Evidence from French Municipalities

Opportunistic Policies and Election Outcomes: Evidence from French Municipalities Mamadou BOUKARI July 22, 2015 Abstract This paper implements a new test of the Rational Political Business Cycle (RPBC) proposed by Aidt et al. (2011), on French municipalities. Using GMM- System estimation, it investigates the presence of a two-way relationship between the win margin of victory and the size of the opportunistic distortion of fiscal policy. The results indicate that there is a non-monotonic relationship between the two variables. That is, under a win margin of around 31% at the first round of municipal elections, incumbent mayors tend to behave more opportunistically. However, opportunism seems not to pay off. Hence, Aidt et al. (2011) s findings cannot be generalized. Keywords: Vote function, Opportunism, Rational Political Business Cycles, System estimation, Municipalities, France. JEL Classification: D72, E32, H72. I thank Aurelie Cassette and Etienne Farvaque for helpful guidance. I thank Jean-Sébastien Pentecôte, Olivier Beaumais, Arnaud Rioual, Jean Baptiste Desquilbet, Stéphane Vigeant, Abdoulaye Papa Diop and the participants in seminars in the Univerity of Lehavre and the University of Lille for useful comments. LEM-University of Lille, Cité Scientifique, Bât. SH2 - Bureau 101, 59655 Villeneuve d Ascq, E-mail: mamadou.boukari@ed.univ-lille1.fr. 1

1 Introduction Do local governments spendings follow a different pattern in elections years? If yes, do electoral budget cycles impact on the re-election prospects of the incumbent mayor and her political party? Then, what determines the magnitude of cycles? Answers to the above questions come from two strands of the literature. On the one hand, the literature on vote functions suggests that economic conditions systematically affect election outcomes. On the other hand, the literature on Political Business Cycles predicts that incumbent governments engage in pre-electoral policy manipulations to influence voters and maximize their re-election prospects. Aidt et al. (2011) proposed an integrated approach of the two strands of literature and derived a two-way relationship between the win margin of the incumbent politician and the size of the opportunistic distortion of fiscal policy. Testing their theoretical model on Portuguese municipalities, they show that opportunism leads to a large win margin for the incumbent and that incumbents behave more opportunistically when their win margin is small. The objective of this paper is to build on their study, but using a dataset of French municipalities to see whether their conclusions can be generalized. However, the paper goes further than a replication by departing from their model by adding some variables, to take into account the specificities of French municipalities. Another motivation of this paper comes from the fact that, even if the existence of a Political Budget Cycle has been observed for French municipalities, there is no study that examines the joint determination of the win margin of the incumbent and of the fiscal distortion in this country. Thus, the main contribution of this paper is to fill this gap. Using a large dataset of French municipalities of only more than 3,500 inhabitants, the estimations reveal that theoretical predictions are partially supported in this case. The data support the hypothesis of a non-monotonic relationship between the win margin of victory and the opportunistic distortion. This means there is a threshold of the win margin under which the opportunistic distortion increases, and decreases 2

above it. This threshold is about 31% of the difference in vote shares between the incumbent mayor and her best challenger at the first round of municipal elections. Note also that the number of candidates increases the win margin of the incumbent. The results indicate also that voters express some kind of attachment to their mayors, but dislike fiscal distortion. Indeed, according to the estimations, the win margin is persistent in municipal elections in France. However, voters punish mayors who belong to the government s party for bad economic conditions. Also, in line with empirical findings, the municipal debt impacts negatively the win margin. A look at the determinants of opportunistic behavior reveal that mayors belonging to the government s party are those who behave more opportunistically contrary to those belonging to the right-wing party or with long tenure in office. Moreover, the increase in capital transfers from the central government to municipalities make mayors behave more opportunistically. Finally, the inclination of incumbent mayors to behave opportunistically depends also on voters awareness. This paper is organized as follows. The next section describes the institutional setting of French municipalities. Section 3 presents briefly the related literature and the main features of the theoretical model. Section 4 introduces the data sources and the empirical model. Results are discussed in section 5. Finally, section 6 concludes. 2 Legal context and role of municipalities in France This section presents the local political context, highlighting the specificities of French municipalities. It starts with a brief description of the institutional context followed by a discussion of municipal officials behavior and ability to affect local public spending. Municipal governments constitute the lowest level of government in France. They were established formally by a law of 14 th December 1789. Their legal context has historically evolved several times, but has remained relatively fixed since 1983. The political system in French municipalities can be characterized as follows. Elections normally take place every six years 1 and election dates are fixed nationally and 1 Except between 2001 and 2008, the election being postponed to avoid electoral fatigue in 2007, 3

