Voters as Fiscal Liberals

Transcription

1 Preliminary Draft May 16, 2007 Voters as Fiscal Liberals Mark P. Jones Osvaldo Meloni Mariano Tommasi Rice University Universidad Nacional de Tucumán, Argentina Universidad de San Andrés, Argentina Abstract This paper presents new evidence on the effects of fiscal performance on the incumbent s party probability of reelection. We study all gubernatorial elections in Argentina from the recovery of democracy in Unlike Peltzman (1992) findings for the U.S. and more recent evidence for Israel and Colombia, we find that voters are not conservatives but liberals. They reward fiscal profligacy. After controlling for political and socioeconomic variables, we find that incumbent parties that increase real per capita public expenditures augment their probabilities of retaining in power. We conjecture that the tax sharing agreement and bailouts provide powerful incentives to voters to reward local politicians who are effective at extracting resources from the federal government. Keywords: elections, public expenditure, fiscal performance JEL Classification Codes: D72, P16

2 Preliminary Draft May 16, 2007 Voters as Fiscal Liberals Mark P. Jones Osvaldo Meloni Mariano Tommasi Rice University Universidad Nacional de Tucumán, Argentina Universidad de San Andrés, Argentina In our day we have seen great things accomplished by none except men who have been reputed parsimonious, the other have been driven from power. Niccolo Machiavelli The Prince Bantam Books, Page 58 I. Introduction In one of the most vivid paragraph of The Prince, Niccolo Machiavelli advice the prince to be parsimonious in its expenditures to avoid been overthrown from power. Machiavelli s argument is simple: if the prince burdens his people to keep his reputation of generosity will be termed as odious and will be exposed to danger at the slightest sign of trouble. Extrapolating this recommendation to the democratic ground seems, at least, risky for the prospects of reelection of the incumbent s party. The common wisdom suggest authorities to increase government spending, particularly in the proximity of elections, to improve their chances (or their party chances) of reelection. Several theories explain why voters reward fiscal profligacy; for instance, Buchanan s theory of fiscal illusion sustains that constituencies support politicians that promise new expenditure programs financed through higher taxes and debt because they fail to perceive them or just underestimate them. Adverse selectiontype models like Rogoff and Sibert (1988) and moral hazard-type models as in Person and Tabellini (2000) also predict that incumbents will undertake expansionary fiscal policy although not necessarily will be rewarded by voters. While in the first type of models the incumbent tries to signal her aptitudes for the government administration by providing more public goods to an uninformed electorate; in the second paper the incumbent is able to show 1

3 increases in public expenses to gain the electorate favor and delaying or hiding transitorily the correspondent augment in taxes or borrowing. Despite these theories look convincing and also rooted in the common sense, the empirical evidence is not so conclusive about constituencies recompensing public sector growth. On the contrary, a number of country studies find that voters penalize incumbent s loose fiscal policy at the polls. Papers by Peltzman (1992) Alesina, Perotti and Tavares (1998), Brender (2003) and Drazen and Eslava (2005) report decreasing incumbent s share of votes as government spending or fiscal deficits rise. Their result is attributed to voters preferences. For unrevealed reasons, voters prefer frugal governments instead of wasteful ones. Our paper presents new evidence on the effects of fiscal performance on the incumbent s party probability of reelection. We study all gubernatorial elections in Argentina from the recovery of democracy in The case of Argentina is particularly interesting since its central rule (that is, the constitution) resembles that of the US and other developed federal republics, but second order laws, like the tax-sharing system, that distributes tax revenues among the federal government and the provinces, encourage political actors, particularly governors, to engage in expenditure hiking to get the support of the electorate. The taxsharing system generates a common pool of resources that is distributed in a first stage among the federal government and the provinces, and in a second stage among the 23 provinces. This mechanism produces high degree of vertical imbalance in most of the provinces, which in turn weakens citizen s incentives to control both, the level and composition of provincial outlays. Governor s incentives to carry out loose fiscal policies are reinforced by the high probability of getting federal government bailouts in the face of financial distress. As reported in Nicolini et al. (2002) the overwhelming evidence shows that higher levels of government are likely to bailout lower levels of government, stimulating rather than discouraging imprudent fiscal behavior. Clearly, financial flows from central to sub-national governments soften the provincial budget constraints and exacerbate the already high level of vertical imbalance. We can conjecture that voters facing less than the full cost of domestic public goods are less likely to remove inefficient incumbent parties. Furthermore, there is a chance that they will behave as fiscal liberals rather than fiscal conservative ones. The paper is organized as follows: In the next section we discuss the previous empirical literature. Section III analyzes Argentine fiscal performance and fiscal incentives for both, local incumbent parties and their electorate. Section IV is dedicated to data description and the empirical specification while section V presents the results of the empirical tests. Finally, section VI concludes. 2

