(1 + r){et + Ö + (^(Lf)) + (J + 0(Lf+i)) + / - T} = 72/^, if et < h. Recall that /it = Jh^H{et) and that Lf = fit-i since investment in human capital is made one period in advance. Hence, w^i = 6 -{- (/>(/xt)- In period 1, however, Lf = 0 because industrialization begins in period 1 so that rich individuals just start to invest in human capital from period 1. That is, under pre-industrialization regime, no one had an incentive to invest in human capital. For any given / and T, we define the threshold, /x G [0,1] such that 7{(1 -h r)h + (2 -h r){6 + (ßifi)) + f-T}
= h.
(4.11)
Now we characterize the distributions of wealth and income in the economies with / / ! < / £ and /ii > //. Case 1: /xi < /x. In this economy, //i = ßt+i ^ [£, for Vt > 1.^^ The fractions of the rich and the poor in the population remain the same, even after industrialization. ^^ To prevent the trivial case of y^ > y^ in which everyone wants to remain unskilled, we assume that wf > (1 + r)(/i + w^_i) + w^ + / for given g and / . Also, wf is assumed to be large enough for each skilled worker to leave a bequest greater than h. That is, ^{{1 + r){e — h) + wf — T} > h if e > h. ^^ From equations (4.10) and (4.11), one can check the following. For Vt > 1,
4.2 The Economy
55
The major benefit of industriaHzation is Umited to only the rich with e\> h and their descendants, except to the extent that the unskilled benefit from the externality that raises their productivity. However, the size of the initial fraction, /ii, still matters for the subsequent evolution of income distribution. Using equations (4.10) and (4.11), we can compute the steady-state levels of wealth, e^ and e'^, for skilled and unskilled workers, respectively. To get the stable path, it is assumed that 1 > 7(1 + r). ^'=,_^l
(4.12)
+ r)^^'-T-il+r)h],
The wealth distribution in the steady state becomes more equitable as the initial fraction of the skilled (ßi) is higher.-''* The income inequality (measured as a ratio of y^ and y^) also declines as ^i increases: / y^
[{l + r){e^-h) + w^-T] [(1 + r)e'^(Mi) + (2 + r)(<5 +
(4.13)
Case 2: /ii > /i. In this economy, the poor can gradually invest in human capital generation after generation, contributing to a reduction in income inequality with industrialization. This is because the externality becomes large enough in each period for some unskilled workers in the traditional sector to leave a bequest et+i = 7{(1 + r){et - h) + wf+i -T}>h, if a > h, = 7{(1 + r){et + (5 + (t>{ßt-i)) + ((^ + (t>M) -\-f-T}
< h, \i Ct < h.
Note that if ei < /i, then 62 < h and hence /i2 = Mi- That is, given that ßi < n and Lf = Oj ei < h implies that 62 = 7{(1 + r){ei +Ö) + (S + (t>{yii)) -¥ f-T} < 7{(1 + r)h + (2 + r){5 + (t>{ß)) + f-T} = h. Similarly, if 62 < h, then es = 7{(1 + r){e2 + (5 + (t>(fii)) -\-{Ö + (l>{f^2)) + / - T} < 7{(1 + r)h + (2 + r){S + >(/£)) + / - T} = /i, and hence /xs = M2 = A*i < /£• Repeating the same procedure establishes that if /xi < /x, then /xi = /it+i < M, Vt > 1. ^'^ Note that e^ = e^(fii) and e^'ißi) = i z : ^ ^ [ ( 2 + r)(t>'{fii)] > 0. Also e^ declines in fii for given r, T, g, and technology because w^ = FL{L^ , K)-\-g/L^ = FL{fii,K)+g/fii,
56
4 Social Polarization, Industrialization, and Fiscal Instability
larger than h. Since Lf = 0 and (t>{Lf) = 0, there can exist an unskilled worker endowed with ei < h who leaves a bequest bigger than h in period 2 only if (/>(/ii) > (2 + r)(p{jjL). Otherwise, there will be none. In either case, however, /X2 cannot be smaller than ßi. Since 0(^*2) > 0(/^i) > ^(/f)? unskilled workers who inherit 62 € [62, h) now leave bequests greater than /i, and their descendants in all future generations become skilled, where 62 is a level of endowment such that 63 = 7{(1 -f-r)(e2 -\-S-\->(/xi)) -\-o-\- 0(//2) -\-f — T} = h. Thus, /is = j^dH{e^) > /i2 = J^dH{e2). Similarly, unskilled workers who inherit 63 € [es,h) leave bequests larger than h in period 4, where 63 is defined to satisfy 64 = 7{(1 + 0(^3 + ^ + >(M2)) 4- (^ + (/)(/i3) + / - T} = /i, and so on.^^ Therefore, if //i > //, then /i2 > /^i and //t+i > A*t, for Vt > 2. Once industrialization begins, the inequalities of income and wealth continue to decrease, converging to unique steady-state levels of income and wealth (see Fig. 4.1).
et+i
e^(/.f)
e\^if)
e^(/if) = e^(/.f)
Fig. 4 . 1 . Countries A and B fall into Case 1 (i.e., ßi < ßt < ß)^ whereas country C belongs to Case 2 (i.e., /£ < /if). For simplicity of the graph, we assume that these countries are identical except for /ii and that f = g — 0.
^^ There is a finite maximum level on / that can be sustained as an endogenous equilibrium as shown in Sect 4.3. Thus, we can rule out the trivial case that everyone becomes educated after the transfer regardless of her initial endowment by choosing parameters h, T, r, 5, and (j){') so that the threshold is binding for a given income distribution.
4.2 The Economy
57
4.2.4 Income Inequality and Polarization Industrialization can engender not only earning differentials but also potential conflicts of interest between income groups over the composition of government spending. Suppose the economy ends up with a permanent segmentation and a wide sectoral income gap (the case of /xi < jj). The resulting sectoral income gap may induce workers in the traditional sector to demand higher redistributive spending /.^^ On the other hand, the permanent segmentation is likely to encourage the rich to strongly influence the fiscal policy in their interest, leading to a greater amount of transfer and subsidy g. Thus, the polarization of preferences for the composition of government spending would be sharper in this economy. To illustrate this point, we use the indirect utility function, V^(y^ : •) = lny% i = 5, N}'^ The welfare gap between sectors is then: V ^ ( / : e>h)-
V'iy''
: e < /i) = In
( ^ )
V[(l + r-)e^(w) + (2 + r)(5 + ,^(Mi)) + / - T ] ; -
^'
'
Note that either a higher / or a higher initial /xi reduces this welfare gap at the steady state for any given level of g and T. For any given level of g, the incentives for unskilled workers to demand higher welfare spending / (as measured by their marginal utility of / ) become much stronger when the initial income distribution is more unequal: dV^/df = l/y^ is greater if //I is lower (i.e., d'^V^/dfdfii < 0). On the other hand, the skilled worker's marginal utility of g is dV^/dg = l/{y^ * Mi)- Again, the more unequal the initial income inequality and hence the lower the //i, the higher the marginal utility of the manufacturing subsidy (i.e., d'^V^/dgdfii < 0). On the contrary, the economy with fii > /i will converge to a unique distribution of income and wealth, causing the welfare gap to decline towards zero. Unskilled workers would be less likely to demand higher redistributive spending as their marginal utility with respect to / becomes smaller due to declining income inequality. By assumption, policymaker S prefers p to / , whereas policymaker N prefers / to g. Since the marginal utility of g is negatively related to //i, policymaker S, representing the manufacturing sector, will put a higher value of a^ on the utility she derives from the transfer g when //i is smaller. Similarly, the policymaker N will put a greater value of (1 — a^) on her utility derived ^^ Related studies on redistributive politics are Meltzer and Richards (1981) and Benabou and Ok (2001). In these models, income inequality or lack of upward social mobility favorably influences the preference of the majority for redistributive policies, leading to a larger amount of redistribution. ^^ Since we assume a Cobb-Douglas utility function and the price is normalized to one, the indirect utility function is in a simple logarithmic form of ln(y).
58
4 Social Polarization, Industrialization, and Fiscal Instability
from / when //i is smaller. The degree of polarization 0 = a^ — a^ G [0,1] will then be most likely higher when //i is lower since both a^ and (1 — a^) are inversely associated with /ii. Therefore, people in a more unequal society would have much more divergent preferences for the composition of fiscal spending because there is a sharp conflict of interests among them. As Birdsall et al. (1995) note, the likelihood that fiscal policies are polarized between the extreme of serving the narrowly defined myopic of the rich and an equally myopic populist extreme will be higher where higher levels of income inequality widen the distance between the interests of the rich and the poor. Now we look at the role of this polarization in the evolution of fiscal spending and deficit.
4.3 The Fiscal Policy The fiscal policy consists of two different types of government spending {gt, ft}u=i and taxation T. While the two different types of policymakers play a non-cooperative dynamic game, each policymaker of a generation also cares about the fiscal spending levels of the next generation, which cares about the government spending of the following generation and so on. Altruism is limited in our model in the sense that each generation derives utility only from its own spending and that of its immediate successor. ^^ This game-theoretic setup introduces two critical driving forces for the fiscal outcome: polarization of sector preference and limited altruism towards the next generation. For simplicity, we assume a lump-sum tax, T. However, the rise of an endogenous fiscal deficit does not rely on the level of tax revenue, as becomes clear later. 4.3.1 Polarization and Endogenous Fiscal Deficit Assuming a simple Cobb-Douglas utility function for v{'), we can write the optimization problem for the policymaker i of generation t as follows: 1
Max J' = Y1 ^'[^' l^SPt+i + (1 - a') log /,+,], i = S,N
(4.15)
3=0
subject to h+j - h+j-i
= rbt+j-i + gt+j + ft+j -T,
j = 0,1.
For a concrete result, we focus on the linear stationary Markov perfect equilibrium that depends only on the current state variable, bt.^^ Let us ^^ Limited altruism models have been explored in a variety of contexts. See Bernheim and Ray (1989), Leininger (1986) and the references therein. ^^ In a Markov strategy set, past decisions influence current play only through their effect on a state variable that summarizes the direct effect of the past on the
4.3 The Fiscal Policy
59
write the government net revenue as Rt = T — rbt-i and consider the linear strategies gt = X^^t and ft = X^^t, where the set of strategies is X* € [0,00), i = S,N. By focusing on stationary equilibria, we describe the evolution of fiscal policy decisions when all generations of each type of policymaker select the same Markov strategy x*Policymaker S solves max{(l+/?)[a^logx^ + (l-c^^)logx'^ + logi^t] + / ? [ l o g ( l - r ( x ^ + x ' ^ - l ) ) ] } . (4.16) The first-order condition for her optimization problem is ßrx^ = (1 + ß)a^[l - rix^ + x ^ - !)]•
(4.17)
Similarly, the first-order condition for policymaker N is ßrx"" = (1 + /3)(1 - a ^ ) [ l - r(x^ + x'^ - 1)1-
(4.18)
The Markov perfect equilibrium is then: _
(l+/?)(l+r)a^
_ (l + / ? ) ( l + r ) ( l - a ^ )
Substituting this solution and Rt = Tt — rbt-i into the government budget constraint (4.7) yields bt-bt-i = k{-){T-rht-i), where k{') = ri"'^^^it^^
(4.20)
> 0. We establish the following proposition.
Proposition 4.1. (i) The higher the degree of polarization, the larger the fiscal deficit, (ii) The less patient the policymakers, the bigger the deficit. (Hi) When there is polarization (i.e. 6 > 0), the spending level and the size of deficit are always greater than the social optimum. Proof, (i) dk{-)/dO > 0. (ii) dk{')/dß
< 0. (iii) See Appendix A.
Given that two policymakers share the government budget, whatever resources one does not exploit may or may not be left, depending on the other's behavior. When they disagree on the ideal composition of government spending, each has an incentive to overexploit the common resource today. That is, the polarization of preference leads each policymaker to insist on a higher current environment. A Markov perfect equilibrium (MPE) is a profile of Markov strategies that yields a Nash equilibrium in every proper subgame. Thus, the MPE is subgame perfect and time consistent. See Leininger (1986) and Bernheim and Ray (1989) for the existence of Markov perfect equilibria in a model of growth with altruism between generations (so-called bequest games).
60
4 Social Polarization, Industrialization, and Fiscal Instability
spending for her favorite sector and to exert negative externality on the other, contributing to bigger overall spending and a larger deficit than the social optimum. In Sect. 4.2.4, we saw that the more unequal the income distribution, the more likely the polarization of sector preference over different types of fiscal spending. Thus, the society with greater income inequality is more likely to run larger deficits. When policymakers care less about the next generation and discount the future more heavily, they also have a deficit bias.^^ 4.3.2 Polarization and Volatility of Fiscal Outcomes In this subsection, we establish that a country with a higher degree of polarization will experience greater fluctuations in its fiscal outcomes. We begin with a simple case of lump-sum taxation. For any given tax revenue, a rise in polarization 6 has two effects. The first, which we call polarization effect, is that the level of government spending (pt = 9t -^ ft) rises (see Proposition 1). The second effect is that government spending falls more quickly because the increase in debt lowers Rt, the pool of the government resources.^^ Thus, greater polarization means larger fiscal deficits today and more drastic spending cuts tomorrow. The country with higher 6 will exhibit a bigger swing of its fiscal spending path for a given level of tax revenue.^^ 20
If we view a high subjective discount rate as reflecting political uncertainty facing the policymakers, this result reaches the similar point made by Alesina and Tabellini (1990). They showed that the incumbent government, faced with the uncertainty over re-election, may have a debt bias because the costs of additional debt are borne only if re-elected. A country with high 6 will spend more until some finite time; thereafter, it will spend less than a country with low 0. Let us say 6i > 62- The ratio of gu to p2t is greater than one until for some finite time t, and thereafter becomes less than one, where the ratio is given by 9u ^ (l + ^i)[/3 + ( l + / 3 ) ( l + ^2)](l-fe(^i)r)*-^ 92t (1 + 02)[ß + (1 + ß){l + Öi)](l k{02)ry-^'
22
In practice, it does not have to be government spending that must adjust. The government can use taxes or seigniorage to pay the debt. No matter which instruments are used, however, today's larger deficits still mean more drastic adjustments in fiscal policies in the future, making fiscal outcomes volatile over time. In most hyper inflationary episodes, large budget deficits were often an initial cause. In Chap. 3, we studied the inflationary consequence of budget deficits in relation to the polarization effect. On the other hand, it is a commonly shared view that developing countries in general have much narrower tax bases relative to those of industrial countries. Also, the collection of tax revenue in developing countries is often hindered by limited administrative capacity and political constraints (Agenor and Montiel 1999). This implies that developing countries might find it harder to raise taxes in order to cut budget deficits than cutting expenditures. Consistent with this line of reasoning, total expenditure and its
4.3 The Fiscal Policy
61
More importantly, a more polarized society would suffer from a greater fluctuation in its government spending in response to a shock to tax revenue of the same magnitude. Note that the fiscal spending is given by 9t = p^^q^^5ji^p^(^* - rbt-i) from equation (4.19). Hence, for any given
where /(•) = m]|!(i-^m(i-}-^)]r ^ ^' ^^^ amount of fiscal spending change in response to a shock to tax revenue is increasing in the degree of polarization (i.e., dl{')/d6 > 0). An increase in tax revenue is translated into a more than proportional increase in spending through the polarization effect. Moreover, the higher the polarization, the bigger the increase in p. But this leads to a sharper reduction in subsequent spending because the increase in tax revenue is dissipated more quickly. Similarly, the absolute size of spending change resulting from a shock to tax revenue is bigger when the policymakers are less patient (i.e., dl{')/dß < 0). Proposition 4.2. (i) The higher the polarization, the more volatile the fiscal spending and the fiscal balance, (ii) The less patient the policymakers, the more volatile the fiscal spending and the fiscal balance. (Hi) When there is polarization, the fluctuations in fiscal outcomes are always greater than the social optimum. Proof, (i) Prom equation (4.19), it is straightforward to see that for any given bt-u \{d{bt - bt-i))/dT\ = k{-) and dk{-)/de > 0. (ii) dk{-)/dß < 0. (iii) For any given bt-u \^/depolarization = KO, and |c^/^T|^^^.^;^^„^^^^ = ((1 H- ß){l -h r))/{[ß + (1 + ß)]r) (see Appendix A). One can easily show that \d9ldT\^^,^^,^^,,^^ > \d9/dTl^^,^,^,^^^^^ when ^ / 0. If Ö = 0, \dg/dT\^^i^^^^^^^^^ = \dg/dTl^^.^ipi^^^^^. Similarly, for any given 6t-i, \{d{bt - bt-i))/dT\^oiarization > Mbt " &t-i))/rfT|,,,,,,^^^^^,, with equality when ^ = 0. 4.3.3 Country Experiences It is a well-established fact that education is the most important determinant of wage differentials in developing countries (for instance, see Gillis et al. 1996). Moreover, the wage differentials by educational level are much greater in developing countries than in developed countries (Psacharopoulos 1985). On the other hand, school enrollment is found to be strongly and negatively correlated with income inequality (Birdsall et al. 1995; Clarke 1995). Indeed, a large wage gap has been reported in Latin America and sub-Saharan Africa components as well as fiscal balance have been more volatile than tax revenues across developing regions (see Table 4.1).
62
4 Social Polarization, Industrialization, and Fiscal Instability
where initial income inequalities were greater. In Colombia and Jamaica, workers earn nonagricultural wages that exceed agricultural wages by 150% (Fields 1994). According to Berthelemy and Bourguignon (1996), in Cote d'lvoire the average wage of workers in the manufacturing sector was 364% higher than the average wage in the informal sector in 1984-85. In sharp contrast, this sectoral wage gap in East Asia is only about 20% (Fields 1994). The favorable initial conditions in East Asia helped to set up a virtuous circle: initial low inequality in income led to educational expansion, which reinforced narrowing wage gap and low inequality. Table 4.2. Social Welfare Spending and Subsidies and Transfers as a Percentage of GDP in Selected Countries and Years Social security and welfare*^ Subsidies and transfers 1975 1980 1985 1990 1995 1975 1980 1985 1990 1995 E a s t Asia Indonesia Korea Malaysia Singapore Thailand Average
N/A 0.83 0.71 0.33 0.65 0.63
N/A 1.09 1.14 0.27 0.51 0.75
N/A 0.92 1.04 0.43 0.62 0.75
N/A 1.46 1.08 0.44 0.51 0.87
0.91 1.54 1.42 0.79 0.55 1.04
3.42 5.04 7.29 1.37 2.48 3.92
5.33 5.88 5.43 1.11 2.69 4.09
5.11 6.23 3.78 2.65 1.60 3.87
3.78 7.44 4.78 2.44 1.22 3.93
Latin America Argentina Brazil Chile Colombia Mexico Panama Uruguay Venezuela Average
N/A N/A 8.20 N/A 3.39 3.04 10.7 1.33 5.33
5.21 6.46 9.04 N/A 3.15 3.17 10.6 1.44 5.58
5.76 5.94 11.8 2.73 2.23 3.78 12.0 1.16 5.68
4.84 8.81 7.24 0.76 2.34 5.22 13.0 N/A 6.03
7.79 N/A 6.71 N/A 3.19 5.73 18.5 N/A 8.38
N/A N/A 11.2 N/A 3.82 3.62 12.3 4.76 7.74
7.81 12.9 13.0 5.04 4.99 4.30 9.38 4.08 8.20
10.4 10.7 15.6 6.74 5.02 4.65 10.5 6.35 9.09
5.97 8.73 13.6 N/A 10.5 10.3 4.88 N/A 2.98 6.89 6.20 6.65 12.9 18.7 7.70 6.02 8.15 10.25
Sub-Saharan Africa N/A 0.97 7.31 Cote d'lvoire Zambia 0.19 0.70 0.54 N/A 2.02 1.16 Zimbabwe Average 0.19 1.23 3.00
0.23 0.30 N/A 0.27
N/A 0.70 N/A 0.70
N/A 7.03 N/A 7.03
4.26 6.35 9.37 N/A 9.06 10.57 7.56 8.46
7.34 N/A 4.73 4.37 4.80 N/A 5.62 4.37
Egypt
4.68 4.94 5.22 3.58 4.54 25.0 16.2
Source: IMF (various issues). ^ Does not include spending on Education and Health.
13.9
7.34
2.34 8.65 4.64 1.83 1.03 3.70
8.72
4.3 The Fiscal Policy
63
Large income gaps across sectors and skills in Latin American and subSaharan African countries may have contributed to political pressures for the government to raise redistributive spending in various forms such as social welfare spending, in-kind transfers, public employment, and public-works programs.^^ The segmentation between the rich and the poor due to income gap may also have helped the economic elite to strongly influence fiscal policy in their interest.^^ In contrast to East Asia, fiscal expenditures in Latin America and sub-Saharan Africa in the past decades are characterized by (i) higher share of social welfare spending in GDP, (ii) higher share of subsidy and transfer in GDP, and (iii) rapid expansion of the public sector such as public sector employment. Social welfare spending (as a percentage of GDP) in Latin American countries was more than six times greater than that of East Asia for the period of 1975-95 (see Table 4.2). During the same period, the subsidy and transfer (as a percentage of GDP) was often much larger in Latin American countries than in East Asia. On the other hand, the state-owned enterprise (SOE) deficit has been important for fiscal instability. According to World Bank (1995), it accounted for 35% of fiscal deficits in 34 developing countries for the period of 1978-92. It is widely accepted that a substantial portion of government subsidy and transfer has been devoted to covering large deficits of SOEs. For example, diverting SOE operating subsidies to basic education would increase central government education expenditures by 50% in Mexico, 74% in Tanzania, and 160% in Tunisia. This stands in stark contrast to 0.1% in Thailand (World Bank 1995). This warrants some special attention because the macroeconomic role of SOEs in East Asia was not smaller than in other developing regions where SOEs' expansion often led to greater fiscal deficits and inefficiency in the form of patronage, subsidies, and overstaffing. The provision of public employment to satisfy the excess demand for highwage employment is also an all-too-common example of populist policies. The public sector employment grew more rapidly than employment in the private sector in Latin America and sub-Saharan Africa. The median share of the public sector in the increase in total employment was 71-87% in Latin America and sub-Saharan Africa, compared to 10% in Taiwan and 33% in Thailand (Gelb et al. 1991). The macroeconomic costs of public employment were often very large, and cutting the wage bill in the public sector was a key ingredient of the fiscal reform package.^^ ^^ In Latin America and sub-Saharan Africa, however, income-transfer programs are still rudimentary when compared to industrialized countries. Consequently, their governments tend to rely on public employment, in-kind transfers, and publicworks programs as well. ^^ The allocation of limited fiscal resources for tertiary education, so common in Latin America, is another example (Birdsall et al. 1995). ^^ However, it has been difficult to restructure the civil service due to strong opposition and lobbies from the insiders. Mackenzie et al. (1997) found that Chile and Ghana were the only two countries to achieve a substantial restructuring of the
64
4 Social Polarization, Industrialization, and Fiscal Instability
4.4 Econometric Evidence The basic implication of the model is that countries with greater polarization are more likely to run larger fiscal deficits and exhibit more volatile fiscal outcomes. We test these implications of social polarization for fiscal outcomes by running panel regressions of fiscal balance and its volatility. Many of the empirical studies on budget deficits use data on the central government's fiscal balance. However, using only the central government deficit data can be problematic because a large part of the fiscal deficits in developing countries is attributable to the rest of the public sector (see Easterly et al. 1994). Here we use an expanded panel data set for 90 developed and developing countries, which combines consolidated public balance data of 57 countries from Easterly et al. (1994) with general government balance data of an additional 33 countries from IMF's Government Finance Statistics (GFS). All other data are from World Bank's World Development Indicators CD-ROM unless specified. Our panel is comprised of decade averages of variables for 1970-79 and 1980-90. See Appendix B for a complete list of these 90 countries. Our basic regression equation for fiscal balance is as follows: SURPit = aiLRGDPit
+ a2GRGDPit + a^INFLATu
+ ßXu
(4.22)
where i and t denote the country and decade (1970-79 and 1980-90), and e is an error term. SURP is the average of fiscal surplus (as a percentage of GDP). We follow the existing studies (for example, Edwards and Tabellini 1991; Roubini 1991) to include the initial income level, the real GDP growth rate, and the inflation rate in our basic regression. According to the tax-smoothing model of fiscal deficits, budget deficits will emerge when output is temporarily low relative to its permanent level. We include the decade average rate of real GDP growth (GRGDP) as a proxy for the cyclical effects on budget deficits (such as booms and recessions), which might proxy for the different degree to which the countries in question faced economic recessions during the sample period. The log of real per-capita GDP at the start of each decade, LRGDP, is introduced to control for potential effects of economic backwardness on the public deficits. Poor countries tend to have relatively inefficient tax and spending systems and may therefore be more prone to budget deficits. We also include the decade average of the rate of inflation of the consumer price index (INFLAT). Inflation can affect fiscal deficits through different channels. Rapidly rising inflation can raise flscal deficits through higher nominal interest payments. High inflation can also lead to lower real tax rev-
civil service through restraining wage and cutting employment in the period of 1978-93 among the eight countries in their study.
4.4 Econometric Evidence
65
enue.^^ We try to capture the effects of inflation on fiscal deficits by including INFLAT. Besides decade dummies, Dt, we include regional dummies, RDi, for East Asia, Latin America and the Caribbean, and sub-Saharan Africa. Regional dummies are intended to control for structural characteristics related to geographical location. We always include the aforementioned economic variables throughout our econometric exercise. Finally, Xu represents other variables, which we specify later. Table 4.3 presents the direct evidence supporting the model's prediction that the fiscal deficit is a positive function of degree of polarization multiplied by net tax revenue as indicated by equation (4.20). As suggested in this chapter, we proxy polarization using the Gini coeflScient, AGINI (see below for detailed explanation). The net tax revenue (a percentage of GDP), NTR, is central government tax revenue minus interest payment. The coefficients of AGINI*NTR have the correct sign (-) and are all significant at the 1-5% significance level. The coefficients of LRGDP, GRGDP and INFLAT all have the expected signs, but only that of LRGDP is consistently significant. In columns (4) and (5), we also include the multiproduct of growth rate of terms of trade and trade share in GDP (EXT) to control for the external shocks to the economy. External shocks can be an important source of fiscal instability, especially in developing countries. Changes in export and import prices can affect the public sector balance either through the profits of exporting public enterprises or through import tariffs and taxes on exports. The growth of terms of trade is expected to be associated with smaller budget deficits and to have a greater impact in economies that are more open to trade. Although the coefficients of EXT have the expected sign (-h), none of them are significant. In column (5), we include frequencies of cabinet changes (CABCHG) and military coups (COUPS) as proxies for political instability. Recent studies on fiscal politics emphasize political instability as an important cause of fiscal deficits (for example, see Alesina and Tabellini 1990). They suggest that if faced with uncertainty about the government's survival, the incumbent government may have a bias towards overspending and fiscal deficits. Edwards and Tabellini (1991) and Roubini (1991) found evidence in support of this political instability channel.^'^ Both the coefficients of CABCHG and COUPS are of the expected sign (-), but they are insignificant. We will have more to say about this later. Table 4.4 reports our main finding that countries with highly polarized societies tend to run larger fiscal deficits. As an indicator of polarization, we use GINI and AGINI. (i) GINI: high-quality data on income inequality measured ^^ If, however, income taxes are not indexed to inflation, the above effects of inflation on deficits can be at least partially offset by the positive effect of bracket creep on income tax revenue. ^^ Based on consolidated public sector balance data of 57 countries, Chap. 5 provides a comprehensive empirical evaluation of different theories of fiscal deficits.
66
4 Social Polarization, Industrialization, and Fiscal Instability
Table 4.3. Pooled Regressions Based on the Model. Dependent variable: Consolidated Public Sector Surplus (% of GDP) (1) 2.336** East Asia (2.171) 0.314 Latin America (0.407) Sub-Saharan Africa 2.167*** (1.852) 2.392* LRGDP (5.324) -0.003** AGINI*NTR (-2.406) GRGDP
(2)
(3)
(4)
(5)
1.328 (1.081) 0.521 (0.679) 2.614** (2.278) 2.690* (5.029) -0.004* (-2.652) 0.330 (1.549)
1.576 (1.249) 0.954 (1.173) 2.773** (2.322) 2.674* (5.008) -0.004* (-2.817) 0.227 (0.979) -0.003* (-2.766)
1.541 (1.218) 1.121 (1.291) 2.791** (2.299) 2.658* (4.852) -0.004* (-2.642) 0.231 (0.985) -0.004* (-2.756) 0.085 (0.428)
1.522 (1.144) 1.155 (1.395) 3.233* (2.804) 2.938* (5.759) -0.004* (-2.624) 0.354 (1.487) -0.002 (-1.309) 0.114 (0.593) -0.773 (-1.098) -1.144 (-0.833)
115 0.63
113 0.63
109 0.62
107 0.64
INFLAT EXT CABCHG COUPS No. of Observarions Adj. R^
115 0.62
The decade dummies were included in the regressions (not shown to save space). White heteroskedasticity-consistent t-statistics are in parentheses. Levels of significance are indicated by asterisks: * 1%, ** 5%, *** 10%.
at the start of each decade from Deininger and Squire (1996); and (ii) ACINI: the decade average of all available data of Gini coefficients from Deininger and Squire (1996).^^ We expect the high degree of social polarization as proxied by 28
To avoid reverse causality from public deficits to income inequality, GINI was limited to countries for which we have high-quality data measured close either to the end of each previous decade, the 1960s and 1970s, or to the start of each decade, the 1970s and 1980s. However, income inequality measured by Gini coefficients is very persistent over time as evidenced by high correlation between GINI and AGINI. The pairwise correlation is 0.92. When GINI is used, the results are often stronger than when AGINI is used. Moreover, if one is concerned with measurement error in a specific year, using a ten-year average of Gini coefficients like
c»
o
(-1
OH
t+-t
Q O
CO
E
'>< O o CL.
CQ
Co
.2 *-+-' SH
3. ^
Co
H:»
;3 (1.
,^
p ^
o
0) CO
-1-3
CO
1 'o o o o
3 *o CO
CO
'? Ö
w
.2
o
j3
0) bO O
T3
Ö 0)
73
^ 3 3 .^ o *5 CL, g ^ ^ Ti5
-2
a o H Q
3Co
CO CO CO CO
(X5
00 to
^
o o o
o o
o o Ö Ö o^ o^ Ö1 o
Ö1 Ö1 Ö Ö Ö Ö O^ CO an r^ y ^ r^ 10 P c<j
Ö d Ö d 9 r^
" ^ 10 -Tt
t- 22 S t^ o b-
oocooiTi^i-iootU';^ l O C O C O O O C O ^r HJ S CO
1-1
00
C5 10 C^ Ö
CO 00 o (M O O CD CO
^ '"J rH
rH iO ;ZJ $2
l^ 10 O Ö
CO 00
CO CO ^.
^.
. ö r-; P rH
CO
0
^ Z^^ C.^ S'ö d (M
T-H
^
CO CO CO ^ O ^ CO CO CO 9 r-i C
* 00
o t-
t^ CO
O O "^
S ^ 2 g:^
^ CO iO O
^
" S ^ CO 1 0 C ^ C^ 2 ffo 1 0 o t--
Ö
iO -
iO
S
CO
"^
-^ LO ^ CO O CM CO 0CO 0 iO Oi CUi CO O rö r»^ "^ 00 Ö d d CM d rH
"
^ 00 '^ ^ r^ CM
o
hS
Oi
rH CX) 1—1 C i
CO
(M CO rH CO
10 CO CM brH (M
CO CT) O l> Ö rH
o CO
ILO T;f
00 00 (M r-1
00 10 O 00
0:1
1
CO
CO CO 1—1
1
CO
T-H
P
4.4 Econometric Evidence
1—1
t>- CO O O
N»
CO t^ CO rA
67
68
4 Social Polarization, Industrialization, and Fiscal Instability
Gini coefficients to be associated with larger public deficits. Columns (2) and (3) show that even after controlling for the basic economic variables, proxies for polarization (GINI and AGINI) are statistically significant at 1-5% levels and have the expected sign, (—). The adjusted R^s are between 0.62 and 0.66. According to the point estimates, a difference in income inequality of ten Gini points is associated with an increase in public deficit of 1.3-1.7 percentage points of GDP per year. We confirm this finding by using alternative proxies for polarization. Column (4) uses GINILA, Gini coefficient of land ownership around 1970, while column (5) employes an index of ethno-linguistic fragmentation (ETHNIC) that measures the likelihood that any two randomly chosen individuals in a country will not belong to the same ethno-linguistic groups (both of them are from Taylor and Hudson 1972). The coefficients of GINILA and ETHNIC are of the correct sign (-) and significant at 10%. When both AGINI and ETHNIC are included, the coefficient of ETHNIC is no longer significant (see column 6). In columns (7)-(9) of Table 4.4, we include a broad measure of quality of government institutions, ICRGE, which has received much attention in the growth literature. This index is based on underlying numerical evaluations relating to the rule of law, bureaucratic quality, corruption, expropriation risk, and government repudiation of contracts (Knack and Keefer 1995). It ranges between 0 and 10 with 10 indicating the best institutions. The idea is as follows. Thus far in our model of fiscal deficits, we have implicitly assumed that these policymakers can exploit the common revenue to spend more on their preferred items. In reality, this may not be true if there is a well-functioning mechanism of "checks and balances" within the government. Therefore, even in a highly polarized society, a good conflict-solving institution may help the country avoid playing the noncooperative spending game. As expected, the coefficients of ICRGE are of the correct sign (-h), but they are not always significant. Interestingly, the coefficients of proxies for polarization become statistically weaker when ICRGE is included in the regressions. This confirms the above idea that when the institutions of conflict management are well-established and working properly enough to suppress potential conflicts of interest over government spending, the extent of the polarization effect might be less important in determining fiscal outcomes. We also tried other measure of institutional quality such as BERI from Knack and Keefer (1995), but the results do not significantly change. Since the effects of polarization on public sector deficits may vary with the quality levels of institutions, we construct the following composite index of polarization that takes the institutional quality into account. Polar = AGINI X (10 — ICRGE). The composite index enters with a negative and strongly significant coefficient (column 10). According to the point estimate, an increase of one standard deviation in the polarization indices is associated AGINI might be preferable. Thus, we report regression results, primarily using AGINI to maintain the largest number of observations possible.