are therefore exogenous from the perspective of municipalities. During the period under study (2000-2014), all elections took place in March, and there were no legal restrictions on the number of times a mayor could stand for election. Note also that there was a 3,500-inhabitant threshold 2 with a different voting system for smaller cities. For the towns with population numbers above the threshold, the poll competition is organized by lists, with two rounds (possibly) taking place. The winning list receives half the seats to be filled in the town council, the other half of the seats being distributed proportionnally between all the lists (including the winning list) that have received more than 5% of the votes. If a second round is necessary, all the lists with more than 10% of the votes can compete but the lists with more than 5% of the votes can merge between the two rounds. The affairs of the municipality are in the joint responsibility of the mayor and the municipal council. The latter elects the mayor within its members. These members are elected directly by citizens who vote for party or independent lists of candidates. In local politics, the mayor heads the municipal council, presides over all council committees and sets the local policy agenda. Concerning the competences, French municipalities are responsible for multiple activities. The main local public services under their control include distribution of water, local transportation, elementary education, property maintenance, promotion of culture and science, provision of recreation and sports facilities, local health care, social housing, environmental protection and municipal policing. Note also that they operate under the same financial regime. Since 2003, French municipalities gained more autonomy in setting local public policy. For instance, they are autonomous in defining their own budgets, and collect the revenues 3 they are entitled by law and allocate expenditures. However, mayors discretionary power is not the same according to the budget components. For instance, while current expenditures, such as municipal employee a year in which both the Presidential and Assembly elections were taking place. 2 Since 2013 the threshold is reduced to 1000 inhabitants. 3 Municipal resources come from local taxation (property tax and local business tax), loans and transfers from the central government. They can only borrow to fund equipment expenditures 4

salaries, are non-discretionary and hard to manipulate, the mayors can control the level and timing of capital expenditures of which equipment expenditures are the main component. This is the reason why the study focuses on that variable. 3 Related literature and theoretical framework 3.1 Related literature As indicated in the introduction, this paper relates to two strands of literature: the political budget cycles and the literature on vote and popularity functions. Studies of political budget cycles are interested in the fiscal impact of elections. These studies shed light on whether elections lead politicians to adopt short-term policy measures in an effort to increase their chances of re-election. But a subsequent question is whether a political budget cycle does exist with rational and foward-looking voters? If so, what is the effect of the opportunistic behavior on the re-election chances of incumbent governments? From the theoretical point of view, asymmetric information between voters and politicians (Rogoff & Sibert, 1988; Rogoff, 1990), voters awareness (Shi & Svensson, 2006), and immaturity of democracy (Gonzalez, 2002) have been shown to magnify the size of opportunistic cycles. For example, Rogoff (1990) formulated the argument that some characteristic of politicians (level of competency) is unobserved, and voters have to infer it according to economic outcomes. Assuming that more competent politicians produce better outcomes, then good outcomes before elections may signal high competence of the incumbent policymaker. Hence, if competence is expected to persist after the election, it may be optimal for voters to reappoint the incumbent. According to Shi & Svensson (2006), the magnitude of electoral budget cycles increases with the size of the rent that politician can earn by remaining in office and with the share of uninformed voters in the electorate. Brender & Drazen (2005) argue that the electoral budget cycles reflects the experience and interactions of all actors with the electoral system. Experienced voters know that policymakers have 5

incentives to inflate the economy in election years, and as a result fiscal manipulation loses its attractiveness. A long experience of competitive elections should thus make a democracy less susceptible to budget cycles. These models and their extensions have been tested empirically both at the national and local levels. At the national level, except for new democracies, political expenditure cycles do not seem to exist. Also, election-year deficit spending does not lead to a higher re-election probability which may even decrease (e.g., see Brender & Drazen, 2005). However, at the local level, there is some evidence of political budget cycles on the spending side, in particular for what concerns public investment. It comes in many forms: increase in public spending or employment, reduction in taxes (e.g., see, Sakurai & Menezes-Filho, 2011; Meloni & Tommasi, 2012), or moving spending from less visible public services to more visible ones: Kneebone & McKenzie (2001) coined the expression visible expenditures to catch the positive impact of some types of spending on politicians reelection chances. In their work, they found evidence of local political budget cycles about Canadian Provinces. Likewise, in the case of Portuguese municipalities, Veiga & Veiga (2007), show that increases in investment expenditures and changes in the composition of spending favouring highly visible items such as investment expenditures on overpasses, streets and complementary works, and on rural roads are associated with higher vote percentages for incumbent mayors seeking reelection. Akhmedov & Zhuravskaya (2004) evaluate local Russian governmental entities and find that pre-electoral manipulation of fiscal instruments increases the incumbent s chances of reelection. Using a sample of Columbian municipalities, Drazen & Eslava (2010) also bring evidence that governments, in their attempt to remain in office, tend to increase visible expenditures on housing, health, water and energy to target voters. In contrast to most previous studies, Balaguer-Colla et al. (2015) use Bayesian techniques to analyze the effect of public spending on municipal re-election in Spain. They find that, in general, increases in local government spending positively impact 6

