Budget Deficits in Asia : What do Political Variables Explain?
An Empirical Study of the Political Determinants of Budget Deficits in Asia during 1985 - 2006 Alfa Farah (s1711288) Supervisor Dr. R. M. Jong-A-Pin MSc. Economics
Faculty of Economics and Business Rijksuniversiteit Groningen
Acknowledgement
It was June 6, 2008 when I started writing this thesis and I only had less than two months to finish this and hence finish my program! And we finished it! Naturally, this thesis would not have been finished (even written!) without supportive interaction and communication with nice people and friends.
A special thank to Dr. R.M. Jong-A-Pin, my supervisor, for being such a very supportive supervisor, for inspiring discussions and valuable comments. We did it!! My thanks to Prof. J. de Haan for reviewing my paper and giving valuable comments and to Dr. G.H.Kuper for ’handing’ me in to the right person, my supervisor. Also to Kadek Sutrisna, who I believe was very busy; yet always let me interrupt him with silly questions trough YM; to Maria Ulpah, my UB friend, for our econometrics and statistics discussion; to Arviansyah for taking care of everything related to my laptop and also to Steven Shelton and Zulkarnaen Nasution for reading my final draft.
My thanks to Yorieke Deen, M.V Rothengathher-Lake and Heidi Scholtz for supporting me during the hardest days in my academic life in Groningen. Also to Siska Aprilianti for always replying my emails with such a warm touch.
I also have been very fortunate to be accompanied by nice friends who always being there when needed. My thanks to friends in PPIGW, especially; Vivi, Asmi, Iya, Zul, Wulan, Arvi, Naga, Windy, Kenna and Sissy and also to my house mate Julia for always asking how it is going with your thesis and exams. To my friends Daniela and Anca who due to my ‘rush-hour’ academic life I seldom visit even for only chat and drink. We should have gone out a little more!
Finally, among those who are the closest to me, Mama Zahroh and Abah Sobirin, thank you for never stop praying and believing; Mas Anton and Mas Helly for supporting me in their own fashions. And although it sounds strange, I really would like to thank to my self, for keep on working and standing still even when I was terribly exhausted and lost hope. Finally, the most important one, to Allah The Almighty for making something possible to happen really does happens!
Abstract
This paper attempts to analyze the effect of political variables on budget deficits in Asia. We exploit a time series and cross sectional data of 30 Asia countries over 1985 -2006 and build an empirical model based on proceeding literature consisting of economic and political explanatory variables. We focus on four groups of political variables. First, to capture political system, we differentiate countries under presidential system and parliamentary system and we also use a measure of democracy. Second, to measure government fragmentation, we use the effective number of parties in the coalition, the seats held by the government in the parliament and also the number of veto players. Third, we also include a measurement of ideological complexion of government and fourth we include a dummy variable to represent political budget cycles. Using a fixed effect model we find that political variables are not robust enough to explain budget deficits in Asia. Variables related to the government fragmentation and to the ideological complexion of government are less likely to explain budget deficits in Asia, although we find a little evidence that effective number of government parties might be positively related to budget deficits in democracies. We find a weak evidence of the existence of political budget cycle in Asia. We also find a weak tendency for a democracy to generate higher budget deficits than autocracy. Furthermore, it seems that a more democratic regime is likely to have lower budget deficits than a less democratic regime.
I. INTRODUCTION
We will start the discussion with the famous Ricardian equivalence theorem. According to this theorem, the existence of budget deficits is a reflection of inter-temporal choice made between taxing and deficit spending. Budget deficits are irrelevant since deficits today will be compensated by surpluses tomorrow, leaving real output unaffected. However, the Ricardian view about the inter-temporal choice between taxes and deficit spending leave an absence of a theory of public debt creation (Barro, 1979). While accepting the theorem, Barro (1979) further develops the model by taking into account the excess burden of taxation to obtain an optimal amount of debt creation. The model is later known as the equilibrium approach to fiscal policy or the tax smoothing model. Barro (1979) in his paper “On the Determination of the Public Debt” offers an explanation of creation of public debts. According to him, in order to minimize the distortionary effect of taxes, the budgetary authorities should hold the tax rates constant over time. By keeping the tax rates smooth, a deficit will emerge during a recession and a surplus will emerge during an expansion. The deficits are then compensated by surpluses leaving the budget balanced through business cycles.
While Barro’s view of the tax smoothing model succeeds to explain deficits in developed countries during wartime, it fails to explain the persistence of the deficits during peace time (Persson and Tabellini, 2000). Several studies on the budget deficits of developed countries have shown that while those countries have similar economic characteristics, their fiscal performances differ a lot. Hence, it is likely that budget deficits might not be explained by economic variables only. In this way, the political economy of fiscal policy gives its basic contribution. Political economy considers institutions as an important determinant of policy. It emphasizes how private agents’ preferences influence public policies (Persson and Tabellini, 1997). In other words, public policies, in particular budget deficits, are partly a reflection of political behavior of policymakers.
majoritarian governments; de Haan and Strum (1994) have found that the frequency of government changes are positively related to budget deficits; and Perotti and Kontopoulos (2002) have found that the more fragmented governments have higher budget deficits. Those findings suggest that political factors might play an important role in shaping budget deficits.
While more literature pays attention to the importance of political variables for explaining the variation of budget deficits, most of them deal with industrialized countries. Very little empirical work has been done for developing countries, and in particular, for Asia. To mention a few; Roubini (1991) studies the relation between political instability and budget deficits in a large sample of developing countries; Edwards and Tabellini (1992) investigate the effect of political stability and political polarizations in a panel of developing countries over 1963 -1988 and Woo (2003) employs a large set of political explanatory variables to explain budget deficits in 57 developed and developing countries. Furthermore, to our knowledge, there is no study which employs a large set of political explanatory variables on a sample of Asian countries. In this way, our paper contribute to the existing literature by employing a large set of political variables specifically on a sample of Asian countries
In this paper, we ask whether political determinants are capable to explain budget deficits in Asia. To answer the question we build an empirical model based on precedents studies, consisting of economic and political explanatory variables. We include for groups of political variables, namely; political system, the government fragmentation, the ideological complexion of government and also political budget cycle. We perform a pooled time series cross sectional data analysis; i.e. the fixed effect model on a panel of 30 Asian countries over the period of 1985 – 2006.
government parties might be positively related to budget deficits in democracies. We also find a weak evidence of the existence of political budget cycle in Asia and a weak tendency for a democracy to generate higher budget deficits than autocracy. Furthermore, it seems that a more democratic regime is likely to have lower budget deficits than a less democratic regime.
The remainder of the paper is organized as follows. Following the introduction, the second and third section discuss the existing theoretical and empirical literature on the determinants of budget deficits. The fourth section describes the data and statistics. The fifth section presents the method employed in the study. The sixth section discusses the estimation results and the seventh section summarizes and concludes the study.
II. LITERATURE REVIEW
In this section we intend to present the theoretical literature that is related to our study. It then will be used as the basis for our empirical model. Although our main emphasis is on the political explanation of budget deficits, we will start the literature review with economic explanations of budget deficits.
Economic Determinants
Government has to finance its expenditures by imposing taxes or issuing public debts. The level of taxes is determined by the inter-temporal budget constraint which implies that the present value of government expenditures should be equal to the preset value of taxes. According to Ricardian equivalence, an inter-temporal choice between tax and deficits will leave output unaffected. However, according to Barro (1979) taxes are distortionary. For instance, a higher tax on labor income will reduce workers’ after-tax income. It creates disincentives to work so that people prefer leisure more. Thus, it reduces total labor supply1. The aim of government is to minimize this distortionary effect of taxation. The government optimization problem is then to choose the level of taxes, to minimize the distortionary effects of taxation subject to the inter-temporal budget constraint. These conditions require the tax rates to be equal over time.