4.4 Econometric Evidence
69
with a deterioration of public sector deficits of the magnitude of -2.62% of GDP. As mentioned above, poUtical uncertainty can be an important cause of fiscal deficits. In relation to the uncertainty about government survival, we consider political assassinations (ASSASSIN) and government crises (GOVTCRIS) as well as CABCHG and COUPS—^the political variables are all from Banks (1997). High levels of socio-political unrest may be strong expressions of dissatisfaction with the current government and its policies, which may not only make the downfall of the government more likely but also may drastically shorten the horizons of politicians. With a shortened expected tenure in office, the government is more likely to engage in myopic short-term policies. Columns (l)-(5) in Table 4.5 report the regression results. The coefficients of these variables (except for GOVTCRIS) are insignificant, although they have the expected sign (-). In reality, political instability is a multidimensional phenomenon that cannot be captured with a single variable. Thus, we combine these four variables by using the principal components method to create a composite index PINSTAB.^^ The coefficient of PINSTAB is of the correct sign (-) and significant at the 10% level (column 6). LRGDP and GRGDP have the correct sign (+), and the values of LRGDP are highly significant. It is remarkable that all the coefficients of AGINI are highly significant. The composite index of polarization also enters with a highly significant negative sign (column 7). We include additional variables as a way to check the robustness of our regression results. Again, we include the indicator EXT to control for external shocks to the economy. Some economists (for example, Edwards and Tabellini 1991; Cukierman et al. 1992) argue that the degree of urbanization is important. It might be relatively easier to tax the urban population than the rural population. Low urbanization ratio will then be associated with low tax revenue or with a tendency to use seigniorage. On the other hand, political conflicts are likely to be more intense and disruptive in urban areas than in rural societies. This means that a high urbanization ratio may be associated with large deficits. Thus, the sign of the coefficient of urbanization is theoretically inconclusive. We also include URBAN (urbanization ratio at the start of each decade) in the regressions (9)-(12). They are all negative but are not significantly different from zero. Now we test our hypothesis on fiscal volatility. According to equation (4.20), the standard deviation of fiscal balance should be a positive function of degree of polarization multiplied by the standard deviation of net tax revenue ^^ The principal components analysis is a statistical technique that helps us combine many correlated variables into a variable that contains most of the information (i.e., a linear combination with the greatest variance). All the variables that are included in PINSTAB are standardized so that they have a mean of zero and standard deviation of one at the outset. PINSTAB=0.22xCABCHG+0.26xCOUPS+0.22xASSASSIN H-0.38X GOVTCRIS.
70
CO CO
-5 (X •^ Q J^ e+HO > >. O ^ O
-(J
X
^ 3. ;3
w O^
-^
.2 -»-'
O i=! 0) +^ CO TJ o T3
0.
<1 3 n^
CO T3
O
'5T3-H w Ö o
(-1
1
'6
hO o Ü
Ti
o
1
Ö 0?
Q
a
kO 4J
^ ^
3 cd H
rH
o
(M
o o o TH Ö Ö
'^"-^"*~^ oT r-l
1
CM -Jf
to
LO"
rH
l> CM
1
CO 1—1
*
CM CM rH
Ö
CJi 05 CM CM
co" 1—1
CM
CO
1
CM
^ CM" lO
CM CM
rH
r-1
'^
^
CO CO LO 0 0
O ffi
o m < o
t- o
CO b LO ^ CO CO
CO
Ü
O
13
OH
LO O O LO CM CO
Ö Ö
CO CO
<
CO CO
<:
LO
CO
1
Ö
CO"
CO 00 00 LO
•X-
1
lO
CO
PQ
00 CO CO CD
•X-
LO CD
o
Ö
CD
•X-
1—t
CN rH
CO
iO
CO
Ö
Ö
LO CO
T—1
CM CD
sg
-
O
1—1
O
Gi
O
'"' Ö
CM CO
CM
o Ö Ö
l^
o" ^
1
iO 00
co"
^-^ ^ o Ö
Oi to 00 iO CM CM to rH 1
CM
1—1
Gi
o o Ö d> Ö Ö CM
co" CO
o
Gi
CO l>
00 O
•5 ^ 'S t
+5 LO
CO w c»
Ö LO LO
Ö
r-3
O
Ti
cC
0)
W
^1
?^ s
CO M 0) -M T3 4^
*-• s
0) -M
M
-2 Ö -0 -^ CD
^ > 73 ^
T3
^°
Is
^
CO flJ
Ji
I^ O 73
rH
Ö
LO CO
r-i
LO CO
Ö
^ ^ <
Ö
O
O CS ^-( ö^
rO
1^
M Ö O
CO rH CO
CO rH rH
CO
^ ?s
CO
:z;
< CQ p:^
^
11
H X Ö
1
1
CJ5 00
CJ
CO o lO T-H CO Ö ^ Ö
4 Social Polarization, Industrialization, and Fiscal Instability
CO
00 00 00 CN O O CTi 00 CX) 00 (M CO
^
Ö ö^ Ö C7i
1
lO 00 o lO lO r-^
CO T-H
o ^ 5 o ^ ^:: T—t 1—I O
O
cq
CO ^ Oq 9 O
T-H
^
00 05 l>
- 1 (^i
^
O
O
.
o
^ O CO CO ^ ^
^ g P CO
rH
CO Oi ^ O ITD O) O) CO '^
0 0 CO LO 0 0 T}< b CM CO CO CO O
C^ Ö
^
CM ^ LO O
CM '«^J^ Oi CO CM CO o O rH O•
rH Oi Tt O
TH
^ CM CM
Ö
CO
o o
o OJ CM rH
^_^••JXt- ^^
Ö1 Ö1
O CD
rH 2 -x- S o o 1u ' Ö
LO rH LO 0 0
Ö
9
• 1-)
1
oT
Ö1
o rH CD 00 CD
CM" o o
Tf rH
CO 00
d>
s
LO" 1—1 •X•XCM o CO T H 05 o CO CO Ö iH 1-5
IJH
1—H
a <
(^ < ^ o Cu
^~^ "—'Ö '—'
CO CM co" CM O CO CM CM CD « ^ « ^ CO O 1
CM CD CM
s ^ g s §2 2
TU P 00 ^
^
^ 2^ b^CM ^
O
CO
O 5 CO
CO ^
•X-
CO £ ; ; k o CM ^ t-
9 o
Ö Ö 0)
Q
^
^ (N (M
o
r-
(M CD CD i O
Ö
O (M O rH r-H t CJi ( M CO
Ö Ö
00 O CO
Ö CO CO l^ O CT) ^
R
00 ID CO
rH
2 S fe
T^t*
Or-;? -^
lO (M
Ö
00
CO O
O
(30 CO
Zl
^
rH (M
o
lO (N
1-H
o ^^
00 ^
<0
CM <=^ CO r-^
Ö ""^
^
ID
^
?ä S o q CM
O
Ö Ö 9 9 CT "^
S S^ ^ ^ •^
CO "^
CD
o 9 Ö Ö CT
;0
'"J
CO CT
<M CO 9
^
•^ CM JS S
CM O - ^ CT CO CO CD CM CD O r H CT CM CD o O Ö rH Ö C) Ö Ö Ö Ö
^ ^
55 2 <^ rH CT
CO
QQ
O
CM CO 0 0 ID rH O
^ ID O
d
O r-^ 9 ^-^ o CD oT J
' ^
N. rH Ö
ID S^ ^
° ?5 § 9 P o i2 ^ CM '"J O rH P 9 CM
fM
S 9 P CO
CO jt
CM O
CT
ID
CT CD
^
,-H 00 00 O 5o CM CO 1-1 CM S Ö r-; 9
il
CO
t > CO CD CD r-i CO
CO r H CM rj<
CT t^
O
Ö Ö 9 9
CM CO
O
Ö O cJ 00 CM t o
rH
00 o O
CT CD CT 0 0 O rH
^
rH O
»-^
ID
o
CO O r^
^ S ^
Ö
CT ^ t^ O UO CO
00
Ö
T-l O (N
Ö
CM
j;^ K2 CM CO b - CO
CT
9 ^ ^ o o Ö rH
O
00 CD ?5 O
Ö
O
rH
Is * CM CM 0 0 ID ^ O ^4
CM CO
r-i CM CD ID
^ ID
o ^;
•XCO
C^l 1-H
o
rH rH CT
2 f^
^ C5
S2
t^ CT 1-H
00 CT
00 CD (N Ö
Ö
ID ID rH rH
^ S
§S
o
0 0 CM CT ID O CD
00 o
CD S
rH rH
Ö
2 f^
2 S
ID j-t
O
0 5 CO CO rH
4.4 E c o n o m e t r i c Evidence
0 0 -XCO tLi ^ J^
tID 00
CO
ID O ID
ID 00 rH CO
ss
^
* CD CO
t^ ^ ,-( ;^ CO ^ rH ^
rH ID ID
CO ^ 00 S ^ ^
csj C5 o i> r^ °9 r^ R ^ "^
CM
rH ID O
ID ID O
ID Tl< O Ö
CO CM ^ CT O CM
Qcj
2 §
^
71
ID
^2
^
^ \ ^^ Q) r H
0)
'^
O 'Td
CO
i::^ CO
bO Jo
CM
^ ca
• rH
Ö
§ ^ ^
CO
TO
^ 53 Ö
O
•5b .S^
si
72
4 Social Polarization, Industrialization, and Fiscal Instability
(assuming that the degree of polarization is constant). Columns (l)-(5) in Table 4.6 report regression results that are based on this specification. As our theory suggests, the multiproduct of AGINI and the standard deviation of net tax revenue (NTRSD) enters with the correct sign (+) and remains significant at the 1 percent level after controlling for initial income levels, volatility of GDP, external shocks, and political instability. Again, the initial income level (LRGDP) is included because poor countries may have relatively inefficient fiscal systems; hence, their fiscal policies may be more prone to inconsistency and volatility. To the extent that cyclical effects on budget deficits are important, the shocks to GDP may be important in explaining volatility of fiscal balance. Thus, we add standard deviation of real GDP (VGDP) to control for shocks to output. In addition, external shocks in the form of substantial changes in the terms of trade have often triggered domestic macroeconomic instability in many developing countries. Here we use a bit different proxy for the external shocks, EXTSH, which is the multiproduct of standard deviation of the first log difference of terms of trade and average share of trade in GDP. In relation to volatility of fiscal balance, whether the external shocks are positive or negative may matter less. Rather, the magnitude of the shocks may be more important. Finally, we control for political instability because changes in government are often accompanied with changes in economic policies as well as changes in ruling ideologies. We expect VGDP, EXTSH, and PINSTAB to be positively associated with greater volatility in fiscal balance. Only the coefficients of EXTSH are highly significant, although they all have the expected sign (+). Columns (6)-(10) in Table 4.6 show that income inequality is positively associated with greater volatility in fiscal balance. The coefficients of AGINI are all significant at 5-10%. So are the coefläcients of the composite index of polarization POLAR. Although the regression results provide some ideas of the most influential factors affecting budget deficits and fiscal volatility, they tell us little about the relative importance of each independent variable. We can, however, use these results to determine the most important factors that contribute to the differences in fiscal outcomes across regions. The average size of public deficits (as a percent of GDP) in East Asia was 2.13 percentage points lower than that of Latin America and 2.61% smaller than that of sub-Saharan Africa for the period of 1970-90. According to column (10) of Table 4.5, slow GDP growth in Latin America and sub-Saharan Africa increases deficits by 1.7 percentage points and 1.77 percentage points relative to East Asia, respectively. Income inequality also explains a substantial amount of the differences in fiscal deficits across regions. Greater income inequality of Latin America and sub-Saharan Africa adds to deficits by 1.3 percentage points and 1.94 percentage points compared to East Asia each. Volatility of fiscal balance as measured by its standard deviation was also lower in East Asia than Latin America and sub-Saharan Africa by 1.692 and 1.688, respectively. Income inequality is responsible for 20% of the gap between
4.5 Concluding Remarks
73
Latin America and East Asia and 30% of the gap between sub-Saharan Africa and East Asia (based on column 10 of Table 4.6). External shock contributes the most to the cross-region differences in volatility, accounting for 21% of the gap between Latin America and East Asia and 71% of the gap between subSaharan Africa and East Asia. Political instability, inflation, and GDP growth volatility (VGDP) also contribute to the differences in fiscal outcomes, but the magnitude of their impact is much smaller.
4.5 Concluding R e m a r k s Given the strikingly different fiscal policy paths among regions, we have argued that the volatile and unsustainable fiscal paths often adopted in Latin American and sub-Saharan African countries can be attributed to social polarization that arises from unequitable distribution of benefits from industrialization, which is rooted in unequal initial income distribution. A highly polarized society may have more acute struggles over government spending, and this may lead each representative policymaker to spend more for her favorite sector, producing bigger deficits and more volatile fiscal outcomes. Econometric results provide support for these predictions. As suggested in the model, income inequality as a proxy of social polarization explains a substantial part of differences in fiscal outcomes across countries. Countries that have suffered the greatest fiscal instability are those with highly polarized economic societies as measured by indicators of income inequality. However, this does not necessarily mean that countries with unequal income distributions must inevitably experience fiscal instability. An important lesson from our empirical results is that good institutions that can resolve confiicts of interest over fiscal policy matter for fiscal outcomes. When the institutions of conflict management are well-established and working properly, suppressing potential conflicts of interest over government spending, the extent of the polarization effect might be less important in determining fiscal outcomes. Our work also presents the first evidence in support of this idea, although recent studies on budgetary institutions found that strong budgetary rules and procedures are conducive to fiscal disciplined^ We suggest that the model developed in this chapter may provide an alternative channel that negatively links income distribution and growth: a fiscal instability channel through which unequal income distribution affects economic growth.^^ In Chap. 6, we empirically explore this fiscal instability channel in relation to long-term economic growth. ^° See von Hagen and Harden (1994) for evidence on Europe and Alesina et al. (1999) on Latin America for example. ^^ This is a channel distinct from the redistributive fiscal policy channel proposed by Alesina and Rodrik (1994) and Persson and Tabellini (1994), whose approaches seem to lack empirical support (Perotti 1996). In Alesina and Rodrik (1994) and Persson and Tabellini (1994), what matters for growth is the distortion caused
74
4 Social Polarization, Industrialization, and Fiscal Instability
4.6 Appendix A. The Social Planner's Solution A social planner is assumed to choose g^ and f^ to maximize the weighted average of the two policymakers' utility functions. As in the previous section, we confine the strategies to linear Markov strategies, gt = X^^t and f^ = x^Rt' The social planner's problem is to then maximize the following objective function, W{x^,x^)^ with respect to x^ and x^^ subject to the government budget (4.7): Wix^x"")
= {alogx^Rt
+ (1 - a)\ogx''Rt}+ß{a\ogx^Rt+i (4.23) + (l-a)logx^i2m},
where a = ^ ^^ . The social planner's optimization problem can be computed in a way similar to each policymaker's maximization problem. max
{(1 + ß)[a\ogx^
+ (1 - S)logx"^ + logi^t]
(4.24)
+ ^[log(l-r(x^+X^-l))]}. The social planner's solution is _ (l+ß){\+T)a
^* -
[l+2/3]r
_ (l + ^)(l + r ) ( l - a )
^' ^""^ ^*
\n;Wr
*'
^^^
Since a^ > a^ and hence a < a^ (similarly, (1 — a) < (1 — a^)), it is straightforward to see that g^ + f^ > gt + ft- Therefore, the social planner's optimal spending level is always lower than the non-cooperative solution of the polarized policymakers, except when a*^ = a^ = \ (i.e., no polarization, 6 = 0). Similarly, the social planner's optimal size of budget deficit is lower than the non-cooperative solution. Only when there is no polarization does the non-cooperative solution coincide with that of the social planner. Note that the social planner's solution does not require the budget to be balanced. It is because the social planner also shares limited altruism towards the next generation. In our model, altruism is limited since each generation derives utility only from its own spending and that of its immediate successor. If the social planner cares about the fiscal spending of all subsequent generations, not merely that of the immediate successor, then the social planner's solution requires that the budget be balanced (see Chap. 2).
by income tax that accompanies redistributive spending. But Perotti (1996) does not find any negative relationship between tax variables and growth.
4.7 Appendix B. Data
4.7 Appendix B. Data
Table 4.7. Original Data Set of Consolidated Public Sector Deficits: 57 Developed and Developing Countries Country 1 Argentina 2 Australia 3 Austria 4 Bangladesh* 5 Belgium 6 Bolivia 7 Brazil* 8 Bulgaria* 9 Burkina Faso* 10 Canada 11 Chile 12 Columbia 13 Cote d'lvorie 14 Denmark 15 Dominica* 16 Dominican Republic 17 Ecuador 18 Finland 19 France
Country 20 Germany 21 Ghana 22 Greece 23 Honduras 24 India 25 Indonesia 26 Ireland 27 Italy 28 Jamaica 29 Japan 30 Jordan* 31 Kenya 32 Republic of Korea 33 Malawi 34 Malaysia 35 Mexico 36 Morocco 37 Netherlands 38 Nigeria
Country 39 Norway 40 Pakistan 41 Paraguay 42 Peru 43 Philippines* 44 Poland 45 Sierra Leone 46 Spain 47 Sri Lanka 48 Sweden 49 Thailand 50 Trinidad k Tobago* 51 Turkey 52 United Kingdom 53 United States 54 Venezuela 55 Zaire 56 Zambia 57 Zimbabwe
* Indicates that data are not available for the period of 1970-79.
75
76
4 Social Polarization, Industrialization, and Fiscal Instability
Table 4.8. Expanded Data Set of Public Sector Deficits: 33 Additional Countries Country 58 Bahamas 59 Barbados 60 Costa Rica 61 Czech Republic* 62 Egypt 63 Ethiopia 64 Gambia 65 Hungary* 66 Iceland 67 Israel 68 Lesotho**
Country 69 70 71 72 73 74 75 76 77 78 79
Luxembourg Malta Mauritius* Myanmar New Zealand Nicaragua* Panama Romania Seychelles Singapore South Africa
Country 80 81 82 83 84 85 86 87 88 89 90
St. Kitts and Nevis* St. Lucia St. Vincent Sz the Grenadines* Surinam Swaziland Switzerland Tonga* Tunisia Uruguay Yemen Yugoslavia
Note that in the expanded data set, the general government deficit data of these 33 additional countries were added to the original data set of consolidated public sector deficits. * Indicates that data are not available for the period of 1970-79. ** Indicates that data are not available for the period of 1980-89. Table 4.9. Pairwise Correlation of Consolidated Public Sector Balance (% of GDP) with Other Variables Series
Correlation P-value
Initial Real Per capita GDP 0.262* Real GDP Growth 0.130 -0.233* Infiation 0.099 External Shocks GINI (High Quality) -0.280* -0.226* ACINI -0.273** Land Gini -0.225* ETHNIC 0.398* ICRGE 0.382* BERI Cabinet Changes -0.201** -0.241* Coups -0.186** PINSTAB Levels of significance are indicated by asterisks: * 1%, ** 5%, *** 10%.
0.001 0.106 0.004 0.245 0.005 0.010 0.017 0.004 0.000 0.000 0.013 0.003 0.023
Economic, Political, and Institutional Determinants of Public Deficits
5.1 Introduction One of the most striking macroeconomic developments during the last three decades is the rise and persistence of large fiscal deficits in a number of developed and developing countries. Many OECD countries have rapidly accumulated government debts by running large deficits since the mid-1970s until very recently, yet there has been a wide variation in the size of public deficits. In the early 1990s, the public deficits among OECD countries ranged from 17.2 % of GDP in Greece and 10.2% in Italy to a surplus of 4.6% in Sweden. This feature is largely true of the developing countries. In 1990, Peru's public deficit was 8.1% of GDP, and Zambia's 8.2 percent, in contrast to Korea's 0.7%.^ Despite the substantial progress made in fiscal reform since the mid1980s, many developing countries still suffer from chronically recurrent large deficits. Brazil's recent financial crisis was closely related to its high budget deficit, which was 8.4% of its GDP (as of January in 1999).^ According to theory of debt (such as the tax-smoothing model of Barro (1979)), we should expect to see fiscal deficits arise when government spending is temporarily high or when output is temporarily low. However, this prediction is hard to reconcile with the reality of large deficit size and wide variations among countries. A growing literature on fiscal politics has developed theories to explain how political and institutional factors determine fiscal outcomes and has provided supporting empirical evidence; see, for example, Alesina and Perotti Reprinted from Journal of Public Economics, Vol. 87, No.3, March, 2003, pp 387-426, Woo J, "Economic, Political, and Institutional Determinants of Public Deficits", with permission from Elsevier. The fiscal stances of many developing countries were even worse in the 1980s. For example, Mexico's public deficit was 15.4 percent of GDP in 1982, Zambia's 28.5 percent in 1986, and Cote d'lvoire's 14.4 percent in 1989, compared to Korea's largest deficit of 4.3 percent in 1982. See Newsweek (January 25, 1999) on Brazil's currency crisis.
78
5 Economic, Political, and Institutional Determinants of Public Deficits
(1995) and Persson and Tabellini (1997) for a literature survey. However, the existing empirical studies are mostly on OECD countries. Moreover, there is no comprehensive empirical evaluation that compares different theoretical explanations with each other.^ The objective of the chapter is two-fold. First, we provide a comprehensive empirical test on a large pool of potential explanatory variables of public deficits in a panel of 57 developed and developing countries over the period of 1970-90, trying to derive robust conclusions about which of these variables are important in explaining cross-country diflPerences in deficits. We consider three types of structural determinants of deficits: (i) political factors such as political instability, government fragmentation, and political institutions, (ii) social polarization such as income inequality and ethnic divisions, and (iii) institutional factors such as budgetary procedures and rules, bureaucratic eflSciency, and democracy. This chapter not only examines a larger set of the political and institutional variables than the existing literature does, but also introduces a new channel that links social polarization to fiscal deficits. Perhaps social polarization is one of the oldest ideas found in the political economy literature, but it has largely been ignored in the empirical studies of fiscal deficits.^ We present strong econometric evidence in support of this channel. In fact, income inequality is among a handful of consistently significant determinants of fiscal deficits which include financial depth, assassinations, cabinet size, and centralization of authority in budgetary decisions. A comparison of the magnitudes of the effects these variables exert on deficits yields interesting policy implications. For example, a decrease in income inequality of ten Gini points is associated with a reduction of deficits of 1.9% of GDP. A reduction of cabinet size by ten ministers brings about a similar
A notable partial exception to this generalization can be found in the work of Kontopoulos and Perotti (1999) and Pranzese (1999). But these papers also study only OECD countries and significantly differ in the coverage and method. Among others, Roubini and Sachs (1989a, b), Grilli et al. (1991), Edin and Ohlsson (1991), De Haan and Sturm (1994, 1997), and Alesina et al. (1997) discuss political and institutional characteristics of OECD countries such as the government regime type, t*he party structure, and the degree of stability of the political system. Roubini (1991) and Edwards and Tabellini (1991) focus on the degree of political instability in relation to budget deficits in developing countries, von Hagen (1992), von Hagen and Harden (1994), and Hallerberg and von Hagen (1999) study budgetary institutions in the European Union. Alesina et al. (1999) discuss budgetary procedures in Latin America, while Lao-Araya (1998) studies Asian countries. A partial exception is Berg and Sachs (1988), who find that income inequality is positively associated with the likelihood of having rescheduled external debt. Chap. 4 presents the first econometric evidence that income inequality is a significant determinant of fiscal deficits in a panel of 90 countries.
5.1 Introduction
79
effect on deficits.^ Given the difficulty of improving the income distribution in the short term, deficit reduction may be achieved relatively easily by reducing cabinet size. Secondly, our study goes beyond testing different theories of fiscal deficits. We also propose a working hypothesis and provide supporting econometric evidence: sociopolitical polarization is important in explaining differences in fiscal outcomes across countries, yet its effects may be even more pronounced or suppressed, depending on the political and institutional structures through which social polarization is linked to the fiscal policy-making process. Indeed, there is a rich interaction between sociopolitical and institutional factors in determining fiscal outcomes. Effects on public deficits of the sociopolitical variables tend to be smaller in countries with better institutional arrangements. Conversely, sociopolitical polarization exerts very strong effects on deficits in the presence of poor institutions. Most of the existing empirical studies have focused on a relatively small set of explanatory variables. However, one potential problem with drawing strong conclusions from a cross-section study on a large set of explanatory variables is that the partial relationship between a variable and the public deficit may be sensitive to the inclusion of other variables. This is well illustrated by Levine and Renelt (1992), who apply the extreme bounds analysis (EBA) to identify "robust" empirical relationships in the economic growth literature, which consists of running a large number of regressions with different combinations of conditioning variables. They conclude that very few (or no) variables are robust. Given potential multicoUinearity, simultaneity, and measurement errors among many of the variables, the finding of Levine and Renelt (1992) is hardly surprising. As Sala-i-Martin (1994) notes, this approach may amount to having a data under-mining problem. Although neither of these problems can be completely solved, we address the robustness of our econometric results by conducting an extensive sensitivity analysis. Ordinary least squares (OLS) estimates tend to be sensitive to outliers, either observations with unusually large errors or influential observations with unusual values of explanatory variables (often called leverage points). One of the most common ways to deal with outliers is to drop observations one at a time. But this is often inadequate because it may miss a group of outliers due to the masking effect. Similarly, single-case diagnostics that are often used to detect outliers (such as Cook's distance measure, the studentized residual, and DFITS) may fail to identify a group of outliers. We check the robustness to outliers by employing a robust estimation procedure that is based on the least median of squares (LMS), due to Rousseeuw (1984). In the next section, we briefly describe our data and begin with economic variables. In Sects. 5.3-5.5, we present the main results bearing on the alterna^ This comparison is based on the weighted average of the estimates of each variable in the 560 regressions with different combinations of conditioning variables (Table 5.8).
80
5 Economic, Political, and Institutional Determinants of Public Deficits
tive theories. We report the robustness test results in Sect. 5.6, and conclude in Sect. 5.7. We provide detailed information on data including the countrylist, source, and summary statistics in the data appendices.
5.2 Economic Variables on Public Deficits We begin by quantifying the empirical relationship between long-term public sector balance and a wide range of economic variables. We then broaden our scope by examining sociopolitical and institutional variables. By doing this, we can be more confident that our results shown later do not merely capture residual effects of other economic variables. Here we use panel data on the consolidated public sector balance from Easterly et al. (1994). Many of the existing studies on fiscal deficits rely on the central government balance data from IMF's Government Finance Statistics, but the coverage of developing countries has some important gaps. Moreover, using only the central government deficit data can be problematic because a large part of the fiscal deficits in developing countries is attributable to the rest of the public sector, such as state-owned enterprises.^ Our panel is comprised of decade averages of variables for 1970-79 and 1980-90 for 57 developed and developing countries.^ 5.2.1 The Benchmark Framework Our basic regression specification is as follows: CPSURPit
= aiLRGDPit
+ a2GRGDPit + a^INFLATu
+ ßXu (5.1)
where i and t denote the country and decade (1970-79 and 1980-90), and e is an error term. CPSURP is the average of the public sector surplus (as a percentage of GDP). According to the tax-smoothing model of fiscal deficits, budget deficits will emerge when output is temporarily low or when government spending is temporarily high (compared to their permanent levels). We Easterly et al. (1994) illustrate how misguided a narrow measure of the fiscal deficit can be. Alternatively, some studies use as a measure of deficit the yearto-year change in the government debt to GDP ratio (for example, Roubini and Sachs 1989a, 1989b; Alesina et al. 1997 on OECD country deficits). But data on government debt are not available for many developing countries. We use decade averages of variables for a number of reasons. First, we focus on the long-term behavior of fiscal deficits in relation to structural variables. Second, for many of the variables we consider, the original data coverage especially for developing countries is quite often spotty with many annual observations missing. More importantly, some variables—institutional quality indices, various indices for budgetary institutions, and political regimes, for example—are not available for time intervals finer than ten years. As a compromise with the purpose of expanding the sample information we settled on the period of ten years.
5.2 Economic Variables on Public Deficits
81
follow the existing studies to include the decade average rate of real GDP growth (GRGDP), which may be a proxy for the different degree to which countries in question faced economic recessions during the sample period. We expect GRGDP to have a positive (+) sign in the regressions.^ However, it is theoretically possible that GRGDP is negatively associated with the public surplus if the successful pressures for higher public expenditures accompany the growing tax revenue due to higher economic growth. Thus, the sign of the coefficient of GRGDP is an empirical question. We also include the decade average of the rate of inflation of the consumer price index (INFLAT). Inflation can affect fiscal deficits through various channels. Rapidly rising inflation can raise fiscal deficits through higher nominal interest payments. High inflation can also lead to lower real tax revenues. If, however, income taxes are not indexed to inflation, the above effects of inflation on deficits can be at least partially offset by the positive effect of bracket creep on income tax revenue. Given the data availability, it is hard to distinguish between the nominal and real components of interest payments in government spending. Hence, we try to capture the effects of inflation on flscal deficits by including INFLAT, and expect it to have a negative (-) sign.^ The log of real per capita GDP at the start of each decade, LRGDP, is introduced in order to control for potential effects of economic backwardness on public deficits. Poor countries may have relatively inefficient tax and spending systems and may therefore be more prone to budget deficits. Alternatively, LRGDP may capture some sociopolitical effects if social conflicts are greater in poor countries (Roubini 1991). In addition to decade dummies, A (DUM70 and DUM80), we include three regional dummies, RDi: EASIA for East Asia, LATINCA for Latin America and the Caribbean, and AFRICA for sub-Saharan Africa. Regional dummies are intended to control for structural characteristics related to geographical location but not to sociopolitical and institutional factors. We al^ One can think of another channel that a growing economy has more resources and may be better positioned to solve socio-economic distributional problems, which may help reduce deficits. ^ Using INFLAT as a regressor is somewhat subtle because inflation can be related to seigniorage, which can in turn be viewed as being jointly determined with the level of fiscal deficits. However, the view that inflation is determined largely by the monetary financing requirements of the government is empirically weak. First, seigniorage itself is quite small in most countries except during times of crisis. Even this temporary seigniorage increase during those times is only weakly related to inflation. Easterly et al. (1994, p. 47) conclude that "a surprising number of episodes of high seigniorage are attributable to increases in real money balances instead of to accelerating inflation." Yet inflation can affect fiscal deficits in various ways mentioned above, so we include INFLAT to control for the effects of inflation on deficits as in Kontopoulos and Perotti (1999). It should be noted that our main results do not change qualitatively even if INFLAT is dropped from the basic specification.
82
5 Economic, Political, and Institutional Determinants of Public Deficits
ways include the aforementioned economic variables in our entire econometric investigation. Finally, Xu represents other variables, which we specify later. ^^ Table 5.1 reports the regression results based on the basic specification (5.1). Because heteroskedasticity may be more important in a cross-country sample, the standard errors of the coefläcients are based on White's (1980) heteroskedasticity-consistent covariance matrix, which reduces the sensitivity of inference and hypothesis test using OLS estimator to general form of heteroskedasticity. However, these standard errors are not greatly diflferent from those obtained from ordinary least squares. As a more complete way of robustness check, we also present the sensitivity analysis and the robust estimation results later. The estimated coeflficients of LRGDP and GRGDP are significant at the conventional level and have the expected sign (+). CoeflScients on the regional dummies are largely insignificant, but EASIA is significant at 10% in columns (3) and (4) and positive (+), reflecting the fact that most of the East Asian countries have run smaller deficits than other developing countries. INFLAT enters with a negative sign (-), but is insignificant. It is significant only in the developing country sample (column 6). Column (2) includes the ratio of liquid liabilities of the financial system to GDP at the start of each decade (ILLY) as a proxy for the financial market development level, "financial depth." One can imagine that countries with highly developed financial markets can more easily finance the fiscal deficits by issuing bonds without having to resort to inflationary finance. ^^ The coefficient of ILLY is highly significant and of the correct sign (-). The estimate suggests that a 10 percentage point increase in ILLY is associated with an additional deficit of 0.6 percent (as a percent of GDP). In column (3), we include POP65 (share of the population over 65 years of age). Larger deficits are expected to be seen in countries with high POP65, presumably due to greater social welfare spending on pensions and health care. Not only is the POP65 value insignificant, but it also has the wrong sign (+). ^° Regarding the basic regression specification, it may be desirable to include a measure of debt servicing costs. However, the public debt data that is consistent with our public deficit measure is largely unavailable. One can, however, try the central government debt data from World Development Indicators, although it is available for a limited number of countries, especially developing countries. A measure of debt service costs (real interest costs) using the central government debt data turns out to be largely insignificant, while it substantially reduces the number of observations. More importantly, it does not significantly change our main results. ^^ Related to this point, Caballero and Krishnamurthy (2004) argue that the financial depth is an important factor in determining whether a certain size of fiscal deficit can be sustained without necessarily leading to a financial crisis. They present evidence that advanced and emerging economies differ in their financial depth, and show that lack of financial depth constrains fiscal policy in a way that can overturn standard Keynesian fiscal policy prescriptions.
5.2 Economic Variables on Public Deficits
83
Table 5 . 1 . Economic Variables on Public Surplus: Pooled Decades (1970s and 1980s). Dependent Variable: Consolidated Public Surplus (Percent of GDP)
Variables DUM70 DUM80 EASIA LATINCA AFRICA LRGDP GRGDP INFLAT
(1) -24.17* (-5.43) -25.51* (-5.92) 0.57 (0.49) 0.26 (0.31) 1.21 (1.03) 2.26* (4.74) 0.48* (2.66) -0.001 (-0.81)
(2) -27.72* (-6.15) -28.92* (-6.61) 1.20 (1.01) -0.69 (-0.67) 0.83 (0.68) 3.05* (5.38) 0.43** (2.47) -0.002 (-1.53) -0.06** (-2.2)
(3) -23.15* (-4.27) -24.27* (-4.49) 2.14*** (1.84) 0.45 (0.41) 1.37 (1.17) 2.19* (2.84) 0.44** (2.52) -0.002 (-1.49) -0.06** (-2.28) 0.29 (1.65)
99 0.70
95 0.71
95 0.72
ILLY POP65 EXT WAR No. of Observations Adj.Ä^
Developing Countries (4) (7) (5) -23.52* -23.68* -15.26* -14.34** (-4.30) (-4.30) (-2.38) (-2.30) -24.63* -24.82* -14.65* -13.91** (-4.52) (-4.52) (-2.29) (-2.22) 1.98*** 1.96 1.42 1.62*** (1.68) (1.64) (1.43) (1.69) 0.32 0.35 1.28 2.00 (0.79) (0.29) (1.01) (1.37) 1.33 1.39 0.05 0.36 (0.04) (0.26) (1.12) (0.25) 2.24* 2.25* 1.60*** 1.43 (2.87) (2.86) (1.54) (1.75) 0.43** 0.44** 0.43** 0.42** (2.51) (2.49) (2.65) (2.58) -0.002 -0.002 -0.002*** -0.003** (-1.35) (-1.38) (-1.76) (-2.19) -0.06** -0.06** -0.11* -0.10** (-2.20) (-2.05) (-3.16) (-2.46) 0.28 0.29 -0.55 -0.74*** (1.59) (1.58) (-1.42) (-1.77) 0.23 0.23 0.27*** 0.25 (0.19) (1.3) (1.81) (1.60) 0.20 1.18 (1.32) (0.23) 94 62 94 62 0.72 0.84 0.85 0.71
~W)
White heteroskedasticity-consistent t-statistics are reported in parentheses. Levels of significance are indicated by asterisks: * 1%, ** 5%, *** 10%. See data appendices for definitions and sources.