on local governments chances of re-election. Moreover, the capital expenditure over the whole term positively affects the re-election probability, although the pre-electoral capital expenditure has a stronger effect on the chances of re-election. They also find that the electorate only rewards increases in current expenditures made in the pre-election period. However, other studies indicate that the opportunistic behavior has also a negative effect on the probability of re-election. For instance, Peltzman (1992) find that the US electorate penalizes governments that increase public expenditure in the run-up to elections. Brender (2003) for the case of local elections in Israel, find that a larger deficit in the year prior to elections reduces the probability of the incumbent party s re-election. Similarly, Brender & Drazen (2008) observe that in the more developed countries and advanced democracies, governments in a situation of deficit and that introduce tax cuts in an election year have lower chances of re-election. Analyzing Argentine electoral districts, Meloni (2001) provide additional evidence in this regard, revealing that an increase in public expenditure negatively affects the percentage of votes obtained by the government party. In France, there are also studies that analyze the electoral budget cycles at the local level. For instance, Binet & Pentecôte (2004) demonstrate the existence of an opportunistic cycle of capital expenditure on 883 French towns of more than 10,000 inhabitants from 1988 to 1999. They also confirm a similar intensity cycle of public debt. Likewise, using a panel of 91 French municipalities over the period 1977-2001, Foucault & François (2005) find the presence of an electoral budget cycle on the spending side. Concerning the political determinants of such a cycle, they find that political changeover reduces the opportunity to behave opportunistically. Dealing with the specificities induced by the two-round process of the French electoral rule, Cassette et al. (2013) establish three results. Firstly, they show that the municipal budget structure has an impact on the share of votes of the incumbent s party, but only in the first round. Secondly, local economic conditions, considered through personal income and unemployment, do not seem to pay a significant part in 7

voters decisions. Finally, political variables play a role at both rounds, especially the number of candidates, and national partisan waves. Concerning the vote functions, the main issue addressed is to look at the impact of economic conditions on voting behavior and election results 4. Empirical results indicate that the performance of the national economy is important especially if local governments are of the same party as the central government (Martins & Veiga, 2013). Hereafter, I can signal a non-exhaustive list of articles dealing with this topic in the French case. Jérôme-Speziari & Jérôme (2002) test empirically a municipal vote function using a pooled time series for the 236 municipalities over 30,000 people. Their model identifies the factors that generate an electoral bonus and those that give an electoral dement point to the outgoing municipal teams. Although it is not so obvious to conclude between punishment and reward hypothesis, they were able to assert that the 2001 vote has been marked by a real grievance asymmetry. According to Auberger & Dubois (2005) the national (the real growth of GDP) and local economic conditions (the growth rate of the number of job-seekers in every department) plays an important role on the outcomes of the French legislatives elections. Farvaque & Jean (2007) study the impact of macroeconomic conditions and of party endorsements on the result of parties candidates in local elections in France. They test for economic variables, and find that the electorate penalizes the incumbent party for unemployment or for a high misery index. They also find the presence of an incumbency premium and a strong impact of endorsement, confirming the presence of partisanship in local election results. Relatively to tax policy, Dubois & Paty (2010) estimate a vote function on a panel data set of 104 municipalities (including both political and economic determinants at the both national and local levels). They show that voters sanction the incumbent if their own local housing tax is higher than that of their close geographical neighbors. 4 Seminal papers are Key (1966) and Kramer (1971). See Nannestad & Paldam (1994) for a survey. 8

3.2 Theoretical framework This section introduces the theoretical model that is going to be tested empirically. It relates to the version of Rogoff (1990) s signaling model proposed by Aidt et al. (2011). The main features of the model are the following. It considers a two-period economy in which voters have the same policy preference as politicians. They derive utility from a private consumption good (C t ) and two types of public goods (g 1,t and g 2,t+1 ). The utility function is given by: U υ = C 1 + lng 1,1 + θlng 2,1 + β (C 2 + lng 1,2 + θlng 2,2 ) (1) where θ is the relative importance of unobservable public good (g 2,1 = 1). As in Rogoff (1990), a political budget cycles arises due to information asymmetry about the incumbent s competence in achieving the production of public goods. Politicians have either a high level of competence with probability ρ or a low level with probability (1 ρ). Thus, voters have to infer the incumbent s competence. The production function help to describe this inference mechanism. Indeed, the production technology of public goods is given by the following equation: g 1,t + g 2,t+1 = τ t + ε t (2) where ε t is a stochastic competency term. Spending on public good 1 (g 1 ) is visible and observed immediately within the period, t, while spending on public good 2 (g 2 ) is hard to observe. As a consequence, citizen-voters cannot infer how much was spent on good 2 until later (with a one-period time lag, t + 1) when they observe the provision levels generated by past spending. Thus, they infer the politician s competence from the quantity of the visible public good provided in the first period. Note that the production of public goods depends on the availability of funds (lump-sum tax) and the politician s competence. In addition to competency, voters care about the ideology of their elected politician. 9