4 II. Evidence on voting behavior: Fiscal liberals vs. fiscal conservatives Do voters prefer fiscal liberal incumbents? It is widely whispered that constituencies favor high spending governments. Moreover, fiscal adjustments are believed to be unpopular, particularly in Latin America, where the sole mention of the word might provoke strong reactions against the government. Nonetheless, in a recent survey, Eslava (2006) concludes that incumbents who adopted loose fiscal policies do not receive greater voter support than fiscally conservative ones. Cross country studies as well as case studies for both, developed and emergent economies, show either electorate proclivity to favor fiscal conservative incumbents or resistance to support fiscal liberals. When it comes to fiscal conservative voters, the obliged reference is Peltzman (1992) whose study covering U.S. presidential, senatorial and gubernatorial elections for most of the second half of the 20 th century concludes that voters penalize federal and state spending growth. The reason is straightforward: well-informed, self-interested voters oppose marginal expansions of government budgets because federal and state fiscal systems are progressive, and voters are wealthier than non-voters. Another paper that deserves close attention is Brender s case study of local elections in Israel. Brender (2003) finds that fiscal performance of mayors, characterized by high current deficits and large accumulation of debt significantly reduces their probability of reelection in the 1998 voting. This result is attributed to the changes in the political environment, most of them staring in 1994, like enforcement of audit and financial reporting requirements, tougher imposition of hard budget constraints and the development of local media. Interestingly, the 1989 and 1993 elections give the opposite results although the estimated coefficients were statistically not significant. Cross-country studies also report evidence favoring fiscal prudent incumbents. Alesina, Perotti and Tavares (1998) analyzed OECD countries fiscal behavior and find that voters did not punish governments that implemented sound fiscal policies. Similarly, Brender and Drazen (2005a) work with a sample of 74 countries over the period and find no evidence that fiscal deficits help incumbent s reelection. For some group of countries deficits lower the probability of reelection and for other have no statistically significant effect. In an attempt to explore further the reaction of the electorate to the incumbent outlays stimulus, Drazen and Eslava (2005) classify public expenditures into two groups, those that are targeted to voters (mainly investment) and those that are not. They study municipalities in Colombia for the period reporting that the share of votes received by the incumbent party in Colombian local elections increases with investment expenditures but 3

5 only to the extent that they do so without running large election-year deficits. The authors interpret this finding as supporting the voter s fiscal conservatism hypothesis. In general, these results are attributed to voter s preference for tight fiscal policies and to the transparency prevalent in developed and consolidated democracies, which prevent incumbents to deceit voters. Hence, the places to look for voters support to incumbents fiscal profligacy are new democracies given their relative lack of transparency. In that sense, Israel as well as Colombia should be considered established democracies. In the first case democracy exist since the creation of the State of Israel, while Colombia has the longest uninterrupted democracy in Latin America. This comment is pertinent to interpret Akhmedov et al. (2002) results in their paper on Russia s young democracy for the years They find that incumbent governors practicing opportunistic loose fiscal policies increase political popularity and chances of reelection. The authors conclude that the maturity of democracy in Russia as well as rationality and awareness of the electorate are key factors to determinate the scope for opportunistic cycles. Similar conclusions are drawn by Goncalves Veiga and Veiga (2006) from a panel of 275 Portuguese municipalities from 1979 to That is, mayors opportunism leads to more votes. Nonetheless, in contrast with Akhmedov et al., in the early stages of democracy (elections from 1979 to 1989) opportunism has no effect on electoral results while in the more recent ballots, (1993, 1997 and 2001) incumbents spending paid off, contradicting also Brender and Drazen (2005a) findings. The only paper that deals with Argentina s recent democracy is Porto and Porto (2000) that studies 125 municipalities belonging to the Province of Buenos Aires (the largest in Argentina) They found that the probability of changing the incumbent party depends negatively on the variation in capital expenditures which is interpreted as supporting Peltzman s results for the U.S. III. Fiscal performance and fiscal incentives in Argentine provinces 1 In Argentina, national and sub national politics and policies are intertwined to a much larger (and convoluted) extent than in other federal polities. The main links are electoral and fiscal. Provinces undertake a large fraction of total spending; yet collect only a small fraction of taxes. Provincial spending amounts to 50 percent of total consolidated public sector 1 This section borrows freely from Spiller and Tommasi (2007) 4