An important implication of this model concerns the cyclical fluctuation of tax revenues due to the business cycle. During a recession, tax revenue is expected to be lower than it is during an expansion. Furthermore, government spending is also expected to be higher during a recession than it is during an expansion. Therefore, deficits should be observed in recessions and compensated by surpluses in expansions.
The effect of economic slowdown to budget deficits also can be captured via unemployment-related spending. Many areas of public spending (e.g. unemployment compensation, social welfare expenditure, early retirement benefits, job retraining, and subsidies) are inherently countercyclical, so that portions of government spending actually tend to rise automatically when growth slows down and unemployment increases. For example, during a recession, unemployment is high. To overcome this, the government raises the unemployment-related spending. If the revenue is constant, an increase in government spending will lead to a higher budget deficit.
We further analyze the effect of inflation on budget deficits. According to Dornbusch, et al. (2004), there are two main mechanisms through which inflation increases budget deficits; namely, tax collection effects and increases in nominal payments on national debt. The first mechanism is also known as the Tanzi-Olievera effect. According to the Tanzi-Olievera effect, when inflation becomes high budget deficits typically become large as well (Blanchard, 2006). The argument is that there are lags in both the calculation and payment of taxes. Taxes are collected on past nominal income so that their real value goes down as inflation increases. Therefore, high inflation reduces real government revenue. This problem is often accompanied by other effect on the expenditure side. For instance, to reduce the inflation, government often forbids its state companies to increase their prices. Since their cost increases due to inflation, the state companies suffer loss. In turn, the loss must be financed by government subsidies, which at the end push the budget deficits up.
However, inflation may negatively affect fiscal deficits by raising revenue via income tax bracket creep2. For instance, the wage rate goes up due to inflation. When the tax
2
rates are not indexed for inflation, an increase in inflation may push the lower bracket tax payers to creep to a higher bracket, resulting in a higher tax revenue.
Political Determinants
The tax smoothing model is a normative benchmark from which political economy models of budget deficit diverge (Alesina and Perotti, 1994). It relies on the assumption that government is a benevolent social planner that only wants to maximize the utility of the society. In fact, policies are made by opportunistic agents who have their own preferences.
The general approach of the political model in budget deficit is to explain the deviation of observed economic policies from the normative benchmark by including specific incentive constraints in the decision making process (Persson and Tabellini, 1997). De Haan and Sturm (1994) classify the constraints in four political and institutional models of fiscal outcomes, namely; (1) models which focus on political system, (2) models which focus on the disagreement between various decision makers, (3) models which focus on ideological differences and (4) models which focus on budgeting procedures. We will examine these models below.
1) Political System
The first class of models investigates how political system affects the behavior of policy makers. It is argued that the political system of a country plays a role in shaping its budget deficits. In general, economic policy (in particular budget policy) is easier to formulate and implement under a presidential system than under a parliamentary system. The reason is that under a presidential system, the government has greater independence and less interference from legislature than under a parliamentary system (Woo, 2003).
services and hence government expenditures. Furthermore, the policy-making body in non-democratic regimes is much smaller than in democratic regimes. Because the policy making body is smaller, the costs and benefits of government policies will be more internalized. Second, the absence of political competition among potential political suppliers to obtain a temporary contract from voters may reduce the budget deficits. In democratic regimes, we may consider voters as public goods’ demanders and politicians as potential public goods’ suppliers. Through election, politicians compete to win voters’ temporary contract of producing public goods. The uncertainty of re-election and the possibility for inter-temporal transfer of deficits trigger the incumbent government to raise deficits, leaving debt to his successor. Furthermore, non-democratic governments do not face an election constraint. Therefore, there is no incentive to attract voters in the next election with deficit spending.
2) Government Fragmentation
The second model of political economy of the budget deficit focuses on disagreement among various decisions makers. Roubini and Sachs (1989a) argue that governments are not monolithic entities who have full control over policy instruments in order to achieve a specific well-defined goal. In fact, the decision making process is often fragmented among several political agents. Examples of fragmented governments are coalition governments, several numbers of veto players in the decision-making process, and ideological preferences of government parties.
expenditures in a coalition government also reflect the government’s effort to maintain coalition and avoid internal conflict by supplying budgetary needs of each coalition member party (Roubini and Sachs, 1989a).
3) Ideology
The existing literature suggests that ideological representation of government might affect the size of budget deficits. It is often argued that left wing government aims for a higher government spending, hence a larger budget deficit.
According to Hibbs (1977) the right wing and right wing governments’ economic platform are class-related. Left wing governments typically weight the unemployment problem more heavily. In contrast, right wing governments favor a relatively low inflation. This is because left wing supporters are mainly middle-lower income class (labor owner groups) suffering the most from the costs of unemployment whereas right wing supporters are mainly upper income class (business oriented/ capital owner groups) suffering more seriously the costs of inflation.
These interests over the inflation-unemployment issue are reflected in budget policy. When left wing governments are in office, they tend to perform loose fiscal policy. The opposite is also true, right wing governments tend to conduct tight fiscal policy. The differences over inflation and unemployment suggest a trade-off between inflation and unemployment, well known as the Phillips curve. To overcome the unemployment problem, left wing governments will conduct an expansionary policy such as increasing government spending. An increase in government spending will lead to a higher output, and a higher employment rate. As employment increases, labor supply declines, the wage rate increases. A higher wage rate leads to a higher inflation. This reverse relationship is more favorable in the short run, which is in line with political decision that is also typically short run. However, according to political business cycle theory proposed by Nordhaus (1975), governments generally will inflate during election years in order to exploit a Phillips curve tradeoff.
4) Budget procedures
the budget and (2) procedural rules. Laws that set certain numerical targets or legislated limit rules are formal laws or rules that restrict the budget outcomes on certain numerical targets. These include balanced budget laws, expenditure ceilings, numerical targets for fiscal variables and restrictions on issuance of debt (Drazen, 2000). Alesina and Perotti (1996) argue that such numerical targets are not necessarily needed to generate fiscal discipline.
The second budget institutions; procedural rule includes three stages in the budgeting process, namely; (1) formulation, (2) approval and (3) implementation. Two essential issues are the voting procedure leading to the formulation and approval of the budget and the degree of the transparency of the budget. Concerning voting procedures in the budget formulation state, there are two types of voting that might impact the budget outcomes. First the hierarchical procedure (a budget procedure that attributes a strong prerogative power to the prime minister or finance minister) and second the collegial procedure (a budget procedure that gives each spending minister a significant power). The collegial procedures emphasize more on the democratic process in decision making. Alesina and Perotti (1996) argue that the hierarchical procedure tends to generate a relatively more stringent fiscal policy hence a lower fiscal deficit. In contrast, due to its egalitarian features which give each spending minister in the cabinet more power to set their desirable budgetary needs, budget deficits are higher under collegial procedure. On the empirical side, von Hagen (1991), Von Hagen and Harden (1996), Halleberg and von Hagen (1997) and Perotti and Kantopoulos (2002) have studied the effect of budget procedure to the budget deficit.
Von Hagen and Harden (1996) also construct an index of centralization3 measuring centralization in budget process. The index includes four important items that further divided in several sub items. The four items are (1) the structure of decisions of government stage, (2) the structure decisions at parliamentary stage, (3) the flexibility of implementation and (4) the informativeness of budget documents. Using a sample of 12 European countries over 1970 – 1990, they find a strong correlation between centralization and fiscal discipline. In particular, countries with high score of the centralization index have lower budget deficits.