Column (4) adds growth rate of terms of trade multiplied by trade openness (EXT) as a proxy for external shocks to the economy. External shocks can be a source of fiscal instability, especially in many developing countries. Changes in export and import prices can aflPect the public sector balance either through the profits of exporting public enterprises or through import tariflPs and taxes on exports. The growth of terms of trade is expected to be associated with smaller budget deficits and to have a greater impact in economies that are more open to trade. EXT has the expected sign (-h), but is not sig-
84
5 Economic, Political, and Institutional Determinants of Public Deficits
nificant.^^ Column (5) shows the regression that includes the dummy for war on national territory during the decade (WAR). Higher government spending during a transitory period such as war can cause deficits to rise, but WAR is not significant and has the wrong sign (+). Finally, columns (6) and (7) show the regressions using the developing country sample only. Now the coefläcients of INFLAT become significant at the 5-10% level and are of the expected sign (-). Adjusted R^ rises to 0.840.85. GRGDP, and ILLY are still significant, suggesting that deficits in the developing countries are equally well explained by these economic variables. Note that the statistical significance of LRGDP becomes weaker. This makes sense because diflPerences in the initial income are smaller among relatively poor developing countries and hence its standard error may be larger. Since the indicator ILLY is consistently associated with larger deficits, our benchmark regression specification includes ILLY in addition to basic economic variables from now on.^^
5.3 Fiscal Politics 5.3.1 Political Instability The political instability approach suggests that the less likely is re-election, the more likely the incumbent government is to leave a large stock of public debt. Faced with the uncertainty over re-election, the incumbent government may fail to internalize the costs of additional debt because these costs are borne by the succeeding government that may be controlled by the opposition party with different preferences. Public deficits should thus be larger in countries with more frequent changes from one party or leading group to another. ^^ ^^ One can think of an alternative measure of external shocks: the standard deviation of the growth rates of the terms of trade multiplied by trade openness (EXTSH). This measure, however, treats positive terms of trade shocks symmetrically with negative shocks; hence, it may not be the best measure for capturing the external effects on government revenue or spending. Still, it may be related to public deficits. For example, large and volatile external shocks can damage economic activity and decrease GDP growth, which in turn affects the deficit. EXTSH is not significant, either, although it has the expected minus (-) sign (not reported to save space). ^^ We tried other variables such as black market premium, agricultural share in GDP, and urban population ratio that were considered by other studies (for example, Edwards and Tabellini 1991). They are all insignificant in our basic regressions. ^^ For example, see Alesina and Tabellini (1990) and Persson and Svensson (1989). The earlier empirical evidence is somewhat mixed. Edwards and Tabellini (1991) and Roubini (1991) find that political instability proxied by the frequency of government changes leads to larger central government deficits in developing country samples. But Grilli et al. (1991) find no evidence for this in a sample of 18 OECD countries.
5.3 Fiscal Politics
85
The first five columns in Table 5.2 report the regressions using frequency of cabinet changes (CABCHG), coups d'etat (COUPS), changes in effective executive (EXECHG), and major constitutional changes (CONSTCHG). The coefficients on all of these indicators have the expected sign (-), but only those of COUPS and CONSTCHG are significant at the 5% level. Political uncertainty is a multidimensional phenomenon that cannot be captured by a single variable. In relation to the uncertainty about government survival, we include broad measures of both violent and nonviolent social unrest which were not considered in the existing literature: political assassinations (ASSASSIN), government crises (GOVTCRIS), and revolutions (REVOLS). High levels of social and political unrest might be strong expressions of dissatisfaction with the current government and its policies, which may not only make the downfall of the present government more likely but also may dramatically shorten the horizons of politicians. With a shortened expected tenure in office, the government would be more likely to engage in short-term policies at the expense of macroeconomic stability. Columns (6)(8) in Table 5.2 display the regressions when alternative indicators of political unrest are added to our basic regression. Again, all have the expected sign (-) and are statistically significant at various levels. To capture this multidimensional political instability, we construct a composite index by applying the principal components analysis to the five variables found to be significant above: PINSTAB=0.31COUPS+0.02CONSTCHG-h0.19GOVTCRIS -|-0.18ASSASSINH-0.4REVOLS.i^ Column (9) shows that the coefficient of PINSTAB is highly significant and of the expected sign (-).^^ The estimate implies that one standard deviation increase in PINSTAB raises public deficits by 1.02% of GDP. The statistical significance and sign of the coefficient of PINSTAB remain unchanged even if we include other variables such as POP65 and EXT in the regressions. The last two columns show the regressions in the developing country sample. They are much the same as in the entire sample except that the coefficient of INFLAT is now significant at the 10% level. One might be concerned with potential reversed causality from deficits to sociopolitical instability. However, this case seems to be fairly weak. Of course, to the extent that public deficits can cause serious macroeconomic Principal components analysis is a statistical technique that helps us to reduce the number of variables in an analysis by describing linear combinations of the variables that contain most of the information (i.e., linear combinations with the greatest variance). All the variables that are included in PINSTAB are standardized so that they have a mean of zero and standard deviation of one at the outset. Even if we include CABCHG and EXECHG in the composite index PINSTAB, it does not increase its explanatory power. Conversely, the statistical significance of PINSTAB is driven by those five significant variables.
86
Q
«-I-H
o
CO OH
o 00
-4-3 Ö CD
T3 o
§
^ ^
c« CD O M o; b- OH CT) CD J3 CO
CO ^~-^ CQ
cu
CO
T3
3 ^ CD
T3
PH
Q _o o OH
CO
^
1
3Co
h^
le +>. 3 r2 ^CO CoÖO ^M ÜO ^o
c^ Iß 0
CD T3 Ö CD
a
a;
H Q
3Co
00
O
^
•X-
Ö
•X-
w
s r-^
f l>- -XC^ CO
C;:^
»>
00 (M
rH
* CD
•XCO
o
CD
CO
P
CO
C35
rH 1
T-l
^ p <-i
CNI
CO
^ 1^o
s ^ CM
lO
C^ o
O ^
O
^ - v -X-
o 3_
O O
c
CO
00 rH
o --^
^
O
^
^
*^
i^ o 2 IL S' CO
O o
o
lo
oo
CO 00 CD
o ffi o m u Ü-<
CD T:JH
o^ c^
Ü
^ O
CU
CO
00 ^ CS O i t-H " ^
Ö 9 <^ Ö ^
«^i d 9 S <^^ CO lO
S
^
CO 1-1 Cq
o w X w
Ü ffi
o K
Ü
H
CO
^ O Ü
O O
>
p^ Ü
^ ^
g
00 CO
<^
p:^
o >
CO
o ^
m CO
g
s
en
O "o 6
o
^ <
CS
CD
CO
O . CQ M •^
CD ^ W Ö CD Co
CD ; 3
H
X! CD
Co U
0) T3
G
O
CD
A
S
*
^
CD CO
Co T3
S?'
^ -B S
«^
^ :S ^
^ CD XJ
CD -03 S
CD «-<
bib's
CD
.2 ö
M
o *
2^^
5 Economic, Political, and Institutional Determinants of Public Deficits
^
•XC:5
•X-
CO
CO
05 "^ "^
CO
CO ^f CO CO
Q
9 o
5.3 Fiscal Politics
87
instability such as hyperinflation or poor economic growth, they may foster violent behaviors of both political and non-political motives. Yet one can still argue that public deficits would have only an indirect effect (if any) on sociopolitical unrest. Perhaps a more serious problem in interpreting our regression results is the interaction between the indicators of sociopolitical instability and the indicators of social polarization (such as income inequality and ethnic division). Alesina and Perotti (1996b) find evidence that income inequality increases political instability. In a society with high income inequality, there might be stronger incentives for different groups to organize and engage in activities outside the market and outside the normal channels of political representation to appropriate some of the resources of the other groups. If so, some of the variation in public deficits captured by sociopolitical instability measures merely reflects the effects of income inequality. We return to this issue in Sect 5.4. 5.3.2 Government Fragmentation The government fragmentation approach stresses the distributional conflict among different parties or policymakers, which is often associated with government weakness such as coalition or minority governments. Coalition governments may delay the necessary fiscal adjustments after the build-up of large deficits due to some adverse exogenous shocks because political parties struggle over the distribution of adjustment burden (Alesina and Drazen 1991). Government weakness, therefore, implies that public debt is a residual source of finance that simply reflects the inability of the government to cut expenditures or raise taxes.^^ To test this prediction, we first consider direct measures of government weakness, the number of seats held by the largest party in the lower house (SEAT) and the party fractionalization index (PARFRACT). We expect deficits to increase with the degree of government weakness.^^ Many studies have suggested that coalition governments not only tend to be weaker than single-party majority governments, but also may have greater difficulty in reaching consensus. To capture this, we employ a variable COAL, which ^^ Roubini and Sachs (1989a, b), Grilli et al. (1991), and Alesina et al. (1997) present supporting evidence of the weak government argument. Sachs and Roubini (1989a, b) find that in a panel of 15 OECD countries the size and persistence of budget deficits is greatest where there have been multiparty coalition governments. However, Edin and Ohlsson (1991) find that this result is mainly due to minority governments. De Haan and Sturm (1994, 1997) do not find even this. ^^ Note that the indicator SEAT is the ratio of legislature size to the number of seats held by the largest party; therefore, higher SEAT values represent a weaker position of the largest party in the legislature. PARFRACT is defined as the probability that two randomly chosen legislators belong to different parties. The higher the index, the larger the number of parties in the legislature.
88
5 Economic, Political, and Institutional Determinants of Public Deficits
is a decade average of annual observations that take the following values: 0 for a one-party government with no major opposition party in the legislature; 1 for a coalition government with more than one party but with no major opposition party; 2 for a coalition government with more than one party but with a major opposition party in the legislature; and 3 for a minority government. This index is similar to the one used by Roubini and Sachs (1989a, b), but our variable does not distinguish between the presidential system and the parliamentary system. This distinction is discussed in the next subsection. Table 5.3 shows somewhat disappointing results. Only the coefläcient of COAL has the expected sign (-), but it is not statistically significant.^^ Edin and Ohlsson (1991) and Kontopoulos and Perotti (1999) argue that the evidence that weaker governments are associated with larger deficits rests mostly on the inclusion of minority governments. Column (4) replicates the regression using MINOR, which is intended to capture only minority governments. The indicator MINOR is a decade average of minority government dummy variables that take the value of 1 for minority governments, and 0 otherwise. The coefficient of MINOR is insignificant, although it is of the correct sign (-). This is consistent with De Haan and Sturm (1994, 1997) who find that government weakness indicators (including minority government dummy) are not significant for OECD countries. In related work, Weingast et al. (1981) and Velasco (1999) view deficits or overspending as arising from the pork barrel or common pool problem. Under this approach, a deficit can arise because individual policymakers fail to internalize the full cost of their own spending financed through common tax revenues. This lack of coordination may be greater when there are more participants in the decision process—so-called size fragmentation (Kontopoulos and Perotti 1999).^^ Column (5) tests this hypothesis using a measure of size fragmentation, CABSIZE (the number of ministers in the cabinet). The coefficient of CABSIZE is highly significant and has the expected sign (-). Consistent with the finding of Kontopoulos and Perotti (1999) for OECD countries, larger cabinet size is strongly associated with larger public deficits. The magnitude of this ^^ However, the result seems quite sensitive to the way of coding. If we use a coalition government dummy that takes the value of 1 if COAL> 0 (i.e. coalition or minority governments), and 0 otherwise (i.e. single-party majority governments), then its coefficient is of the correct sign (-) and significant at the 5% level (in the basic regression). ^^ However, the theoretical relationship between the number of decisionmakers and the fiscal outcome is not monotonic. For instance, Tornell and Lane (1998) show that the overexploitation of the common resource is greatest when there are only two groups. It then decreases as the number of groups increases because each group's power to extract the common asset declines as more groups try to do so. Clearly, this coordination failure is not government weakness per se. It might, however, be associated with either lack of a strong unitary leadership in a coalition government or the existence of powerful interest groups.
5.3 Fiscal Politics
89
Table 5.3. Government Fragmentation: Pooled Decades (1970s and 1980s). Dependent Variable: Consolidated Public Surplus (Percent of GDP) Variables LRGDP GRGDP INFLAT ILLY SEAT
(1) 3.01* (5.07) 0.43** (2.45) -0.002 (-1.53) -0.06** (-2.15) 0.08 (0.16)
PARFRACT
(2) 2.6* (4.25) 0.42** (2.38) -0.002** (-2.05) -0.05*** (-1.93)
(3) 3.13* (5.10) 0.43** (2.48) -0.002 (-1.52) -0.06** (-2.17)
(4) 3.05* (5.38) 0.43** (2.47) -0.002 (-1.55) -0.06** (-2.17)
Developing Countries (7) 1.31 1.44 (1.62) (1.61) 0.45* 0.45* (2.65) (2.67) -0.003** -0.003** (-2.21) (-2.3) -0.12* -0.12* (-3.09) (-3.07)
"16)
3.17 (1.63) -0.24 (-0.59)
COAL
-0.31 (-0.76) -0.13 (-0.16)
MINOR CABSIZE No. of Observations Adj. B?
(5) 2.99* (5.43) 0.44* (2.66) -0.002 (-1.54) -0.07* (-2.91)
95 0.71
95 0.72
95 0.71
95 0.71
-0.22* (-3.39) 95 0.74
63 0.84
-0.14 (-1.64) 63 0.84
The decade and regional dummies were included in the regressions (not shown to save space). White heteroskedasticity-consistent t-statistics are reported in parentheses. Levels of significance are indicated by asterisks: * 1%, ** 5%, *** 10%. See data appendices for definitions and sources.
effect on deficits is quite substantial. In particular, an increase in cabinet size by one minister is associated with a larger deficit of 0.2% of GDP. Columns (6) and (7) display regressions for the developing countries. Once again, the coefficient of COAL is insignificant, suggesting that the weak government argument is weak. Even the coefficient of CABSIZE is now insignificant. CABSIZE is thus less strongly associated with higher deficits in the developing country sample than in the whole sample. 5.3.3 Political Regime and Electoral Law The political regime and electoral law approach is related to comparative politics. Persson and Tabellini (1999b) study how political institutions affect the size of government spending. Although not aiming to explain fiscal
90
5 Economic, Political, and Institutional Determinants of Public Deficits
deficits, they find that majoritarian elections lead to more redistribution and larger governments and that presidential regimes lead to less redistribution and smaller governments.^^ Under presidential systems in general, the government has greater independent and centralized authority as well as accountability (Shugart and Carey 1992). Hence economic policy can be formulated and implemented without much delay or interference from the legislature and can be more easily reversed if it delivers undesirable outcomes. The opposite may be true of the parliamentary system, depending on the electoral laws and their degree of proportionality. Therefore, we expect different fiscal outcomes across regimes (presidential versus parliamentary) and electoral systems (proportional versus majoritarian). Our indicator of regime type, PRES—a dummy variable taking the value of 1 for a presidential regime—is constructed mainly based on Mainwaring (1993), which is augmented and cross-checked with other sources such as Powell (1982), Shugart and Carey (1992), and Cox (1997).22 The first column in Table 5.4 displays the regression for the entire sample period (1970-90) when PRES is used. The coeflScient of PRES is not statistically significant, although it has the expected sign (+). According to the implied magnitude, public surpluses are 0.9% of GDP larger in presidential regimes than in parliamentary systems. The PRES is again insignificant in the developing country sample (column 2). For lack of data on electoral laws with a comprehensive coverage for the relevant countries and time periods, we use data from Persson and Tabellini (1999b) for the subperiod of the 1980s only. Our indicator for the electoral rule is MA J, which takes the value of 1 for majoritarian elections and 0 for proportional elections. Column (3) of Table 5.4 shows that the coeflicient on MAJ is significantly negative. Political scientists often argue that coalition or minority governments are more likely to emerge under a parliamentary system with proportional electoral laws. Although the government type indicator (COAL or MINOR) itself turned out to be statistically insignificant, this suggests that there may still ^^ Grilli et al. (1991) find that public debts and primary deficits are much greater in the proportional parliamentary regimes in 18 OECD countries. ^^ Persson and Tabellini (1999b) construct a regime type index (PRESPT) based on Cox (1997), but its sample period roughly corresponds to the 1980s. In contrast, Mainwaring (1993) provides a list of 31 stable (1967-92) and a number of unstable democratic countries since 1945 with the classification of regime types. The PRES contains four more observations than the PRESPT in our sample for the 1980s. The difference between the indicators PRES and PRESPT for the countries that are covered by both of them is Prance. We follow Powell (1982) in classifying France as a presidential regime, whereas Persson and Tabellini (1999b) classify France as a parliamentary regime. In fact, somewhat controversial is the coding of regime types for countries with premier-presidential systems such as Austria, Bolivia, Ecuador, Finland and France (Shugart and Carey 1992). We follow Powell (1982) and Mainwaring (1993) in classifying these countries.
5.3 Fiscal Politics
91
be an interesting interaction between regime types and electoral rules. This exercise is reported in columns (4) and (5).^^ The estimated coefficients of the presidential regime dummies (PRES and PRESPT) are both significantly positive once we control for MA J and other variables. Conversely, this implies that—holding the electoral law dummy and other variables constant—the parliamentary regime is significantly, negatively associated with public surpluses. Table 5.4. Presidential vs. Parliamentary System and Proportional vs. Majoritarian Electoral Laws. Dependent Variable: Consolidated Public Surplus (Percent of GDP) Columns (l)-(2) for Pooled Decades (1970s and 1980s) and (3)-(5) for the 1980s
Variables LRGDP GRGDP INFLAT ILLY PRES
(1) 3.36* (4.26) 0.50*** (1.95) -0.001 (-0.97) -0.06*** (-1.71) 0.93 (0.94)
Developing Countries (2) 1.48 (1.41) 0.42 (1.65) -0.003** (-2.09) -0.13** (-2.26) 1.07 (0.63)
67 0.67
37 0.82
PRESPT MAJ MAJPRES No. of Observations Adj. R^
r h e 1980s (4) (5) (3) 4.11* 4.07* 3.87* (3.49) (3.32) (4.19) 0.92*** 0.64 0.63 (1.44) (1.99) (1.56) -0.001 -0.003** -0.003** (-0.99) (-2.17) (-2.23) -0.07 -0.05 -0.05 (-1.56) (-1.09) (-1.09) 5.62** (2.73) 5.84** (2.76) -2.27*** -0.31 -0.77 (-1.71) (-0.15) (-0.36) -5 47*** -4.47 (-1.96) (-1.5) 36 35 35 0.28 0.29 0.29
The decade and regional dummies were included in the regressions (not shown to save space). White heteroskedasticity-consistent t-statistics are reported in parentheses. Levels of significance are indicated by asterisks: * 1%, ** 5%, *** 10%. See data appendices for definitions and sources.
The magnitude of effects on deficits of regime types varies somewhat systematically across electoral rules. For example, the presidential regime (par^^ In column (4), PRESPT and MAJ from Persson and Tabellini (1999b) were used, whereas PRES instead of PRESPT was used in column (5).
92
5 Economic, Political, and Institutional Determinants of Public Deficits
liamentary) is associated with larger surpluses (larger deficits) under proportional law than under majoritarian law. According to column (4) in Table 5.4, the partial eflFect on deficits of PRESPT, holding all other controls constant, is (5.84-5.47MAJ)*PRESPT. Thus, it becomes 0.37PRESPT under majoritarian rule (MAJ = 1). Under majoritarian rule, being in a presidential regime increases the public surplus by 0.37% of GDP. We can rewrite the regression in column (4) with respect to the parliamentary system (PARL =1-PRESPT), where PARL is a dummy variable taking the value of 1 for a parliamentary regime. The partial effect of a parliamentary regime under majoritarian elections is -0.37PARL. The partial eflPects of PRESPT and PARL under proportional law are 5.84PRESPT and -5.84PARL, respectively. This implies that parliamentary regimes tend to run larger deficits under proportional laws than under majoritarian laws if other variables are held constant. This may be explained by Tsebelis's veto player hypothesis that regime instability is associated with a larger number of veto players that lack ideological cohesion (Tsebelis 1995).^^ According to this hypothesis, one can expect more veto players to be associated with larger fiscal deficits. Majoritarian rule tends to have two parties, and therefore, only one veto player in the parliamentary regime. By contrast, proportional parliamentary systems tend to have a greater number of veto players. This may induce proportional parliamentary regimes to run larger deficits.^^ In short, public deficits are lower in the presidential system than in the parliamentary system once we control for electoral law and other variables. We also find econometric evidence that the combination of a proportional electoral law and a parliamentary system can be a dangerous mix for fiscal outcomes.
5.4 Social Polarization: Income Inequality and Ethnic Divisions Income inequality has long been mentioned as an important source of social conflict, which may lead to populist fiscal policies and poor macroeconomic performance. This has been extensively documented in studies on Latin America and sub-Saharan Africa.^^ Yet there are very few theories that explain why unequal income distributions can lead to large deficits. Alesina and Rodrik ^^ I thank an anonymous referee for suggesting this interpretation, and George Tsebelis for further discussion in a private correspondence. ^^ However, the same reasoning is less directly applicable to presidential regimes because presidential regimes have much the same number of institutional veto players across electoral rules. Whether there are actually more veto players or not depends on which veto players get 'absorbed.' For example, in a presidential system in which the president and the majority leader in the congress belong to the same party, there is only one veto player. 2^ Rodrik (1996), Kauffman and Stallings (1991), and Berg and Sachs (1988) among others.
5.4 Social Polarization: Income Inequality and Ethnic Divisions
93
(1994) and Persson and Tabellini (1994) suggest that there may be a tendency of the majority to vote for large redistributive spending in a democratic country with an unequal income distribution. In Chap. 4, we develop a model of fiscal deficits in which the polarization of preferences play a crucial role in the evolution of fiscal deficits. The idea is that in a two-sector economy, the more unequal the initial income distribution, the larger the sectoral income gap during industrialization, and the more likely the polarization of sector preferences for different types of government spending. In a highly polarized society, policymakers face greater incentives to insist on higher spending for their preferred sectors, leading to larger deficits. This model yields a sharp empirical prediction that fiscal deficits are larger in countries with highly polarized societies (as measured by income inequality). There are very few empirical studies on the impact of income inequality on fiscal deficits.^^ Chap. 4 provides the first econometric evidence that income inequality is a significant determinant of public deficits, even after controlling for a number of other variables in a sample of 90 countries for the period of 1970-90. As an indicator of social polarization, we first consider the three measures of income inequality: GINIHI, AGINIHI, and AGINI from Deininger and Squire (1996). The indicator GINIHI is high-quality data of Gini coefficients measured as close either to the end of previous decade or to the start of each decade, 1970s and 1980s, as possible. The indicator AGINIHI is the decade average of all high-quality data of Gini coeflBcients. Finally, the indicator AGINI is the decade average of all available data of Gini coefficients.^^ We expect a high degree of social polarization as proxied by Gini coefficients to be associated with larger public deficits. Table 5.5 delivers a striking result: the coefficients of the income inequality indicators are all highly significant. This remains true even after controlling for other variables, as will be shown later. The first three columns show the Perhaps this is partly due to the unavailability of comprehensive high quality data on income inequality until recently when Deininger and Squire (1996) made such a data set available. A partial exception is Berg and Sachs (1988), who study the relationship between income inequality and external debt crises. To avoid reverse causality from public deficits to income inequality, GINIHI was limited to countries for which we have high-quality data measured close either to the end of each previous decade, the 1960s and 1970s, or to the start of each decade, the 1970s and 1980s. However, income inequality measured by Gini coefficients is very persistent over time as evidenced by high correlation among GINIHI, AGINIHI, and AGINI. The pairwise correlation between GINIHI and AGINIHI is 0.93, and the correlation between GINIHI and AGINI is 0.92. The use of GINIHI tends to produce even stronger results that support the social polarization hypothesis. Moreover, if one is concerned with measurement error (especially in a specific year), using a 10-year average of Gini coefficients like AGINIHI or AGINI might be preferable. Therefore, we report regression results, primarily using AGINI to maintain the largest number of observations possible.
94
5 Economic, Political, and Institutional Determinants of Public Deficits
regressions, using each of GINIHI, AGINIHI, or AGINI, while controlling for other economic variables. Their coefficients are all significant at the 1% level, and the size of coefficients is between -0.15 and -0.19. According to the point estimate, an increase in inequality of ten Gini points is associated with an increase in the public deficit of 1.5%-1.9% of GDP. There are a couple of things to note regarding the interplay between income inequality measures and other variables. The age structure of the population may be correlated with income inequality, and of course, with initial income levels as well. While income inequality would be lower among the elderly, income inequality may affect the life expectancy and presumably reduce the ratio of the elderly in the population. Indeed, Gini coefficients and POP65 are strongly and negatively correlated. Including both POP65 and inequality indicators may thus cause high variances of their estimated coefficients. However, inequality is still significant at 1%, although POP65 remains insignificant (not reported to save space). Similarly, the level of initial income may be correlated with the income distribution (the Kuznets hypothesis), yet the statistical significance of Gini coefficients remains the same. Next, we consider ethnic divisions, which are another source of social polarization. Our indicator of ethno-linguistic fragmentation, ETHNIC, measures the likelihood that any two randomly chosen individuals in a country will not belong to the same ethno-linguistic groups (Taylor and Hudson 1972). The value of ETHNIC increases with the number of ethnic groups in a country. Column (4) in Table 5.5 shows the regression result when ETHNIC is included. It enters with the expected sign (-) and its coefficient is significant at 10%. Since income inequality may reflect a different dimension of social polarization than ETHNIC does, we include both AGINI and ETHNIC in the regressions (column 5). The coefficient of AGINI is of the correct sign (-) and is significant at the 1% level, but the coefficient of ETHNIC becomes insignificant.^^ We apply the principal components method also to AGINI and ETHNIC and create a social polarization composite index, SOCPOLA. Column (6) confirms that social polarization is important in explaining the deficits. Now we return to the issue of potential interaction between social polarization and socio-political instability. First of all, there is not much correlation between income inequality and the indicators of sociopolitical instability. Among sociopolitical instability indicators, ASSASSIN is the only variable whose correlations with Gini indices are significant, yet their pairwise correlations are less than 0.29. Thus, income inequality and sociopolitical instability seem to have quite independent effects on public deficits.^^ Columns (7)-(9) display regression results when AGINI and indicators of sociopolitical instability are included together. It is remarkable that AGINI, ^^ Other income inequality indicators are all significant at 1%, while ETHNIC is not. ^° If we regress PINSTAB on Gini coefficients and dummies (with or without other variables), the estimated coefficients of income inequality are all insignificant.
T3
o o
5
13
(V
a; ^
O l-H
II o
;>
:?? CO
Co
d
'S ^
o
O
«? ^
1:0 '^
O
CO
q
"^
rH
rH
q 1o q
CO *
s^
a
l-J
>3
^
CO
CO CO CNI
CD T-H
^ ^
'^
w
Ü
q
Ü
r-i
<
^_l
<
Ü
s s 0g
ss
^
I
m CO
< H g E
CO
q
" ^
<^! t^
T-l
^*
2
CQ
^
S
It,
^—' ^ c^ ^
CO
"^^
^
w N
PH
CO
CQ
< 0
Ü
;:^ 0
CO
r-^ '^i' Ö 9
§
O «? 2
CO I L
r- ^
*
OJ ^
PH
O Ü
CO
O
00 CO r-A
O ^i
10 00
!H
> Co
L. •
0
^ CO CO
Ö ^ ^<
< <^ CO CO
g
CO r-i
5.4 Social Polarization: Income Inequality and Ethnic Divisions
(M
CO o O
(M
^ r t d £^. o
II ^
CO -X-
q
l^ O
o O
CO
Ö
(-J
C75 ^ QÖ ~
Ö m!
ö
° C o 3 ö 2.9
^ q § ^ § ;i •;
10
Tti
*
. S ^ q
lO
^
^ CO CO CO
O O
O) ^
lO
^ ^
^ to
•«•
00 nn
* rr
CO CO
O) CO
(M -XCO LO CO CM
CO
•)«• rH
CO O
;i! §
CO ^ lO " ^
*
00 t-
IIM
5 ^ 9 9 § <^
^ ^
CO
PH
Q Ü
Ü
95
o *
1^^ CD v p
-1-3
CD - t J -X-
^ "^ Co CD cn J j O '^
li1^
.0
O
2:^ 9^ «!^
^
Co X )
^
cp 0)
T3
CO cc A,
0 0
* -x-
1*
^2
CD
CD
CO
T3
•73 CD
96
5 Economic, Political, and Institutional Determinants of Public Deficits
SOCPOLA, PINSTAB, and CABSIZE are all significant at the l%-5% level. If we use indicators of sociopolitical instability instead of PINSTAB, some are not significant. For example, COUPS, CONSTCHG, and REVOLS are now completely insignificant when each of them is included with AGINI in our basic regression specification (not reported to save space); see also column (9). This result is broadly the same in the developing country sample.
5.5 Institutions Recently, a growing number of studies have paid a great deal of attention to the idea that budgetary procedures—^the rules according to which budgets are drafted by a government, amended and passed by parliament, and implemented by the government—have a direct influence on fiscal outcomes (see Alesina and Perotti 1996a for a literature survey).^^ In general, these studies find that a budgeting process that allows the prime minister or finance minister to have a dominant position over the spending ministers, and limits the amendment power of parliament is conducive to fiscal discipline. However, empirical studies have been largely limited to only a small sample of countries. We first examine some indicators of budgetary institutions, and then consider broad measures of institutional quality for a number of reasons. First, high-quality institutions can make a diflFerence for public finance: a more efficient tax-collection system and better monitoring on disbursement should strengthen the fiscal position of the government. Second, good institutions are found to be associated with higher growth, which may aflPect sociopolitical variables. Moreover, when institutions of conflict management are wellestablished and work well enough to suppress conflicts of interest among different groups, the social polarization effect we found earlier may be less important in determining the fiscal outcomes. Third, institutional quality indicators such as the efläciency of public sector, the degree of corruption, and the rule of law can be a good measure of the quality of budgetary institutions. This is because government institutions tend to be shaped by common factors such as economic, political, cultural, and historical circumstances and change only slowly (La Porta et al. 1999). As for the budgetary procedures, we make the most of the existing data, especially those from von Hagen (1992) and Lao-Araya (1998) for a combined total of 21 European and East Asian countries.^^ These two studies apply Among others, von Hagen (1992), von Hagen and Harden (1994), and Hallerberg and von Hagen (1999) study European Union countries, and Kontopoulos and Perotti (1999) examine OECD countries. Alesina, Hausmann, Hommes, and Stein (1999) focus on Latin America, while Lao-Araya (1998) deals with Asian countries. Poterba and von Hagen (1999) provide a collection of studies on fiscal institutions. I am grateful to Jürgen von Hagen for bringing Lao-Araya (1998) to my attention and for kindly providing me with a copy of the paper.
5.5 Institutions
97
almost the same methods and criteria, so that the problem of compatibility (if any) should not be a big issue.^^ After a close examination, we reconstruct indices of centralization of the budget process. An index of centralization for the preparation stage, MF, identifies who sets (or proposes) budget targets for negotiations within the government: 0, if the finance minister or cabinet collects bids from spending ministers; 1, if the finance minister or cabinet collects bids subject to pre-agreed guidelines; 2, if the cabinet decides on budget norms first; 3, if the finance minister proposes budget targets to be voted on by the cabinet; 4, if the finance or prime minister sets budget targets to be observed by spending ministers.^"^ Table 5.6 reports the regression results using MF. Despite the small sample size, it yields a striking result. The coefficient of MF is highly significant and of the expected sign (+). The adjusted R^ is quite high. The more centralized is the budgetary authority in the finance minister, the smaller are the deficits. Alesina et al. (1999) report a similar result in the Latin American country sample. In addition, we considered measures for the possibility to amend the executive's budget (AMEND), flexibility of budget execution (INFLEX), and the degree of transparency of budget documents (TRANS). Although TRANS and AMEND have the expected sign (+), they are not significant in our sample. However, it should be noted again that our indices are far from perfect. In fact, it is possible to create dummy variables out of these indices, which become significant in our basic regression.^^ Columns (2)-(5) in Table 5.6 display the regressions that include MF and variables that were found to be consistently associated with public deficits. It is interesting to note that ILLY and AGINI become weaker or insignificant once we include MF. This makes sense since we expect the negative effects of social polarization on deficits to be smaller when tighter budgetary procedures ^^ Alesina et al. (1999) construct indices on 20 Latin American countries in quite a different way from the two studies. Kontopoulos and Perotti (1999) provide their own indices of budgetary procedures for OECD countries, drawing on OECD publications. However, it is not as comprehensive as the aforementioned studies. ^^ Hallerberg and von Hagen (1999) argue that centralization of the budget process can be achieved either by delegating strong power to a finance minister or by engaging all of the members of the coalition in binding contracts on fiscal targets. This argument suggests that the coding for the index MF should change for countries such as Denmark and Ireland, where the budgeting process relies strongly on fiscal contracts. While the regression result remains almost the same regardless of whether such a change is made or not, one may be concerned about its compatibility with Lao-Araya (1998), who does not provide data for the incidence of fiscal contracts in Asian countries in question. Therefore, we use the code based on von Hagen (1992). ^^ A dummy variable for INFLEX (or TRANS), which takes 1 if INFLEX> 0 (or if TRANS> 0), and 0 otherwise, is significant at the 5% level in the basic regression. Higher values of INFLEX and TRANS represent greater inflexibility of budget execution and greater transparency, respectively. Both range from 0 to 4.
98
'S o
p0) T3 r^
cu
'o o ra
p
'•4^
T}* b -
*
Tf
CO (M
1—1
;0
^8 29
>^
»n
oo'
O JIJ 00 ^ ^ d
co" o co" #
CO
1-t
"--^ 9
<M 0 5
^-^ d Tt*
>^ d
csi 9 --^ o Ä c5
8 ^- S
CO tie
O
c^i ^ --^ o CO iO iC O ; 0 O CO d d c4
c6 isS e o
S^3^9S§" cv5i«de°:iL§:i
ci
°Oc^^c4^qOc5
O u:) •<* o o> CO C O (M CO O
O
^ S « 8 S =5 iS d ^ d 5, d
d^
co^
CM
d
^ o^ CO ^ CO ^ ; Q -: CO
d cT o
CD
o
<^ 5£ ^ ^
CO
?