This leads to an advantage (or disadvantage) for the incumbent that decomposes into: an average incumbency advantage, µ, which is constant over time and captures the incumbent s underlying popularity with voters; a random ideological shock, υ which is drawn before each election from a symmetric unimodal distribution F (υ t ) with zero mean and variance one 5. Formally, one has: α t = µ συ t (3) where µ and σ are parameters and σ > 0. Concerning the timing of events, it is described as a sequential game of incomplete information. At the beginning of period 1, each type of politician takes a first period action that reveals their type. Voters observe α 1 and how much is provided of the observable good (g 1,1 ). At the end of period 1, an election takes place where the incumbent runs against a randomly chosen challenger. Taking into account the amount of visible public good provided in period 1, voters update their beliefs about the incumbent s competency and cast their vote. The incumbent is re-elected if she is supported by a majority of voters; otherwise the challenger takes office. At the beginning of period 2, the incumbent, if reelected, decides how much to spend on the observable public good. If the challenger is elected, he observes his competency (ε 2 ) and decides how much to spend on the observable public good. Two useful tools help deriving the equilibrium. First, the politician obtains utility from being in office ( ego rent ). Second, she faces a trade-off between the cost of deviating from the optimum (opportunistic distortion) and the expected returns of staying in office. Since distorting the fiscal policy is less costly the more competent the politician is, she inflates the economy to signal her own high competency. Applying the concept of Perfect Bayesian Equilibrium, the resolution of the model reveals that a separating equilibrium exists. The more competent incumbent tries to 5 For example, Solé-Ollé & Sorribas-Navarro (2014) find that corruption scandals have a marked negative effect on trust in local politicians. This effect is even more marked in the case of individuals that have no ideological attachment to the party accused of corruption and/or who obtain their information from the media. 10

distinguish herself from the politician of the other type by distorting the economy. This is the Rational Political Business Cycle. Its size is called the opportunistic distortion and is the first important object of the model. The second one is the average reelection probability of the incumbent. It can be defined as the type-weighted ex ante reelection probability of the incumbent. For simplicity, Aidt et al. (2011) call it the win margin of victory. These two objects are jointly determined at equilibrium. On the one hand, since the opportunistic distortion affects the equilibrium beliefs, it appears to be a determinant of the win-margin of victory. On the other hand, the win-margin, through its effect on the re-election differential between competent and incompetent politicians, is a determinant of the size of the opportunistic distortion. Prior to the emprical specification, it is interesting to check how these two objects feed each other. This two-way relationship is stated in the two main hypotheses of the study. H1: Opportunism pays off When the politician with high level of competency wants to separate out, she has to increase the opportunistic distortion. Behaving so increases her win margin. Thus, the win-margin is at least increasing in the size of the opportunistic distortion. H2: There is a non monotonic relationship between the win margin and the fiscal policy distortion. The second theoretical result, established by Aidt et al. (2011), suggests that the oppotunistic distortion is a quasi-concave function of the win-margin, increasing at first, and decreasing. Therefore, one can expect the impact of the win-margin on the size of the opportunistic distortion to be non-monotonic. 11

4 Data and econometric model 4.1 Data The model encompasses annual time series data for 2704 French municipalities of more than 3,500 inhabitants over the period 2000-2014. With this size, the dataset is larger than in any of the previous studies of municipal budget cycles in France. Two thirds of cities are populated by a range of 3,500 to 10,000 inhabitants and one third by more than 10,000 inhabitants. In the sample, all the departments in metropolitan France are represented. [Insert Table 1 about here] In the time period under review, three municipal elections were held (2001, 2008 and 2014). According to different electoral rules, cities of Paris, Marseille and Lyon were excluded from the sample. Table 1 presents data sources and the descriptive statistics of the variables mentioned in this study. In the full sample, 7 percent of all elections are single-candidate, while more than 35 percent had 2 candidates, 30 percent had 3 candidates, 16 percent 4 candidates and 11 percent had between 5 and 11 candidates. To exclude single-candidate elections, the general sample is restricted to be within 80% winning margin. Table 2 presents the frequency and percentage distributions of the win margin of victory for the resulting subsample. [Insert Table 2 about here] Dependent variables: Win margin (WM) of the incumbent: It is the difference between the vote share of the mayor s party and that of the largest opposition party at the first round of municipal election. In case of defeat of the incumbent, the win margin takes a negative value. 12

Opportunistic Distortion (OD): The opportunistic distortion is measured as the percentage deviation of expenditures in the year preceeding elections from the election term average (EExptm). As municipal elections take place in March, this study considers that the opportunistic distortion could be higher in the year before the election 6. This study considers the equipment expenditures (EExp). This choice is motivated by the fact that opportunistic distortions are, in practice, most likely to show up in budgetary items whose timing of implementation is controlled by the mayor and which are visible to the electorate. Independent variables include the following: The square of the win margin (WM2) is included in the estimates in order to check for potential non-linearities in the win margin effect. Years in Office (YO): the number of years the incumbent has been in office. Run for Reelection (RR): a dummy variable that measures whether the incumbent mayor or her party runs for reelection (1) and or not (0). Government Party (GovP): A dummy variable equal to 1 if the incumbent mayor belongs to the same party as the French President and 0 otherwise. Municipal employment (EMP) designates the number of people employed in the private sectorin each municipality. Average municipal real wages (WAGES), euro per capita. This variable measures the private sector employee salaries in the municipality. Average capital transfer from the national government to the municipality during the preceding election term (CAPTtm); euro per capita. Election year change in the capital transfers (VCAP) in percentage. Percentage of the population over 65 years old (POP65), used to proxy the low average education levels. 6 According to Foucault & François (2005), the implementation of Local Political Business Cycle (LPBC) on the French municipalities raises some difficulties in terms of agenda. While the municipal election is usually planned in March, the budget of year t is voted in December of t 1 year and is theoretically applicable for year t whatever the result of election. This causes a real ambiguity concerning the importance of LPBC analysis. To limit this ambiguity, they suggest to consider that opportunistic cycles are likely to occur during the year before the election (t 1) and/or during the year of election. 13