Besides budget procedure, another important issue is transparency. According to Alesina and Perotti (1996) politicians tend to produce complex, unclear and less transparent budget. The more complex, unclear and less transparent budget may lead to voters’ confusion and reduce politicians’ incentives to be more fiscally disciplined. 5) Political Budget Cycle
Besides the four models described earlier, the political budget cycle is also used to explain budget deficits. Mink and de Haan (2006) distinguish three generations of theoretical political budget cycle models. The first generation model which was first proposed by Nordhaus (1975) is a part of a broader literature on political business cycles. According to political business cycle theory, in order to maximize the re-election probability, the incumbent governments perform fiscal manipulations. They often use expansionary economic policy to stimulate aggregate demand in order to signal a sound economic performance (i.e GDP growth and unemployment rate) to the voters. Because of the lack of empirical evidence, the political business cycle theory studies have shifted their focus from the real effects of elections to the policy makers’ instruments, in particular fiscal expansion in election years or generally known as the political budget cycle.
The second generation of political budget cycle (the adverse selection type) is first developed by Rogoff and Sibert (1988). Political agents are assumed to have a certain level of competence (high or low) that is known only by the politicians and not by the
3 The index is based on von Hagen previous paper “ Budget procedures and fiscal performance in the
voters. Voters are assumed to prefer more competent politicians and evaluate the competence from current observable fiscal outcomes. The high competence politicians will signal their type (high performance) by doing a loose fiscal policy resulting in a higher budget deficit prior to the election. The incumbent government also can signal their type by shifting expenditure to easily observed consumption spending and away from investment. According to Shi and Svensson (2003) this separating equilibrium implies that only competence politician will inflate prior to the election and as voters are rational to choose the most competence politicians, only high competence will be elected.
III. EMPIRICAL EVIDENCE
There are various empirical studies on the impact of the political variables on budget deficits which are conducted for a variety of countries and time periods. However, most of the empirical studies in this field deal with the experience of developed countries, particularly OECD countries. To our knowledge, there are only two studies focusing on developing countries; namely Roubini (1991) and Edwards and Tabellini (1990). Woo (2003) focuses on both developed and developing countries.
Because typical studies analyze a group of countries during a certain period of time, most of them exploit a pooled time series cross section dataset. To our knowledge, only the work of Roubini (1991) uses a cross section dataset.
Various ways to measure political variables are used in the literature. Roubini and Sachs (1989a) introduce an index4 -which is later known as Roubini-Sachs political cohesion index- to measure the degree of political cohesion of the national government. This index assigns 0 for a one-party majority parliamentary government or a presidential government with the same party in the majority in the executive and legislative branch; 1 for a coalition parliamentary government with 2 coalition partners or a presidential government with different parties in control of the executive and legislative branch; 2 for a coalition parliamentary government with 3 or more coalition partners; and 3 for a minority parliamentary government. Using the index, Roubini and Sachs (1989a) find a clear tendency for larger deficits in countries with a relatively large number of political parties in government. Yet, their results have been questioned by Edin and Ohlsson (1991). Edin and Ohlsson (1991) argue why one should believe that the budget-effect of a minority government is three times as large as a two-party majority coalition. After testing the robustness of the results of Roubini and Sachs (1989a), they replace the Roubini-Sachs political cohesion index with a dummy variable for each political class. They find that the index captures the effects of minority governments rather than majority coalition governments. Re-examining the effect of the Roubini and Sachs
4 In another paper, Roubini and Sachs (1989b) redefine the index by assigning 1 for a coalition
political cohesion index5, De Haan and Sturm (1997) do not find any significant relationship between the index and government debt.
Volkerink and De Haan (2001) use a measure government fragmentation, as their political explanatory variables. They use effective number of government parties in the coalition and the total number of spending ministers in the cabinet. Moreover, they include the position of government vis á vis parliament which is measured by the number of seats above those needed for a simple majority and the effective number of parties in the parliament. They also include the ideological complexion of government and parliament in their explanatory variables. They find that more fragmented governments are likely to have higher budget deficits.
Woo (2003) employs a large set of political variables in his study. To measure government fragmentation, he uses the number of seats held by the largest party in the lower house, the party fractionalization index6 and the number of ministers in the cabinet. He also includes a variant of Roubini and Sachs political cohesion index7 and tests the index by including a dummy for minority governments as suggested by Edin and Ohlsson (1991). To measure political regime, a dummy taking value 1 for a presidential system government and 0 otherwise is included in the model. The study shows that a large size of the cabinet and lack of central authority are strongly negatively related to public surplus. The study also shows that proportional parliamentary regimes tend to run higher deficits and that a government weakness or regime type does not seem to be consistently related to budget deficits.
Shi and Svensson (2002) show that during election periods, government expenditures rise and revenues fall, thus creating higher budget deficits. The result is observed in
5 De Haan and Sturm (1997) assign 0 for a one-party majority parliamentary government; 1 for a coalition
parliamentary government with two-to-three coalition partners 4; 2 for a coalition parliamentary government with four or more coalition partners; and 3 for a minority government.
6 The party fractionalization index is defined as the probability that two randomly chosen legislators
belong to different parties.
7 Woo (2003) modify the index by not distinguishing the presidential and parliamentary system. He
Table 1 : The Studies in the Effects of Political Variables on Budget Deficits
No Author Sample Political Variables Main Findings
1 Roubini and Sachs (1989a)
A panel of 13 industrial countries over 1972 -1985
Roubini-Sachs political cohesion index (1989a)
Larger coalition governments have higher budget deficits than one party, majoritarian governments. 2 Roubini and Sachs (1989b) A panel of OECD countries (1960 – 1985)
Roubini-Sachs political cohesion index (1989b)
Budget deficits are larger in countries with weak government (a short average tenure of government) and with many political parties in the coalition. 3 Edwards and Tabellini (1990) A panel of developing countries over 1963 -1988
Political instability which is measured by the frequency of government change
Politically unstable countries tend to have larger budget deficits
4 Edin and Ohlsson (1991) A panel of 13 OECD countries covering a maximum period of 1964 - 1985 (minimum 1972 - 1985)
Roubini-Sachs political cohesion index (1989a)
- The finding of Roubini and Sachs (1989b) that coalition government less capable of budgetary deficits is robust.
- Disaggregating the power dispersion index (using a dummy for each political class) shows that the index captures the effects of minority governments rather than majority coalition governments.
5 Roubini (1991) A cross section of 77 countries during the 1971 - 1982
The frequency of government change - The greater the frequency of government change, the greater the deficits. - Military regimes are not more likely to run budget deficits than democratic
regimes. 6 De Haan and
Sturm (1994) A panel of European Community member countries during 1980s.
1. The number of government changes, 2. The share of cabinet portfolios or seats in
parliament held by social democratic and other leftist parties,
3. The (corrected) Roubini Sachs political cohesion index8
4. A variant of the Von Hagen budgetary process variable
- The frequency of government changes are positively related to government debts.
- Good performances of budgetary procedure are negatively related to government debts. 7 De Haan and Sturm (1997) A panel of 21 OECD countries over 1982 – 1992
The (corrected) Roubini-Sachs political cohesion index
- The political variables are not significant.
- The growth of government debt is not related to Roubini-Sachs political cohesion index nor to the variant suggested by Edin and Ohlsson (1991)
Continued Table 1
No Author Sample Political Variables Main Findings
8 De Haan, et al.
(1999) A panel of 21 OECD countries over 1979 – 1995
1. The (corrected) Roubini-Sachs political cohesion index
2. Roubini-Sachs political cohesion index as suggested by Edin and Ohlsson (1991) 3. The number of coalition parties
- Type of government does not seem to affect fiscal policy
- Number of political parties is not significant to explain general government debt. However, it is significant to explain central government debt
9 Volkerink and De Haan (2001)
A panel of 22 OECD countries over the period of 1971 -1996.