6 spending. This figure rises to close to 70 percent if we exclude the pension system and focus on more discretionary spending. Furthermore, the type of spending in the hands of provincial governments tends to be politically attractive (such as public employment and social programs) because it is close to the interests of territorially based constituencies. Yet, on average, provinces finance only 35 percent of that spending with their own revenues. The rest of their spending is financed from a common pool of resources, according to the Federal Tax-Sharing Agreement. In a large number of small provinces the proportion of funds from this common pool constitutes over 80 percent of their funding. Local politicians, then, enjoy a large share of the political benefit of spending and pay only a small fraction of the political cost of taxation. This fiscal structure at the provincial level is one reason why many professional politicians are more interested in pursuing a career through appointments in the provincial government (or even the party at the provincial level) than in the National Congress. But the powerful provincial brokers that is, the governors depend heavily on the allocation of central monies to their provinces to run both their political and their policy businesses. That is, they need central money to deliver particularistic political goods, as well as to provide general public goods in their province. There are several channels for funneling funds to the provinces, the main ones being the geographic allocation of the national budget and the Federal Tax-Sharing Agreement. The game in which these allocations are determined is the source of many political and policy distortions, at both the national and the provincial level. The game even affects the quality of democracy at the local level. The Argentine voter at the provincial level has an incentive to reward politicians who are effective at extracting resources from the center. These are not necessarily the most competent or honest administrators. Given the political mechanisms by which funds are allocated, this also adds uncertainty to provincial public finances, since it is not easy to project future allocations. The history and evolution of the Tax-Sharing Agreement is fraught with examples of opportunistic manipulation, occasionally curtailed by fairly rigid and inefficient mechanisms (Tommasi, 2006, and Iaryczower, Saiegh and Tommasi, 1999). Unilateral, bilateral, and coalitional opportunism (by the national government, by a province, or by a set of provinces that turns out to be pivotal for an important vote in Congress or for some other reason) has been common in the allocation of central government monies to the provinces. The national executive has enjoyed substantial discretion to allocate items in the federal budget geographically (See Bercoff and Meloni, 2007). In an attempt to prevent adverse changes in the future (for instance, a reduction in the amounts going to any specific province), political 5

7 actors have tended to impose greater rigidity on the Tax-Sharing Agreement, reducing the government s capacity to adjust fiscal policy to changed economic circumstances. One example is the earmarking of taxes for specific programs with clear regional distributional effects. This earmarking led to a rigid and convoluted system of federal tax collection and distribution, which has been christened the Argentine fiscal labyrinth. Recent attempts to simplify that labyrinth, which also reflect the inability to strike efficient intertemporal agreements, led to the 1999 and 2000 fiscal pacts between the national and provincial governments. Those pacts generated a rigid commitment to a minimum of revenues from the center to the provinces, which turned out to be a very costly straightjacket for the De la Rúa government during the lead-up to the 2001 crisis. Similarly, the lack of cooperation from the provinces has been credited as the immediate cause of the country s move to default (see Eaton 2005, and Tommasi 2006). 2 That episode of the Argentine federal fiscal drama, which led to one of the largest defaults in modern world economic history, was a clear manifestation of one of the central points in our argument. Provincial governors, who are crucial players in national politics and policymaking, have only secondary interest in national public goods (such as macroeconomic stability), in the quality of national policies, and (hence) in investing in institutions (a professional Congress, a stronger civil service) that might improve the quality of policies. The primary interest on the basis of which they grant or withdraw support to national governments and their policies is the access to common-pool fiscal resources. Other source of sub-national fiscal expansion is found in bailouts that introduce modifications in the provinces budget constraints. Provincial governments could behave opportunistically if they anticipate that federal authorities have ex-post incentives to deviate from implicit or explicit ex-ante rules. In Argentina, this opportunistic behavior usually takes the form of unsustainable fiscal behavior that increases the district s exposure to a crisis in the event of an exogenous shock. As shown by Nicolini et al. (2002) that analyzed salient episodes of bailouts during the 90 s, small and poor jurisdictions rather than larger ones benefited from federal money in times of financial distress. Summarizing, when voters evaluate candidates running for governor, prodigal rather than parsimonious incumbents rank high. The tax sharing agreement and bailouts provide powerful incentives to voters to reward politicians who are effective at extracting resources from the federal government. 2 The buildup of a very dangerous aggregate fiscal position in the late 1990s was also related to political gaming between the national government and the provinces (Tommasi 2006). 6

8 IV. Data Description and empirical specification Our dependent variable, WINNER, is a binary variable with the value 1 if the incumbent party wins the gubernatorial election and 0 if it is defeated. Absolute victory, not the change in the percentage of the vote won, is employed as the dependent variable for three principal reasons. First, the goal of candidates in gubernatorial elections is to win the election, not necessarily to maximize the percentage of the vote received. In Argentina, for instance, it is not uncommon for gubernatorial candidates who are confident of victory in the general election to lend material and logistical support to an opposition candidate (where the same funds can achieve more votes than if used in support of the governor's party) with the goal that this candidate's coattails will provide his/her party with seats in the legislature, seats that will be occupied by legislators who in turn will support the governor. At the same time, governors who are confident of victory often do not expend all available resources on the election campaign (preferring to conserve them for post-election bargaining purposes or utilize them to gain the support of co-partisans). Sometimes, it pays (for incumbent governors) to spend resources at splitting the opposition (or avoid the coalition of opposition parties) rather than use them up in trying to improve or equate their previous results at the polls. Second, inter-party alliances are very common in Argentina. Furthermore, rarely is the alliance of parties that supported a candidate in one election identical to that in the election immediately anterior or posterior to that election. As a consequence calculating vote shifts is at times anything but straightforward. Third, all Argentine gubernatorial elections feature more than two candidates, and many involve three or four viable candidates. As a consequence the vote-based measure most commonly employed in the United States (share of the two-party vote) is not easily transferred to this multi-party context. While in the United States a vote increase for the Democratic candidate (compared to the previous election) is virtually synonymous with a vote decrease for the Republican candidate, within a multi-party context such logic does not hold. Considering the total number of observations (112), the incumbent party that increases the percentage of votes with regard to the previous election was reelected in 51 elections (45.5%) and was defeated in 5 ballot vote (4.5%). That is, improving the performance of the previous election does not guarantee the victory. Furthermore, a decline in the percentage of votes does not mean to cede the administration to the opposition: in 39 out of 112 elections (34.8%) the incumbent party diminished the percentage of votes but won the election. In the remaining 17 ballots (15.2%), the incumbent party lost popular support and also lost the gubernatorial election. 7