Population density, used to proxy urbanization (DENS); inhab/km 2. Municipal Debt (Debt); euro per capita. The introduction of the variable debt, reflecting the financial liabilities generated in the years analysed, allows to verify whether the electorate punishes high levels of local debt or whether, on the contrary, it supports certain levels of debt that may derive from higher municipal expenditure Prior to the econometric analysis, Figure 1 displays the evolution of the main variables used in this study (Equipment expenditures, current expenditures and capital transfers). Two main points can be noticed. First, there is an increasing trend in all these variables. Second, graphically, a budget cycle does exist, according to the peaks in 2001 and 2007. [Insert Figure 1 about here] These observations are in line with Foucault et al. (2008) who confirm the opportunistic behavior of local governments. They note an increase in all categories of public spending in pre-electoral periods. However, this does not reveal if there is a link between the opportunistic distortion (OD) and the win margin of victory (WM), something towards which I now turn. 4.2 Econometric model Following Aidt et al. (2011), this study estimates a system of two simultaneous equations. The first equation represents a vote function and the second one is for the opportunistic distortion (OD). This specification allows analyzing the existence of a two-way relationship between the win margin and the opportunistic distortion. I thus write: 14

W M it = α 1 OD it + α 2 IE it + α 3 Y M it + α 4 RR it + α 5 W MP it + α 6 GovP it +α 7 URG it + α 8 GovP it URG it + α 9 W AGES it + α 10 Debt it 1 + υ i + ɛ it (4) OD it = β 0 W M it + β 1 (W M it ) 2 + β 2 IE it + β 3 Y M it + β 4 RR it + β 5 CAP tm it +β 6 V CAP it + β 7 P op65 it + β 8 DENS it + β 9 Right + γ i + γ t + µ it (5) where i is the index for municipalities and t indicates election years 7. Municipal fixed effects (υ i and γ i ) into the two equations and election year fixed effects (γ t ) are incorporated into the equation 5. α 1 to α 10 and β 0 to β 9 are parameters to be estimated and ɛ it and µ it are random error terms with E (µ it ) = E (ɛ it ) = 0. According to the theoretical model, expectations about the sign of interest variables are α 1 > 0 (hypothesis 1) and β 0 > 0 and β 1 < 0 (Hypothesis 2). There are common variables to the two equations. This concerns Equipment Expenditures, Years in office and Run for reelection. As voters reward mayors for keeping investments high on average throughout the term, one expects that equipment expenditures are positively related to the win margin (α 2 > 0). As documented by Mueller (1970) and Veiga & Veiga (2004) for example, the number of years the incumbent has been in office (YO) reduces the win margin. Thus, its coefficient is expected to be negative (α 3 < 0). Likewise, one expects that low average investment expenditures (IE) make it easier to be opportunistic and to create large percentage deviations of investment expenditures from the average at election times (β 2 < 0). Concerning the fiscal distortion ability, it is assumed that mayors with long tenures are more experienced and so are more able to manage investment expenditures opportunistically (β 3 > 0). In the same way, the party of the incumbent mayor is expected to do better when the mayor runs for reelection than when a new candidate of unknown type is presented (α 4 > 0). Running for reelection entails the will to incur the cost of 7 The election years are 2001, 2008 and 2014. The election of 2001 is not included in the analysis whenever lags, term averages or deviations from term averages are included. 15

signalling and, thus, the attempt to increase investments opportunistically: (β 4 > 0). In the distortion equation, four other variables are added. On the one hand, we have the average capital transfer from the national government during the preceding election term (CAPTtm) and the election year change in the capital transfer (VCAP). These two variables are related to the availability of funds. The theory suggests that capital transfers increase the opportunistic distortion in election years without having a direct effect on the win-margin. Thus, β 5 and β 6 are expected to be positive. On the other hand, two variables that are related to the voters awareness are also included because the theory suggests that voters awareness tends to reduce the magnitude of the Political Business Cycle. One way to seize voter awareness is to use a measurement based on education and urbanization as in Akhmedov & Zhuravskaya (2004). Like Aidt et al. (2011), here the percentage of the population over 65 years of age (POP65) is used to proxy for low average education levels and the population density (DENS) for urbanization. POP65 is expected to be associated with low levels of awareness while DENS is expected to be asssociated with high levels of voters awareness. Thus, predicted signs of their coefficients are β 7 > 0 and β 8 < 0. The last variable in this equation is RIGHT, a dummy variable that is equal to 1 if the mayor belongs to a right-wing party. There is no prior on the sign of β 9. Concerning the vote function, it also includes other variables such as local economic conditions (WAGES), incumbent s personal characteristics (WMP) and an interaction variable between the Government s Party (GovP) and the unemployment rate gap (URG). The latter helps to test for the responsibility hypothesis. A negative sign is expected for α 6 because the electorate may wish to reward, or to punish, the national government in second tier (local) elections (Carsey & Wright, 1998). Given the fact that voters tend to punish the national government for bad economic outcomes, higher unemployment rates should lead to a lower percentage of votes for the incumbent mayors who belong to the same party as the national government. Likewise, since voters are expected to reward mayors who achieve high level of high average municipal 16