1. The size fragmentation (the effective number of parties in the coalition and the number of the spending ministers), 2. The position of government vis a vis
parliament (the number of seats above needed for a simple majority and the effective number of parties in parliament), 3. The ideological complexion
4. The political fragmentation of government.
- Concerning the size fragmentation, the number of ministers is stronger and more robust to explain budget deficits than the effective number of parties in
government.
- Concerning the position of government vis a vis parliament, the effective number of parties in the parliament is the most robust to explain budget deficits. - The political fragmentation does not seem to influence budget deficits. - Left wing governments have higher budget deficits.
10 Perotti and Kontopoulos (2002) A panel of 19 OECD countries over 1970 – 1995
1. Roubini-Sachs political cohesion index (1989a)
2. The number of spending ministers 3. The number of parties in the coalition 4. The ideological complexion
5. The Budget Procedure
- The type of government is fragile in explaining deficits (inference depends on the coding)
- Cabinet size is significant to explain fiscal outcomes - Coalition size is also significant to explain fiscal outcomes - Ideology is also significant to explain fiscal outcomes 11 Woo (2003) A panel comprised of
averages of variables for 1970-1979 and 1980 - 1990 for 57 developed and developing countries
1. Political instability (cabinet change, coups, changes in effective executive, and major constitutional change, political assassination, government crisis, and revolution)
2. Government fragmentation (the number of seats held by the largest party in the lower house, the party fractionalization index, the variant of Roubini Sachs Index, Minority dummy variables and Cabinet Size)
3. Political Regime and Electoral Law 4. Social Polarization (Income inequality
and Ethnic divisions)
- Sociopolitical instability, income inequality, a large size of the cabinet and lack of central authority are strongly negatively associated with public surplus.
- Proportional parliamentary regimes tend to run higher deficits,
- Government weakness or regime type does not seem to explain budget surplus. - The budgetary institutions and government institutions in general influence the
fiscal outcomes
Continued Table 1
No Author Sample Political Variables Main Findings
11 Huber, et al.
(2003) 21 OECD countries in time period 1970-1999 1. The government strength (The sum of the Banzhaf indices9 of parties in government
2. The government dispersion (The standard deviation Banzhaf indices of parties government,
3. Dummy for election years 4. The membership in EMU.
- The stronger government are not prone to have lower budget deficits
- Equally strong coalition partners tend to block any cooperative outcome by using their veto power.
- Coalition governments composed of parties which differ considerably in their voting power are better in achieving a successful stabilization of their debt levels.
IV. DATA AND STATISTICS
Data
We compile a dataset from many sources, namely; Key Indicator Statistics (Asia Development Bank, http://www.adb.org/statistics), The United Nation Statistics Division ( The United Nation, http://unstats.un.org/unsd/databases.htm) , IMF Statistical Databases - Online Browsers and Documentation (International Monetary Funds), The Database of Political Institution 2006 (Beck, et al., 2007) and The Polity IV Project: Political Regime Characteristics and Transitions, 1800-2006 (Marshall and Jaggers, 2007). Data descriptions and sources can be found in appendix A.
Statistics
Table 2.a and Table 2.b report the descriptive statistic of the data. In a sample of 22 countries, one should be aware that ENPG and INF have a large range of data. The minimum value of ENPG is 0.01 and the maximum value is 150.00. The reported standard deviation is 32.07. Meanwhile, the minimum value of INF is -2.29, the maximum value is 128.42 and the standard deviation is 10.84
Table 2.a
Descriptive Statistic in Sample of 22 Countries
Variables Mean Median Max Min Std. Dev. Obs. Crosec
DEF -2.64 -2.62 25.22 -20.01 5.08 451 22 GDP 5.52 5.80 39.50 -33.60 5.24 484 22 INF 7.24 4.70 128.42 -2.29 10.84 406 21 UNEMP 4.52 3.60 15.90 0.60 2.84 331 22 PRES 0.86 1.00 2.00 0.00 0.86 482 22 DEM 0.06 0.00 9.00 -8.00 6.69 420 20 ENPG 10.67 1.00 150.00 0.01 32.07 423 21 CHECKS 2.48 1.00 18.00 1.00 2.27 471 22 MAJ 0.77 0.83 1.00 0.21 0.23 419 21 ICP -0.18 -0.17 1.00 -1.00 0.80 311 17 ELEC 0.05 0.00 1.00 0.00 0.23 484 22
Max : Maximum Obs. : Observations Min : Minimum Crosec : Cross sections Std. Dev. : Standard Deviation
Meanwhile, the minimum value of INF is -8.52 while the maximum value is 5244.20 and the standard deviation value of 297.22.
Table 2.b
Descriptive Statistic in Sample of 30 Countries
Variables Mean Median Max Min Std. Dev. Obs. Crosec
DEF -2.31 -2.36 25.22 -31.35 4.81 449 30 GDP 4.51 5.80 39.50 -44.90 7.87 480 30 INF 37.18 5.67 5244.20 -8.52 297.22 405 29 UNEMP 4.80 3.81 14.70 0.10 3.37 348 30 PRES 0.65 0.00 2.00 0.00 0.85 472 30 DEM 0.37 0.50 10.00 -9.00 6.81 442 28 ENPG 8.30 1.04 150.00 0.01 27.63 400 28 CHECKS 2.42 1.00 18.00 1.00 2.27 452 30 MAJ 0.73 0.75 1.00 0.21 0.23 399 28 ICP -0.21 -0.19 1.00 -1.00 0.74 297 23 ELEC 0.09 0.00 1.00 0.00 0.29 472 30
Max : Maximum Obs. : Observations Min : Minimum Crosec : Cross sections Std. Dev. : Standard Deviation
Table 3.a and 3.b report the correlations between explanatory variables. One should be aware that some variables have relatively high correlation which in turn may influence the regression results. For instance; the correlation between DEM and MAJ are -0.713 and -0.613 for a sample of 22 countries and a sample of 30 countries respectively (significant at 1% level of significant). According to Hill et al. (2000) collinear variables do not provide sufficient information to estimate their separate effects. Therefore, it causes difficulties in isolating the individual effect of each variable.
Table 3.a
Correlations (Sample of 22 Countries)
GDP INF UNEMP PRES DEM ENPG CHECKS MAJ ICG ELEC
GDP 1 INF -0.019 1 UNEMP -0.162** 0.159** 1 PRES 0.071 -0.156** -0.367** 1 DEM -0.189** -0.270** 0.266** 0.241** 1 ENPG 0.159** 0.013 -0.044 -0.326** -0.315** 1 CHECKS -0.042 -0.109* 0.127* 0.415** 0.613** -0.198** 1 MAJ 0.137** 0.102* -0.258** -0.108* -0.713** 0.286** -0.639** 1 ICG -0.028 -0.289** -0.108 0.126* 0.596** 0.110* 0.265** -0.535** 1 ELEC -0.003 -0.010 0.102* -0.176** 0.107* -0.027 -0.015 -0.072 0.044 1 ** Correlation is significant at the 0.01 level (1-tailed).
Table 3.b
Correlations (Sample of 30 Countries)
GDP INF UNEMP PRES DEM ENPG CHECKS MAJ ICP ELEC
GDP 1 INF -0.232** 1 UNEMP 0.034 -0.068 1 PRES 0.111** -0.088* -0.297** 1 DEM -0.061 0.007 0.286** 0.304** 1 ENPG 0.048 -0.021 -0.025 -0.210** -0.244** 1 CHECKS -0.026 -0.014 0.065 0.427** 0.588** -0.154** 1 MAJ 0.010 -0.037 -0.406** 0.004 -0.613** 0.276** -0.534** 1 ICP 0.048 -0.101 0.076 0.166** 0.605** 0.122* 0.231** -0.579** 1 ELEC -0.107** 0.042 0.141** -0.198** 0.047 -0.045 -0.057 -0.064 0.024 1 ** Correlation is significant at the 0.01 level (1-tailed.)