9 Since the return to democracy in 1983, Argentina had gubernatorial elections regularly in most of its 24 districts (23 provinces plus the federal district) every four years: 1987, 1991, 1995, 1999 and The exceptions, which originate missing data in our panel, are the districts of Corrientes, Tierra del Fuego and the City of Buenos Aires (Federal District). In the case of the Province of Corrientes we only consider the elections of 1983, 1987 and 1991 since provincial authorities were intervened twice by the federal government in the period , amounting for four years of intervention 3. The cases of Tierra del Fuego and the Federal District are different: their people elected directly their governors for the first time in 1991 and 1996 respectively. Before those dates, the President of the Nation appointed the governors 4. Notice that we took the 1983 election as an starting point for our empirical analysis, since before that date the military was in charge and there was no party allied with the military regime so we cannot treat any party as the incumbent in that election. Table 1. Incumbent Party Victories in the Argentine Districts Party 5 out of 5 elections Peronist UCR/Alianza Provincial Number of Districts Formosa, Jujuy, La Rioja, La Pampa, San Luis, Santa Fe, Santiago del Estero Federal District*, Río Negro Neuquén The Incumbent party of the following districts won: 4 out of 5 elections Buenos Aires (4), Catamarca (1), Córdoba (1), Chaco (1), Entre Ríos (2), Mendoza (2), Misiones (4) Catamarca (3), Córdoba (3), Chaco (2), Entre Ríos (1), Mendoza (1) 3 out of 5 elections Salta, Tucumán 2 out of 5 elections Chubut San Juan (1) Corrientes**, San Juan (1) 1 of 5 elections*** Tierra del Fuego*** Notes: number of elections won by the respective incumbent party in parenthesis * Before 1996 the Chief of Government of the Federal District used to be appointed by the President of the Nation, so only two elections are considered (besides the initial point, 1996): 2000 and ** Due to federal interventions we only consider two elections, 1987 and 1991, with 1983 as an starting point. *** Tierra del Fuego acquired the status of province in 1991, so only three elections are taken into account, 1995, 1999 and 2003 (with 1991 as starting point) The importance of the incumbency in the Argentine provinces is reflected in the fact that the incumbent party won 80% of the 112 gubernatorial elections analyzed. Moreover, in 11 out of 24 districts, the incumbent party always won the gubernatorial elections. The Peronist party remains undefeated in 8 provinces while the UCR and the provincial party Movimiento 3 In the period considered, the provinces of Tucumán and Santiago del Estero were also federal intervened. In both cases the intervention lasted a few months so the electoral calendar was not altered. 4 Tierra del Fuego had the status of National Territory, dependent of the Federal government, until

10 Popular Neuquino always won the elections in the provinces of Río Negro and Neuquén respectively. The case of the Federal District deserves some comment: the incumbent party managed to remain unbeaten appealing to alliances with incumbent Presidents de la Rúa (UCR party) and Kirchner (Peronist party). Table 1 shows the record of victories of the incumbent parties in the provinces. Interestingly, incumbent governors that run for reelection obtained the victory in 91% of the cases. It is worth remarking that some provinces have constitutional provisions that ban immediate reelection. Nonetheless, during the 90s some provinces reformed the constitution to allow incumbent governor reelection. Our key explanatory variable, the growth rate of the real per capita public expenditure moves procyclically: for the majority of provinces expansions are characterized by large increases in public expenditure and recessions by huge declines in that variable. As noted earlier, besides political incentives (for instance, incumbent party reelection), the combination of tax sharing system ( coparticipación federal de impuestos ) and procyclical revenues plus the chance of borrowing overseas fuel public expenditures during expansions. In turn, this increase in public expenditure contributes to further augment the rate of growth of the GDP. On the contrary, reversals in international capital flows, decline in provincial revenue collection and fall in transfers from the national government are the main forces behind the diminishing expenditures. The behavior of the provincial real per capita public expenditure in each gubernatorial period shows high dispersion which might indicate autonomous conduct of governors, although we still can trace out some impact of the overall country performance (see Table 2). During the 80 s the argentine economy enjoyed a modest expansion after the stabilization plan called Austral, in 1985, which may explain the high rates of growth of public expenditures (more than 20%) in eleven provinces in the period Conversely, in the years , featuring hyperinflation and recession in the country, twenty-one provinces had diminishing per capita public expenditure, with eighteen of them retreating more than 20%. Only the province of Jujuy managed to expand public expenditures under those circumstances. During the 90 s the business cycle favored expansion of the public expenditure. Well after the implementation of the currency board, known as Convertibility Plan, in 1991, the economy stabilized and began a rapid growth. Not surprisingly, in the period sixteen districts had growth rates of the per capita public outlays that rose above 30%. Surprisingly, among the few districts that increased real per capita public expenditure less than 20% was La Rioja, the home province of former president Carlos Menem. It is expected that small 9