wages (WAGES) during their tenure, a positive sign is also expected for α 9. Also empirical findings suggest that the Municipal debt is associated to a negative sign (α 10 < 0). Equation 4 also includes the win margin in the previous election (WMP). This variable picks up unobserved factors such as the mayor s personal characteristics and ideologies, as well as party affiliations of voters. Persistence in voter preferences (and thus in voting behavior) is expected and a positive sign is predicted for α 5. In addition to variables retained by Aidt et al. (2011), I use two other variables in the vote function: one year lagged municipal debt and the number of lists running for the election. The primer is to include the fact that debt is used to fund equipment spending. The latter allows to take into account the impact of the multi-party system at the first round of municipal elections. To estimate this system of two equations, various methods exist and include: Two Stage Least Squares (2SLS), Three Stage Least Squares (3SLS), Full Information Maximum Likelihood (FIML) and Generalized Method of Moments (GMM). For Aidt et al. (2011), the GMM estimator selects parameter estimates so that the correlations between the instruments and disturbances are as close to zero as possible, as defined by a criterion function. By choosing the weighting matrix in the criterion appropriately, GMM can be made robust to heteroskedasticity and/or autocorrelation of unknown form. These efficiency gains in the presence of heteroskedasticity are the relative advantage to 3SLS. Given these elements, the GMM estimator is applied and results are shown in the next section. To diagnose multicollinearity between regressors, the Variance Inflation Factor (VIF) is computed. Generally a large value of the VIF means that the variables are more collinear. A common criterion is to consider a VIF exceeding 10 as an indication of high collinearity (Gujarati, 2004). All the variables used in this study seem not to suffer from any multicollinearity because the VIF ranges between 1.12 and 9.58 8. 8 Results of VIF tests available upon request. 17

5 Results This section contains two parts. In the first part, the main results of the voting function (Equation 4) and of the Opportunistic distortion s equation (Equation 5) are discussed. The second part deals with the robustness check. Tables 3 report findings of opportunistic distortions in equipment expenditures. The results of the main estimation (whole sample) are displayed in the first two columns. As the time size is very small, only municipal fixed effects were taken into account when running the estimations. The overidentification test of all instruments (Hansen J-Statistic) suggest that instruments are valid. Thus, results can be discussed. [Insert Table 3 about here] 5.1 Main results In table 3, the most general estimation in column (1) includes all the explanatory variables of the two equations. In column (2), the least significant variables are eliminated. The main variables of interest are highly significant. But, in the vote function, the Opportunistic distortion has an unexpected sign. More precisely, contrary to the theoretical prediction and Portuguese evidence, opportunism reduces the win margin of victory. This result is not in line with the findings of Aidt et al. (2011) on Portuguese municipalities. An explanation of this result is to consider that voters recognize the induced cost of fiscal distortion on the spending side. That is, an increase in spending leads to higher taxes after elections or debt accumulation. For instance, Cassette & Farvaque (2014) show that voters are fiscal conservatives, although in the long run. Indeed, the negative impact of the municipal debt on the win margin is confirmed in this study. Moreover, the fact that voters dislike fiscal distortion is in line with conventional wisdom and some research holding that voters dislike government spending and big government (Peltzman, 1992; Meloni & Tommasi, 2012). Peltzman (1992) shows that increases in aggregate expenditures are politically damaging to governors, lowering 18

their vote share. This should decrease the likelihood of political cycles in aggregate expenditures at the state-level in the United States. For the effects, a one-percentage point increase in the opportunistic distortion reduces the win margin of victory by approximately 2 to 5%. From the estimates of equation (4), the number of years the incumbent has been in office increases her win margin. This suggests that voters are attached to their mayors. One notes also that the win margin is persistent. This gives a strong support to the fact that voters are attached to their mayors. This result may also reflect a strong degree of partisanship. This result is in line with Farvaque & Jean (2007). As expected, the term mean equipment expenditures appear to increase the win margin of victory. All the contrary, the estimations show also that the unemployment rate gap impacts negatively the win margin of victory. This means that bad local economic conditions reduce the incumbent support. More specifically, a one percentage point increase in the local unemployment rate above the national one is associated with a 2.4 percentage point decrease in the vote share of an incumbent belonging to the Government s Party. Thus, voters may use municipal elections to express their discontent with regard to the national government. Note also that the number of candidates increases the win margin of the incumbent. This result is consistent with the fact that a dispersion of the opposition from all sides benefits to the incumbent. Estimates of equation 5 support the hypothesis that there is a non-monotonic relationship between the win-margin and the opportunistic distortion of the fiscal policy. One notes that the win margin has a highly statistically significant positive effect on the opportunistic distortion and that the quadratic term is negative and significant. Thus, if an increase in the win margin increases the opportunistic distortion, this effect decreases while the former still increases. Then, there is a threshold 9 of the win-margin under which the inclination to distort fiscal policy increases. The values of the threshold for each subsample used in the estimations are displayed in table 4. 9 From the first partial derivative of OD over WM in equation 5 one obtains the threshold equal to W M = β 0 2β 1. 19