* Correlation is significant at the 0.05 level (1-tailed.)
V. METHOD
The empirical analysis is conducted by using a pooled time-series cross sectional data analysis; i.e. fixed effect model, on a sample of 30 Asian countries10 over the period of 1985-200611. This model specifies that only the intercept parameter varies, not the
response parameters. Furthermore, the intercept varies only across countries not over time. Therefore, all behavioral differences between countries and over time are captured by the intercept. We develop a model based on several precedent studies discussed earlier in the literature review. The estimated equation is:
DEFit = α0 + α1 DEFL it + α2ECOit + α3POLit + μi + eit (1)
,where i and t denote the country and year. The dependent variable (DEF) is the budget balance-to-GDP ratio of the central government12. The explanatory variables consist of economic explanatory variables (ECO) and political explanatory variables (POL). The
10 The sample includes countries in Eastern Asia (China, Japan, Mongolia, South Korea and Taiwan) ,
Southeastern Asia (Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Thailand and Vietnam), Southern Asia (Afghanistan, Bangladesh, Bhutan, India, Maldives, Pakistan and Sri Lanka), Central ( Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan) and Western Asia (Armenia, Azerbaijan and Georgia)
11 However, as some countries were established only after 1990, we plan to split the sample into two. First
we will conduct a regression with full coverage of the period (22 countries over 1985 - 2006). We then regress with full coverage of the sample (30 countries over period 1991-2006). For a detailed sample see appendix B.
12 We define budget balance as the difference between total revenue and total expenditure. When the
economic explanatory variables are; the real GDP growth rate (GDP), the rate of inflation (INF) and the rate of unemployment (UNEMP). The political explanatory variables are clustered into four groups which we will explain further later. We also include the lag of the dependent variable (DEFL) in our set of explanatory variables to allow for autocorrelation problem in the model. Finally, μi is a country fixed effect and eit is an error term.
We expect that the GDP will be positively related to DEF (budget deficits will emerge when output is temporarily low), UNEMP will be negatively related to DEF (as the unemployment rate increases, government might increase the unemployment-related spending. Hence, it will raise the budget deficit) and INF is also negatively related to DEF.
We include four groups of political variables in the model, namely; (1) political regime, (2) government fragmentation, (3) the ideological complexion of government and (4) political budget cycle. These political variables are included for several reasons. First, concerning the political regime, we would like to test whether countries under a presidential system are less likely to generate higher budget deficits than countries under a parliamentary system. Furthermore, we also would like to test whether democracies are prone to higher budget deficits. Second, concerning the government fragmentation, we would like to test whether a higher number of decision makers in the budget decision process lead to a higher budget deficit. Third, concerning the ideological complexion of government, we would like to test whether left wing governments are in favor of larger budget deficits. Fourth, concerning the political budget cycle, we would like to test whether higher budget deficits will be found around the election years.
parliamentary system are twice as large as in a presidential system. Therefore, we will also use a dummy variable for each political system, namely; PRES0 (scoring 1 for a presidential system and 0 otherwise), PRES1 (scoring 1 for an assembly-elected president system and 0 otherwise) and PRES2 (scoring 1 for a parliamentary system and 0 otherwise). To measure democratic regimes, we collect data from The Polity IV Project (Marshal and Jaggers, 2007) which measures democracies in a range of +10 (full democracy) to – 10 (full autocracy). We expect that both PRES and DEM will be negatively related to DEF.
For the second group, the government fragmentation, De Haan and Sturm (1994) argue that one possible measure for the fragmentation of fiscal policy-making is the number of decision makers. In particular, they measure fragmentation by using the number of parties in coalition and the number of spending ministers in cabinet. Following De Haan and Sturm (1994), we include ENPG (effective number of parties in the coalition) which is measured as follows:
∑
= = n i ip
ENPG 1 2 1 (2),where pi denotes the share of ministers from party i as a proportion of the total number
of ministers and n is the number of coalition parties. This concept is the inverse of the Government Herfindahl index.
Besides ENPG, we also include other measurements of government fragmentation, namely; MAJ and CHECKS. MAJ calculates the fraction of seats held by the government and CHECKS calculates the number of veto players in the decision making process.
budget deficits after adverse shocks, since each party in the coalition will protect parts of the budget that are vital to them so that they will veto a spending cut or tax increase (De Haan and Sturm, 1994).
For the third group, the ideological complexion of government, following Volkerink and De Haan (2001), we measure the ideological complexion of government (ICG)13 as follows i n i i Ide Total Seat ICG=
∑
× =1 (3),where i denotes the government party i, Seati denotes the number of seats held by the
government party i, Idei is the ideology of party i and Total denotes the total seats held
by all government parties. We collect ideology preferences data from The Database of Political Institution 2006 (Beck. et al., 2007) and we score -1 for left wing parties, 0 for centrists and 1 for right wing parties. We expect that ICG will be positively related to DEF (left wing parties prefer higher deficits)
Finally, to capture the existence of a political budget cycle, we include ELEC, which is a dummy variable with a score of 1 for election years and 0 otherwise. We expect that ELEC will be negatively related to DEF (deficits are higher during election years) For convenience, we summarize the explanatory variables and the expected sign in the regression in Table 4.
13While we include the ideological complexion variable in our model, we are fully aware of what Huber
Table 4
The Explanatory Variables
Variables Description Expected Sign to
budget balance
GDP Annual GDP growth +
INF Inflation Rate -
UEMP Unemployment Rate -
PRES Political System -
DEM Democracy Regime -
ENPG Effective Number of Parties in The Coalition -
CHECKS Number of Veto Players -
MAJ Seats held by the government +
ICG Ideological Complexion of Government +
ELEC Election -
For a detailed data description and source see appendix A
VI. ESTIMATION RESULTS
Before discussing the political determinants of budget deficits in Asia, we present the results of regression when we include only the economic variables. The reported Durbin Watson Statistics show that we can not reject the null hypothesis of no autocorrelation. Hence, to take into account the serial correlation problem in the estimation, we will include a lag of dependent variable in the set of explanatory variables. We are also aware of the endogeneity problem in the model (the two way causality between dependent and independent variables may exist). Therefore, we also conduct regressions with the lag of economic explanatory variables.
Furthermore, we conduct an F-test to test for fixed effects. The reported F statistic and the associated p value show that we can reject the null hypothesis that the intercept parameters for all countries are equal. Therefore, country dummies are included in all regressions.
The estimated coefficient of INF is positive and significant at 5% level of significant in sample 22. However, in sample 30 INF is negative and insignificant. The estimated coefficient of UNEMP enters with the expected sign (negative) in both samples but only is significant at 10% level of significant in sample 30.
We are concerned that we include two variables that more or less measure the same thing (economic performance); namely, unemployment rate and annual GDP growth. However, the idea to include unemployment rate in the regression is to know how the effect of economic growth on budget deficits is captured via unemployment-related spending. Furthermore, if the trade off between inflation-unemployment exists, including inflation and unemployment rate at the same time may suggest the existence of a collinearity problem. Therefore, in column (2) and (5) we exclude UNEMP from the regression.
When we exclude UNEMP from the regression, GDP is still positive and significant at 1% level of significant in both samples. INF is positive in both samples and remains significant in sample 22 (but now at 10% level of significant) and insignificant in sample 30. By excluding UNEMP from the regression, we see that the significance of GDP and INF are unchanged in both samples. Considering this and the above reasons, we therefore decide to drop UNEMP from subsequent regressions14.
Next, we employ the lag of economic explanatory variables in our regressions. Column (3) and column (6) show that the sign and the significance of GDPL is the same as GDP (positive and significant at 1% level of significant). INFL is positive and significant at 5% level of significant in sample 22 while negative and insignificant in sample 30. DEFL enters both samples with a coefficient of about 0.45 (significant at 1% level of significant), suggesting that about 45 percent of the previous period budget balance persists to the next period budget balance.