•x00
o
" q CM
CM
•X-
CM
o
co"
5I
00^ 00
CM Tf
CO
•X-
ci
-^ Tf
CM CM
•X-
CO CO CM CM i n
CM
CM
« i6^ C3 C5 CM CO £i o d ^ o d •x^^ ^^ ^_^00 ^^ cI? CM CSJ
csi
CO
csi d
3,
^
t-H
r-i d
CO iO^ CO CM CJ CO CM
o
^ ^^ d ^--^o f^ ^
i5^ o
d
00
d d
00
d
3
C
lO (M t ^ kC 0? 00 <M CO T-H 00 CO CO d d^ d
* ^csi *o *CO CO CO
ci Ci d
8
CO
^
ö
oT I j
^
"T
fe S ?5 o^ d ^
Ö
«) "^ cJ ^
*
H:
(M
a.
0
^ tf <: J
o d
>J^ o
<M (M O
CO <M Tf
^ 2 ^ d
< 0
PQ
CO
H N
^ ^ 3 0
m CO
cu
^
^ CO ( M CO t ^ l CX) ^ d ^ o ^ g ^ o c o
6 S
^
. ^
(M CO
(M
Pi
a,
0
< J
t* ^
d 00
CO S
s^sS
^
•
O
o
CO ?2
°° d CO ?2
°° d
Of,
H
CO
g O
S «5
93
^i» CO
•^ «
S £ «
•2 « ^ 1
?•-!
CO ^ •3 Ö
^
• ^ l_-J CO
.3
^ ?;»
^:t
-§^
5 Economic, Political, and Institutional Determinants of Public Deficits
^ '^—^ CQ
O
<^-H
o
Ö .2 ^P^H^.^
P
-M 02
^ Ö 0?
-t^
;-l
Ö
a ^'E
(-1
J3 CO
.2
S?
§
o ^ ^
w Ö
o
1e
TJ Ö CO QH
CD c:> O PL,
1 ^
1
b ^o '6 bO
n3 0
?0 lO P2
a Q
T3 Ö
H P
3cd
5.5 Institutions
99
are in place.^^ This is consistent with the argument that a budgeting process that gives the prime or finance minister a dominant position over the spending ministers is conducive to fiscal discipHne. In fine with this argument, we do one more exercise using a proxy for the prime minister's position in the legislature. The indicator PM measures the degree to which a premier must depend on the support of a majority in the legislature in order to remain in ofläce. It ranges from 0 to 3, with higher values indicating greater independence. We expect higher values of PM to be associated with smaller deficits. This time the results do not suffer from the small sample problem. The coefficient of PM is significant at the 10% level in the basic regression (column 6 of Table 5.6). When either AGINI or PINSTAB or both are included with PM, the coefficient of PM is still significant at the 5-10% level and of the expected sign (+). Only when CABSIZE is added does PM become insignificant, but it still maintains the correct sign (as in column 7). Now we turn to a broad measure of the quality of government institutions, ICRGE, which has received a lot of attention in the growth literature. ICRGE is based on underlying numerical evaluations regarding the rule of law, bureaucratic quality, corruption, expropriation risk, and government repudiation of contracts (Knack and Keefer 1995). It ranges between 0 and 10, with high values representing better institutions. Columns (8)-(9) in Table 5.6 show the results. In column (8), the coefficient of ICRGE is significant at the 1% level and is of the expected sign (+). It becomes insignificant only when CABSIZE is also included (as in column 9).^*^ At the beginning, we proposed a working hypothesis that sociopolitical polarization is important, yet its effect on public deficits might be more pronounced or suppressed, depending on the political and institutional environment. We stress the importance of the interplay between social polarization and poor institutions in understanding public deficits by constructing composite indices of sociopolitical polarization that take institutional constraints into account, borrowing the idea from Rodrik (1999). Here we consider our main indicator of social polarization, AGINI; two institutional variables, PM Indeed, the coefficient of AGINI is significant at the 5% level in the regression excluding MF but using the same 29 observations as in column (1) of Table 5.6. In addition, we can think of a mesisure of democracy with the presumption that a democratic government can be a high-quality institution of conflict management. For example, the negative effects of social polarization on deficits may be mitigated if there are well-functioning 'checks and balances' mechanisms in the budgeting process. However, Alesina and Rodrik (1994) and Persson and Tabellini (1994) suggest that there may be a tendency of the majority to vote for bigger redistributive spending in democratic countries. Thus, the net effect is theoretically inconclusive. We tried DEMOC, a dummy variable taking the value of 1 if democracy and 0 if not, from the widely used Polity III data set (Jaggers and Gurr 1996). Democracy is weakly associated with larger deficits, but the coefficient of DEMOC is not significantly different from zero.
100
5 Economic, Political, and Institutional Determinants of Public Deficits
and ICRGE; and the size fragmentation indicator CABSIZE. Higher values of PM and ICRGE indicate greater centralized authority of the prime minister and greater efficiency of institutions, respectively. Polarl=AGINIx(3-PM); Polar2=AGINIx(10-ICRGE); and Polar3=AGINIxCABSIZEx(10-ICRGE). The idea behind Polar3 is that the effect on deficits of social polarization is stronger when there are more policymakers representing potentially conflicting interests of socio-economic groups in poor institutional settings. These indices are expected to be associated with larger public deficits, which is supported by regressions (10)-(15) of Table 5.6. They are all significant at the 1-5% level and are associated with greater deficits, even after controlling for all the important and significant variables. This is also largely true for the developing countries. Although the coefficient of Polar2 becomes insignificant (see column 17), the result is quite sensitive to the inclusion of some observations. For example, if we exclude Peru and Nigeria for the 1970s from the developing country sample, the coefficient of Polar2 becomes significant at the 5% level (p-value is 0.012). In fact, Peru and Nigeria are ranked first and second in terms of Polar2 in the entire sample (not only in the developing country sample), yet their average size of deficits was only modest (-2.5% of GDP in Peru and -1.58% of GDP in Nigeria), compared to the average public deficits of developing countries (-5.2% of GDP) in the 1970s. We confirm our main findings by using the seemingly unrelated regression (SUR) method. If errors are correlated across decades, the SUR estimator will be more efficient. However, Table 5.7 shows that the SUR results are remarkably similar to the least squares estimates.
5.6 Robustness and Sensitivity Analysis We check the robustness of our regression results in two ways. The first is to check if the partial correlation between the public deficit and an explanatory variable of interest is robust for various combinations of the conditioning variables, employing Leamer's (1985) extreme bounds analysis (EBA). This test involves estimation of regressions of the form CPSURP
= ßijl + ß,jZ -h ßa^jXj 4- e,
(5.2)
where I is a vector of variables always included in the regression, Z is the variable of interest, and Xj is a vector of three variables chosen from a pool X of potential explanatory variables of public deficits. We need to estimate this regression for all the possible M combinations of X^ € x- For each model j , we can find an estimate ßzj and the corresponding standard deviation azj-
5.6 Robustness and Sensitivity Analysis
101
The lower extreme bound is defined to be the lowest value of ßzj —2azj, and the upper extreme bound is defined to be the highest value of ßzj -{-2GZJ. If the lower and upper bounds are not of the same sign, the variable Z is not robust. In fact, this test is too strong. Levine and Renelt (1992) conclude that very few (or no) variables are robust after applying EBA to a pool of potential variables of growth. That is, it is possible to find a regression where inclusion of a certain subset from the pool renders the estimated coefficient of a variable of interest statistically insignificant. Given potential multicollinearity, simultaneity, and measurement errors among many of the variables, Levine and Renelt (1992)'s finding is hardly surprising. As Sala-i-Martin (1994) notes, this approach amounts to having a data under-mining problem. Following Sala-i-Martin (1997), we instead look at the whole distribution of the estimates of ßz by running the regressions above with all possible triplets of conditioning variables. We are interested in the fraction of the cumulative distribution of ßz lying to the right (or left) of zero (CDF), although we also report the EBA results and the fraction of the statistically significant estimates.^^ We assume that the distribution of the estimates of ßz is normal. The mean and the standard deviation of this distribution are computed by taking the averages of the estimated ßz^ and cr^s, respectively. The averages are computed using two different weighting schemes, equal weights and log likelihood weights. The log likelihood weighting scheme gives more weight to the regressions that are more likely to be the true model. We estimate models like equation (5.2) to conduct the sensitivity test on 17 variables that were found to be significant in at least one regression. Model j combines some fixed variables I (decade and regional dummies, LRGDP, GRGDP) that appear in all regressions, the variable of interest Z, with the trio Xj from the pool of the remaining 16 variables. For each variable Z, we need to run 560 regressions.^^ The sensitivity test results are reported in Table 5.8. First, we note that MF, AGINI, CABSIZE, and ILLY have 97 percent or more of their cumulative distribution of ßz lying to the left (or right) of zero (depending on the sign of their coefficients) under equal weight schemes (column 8). When log likelihood weights are used, almost 99 percent or more of the cumulative distribution of ßz is estimated to lie to one of side of zero (column 7). In particular, the coefficients of MF and AGINI are significant at the 5% level for more than 85 percent of the time and at the 10% level 95 percent of the time or more, as shown in columns (3) and (4). Interestingly, ASSASSIN, PM, and ICRGE also have a high proportion (95 percent or more) of the estimated cumulative distribution to the left or right of zero (see column 7). Columns (1) and (2) show the upper and lower extreme ^^ See Sala-i-Martin (1997) for more details. ^® The possible combinations of X^ are 16Cr3 =16!/(3!13!)=560, so we ran 9,520 (=17x560) regressions in total. To report the sensitivity test result for the fixed variables, LRGDP and GRGDP, we ran additional 2380 regressions that combine the fixed variables I with four variables out of the pool of 17 variables (17Cr4=2380 regressions).
102
Q
OH
O o a»
u
CO
0)
CD
T-H
2
d
ä
CO CO LO
LO * o (N
• ^o-o
IV rf
O -XO^ l ^
00 t L
2QO-.^: 3- 1
Ö
°0 '7*
c4 ^
CO CD
O ^.
CS
^
o q ^
rH
CO
o ^
3 S -f o s l ^ CO • *
^ csi b-- co" ?
Ö
O ^
^ ^
O) CO
"ctj
:>. ^
^ '^
5 Economic, Political, and Institutional Determinants of Public Deficits
O ^
It t^
IC U5 >0 * 3 »H ^ CO d
^ eo ^ T c^
^'T^
Sgo
t^ o S ^ !O> t- o ^ o CQ ö 9 Ö c4
^:
^- ^ o 2 c:^
t-H •?(•
t^ T-! ^ vJ^ O v i . P
S3 2 CO ^ o
^
o o
^% iS Ö ^ 9 ? 9
l5 ^ I S § I I f
3^8SgS
IC
q o 09 CO Ö 9 Ö
d ^ ci
IC
ö s o 9,0 2.
co^ 00 5q c4 Tf ^
q
^ ^
§5 S3dS§9§ä§!^ (M CO lO ^ 00 ^
g
5.7 Concluding Remarks
103
bounds. According to the stringent EBA criterion, none of the variables are 'robust.' However, in view of potential multicollinearity among explanatory variables and the interaction between sociopolitical variables and institutional variables, we believe that our results are strong. We check the robustness of our results also in terms of the observations, not the conditioning variables, by using a robust regression method. OLS estimates tend to be sensitive to outliers, either observations with unusually large errors or influential observations with unusual values of explanatory variables (often called leverage points). One of the most common ways to deal with outliers is to drop observations one at a time. But this is often inadequate because it may miss a group of outliers due to the masking effect. Similarly, single-case diagnostics such as Cook's distance measure, the studentized residual, and DFITS that are often used to detect outliers may fail to identify a group of outliers. Robust estimation is intended to obtain estimates that are not sensitive to outliers, and hence allows us to characterize the most coherent part of the data set. We use the least median of squares (LMS) estimator due to Rousseeuw (1984), which is given by min median ef,
(5.3)
ß
where e^ is the residual of the 2th observation with respect to the LMS fit. This LMS estimator, typically computed by approximate algorithms, can resist the effect of nearly 50% of contamination in the data. A disadvantage of the LMS method is its lack of efficiency because of its unusually slow convergence, making it unsuitable for inference. To deal with this problem, we use the LMS estimates to classify some of the observations as outliers, and then carry out a simple reweighted least squares (RWLS) procedure by assigning zero weight to outliers and full weight to the rest of the observations, as recommended by Rousseeuw and Leroy (1987). Table 5.9 displays the robust estimates of the coefficients of the explanatory variables that were found to be highly significant in the panel regressions. Once again, they are very close to the results presented in the previous sections, and in fact, the results are much stronger. They are all significant at the 1% level.^^
5.7 Concluding R e m a r k s We tested different theories of public deficits, using more than 40 economic, sociopolitical and institutional variables, and could identify which of the vari^^ Among the most frequently identified outlying countries are Italy, India, Belgium, Zaire (now Congo), Japan, Ghana, Malaysia, Greece, Philippines, Argentina, Spain, Mexico, Indonesia, Malawi in decreasing order of frequency. All of these countries (except India, Mexico and Philippines) were among the 15 countries with the largest (or smallest) deficits either for the 1970s or the 1980s or both. See Data appendices.
Q
O
X5 OH
CO
O Ü
'S
I
Ö
-^
00
CO
3
104
°° p
^ fe
t>05 Oi C5
<0 T-H 03 (35
iO OO - ^ !>Cl bCJ5 <Ji
Tj< CO
05 CO
(M (M 00
CO 00 CO 05
Tf O T—I LO
ro (N 05
(M 00 iO
t^ lO CO 05
rt^
lO CO CS 05
CO CO CO
o^ T—1
(M
o
lO LO lO
q o
00
Oi
^ l>00
C75 CO 00
oo
T—1
CO (M
o
-^ ,_( ^
Tf l^ (M 03 (M LO b- CO CO c^ CO l^ t^ t- t-- t-
^ Tt^
05 00
rt< (M l>-
CO
o
Tf«
LO
00
(M
NCO 05
t^ C5
CQ CO 00
05
^ CO CO (M
CO CO 00 l>CO I:CO b-
CO
t~ CO LO
T—5
CO
LO LO
^ - t^ LO lO o
CO
o CO
c^
CO
CO ^ o0LO0
t--
lO CO CO rH
c> ö ö 05 CQ CO 05
T—1 o cq T-H co CO o
^ LO 05
1—1
CS
1> <Ji 00
c^ ^ o
CS
^
00 05
CS
C75
o
CS 00
o
TH
CO
^
lO CO CO 05
r-i
CO 05 05
05 05
t-
ö ö
^ !>•
CO 1>05
LO LO
CD
N-
T-H
t>- rf LO CO CO
ö
o
LO CO CO
CO
LO CO
t-
00 00
^ 05 CS
LO CO
^^
,-i 03
Q Q
a=! Jö ; >^ o3
3CO
o
o
o
05 LO
1—1
^ ö CS
00 CO 00 00 CS
LO CO LO CS
J
CO
O H:I > < W O
0)
ö ö
05 05 C75
t-
CO LO 00
o rH o o q '"!
CO
05 CS
^
$ J C S ° - : C S ? o LT^ i s j « ^ „„ , ^ , _. CS <^ Ni-^ 05 oL^* I ^ ^__| O JÜ l JJ L/-S.-I o ' l :
^
00 00
r H L O g ; O 5 ° > ° I : ^ g ^ 0 0
LÖ
b-
,_( TH
b00 05 CO
CO CO
00 00 CO
C75
c:5 Ires T—5
^ ,_)^ CO o LO o CO LO
TH
00
cs CO
00 b-
b00
^
TH
ö
LO
T-H
05
o
CO CS
C75
T-H
NlO CO LO
CO
r-
o6
o (Ji
CO
co
N^
CO
^
o
'"'CS ^
CO
o
CS LO
CO
T—1
05
1—(
CS
ü
r^
ö
!>•ö ö ö o c5 ö Ö ö ö ö d> ö ö ö ö ö ö 05
Tt<
CO C35
*^ 00 ^ C35 ^ 05
00 I>CO 05
o 00
ö ö ö
03 lO lO > — I 05 05 CO lO CO 03 05 05 00 t^ 05 05 05 05 05
00 00 CO T—1
Ö
00
CO 00
CO CO CQ CO o (Zi ^ q ^ ^ ö ^ ö ö ö ö ö '-'^ ^ ^ ^ ö
oq ^! > •^ o
T—i
CO (M
Ö Ö Ö Ö d> Ö ö
q
lO -^ N- LO o lo r- o 00 O 00 CJ5 05 t^ CO (M
io o q
ö c5 o ö 00 <M CO O CO ^ oq CN o Tl^ oo O 05 -«^
2 2 ^ g'"!r- bS^^ C 0 0\^j 5 C S^j : 5 T - I C O C ? 5 0 5 r H C 0 0 5 - - l CO CO 0 0 f QC i c^ lT S/^T O Ö^-H T j f-^ ^ C O
/^/^ i-sr\ r^r\O C ^
C s i O O O C 0 o o O ' - ' Ö ' ~ ' ' ~ ' ' ~ ' ' ~ ' Ö ^ ^
C O VA-/ CJ5 /^NO C
T-l 1>
o o
ö
t^
,_i LO LO t-^ CS CO o CO CS ^ CS '-''-''—'CO CS '"'
l^
05
o o
o o o
t^ CS
LO 00 CS CO
•<*
0 3 C J 5 0 0 1 > C O C O L O C O C O T - ( C O T - I " " T - | ' ^ C O
rH b-
CO LO
cyi
CO
o
o CS t~ Oi 00
CO
ö ö ö od
LO CO ^ ^ C:J5
'—t
N-
Nl
J
ä
CO
_tß . 03 CO
s .s ^
^ CS!
(1)
o S3
LO
•r-t
^
+ ö
CO
•rH
> 2
3
^
Co
^
O
1 ö Co
bß (H . _ en O ^
••§ ^ .22 hO l^ O Ä
^
cö cu
ö
Ö Co
-M
5 Economic, Political, and Institutional Determinants of Public Deficits
g^
lö
hjO
o & ^ ^ ^ H:I
^ fe ^^Q
^
^
o
lO "o
cö
^ ü
O T—1
Ca
CC
5^ '3 fs m ^
ö ö o o cö xo
•g
"^ Ä
^^
z; ^ <:
5.7 Concluding Remarks
105
ables are important in explaining the different size of public deficits across countries. Socio-political instability, income inequality, a large size of the cabinet, and lack of central authority in the fiscal decision-making process are strongly negatively associated with the public surplus. While proportional parliamentary regimes tend to run larger deficits, government weakness or regime type does not seem to be consistently associated with deficits. Also the budgetary institutions and government institutions in general matter for the fiscal stance. In particular, countries with highly polarized societies or greater socio-political instability may achieve fiscal prudence by improving budgetary procedures and rules (for example, delegating strong power to the finance minister). Indeed, the negative effects of socio-political instability or social polarization on deficits can be mitigated by institutional arrangements such as stringent budgetary rules. While it is very important to create conditions conducive to fiscal prudence by promoting socio-political stability and/or equitable income distribution, socio-political instability or an unequal income distribution may be much harder to address than cabinet size or budgetary institutions in order to reduce budget deficits. Therefore, it is all the more important to build better budgetary institutions.
106
fo
O
^ o
CO
2 ^
CO b-
Ö
O
.^ 00 g N: CO
Ö "? ^
CO ^
CO^
CO
^ CD
Ö
q 2 g S .^ _:
It,
• CO
O
CO
CO CO
CO
CO
O c:)
|>
Ö
t>.
b-
T-H
-"
)5 00 i> ^ Ö
^
lO
t °° Ö
CO
<^ Ö
Gt)
S o g t^ ^ CS
o <^ P N PQ
hH CO
Ü
<:
^ It* 'i CO \<^ CO ' '
^
CO
lO
o ^w
I I ^.s ^ O Co * ^ 0) CO O
^ s
3 ^
ca
O
^I CO
CO CQ
p ^
s *S
0) bO CO Ö
D
-ü C
^
CO * M
*-+3 ^bO
?3
^2 <x> , 0
.£3
10
(M
CN
VI (-1
CD
*s .a 0 0
TJ
0
0
^ r:^ S
'§
.23
4^
^
CO
^ 'S t^ 0
>i
•^ > ^;''
9}"
11
P M
'^ 0 /-^^ ' ^
1 S
1
-se (V
S 3 \ 0 10 II
0
<73 Ö CO .^ W
+ J;
• . CO (M 00 ^
II '^
'^
>. Ö
11 «Ö
^ ^ a;
0
c^ 0
^
(-1
CQ
3CD
•+j"
g
•^
(-1
M
CD
df
0 CO
3.
^ i-:i >
*5b ,£3OJ
>
M
0
r^ •4J
M 0)
0
0 rO
CO
4 ^
CO
^ 73 §
T3
(-4
•<s>
4-3
.a
^0
M
CD
'+3
.2 ^ >
*0?
H-3
rilS
a;
CO
^CQ
PQ
CO
5 Economic, Political, and Institutional Determinants of Public Deficits
O
P O
Ö Ö
P
Ö
3
^
^ ho k^
O
gl^co
(^ O T3
;2; <: 12;
5.8 Data Appendix
5.8 Data Appendix
Table 5.10. Data Set of Consolidated Public Sector Deficits: 57 Developed and Developing Countries Code
Country
Code
Country
Argentina 1 ARC 30 JOR Jordan* Australia 31 KEN Kenya 2 AUS Austria 32 KOR Republic of Korea 3 AUT Bangladesh* 33 MWI Malawi 4 BDG Belgium 34 MYS 5 BEL Malaysia Bolivia 35MEX 6 BOL Mexico Brazil* 36MOR 7 BRA Morocco 37 NLD Bulgaria* 8 BGR Netherlands Burkina Faso* 38 NGA Nigeria 9 BFA Canada 39 NOR Norway 10 CAN Chile 40 PAK Pakistan 11 CHL Columbia 41 PRY 12 COL Paraguay 42 PER Cote d'lvoire Peru 13 CIV Denmark 43 PHL 14 DNK Philippines* 44 POL Dominica* Poland 15 DMA 16 DOM Dominican Republi(z45 SLE Sierra Leone Ecuador 46 ESP Spain 17 ECU Finland 47 LKA Sri Lanka 18 FIN Prance 48 SWE Sweden 19 FRA Germany 49 THA Thailand 20 DEU Ghana 50 T T O Trinidad & Tobago* 21 GHA Greece 51 TUR Turkey 22 CRC 52 GBR United Kingdom Honduras 23 HND India 53 USA 24 IND United States Indonesia 54 VEN 25 IDN Venezuela Ireland 55 ZAR Zaire 26 IRL Italy 56 ZMB Zambia 27 ITA Jamaica 57 ZWE 28 JAM Zimbabwe Japan 29 JPN * Indicates that data are not available for the period of 1970-79.
107
108
5 Economic, Political, and Institutional Determinants of Public Deficits Table 5.11: Data Description and Source
Description and Source Dummy variables for Sub-Saharan African countries (according to the World Bank definition). Decade average of Gini coefficients obtained using all the AGINI data points available. Source: Deininger and Squire (1996). Decade average of Gini coefficients obtained using only high AGINIHI quality data points. Source: Deininger and Squire (1996). AMEND indicates whether amendments of the executive's budget AMEND in the adoption stage are limited or not: 0 if unlimited; 4 if limited. Source: von Hagen (1992) and Lao-Araya (1998). ASSASSIN Assassinations: number of assassinations, defined as any politically motivated murder or attempted murder of a high government official or politician, per 10 million population, decade average. Source: Banks (1997). Major cabinet changes: The number of times in a year that a new CABCHG premier is named and/or 50% of the cabinet posts are occupied by new ministers. Source: Banks (1997). Size of cabinet: The number of ministers of 'cabinet rank', excluding CABSIZE undersecretaries, parliamentary secretaries, ministerial alternates, etc. Includes president and vice-president under a presidential system, but not under a parliamentary system. Chief of state excluded, except under presidential system. Source: Banks (1997). Party coalition, decade average: 0 = no coalition and no opposition; COAL 1 = more than one party, coalition government and no opposition; 2 = more than one party, coalition government and opposition; 3 = minority government. Source: Banks (1997). CONSTCHG Major constitutional changes: The number of basic alternations in a state's constitutional structure, the extreme case being the adoption of a new constitution that significantly alters the prerogatives of the various branches of government Source: Banks (1997). Number of coups d'etat: The number of extraconstitutional or COUPS forced changes in the top government elite and/or its effective control of the nations power structure in a given year. Unsuccessful coups are not counted. Source: Banks (1997). Decade average of consolidated public sector surplus. CPSURP Source: Easterly, et al. (1994). Dummy variable for democratic regime: 1 if democracy and 0 DEMOC otherwise. Source: Jaggers and Gurr (1996). Dummy variable for East Asian countries (according to the EASIA World Bank definition). Continued on next page Variables AFRICA
5.8 Data Appendix
109
Table 5.11 — continued from previous page Variables Description and Source ETHNIC Index of ethno-linguistic fractionalization, 1960, which measures the probability that two randomly selected people from a given country will not belong to the same ethnolinguistic groups. Source: Taylor and Hudson (1972). EXECHG Changes in effective executive: The number of times in a year that effective control of the executive power changes hands. Such a change requires that the new executive be independent of his predecessor. Source: Banks (1997). EXT External shock: Product of growth rate of terms of trade and trade openness, where trade openness = (export + import)/GDP. Source: terms of trade from Easterly and Yu (1999), and trade Openness from World Bank (1997). EXTSH External shock: Product of the standard deviation of the first log-differences of terms of trade and the trade openness, where trade openness = (Export -f Import)/GDP. Source: Terms of trade from Easterly and Yu (1999), and trade openness from World Bank (1997). GINIHI Gini coefficients (high quality data) measured as close to the end of the previous decade (1969 and 1979) or the start of each decade (1970 and 1980) as possible. Source: Deininger and Squire (1996). GOVTCRIS Major government crises: Any rapidly developing situation that threatens to bring the downfall of the present regime— excluding situations of revolt aimed at such overthrow. Source: Banks (1997). GRGDP Growth rate of real GDP, decade average Source: World Bank (1997). ICRGE Quality of institutions: Average of (a) government repudiation of contracts, (b) risk of expropriation, (c) rule of law (first multiplied by 5/3), (d) bureaucratic quality (first multiplied by 5/3) over the 1980s. Source: Easterly and Levine (1997). Original source: Knack and Keefer (1995), which originates in International Country Risk Guide (ICRGE). ILLY Financial depth: The ratio of liquid liabilities of the financial system to GDP at the start of each decade. Liquid liabilities consist of currency held outside the banking system -h demand- and interest-bearing liabilities of banks and non-bank financial intermediaries. Source: World Bank (1997). INFLAT Inflation: Decade average of the consumer price index. Source: World Bank (1997). INFLEX Flexibility of budget execution: Range 0-4. Higher value Continued on next page
110
5 Economic, Political, and Institutional Determinants of Public Deficits
Table 5.11 — continued from previous page Variables Description and Source corresponds to higher degree of centralization and rigidity in budget execution. Source: von Hagen (1992) and Lao-Araya (1998). LATINCA Dummy variable for Latin America and the Caribbean (according to the World Bank definition). LRGDP Log of initial income: Log (real per capita GDP) measured at the start of each decade (1970,1980). Source: Summers and Heston (1988). MA J Dummy taking 1 if majoritarian electoral rule: If MA J = 0, it indicates proportional system. Source: Persson and Tabellini (1999b). Original source: Cox (1997). MAJPRES Product of MAJ and PRESPT: Dummy taking 1 if majoritarian electoral rule and presidential system. MF Budget agenda setter: 0 if minister of finance or cabinet collects bids from spending ministers; 1 if minister of finance or cabinets collect bids subject to pre-agreed guidelines; 2 if cabinet decides on budget norms first; 3 if minister of finance proposes budget targets to be voted on by cabinets; 4 if minister of finance or prime minister determines budget targets to be observed by spending ministers. Source: von Hagen (1992) and Lao-Araya (1998). MINOR Minority government, decade average: 1 if minority government; 0 otherwise. Source: Banks (1997). PARFRACT Party fractionalization index = 1 — X)(^i)^? where U = proportion of members associated with the ith party in the lower house of the legislature. It ranges between 0 and 1. A value of 0.5 is associated with a perfectly balanced two-party system and a value greater than 0.5 is associated with more than two parties. Source: Banks (1997). PM Degree of parliamentary responsibility: Refers to the degree to which a premier must depend on the support of a majority in the lower house of a legislature in order to remain in office. 3 = irrelevant. Office of premier does not exist; 2 = absent. Office exists, but there is no parliamentary responsibility; 1 = incomplete. The premier is, at least to some extent, constitutionally responsible to the legislature; 0 = complete. The premier is constitutionally and effectively dependent on a legislature majority for continuance in office. Source: Banks (1997). POP65 Average share of population aged over 65 in each decade. Source: World Bank (1997). PRES Dummy for presidential system: 1 if presidential and 0 if parliamentary. Continued on next page
5.8 Data Appendix Table 5.11 - continued from previous page Variables Description and Source Source: Mainwaring (1993). Augmented by Cox (1997) and Shugart and Carey (1992). PRESPT Dummy for presidential system: 1 if presidential and 0 if parliamentary. Source: Persson and Tabellini (1999b). Original source: Cox (1997) and Shugart and Carey (1992). REVOLS Revolutions: Any illegal or forced change in the top government elite, any attempt at such a change, or any successful or unsuccessful armed rebellion whose aim is independence from the central government. Source: Banks (1997). SEAT Size of legislature divided by number of the seats held by the largest party: A country with no party (no legislature) has a score of 0; a one-party system has a score of 1; and a system with 40 out of 100 seats held by the majority party has a score of 2.5. Thus, a higher score means a smaller number of seats held by the largest party in legislature. Source: Banks (1997). TRANS Transparency of budget documents: TRANS is between 0 and 4, where higher value of TRANS represents greater transparency. Source: von Hagen (1992) and Lao-Araya (1998) WAR Dummy variable for war on national territory during the decades. Source: Easterly and Levine (1997).