For the all sample, that value is about 31% and is slightly higher than the mean of the win-margin (see Table 1). [Insert Table 4 about here] The data support some secondary hypotheses. Firstly, opportunism is greater when the incumbent mayor belongs to the government s party, and when there are increases in capital transfers from the national government in the year prior to elections. Secondly, mayors who belong to the right-wing party and those who have a long tenure in office tend to behave less opportunistically. The population density has a negative effect on the opportunistic behavior of the incumbent. This is line with the theoretical findings of the literature according to which voters awareness (education level and urbanization) reduces politicians opportunism. Since the percentage of population over 65 years of age is not significant, it was dropped from the other columns. 5.2 Subgroup analyses For robustness checks, the estimation is repeated on two subsamples. The first subgroup (column 3) relates to cities with a population size comprised between 3,500 and 10,000 inhabitants, and the second one, those of more than 10,000 (column 4). Note that the first four columns are based on the opportunistic distortion defined as in section 4. In the last three columns (5, 6 and 7), the estimations are based on the opportunistic distortion measured as the percentage deviation of equipment expenditures from their trend obtained using the Hodrick-Prescott filter. The results obtained on the whole sample appear to be robust. Results on subsamples are similar to those on the whole sample. In the equation 5, the non-linearity effect of the win-margin on the opportunistic distortion is confirmed. The threshold under which mayors behave opportunistically is around 31% (see table 4), close to the mean of the win margin (27.6%). The result according to which mayors who belong to the right-wing party and those who have a long tenure in office tend to behave less opportunistically is also robust. 20

Finally, population density seems to reduce the mayors inclination to distort fiscal policy prior to elections. This means that voters awareness matters. This result is in line with Shi & Svensson (2006) who argue that a large share of informed voters renders fiscal policy manipulation less effective. 6 Conclusion Aidt et al. (2011) introduced an innovation which is to acknowdge the interaction between the incentive to generate cycles and the prospect of winning elections. This paper adds an empirical evidence to this innovation by taking advantage of a large dataset of all French municipalities of more than 3,500 inhabitants. Relatively to the first hypothesis under study, the estimations reveal that opportunism seems not to pay off in the French case. This result lies in contradiction with Aidt et al. (2011) and implies that politicians should not systematically engage in electorally motivated spending as it is an ineffective means to enhance re-election chances. The second main hypothesis is confirmed: this paper highlights a non-linear effect of the win margin on the opportunistic distortion. At a first time, opportunism increases until the win margin reaches a threshold, and decreases after. This threshold is around 31%. This means that mayors behave more opportunistically when their win margin is small, i.e. under the threshold. Beyond the main findings of this paper, it would be interesting to check whether partisanship matters in central government capital transfers. If so, what is the effect on election results? Moreover, as voters preferences are persistent over time, it will be also interesting to examine whether incumbency advantages are heterogeneous accross parties. 21

References Aidt, Toke, Veiga, Francisco, & Veiga, Linda. 2011. Election results and opportunistic policies: a new test of the rational political business cycle model. Public Choice, 148, 21 44. Akhmedov, A., & Zhuravskaya, E. 2004. Opportunistic political cycles: test in a young democracy setting. Quaterly Journal of Economics, 119, 1301 1338. Auberger, A., & Dubois, E. 2005. The influence of local and national economic conditions on French legislative elections. Public Choice, 125(3), 363 383. Balaguer-Colla, M. T., Brun-Martosa, M. I., Forteb, A., & Tortosa-Ausina, E. 2015. Local governments re-election and its determinants: New evidence based on a Bayesian approach. European Journal of Political Economy, 39(1), 94 108. Binet, M.-E., & Pentecôte, J.-S. 2004. Tax degression and Political budget cycle in French Municipalities. Applied Economics Letters, 11, 905 908. Brender, A. 2003. The effect of fiscal performance on local government election results in Israel: 1989-1998. Journal of Public Economics, 87, 2187 2205. Brender, A, & Drazen, A. 2005. Political budget cycles in new versus established democracies. Journal of Monetary Economics, 52(7), 1271 1295. Brender, A., & Drazen, A. 2008. How do budget deficits and economic growth affect reelection prospects? Evidence from a large panel of countries. American Economic Review, 1993, 2203 2220. Carsey, T., & Wright, G. 1998. State and national factors in gubernatorial elections. American Journal of Political Science, 42(3), 994 1002. Cassette, A., & Farvaque, E. 2014. Are Elections Debt Brakes? Evidence from French Municipalities. Economic Letters, 122(2), 314 316. 22