14 Unemployment rate data also contains a lot of missing values (31.61 % and 27.50 % for sample 22 and
Finally, the reported adjusted R square is around 0.68 (column (6)) and 0.79 (column (1)). This suggests that around 68 percent to 79 percent variation in the budget deficits can be explained by our set of economic explanatory variables.
Table 5
Economic Variables on Budget Balance in Asia Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) N = 30 (1991 – 2006) Variables (1) (2) (3) (4) (5) (6) DEFL 0.57* 0.46* 0.44* 0.53* 0.46* 0.44* (0.04) (0.06) (0.05) (0.03) (0.04) (0.04) GDP 0.19* 0.17* 0.11* 0.13* (0.06) (0.05) (0.04) (0.13) INF 0.04** 0.02*** -0.00 0.00 (0.02) (0.01) (0.00) (0.00) UNEMP -0.06 -0.13*** (0.07) (0.08) GDPL 0.11* 0.09* (0.03) (0.02) INFL 0.01** -0.00 (0.05) (0.00) Obs. 273 371 364 291 373 364 Adj. R2 0.79 0.72 0.71 0.76 0.68 0.67 F Stat 43.52 41.71 39.52 29.16 26.93 25.03 DW 1.84 1.88 1.88 1.88 1.96 1.94
F Stat (Fixed Effect) 3.27 4.96 5.104 2.77 3.42 3.15
p value 0.00 0.00 0.00 0.00 0.00 0.00
- Level of significant is indicated by asterisks : * 1%, ** 5% and *** 10 % - Standard errors white consistent covariance matrix are reported in parentheses - To know what causes the differences between two samples’ results, we conduct
regressions for sample with 22 countries over 1991 – 2006. The results show that the differences are due to the inclusion of more countries in the sample
Table 6 reports the results when we include variables related to the political system in our regressions. As expected, in both samples PRES and DEM are negatively related to DEF. However, only in sample 30 PRES and DEM are significant (at 1% level of significant). This implies that in sample 30, countries under a parliamentary system tend to have higher deficits. Furthermore, democratic countries also tend to have higher budget deficits.
variables (PRES0, PRES1 and PRES2) are significant in sample 22. These results are consistent with our results in column (1) and column (6).
As expected, in sample 30 all of the dummy variables are significant at 1% level of significant. The estimated coefficient of PRES0 is positive whereas the estimated coefficients of PRES1 and PRES2 are negative. It implies that presidential system tends to generate lower budget deficits. Again, these results are consistent with our results in column (11) and column (16).
Table 6
Political System on Budget Deficit in Asia
Dependent variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) N = 33 (1991 – 2006) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) DEFL 0.48* 0.50* 0.48* 0.48* 0.48* 0.44* 0.44* 0.44* 0.44* 0.44* 0.46* 0.49* 0.46* 0.46* 0.46* 0.43* 0.45* 0.43* 0.44* 0.44* (0.05) (0.09) (0.06) (0.06) (0.06) (0.05) (0.09) (0.05) (0.05) (0.05) (0.04) (0.07) (0.04) (0.04) (0.04) (0.04) (0.08) (0.04) (0.04) (0.04) GDP 0.17* 0.18* 0.017* 0.17* 0.17* 0.13* 0.13* 0.13* 0.13* 0.13* (0.05) (0.05) (0.05) (0.05) (0.05) (0.04) (0.04) (0.04) (0.04) (0.04) INF 0.02** 0.02*** 0.02*** 0.02** 0.02** 0.00 0.00 0.00 0.00 0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) (0.00) (0.00) GDPL 0.11* 0.13* 0.11* 0.11* 0.11* 0.09* 0.10* 0.10* 0.09* 0.09* (0.03) (0.03) (0.03) (0.34) (0.34) (0.02) (0,02) (0.01) (0.02) (0.02) INFL 0.02** 0.02* 0.02** 0.02** 0.02** -0.00 -0.00 -0.00 -0.00 -0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) (0.00) (0.00) PRES -0.02 -0.08 -0.58* -0.65* (0.17) (0.17) (0.08) (0.05) DEM -0.03 -0.03 -0.07* -0.08* (0.03) (0.04) (0.02) (0.02) PRES0 -0.04 0.04 0.95* 1.03* (0.33) (0.24) (0.00) (0.22) PRES1 0.26 0.19 -0.78* -0.71 (0.56) (0.17) (0.12) (0.08) PRES2 0.26 -0.31 -1.10* -1.28* (0.56) (0.39) (0.17) (0.11) OBS 371 329 371 371 371 364 324 364 364 364 373 341 373 373 373 364 333 364 364 364 Adj. R2 0.72 0.80 0.72 0.72 0.72 0.71 0.80 0.73 0.71 0.71 0.68 0.78 0.68 0.68 0.68 0.67 0.77 0.67 0.67 0.67 F Stat 39.85 65.89 39.85 39.87 39.87 37.77 61.24 37.76 37.77 37.78 26.11 43.37 26.11 26.04 26.08 24.29 40.00 24.29 24.20 24.27 DW 1.88 2.08 1.89 1.88 1.88 1.88 2.01 1.88 1.88 1.88 1.96 2.20 1.96 1.96 1.96 1.94 2.11 1.94 1.94 1.94 F Stat (Fixed Effect) 4.91 5.36 4.93 4.93 4.93 5.07 5.80 5.08 5.08 5.06 3.45 3.91 3.44 3.37 3.43 3.20 3.69 3.19 3.13 3.19 p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Table 7 reports the results when we include variables related to the government fragmentation in our regressions. Column (1) to column (6) report regression results for sample 22 while column (7) to column (12) report regression results for sample 30. None of the variables measuring government fragmentation are significant. It implies that government fragmentation is less likely to explain budget deficits in Asia.
Table 7
Government Fragmentation on Budget Deficit (1985 – 2006) Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) N = 33 (1991 – 2006) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) DEFL 0.52* 0.48* 0.53* 0.46* 0.45* 0.47* 0.53* 0.46* 0.53* 0.51* 0.44* 0.51* (0.07) (0.02) (0.07) (0.07) (0.05) (0.07) (0.05) (0.05) (0.05) (0.06) (0.04) (0.05) GDP 0.21* 0.17* 0.22* 0.16* 0.13* 0.16* (0.04) (0.05) (0.04) (0.04) (0.04) (0.04) INF 0.02*** 0.02*** 0.02*** 0.00*** 0.00*** 0.00*** (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) GDPL 0.13* 0.11* 0.13* 0.06* 0.07* 0.06* (0.04) (0.04) (0.04) (0.02) (0.02) (0.02) INFL 0.02** 0.02** 0.02** -0.00*** -0.00 -0.00*** (0.00) (0.01) (0.00) (0.00) (0.00) (0.00) ENPG 0.15 0.17 -0.04 -0.03 (0.13) (0.16) (0.11) (0.11) CHECKS -0.00 -0.00 -0.05 -0.05 (0.00) (0.02) (0.03) (0.03) MAJ -0.26 -0.07 0.01 -0.13 (0.47) (0.55) (0.82) (0.82) OBS 333 365 330 327 359 324 329 357 328 323 350 322 Adj. R2 0.82 0.72 0.82 0.80 0.71 0.80 0.80 0.68 0.80 0.78 0.67 0.78 F Stat 69.23 39.50 68.35 60.41 37.42 59.50 47.10 24.95 46.85 41.20 23.34 40.99 DW 1.85 1.89 1.87 1.82 1.88 1.84 1.95 1.97 1.95 1.97 1.96 1.97 F Stat (Fixed effect) 5.55 4.888 5.63 5.33 4.92 5.32 3.04 3.15 3.23 2.79 2.90 2.94 p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
- Level of significant is indicated by asterisks : * 1%, ** 5% and *** 10 % - Standard errors white consistent covariance matrix are reported in parentheses
rightists15. The parties’ platforms are not necessarily related to certain economic issues. According to Shi and Svensson (2003), the ideological preference (the partisan model) which has been partly successful in explaining the macroeconomic fluctuation in OECD countries where a party’s social and economic orientation can be relatively easily identified is unlikely useful to explain electoral policy cycles in developing countries where the differences in economic and ideological preferences among parties are much harder to pin down and the distinction frequently does not exhibit the typical western left right pattern. Furthermore, in fact, no study on the experience of developing countries used a partisan model.