111
112
5 Economic, Political, and Institutional Determinants of Public Deficits
Table 5.12. Comparison of Selected Indicators Between the 15 Countries with Largest Public Deficits and 15 Smallest Deficits: 1970-79 Country Largest deficits 1970-79 Zambia Jamaica Zaire Italy Sierra Leone Bolivia Morocco Honduras Malawi Argentina Sri Lanka Pakistan Malaysia Ireland Ghana Average Smallest deficits 1970-79 Finland Norway Sweden Denmark Venezuela Australia France Austria Spain Canada USA Dominica Rep. Paraguay Korea Nigeria Average
CPSURP ACINI ETHNIC ASSASSIN ICRGE CABSIZE PARFRAC PM
-12.4 -9.3 -8.9 -8.6 -8.4 -8.4 -8.1 -7.5 -7.4 -7.3 -7.1 -7.1 -7.0 -6.6 -5.4 -8.0
59.5 48.9 N/A 35.8 49.0 N/A 54.0 N/A 51.8 37.6 36.4 31.3 49.2 37.7 N/A 44.7
0.8 0.1 0.9 0.0 0.8 0.7 0.5 0.2 0.6 0.3 0.5 0.6 0.7 0.0 0.7 0.5
0.0 1.7 0.0 0.5 0.3 0.6 0.1 0.0 0.0 1.9 0.0 0.1 0.1 0.6 0.0 0.4
4.1 4.7 3.0 8.2 5.4 2.3 4.3 3.4 4.5 4.3 4.3 4.1 6.9 8.3 3.7 4.8
19.7 19.5 22.4 25.7 20.4 16.0 20.2 11.5 15.0 13.0 25.3 16.6 24.1 16.0 20.9 19.1
0.1 0.4 0.0 0.7 0.1 0.1 0.2 0.1 0.0 0.2 0.5 0.2 0.2 0.6 0.1 0.2
2.3 0.0 2.7 0.0 1.4 3.0 2.0 3.0 3 3.0 0.2 2.6 1.0 0.0 2.4 1.8
3.6 3.1 2.6 1.5 0.2 0.0 -0.3 -0.7 -0.8 -0.8 -1.0 -1.2 -1.5 -1.5 -1.6 0.1
28.1 31.1 37.0 32.0 47.4 35.9 37.2 30.3 36.2 31.5 35.3 45.0 N/A 36.3 51.8 34.3
0.2 0.0 0.1 0.1 0.1 0.3 0.3 0.1 0.4 0.8 0.5 0.0 0.1 0.0 0.9 0.3
0.0 0.0 0.0 0.0 0.1 0.1 0.0 0.0 0.8 0.1 0.0 0.2 0.0 0.1 0.0 0.1
9.7 9.6 9.6 9.7 5.6 9.4 9.3 9.4 7.6 9.7 9.8 4.5 4.4 6.4 3.1 7.9
16.3 15.6 18.8 18.0 14.9 25.3 25.2 15.5 19.4 29.8 13.5 16.0 11.2 20.1 21.4 18.7
0.8 0.7 0.7 0.8 0.7 0.6 0.7 0.5 0.2 0.6 0.5 0.4 0.5 0.4 0.1 0.5
0.0 0.0 0.0 0.0 3.0 0.0 0.0 0.0 1.3 0.0 3.0 3.0 3.0 1.0 3.0 1.2
5.8 D a t a A p p e n d i x
113
Table 5.13. Comparison of Selected Indicators Between the 15 Countries with Largest Public Deficits and 15 Smallest Deficits: 1980-90 Country Largest Deficits 1980-90 Zambia Zimbabwez Greece Jamaica Sri Lanka Italy Bangladesh Boliva Malaysia Jordan Ireland Belgium Cote d'lvoire Morocco Honduras Average Smallest Deficits 1980-90 Norway Finland Chile Dominica Venezuela Bulgaria Japan Sweden Australia Indonesia Korea Ghana UK Germany France Average
CPSURP ACINI ETHNIC ASSASSIN ICRGE CABSIZE PARFRAC PM
-14.2 -12.1 -11.7 -11.5 -11.1 -11.0 -10.4 -10.2 -9.8 -9.6 -9.2 -8.9 -8.9 -8.6 -8.6 -10.3
N/A 56.8 35.7 50.7 43.9 31.9 35.7 52.0 48.2 34.6 35.8 24.6 42.4 40.7 55.0 42.0
0.8 0.5 0.1 0.1 0.5 0.0 N/A 0.7 0.7 0.1 0.0 0.6 0.9 0.5 0.2 0.4
0.0 1.1 0.2 0.8 0.2 0.3 0.0 0.3 0.0 0.0 0.0 0.0 0.0 0.0 0.4 0.2
4.1 4.4 5.5 4.7 4.3 8.2 2.7 2.3 6.9 4.1 8.3 9.7 6.7 4.3 3.4 5.3
23.7 25.3 28.2 19.5 38.1 29.8 29.8 19.7 26.5 23.5 17.4 21.7 32.7 30 13.1 25.3
0.0 0.4 0.6 0.2 0.3 0.7 0.2 0.6 0.3 0.0 0.6 0.9 0.0 0.8 0.4 0.4
2.0 1.1 0.0 0.0 1.0 0.0 2.4 3.0 1.0 2.0 0.0 0.0 2.9 2.0 3.0 1.4
4.8 0.5 -0.3 -0.4 -0.6 -0.7 -1.0 -1.1 -1.4 -1.5 -1.6 -1.6 -1.7 -1.9 -2.1 -0.6
27.0 25.5 53.8 N/A 43.0 23.2 35.2 30.3 34.3 32.8 34.9 36.3 29.6 30.5 32.1 33.5
0.0 0.2 0.1 N/A 0.1 N/A 0.0 0.1 0.3 0.8 0.0 0.7 0.3 0.0 0.3 0.2
0.0 0.0 0.6 N/A 0.0 N/A 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 0.1 0.1
9.6 9.7 6.3 N/A 5.6 7.1 9.4 9.6 9.4 3.7 6.4 3.7 9.3 9.6 9.3 7.8
18.0 17.7 17.9 8.8 19.2 27.0 13.5 19.7 29.3 25.4 22.6 22.8 22.6 19.0 27.7 20.7
0.7 0.8 0.1 0.4 0.6 0.1 0.6 0.7 0.5 0.7 0.6 0.1 0.5 0.6 0.8 0.5
0.0 0.0 3.0 0.0 3.0 1.9 0.0 0.0 0.0 3.0 1.0 3.0 0.0 0.0 0.0 1.0
114
5 Economic, Political, a n d I n s t i t u t i o n a l D e t e r m i n a n t s of P u b l i c Deficits Table 5.14. Summary Statistics for the 1970s
Variable E c o n o m i c Variables Consolidated Public Deficits (CPSURP) Consolidated Public Deficits (CPSURP) for Developing Countries only Log of Initial Income (LRGDP) Growth Rate of Real GDP (GRGDP) Inflation Rate (INFLAT) Population of over 65 years old (POP65) Financial Depth (ILLY) External Shocks (EXT) War (WAR) Socio-Political Variables Assassinations (ASSASSIN) Cabinet Change (CABCHG) Coups D'etat (COUPS) Effective Executive Changes (EXECHG) Government Crisis (GOVTCRIS) Political Instability Index (PINSTAB) Revolutions (REVOLS) Coalition Governments (COAL) Minority Governments (MINOR) Party Fractionalization (PARFRACT) Size of Cabinet (CABSIZE) Size of Legislature/Seats of Largest Party (SEAT) Dummy for Presidential System (PRES) Income Inequality (GINIHI) Income Inequality (AGINIHI) Income Inequality (AGINI) Ethno-hnguistic Fractionalization (ETHNIC) Social Polarization Index (SOCPOL) Institutional Variables Amendments of Executive's Budget Proposal (AMEND) Budget Agenda Setter (MF) Dummy for Democracy (DEMOC) Institutional Quality (ICRGE) Prime Minister's Power (PM) Flexibility of Budget Execution (INFLEX) Transparency of Budget Documents (TRANS) Composite Index of Polarization-Institution (POLARl) Composite Index of Polarization-Institution (P0LAR2) Composite Index of Polarization-Institution (P0LAR3)
Obs Mean Std. Dev. Min
49 31
-3.68 -5.11
Max
3.55 2.99
-12.39 -12.39
48 7.95 47 4.67 48 19.40 49 6.46 44 36.57 45 0.15 49 0.18
0.92 2.25 32.69 4.24 20.74 1.74 0.39
6.16 9.47 0.61 10.49 4.14 190.26 2.23 14.48 102 8.57 -5.19 6.36
48 46 46 49 45 44 45 49 49 49 49 49 31 34 34 43 48 42
0.41 0.69 0.21 0.23 1.67 1.02 0.59 1.08 0.42 0.26 4.97 0.83 0.49 10.42 8.46 10.14 0.30 0.56
0.20 0.52 0.04 0.26 1.38 0.13 0.20 1.74 0.38 0.39 19.20 1.54 0.35 39.55 38.02 41.47 0.39 0.04
3.56 0.20
0
1
0 0 0 0 0
1.90
-0.40
4.73
0 0 0 0
3 3 1
3 1 1 7
11.20 0.24
0.82 36.10 3.74
0
1
24.90 61.88 24.30 55.00 24.70 65.38 0 0.90 -0.74 1.29
17 2.82 1.88 4 0 16 2.38 1.02 1 4 49 0.49 0.51 1 0 49 6.34 2.40 2.27 9.81 49 1.48 1.32 0 3 1.42 17 2.29 0 4 17 2.85 1.07 1 4 43 60.86 0 146.55 47.68 43 148.86 115.89 5.60 365.89 43 2846.13 2225.50 91.29 7658.13
115
5.8 D a t a A p p e n d i x Table 5.15. Summary Statistics for the 1980s Variable E c o n o m i c Variables Consolidated Public Deficits (CPSURP) Consolidated Public Deficits (CPSURP) for Developing Countries only Log of Initial Income (LRGDP) Growth Rate of Real GDP (GRGDP) Infiation (INFLAT) Population of over 65 years old (POP65) Financial Depth (ILLY) External Shocks (EXT) War (WAR) Socio-Political Variables Assassinations (ASSASSIN) Cabinet Change (CABCHG) Coups D'etat (COUPS) Effective Executive Changes (EXECHG) Government Crisis (GOVTCRIS) PoHtical Instability Index (PINSTAB) Revolutions (REVOLS) Coalition Governments (COAL) Minority Governments (MINOR) Party Fractionahzation (PARFRACT) Size of Cabinet (CABSIZE) Size of Legislature/Seats of Largest Party (SEAT) Dummy for Majoritarian Electoral Law (MAJ) Dummy for Presidential System (PRES) Dummy for Presidential System (PRESPT) Income Inequality (GINIHI) Income Inequality (AGINIHI) Income Inequality (AGINI) Ethno-linguistic Fractionahzation (ETHNIC) Social Polarization Index (SOCPOL) Institutional Variables Amendments of Executive's Budget Proposal (AMEND) Budget Agenda Setter (MF) Dummy for Democracy (DEMOC) Institutional Quality (ICRGE) Prime Minister's Power (PM) Flexibihty of Budget Execution (INFLEX) Transparency of Budget Documents (TRANS) Composite Index of Polarization-Institution (POLARl) Composite Index of Polarization-Institution (P0LAR2) Composite Index of Polarization-Institution (P0LAR3)
Obs Mean Std. Dev. Min
57 -5.33 39 -6.08
3.95 3.63
54 8.11 56 2.88 56 82.01 56 6.81 54 42.80 56 -0.02 57 0.26
1.02 2.06 246.02 4.66 24.11 1.51 0.44
54 55 55 57 54 51 54 57 57 57 57 57 39 42 38 40 44 51 53 48
0.12 0.58 0.04 0.21 0.28 -0.18 0.07 1.88 0.39 0.44 22.29 1.74 0.38 0.43 0.42 38.05 37.56 38.77 0.40 -0.06
0.23 0.69 0.19 0.16 0.68 0.57 0.26 1.06 0.42 0.27 6.61 0.94 0.49 0.50 0.50 8.83 8.10 9.50 0.30 0.49
Max
-14.16 4.81 -14.16 -0.31 9.64 5.86 -1.12 8.01 2.58 1258.13 2.41 17.08 8.49 140 -4.61 5.01
0
1
0 0 0 0 0
1.10
2 1 0.73
3
-0.40
2.44
0 0 0 0
1 3 1
8.82
0 0 0 0
0.86 45.09 4.85
1 1 1
24.40 59.44 23.20 55.55 23.20 58.30 0 0.90 -0.84 1.07
19 2.74 1.91 4 0 1 4 17 2.41 1.00 56 0.57 0.50 1 0 56 6.16 2.36 2.27 9.81 1.31 57 1.48 0 3 1.41 19 2.44 4 0 19 2.87 1.07 1 4 46.32 51 53.55 0 152.19 51 154.41 110.88 5.47 401.99 51 3383.94 2500.73 92.48 9482.33 '
116
5 Economic, Political, and Institutional Determinants of Public Deficits Table 5.16. Correlation Matrix, Selected Series CPSURP LRGDP GRGDP INFLAT ILLY COAL PINSTAB
CPSURP LRGDP GRGDP INFLAT ILLY COAL PINSTAB CABSIZE ACINI ELF60 ICRGE PM PRES MF*
i 0.39 0.17 -0.26 0.13 0.05 -0.25 -0.20 -0.30 -0.26 0.55 -0.11 -0.02 0.37
1 -0.30 -0.15 0.56 0.34 -0.15 0.08 -0.41 -0.41 0.81 -0.40 -0.14 -0.35
CABSIZE AGINI CABSIZE ACINI ELF60 ICRCE PM PRES MF*
1 -0.39 0.03 -0.19 0.05 -0.04 0.14 0.05 0.05 0.01 -0.09 0.25
1 -0.28 -0.05 0.29 -0.14 0.27 0.15 -0.39 0.32 0.35 -0.10
ELF60
ICRGE
1 -0.29 0.17 0.00 0.39
1 -0.61 -0.42 -0.23
1 0.32 -0.25 0.13 -0.51 -0.21 0.64 -0.52 -0.40 -0.32
1 -0.26 0.12 -0.21 -0.14 0.28 -0.28 -0.05 0.07
PM PRES
1 -0.10 0.22 0.01 -0.29 0.18 0.18 -0.37 MF
i -0.27 0.35 0.14 -0.43 -0.43 0.16
1 -0.04 -0.61 0.62 0.57 0.17
1 0.80 0.30
1 0.50
1
* With regard to correlation with MF, we report the pairwise correlation because of its small number of observations.
6
Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
6.1 Introduction In the previous chapters, we have presented theoretical models and empirical evidence regarding the linkage between social polarization and fiscal problems such as fiscal deficits and volatility (of fiscal balance) in different contexts and samples of countries. In particular, we have presented a dynamic model of fiscal policy in a simple growth framework where social polarization (of preferences) plays a central role in generating poor growth through a fiscal instability channel in Chap. 2. In a highly polarized society, a deficit occurs endogenously, fiscal spending path becomes more volatile, output collapses, and economic growth rate is reduced along the transition path to a new steady state with a lower level of output. That is, we have advanced the fiscal instability channel that negatively links social polarization and growth, which is a distinct explanation for the common empirical finding that social polarization, as measured by income inequality or ethnic divisions, is harmful to growth (see Easterly 2002; Rodrik 1999; Perotti 1996; Benabou 1996 among others). We have shown strong econometric evidence that income inequality is important in explaining cross-country differences in public sector deficits in the period of 1970-1990 in Chap. 5. Also, many studies find that fiscal deficits are negatively associated with growth (Fischer 1993; Easterly and Levine 1997; Adam and Bevan 2005 among others). Combining these two, one can suspect that fiscal instability is a channel that negatively links income inequality and growth. Yet we have not fully investigated this fiscal instability channel in the context of growth regression. The objective of this chapter is to provide empirical evidence in support of the fiscal instability channel that relates income inequality to poor growth through a cross-country analysis for the period of 1970-2000 in a sample of 93 countries, exploiting both cross-sectional and time-series dimensions of the data. There are numerous empirical papers on the relationship between income distribution and growth. However, to our best knowledge, this is the first study to present such an evidence. As a matter of fact, most studies
118
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
have focused on the reduced-form growth regression that includes an income distributional measure as an explanatory variable, rather than investigating specific channels through which income distribution affects growth as we do in this chapter. The notable exception is Perotti (1996) in which he examines four specific channels in the period of 1960-1985, yet he does not consider our fiscal instability channel either. The four channels are (1) redistributive fiscal spending and taxes, (2) sociopolitical instability, (3) borrowing constraints and education, (4) education and fertility decision. He concludes that there is strong empirical support for sociopolitical instability and education/fertility decision channels. Previously, we have used the term fiscal instability to mean both large fiscal deficits and volatility of fiscal outcomes. Since we already provided comprehensive econometric analysis on the relationship between income inequality and fiscal deficits in Chap. 5, we focus on fiscal volatility in this chapter. Accordingly, we use fiscal instability and fiscal volatility interchangeably from now on. In our empirical analysis, we pay careful attention to robustness and consistency of our results with respect to various issues in running a regression, such as outlier and endogeneity problems. As the empirical growth literature has explosively grown, some shortcomings of growth regressions have become apparent (see Temple 1999; Durlauf et al. 2004 for a critical survey). A dominant concern has been the robustness. Many growth studies have regressed (various indicators of) output growth on a vast array of potential determinants. But the usefulness of this approach has increasingly been called into question, largely because the resulting parameter estimates are often sensitive to other conditional variables (Levine and Renelt 1992; Sala-i-Martin 1997). This problem is often compounded by heterogeneity of sampled population of countries and time periods under study. More recent studies such as Bosworth and Collins (2003) suggest that we focus a core set of explanatory variables that have been shown to be consistently associated with growth and evaluate the importance of other variables conditional on inclusion of the core set. Instead of pursuing an extensive sensitivity analysis that involves running a large number of regressions with various combinations of potential determinants as in Chap. 5, we rather focus on a "core set" of growth determinants while investigating our fiscal instability channel. On the other hand, the ordinary least squares (OLS) estimates tend to be sensitive to outliers, either observations with unusually large errors or influential observations with unusual values of explanatory variables (often called leverage points). For example. Easterly (2004) emphasizes that the large policy effects uncovered in growth regressions are typically driven by outliers which often represent instances of extremely "bad" policies. One of the most common ways to deal with outliers is to drop observations one at a time. But this is often inadequate because it may miss a group of outliers due to the masking effect. Similarly, single-case diagnostics that are often used to detect outliers such as Cook's distance measure may fail to identify
6.2 Inequality and Growth: Cross-Country Regression
119
a group of outliers. We address the outlier problem by employing a robust estimation procedure that is based on the least median of squares (LMS), due to Rousseeuw (1984). Since the robust estimation obtains estimates that are not sensitive to outliers, it allows us to characterize the most coherent part of the data set. Another concern with growth regressions is that many explanatory variables are most likely to be endogenous. This is a reason that we focus on a small core set of growth determinants which mostly includes initial conditions that can be viewed as predetermined for each country. For other variables such as trade openness and institutional quality, we employ instrumental variables (IV) method to address the endogeneity problem. Recent work has identified a few instrumental variables for institutional quality and trade openness (Hall and Jones 1999; Frankel and Romer 1999; Acemoglu et al. 2001; Glaeser et al. 2004). While there are no effective instruments for macroeconomic policy variables in general (see Rodrik 2005), we follow our theory to use initial inequality to instrument fiscal instability indicators that is our key variable in the growth regression. As a part of our empirical investigation, we also show that initial income inequality is consistently associated with fiscal instability measured over the subsequent years. The plan of the chapter is as follows. Sect. 6.2 presents reduced-form growth regression on income inequality. Sects. 6.3 and 6.4 separately examine the two pieces of fiscal instability channel: link between income inequality and fiscal volatility, on the one hand, and link between fiscal volatility and growth, on the other hand,. Next, we further check the robustness of our results. We conclude in Sect. 6.6. We provide detailed information on data including the country list, sources and summary statistics in the data appendices.
6.2 Inequality and Growth: Cross-Country Regression We begin by quantifying the empirical relationship between long-term economic growth and social polarization as measured by income inequality in a reduced-form cross-country regression.-^ To achieve this, we add an inequality measure to a standard growth regression equation in which the economic ^ Although most studies find a negative correlation between growth rate and a number of measures of income inequality in a cross-country regressions, some studies using panel data report a zero, non-linear, or even positive relationship between income inequality and growth. Related studies are Forbes (2000) and Barro (2000). However, it is not clear whether panel methods using relatively high frequency data are the appropriate test of a relationship whose mechanisms seem to be long run characteristics that are fairly stable over time. Easterly (2002) discusses why there are conflicting results in the empirical literature on inequality and growth, and finds high inequality to have a large and statistically significant negative effects on human capital accumulation and institutional quality and hence on economic growth through these channels.
120
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
growth is explained by initial income per capita, country size, h u m a n capital, and other variables as t r a d e openness and institutional quality, which make u p our core set of growth determinants. This regression is intended t o provide some benchmark before we t u r n t o our fiscal instability channel in t h e following sections. In t h e growth regression, we consider two types of dependent variables: one is t h e average growth r a t e of real G D P per capita in 1970-2000, which is obtained from Penn World Table ( P W T ) version 6.1 of Heston et al. (2002), and t h e other is t h e average growth r a t e of total factor productivity ( T F P ) for t h e same period, which is derived from Woo (2004). Our basic regression specification is as follows:
Growth^ = ao-\ra\n(initial
income per capita)^ + / ? l n ( h u m a n capital)^
+ 7 In (population) i + Jincome inequality^ + ^ X j + ci,
(6.1)
where i denotes t h e country, and e is an unobserved error term. To capture t h e catching-up process, we include t h e log of real G D P per capita in 1970 (LRGDPCH70). Also, we include h u m a n capital and country size as basic explanatory variables in t h e benchmark specification. Countries with an abundance of h u m a n capital and large country size (also, capturing potentially large market extents and aggregate scale effects) more likely have a greater ability t o a t t r a c t investors, t o absorb ideas from t h e rest of t h e world, and t o engage in innovation activities (Grossman and Helpman, 1991).^ We use t h e log of average years of secondary schooling in t h e population over age 15 in 1970 (LSYR1570) from Barro and Lee (2000) as a proxy for h u m a n capital, and t h e log of population in 1970 (LPOP70) from P W T 6.1 as a proxy for country size.^ Inclusion of country size has multi-purposes. It is originally intended t o capture t h e potential market extents or aggregate scale effects ^ Endogenous growth theory typically predicts that population size (scale effect) or population growth is positively related to technological progress and hence economic growth. See Jones (1999) for a comparison of variants of endogenous growth theory. Intuitively speaking, as for the scientific discovery, a country with large population is more likely to have an Einstein than a country with a small one. ^ Recently, Cohen and Soto (2001) constructed a new data set on human capital for 95 countries at the beginning of each decade of the period 1960-2000. They try to improve upon Barro-Lee data (2000) by addressing some inconsistency in Barro-Lee data, and notably use age-specific data in the available census to construct estimates of educational attainment for each age-cohort in other years for which direct observations are missing. However, the regression result using Cohen-Soto data is similar to that using Barro-Lee data. Moreover, initial human capital variable (for example, average years of schooling of the total population aged 15 or over in 1970) is available only for 78 countries in our sample. Therefore, we report regression results, using Barro-Lee data in order to maintain the largest number of observations possible.
6.2 Inequality and Growth: Cross-Country Regression
121
that provide an incentive for innovation or adoption of better technologies. But recent studies have shown there are some interactions among country size, trade openness, and government size (see Ades and Glaeser, 1999; Alesina and Wacziarg, 1998 among others). For instance, Alesina and Wacziarg (1998) find the country size determines the degree of trade openness and the size of government. So it is important to control for the country size. Our main indicator of social polarization is the income inequality measure. We use indicators AGINI60 and AGINI70 that are averages of all available data of Gini coefficients in the 1960s and 1970s, respectively, which are obtained from Deininger and Squire (1996). Finally, Xi represents other variables such as trade openness and quality of institutions, which we discuss later. Because heteroskedasticity may be more important in a cross-country sample, the reported standard errors of the coefficients are based on White's (1980) heteroskedasticity-consistent covariance matrix, which reduces the sensitivity of inference and hypothesis test using OLS estimator to general form of heteroskedasticity. Columns (l)-(5) in Table 6.1 present OLS regression results based on the above specification. As in the growth literature, we find evidence of conditional convergence that other things being equal, poorer countries tend to grow faster. The coefficients of LRGDPCH70 are all significant at the 1% level and have the expected sign (-). The initial level of human capital (LSYR1570) enters the regression with various levels of significance, indicating that countries with rich endowment in human capital tend to grow faster, other variables being controlled for. But its significance is somewhat sensitive to inclusion of other variables, notably institutional quality measures, which will be discussed shortly. The estimated coefficients of country size are occasionally significant and mostly of the expected sign (-h). Columns (1) and (2) show that coefficients of AGINI60 and AGINI70 are both significant at the 1-5% level and of the expected sign (-). According to the point estimates, an increase in inequality of 10 Gini points is associated with a decrease in growth rate of income per capita of 0.5-0.53% per year on average. In columns (3)-(5), we add two important variables to our basic regression, trade openness (TRADE) and institutional quality (GADP and POLCON) indicators, which are usually included in growth regression. TRADE is the sum of exports and imports as a share of GDP, averaged over the period of 1970-2000, which is from World Development Indicators CDRom of World Bank (2005). As an overall institutional quality measure, we use Government Anti-Diversion Policy (GADP) from Hall and Jones (1999). It measures on a 0-1 scale how effective government policies and institutions are in supporting production over diversion. The original source of the data is International Country Risk Guide (ICRGE) data by Political Risk Services that rates 130 countries according 24 categories, which has been very popular in the empirical growth literature. It is the average of the following 5 categories for the year of 1986-1995: (i) law and order, (ii) bureaucratic quality, (iii) corruption, (iv) risk of expropriation, and (v) risk of government repu-
o o o o 03
OH
Q
o Ü
p:^
Co
I o
I T3
PQ
I O
3
122
2i
00
CO
^
^
HJ
CO
^^ o
00
CO
'O CO
'O
^
a
^
CO CO (M * T-l ^ O O O (M O O O O O O O (M
o o o o l t j O ^ , 5 o o o o g o ^ P Ö Ö c5 Ö o cj
00 O O
Ö O Ö Ö '^^ Ö P
T—I C<J 1—I 1—I
O
O
r
t
o o ^
-Jt ^ CO O
O
H
g O O O C N O C O r H
CN
O
r
O o o o o o o o o ö ö ö ^ ö ö ö ^5
O
g o o o o o P o o o o o 9 c5 Ö Ö Ö c5
^
o
•^ O
^ lO
05 ^ C O T""
o o Ö O
o o o o Ö Ö
O * rH ^
^
O
lO
Ö <=? g
o o
CTi C O
o o
hs
o o Ö
00
I
s^
CJi
^
Ö
O
Ö 3
o
2 ^ ^
.
Sg
'-' ' ^ Ö
o
<^
ss
O (M 2 ^ ^ Ö
^
g S °0 5 CO
Ö
S S ;^ S ^
CO
o o ""^ Ö
CN O
1-1 CO r-( O Ö CD
^
q 3
Ö
o O
2 Ö
0 0 CO
O
PH
o o
00 I—'
Ü
i:^ o
p
0. Q
o 'P o Ö 3 Ö
^ 3
cr^
^ g ^
I t i 00 £2 ^
g o o o o o o o • ö ö ö ö ö ö ö o o
lO
T-H O
1-i
Ö
S'CO ^ ^ »o '—' q o q ^ Ö <^
Ö
o
o <
CO
o
g 1—H
o <<
EH
o
o o 3 § o o P W Ö c=5 9 o ^
1-i
o o o
-XCO
CO (M T-H o ^
o
o o o Ö CO O
^
g S ^ 9
oo o o o o
o o p
o 5^ o o Ö
o PLH
^-3
o
-
PH
W
CO
Ol
^
'M
CD
>
^
bO
CO
CT* ^ J
.=" ^ ^
a^
bJO Ö
Ö CO
03 4:5 ^ •—' o '^
O)
Co CO
4J
ö;
^ CO hJ
^^
:3
^
*> s^ !3
3 ^ «^
T3 CO Ö f-i O CD CO
O
o
CD n 3 *
-a-
^a 2 T? 0
"^ ^ CO g CO
CQ O) T3
O •-; T3 s-i p 0)
S OH'
w
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
w
H lO
o o
^ o
^ o
^
o
2 ^
CO
6.2 Inequality and Growth: Cross-Country Regression
123
diation of contracts. The indicator POLCON, which is obtained from Henisz (2002), captures the extent to which the executives face political constraints in implementing his or her policy. It is based on the number of institutionally embedded veto players among various branches of government.^ As we have argued in previous chapters, social polarization effects and hence resulting fiscal instability may be greater in the absence of institutional constraints on heterogenous policymakers with potentially conflicting interests in decisionmaking process. Persson et al. (1997) show that separation of powers with appropriate checks and balances can lead to significant improvement in equilibrium outcomes by reducing the rents extracted by politicians. Thus, one can argue that better checks and balances and greater power dispersion may lead to more sound economic policy and therefore to higher growth. The estimated coefficients of trade openness and institutional quality are highly significant and of the expected sign (+), as the empirical growth literature typically finds. Even after controlling for international trade and institutional quality, the coefficients of AGINI70 remain statistically significant at the 1-5% level except for column (4) in which it becomes insignificant. However, Easterly (2002) finds that high income inequality has a large and significant negative effects on institutional quality and human capital accumulation. Indeed, the correlation between AGINI70 and GADP is substantially high at -0.54.^ At the same time, better institutional arrangements may mitigate harmful effects on growth that income inequality exerts through various channels. Therefore, column (4) does not necessarily imply that income inequality is not important in determining economic growth simply because the coefficient of income inequality becomes insignificant. We have more to say about this when we turn to robust estimation results. Next, we check the robustness of our results in terms of the observations by using a robust estimation method. The OLS estimates tend to be sensitive to outliers, either observations with unusually large errors or influential observations with unusual values of explanatory variables (often called leverGlaeser et al. (2004) criticize the use of measures of institutions used in the current economic growth literature such as measures from the ICRGE, arguing that these measures are based on subjective assessments of risk for international investors along such dimensions as law and order, bureaucratic quality, corruption, risk of expropriation by the government, and risk of government contract repudiation. They argue that measures of contraint on the executives are probably the best of the measures commonly used in the literature, although even these seem to be based on political outcomes. In that sense, therefore, POLCON may be a better indicator of institutional quality. In a regression of GADP on AGINI70 along with initial income level, initial human capital and country size as measured by initial population, the coefficient of AGINI70 is statistically significant at 5% and of a negative sign. In a comprehensive examination of the channels linking income inequality and growth, Perotti (1996) also reports strong empirical support for the linkage of income inequality to the education/fertility decision.
124
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
age points). In a recent evaluation of growth regressions in relation to policy variables, Easterly (2002) argues that some of the large effects of a policy variable(s) are often caused by outliers. Thus, it is important to make sure that some of our results are not unduly driven by outlier observations. One of the most common ways to deal with outliers is to drop observations one at a time or to use single-case diagnostics such as Cook's distance measure, the studentized residual, or DFIT. But this is often inadequate because it may miss a group of outliers due to the masking effect. Instead, we employ robust estimation to obtain estimates that are not sensitive to outliers, and hence to characterize the most coherent part of the data set. This estimation involves a reweighted least squares (RWLS) procedure. We first use the least median of squares (LMS) estimator due to Rousseeuw (1984) to detect outliers, which is given by min median ef,
(6.2)
ß
where Fj is the residual of the zth observation with respect to the LMS fit. This LMS estimator, typically computed by approximate algorithms, can resist the effect of nearly 50% of contamination in the data. A disadvantage of the LMS method is its lack of efficiency because of its unusually slow convergence, making it unsuitable for inference. To deal with this problem, we use the LMS estimates to classify some of the observations as outliers, and then carry out a simple reweighted least squares (RWLS) procedure by assigning zero weight to outliers and full weight to the rest of the observations, as recommended by Rousseeuw and Leroy (1987).^ Columns (6)-(10) of Table 6.1 display the robust estimates of the coefficients of the explanatory variables in the growth regressions. They are largely similar to the OLS results, but all of the coefficients of LSYR1570 become insignificant. Trade openness is now a bit sensitive to inclusion of other variables, whereas country size often enters with significant and positive coefficients. By contrast, it is noteworthy that coefficients of income inequality indicators and institutional quality remain statistically significant at various levels of significance. As a matter of fact, the statistical significance of income inequality coefficients strengthens.^ Figs. A weight Wi is assigned for each observation as follows: Wi = l\iii
< (2.56-)^
= 0 otherwise where the robust standard error, a, is given (after running the LMS) by a = 1.4826 [1 -f 5/(n - k)] y^mediaui ff; i is the residual of the ith observation with respect to the LMS result; n = number of observations; and k = number of explanatory variables. As a further robustness check, we tried the Least Absolute Deviations (LAD) regression method. Since the LAD is designed to estimate the effects of explanatory variables on the conditional median, rather than the conditional mean as in the
6.2 Inequality and Growth: Cross-Country Regression
125
6.1 and 6.2 show scatter plots of growth rate of real GDP per capita against AGINI70 and AGINI60, respectively. The figures are based on the samples of countries in columns (6) and (7), which exclude outliers as were identified by the LMS. A negative correlation between the two measures is evident in the figures.^ Indonesi
.041117-
Barbados Don?Wa»Ja Portugal
-
Norway India Chile
^a«-i Carteawl JUniMili#|6lwy
te^'" Belgium 'taly ** France
Austral! Denmark Uganäi'^f:;^,^,
Turkey
Greece
Braz"
TririßokWiftlö^'
u5|uay P-arr,«
Mexico Ken>Bcuador
NewZeaF°='^'^
''""'"
^"^*«"«'
Switzeri Argentin
South Af
El Salva
Senegal Jamaica
.001533 -
S .281
1
1
p^^
\
r~ .6538
AGINI70n
Fig. 6 . 1 . Scatter Plot of real GDP per capita growth against AGINITOn Note: This scatter plot corresponds to the sample for Column (6) in Table 6.1.
Table 6.2 presents the OLS and robust estimation results for the TFP growth. The explanatory variables in TFP growth regression are identical to those used in income growth regression with the only exception being the log of initial level of TFP relative to USA (LTFP70) in place of LRGDPCH70. The results are quite similar to those of income growth regression. Income inequality is consistently associated with slower TFP growth in our sample, and again the robust estimation yields even stronger negative effects of income inequality on growth than the OLS. The coefficients of trade openness and OLS, of the dependent variable, it is also known to be resistant to outliers. The LAD regression outcomes are consistent with our robust estimation results. So we do not report LAD results. ^ Even if we use all the available observations, a negative correlation between the two are still relatively high. The pairwise correlation coefficients between AGINI60 and GRGDPCH7000 and between AGINI70 and GRGDPCH7000 are -0.35 and -0.25, respectively. Based on the samples excluding outliers, they are -0.63 and -0.39 each. Simple comparison of the above results suggests that our LMS procedure to detect outliers and hence to get the most coherent part of the data is highly effective.
126
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence .041117
I
Barbados IWlfiiflca Norway Pakistan
O O O X Ü Q. Q Ü Q
EgyfiPP^n
I^HlWhlle Brazil Turkey
Candal«^?'-^'"'" United K 1
Genmany Australl
Netheria
France
Greece Malawi Trinidad
Colombia
Denmark I
Banglade
Mexico
Uruguay 'PQvwswten " ' Ecuador COSUR,™««-
Kenya ^
^
Argentin B^j^|,duras
South Af Senp%^
-.001533 H .3068^
.628 AGINieOn
Fig. 6.2. Scatter Plot of real GDP per capita growth against AGINI60n Note: This scatter plot corresponds to the sample for Column (7) in Table 6.1. institutional quality indicators are highly positively significant as well. But it is a bit surprising that human capital as measured by average years of school turns out not to be consistently associated with growth. Next, we turn to potential endogeneity problem of trade openness and institutional quality using the instrumental-variable (IV) method. By using some exogenous components of instrumental variables largely determined by characteristics of countries (that are unrelated to the unobserved error), we investigate whether our earlier OLS regression and robust estimation results are still valid. Our main instrument for trade is the gravity-model-predictedtrade-share from Frankel and Rose (2002) that extends and revises the data in Frankel and Romer (1999) by adding common languages, common borders, and common colonizers to the original two variables of bilateral distance and landlockedness in estimating the gravity model. This is widely used as an instrument for trade in estimating the impact of trade on growth or levels of income in the literature. Our main instruments for GADP are EURFRAC (fraction of a country's population speaking one of the five primary Western European languages-English, French, German, Portuguese, and Spanish-as a mother tongue) and DISTANCE (distance from the equator as measured by absolute value of latitude in degrees divided by 90), which are used by Hall and Jones (1998).^ Similarly, DISTANCE and POLCON6069 (POLCON av^ The European settlers' mortality rate in their colonies from Acemoglu et al. (2001) is arguably the best instrument for institutional quality across the countries in the world today. Yet we do not use the settlers' mortality rates as the instrument in our IV regression for two reasons. First, this variable does not work well in growth
4
o o o
H
ÜH
d.
O
. 2 CU
en Co
^s O 2
go ^ Jrk h o -r«
;5
25
O
JD O
^ ^
00 ^ ^
^
P^
O
--^
^-^
CO
'O
CO
to i-J
^ ^
CO
•o
I o 1ö
& H
w
H
W Co
H
So o
Ö
^^
2 § o •X-
Ö q
o o
g§ •L, (N "^ iO
^ <=> 2 o q o Ö 3 Ö
o ^ 3
M
-)f •X•X-
•X-
lO CM
o d
CM ^ O
^.^
00
s.
oo' o o d^
o o d^
ooT q d
q d
T—1
o d co" o d^ TH
d
TH
o O
+3
-!.:>
d
CM O
Oi O
O
o O d 3^
1—1
o o o d d ^„.^
O
CO
'
o d
_^ ^._^ ^ t - t^ 1—t
d
oo" o o d^
CO to
CO LO
to
Oi
CO
d
c^ d
LO
LO
d
CO
d
CO
d
05 CO
d
d CD CO
d
d CM
d
Ö ^ ^ <^
O
w
CM CD
LO
CM CD
CM CD
CD
o
LO LO
CO LO
=! ;a :2;
o
O
l>
T}^
l^
O ^ o CO o t ^ O o o d 3 q d d1 ^ ^—'-
CD 1
6.2 Inequality and Growth: Cross-Country Regression
O
O
O
;_
°s Q
Ö ^
o
q o
o Q o <=^
5 o o 9 p^ o
^ o
(
O
O
^ O o h^ o a.
d
O
CD
•X-
d
9 9
•xo q d
00
IH
O
3
O
CS| O
O
o
Q
Ö
J-J O
rH O
0
d 3
g 3 ö
O
s og s
q
(M O
C
CM '~'
II Ö
3
q q o q o 9 3 ö 3. ^
O
t^ CO
o d
TH
CVJ
^ O
CO O
I
rH O
° P
s § s <=?
i S ^2
rH O
o <:
CD
o
O O
rH rH
-
o q q q o q Ö o Ö Ö Ö Ö
T
CM O
'^^ C^ O
o o Ö
o o o q o o o o Ö ^ Ö Ö Ö Ö Ö
CO O O
,5 o g o o q Ö 9 '-^ CO O
(M O
o o o o o 9 o d> p> d> d> ci> P^ ri
-^ rH
CO O
o d
CO ^
o q q q q q o o ö o ö ö ö ö ö ö O O O
r• H •
O O
CM cb O o O Ö o q. ;::z ^
TH
CM
^ 22 '-^ Q W q, -x-
QO 00
CO
^ O ^ O CO o q o qq q o Ö' 9 CD Ö
a <:
o
^^ o o '"' 9
o
^ o o o Ö
CO
o 5?
hJ
o
PLH 1-3
CO
CU
"T?