11 districts be more incline to fiscal profligacy since they are more favored by the secondary coparticipation from the common pool of resources. Likewise, it can be anticipated a more prudent behavior in larger provinces because they internalize more the federal tax costs of their spending. Nonetheless, it must also be taken into account the national executive pressures to governors from the same party (as La Rioja) to have fiscal discipline in order to make them internalize some of these costs. Table 2. Number of Provinces with growth rate of Real per capita public expenditures in a given interval Real per capita public Expenditure Interval Gubernatorial period More than -20% From -11 to 19.9% From 10.9 to 0% From 0.1 to 10% From 10.1 to 20% From 20.1 to 30% More than 30% Total provinces 22* 22* 22** 23*** 23*** Notes: * Federal District and Tierra del Fuego were excluded ** Federal District and Corrientes were excluded *** Corrientes is excluded. After the short 1995 crisis, known as Tequila, the national GDP expanded again, which generated renewed impulse in the public expenditure of the majority of provinces in In opposition, the 2001/02 crisis, one of the most deleterious in the country s history, was the responsible of the huge adjustment in the public outlays of all provinces, which showed negative growth rates in the gubernatorial period Moreover, in 14 districts real per capita public expenditures declined more than 40% 5. Data The dataset used in this study comes from several sources. Real expenditures as well as real revenues at constant 2003 prices were drawn from the Secretaría de Hacienda, Ministerio de Economía de la Nación. Population data and provincial GDP at constant prices were estimated by the Universidad Nacional de La Plata, based on official census. The crime rate figures were collected from the Dirección Nacional de Política Criminal, Ministerio de Justicia y Derechos Humanos. The unemployment data were taken from the INDEC (the 5 Table 1A in the Appendix shows the coefficient of variation of per capita Spending for each province in

12 Argentinean Bureau of Statistics). Political and electoral data comes from Dirección Nacional Electoral. A detailed description of the data sources appears in the Appendix. Empirical specification Our basic estimating equation is: WINNER it = α 0 + α 1 POLITICAL it + α 2 SOCIOECONOMIC it + α 3 EXPENDITURE it + ε it We test for the effect of the growth rate of real per capita public expenditures on the incumbent party s probability of winning the gubernatorial election t in province i. We control for traditional socioeconomic influences, such as provincial GDP per capita and unemployment. We expect an increase in the incumbent party s probability of victory as provincial GDP per capita goes up and unemployment decreases. Another socioeconomic control is the rate of growth of property crime, associated negatively with the dependent variable. In the last decade or so, the growing unemployment and crime have become central topics for Argentineans, ranking at the top in the mass-media coverage. We also control for the business cycle and the inflation rate. We conjecture that a growing national economy improves the probability of victory of all provincial incumbent parties affiliated to the same party of the President. There are at least two channels through which a good economic performance of the country favors, in electoral terms, the provincial party affiliated at the president s party even though we assume that voters distinguish between the responsibilities of the federal and provincial government. The most obvious are discretionary transfers from the federal government to its allied provinces. Argentina s revenue collection features high procyclicality, which gives extra funds to the federal government that can be discretionary distributed among allied incumbents at the provinces. Voters expect the President to help provinces administered by allied governors not only with federal funds but also influencing on private investments or through regulations, particularly international trade policies. On the other hand, voters might be prone to support provincial incumbent party affiliated to the same party of the President, as a way to sustain a successful (in terms of GDP growth) national economic policy. Conversely, we speculate that high inflation rates negatively affect the electoral performance of provincial incumbent parties that respond to the President. To capture these effects, we construct two variables named CYCLE and INFLATION, that interact a dummy variable that takes the value 1 if the incumbent party is affiliated to the same party of the president and 0 otherwise, with the growth rate of the national GDP and the inflation rate, respectively. 11