Table 8
Ideology on Budget Deficit (1985 – 2006)
Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) N = 33 (1991 – 2006) Variables (1) (2) (3) (4) DEFL 0.57* 0.53* 0.53* 0.48* (0.05) (0.05) (0.07) (0.06) GDP 0.17* 0.16* (0.03) (0.03) INF 0.01 0.00* (0.00) (0.00) GDPL 0.11* 0.10* (0.03) (0.03) INFL 0.02* -0.00 (0.00) (0.00) ICG -0.10 -0.08 0.02 -0.04 (0.17) (0.18) (0.17) (0.20) OBS 247 242 239 233 Adj. R2 0.82 0.81 0.78 0.77 F Stat 63.92 57.35 36.78 33.18 DW 2.10 2.04 2.09 1.98
F Stat (Fixed Effect 4.06 4.28 2.78 2.97
P value 0.00 0.00 0.00 0.00
Level of significant is indicated by asterisks : * 1%, ** 5% and *** 10 % Standard errors white consistent covariance matrix are reported in parentheses
To add the explanation, in a relatively new democracy, parties use voters’ sentiments rather than specific programs to gain votes in the elections. Moreover, in some countries (particularly in countries from the former Russia), democracy is only recently established so that the new established parties in those countries are less likely to have a
15 The left and right wings definition is according to The Database of Political Institution (Beck, et.al,
well-defined economic platform. We also note that there are many missing values for this variable so that the number of observations drops so much. Because of this, we will exclude ICG in the overall model.
Next, we will evaluate the existence of political budget cycles in Asia. It is expected that during election years, budget deficits become higher and vice versa. According to Brenden and Drazen (2005), fiscal manipulations only make sense to occur in countries where elections are competitive. In other words, we expect that it is more likely to occur in democratic countries. Considering the argument, as our initial samples include both democratic and non democratic countries (i.e Brunei and Bhutan), we filter our sample by including only the countries which have score for democratic measure above 0 (DEM>0).
Table 9
Political Budget Cycle on Budget Deficit (1985 – 2006) Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) N = 30 (1991 – 2006) Variables (1) (2) (3) (4) DEFL 0.63* 0.58* 0.65* 0.62* (0.05) (0.04) (0.05) (0.04) GDP 0.22* 0.17* (0.02) (0.03) INF 0.07*** -0.01 (0.04) (0.04) GDPL 0.14* 0.08* (0.04) (0.02) INFL 0.07** -0.02 (0.03) (0.00) ELEC -0.69** -0.82* 0.24 0.06 (0.31) (0.16) (0.56) (0.48) Obs. 192 191 176 175 Adj. R2 0.88 0.86 0.85 0.84 F Stat 93.49 76.72 57.69 51.11 DW 2.19 2.05 2.04 2.01
F Stat (Fixed Effect 4.62 4.18 2.66 2.38
p value 0.00 0.00 0.00 0.00
The results reported in Table 9 show that ELEC is significant and negatively related to DEF only in sample 2216. In other words, the evidence of political business cycles is found only in sample 22 not in sample 30. This implies that political budget cycle exists in democracies in Asia over 1985-2006, although the result is sensitive to the time period covered in each sample17. In a longer time span, it is likely more elections are to be held in each country. The more elections are held in each country, the more likely we may see their impact to budget deficits. Moreover, in sample 30 we include 8 newly established countries (from the former Russia) which are likely to have fewer elections than those older countries in sample 22.
Finally, we construct a final model that includes all political variables. As we include ELEC, the samples are filtered to democracy. The results are reported in Table 10.a for sample 22 and Table 10.b for sample 30.
In Table 10.a, the adjusted R square is relatively high (in a range of 0.86 to 0.90). It means around 86 – 90 percent variation of the dependent variable can be explained by our explanatory variables. DEFL remains significant (at 1% level of significant) and is positively related to DEF. The estimated coefficients of DEFL are also relatively high (in a range from 0.58 to 0.64). It implies that a large part of previous budget deficits persist in the next period. GDP and INF are also positive and significant at 1% level of significant and 10% level of significant respectively.
Meanwhile, our political explanatory variables are rather unsuccessful in explaining budget deficits. Variables related to the political system (PRES and DEM) are not significantly different from zero. Variables related to government fragmentation, namely; ENPG, CHECKS and MAJ, are also not significant. In contrast, variable measuring the political budget cycles; ELEC, is significant (at 1% level of significant and 5% level of significant) and as expected negatively related to DEF. It implies that during election years budget deficits tend to increase.
16 We also conduct regressions without democracy filter. The results are the same. However, the
estimated coefficients of ELEC in sample 22 are lower (-0.57 for regression with economic variables and -0.59 for regression with lag of economic variables). The adjusted R squares are also lower.
17 We conduct regressions for sample with 22 countries over 1991 – 2006 to know what causes the
Regression results with the lag of economic explanatory variables are reported in column (7) to column (12). DEFL remains significant (at 1% level of significant) and is positively related to DEF. GDPL and INFL are positive and significant at 1% level of significant and at 5% level of significant respectively. The same as before, only ELEC is significant at 1% level of significant and is negatively related to DEF.