CD
TH
127
H
W
f-i
••g-s
O
O
*4J
CO
:a <:
•5^ •X-
0
2
a;
ac2 13
R f^ a Co
^
M
CO
1^ c6
4J
ö
.2 -^ 2 0
'" CO
^ gCJ CO
Ö g CO
1—^
M
s*. ä
'^ 'S
128
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
erage in 1960-1969) for indicator POLCON. Finally, we use PRIMCOMP1960 (percentage of "primary school complete" in the total population over age 25) and AYS 1960 (average years of schooling of population over age 25 in 1960) as instruments for our income inequality measure, which is consistent with the human capital model of income distribution including the work of Schultz, Becker and Mincer that the distribution of income is determined by the level and the distribution of schooling across the population (see De Gregorio and Lee 2002).^^ The second-stage regression in the IV estimation always include LRGDPCH70, LSYR1570, and LPOP70 as basic explanatory variables. As an IV method, we employ two-step feasible efficient GMM (generalized method of moments), which produces a consistent and also efficient estimator in the presence of heteroskedasticity that is more likely in a cross-country study. The conventional IV coefficient estimates are still consistent, whereas its estimates of the standard errors are inconsistent. The latter can be partially addressed by using heteroskedasticity-consistent Huber-White standard errors, yet this conventional IV estimator is still inefficient when there is heteroskedasticity. Table 6.3 presents the IV regression results. They are largely consistent with our previous OLS results except for the fact that coefficients of GADP and POLCON are now insignificant. AGINI70 and TRADE enter the growth regression with statistically significant coefficients, indicating that better income distribution and greater international trade lead to faster growth, rather than the other way around. The instrumental variable must satisfy two requirements: it must be correlated with the included endogenous variable (s), and orthogonal to the error process. Over-identification test (Hansen J-test statistic) is employed to test the validity of instrument (s) that the instrument (s) is orthogonal to the error process. The null hypothesis is that the instrument is orthogonal to the error. A rejection of the null hypothesis implies that the instrument (s) are not satisfying the orthogonality conditions required for its employment. Hansen J-statistic is consistent in the presence of general form of heteroskedasticity. As one can see, they are all accepted, indicating that these IVs satisfy the orthogonality condition. The last columns show Ftest statistics from the first-stage regressions, a test of joint significance of the excluded IVs. The first-stage regressions are reduced form regressions of the endogenous variable on the full set of instruments, so that the relevant test statistics relate to the explanatory power of the excluded instruments in these regression. The coefficients of institutional quality indicators in the second-stage regressions are not significant. Originally it was used in income level regression in their paper. Secondly, Glaeser et al. (2004) make a case that it is not clear whether the settlers' mortality rates should be viewed as an effective instrumental variable for quality of institution. ^^ Initial income distribution can be viewed as a predetermined variable. Thus, there is less concern with reverse causality or simultaneity problem. Yet Lundberg and Squire (2001) argue that inequality is an endogenous variable jointly determined with growth and hence one cannot expect a stable relationship between the two.
6.3 Income Inequality and Fiscal Volatility
129
regressions. The F-test results indicate that our instruments are significantly correlated with the endogenous variable.
6.3 Income Inequality and Fiscal Volatility Since we have seen evidence that income inequality is consistently negatively associated with economic growth, we now turn to the fiscal instability mechanism through which income inequality is negatively linked to growth. We separately examine the relation between income inequality and fiscal volatility, on the one hand, and the relation between fiscal volatility and growth, on the other hand. In implementing an econometric exercise to test the hypothesis that fiscal volatility will be greater in a country with highly polarized society, we have to find an appropriate measure of fiscal volatility. Our theoretical models in the previous chapters highlight strategic behaviors of the policymakers in determining fiscal spending. In Chap 4, we have presented evidence that government spending and its components, not tax revenue, tend to exhibit much wider cross-country variation in the magnitude of their fluctuations over time. Moreover, fiscal spending is a policy variable that is directly influenced by policy decisions of the government. Thus, it is natural to consider the standard deviation of fiscal spending (or of its annual growth rates) over our sample period as a measure of fiscal volatility. However, it may be useful to further distinguish discretionary fiscal spending from that used for business cycle management because the former is more likely to destabilize the economy and its volatility can itself be an indicator of aggressiveness in using discretionary policy. Therefore, we construct two indicators of fiscal volatility, using time series data on fiscal spending of each country in 1970-2000. The indicator FISVOLl is the log of standard deviation of annual growth rates of real general government final expenditure, which is the nominal expenditure deflated by GDP deflator. Raw data are from World Bank (2005). The indicator FISV0L2 is the log of standard deviation of the residuals from time-series regression of real government expenditure growth on macroeconomic variables for each country for the period of 1970-2000, which is derived from Woo (2005b).^^ The idea is to obtain a measure of fiscal volatility that reflects aggressiveness in using discretionary fiscal spending which is not used for smoothing out the output fluctuations over the business cycle. That is, the residual captures changes in fiscal policy that is implemented for reasons other than current macroeconomic condition such as booms or recessions.^^ ^^ We could have simply used standard deviation measures, rather than the log of the standard deviations, but we get quite similar results. Yet the logarithm of the standard deviation fits the data better, so we use the log form. ^^ Specifically, we estimate the following regression equation for each country in the periods of 1970-2000: AlogGit = ai + ßiAlogRGDPit + SiAlogGu-i
+ Ci^u + en,
Table 6.3. Growth Regression: IV Estimation by GMW Dep. var: Real GDP per capita growth Dep. var: T F P growth Explanatory Variables (2) (4) (5) (3) (6) (1) -0.009** -0.004 -0.002 LRGDPCH70 (0.004) (0.007) (0.006) -0.013* -0.013* -0.01** LTFP70 (0.004) (0.003) (0.004) (relative to US) 0.005 0.005 0.004 0.002 0.002 0.001 LSYR1570 (0.003) (0.003) (0.004) (0.002) (0.002) (0.003) 0.002 0.002 0.003*** 0.001 0.001 0.0001 LPOP70 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) -0.099* -0.186*** -0.150** -0.076** -0.098*** -0.132* AGINI70n^ (0.038) (0.109) (0.059) (0.032) (0.06) (0.051) 0.015* 0.016* 0.013* 0.009* 0.008* TRADE 0.005 (0.002) (0.002) (0.003) (0.003) (0.003) (0.003) -0.053 -0.011 GADP (0.058) (0.023) -0.04 -0.022 POLCON (0.025) (0.016) 0.015 0.103** 0.136** 0.035 0.061*** Constant 0.085 (0.046) (0.061) (0.053) (0.019) (0.047) (0.035) 0.122 Over-identification 0.22 0.111 0.36 0.183 0.763 Accept Accept Accept Accept Accept Accept J-statistics 0.000 F-test (p-value) on 0.026 0.000 0.000 0.003 0.001 0.000 0.000 0.001 0.000 joint significance of 0.000 0.001 excluded instruments 0.000 0.000 0.000 0.000 60 56 No. Obs. 60 60 60 56 Levels of significance are indicated by asterisks: * 1 percent, ** 5 percent, *** 10 percent. White heteroskedasticity-consistent standard errors are reported in parentheses. See data appendices for definitions and sources. " We employ a two-step feasible efficient GMM as an IV method, which produces a consistent and also efficient estimator in the presence of heteroskedasticity that is more likely in a cross-country study. The excluded instruments are AYS 1960 (average years of schooling of population over age 25) and PRIMCOMP1960 (percentage of "primary school complete" in the total population over age 25) in 1960 for AGINI70; gravity-predicted-trade-share from Frankel and Rose (2002) for TRADE; EURFRAC (fraction of a country's population speaking one of the five primary Western European languages - English, French, German, Portuguese, and Spanish - as a mother tongue) and DISTANCE (distance from equator - absolute value of latitude in degrees divided by 90) for GADP; DISTANCE and POLCON6069 for indicator POLCON. The second-stage regression in the IV estimation always includes LRGCPCH70 (log of initial income per capita), LSYR1570 (log of years of schooling in 1970), and LPOP70 (log of population in 1970) as basic explanatory variables. Over-identification test (Hansen J statistic) is employed to test the validity of instruments that the instrument (s) is orthogonal to the error process. The null hypothesis is that the instrument is orthogonal to the error. A rejection of the null hypothesis implies that the instrument(s) are not satisfying the orthogonality conditions required for its employment. Hansen J statistic is consistent in the presence of general form of heteroskedasticity. Reported p-values of F-test are in the order of the instrumented variables, AGINI70n, Trade, and GADP (or POLCON). ^ AGINI70n is expressed as a fraction by dividing AGINI70 by 100.
6.3 Income Inequality and Fiscal Volatility
131
The two measures are vrey highly correlated to each other (their correlation coefficient is 0.97). Fig. 6.3 shows the scatter plot between the two indicators. Congo, D
.777859 Argentin Banglade Nicarag9"y3"«
Portuaal India l^nd
-4.52899 1
1
1
1
-4.18274
-.691635
FISV0L1
Fig. 6.3. Scatter plot of FISVOLl against FISV0L2 Table 6.4 reports regression results for fiscal volatility as measured by FISVOLl. The log of initial real GDP per capita in 1970, LRGDPCH70, is introduced in order to control for potential effects of economic backwardness on fiscal policy. For example, poor countries may have relatively inefficient tax and spending systems and may therefore be more prone to poor fiscal outcomes such as frequent uses of discretionary policy. Also, LRGDPCH70 may capture some socio-political effects on the fiscal outcomes if social confiicts are greater in poor countries. LRGDPCH70 tends to enter the OLS regression with negative coefficients, indicating that poorer countries tend to experience greater fiscal volatility. Its coefficients are significant at the 1% level in columns (1) and (2), but loses statistical significance once we control for institutional quality, which we will return to shortly. Government size is included to control for the stabilizing effects of government size on GDP volatility: larger governments stabilize output (see Gali, where Gu is real general government final expenditure, RGDP is real GDP and "Kit is a vector of other explanatory variables such as inflation and inflation squared. Our country-specific measure of discretionary policy is the standard deviation of the residuals (s.d. of Su) from the above regression, which is denoted by aei. That is, we measure the aggressiveness of discretionary policy by the magnitude of discretionary spending shock volatility that is not accounted for by macroeconomic variables such as real GDP and inflation. See Woo (2005b) for more details.
+= 03
3
bJO O
o
T3
Q
O
Co
o Ü
X
132
1 CO
o
CO
o
CO
>
fo O «H-l
3
(-1
Ö [JLH O w CO r ^ CD H CO bO
^ S 6 T3 Ö CD
a
Q;
H Q
^ 3Co
a CO
w Ö
Co
.2 c«
a H
(M
o
I>
CD O C3 CD Ö •X•X-
CD
o
Ol
t^
o o^ o o_^
o
3
Ö 3 tin Ö
rH
^ '-^ <^
CNJ
^
^
CO CO ^" ^ l O O <' "^' r-H ^ CD o Ö o
NCM
CZ5 o
C^ i-M
O
O
CO S2 CD O P r-(
CM O
lO O ^ O
. ^ d
^
Ö
. co" o Ö
o:,
co" o 9 Ö ^
CM CM g
q o
CD i-j CI3 tvO O O CM
T:J< 0 0 CD O C^ CD
CO O
^ l O CM CO ^t CM 0 5 ^ ^ ^
'"1 O O O -^
^
q / O
'-I ;S! 1^ S »o o^ ^ p d
CM ^1 ^ « CM W CO (,N ^ o CO «-I H o CJ d CD
9
r^ cT ^
Ö p '" Ö 9
CM S 00 l O CD CO O CO o < ^ o
^ s s
,^ ^ O
<M £ 2 C:5 P T—I
O
O
^r^
S
CM
P
CM
^ C7>
O
^
•^
w H
< ec^ ^
<^ O
0. Q
lt.,
p^
X
<
w o
Q
p d P d p d
Ü
Q O
1—1 CJ5
CX3CMU0 ^ l O C D
o Ü
o
Ö
o Ö
crj Ö
o C5
CD
a o o o o ( oD Oo Oo O O o
o Ö
u ^ o O
o Ö
a . I—I o CP O O
S ^
:3
O CD
'
^ ^ <
-5 o
JD
CO
<
O
:a < 1:; to
)t
""•
2
Ö
CO
a »^
X- Va ^•^ Ö
^O X3 ^
CO
CD
Ci^cs
CO
1—1
^
n ^
'^ O ^
CO
Co
Ü
•4J
o o
+3 rO
^
-d CD
I ^
Ü
^ -> J^ «1 •* ^
^ " ö^
p
Ö OJ "
a^ hO
^ ^ 2
^
.^ t: -^
Ö
S ^ § ° >.
g
£sgi I t 2 &^"
o .S 5 •« 5
^ -ij
V. ^ ^ ^ ^
M
^
a
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
^ ^^ ^ O
C3
^ o" ^ ^~^o ^ „_^ 05
4J CO ;3 ^
^-^ o p^
^-v CD
'—^ o f^
'O
>
o
6.3 Income Inequality and Fiscal Volatility
133
1994; Fatas and Mihov, 2001 for evidence in the OECD country sample). Government size is proxied by the general government expenditures as a percent of GDP (AGEXP), averaged over the period of 1970-2000. One can associate the size of government with the strength of automatic stabilizer. If the built-in automatic stabilizer is fully functional, there may be less need for discretionary fiscal policy, which would then result in smaller fluctuations in fiscal spending. This is particularly true of the case of FISV0L2. But if large government means large fiscal deficits, there might be bigger fluctuations in fiscal spending over time as bigger fiscal adjustments become necessary later. Thus, the sign of the coefficient of AGEXP is ambiguous. Trade openness (TRADE) is also included in our basic specification. In his widely cited paper, Rodrik (1998) argues that given that the government attempts to facilitate consumption smoothing by conducting a counter-cyclical policy, more open economies tend to have larger governments because trade openness exposes a country to external shocks. Similarly, greater exposure to external shocks may lead to greater fluctuations in fiscal spending if the government tries to offset the shocks by fiscal policy tools. Then, trade openness is expected to enter the regression with a positive sign. However, the OLS coefficients of AGEXP and TRADE are insignificant and even change their signs. Also, we alternatively include measures of institutional quality, GADP and POLCON, for a couple of reasons. First, high-quality institutions can make a difference for public finance: a more efficient tax-collection system and better monitoring on disbursement should strengthen the fiscal position of the government and would avoid drastic changes in fiscal policy. Second, when institutions of confiict management are well-established and work well enough to suppress conflicts of interest among different groups, the social polarization effect we found earlier may be less important in determining the fiscal outcomes. Third, institutional quality indicators such as the efficiency of public sector, the degree of corruption, and the rule of law can be a good measure of the quality of budgetary institutions governing fiscal policy decisions. This is because government institutions tend to be shaped by common factors such as economic, political, cultural, and historical circumstances and change only slowly (La Porta et al. 1999). Indeed, the coefficients of GADP and POLCON are significant and of the expected sign (-), which indicates that better institutional arrangements tend to be associated with smaller fiscal ffuctuations. Even after controlling for the quality of institutions, the coefficients of income inequality indicators AGINI60 and AGINI70 are significant at the conventional levels and of the correct sign (+), as our theory predicts. Countries with unequal income distribution tend to exhibit greater voaltility in fiscal spending. The economic magnitude of income inequality can also be substantial. An increase of 10 Gini point is associated with an 11-23% increase in fiscal volatility as measured by the standard deviation of fiscal spending growth rates. Overall, more than half of the cross-country variation in fiscal volatility is explained by our regression specification.
134
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
As a robustness check, columns (5)-(8) of Table 6.4 report results based on the robust estimation. The robust estimation outcomes largely consistent with the conclusions from the OLS regressions in columns (l)-(4). Yet the robust estimation results are more reliable since it attempts to characterized the most coherent part of the data at hand. In this sense, we are more confident that income inequality is very strongly associated with fiscal volatility, for the statistical significance of income inequality coefläcients strengthens further. Accorindg to the robust estimates, an increase of 10 Gini point is associated with even greater fiscal volatility, ranging from 21% to 32% increase. Figs. 6.4 and 6.5 display scatter plots of fiscal volatility measures FISVOLl and FISV0L2 against AGINI70, respectively. The positive correlation between the two indicators is striking. It is also interesting to note that government size tends to be associated with smaller fiscal volatility. But the coefläcients of TRADE are very imprecisely estimated. Overall, the robust estimation further improves the goodness offit.About 71-79% of the cross-country variation in fiscal volatility is explained by the explanatory variables of our regressions. Banglade
-1.24843
Dominica
sierra L Malawi ''"""''^^'^
Uganda Indonesi
Venezuel Jamaica FiiCoS^Sfflk
EgyghLank
^'«^7afel„
Philipp«
^ja
Korea. R '^^"
iFinland
Peru
Botswana
GrJA^and Piftfigi^"
Tunisia
AU§lyali Canada Japan _ _, r,^i„!,,„
France USA
-4.18274 1
1
1
28.1
1
r—
65.38 AGINI70
Fig. 6.4. Scatter plot of FISVOLl against AGINI70 Note: This scatter plot corresponds to the sample for Column (5) in Table 6.4.
Again, we try to address potential endogeneity problem of trade openness and institutional quality using the instrumental-variable (IV) method. Columns (9)-(ll) apply the IV method. The IV results indicate that an increase in income inequality leads to greater fiscal policy shocks, rather than the other way around. We used the same set of instrumental variables as in the IV regressions for growth (except for AGEXP being instrumented by the 1965-69 average of AGEXP), and they satisfy the two conditions for an appro-
6.4 Fiscal Volatility and Growth
135
Bangtade
1.26102-
Zambia Dominica ^9^"«^^
Sri Lank Indonesi Pal
^gyp^
... . Malawi , Jamaica Sierra L
Venezuel ,, p^f^ ^P,ff"^uatemal Colombia *^'^ cuaiemai ^^^^^ Q^^II
Ecuador
Chile
Korea. R Austral!
^^^1500*1^ New Zeal Tunisia Portugal
Mexico
India Italy'^^'^"^ ' 1inland Germany Norway United K penmaric Spain R^ininm NethdBanada Japan Swedeff^'S'""^ Switzeri Austria USA
-4.4221 S 28.1
1
1 AGIN 170
I
T"
65.38
Fig. 6.5. Scatter plot of FISV0L2 against AGINI70 Note: This scatter plot corresponds to the sample for Column (5) in Table 6.5. priate instrumental variable as indicated by over-identification J-statistic and F-test on joint significance of excluded instruments. However, the coefficients of TRADE, GADP and POLCON are all insignificant. In Table 6.5, we show regression results for our alternative measure of fiscal volatility, FISV0L2 (log of standard deviation of residuals from the government expenditure growth regression), using the same set of explanatory variables. The OLS, robust estimation and GMM results are similar to those for FISVOLl. Notably, the robust estimation that is robust to the presence of outliers yields positive coefficients of income inequality that are all significant at the 1-5% levels. To sum, the above results show that high income inequality is consistently, positively associated with greater fiscal volatility.
6.4 Fiscal Volatility and Growth In this section, we examine the second part of the inequality-growth nexus, that is, relation between fiscal volatility and growth. We adopt the growth regression specification that is essentially the same as equation (6.1), but now replace income inequality with a fiscal volatility measure in the regression. The fiscal volatility indicators are expected to enter the growth regressions with significant, negative coefficients, other variables being controlled for. Tables 6.6 and 6.7 display growth regression results on FISVOLl and FISV0L2, respectively, using various estimation techniques. The estimated coefficients of both fiscal volatility indicators are highly significant and of the
136
CO
a;
O T3
CO
o
& T3
Pf.
^
1
-¥. ^3 o
^
CO
'O
^
° Ö o
v.-, CO CN
t-
g ° a0 s
lö M
+j
.2
Ö
w Co
Ö
w
O
0
C5
O
C<J rH ,-1
T-5
l> 10 CO
Ö
CO 00 ^
Ö
0 O
a o o o 0
0
O
o
O
o
o d d d
0 ) 0 0 C-> O O
JH 0 O
a o o o
d d d d
0 ) 0 0 CJ O O
JH
.S^
i
CD
0
^0
0 -JH
1•I
CO
8
CQ
0
CD
;3
1 1
b- a rH O O 10 cCJu oO oO oO CO
c5
O
(X
u o
c:^
9 3 9 0 9 0 (M (M 22 CO O (M
l>£2 <^
c6
CO --I
d> cS '^
^ (M
rH
00 !>•
d
CO
CO
CO
Ö
^
-sf< 00
^
o
CO
CN O
p
fO
0
CO CO
o
0
q 0
l> O 0
^ 0
CO 00
CM
0
CD
O
^ d
G5 ;i:; CS 1— ^I ^ (3 O o
0
CO
f?
00
CO CM 1-1
PH
Q Ü
00
^ fo § i § g
I
CD (M
^
nl ^
0
C5
CM
d o <^ d
0
T^
'i-T 10
,-(
CM CO O
r:J CO 10 CO
00 CO O
°e s^^ ^ s ?^
^
CT
P
^ o ^ '^ o o
;$; rH
^
O 1-1
2
W P
CO O
^ O
(X
^ H <
0
CM o
O ^
CU
^ hJ
Q 0
;3^ CO
Ö d . S 6. 9
CO
<^
s 0
s <
S
^
CM CO
ILO
l>-
CM CO
1^ 06
^-1 CD
0
0
.2
<
0:5
d
10
d
05 CO
d
CO
CM
CO
T3
•3
4J
.2
CO
a
CO CD
1 0
CD
T3
0)
CD
1
(-1
a CO CD 0
d»
a a
CD
CD
CO
-(J
CX 0)
CO
CO
CD CO
^ Ö
0
CO
'^
CO
o;
^
03 CD
CD
CD S-i
CD
CD 4J
0
SH
CO 4J CO
Ö 2i Ö Co Co "^ a cC •3 .bO
0) 0
Ö CD
a, Co
Ö CD 0
+£
•X•X-
iO
a
CD
SH
c;» T3
-j-T CO Ö CD CD
•X•X-
*
0 1—1
-^ <
u
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
.^ 00 ^ jD
S ^
-M tn
^—' P^ o
---. t-
CO
'O CO
bO
o
1
(M
0
CO
'O
o
"o o
03
Eo
O CO o
>
O
P H ^:]
Co 29
fa C+H
* M W
1 1 .s
&
Ö; bO
M
^0) .2
1—1
T3
50
i6
<x>
o
5 cd o H Q f^
6.4 Fiscal Volatility and Growth
137
expected sign (-), although there is some tendency for their statistical significance to slightly weaken when institutional quality indicators are included. As noted before, this is not surprising because better institutions are expected to offset some of the undesirable volatility effects. Nonetheless, the fiscal volatility continues to exert strong negative effects on growth. Figs. 6.6 and 6.7 visually confirm these by displaying scatter plots of real GDP per capita growth against each of the two volatility indicators. The country samples correspond to those in column (5) of Tables 6.6 and 6.7. Also, our basic growth determinants are mostly significant at the conventional levels. The overall goodness of fit improves substantially in both OLS and robust estimation results. .052178 H Hong Kon
Barbados Syria
^. Poffugal Norway
icelarti? Austria
Japai^rmiand
Pakistan Sri^^g & r a e l Trinidai^alaw'
Kenya ' " f H 8 W Uganda PhilipftoÄteWia Guate»3feroon ^. .. Mall ._, „ . Iran Zimbabwe SoiOl^^aKJa"^« ^^"'" Bolivia
. ^^9«"*'" Gambia.
Mozambiq Venezuel '09° Niger
.018205 -4.18274
GuyBflSglade
FISV0L1
-1.03632
Fig. 6.6. Scatter plot of real GDP per capita growth againts FISVOLl Note: This scatter plot corresponds to the sample for Column (5) in Table 6.6.
In the IV regressions, we used the same set of instrumental variables as in Table 6.3 and instrumented FISVOLl and FISV0L2 with PRIMCOMP1960 and AYS 1960 that were used to instrument income inequality in the previous sections. The basic conditions for good instruments are satisfied as indicated by over-identification and F tests. The coefficients of fiscal volatility measures remain significant and negative, indicating that greater fiscal volatility leads to slower economic growth. We have also tried TFP growth regressions, but they are remarkably similar to the income growth regressions. To save space, we do not report the TFP growth regression results.
138
O
o o
OH
-tJ
5-1
a Co Co ^ *a O >>ÜCo +3
'cö
t M
a;
u a < a
t?
1 ü w
fe CoD
4-5
o O
1
^
i^ ^ 0^ o 5-1 M-H
o Ö
^ 'S P^ >
^^ CO < D
-5 ft
HQ
o
o
g
o
CO
P
^
d 9
o
T:}^
o
*
q
^
o
05
^
T-H
O CO C^ O O ^ r-l S q o lo o o o q ^
si O Ö O r-^
o o o L_, P q P o
o q g q P q o q E o 3 Ö• B
2 § o o Ö p.
— 0 - -
0
^ 0 > CO
E
^ " - - - ^
w Q H
0
^ JI^
O
q
CM
<X3 .-(
O
§
O
9 d o ^-^
o Ü
PH
o
rH
o O
U O O
a, o CD O
o O
O
0 , 0 0 0 CD O O O U O O O
CM CO
i-H CM -^
o
CD o
0
cn
0 S 6 ^ "^ ^ ^ <
CD
0 Ö bp -TS CD 0
i-H
!_;
Ö .a
•^ a
0
IN
>
O
(D
O
s |
-^ ^^
o
s:s s CD . i : ^ CO
^ :t3 CO - ^ 03 CD
CO S
§ d
^ 8 3
r«5 ^ $g
o
^ '—^
H 'S CO 2 o
.2 Ö ^ -
-P^ bO
R CD
CO . 5 CM
^ CD CD >
CD r rTj
.a
a
;=j
CD CO o
2 B§
^
CO 54-1
^
CD
M
-.- fcuO Ö Co Ö a;
F^
a CO
Ö CO
- si
(V)
m g >
§ P. o
(D HH
13 O ,5
tj)
;
CD
P .
^
3
CO
=0 ^
CO
'ops
1 0 CO
O ,5
. ü CD
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
si O
o
'^ ; i ^ o t ^ o c M o O o ^ o ^
qogoooqoQog^
—
O
CO
o ^rH c3 o z: _, o o-. o o q o Q o o d
Tj^ ^
O ,-1 O ^
r-H ^
0
H^ '-I 00
10
PH
P. Q
S 9 S9
TH
d o^ d 3
0 0 0 9
O
— d3—• 9• 3
CM
C^
2 S |
si So
2
^
0 i>
CO
PH
J
s
o o r=! o ifB o q 9 o . o. o. o 9 o 9 d o (-5 o o o d 3 0 3
S o g o q 8 9 3 o
CD 3 Ö
o q 8
'=>- 's^ ooo o ; : : : ^ o £ 4 o ?^ < o q-. ow Q o q o § o o w ' ^ 3 d 3 ' = 3 9 3 d
CO IT
O
' S.Ö c5ri 3^ dr- 3 ( 2 ••§ '<=^ d> a d> ^. Ö '^
'O . CO
'O
>
w Ü
cu Q
Ü HJ
PH
~l-3
cö
'ö; Q O r^
• ^
o
od +^
'^
> CO
'O
'O
cö
^ o
'cö
OH
CD
^
cö
'a O
tClo
o ^
1 l^
r^
?
^ CCJ o ^ O t+H o o ;-l Ö O .2 rä CO 0)
'M
flS
3
^CÖ
O)
C^ >
H Q
-5 ^
*
CO
•)(• lO (M 05 52 CO <M r-l O Ji^ Tf O O O P
<M t t v f^
^
o lo o S o o o o 8 g 8 q S ö ö ö <=' 3 ö p •
CO —
S I CD8 ÖSOO -X-
d o lO
00
r H - ^ C O
o
e.
hs
O ^ O CO O q O o
O O '-" ö CD ° ^-" O C M C O t ^ C S *
r H o S o c o o o o o ö ^ ö o ö o d S
"=>. ^ "R
Q ^ o g o 5 o o o o
"^i ^
d O Q O O o d O ( - 3 0
\ <=><=> S R ^
q o o o o dl d> ö ö d) O O O 00 o o o g ^o ^ <=> ^ o ü: o ^
S
Q
<
(1H
P
<
§ o o o , D2 r-i ; ooi :L: : oo
-^ ^ . d
'^ * CO O <-! O O O O
O
CO
>
o q gg oq o o :? o oo ^ 3 - ^q ^ ^^^ . ^ ö^ ö^ _.' q^ cS d (-^ ö CO rH O
d d d d
(1H
o
CO
Ol (U
CP
O O
O
O O
O
O O
O
a o o o
lO 05
i> o r—I
^
^ _g) X!
fe o o
^ ^ <
0:5
^
ri
2
en
• ^ 1 I—I
139
O
> O r-v ^ lO
CD
Ö
i^
^I CO
S^ .2 O o
'+J
o ^
2S
OD
CD
a
^ a
CÖ
CO
^8
Co ^ - v
3 § d r^ ^ S H 'S 5
>
.2 d ^ c« S C<J
r£3
CD - ^
2 Ö o hO
CD
Hi
Ö 2 ^ .a
5 c6 " - " o
^ S8.S^ c5
!l^
riS
«4-1
j3 <;
CD
a »^ ^
^
cö ^ ;=! ^ O o O ^
la
[ill
CD 2 ^ 2 ^23 o ?3 ^ 2
§ o O
'S ' ^ T3
.§ •- -2 Co
1^
. . CD CO 0) -7^ CO
:3 <
II
6.4 Fiscal Volatility and Growth
t
S 8 o d d ^^
12; O O PLH
O
140
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence SIngapor
.061541 H
Hong Kon Thailand
^^P'y^Maesi
Norway Austria Japan
Portugal , . . ^ India
Syria
Finland ff^u ^
Chff W^^
Italy ---'
lu'rl^ey Austral! Greece
Lank
Quinea-B
Israel
Jordan Q Q I Q ^ ^ ^ W S O ^ . ^ ^ ^
^^xiy^anamgarag^^^
Ua\avA Uganda
3 , ^ 3 ^ Banglade
New^^Pfi^PP^ Switzeri
(awteö»»i Argentin
Zimbabwe El §ai#Whduras Senegal
Gambia,
^ßEH!
RwarK Jamaica
Mozambiq Venezuel Togo Niger
.018205
•1.10543
-4.52899 FISV0L2
Fig. 6.7. Scatter plot of real GDP per capita growth againts FISV0L2 Note: This scatter plot corresponds to the sample for Column (5) in Table 6.7.
6.5 Comparison with Sociopolitical Instability a n d Fertility Decision So far we have presented substantial evidence that support our theoretical predictions that social polarization as measured by income inequality is associated with poor growth through the fiscal instability channel. Also, robust estimation and IV regressions employed to address outlier and endogeneity problems confirm and reinforce the OLS results that income inequality measures are consistently positively associated with fiscal volatility, on the one hand, and the volatile fiscal outcomes are strongly negatively associated with economic growth, on the other hand. We have focused on the core set of growth determinants in the empirical growth literature, and evaluated the impact of income inequality through the fiscal instability channel by adding a fiscal volatility indicator to the standard growth regression. In this section, we check robustness of our empirical results with respect to sensitivity of our estimates to the inclusion of other variables in the growth regression. In particular, we want to compare our fiscal volatility channel to other competing channels that were found to be statistically significant and important by Perotti (1996): sociopolitical instability and fertility. Since we cover a different time period (1970-2000) and sample of countries, we construct two variables according to our time period under investigation. Perotti (1996) uses a composite index of sociopolitical instability derived from Alesina and Perotti (1996), which is constructed by applying the principal
6.5 Comparison with Sociopolitical Instability and Fertility Decision
141
component method to individual indicators of social unrest. In Chap. 5, we also used a composite index by applying the principal components analysis to five variables, military coups, constitutional changes, government crisis, political assassinations, and revolutions, found to be individually significant in fiscal deficit regressions. To capture the multidimensional political instability, once again we construct the following index, PINSTAB, based on the five variables from Banks (2003) for the period of 1970-2000 as follows. ^^ PINSTAB=0.59COUPS+0.52CONSTCHG4-0.17GOVTCRIS +0.25ASSASSIN+0.54REVOLS. Next, we use the measure of fertility that is the average of 1972 and 1987 values of this variable from World Bank (2004). In Perotti (1996), the average of 1965 and 1985 fertility values from Barro and Sala-i-Martin (1995) was used. We tried various measures of fertility, but the results do not change appreciably. Table 6.8 reports the OLS estimation results. First, column (1) shows that the statistical relationship between income inequality and sociopolitical instability is weak, in sharp contrast to Perotti (1996) and Alesina and Perotti (1996) which examine the data in the period 1960-85. Even in the most basic regression specification for PINSTAB, the coefficient of AGINI70 is completely insignificant, whereas the coefficient of LRGDPCH70 is highly significant and of the expected sign (-). This result is surprising, although we already noted it in a panel of 57 countries for a different time period 1970-90 in Chap. 5. We tried each of the five individual indicator of sociopolitical instability instead, yet we obtained much the same results (not shown to save space). Only when we use ASSASSIN as the dependent variable in the basic regression, the coefficient of AGINI70 is significant at the 10% level and of the correct sign (H-). When each of the other variables is used as an dependent variable, the income inequality coefläcient remains completely insignificant.-^^ Thus, previous finding of a significant positive relation between income inequality and instability is questionable, despite the popular notion that high income inequality tends to make a country prone to sociopolitical unrest. Possible explanations may include the difference in time periods and quality of data. Most of the studies that report such a positive relation between inequality and instability examine the period of 1960s-1980s (see Drazen 2000, pp. 500-523). It is also worthwhile to mention that we use income inequality data from Deininger ^^ Perotti (1996) includes democracy dummy variable as well, mainly because in dictatorships episodes of social unrest tend to be under-reported for propaganda purposes. However, including democracy dummy does not change his results. So we do not include it. ^^ If we use AGINI60 instead of AGINI70, the result is exactly the same. Only for ASSASSIN as the dependent variable, the coefficient of AGINI60 is statistically significant (at 5%) and of the correct sign (+).