13 Notice that neither the inflation rate nor the national GDP growth rate vary across the provinces governed by the same party as the President s, but the variables CYCLE and INFLATION do vary across provinces. It is also worth remarking that CYCLE and the growth rate of provincial GDP per capita are included to control for different effects on the incumbent s probability of victory. While the former attempts to capture the impact of the business cycle, as mentioned above; the last is a proxy for the rate of growth of personal income in a given jurisdiction. Moreover, the Pearson coefficient of correlation between both variables is not so high: The impact of political variables is measured by four variables: the normal share of votes of the incumbent party in a given province (PARTY), a dummy variable that is intended to reveal the extra effort that receives a party when the incumbent governor runs for the reelection (REELECTION), the margin of victory in the last gubernatorial election (MARGIN) and another dummy variable that tries to capture the influence of simultaneous elections for President and Governors (NATLCOAT). The normal share of votes of the incumbent party reflects various concepts such as political inertia, ideological affiliation and party loyalties. We approximate this idea by the share of votes in the congressional election held two years before the gubernatorial elections that is, 1985, 1989, 1993, 1997 and 2001 for most provinces. Congressional elections usually captures the party vote since candidates generally have low name recognition, so ballots are cast along party lines rather than personal prestige. We assume that the chance of victory of the incumbent party increases the more rooted the party loyalty. In the same line of reasoning, the incumbent party augments its probability of triumph, the larger the margin of victory in the past gubernatorial election. Large margins might indicate strong incumbents as well as disorganized oppositions. REELECTION assumes that incumbent governors running for reelection devote more effort and resources to remain in office than the ones constitutional impeded to dispute a new period or voluntary excluded 6. On the other hand, we expect the simultaneity of gubernatorial and presidential elections to be positively correlated with our dependent variable, not only because it biases the public discussion towards national topics of interest to the President, in detriment of provincial affairs, but also because it helps to exploit some economies of scale in the use of campaign resources. Before the constitutional reform of 1994, gubernatorial and presidential elections 6 Before 1983, all provinces had constitutional arrangements banning incumbent governors to run for an immediate reelection. Since then there were various provinces amending their constitutions to allow for governor s reelection. See Appendix. 12

14 coincided every twelve years as the President lasted six years in office and governors four years, but after the reform, the president period was shortened to four years, so, from 1995, both, presidential and gubernatorial elections are held the same year for the majority of provinces. Despite that, some provinces constitutions authorize governors to carry out the voting a few months before or after the Presidential election, avoiding simultaneity. Table 3. Definition of variables used in the regressions Abbreviation WINNER it REELECTION it MARGIN it SAME it PARTY it NATLCOAT it EXPENDITURE it GDPPC it UNEMPLOYMENT it: CRIME it INFLATION it CYCLE it Description Dummy variable that takes the value 1 if the incumbent party wins the gubernatorial election t in the province I, and 0 otherwise. Dummy variable that takes the value 1 if incumbent governor runs for reelection in the province i at the gubernatorial election t, and 0 otherwise. Difference between the percentage of votes obtained by the incumbent party and the main rival in the province i at the gubernatorial election t. Dummy variable that takes the value 1 when the incumbent governor is affiliated to the same party as the President and 0 otherwise. The case of the province of Catamarca was coded 1 despite it was intervened by the federal government, but the governor appointed by the Executive to intervene the province was very closed (politically) to the incumbent party. Percentage of votes obtained by the incumbent party in the representatives election carried out two years before the gubernatorial election t in province i. Dummy variable that takes the value 1 when gubernatorial elections are held the same date as presidential elections and the incumbent party is PJ (1995, 2003) or UCR/FREPASO (1999); -1 when gubernatorial elections are held the same date as presidential elections and the incumbent party is different from PJ or UCR/FREPASO, and 0 otherwise Growth rate of the real total expenditures per capita for province i between the end and the beginning of the gubernatorial term. Growth rate of the real provincial GDP per capita for province i in the last year of the gubernatorial term that finishes in election t. Rate of unemployment (level) in the year of the election t in province I growth rate of the property crime rate (offenses per 10,000 inhabitants) for province i in the last year of the gubernatorial term that ended in election t. Stands for the interaction between the variable Same and the annual inflation rate in the year of the election. results from multiplying the variable Same by GDP growth rate of the country in the Election Year. 13