The regression results also show that the estimated coefficient of ELEC is higher than the estimated coefficient of GDP and INF. It is close to the estimated coefficient of DEFL, even higher when we include the lag of economic explanatory variables. It implies that the effect of political budget cycles is stronger than the effect of economic performance. We may suggest budget deficits are more likely to be driven by political forces in our sample 22
Table 10.a
Economic and Political Determinants of Budget Deficits in Asia Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 22 (1985 – 2006) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) DEFL 0.63* 0.63* 0.63* 0.63* 0.64* 0.64* 0.63* 0.63* 0.64* 0.58* 0.58* 0.58* 0.58* 0.59* 0.59* 0.58* 0.58* 0.59* (0.22) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.04) (0.04) (0.05) (0.05) (0.04) (0.04) (0.04) (0.04) (0.04) GDP 0.22* 0.22* 0.22* 0.22* 0.22* 0.22* 0.22* 0.22* 0.22* (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) INF 0.07*** 0.07*** 0.07*** 0.07*** 0.07*** 0.07*** 0.07** 0.07*** 0.07** (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) GDPL 0.14* 0.14* 0.14* 0.14* 0.14* 0.14* 0.13* 0.13* 0.14* (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) INFL 0.07** 0.07** 0.07** 0.07** 0.07** 0.07** 0.06*** 0.06*** 0.07** (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) PRES0 0.13 0.17 0.11 0.08 0.08 0.07 (0.17) (0.16) (0.17) (022) (0.20) (0.22) PRES1 -0.13 -0.17 -0.11 -0.08 -0.08 -0.07 (0.17) (0.16) (0.17) (0.22) (0.20) (0.22) DEM 0.03 0.02 0.05 -0.06 -0.06 -0.04 (0.09) (0.09) (0.09) (0.11) (0.11) (0.11) ENPG 0.05 0.05 0.05 0.08 0.08 0.08 (0.07) (0.07) (0.07) (0.09) (0.09) (0.09) CHECKS -0.02 -0.02 -0.01 -0.00 -0.00 -0.00 (0.03) (0.03) (0.03) (0.02) (0.02) (0.03) MAJ -0.12 -0.12 -0.08 -0.14 -0.14 -0.20 (0.67) (0.67) (0.66) (0.86) (0.86) (0.84) ELEC -0.70* -0.70* -0.70** -0.70** -0.70** -0.70** -0.68** -0.69** -0.68** -0.82* -0.82* -0.82* -0.82* -0.83* -0.83* -0.82* -0.82* -0.83* (0.32) (0.32) (0.32) (0.32) (0.32) (0.32) (0.31) (0.30) (0.31) (0.17) (0.17) (0.16) (0.16) (0.17) (0.17) (0.17) (0.16) (0.16) OBS 192 192 190 190 191 191 192 190 191 191 191 190 190 190 190 191 190 190 Adj. R2 0.88 0.90 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.86 0.86 0.85 0.85 0.86 0.86 0.86 0.85 0.86 F Stat 81.68 81.68 80.91 80.91 81.19 81.19 81.74 80.93 81.32 67.13 67.13 66.29 66.29 66.78 66.78 66.29 66.47 66.86 DW 2.19 2.19 2.18 2.18 2.22 2.22 2.19 2.18 2.23 2.05 2.05 2.04 2.04 2.08 2.08 2.05 2.04 2.08
F Stat (Fixed Effect) 4.52 4.40 4.28 4.19 4.47 4.38 4.42 4.02 4.36 4.16 4.00 3.97 3.85 4.12 3.95 4.19 4.02 4.13
p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Table 10.b reports our regression results for sample 30. The adjusted R square is relatively high in a range of 0.84 – 0.87. As expected, DEFL remains highly significant and is positively related to DEF. The estimated coefficients of GDP and GDPL are still positive and significant at 1% level of significant. INF and INFL show an erratic pattern.
Concerning political variables, DEM is strongly significant (at 1% and 5% level of significant) but with the wrong sign (positive). ENPG surprisingly becomes significant (at 1%, 5% and 10% level of significant) and as expected it is negatively related to DEF. PRES0 and PRES1 become insignificant (only significant in column (10) (11) (14) and (15)). The other variables are insignificant.
The results generate some interesting results. First, in the overall model, DEM remains significant (consistent with our result in Table 6) but in contrary with positive sign. To check this, we conduct regression without democracy filter. We find that DEM is also significant and as expected negatively related to DEF. It may imply that democracy is positively related to budget deficits when it is associated with democracy-autocracy comparison. In other words, comparing to an autocracy, a democracy has higher budget deficits. However, the relation is reversed when the comparison is between less democratic and more democratic countries. The more democratic countries have lower budget deficits than the less democratic countries.
Second, PRES0 and PRES1 now become insignificant which are inconsistent with our results in Table 6. To check this, we conduct regressions without filter and find that PRES0 and PRES1 are significant (consistent with our results in Table 6). This may imply that the exclusion of non democracies in the overall regression affects the results. Third, in contrary to our results in Table 7, ENPG now becomes significant. Therefore, we conduct regressions without democracy filter and find that ENPG is insignificant18. It may suggest that the significance of ENPG is due to the exclusion of non democracies. These results are sensible as the competitive parties are more likely to find in democracies.
In short, the results when we include all political variables in sample 22 are consistent with the results when we include certain political variable in each regression. It may suggest that the results are quite robust and the exclusion of non democratic regimes does not influence the results. In contrary, the results of the overall model of sample 30 reported in Table 10.b seem not consistent with our previous results. However, we have to note that in the overall model we use a democratic filter. Therefore, we also conduct regression with filter by including only certain political variables in each regression and without filter by including all political variables. We find that the results are consistent with the results reported in Table 10.b
Furthermore, we have seen that the results reported in Table 10.a and Table 10.b differ substantially. To check whether this is due to the inclusion of certain countries or the period covered in each sample, we conduct regressions for sample 22 with shorter time period. We find that the difference is likely due to time period covered.
Table 10.b
Economic and Political Determinants of Budget Deficits in Asia Dependent Variable : Budget Balance to GDP Ratio
Explanatory N = 30 (1991 – 2006) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) DEFL 0.65* 0.65* 0.58* 0.58* 0.65* 0.65* 0.63* 0.56* 0.63* 0.63* 0.63* 0.56* 0.56* 0.63* 0.63* 0.61* 0.54* 0.62* (0.05) (0.05) (0.07) (0.07) (0.05) (0.05) (0.05) (0.07) (0.05) (0.04) (0.04) (0.06) (0.06) (0.04) (0.04) (0.04) (0.06) (0.04) GDP 0.17* 0.17* 0.19* 0.19* 0.17* 0.17* 0.17* 0.19* 0.17* (0.03) (0.03) (0.02) (0.02) (0.03) (0.03) (0.04) (0.03) (0.04) INF -0.01 -0.01 0.08** 0.08** -0.01 -0.01 0.00 0.08** -0.00 (0.04) (0.04) (0.04) (0.04) (0.03) (0.04) (0.04) (0.04) (0.04) GDPL 0.08* 0.08* 0.12* 0.12* 0.08* 0.08* 0.09* 0.13* 0.09* (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) INFL -0.02* -0.02* 0.05 0.05 -0.02* -0.02* -0.01* 0.06 -0.01* (0.00) (0.00) (0.04) (0.04) (0.00) (0.00) (0.00) (0.04) (0.00) PRES0 -0.47 0.03 -0.57 -0.60* -0.10 -0.68* (0.32) (0.25) (0.39) (0.16) (0.20) (0.24) PRES1 -0.47 -0.03 0.57 0.60* 0.10 0.68* (0.32) (0.25) (0.39) (0.16) (0.20) (0.24) DEM 0.36* 0.32* 0.34* 0.27** 0.33* 0.25** (0.11) (0.08) (0.11) (0.11) (0.10) (0.11) ENPG -0.13*** -0.13*** -0.14** -0.13** -0.13** -0.14* (0.08) (0.08) (0.07) (0.05) (0.05) (0.05) CHECKS -0.04 -0.04 -0.05*** -0.03 -0.03 -0.04 (0.02) (0.02) (0.03) (0.02) (0.02) (0.03) MAJ -0.75 -0.75 -0.05 -0.57 -0.57 -0.04 (1.22) (1.22) (1.24) (1.13) (1.14) (1.17) ELEC 0.24 0.24 0.11 0.11 0.24 0.24 0.26 0.16 0.28 0.07 0.07 -0.14 -0.14 0.08 0.08 0.07 -0.12 0.09 (0.56) (0.56) (0.54) (0.54) (0.54) (0.54) (0.55) (0.52) (0.53) (0.46) (0.46) (0.40) (0.40) (0.44) (0.44) (0.46) (0.38) (0.44) OBS 176 176 170 170 175 175 176 170 175 175 175 170 170 174 174 175 170 174 Adj. R2 0.85 0.85 0.87 0.87 0.85 0.85 0.86 0.87 0.85 0.84 0.84 0.84 0.84 0.84 0.84 0.84 0.85 0.84 F Stat 51.93 51.93 55.50 55.50 51.07 51.07 53.53 57.07 52.27 46.04 46.04 45.99 45.99 45.20 45.20 46.64 47.12 45.60 DW 2.06 2.06 2.11 2.11 2.07 2.07 2.02 2.09 2.01 2.04 2.04 2.02 2.02 2.04 2.04 2.02 2.03 2.00 F Stat (Fixed Effect) 2.53 2.36 3.49 3.39 2.63 2.46 2.83 3.57 2.79 2.36 2.18 2.79 2.69 2.34 2.17 2.50 3.04 2.45 p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