142
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
and Squire (1996) that provides a first kind of comprehensive and high quality data on income distribution, which was not available in Perotti (1996). His main sources of income distribution data are Jain (1975) and Lecallion et al. (1984). The negative relation between instability and growth is a bit stronger, albeit sensitive to inclusion of other conditioning variables. Yet columns (3) and (4) show that the coefficients of PINSTAB are insignificant, when our core set of growth determinants are controlled for. There are some caveats here. Intuitively, if good institutions are in place and well-functioning enough to suppress the harmful effects of sociopolitical instability, the instability effects on growth may be limited. Then, it is not surprising to find the coefficient of PINSTAB to be insignificant when an institutional quality indicator is also included.^^ More intriguing, correlations between PINSTAB and fiscal volatility indicators FISVOLl and FISV0L2 are high. Their correlation coefficients are 0.60 and 0.59, respectively. This has two implications: one for the relationship between instability and growth and another for the relationship between instability and fiscal outcomes. Because of the high correlation between instability and fiscal volatility and the offsetting effects of institutions, only when we exclude both institutional quality and fiscal volatility indicators from our growth regression, indicator PINSTAB enters the regression with a significant positive-signed coefficient (not reported to save space). Thus, the relation between instability and growth is conditional. Conversely, the harmful effects of instability on growth are stronger in the absence of offsetting institutional arrangements. On the other hand, high correlation between political instability and fiscal volatility is consistent with our earlier finding that political instability is one of a few explanatory variables that were robustly positively associated with public sector deficits in Chap. 5. Theoretically, it is plausible that high level of sociopolitical unrest may make policymakers' time horizon shorten and focus on myopic policies, resulting in aggressive uses of arbitrary fiscal policies and hence greater fiscal volatility. Thus, the causality would run from political instability to volatile fiscal outcomes. Unlike the case of public deficits, however, political instability has significant coefficients only in a very simple fiscal volatility regression that excludes the income inequality and trade openness indicators. When both income inequality and political instability are included in the fiscal volatility regression, only the coefficients of income inequality are significant (not reported to save space). ^^ Similarly, Knack and Keefer (1995) find that, once institutional variables like the fairness and effectiveness of the judicial system and the stability of property rights are taken into account, the negative effect of sociopolitical instability on growth vanishes. More recently, Campos and Nugent (2002) also find that the evidence supporting the hypothesis that high levels of sociopolitical instability cause lower rates of economic growth is much weaker than generally believed, and that once one controls for institutional development or alternatively the terms of trade, the causality results vanish.
o Ü
•§
o
.2
>
*o o en
^
o
00 CO
2 &
Q
CO
hJ o
Jt; o lo o S o o o o R
^
*
O
CM
O
1-1
O
T-H
o o
(M 'Nt^
CM O
-XlO o
CM <-• O O o q
^q
rH O o
o o o o o o CM ^ CM O o o
CI.
O OH
CI
<
§
o
o
Ö
tr i-H
• CD
o o g o Ö
Si is rH Ö J2 o P o cp 3 Ö
S Ö ^
O Ü
H:I
O
•J
OH
o
>
^
CO
o >
S
CO
CO O
Oi CO
O O
tlH
O O
O O
? ^"I^ a,
CO ;H ^—s o -M
O O
o p lo q o q q c5 Ö ö d> Ö d> d>
^ O
K
d '5'
6.5 Comparison with Sociopolitical Instability and Fertility Decision
CO •
lO
2 CO HJ
143
J -5
ii
.
Ö
Co
2 <
O
Ö
I
Ö
cu
o
Co
^
03 T3
CO
H
o -fi H y
il
-^ cd
I t^iö a
CO
fe? -^
^
i_q
144
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
To sum, the sociopolitical instability channel that links the income inequality to growth is not supported by our data. Particularly, our regression results call into question the relation between income inequality and sociopolitical instability that previous studies have found. We turn to the fertility choice channel. Column (2) of Table 6.8 shows a simple OLS regression of fertility on initial income per capita, human capital and income inequality. Their coefficients are all significant at the 1% level and of the expected signs. The higher the income level and the educational level, the lower the fertility rate. The more unequal the income distribution, the higher the fertility rate. Since our main interest is in comparing our fiscal volatility channel with fertility channel, we do not further investigate the determination of fertility decision. It would suffice to show that there is evidence for a significant positive relation between income inequality and fertility rate. Columns (5) and (6) present the OLS regressions that add PERT to our growth regression equation. The coefficients of PERT are significant at the 1% and of the correct sign (-). It is noteworthy that our fiscal volatility indicators enter the regressions with highly significant coefficients. So do LRGDPCH70, LSYR1570 and TRADE. However, the coefficients of POLCON become insignificant. Like many other variables in the empirical growth literature, PERT, POLCON, and LRGDPCH70 are highly correlated to each other, which makes it difficult to isolate the fertility effect from income or institutional effect. It is because the demographic factors including fertility rate tend to exhibit some systematic patterns depending on the stage of economic development in a country. Yet one can still argue that a lower fertility rate is a precondition for economic take-off. Column (7) shows the IV regression result using the same set of instrumental variables for the endogenous variables, PISVOLl, TRADE, POLCON and PERT. (Por PERT, we used PERT in 1967.) The coefficient of PERT is significant and of the correct sign (-). Other variables except POLCON enter the regression with correctly signed coefficients, and largely consistent with our previous results. Remarkably, the coefficient of PISVOLl remains negative and significant at the 1% level. The same is true of the case in which we use PISV0L2 instead of PISVOLl. Once again the data lend support for the fiscal instability channel that links unequal income distribution to slow growth, even after controlling for PERT. Lastly, we tried a few more variables for the growth regression such as urbanization and external shocks (as proxied by standard deviation of first difference in log of terms of trade). But they do not change our main results appreciably. So we do not report to save the space.
6.6 Concluding R e m a r k s We have examined the fiscal instability channel that relates income inequality to poor growth in a cross-country data for 1970-2000. Not only we could confirm the negative relation between income inequality and growth in our
6.6 Concluding Remarks
145
sample, but we also presented substantial evidence both for the relation between income inequality and fiscal volatility and for the relation between fiscal volatility and growth. Moreover, we checked robustness of our results by addressing the problems of outliers and endogeneity and also by comparing other prominent explanations in the existing literature. In short, the evidence consistently supports our theoretical hypothesis that more unequal societies are more likely to run volatile fiscal policies that are harmful to economic growth. As for other competing explanations for the observed negative relation between unequal income distribution and growth, the data also lend support to the fertility decision channel, but do not support the sociopolitical instability channel. In particular, we do not find any evidence that sociopolitical instability is statistically significantly associated with initial income distribution, in sharp contrast to some of the existing studies. On the other hand, there seems to be a negative relation between sociopolitical instability and growth, but it is also weaker than previously reported in the literature. More work needs to be done on this theoretically and empirically in the future.
146
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
6.7 Data Appendix
Table 6.9. List of 93 Developed and Developing Countries Country
Country 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31.
Algeria Argentina Australia Austria Bangladesh Barbados Belgium Benin Bolivia Botswana Brazil Cameroon Canada Central African Republic Chile China Colombia Congo, Dem. Rep. Costa Rica Cyprus Denmark Dominican Republic Ecuador Egypt El Salvador Fiji Finland France Gambia Germany Ghana
32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62.
Country
Greece 63. Pakistan Guatemala 64. Panama Guinea-Bissau 65. Papua New Guinea Guyana 66. Paraguay 67. Peru Honduras Hong Kong 68. Philippines Iceland 69. Portugal India 70. Rwanda 71. Senegal Indonesia Iran 72. Sierra Leone Ireland 73. Singapore 74. South Africa Israel Italy 75. Spain Jamaica 76. Sri Lanka Japan 77. Sweden Jordan 78. Switzerland Kenya 79. Syria Korea, Republic of 80. Taiwan Lesotho 81. Tanzania 82. Thailand Malawi Malaysia 83. Togo Mah 84. Trinidad & Tobago Mauritius 85. Tunisia Mexico 86. Turkey Mozambique 87. USA Nepal 88. Uganda 89. United Kingdom Netherlands New Zealand 90. Uruguay 91. Venezuela Nicaragua Niger 92. Zambia Norway 93. Zimbabwe
6.7 Data Appendix
147
Table 6.10: Data Description and Source Variables AGEXP AGINI60
AGINI70
AYS1960 DISTANCE EURFRAC
PERT FISVOLl
FISV0L2
GADP
GRGDPCH7000 GRTFP7000 LPOP70
Description and Source General government expenditures as a percent of GDP, averaged over the period of 1970-2000. Source: World Bank (2005). Decade average of Gini coefficients obtained using all the data points available for the 1960s. Source: Deininger and Squire (1996). Decade average of Gini coefficients obtained using all the data points available for the 1970s. Source: Deininger and Squire (1996). Average years of schooling of population over age 25 in 1960. Source: Barro and Lee (2000). Distance from the equator. Absolute value of latitude in degrees divided by 90. Source: Hall and Jones (1999). Fraction of a countrys population speaking one of the five primary Western European languages (English, French, German, Portuguese, and Spanish) as a mother tongue. Source: Hall and Jones (1999). Fertility rate (births per woman), average of 1972 and 1987 values of this variable. Source: World Bank (2005). Log of standard deviation of the first log differences in real general government final expenditures in 1970-2000. Source: World Bank (2005). Log of standard deviation of the residuals from time-series regression of real general government expenditure growth. Source: Woo (2005b). Government Anti-Diversion Policy indicator. It measures on a 0-1 scale how effective government policies and institutions are in supporting production over diversion. It is the average of the following 5 categories for the year of 1986-1995: (i) law and order, (ii) bureaucratic quality, (iii) corruption, (iv) risk of expropriation, and (v) risk of government repudiation of contracts. A high value signifies more effective government policies and institutions. Source: Hall and Jones (1999). The original source of the data is International Country Risk Guide (ICRGE) data by Political Risk Services that rates 130 countries according 24 categories. Average compound growth rate of real GDP per capita, 19702000. Source: Heston et al. (2002). Average compound growth rate of TFP, 1970-2000. Source: Woo (2004). Log of population in 1970. Source: Heston et al. (2002). Continued on next page
148
6 Growth, Income Inequality, and Fiscal Volatility: Empirical Evidence
Table 6.10 — continued from previous page Variables Description and Source Log of real GDP per capita in 1970. Source: Heston et LRGDPCH70 al. (2002). Log of average years of secondary schooling in the populLSYR1570 ation over age 15 in 1970. Source: Barro and Lee (2000). Log of T F P relative to USA in 1970. Source: Woo (2004). LTFP70 Sociopolitical instability indicator. PINSTAB PINSTAB = 0.59COUPS -h 0.52CONSTCHG + 0.17GOVTCRIS + 0.25ASSASSIN + 0.54REVOLS. Original data source: Banks (2003). It measures the extent to which the executives face poliPOLCON tical constraints in implementing his or her policy. It is based on the number of institutionally embedded veto players among various branches of government. A high value indicates greater political constraints. Source: Henisz (2002). PRIMCOMP1960 Percentage of "primary school complete" in the total population over age 25 in 1960. Source: Barro and Lee (2000). TRADE The sum of exports and imports as a share of GDP, averaged over the period of 1970-2000. World Bank (2005).
6.7 Data Appendix T a b l e 6 . 1 1 . Summary Statistics
Variable AGEXP AGINI60 AGINI70 AYS1960 DISTANCE EURFRAC FERT FISVOLl FISV0L2 GADP GRGDPCH7000 GRTFP7000 LPOP70 LRGDPCH70 LSYR1570 LTFP70 PINSTAB POLCON PRIMCOMP1960 TRADE
Number of Standard observations Mean deviation Minimum Maximum 91 57 62 85 93 93 92 91 91 92 93 93 93 93 93 93 82 91 85 93
15.214 5.232 44.975 8.616 41.784 8.683 2.533 3.469 0.272 0.186 0.42 0.32 4.4105 1.907 -2.469 0.728 -2.671 0.784 0.635 0.206 0.02 0.017 0.002 0.014 8.914 1.599 8.104 0.989 -0.592 1.252 -0.685 0.525 1.486 0.000 0.38 0.288 17.144 14.287 0.665 0.435
6.358 30.68 28.1 0.07 0.003 0 1.57 -4.183 -4.529 0.225 -0.048 -0.051 5.318 5.804 -4.269 -2.824 -1.786 0 0.1 0.159
33.739 62.8 65.38 9.56 0.71 1.004 8 -0.692 -0.778 1 0.067 0.035 13.615 9.934 1.475 0.191 3.548 0.885 67.1 3.296
149
References
1. Acemoglu, D., Johnson, S., Robinson, J. (2001) The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review 91: 1369-1401 2. Adams, C , Bevan, D. (2005) Fiscal Deficits and Growth in Developing Countries. Journal of Public Economics 89: 571-597 3. Ades, A., Glaeser, E. (1999) Evidence on Growth, Increasing Returns, and the Extent of the Market. Quarterly Journal of Economics 114: 1025-1045 4. Agell, J. (1999) On the Benefits From Rigid Labor Markets: Norms, Market Failures, and Social Insurance. Economic Journal 109: 143-164 5. Agenor, R, Montiel, P. (1999) Development Macroeconomics. Princeton University Press, Princeton 6. Aizenman, J. (1990) Competitive Externalities and the Optimal Seigniorage Segmentation. Journal of Money, Credit and Banking 24: 61-71 7. Alesina, A., Drazen, A. (1991) Why Are Stabilizations Delayed? American Economic Review 81: 1170-1188 8. Alesina, A., Hausmann, R., Hommes, R., Stein, E. (1999) Budget Institutions and Fiscal Performance in Latin America. Journal of Development Economics 59: 253-273 9. Alesina, A., Ozler, S., Roubini, N., Swagel, P. (1996) Political Instability and Economic Growth. Journal of Economic Growth 1: 193-215 10. Alesina, A., Perotti, R. (1995) The Political Economy of Budget Deficits. IMF Staff Papers 42: 1-31 11. Alesina, A., Perotti, R. (1996a) Budget Deficits and Budget Institutions. NBER Working Paper, No. 5556 12. Alesina, A., Perotti, R. (1996b) Income Distribution. Political Instability and Investment. European Economic Review 40: 1203-1228 13. Alesina, A., Rodrik, D. (1994) Distributive Politics and Economic Growth. Quarterly Journal of Economics 109: 465-490 14. Alesina, A., Roubini, N., Cohen, G. (1997) Political Cycles and the Macroeconomy. The MIT Press, Cambridge 15. Alesina, A., Tabellini, G. (1990) A Positive Theory of Fiscal Deficits and Government Debt. Review of Economic Studies 57: 403-414 16. Alesina, A., Wacziarg, R. (1998) Openness, Country Size, and Government. Journal of Public Economics 69: 305-321
152
References
17. Alfaro, L. (2005) Inflation, Openness, and Exchange-Rate Regimes: The Quest for Short-Term Commitment. Journal of Development Economics 77: 229-249 18. Allen, M. (2003) Some Lessons from the Argentine Crisis: A Fund Staff View. In: Teunissen, J.J., Akkerman, A. (eds) The Crisis That Was Not Prevented: Argentina, the IMF, and Globalization. FONDAD, The Hague, pp 120-150 19. Al-Marhubi, F. (1997) A note on the link between income inequality and inflation. Economics Letters 55: 317-319 20. Azariadis, C , Drazen, A. (1990) Threshold Externalities in Economic Development. Quarterly Journal of Economics 105: 501-526 21. Banks, A.S. (1997) Cross-National Time Series Data Archives. State University of New York at Binghamton. 22. Banks, A.S. (2003) Cross-National Time Series Data Archives. Center for Social Analysis. State University of New York at Binghamton 23. Barro, R. (1979) On the Determination of Pubhc Debt. Journal of Political Economy 87: 240-271 24. Barro, R. (1991) Economic Growth in a Cross Section of Countries. Quarterly Journal of Economics 106: 407-443 25. Barro, R. (2000) Inequality and Growth in a Panel of Countries. Journal of Economic Growth 5: 5-32 26. Barro, R., Lee, J. (2000) International Data on Educational Attainment, Updates and Implications. Center for International Development Working Paper 42, Harvard University 27. Barro, R., Sala-i-Martin, X. (1992) Convergence. Journal of Political Economy 100: 223-251 28. Barro, R., Sala-i-Martin, X. (1995) Economic Growth. McGraw-Hill, New York 29. Beetsma, R.M.W., Van Der Ploeg, F. (1996) Does inequality cause inflation? The political economy of inflation, taxation and government debt. Public Choice 87: 143-162 30. Benabou, R. (1996) Inequality and Growth. In: Bernanke, B., Rotemberg, J. (eds) NBER Macroeconomic Annual. The MIT Press, Cambridge, pp 11-74 31. Benabou, R., Ok, E. (2001) Social Mobility and the Demand for Redistribution. Quarterly Journal of Economics 116: 447-487. 32. Berg, A., Sachs, J.D. (1988) The Debt Crisis: Structural Explanations of Country Performance. Journal of Development Economics 29: 271-306 33. Bernheim, B.D., Ray, D. (1989) Markov Perfect Equihbria in Altruistic Growth Economies with Production Uncertainty. Journal of Economic Theory 47: 195202 34. Berthelemy, J.C., Bourguignon, F. (1996) Growth and Crisis in Cote d'lvoire. World Bank, Washington DC 35. Birdsall, N., Ross, D., Sabot, R.H. (1995) Inequality and Growth Reconsidered: Lessons from East Asia. The World Bank Economic Review 9: 477-508 36. Blanchard, O. (2004) Fiscal Dominion and Inflation Targetting: Lessons From Brazil. Department of Economics Working Paper No. 04-13, MIT 37. Bosworth, B., Collins, S. (2003) The Empirics of Growth: An Update. Brookings Papers on Economic Activity (2): 113-206 38. Caballero, R., Krishnamurthy, A. (2004) Fiscal PoHcy and Financial Depth. NBER Working Paper No. 10532 39. Calvo, G. (1985) Macroeconomic Implications of the Government Budget: Some Basic Considerations. Journal of Monetary Economics 15: 95-112
References
153
40. Campos, N., Nugent, J. (2002) Who Is Afraid of Political Instability? Journal of Development Economics 67: 157-172 41. Catao, L., Terrones, M. (2003) Fiscal Deficits and Inflation. Working Papers 03/65, International Monetary Fund 42. Chari, V.V., Cole, H. (1993a) A Contribution to Theory of Pork Barrel Spending. Staff Report, No. 156, Federal Reserve Bank of Minneapolis 43. Chari, V.V., Cole, H. (1993b) Why Are Representative Democracies Fiscally Irresponsible? Staff Report, No. 163, Federal Reserve Bank of Minneapolis 44. Clarke, G. (1995) More Evidence on Income Distribution and Growth. Journal of Development Economics 47: 403-427 45. Cohen, D., Michel, P. (1988) How Should Control Theory Be Used to Calculate a Time Consistent Government Policy? Review of Economic Studies 55: 263274 46. Cohen, D., Soto, M. (2001) Growth and Human Capital: Good Data, Good Results. Technical paper 179, OECD Development Center 47. Cox, G. (1997) Making Votes Count. Cambridge University Press, New York 48. Cukierman, A., Edwards, S., Tabellini, G. (1992) Seigniorage and Political Instability. American Economic Review 82: 537-555 49. De Gregorio, J., Lee, J. (2002) Education and Income Inequality: New Evidence From Country Data. Review of Income and Wealth 48 (3): 395-416 50. De Haan J., Sturm, J. (1994) Political and Institutional Determinants of Fiscal PoHcy in the European Community. PubHc Choice 80: 157-172 51. De Haan J., Sturm, J. (1997) Political and Economic Determinants of OECD Budget Deficits and Government Expenditures: A Reinvestigation. European Journal of Political Economy 13: 739-750 52. Deininger, K., Squire, L. (1996) A New Data Set Measuring Income Inequality. The World Bank Economic Review 10: 565-591 53. Desai, R., Olofsgard, A., Yousef, T. (2005) Inflation and inequality: does political structure matter? Economics Letters 87: 41-46 54. Dornbusch, R., Sturzenegger, F., Wolf, H. (1990) Extreme Inflation: Dynamics and Stabilization. Brookings Papers on Economic Activity 2: 1-84 55. Drazen, A. (2000) Political Economy in Macroeconomics. Princeton University Press, Princeton 56. Drazen, A., Helpman, E. (1990) Inflationary Consequences of Anticipated Macroeconomic Policies. Review of Economic Studies 57: 147-164 57. Durlauf, S., Johnson, P., Temple, J. (2004) Growth Econometrics. In: Aghion, P., Durlauf, S. (eds) Handbook of Economic Growth. North-Holland, Amsterdam (forthcoming) 58. Easterly, W. (1999) When Is Fiscal Adjustment An Illusion? Economic Policy 14: 55-86 59. Easterly, W. (2001) The Elusive Quest for Growth. The MIT Press, Cambridge, pp 255-283 60. Easterly, W. (2002) Does Cause Underdevelopment. Working Paper No. 1, Center for Global Development 61. Easterly, W. (2004) National Policies and Economic Growth: A Reappraisal. In: Aghion, P., Durlauf, S. (eds) Handbook of Economic Growth. North-Holland, Amsterdam (forthcoming) 62. Easterly, W., Levine, R. (1997) Africa's Growth Tragedy: Policies and Ethnic Divisions. Quarterly Journal of Economics 112: 1203-1250
154
References
63. Easterly, W., Rodriguez, C , Schmidt-Hebbel, K. (1994) Public Sector Deficits and Macroeconomic Performance. Oxford University Press, New York 64. Easterly, W., Yu, H. (1999) Global Development Network Growth Database. World Bank, Washington DC 65. Edin, P., Ohlsson, H. (1991) Political Determinants of Budget Deficits: Coalition Eff"ects versus Minority Effects. European Economic Review 35: 1597-1603 66. Edwards, S., Tabellini, G. (1991) Explaining Fiscal Policies and Inflation in Developing Countries. Journal of International Money and Finance 10: S16-48. 67. Engerman, S., Sokoloff, K. (1997) Factor Endowments, Institutions, and Differential Paths of Growth Among New World Economies: A View from Economic Historians of the United States. In: Haber, S. (ed) How Latin America Fell Behind. Stanford University Press, Stanford, pp 260-306 68. Fatas, A., Mihov, I. (2001) Government Size and Automatic Stabilizers: International and Intranational Evidence. Journal of International Economics 55:3-28 69. Fields, G.S. (1994) Changing Labor Market Conditions and Economic Development in Hong Kong, the Republic of Korea, Singapore, and Taiwan, China. The World Bank Economic Review 8: 395-414 70. Fields, G.S., Jakubson, G.H. (1994) New Evidence on the Kuznets Curve. Unpublished Working Paper, Cornell University 71. Fischer, S. (1993) The Role of Macroeconomic Factors in Growth. Journal of Monetary Economics 32: 485-512 72. Fischer, S,, Easterly, W. (1990) The Economics of the Government Budget Constraint. World Bank Research Observer 5: 127-142 73. Fischer, S., Sahay, R., Vegh, C. (2002) Modern Hyper- and High Inflations. Journal of Economic Literature 40: 176-187 74. Forbes, K. (2000) A Reassessment of the Relationship between Inequality and Growth. American Economic Review 90: 869-887 75. Frankel, J., Romer, D. (1999) Does Trade Cause Growth? American Economic Review 89 (3): 379-399. 76. Frankel, J., Rose, A. (2002) An Estimate of the Effect of Common Currencies on Trade and Income. Quarterly Journal of Economics 117 (2): 437-466 77. Franzese, R.J. (1999) Electoral and Partisan Manipulation of Public Debt in Developed Democracies, 1956-90. In Strauch, R.R., von Hagen, J. (eds) Institutions, Politics and Fiscal Policy. Kluwer Academic Publishers, Boston, pp 61-83 78. Fudenberg, D., Tirole, J. (1992) Game Theory. The MIT Press, Cambridge 79. Gali, J. (1994) Government Size and Macroeconomic Stability. European Economic Review 38(1): 117-132 80. Galor, O., Zeira, J. (1993) Income Distribution and Macroeconomics. Review of Economic Studies 60: 35-52 81. Gavin, M., Perotti, R. (1997) Fiscal Policy in Latin America. In: Bernanke, B.S., Rotemberg, J.J. (eds) NBER Macroeconomics Annual 1997. The MIT Press, Cambridge, pp 11-61 82. Gelb, A., Knight, J.B., Sabot, R.H. (1991) Public Sector Employment, Rent Seeking and Economic Growth. The Economic Journal 101: 1186-1199 83. Gillis, M., Perkins, D., Roemer, M., Snodgrass, D. (1996) Economics of Development. W.W. Norton & Company, New York 84. Glaeser, E., La Porta, R., Lopez-de-Silanes, F., Shleifer, A. (2004) Do Institutions Cause Growth? Journal of Economic Growth 9: 271-303
References
155
85. Goldin, C , Katz, L. (1998) The Origins of Technology-Skill Complementarity. Quarterly Journal of Economics 113: 693-732 86. Gosh, A., Guide, A., Wolf, H. (2002) Exchange Rate Regimes: Choice and Consequences. The MIT Press, Cambridge 87. Grilh, v . , Masciandaro, D., Tabellini, G. (1991) Political and Monetary Institutions and Public Financial Pohcies in the Industrial Countries. Economic PoHcy 13: 341-392 88. Grossman, G., Helpman, E. (1991) Innovation and Growth in the Global Economy. The MIT Press, Cambridge 89. Hall, R., Jones, C. (1999) Why Do Some Countries Produce So Much More Output per Worker than Others? Quarterly Journal of Economics 114: 83-116 90. Hallerberg, M., von Hagen, J. (1999) Electoral Institutions, Cabinet Negotiations, and Budget Deficits in the European Union. In: Poterba, J., von Hagen, J. (eds) Fiscal Institutions and Fiscal Performance. The University of Chicago Press, Chicago, pp 209-232 91. Hausmann, R., Velasco, A. (2002) Hard Money's Soft Underbelly: Understanding the Argentine Crisis, Working Paper, Harvard University 92. Henisz, W. (2002) The Institutional Environment for Infrastructure Investment. Industrial and Corporate Change 11(2): 355-389 93. Heston, A., Summers, R., Aten, B. (2002) Penn World Table Version 6.1. Center for International Comparisons at the University of Pennsylvania (CICUP) 94. IMF (1999) World Economic Outlook. May Issue, IMF (International Monetary Fund), Washington DC, pp 27-49 95. IMF (Various issues) Government Finance Statistics Yearbook. IMF, Washington DC 96. Inter-American Development Bank (1998) Facing Up to Inequality. InterAmerican Development Bank, Washington DC 97. Jaggers, K., Gurr, T.R. (1996) Pohty HI: Regime Change and Political Authority. Center for International Development and Conflict Management, University of Maryland 98. Jain, S. (1975) Size Distribution of Income: A Comparison of Data. Unpublished Manuscript, The World Bank 99. Jones, C. (1999) Growth: With or Without Scale Effects? American Economic Review Papers and Proceedings 89: 139-144 100. Kauffman, R., Stallings, B. (1991) The Political Economy of Latin American Populism. In: Dornbusch, R., Edwards, S. (eds) The Macroeconomics of Populism in Latin America. The University of Chicago Press, Chicago, pp 15-34 101. Klein, M., Marion, N. (1994) Explaining the Duration of Exchange Rate Pegs. NBER Working Paper, No. 4651 102. Knack, S., Keefer, P. (1995) Institutions and Economic Performance: CrossCountry Tests Using Alternative Institutional Measures. Economics and Politics 7: 207-227 103. Kontopoulos, Y., Perotti, R. (1999) Government Fragmentation and Fiscal Pohcy Outcomes: Evidence from OECD Countries. In: Poterba, J., von Hagen, J. (eds) Fiscal Institutions and Fiscal Performance. The University of Chicago Press, Chicago, pp 81-102 104. Lao-Araya, K. (1998) Budget Processes in Selected Asian Countries. Unpublished Working Paper, Indiana University 105. La Porta, R., Lopez-de-Silances, F., Shleifer, A., Vishny, R. (1999) The Quality of Government. Journal of Law, Economics and Organization 15: 222-279
156
References
106. Learner, E. (1985) Sensitivity Analyses Would Help. American Economic Review 75: 308-313 107. Lecallion, J., Paukert, F., Morrison, C , Gemiolis, D. (1984) Income Distribution and Economic Development: Analytical Survey. International Labor Office 108. Lee, N., Ha, J., Kang, T., Lim, C. (2004) The Causes of Economic Polarization and Their Policy Implications (in Korean). The Institute for Monetary and Economic Research Working Paper, No. 184, Bank of Korea 109. Leininger, W. (1986) The Existence of Perfect Equilibria in a Model of Growth with Altruism between Generations. Review of Economic Studies 53: 349-367 110. Levhari, D., Mirman, L. (1980) The Great Fish War: An Example Using a Dynamic Cournot-Nash Solution. Bell Journal of Economics 11: 322-334 111. Levine, R., Renelt, D. (1992) A Sensitivity Analysis of Cross-Country Growth Regressions. American Economic Review 82: 942-963 112. Liviatan, N. (1983) Inflation and the Composition of Deficit Finance. In: Adams, F.G., Hickman , B.C. (eds) Global Econometrics: Essays in Honor of Lawrence R. Klein. The MIT Press, Cambridge, pp 84-100 113. Liviatan, N. (1988) On the Interaction between Monetary and Fiscal Pohcies under Perfect Foresight. Oxford Economic Papers 40: 193-203 114. Lundberg, M., Squire, L. (2001) The Simultaneous Evolution of Growth and Inequality. Working Paper, The World Bank 115. Mackenzie, G.A., Orsmond, D., Gerson, P. (1997) The Composition of Fiscal Adjustment and Growth: Lessons from Fiscal Reforms in Eight Economies. IMF Occasional Paper 149, IMF 116. Mainwaring, S. (1993) Presidentialism, Multipartism, and Democracy: The Difficult Combination. Comparative Political Studies 26: 198-228 117. Mankiw, N.G., Romer, D., Weil, D. (1992) A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics 107: 407-437 118. Meltzer, A., Richards, S. (1981) A Rational Theory of the Size of Government. Journal of Political Economy 89: 914-927 119. Missale, A., Blanchard, O. (1994) The Debt Burden and Debt Structure. American Economic Review 84: 309-319 120. OECD (2004) Economic Outlook. OECD (Organisation for Economic Cooperation and Development), Paris 121. OECD (2005a) Economic Survey of The Euro Area. OECD (Organisation for Economic Cooperation and Development), Paris 122. OECD (2005b) Employment Outlook. OECD (Organisation for Economic Cooperation and Development), Paris 123. Patinkin, D. (1993) Israel's Stabilization Program of 1985, Or Some Simple Truths of Monetary Theory. Journal of Economic Perspectives 7: 103-128 124. Perotti, R. (1993) Political Equilibrium, Income Distribution, and Growth. Review of Economic Studies 60: 755-776 125. Perotti, R. (1996) Growth, Income Distribution, and Democracy: What the Data Say? Journal of Economic Growth 1: 149-187 126. Persson, T., Roland, G., TabeUini, G. (1997) Separation of Powers and Political Accountability. Quarterly Journal of Economics 112 (4): 1163-1202 127. Persson, T., Svensson, L.E.O. (1989) Why a Stubborn Conservative Would Run a Deficit: Policy with Time-Inconsistent Preferences, Quarterly Journal of Economics 104: 325-345 128. Persson T., TabeUini, G. (1994) Is Inequality Harmful for Growth?: Theory and Evidence. American Economic Review 84: 600-621
References
157