15 Table 4 presents some descriptive statistics for the full sample, The annual average per capita expenditure is $ 2355 (argentine currency). There is, however, high variation in public outlays, which is reflected in the standard deviation of $ The province of Tierra del Fuego in the period spent, on average, $ 7942 per inhabitant, per year, which is the maximum of the sample while Buenos Aires in reached the minimum with $ per inhabitant per year. The rate of growth of real per capita public expenditures for the gubernatorial period also presents elevated volatility, with huge increases of as much as 76.5% registered by the province of Mendoza from 1991 to 1994 and tremendous falls like the one suffered by the province of La Rioja from 1999 to 2002 that reduced its public expenditures more than 50%. La Rioja finances only 10% of their spending with their own revenues. From 1989 to 1999 the federal government directed copious funds to the President s home province, which allowed extravagant expenditures in La Rioja. As the Menem s star extinguished the federal money diminished and so the provincial outlays. Table 4. Descriptive Statistics Variable Obs Mean Std. Dev. Min Max Dummies Winner Reelection National coat Rates of Growth GDP pc (%) Cycle (%) Inflation (%) Crime (%) Expenditure (%) Levels Margin (%) Party (%) Unemployment (%) Expenditure Level Note: All statistics are calculated on gubernatorial period. Table 1A in the Appendix shows summary statistics computed on the yearly basis. Fiscal Statistics for panel data observations for 24 districts are expressed in 2003 pesos. Crime rate is defined as number of offenses per 10,000 inhabitants. 14

16 V. Results Probit equations were estimated for a panel including five consecutive gubernatorial elections (1987, 1991, 1995, 1999 and 2003) and 24 districts 7. Results for our basic specifications are presented in Table 5. The first column shows only control variables, just to have a benchmark, while the ones numbered (2) to (5) incorporate also our key explanatory variable, the rate of growth of the real per capita public expenditure. In the last four regressions, control variables have the expected sign and they are statistically significant at usual levels except for Margin and National Coat in regression (2). As expected, high unemployment rates and rising property crime rates in the proximities of the gubernatorial elections definitively hurts incumbent party chances to win the election. On the contrary, good economic conditions measured by the rate of growth of the country GDP and also by the provincial GDP per capita help incumbent parties affiliated to the President s party to get reelected. While the significance of the inflation rate and the country GDP behavior might be interpreted as voters ignorance of the governor s responsibilities, it can also be viewed as a refusal to vote incumbent parties that support federal governments that implement wrong economic policies leading to recessions and inflation. These results indicate that Argentineans voters have short memory when casting the ballot for governors, that is, socioeconomic conditions prevailing in the year previous to gubernatorial elections, both at national and provincial level, are relevant to explain the probability of victory of the incumbent party. On the contrary, socioeconomic conditions computed along the full gubernatorial period do not have statistical significance. The growth rate of real per capita public expenditure is always significant and its inclusion definitively improves the fit. Of the regressions in Table 5, the third column is the one that fits the data best and is the one upon which we will base our conclusions. Regressions including the variable Reelection was not included because they resulted statistically non- significant 8. National Coattail also resulted non-significant at usual confidence level. This result can be termed as expected or at least not surprising given the peculiarities of Argentine politics described in section II. Is National Coattail still irrelevant if we take out the peculiar 2003 case (the fact that they were three different peronist candidates that time, plus the peculiar regional logic of that vote)? We run a new set of regressions excluding the year 2003 and our results remain unaltered 9. 7 We also estimated by the Logit procedure with negligible difference in the results. 8 The average expenditure level was also included as explanatory variable but results were not satisfactory. 9 See Appendix 15

17 Table 5. Explaining the Incumbent s probability of Victory in gubernatorial elections. Full Sample estimates. Variables Control Variables (1) Expenditures (2.42)** National Coat (1.58) Party (2.96)*** Margin (1.26) GDP per capita (1.18) Unemployment (electoral year) (-1.86)* Crime rate (-1.26) Cycle (2.09)** Inflation (-3.37)*** Constant (-0.44) (2) (3) (4) (5) (1.39) (2.34)** (1.56) (1.92)* (-2.00)** (-1.97)** (2.56)** (-3.17)*** (0.28) (2.56)*** (2.40)** (1.66)* (2.01)** (-2.04)** (-1.91)* (2.62)*** (-3.29)*** (0.17) (2.35)** (2.69)*** (1.95)** (-2.02)** (-1.71)* (2.49)** (-3.36)*** (0.46) (2.27)** (1.49) (2.60)*** (1.92)* (-2.04)** (-1.82)* (2.50)** (-3.30)*** Observations Districts Observations per group MIN Observations per group MAX Observations per group AVG (0.58) Log Likelihood Note: * significant at the 10% level; ** Significant at the 5% level; *** Significant at the 1% level. z statistics in parenthesis As a robustness check, we run regressions limiting the sample to those provinces with at least five periods of data, and the results are pretty similar. We also run logit fixed effects regressions, what diminishes the number of observations considerably, but still the rate of growth of per capita public expenditures is significant at usual statistical levels See Appendix 16

18 VI. Conclusions Voters reward fiscal profligacy. This is the main result of our paper. After controlling for political and socioeconomic variables, we find that incumbent parties that increase real per capita public expenditures augment their probabilities of remaining in power. Of course, the key is not explaining why politicians want to raise public outlays but why voters do not punish profligacy in Argentina as in the U.S. Israel and Colombia. In other words, what is the fundamental difference between US voters portrayed in Peltzman s paper and the Argentinean voters as appear in this paper? Apparently, none. Both respond to incentives. Both may even dislike prodigal politicians. The real difference is in the set of incentives they face. The argentine voter at the provincial level has incentives to reward politicians who are effective at extracting resources from the federal government. Voters, particular in small and poor districts, and to a lesser extent in large provinces, know that they are not going to pay the full cost of their local authorities imprudent fiscal behavior. The tax sharing agreement and the high probability of bailouts (decreasing in the size of the province) are powerful incentives to both, politicians to spend and to voters to reward big spenders. For example, voters know that their governor engages in unsustainable fiscal behavior that increases the district s exposure to a crisis in the event of an exogenous shock. Nonetheless, they may agree to reelect them. Why? At best, if the shock does not occur, they will enjoy from more public goods. At worst, if the shocks generate financial distress, voters know that they will not pay the full cost of the fiscal adventure. It is highly probable that the federal government will bailout the province, paying part (or even the total) of the costs. Rather than reflecting voters preferences, our results mirror the set of incentives faced by voters. 17

19 References Akhmedov, A. Ravichez, A. and Zhuravskaya, E. (2002) Opportunistic Political Cycles: test in a young democracy setting. Working Paper Alesina, A., Perotti, R. and Tavares, J. (1998) The political economy of fiscal adjustments. Brookings Papers on Economic Activity, 1. Bercoff J. and Meloni, O. (2007) Federal budget allocation in an emergent democracy. Evidence from Argentina. Mimeo. Brender A. (2003) The effects of fiscal performance on Local Government Election Results in Israel: Journal of Public Economics, Vol. 87, pp Brender, A. and Drazen, A. (2005a) How do Budget Deficits and Economic Growth Affect Reelection prospects? Evidence from a Large Cross-Section of Countries. NBER Working Paper. No Brender, A. and Drazen, A. (2005b) Political budget cycles in new versus established democracies. Journal of Monetary Economy. Drazen, A. and Eslava, M. (2005) Electoral manipulation via expenditure composition: theory. NBER Working paper Cambridge, MA National Bureau of economic Research. Eslava (2005) Political budget cycle or voters as fiscal conservatives? Evidence from Colombia Documento CEDE Febrero Eslava, M. (2006) The Political Economy of Fiscal Policy: Survey. Inter American Development Bank. October. Goncalves Veiga, L. and Veiga, F. (2006) Does Opportunity Pay Off? NIPE Working Paper 5. Jones, Mark, Sanguinetti, Pablo y Tommasi, Mariano (1997) Politics, Institutions, and Fiscal Performance in the Argentine Provinces. Universidad Torcuato Di Tella. Mimeo. Nicolini, J., Posadas, J., Sanguinetti, J, Sanguinetti, P., Tommasi, M. (2002) Decentralization, Fiscal Discipline in Sub-National Governments and the Bailout Problem: the case of Argentina. Inter- American Development Bank, Working Paper R-467. Peltzman, Sam (1987) Economic conditions and Gubernatorial Elections. American Economic Review Papers and Proceedings (May): Peltzman, Sam (1992). Voters as fiscal conservatives. Quarterly Journal of Economics. Vol. 107 No. 2 (May): Persson, T. and Tabellini, G. (2000) Political Economics: Explaining Economic Policy. MIT Press, Cambridge MA Porto. Alberto y Porto Natalia (2000) Fiscal Decentralization and Voters Choices as Control. Journal of Applied Economics. Vol. III No. 1: Rogoff, K. and Sibert, A.(1988) Elections and Macroeconomic Policy Cycles. Review of Economic Studies, Vol.55,

20 Appendix Table 1 A. Real Public Expenditures per capita per district: (ranked according to the coefficient of variation) District Average Standard Deviation Coefficient of variation Maximum Minimum Buenos Aires CABA Mendoza Córdoba Tierra del Fuego Salta Misiones Entre Ríos San Juan La Rioja San Luis La Pampa Santiago del Estero Chaco Río Negro Tucumán Catamarca Formosa Corrientes Santa Fe Jujuy Neuquén Chubut Santa Cruz Note: expenditures expressed in pesos of

21 Table 2 A. Logit fixed effects regressions. xtlogit w texppclevel party natlcoat texppclr gdppcsr uelecyear crimesr cyclesr infsrsp, nolog fe note: multiple positive outcomes within groups encountered. note: 12 groups (54 obs) dropped due to all positive or all negative outcomes. Conditional fixed-effects logistic regression Number of obs = 58 Group variable (i): id Number of groups = 12 Obs per group: min = 3 avg = 4.8 max = 5 LR chi2(9) = Log likelihood = Prob > chi2 = w Coef. Std. Err. z P> z [95% Conf. Interval] Expend level Party Natlcoat Expenditure GDP pc Unemployment Crime Cycle Inflation Expend level is the average real per capita total expenditures for province i during the gubernatorial term that ends in election t. 20