129. Persson, T., Tabellini, G. (1997) Political Economy and Macroeconomic Policy. NBER Working Paper, No. 6329 130. Persson, T., Tabellini, G. (1999a) Political Economics and Macroeconomic Policy. In: Taylor, J., Woodford, M. (eds) Handbook of Macroeconomics. Elsevier Science B.V., Amsterdam, pp 1397-1482 131. Persson, T., Tabellini, G. (1999b) The Size and Scope of Government. European Economic Review 43: 699-735 132. Poterba, J., von Hagen, J. (1999) Fiscal Institutions and Fiscal Performance. Chicago: The University of Chicago Press, Chicago 133. Powell, B.G. (1982) Contemporary Democracies: Participation, Stability, and Violence. Harvard University Press, Cambridge 134. Psacharopoulos, G. (1985) Returns to Education: A Further International Updated and Implications. Journal of Human Resources 20: 583-604 135. Radelet, S., Sachs, J.D. (1998) The Onset of the East Asian Financial Crisis. NBER Working Paper, No. 6680 136. Rodden, J., Eskeland, G., Litvack, J. (2003) Fiscal Decentralization and the Challenge of Hard Budget Constraints. The MIT Press, Cambridge 137. Rodrik, D. (1996) Understanding Economic Policy Reform. Journal of Economic Literature 34: 9-41 138. Rodrik, D. (1997) Has Globalization Gone Too Far? Institute for International Economics, Washington DC 139. Rodrik, D. (1998) Why Do More Open Economies Have Bigger Governments. Journal of Political Economy 106: 997-1032 140. Rodrik, D. (1999) Where Did All the Growth Go? External Shocks, Social Conflict and Growth Collapses. Journal of Economic Growth 4: 385-412 141. Rodrik, D. (2005) Why We Learn Nothing from Regressing Economic Growth on Policies. Working Paper, Harvard University 142. Roubini, N. (1991) Economic and Political Determinants of Budget Deficits in Developing Countries. Journal of International Money and Finance 10: 49-72 143. Roubini, N., Sachs, J.D. (1989a) Political and Economic Determinants of Budget Deficits in the Industrial Democracies. European Economic Review 33: 903933 144. Roubini, N., Sachs, J.D. (1989b) Government Spending and Budget Deficits in the Industrialized Countries. Economic Policy 8: 99-132 145. Rousseeuw, P.J. (1984) Least Median of Squares Regression. Journal of the American Statistical Association 79: 871-880 146. Rousseeuw, P.J., Leroy, A.M. (1987) Robust Regression and Outlier Detection. John Wiley Sons, New York 147. Sachs, J.D. (1989) Social Conflict and Populist Policies in Latin America. NBER Working Paper, No. 2897 148. Sala-i-Martin, X. (1994) Cross-sectional Regressions and the Empirics of Economic Growth. European Economic Review 38: 739-747 149. Sala-i-Martin, X. (1997) I Just Ran Two Million Regressions. American Economic Review Papers and Proceedings: 178-183 150. Sargent, T., Wallace, N. (1981) Some Unpleasant Monetarist Arithmetic. Quarterly Review of the Federal Reserve Bank of Minneapolis 5: 1-17 151. Shugart, M.S., Carey, J. (1992) Presidents and Assemblies. Cambridge University Press, New York
158
References
152. Summers, R., Heston, A. (1988) A New Set of International Comparisons of Real Product and Price Levels Estimates for BO countries, 1950-85. Review of Income and Wealth 34: 1-25 153. Talvi, E., Vegh, C. (1996) Can Optimal Fiscal Policy Be Procyclical? Working Paper, U.C.L.A. 154. Taylor, C.L., Hudson, M.C. (1972) World Handbook of Political and Social Indicators. Yale University Press, New Haven 155. Tarantelli, E. (1986) Economia Politica del Lavoro. UTET, Torino 156. Temple, J. (1999) The New Growth Evidence. Journal of Economic Literature 37 (1): 112-156 157. Tornell, A., Lane, P. (1998) Are Windfalls a Curse? A Non-representative Agent Model of the Current Account and Fiscal Policy. Journal of International Economics 44: 83-112 158. Tornell, A., Lane, P. (1999) The Voracity Effect. American Economic Review 89: 22-46 159. Tornell, A., Velasco, A. (2000) Fixed vs. Flexible Exchange Rates: Which Provides More Fiscal Discipline. Journal of Monetary Economics 45: 399-436 160. Tsebelis, G. (1995) Decision Making in Political Systems: Veto Players in Presidentialism. Parliamentarism, Multicameralism and Multipartyism. British Journal of Political Science 25: 289-326 161. Turnovsky, S. (1978) Macroeconomic Dynamics and Growth in a Monetary Economy: A Synthesis. Journal of Money, Credit, and Banking 10: 1-26 162. Velasco, A. (1999) A Model of Endogenous Fiscal Deficits and Delayed Fiscal Reforms. In: Poterba, J., von Hagen, J. (eds) Fiscal Institutions and Fiscal Performance. The University of Chicago Press, Chicago, pp 37-57 163. von Hagen, J. (1992) Budgeting Procedures and Fiscal Performance in the European Communities. Economic Papers, No. 96, Commission of the European Communities 164. von Hagen, J., Harden, I. (1994) National Budget Processes and Fiscal Performance. European Economy, Reports and Studies 3: 311-418 165. Weingast, B.R., Shepsle, K.A., Johnsen, C. (1981) The Political Economy of Benefits and Costs: A New Approach to Distributive Politics. Journal of Political Economy 89: 642-664 166. White, H. (1980) A Heteroskedasticity-Consistent Covariance Matrix Estimator and A Direct Test for Heteroskedasticity. Econometrica 48: 817-838 167. Williamson, J. (1998) Part II. Globalization and the Labor Market: Using History to Inform Policy. In: Aghion, P., Williamson, J. Growth, Inequality and Globalization: Theory, History, and Policy. Cambridge University Press, Cambridge 168. Woo, J. (2003a) Economic, Political, and Institutional Determinants of Public Deficits. Journal of Pubhc Economics 87: 387-426 169. Woo, J. (2003b) Social Polarization, Industrialization, and Fiscal Instability: Theory and Evidence. Journal of Development Economics 72: 223-252 170. Woo, J. (2004) Growth, Government PoHcies and Institutions in the Globahzed World Economy. In: Hahm, J. (ed) Sustaining Korean Economic Growth A Way Forward: Augmenting Growth Potential and Upgrading the Financial System. The Bank of Korea, Seoul, pp 3-105 171. Woo, J. (2005a) Social Polarization, Fiscal Instability, and Growth. European Economic Review 49: 1451-1477
References
159
172. Woo, J. (2005b) The Behavior of Fiscal Policy: Cyclicality and Discretionary Fiscal Decisions. Working paper, DePaul University 173. Woo, S. (2005) Solution for Social Polarization? (in Korean). Weekly Dong-A 495 (March 15): 90 174. World Bank (1993) The East Asian Miracle: Economic Growth and Pubhc Policy. Oxford University Press, New York 175. World Bank (1995) Bureaucrats in Business: The Economics and Politics of Government Ownership. Oxford University Press, New York 176. World Bank (1997) World Development Indicators CD-ROM. World Bank, Washington DC 177. World Bank (1999) World Development Indicators CD-ROM. World Bank, Washington DC 178. World Bank (2000) World Development Indicators CD-ROM. World Bank, Washington DC 179. World Bank (2005) World Development Indicators CD-ROM. World Bank, Washington DC
Subject Index
Africa, sub-Saharan East Asia/Latin America compared fiscal balance/volatility, 47, 4St, 72-73 social welfare spending, 62t, 63 subsidies and transfers, 62t, 63 fiscal deficit in, 47,48/, 72-73, 77 fiscal policies, 2,47 income inequality in, results of, 3, 61-62 public vs. private sector employment, 63 altruism model with polarization, 58-63 Argentina, fiscal deficit in, 2nl B bequest games, 59nl9 bond-financing, 37-44 Brazil, 2,47, 77 budget, balanced capital stock, effect on, 28-30, 29/ as social optimum, 23-24, 27 budgetary procedures and fiscal outcomes, 32, 96-100, 98/
cabinet size and deficit reduction, 79, 88 capital accumulation, 15-16, 27-31, 29/ 3 1 / See also human capital accumulation common pool problem, 11-12, 88
conflict-solving in institutions, 68,73,96, 99n37, 133 consumption in endogenous growth model, 15 polarization increase, impact on, 28-29 coordination failure, 3,7-8,11,32, 88n20 Coted'Ivoire, 62, 77nl country size in inequality-growth regression, 120-129, 122/, 127/, 130/ D data, description and sources of, 75-76, 108-111,146-149 debt, neoclassical theory of, 77 debt accumulation-capital accumulation link, 27-31 deficit bias with political uncertainty, 12n7,24-25, 60n20 deficit finance. See also fiscal deficits inflationary consequences of, 35-36, 39^4 money-bond ratio and degree of polarization, 37-44 democratic countries income inequality-inflation correlation in, 36 as institution of conflict management, 68, 99n37 developing countries. See also specific countries fiscal deficits in, 35-36,47,77,83nl2, 84,100
162
Subject Index
fiscal volatility in, 47, 131 wage differentials by educational levels, 61-62 differential game, 17 discount rate in endogenous growth model, 14, 15-16 political uncertainty and, 9-10,24-25, 30, 60n20 social planner's, 24, 27 E East Asia. See also specific countries financial crisis 1997, causes of, 47nl fiscal balance/volatility, 47,48/, 72-73 fiscal policies, 2, 2n3, 47 income inequality, results of, 3,48-49 Latin America/sub-Saharan Africa compared fiscal balance/volatility, 47, 48/, 72-73 social welfare spending, 62/, 63 subsidies and transfers, 62/, 63 wage/income inequality in, 61-62 economic cleavage, source of, 3 economic growth. See growth education central government expenditures for, 63 as variable, inequality-growth regression, 120-129, 130/ wage/income inequality with, 51, 61-62 electoral law, 89-92 employment, public vs. private sector, 63 endogeneity problem, 126, 128 ethnic fragmentation, 3, 9/ 68, 94, 95/. See also social polarization euro area, fiscal consolidation efforts, 2 exchange rates, fixed and flexible, 38 extreme bounds analysis (EBA), 79, 100-103
feedback Nash equilibrium, 17,18-24,33 feedback strategy in differential games, 18nl7
fertility decision, 144 financial depth, 36, 82 fiscal decentralization, 32 fiscal deficits. See also public deficits capital stock, effect on, 28-30, 29/ causes/critical conditions for, 11,25, 49 in developing countries, 35-36,47,77, 83nl2,84, 100 in East Asia, 47n 1,48/ economic factors of, 80-84, 83/ ethnic fragmentation and, 94, 95/ fiscal volatility and, 17-27, 60n22 growth and, 27-31, 29/, 3 1 / 117 income inequality and, 12n7, 60, 72, 78-79, 92-96, 95/ inflationary consequences of, 35-36, 39-44 institutional determinants of, 96-100, 98/ in Latin America, 2nl, 47,48/, 77, 100 money-bond ratio and degree of polarization, 37-44 neoclassical theory of, 77 polarization and, 9-11,19-20, 22/ 25, 26/ 59-60, 64-69, 66/, 67/, 92-95, 100 political uncertainty and, 10-11,12n7, 24-25, 26/ 64-69, 84-87, 86/ sociopolitical instability and, 69, 95-96 in sub-Saharan Africa, 47,48/, 77, 100 fiscal deficits, variables identifying cross-country differences the benchmark framework, 80-84 data, 64, 80 economic factors, 80-84, 83/ institutional factors, 96-100, 98/ largest and smallest deficits in 1970-90,112-113 main findings, 78 polarization, 92-96, 95/ political factors, 84-92, 86/, 89/, 91/ robustness and sensitivity analysis, 100-103, 102/, 104/, 106/ fiscal discipline fixed vs. flexible exchange rates for, 38
Subject Index growth, importance to, 1-2 money-financing for, 37-38, 44-45 fiscal instability. See fiscal volatility fiscal policy. See also government spending limited altruism model with polarization, 58-63 political uncertainty and, 24-27 procyclicality of, in Latin America, 9 fiscal volatility comparison of developing regions. East Asia, Latin America, sub-Saharan Africa, 48/ fiscal deficit and, 17-27 growth and, 135-140,136/, 137/ 138/, 139/, 140/ income inequality and, 129-135, 131/ 132/, 135/ indicators of, 129 polarization and, 9, 20-22, 60-61, 69-73, 70-71/ political uncertainty and, 8,24-27, 142, 143/ sociopolitical instability and, 142,143/ fi*agmentation. See also social polarization ethnic, 3, 68, 94, 95/ government, 49, 87-89, 89/
GDP, real per capita growth regression, 120-123, 140-144 Gini coefficients, 48, 66-68, 93-94, 108-109 generalized methods of moments (GMM), 128 globalization, results of, 4 government. See also institutions globalization as cause of growth in, 4 policy factors underlying undesirable outcomes, 8 governmentfi-agmentation,49, 87-89, 89/. government size, 131, 132/, 133 government spending, procyclicality of, 22-23 types of, 53, 62-63
163
tax revenue shocks and changes in, 9-10, 21-22, 23/ 27, 50, 60-61 Greece, fiscal deficit in, 77 growth capital accumulation and, 27-31 fiscal deficit and, 27-31, 29/ 3 1 / 117 fiscal volatility and, 135-140, 136/, 137/ 138/, 139/, 140/ income inequality and, 119-129, 122/, 125/ 126/ 127/, 144-145 long-term, determinants of, 120 polarization and, 7, 9-11, 27-31,31/ 117 political uncertainty and, 9-11, 27-31, 141 sociopolitical instability and, 8, 142-144, 143/ growth regression, shortcomings, 118-119 growth theory, endogenous, 120n2 H Hamiltonian function, 19, 33 human capital accumulation. See also capital accumulation growth regression, 120-129, 122/, 127/ post-industrialization economy, 52, 54-56, 56/ pre-industrialization economy, 51 I impatience of policymakers. See also political uncertainty and discount rate income inequality. See also social polarization and Gini coefficients education and, 50n8, 61-62 fertility rate and, 144 fiscal deficits and, 12n7, 60, 72, 78-79, 92-96, 95/ fiscal volatility and, 129-135, 131/ 132/, 135/ globalization and, 4 growth's relationship to, 119-129, 122/, 125/ 126/ 127/, 144-145 industrialization and, 50-51, 55-56 inflation in democratic countries, 36
164
Subject Index
inflation tax and, 2>6,2>lf polarization and, 3-4, 57-58, 92-96 redistributive policies and, 49n4, 57-59 sociopolitical instability and, 140-142, 143^ Indonesia, fiscal deficit, 47nl industrialization defined, 49 income inequality, 50-51, 55-56 post-industrialization economy, 52-58 pre-industrialization economy, 51 industrialized countries fiscal deficits in, 77 polarization in, 4 inflation deficit finance composition, polarization, and 35-37, 39-44 income inequality and, 36, 37/ as variable in fiscal deficit regression, 64-65, 66^,81 inflafion tax, 36-37, 37/ 40-41, 43, 60n22, 81n9 institutions. See also government of conflict management, 68, 73, 96, 99n37, 133 fiscal deficit determinants, 96-100, 98/ growth, 121-129 quality indicators, 96, 123n4, 133 social planner's function, 23-24 Israel, prisoners' dilemma in coalition government, 36 Italy, fiscal deficit in, 77 K Korea, fiscal deficit in, 47, 77 Kuznets inverted-U hypothesis, 50n7
Latin America. See also specific countries East Asia/sub-Saharan Afi*ica compared fiscal balance/volatility, 47,48/, 72-73 social welfare spending, 62/, 63 subsidies and transfers, 62/, 63
fiscal balance/volafility, 2nl, 47,48/, 72-73 fiscal policies, 2, 47 polarization effects in, 32 public vs. private sector employment, 63 wage/income inequality in, 3, 50n8, 61-62 least median of squares (LMS), 79, 103, 119, 124 M macroeconomic instability, 2, 10/ Malaysia, fiscal deficit, 47nl, 51 Markov perfect equilibrium (MPE), 19, 22,59nl9 Markov strategy, 18, 58nl9 Mexico, fiscal deficit in, 47, 77nl money-bond ratio in deficit financing, 37^4 N Nigeria, fiscal deficit in, 100 no-Ponzi-game (NPG) condifion, 14, 18nl6 O open-loop strategy in differential games, 18nl7 ordinary least squares (OLS), 79, 118 outlier problem, 79, 118-119, 124
party fracfionalization index, 87, 110 Peru, fiscal deficit in, 77, 100 polarization. See social polarizafion polarization composite index, 9/ 10/ polarizafion effect, 20-22, 27-29, 36, 59-61 political factors in fiscal deficit, 77-78, 84-92, 89/, 91/. See also specific factors, e.g. political instability political instability. See political uncertainty political uncertainty. See also government fragmentation; sociopolitical instability fiscal deficit and, 10-11, 12n7, 24-25, 26/ 49, 60n20, 65-69, 84^87, 86/
Subject Index fiscal volatility and, 8, 27, 142, 143/ growth and, 9-11, 27-31, 142-144 income inequality and, 140-142 as variable in fiscal deficit regression, 65, 84 in fiscal volatility regression, 72 population growth regression, 130 inequality-growth regression, 120 volatility-growth measurement, 138/, 139/ pork-barrel problem, 11, 88 principal components analysis, 69n29, 85, 141 prisoners' dilemma, 36 public deficits. See fiscal deficits R Real Plan, Brazil, 2nl redistributive poHcies, 13, 49n4, 57-59, 63 re-election uncertainty. See political uncertainty regime type and fiscal deficit, 89-92, 91/ reweighted lest squares (RWLS) procedure, 124 Ricardian equivalence proposition, 20n21 robust estimation, 79, 103, 119, 124
seemingly unrelated regression (SUR), 100 seigniorage. See inflation tax sensitivity analysis, 79, 100-103 social optimum balanced budget as, 23-24, 27 polarization and achieving the, 20, 50, 59,61 social planner's solution, 13, 23-24, 27, 33,74 social polarization. See also fi-agmentation; income inequality collective decision-making, role in, 7-8 consumption, impact of increases in, 28-29 fiscal deficits and, 9-11, 19-20, 22/, 25, 26/ 59-60, 64-69, 66/, 67/,
165
92-95, 100 fiscal volatility and, 9, 20-22, 23/ 60-61,69-73,70-71/ globalization and, 4 growth and, 7, 9-11, 27-31, 31/117 income inequality and, 3-4, 57-58, 92-96 inflation-deficit finance composition relafionship, 35-36, 39-44 social welfare, impact on, 28-29 source of, 3-4, 49 social unrest. See political uncertainty social welfare spending, 28-29, 62/, 63, 82 sociopolifical instability. See also political uncertainty fiscal deficit and, 69, 95-96 fiscal volafility and, 142, 143/ growth and, 8, 142-144, 143/ income inequality and, 140-142, 143/ South Korea, polarization in, 4 state-owned enterprises (SOE), 63 Sweden, fiscal deficit in, 77
tax smoothing model, 78, 80 Thailand, fiscal deficit, 47nl total factor productivity (TFP), 120, 127 trade variable in cross-country differences of fiscal deficit, 83-84, 83/ in inequality-growth regression, 121, 122/, 127/, 130/ in inequality-volatility measurement, 133-135, 132/ transition dynamics, 30n29 U urbanization ratio, 69 V veto players, 92 voracity effect, 21n22 W wage gap. See also income inequality education and, 61-62 industrialization and the, 50, 51
166
Subject Index
public employment and the, 63 war and fiscal deficits, 83/, 84. See also political uncertainty Z Zambia, fiscal deficit in, 47, 77
Author Index
Acemoglu,D., 119, 126n9 Adams, C , 117 Ades,A., 121 Agell, J, 4n5 Agenor, P., 60n22 Aizenman, J., 36n2 Alesina, A , 7, 8, 11, 12n7, 13, 32, 32n30, 36n4,49n4, 60n20, 65, 73n30, 73n31, 77, 78n3, 80n6, 84nl4, 87, 87nl7, 92, 96, 96n31, 97, 97n33,99n37, 121, 140, 141 Alfaro, L., 35nl Allen, M., 2nl Al-Marhubi, F., 36 Azariadis, C, 50n6 B Banks, A.S, 69,108,109,110, 111, 141, 148 Barro, R., 14, 15, 15nlO, 30, 77, 119nl, 120, 120n3, 141,147, 148 Beetsma, R.M.W., 36, 36n4 Benabou, R, 57, 117 Berg, A., 49, 51, 78n4, 92n26, 93n27 Bemheim, B.D., 58nl8, 59nl8 Berthelemy, J.C., 62 Bevan,D., 117 Birdsall, N., 3, 50n7, 58, 61, 63n24 Blanchard, O., 38n5,41n9 Bosworth,B., 118 Bourguignon, F., 62
Caballero, R., 36, 82n 11 Calvo, G.,44nl2 Campos, N., 142nl5 Carey,J.,90,90n22, 111 Catao, L., 35 Chari,V.V, ll,12n7 Clarke, G., 61 Cohen,D., 17,39, 120n3 Cole,H., 11, 12n7 Collins, S., 118 Cox,G.,90,90n22, 110, 111 Cukierman, A., 36n2, 69 D De Gregorio, J., 128 DeHaanJ.,78n3, 87nl7, 88 Deininger, K., 9/, 36n3,48^, 50,66,93, 93n27, 108, 109, 121, 141, 147 Desai, R., 36 Dornbusch, R, 35,41 Drazen, A, 3, 41nl0,49n4, 50n6, 87, 141 Durlauf, S., 118 E Easterly, W., 2, 3, 7, 7nl, 9/ 35, 48^, 49n3, 50, 64, 80, 80n6, 81n9, 108, 109,111,117,119,123,124 Edin,P.,78n3,87nl7,88 Edwards, S., 64, 65, 69, 78n3, 84nl3, 84nl4
168
Author Index
Engerman, S., 3 F Fatas,A., 133 Fields, G.S., 50n7, 62 Fischer,S., 13,35,117 Forbes, K., 119nl Frankel,J, 119, 126, 130 Franzese, RJ., 78 Fudenberg,D., Iln3, 19nl8 G Gali,J., 131 Galor, O., 52 Gavin, M., 10,22,47n2 Gelb, A., 63 Gillis,M.,53,61 Glaeser, E., 119, 121, 123n4, 128n9 Goldin, C, 49 Gosh, A., 35nl Grim, V , 78n3, 84nl4, 87nl7, 90n21 Grossman, G., 120 Gurr, T.R.,99n37, 108 H Hall,R., 119, 121,126, 147 Hallerberg, M , 11,1 ln4,32n30,49n5, 78n3, 97n34 Harden, I., 73n30, 78n3, 96n31 Hausmann, R., 2nl, 96 Helpman, E.,41nl0, 120 Henisz, W, 123, 148 Heston,A., 110, 120, 147, 148 Hommes, R., 96n31 Hudson, M.C., 48/, 68, 94, 109
Jaggers,K., 99n37, 108 Jain, S., 142 Jakubson, G.H., 50n7 Jones, C, 119, 120n2, 121, 126, 147 K Katz, L., 49 Kauffman, R., 3, 12,49, 92n26 Keefer,P.,68,99, 109, 142nl5 Klein, M., 41 Knack, S.,68,99, 109, 142nl5
Kontopoulos, Y., 12, 81n9, 88, 96, 97n33 Krishnamurthy, A., 36, 82nl 1
Lane,P, ll,82nll Lao-Araya, K., 78n3, 96n31, 96n32, 97n34, 108, 110, 111 La Porta, R., 96, 133 Lecalhon, J., 142 Lee, J, 120, 120n3, 128, 147, 148 Lee, N., 4 Leininger, W., 58nl8, 59nl8 Leroy,A.M, 103, 104 Levhari, D., Iln3 Levine, R., 3, 79, 101, 109, 111,117, 118 Liviatan, N., 37,44, 44nl2 Lundberg, M., 128 M Mackenzie, G.A., 63n25 Mainwaring, S., 90, 90n22, 111 Mankiw, N.G., 30n29 Marion, N., 4In 10 Meltzer, A., 57nl6 Michel,P., 17nl4,39 Mihov, L, 133 Mirman, L., Iln3 Montiel, P., 60n22 N Nugent, J., 142nl5 O Ohlsson, H., 78n3, 87nl7, 88 Ok, E., 57
Patinkin, D., 36 Perotti, R., 3, 8, 10, 11, 12,13, 23, 32, 47n2, 50n6, 73n31, 74n31, 77, 78n3, 81n9, 87, 88, 96, 96n31, 97n33, 117, 118, 123n5, 140, 141, 141nl3, 142 Persson,T,3, 11,13,49n4, 73n31, 78, 84nl4, 89, 90, 90n22, 91n23, 93, 99n37, 110, 111,123 Poterba,!, 96n31
Author Index
69
Powell, B.G, 3, 90n22 Psacharopoulos, G., 61
Tumovsky, S., 37,44
R Radelet, S., 2n3,4nl Ray,D.,58nl8,59nl8 Renelt,D, 79, 101,118 Richards, S., 57 Rodden, J.,32n31 Rodrik, D , 3,4, 7, 12, 13, 36n4,49, 49n4, 73n31, 92, 92n26, 99, 99n37, 117,119,133 Romer, D., 119, 126 Rose, A., 126, 130 Roubini, N., 64, 65, 78n3, 80n6, 81, 84nl4, 87nl7, 88 Rousseeuw, P.J., 79, 103, 119, 124
Van Der Ploeg, F., 36, 36n4 Vegh, C, 23n24 Velasco, A., 2nl, 11, 1 ln4, 38, 49n5, 88 von Hagen, J., 11, 1 ln4, 32n30,49n5, 73n30, 78n3, 96, 96n31, 96n32, 97n34, 108, 110, 111
W Wacziarg, R., 121 Weingast, B.R., 11,1 ln4, 49n5, 88 White, H., 66, 82, 83, 89,91,121,128, 130 Williamson, J., 4 Woo, J., 3, 7, 10, 12n7, 13, 32,47, 77, 120,129, 131nl2, 147, 148 Sachs, J.D., 2n3, 3, 3n4, 36,47nl, 49,Woo, S., 4 51nl0, 78n3, 78n4, 80n6, 87nl7, 88, 92n26, 93n27 Sala-i-Martin, X , 14,15, 30, 79, 101, 101n38,118, 141 Shugart,M.S,90n22, 111 Z Sokoloff, K., 3 Zeira, J., 52 Soto,M., 120n3 Squire, L., 9, 36, 48^, 50, 66, 93, 93n27, 108, 109, 121, 128, 142, 147 Stallings, B., 3, 12n6,49, 92n26 Stein, E.,96n31 Sturm, J., 78n3, 87nl7, 88 Summers, R., 110 Svensson, L.E.O., 84
Tabellini, G, 3, 11, 12, 12n7,49n4, 60, 64, 65, 69, 73n31, 78, 78n3, 89, 90, 90n22, 91n23, 93, 99n37, 110, 111,84.13,84.14 Talvi, E., 23n24 Tarantelli, E., 3n4 Taylor, C.L., 48/, 68, 94, 109 Temple,!., 118 Terrones, M., 35 Tomell, A., 11, 1 ln4, 21n22, 38, 49n5 Tsebelis, G., 92
Lecture Notes in Economics and Mathematical Systems For information about Vols. 1-475 please contact your bookseller or Springer-Verlag Vol. 476: R. Demel, Fiscal Policy, Public Debt and the Term Structure of Interest Rates. X, 279 pages. 1999.
Vol. 498: M. Reutter, A Macroeconomic Model of West German Unemployment. X, 125 pages. 2001.
Vol. 477: M. Thkra, R. Tichatschke (Eds.), Ill-posed Variational Problems and Regularization Techniques. VTTT, 274 pages. 1999.
Vol. 499: A. Casajus, Focal Points in Framed Games. XI, 131 pages. 200 1.
Vol. 478: S. Hartmann, Project Scheduling under Limited Resources. XTT, 221 pages. 1999. Vol. 479: L. v. Thadden, Money, Inflation, and Capital Formation. TX, 192 pages. 1999. Vol. 480: M. Grazia Speranza, P. Stahly (Eds.), New Trends in Distribution Logistics. X, 336 pages. 1999. Vol. 481: V. H. Nguyen, J. J. Strodiot, P. Tossings (Eds.). Optitnation. IX, 498 pages. 2000. Vol. 482: W. B. Zhang, A Theory of International Trade. XI, 192 pages. 2000. Vol. 483: M. Kiinigstein, Equity, Efficiency and Evolutionary Stability in Bargaining Games with Joint Production. XTT, 197 pages. 2000. Vol. 484: D. D. Gatti, M. Gallegati, A. Kirman, Interaction and Market Structure. VI, 298 pages. 2000. Vol. 485: A. Garnaev, Search Games and Other Applications of Game Theory. VIII, 145 pages. 2000. Vol. 486: M. Neugart, Nonlinear Labor Market Dynamics. X, 175 pages. 2000. Vol. 487: Y. Y. Haimes, R. E. Steuer (Eds.), Research and Practice in Multiple Criteria Decision Making. XVTT, 553 pages. 2000. Vol. 488: B. Schmolck, Ommitted Variable Tests and Dynamic Specification. X, 144 pages. 2000. Vol. 489: T. Steger, Transitional Dynamics and Economic Growth in Developing Countries. VIII, 151 pages. 2000.
Vol. 500: F. Nardini, Technical Progress and Economic Growth. XVII, 191 pages. 2001. Vol. 501: M. Fleischmann, Quantitative Models for Reverse Logistics. XI, l 8 l pages. 200 1 . Vol. 502: N. Hadjisavvas, J. E. Martinez-Legaz, J.-P. Penot (Eds.), Generalized Convexity and Generalized Monotonicity. TX, 41 0 pages. 2001. Vol. 503: A. Kirman, J.-B. Zimmermann (Eds.), Economics with Heterogenous Interacting Agents. VII, 343 pages. 200 1. Vol. 504: P.-Y. Moix (Ed.), The Measurement of Market Risk. XI, 272 pages. 2001. Vol. 505: S. Vofi, J. R. Daduna (Eds.), Computer-Aided Scheduling of Public Transport. XI, 466 pages. 2001. Vol. 506: B. P. Kellerhals, Financial Pricing Models in Con-tinuous Time and Kalman Filtering. XTV, 247 pages. 2001. Vol. 507: M. Koksalan, S. Zionts, Multiple Criteria Decision Making in the New Millenium. XII, 48 1 pages. 200 1. Vol. 508: K. Neumann, C. Schwindt, J. Zimmermann, Project Scheduling with Time Windows and Scarce Resources. XI, 335 pages. 2002. Vol. 509: D. Hornung, Investment, R&D, and Long-Run Growth. XVT, 194 pages. 2002. Vol. 510: A. S. Tangian, Constructing and Applying Objective Functions. XII, 582 pages. 2002.
Vol. 490: S. Minner, Strategic Safety Stocks in Supply Chains. XI, 214 pages. 2000.
Vol. 51 1: M. Kiilpmann, Stock Market Overreaction and Fundamental Valuation. IX, l98 pages. 2002.
Vol. 49 1 : M. Ehrgott, Multicriteria Optimization. VTTT, 242 pages. 2000.
Vol. 512: W.-B. Zhang, An Economic Theory of Cities.XI, 220 pages. 2002.
Vol. 492: T. Phan Huy, Constraint Propagation in Flexible Manufacturing. TX, 258 pages. 2000.
Vol. 513: K. Marti, Stochastic Optimization Techniques. VIII, 364 pages. 2002.
Vol. 493: J. Zhu, Modular Pricing of Options. X, 170 pages. 2000.
Vol. 514: S. Wang, Y. Xia, Portfolio and Asset Pricing. XII, 200 pages. 2002.
Vol. 494: D. Franzen, Design of Master Agreements for OTC Derivatives. VIII, 175 pages. 2001.
Vol. 515: G. Heisig, Planning Stability in Material Requirements Planning System. XTT, 264 pages. 2002.
Vol. 495: I. Konnov, Combined Relaxation Methods for Variational Inequalities. XI, 18 l pages. 200 1 .
Vol. 516: B. Schmid, Pricing Credit Linked Financial Instruments. X, 246 pages. 2002.
Vol. 496: P. WeiR, Unemployment in Open Economies. XII, 226 pages. 2001.
Vol. 5 17: H. 1. Meinhardt, Cooperative Decision Making in Common Pool Situations. VIII, 205 pages. 2002.
Vol. 497: J. Tnkmann, Conditional Moment Estimation of Nonlinear Equation Systems. VIII, 214 pages. 2001.
Vol. 518: S. Napel, Bilateral Bargaining. VIII, 188 pages. 2002.
Vol. 519: A. Klose, G. Speranza, L. N. Van Wassenhove (Eds.), Quantitative Approaches to Distribution Logistics and Supply Chain Management. XTTT, 421 pages. 2002. Vol. 520: B. Glaser, Efficiency versus Sustainability in Dynamic Decision Making. TX, 252 pages. 2002. Vol. 521: R. Cowan, N. Jonard (Eds.), Heterogenous Agents, Interactions and Economic Performance. XIV, 339 pages. 2003. Vol. 522: C. Neff, Corporate Finance, Innovation, and Strategic Competition. IX, 218 pages. 2003. Vol. 523: W.-B. Zhang, A Theory of Interregional Dynamics. XI, 23 1 pages. 2003. Vol. 524: M. Frolich, Programme Evaluation and Treatment Choise. VTTT, 191 pages. 2003. Vol. 525: S. Spider, Capacity Reservation for CapitalIntensive Technologies. XVI, 139 pages. 2003. Vol. 526: C. F. Daganzo, A Theory of Supply Chains. VTTT, 123 pages. 2003. Vol. 527: C. E. Metz, Information Dissemination in Currency Crises. XI, 231 pages. 2003. Vol. 528: R. Stolletz, Performance Analysis and Optimization of Inbound Call Centers. X, 219 pages. 2003. Vol. 529: W. Krabs, S. W. Pickl, Analysis, Controllability and Optimization of Time-Discrete Systems and Dynamical Games. XII, 187 pages. 2003. Vol. 530: R. Wapler, Unemployment, Market Structure and Growth. XXVII, 207 pages. 2003. Vol. 53 1: M. Gallegati, A. Kirman, M. Marsili (Eds.), The Complex Dynamics of Economic Interaction. XV, 402 pages, 2004. Vol. 532: K. Marti, Y. Ermoliev, G. Pflug (Eds.), Dynamic Stochastic Optimization. VTII, 336 pages. 2004. Vol. 533: G. Dudek, Collaborative Planning in Supply Chains. X, 234 pages. 2004. Vol. 534: M. Runkel, Environmental and Resource Policy for Consumer Durables. X, 197 pages. 2004. Vol. 535: X. Gandibleux, M. Sevaux, K. Soremen, V. T'kindt (Eds.), Metaheuristics for Multiobjective Optimisation. TX, 249 pages. 2004. Vol. 536: R. Briiggemann, Model Reduction Methods for Vector Autoregressive Processes. X, 21 8 pages. 2004. Vol. 537: A. Esser, Pricing in (1n)Complete Markets. XI, 122 pages, 2004. Vol. 538: S. Kokot, The Econometrics of Sequential Trade Models. XI, 193 pages. 2004. Vol. 539: N. Hautsch, Modelling Irregularly Spaced Financial Data. XII, 291 pages. 2004. Vol. 540: H. Kraft, Optimal Portfolios with Stochastic Interest Rates and Defaultable Assets. X, 173 pages. 2004. Vol. 541: G.-y. Chen, X. Huang, X. Yang, Vector Optimization. X, 306 pages. 2005. Vol. 542: J. Lingens, Union Wage Bargaining and Economic Growth. XTTT, 199 pages. 2004. Vol. 543: C. Benkert, Default Risk in Bond and Credit Derivatives Markets. IX, 135 pages. 2004.
Vol. 544: B. Fleischmann, A. Klose, Distribution Logistics. X, 284 pages. 2004. Vol. 545: R. Hafner, Stochastic Implied Volatility. XI, 229 pages. 2004. Vol. 546: D. Quadt, Lot-Sizing and Scheduling for Flexible Flow Lines. XVIII, 227 pages. 2004. Vol. 547: M. Wildi, Signal Extraction. XI, 279 pages. 2005. Vol. 548: D. Kuhn, Generalized Bounds for Convex Multistage Stochastic Programs. XI, 190 pages. 2005. Vol. 549: G. N. Krieg, Kanban-Controlled Manufacturing Systems. IX, 236 pages. 2005. Vol. 550: T. Lux, S. Reitz, E. Samanidou, Nonlinear Dynamics and Heterogeneous Interacting Agents. XTTT, 327 pages. 2005. Vol. 551: J. Leskow, M. Puchet Anyul, L. F. Punzo, New Tools of Economic Dynamics. XIX, 392 pages. 2005. Vol. 552: C. Suerie, Time Continuity in Discrete Time Models. XVIII, 229 pages. 2005. Vol. 553: B. Miinch, Strategic Trading in Tlliquid Markets. XTTT, 1 16 pages. 2005. Vol. 554: R. Foellmi, Consumption Structure and Macroeconomics. M, 152 pages. 2005. Vol. 555: J. Wenzelburger, Learning in Economic Systems with Expectations Feedback (planned) 2005. Vol. 556: R. Branzei, D. Dimitrov, S. Tijs, Models in Cooperative Game Theory. VTTT, 135 pages. 2005. Vol. 557: S. Barbaro, Equity and Efficiency Considerations of Public Higer Education. XII, 128 pages. 2005. Vol. 558: M. Faliva, M. G. Zoia, Topics in Dynamic Model Analysis. X, 144 pages. 2005. Vol. 559: M. Schulmerich, Real Options Valuation. XVI, 357 pages. 2005. Vol. 560: A. von Schemde, Index and Stability in Bimatrix Games. X, 151 pages. 2005. Vol. 561: H. Bobzin, Principles of Network Economics. XX, 390 pages. 2006. Vol. 562: T. Langenberg, Standardization and Expectations. M, 132 pages. 2006. Vol. 563: A. Seeger (Ed.), Recent Advances in Optimization. XI, 455 pages. 2006. Vol. 564: P. Mathieu, B. Beaufils, 0 . Brandouy (Eds.), Artificial Economics. XIII, 237 pages. 2005. Vol. 565: W. Lemke, Term Structure Modeling and Estimation in a State Space Framework. IX, 224 pages. 2006. Vol. 566: M. Genser, A Structural Framework for the Pricing of Corporate Securities. XTX, 176 pages. 2006. Vol. 567: A. Namatame, T. Kaizouji, Y. Aruga (Eds.), The Complex Networks of Economic Interactions. XI, 343 pages. 2006. Vol. 568: M. Caliendo, Microeconometric Evaluation of Labour Market Policies. XVII, 258 pages. 2006. Vol. 569: L. Neubecker, Strategic Competition in Oligopolies with Fluctuating Demand. TX, 233 pages. 2006. Vol. 570: J. Woo, The Political Economy of Fiscal Policy. X, 169 pages. 2006. Vol. 57 1 : T. Herwig, Market Conform Valuation of Options. VIII, 104 pages. 2006.