Cassette, A., Farvaque, E., & Héricourt, J. 2013. Two-round elections, one round determinants? Evidence from the French municipal elections. Public Choice, 156, 563 591. Drazen, A, & Eslava, M. 2010. Electoral manipulation via voter-friendly spending: Theory and evidence. Journal of Development Economics, 92(1), 39 52. Dubois, E., & Paty, S. 2010. Yardstick competition: which neighbors matter? Annals of Regional Science, 44(3), 433 452. Farvaque, E, & Jean, N. 2007. Analyse économique des élections municipales: Le cas de la France (1983-2001). Revue d Économie Régionale et Urbaine, 5, 945 961. Foucault, M., & François, A. 2005. La politique influence-t-elle les décisions publiques locales? Analyse empirique des budgets communaux de 1977 à 2001. Revue Politiques et Management Public, 23(3). Foucault, M., Madies, T., & Paty, S. 2008. Public spending interactions and local politics: Empirical evidence from French municipalities. Public Choice, 137(1-2), 57 80. Gonzalez, M. 2002. Do Changes in Democracy Affect the Political Budget Cycle?: Evidence from Mexico. Review of Development Economics, 6(2), 204 224. Gujarati, D. 2004. Basic Econometrics. Fourth edition, The McGraw-Hill Company. Jérôme-Speziari, V., & Jérôme, B. 2002. Les municipales de 2001: vote récompense ou vote sanction? Revue Française de Science Politique, 52, 251 272. Key, V. O Jr. 1966. The responsible electorate: rationality in presidential Voting. Havard University Press. Kneebone, R., & McKenzie, K. 2001. Electoral and Partisan Cycles in Fiscal Policy: An Examination of Canadian Provinces. International Tax and Public Finance, 8, 753 774. 23

Kramer, G. H. 1971. Short term fluctuations in US Voting behavior. American Political Science Review, 65(1), 131 143. Martins, R., & Veiga, José F. 2013. Economic voting in Portuguese municipal elections. Pubic Choice, 155(3-4), 317 334. Meloni, O. 2001. Gobernadores y elecciones: Es "negocio" ser austero? Evidencia a partir de datos de panel. XXXVI Reunión Anual de la Asociación Argentina de Economia Politica, Buenos Aires. Meloni, O., & Tommasi, M. 2012. Voters as Fiscal Liberals: Incentives and Accountability in Federal Systems. Economics & Politics, 24(2), 135 156. Mueller, J. E. 1970. Presidential popularity from Truman to Johnson. American Political Science Review, 64, 18 23. Nannestad, P., & Paldam, M. 1994. The VP-function: A survey of the literature on vote and popularity functions after 25 years. Public Choice, 79(3), 213 245. Peltzman, S. 1992. Voters as fiscal conservatives. Quaterly Journal of Economics, 107(2), 327 361. Rogoff, K. 1990. Equilibrium political budget cycles. American Economic Review, 80, 21 36. Rogoff, K., & Sibert, A. 1988. Elections and macroeconomic policy cycles. Review of Economic Studies, 55, 1 16. Sakurai, S., & Menezes-Filho, N. 2011. Opportunistic and partisan election cycles in Brazil: new evidence at the municipal level. Public Choice, 148, 233 247. Shi, M., & Svensson, J. 2006. Political budget cycles: do they differ across countries and why? Journal of Public Economics, 90, 1367 1389. Solé-Ollé, A., & Sorribas-Navarro, P. 2014. Does Corruption Erode Trust in Government? Evidence from a Recent Surge of Local Scandals in Spain. Cesifo Working Paper, 4888. 24

Veiga, Francisco J., & Veiga, Linda G. 2004. Popularity functions, partisan effect and support in Parliament. Economics & Politics, 16(1), 101 115. Veiga, Linda G., & Veiga, Francisco J. 2007. Does opportunism pay off? Economics Letters, 96, 177 182. 25

Table 1: Summary statistics Variable name Data Obs Mean Std. Min Max Source Dev. Win Margin Ministry of Internal Affairs 5200 27.60 26.33-34.66 100 Win Margin in the Previous Election Ministry of Internal Affairs 5180-4.12 20.75-100 100 Equipment expenditures Census of Ministry of Finance 5201 387.07 267.43 13.33 6777.47 Equipment expenditures (term mean) Census of Ministry of Finance 5201 387.03 221.65 31.05 3597.27 Opportunistic distortion Census of Ministry of Finance 5186 10.45 49.11-465 178.29 Average real wages INSEE 5199 13566.68 3806.31 78.85 52044 Capital transfers (term mean) Census of Ministry of Finance 5201 70.76 58.97 0 704.32 % change in capital transfers (from previous year) Census of Ministry of Finance 5201 3.49 81.64-1289.4 1321.44 Government s Party Ministry of Internal Affairs 5200.36.48 0 1 Years in Office Ministry of Internal Affairs 5201 9.83 8.14 0 49 Municipal employment INSEE 5198 5883.58 12334.05 8 258255 Municipal debt Census of Ministry of Finance 4590 544.39 639.16 0 7877.47 Population density INSEE 5201 1247.06 2196.17 0 26870.12 Run for re-election Ministry of Internal Affairs 5201.83.38 0 1 Number of Lists Ministry of Internal Affairs 5201 3 1.38 1 11 % Population over 65 years old INSEE 5201 16.66 6.25 0 51.56 Right Ministry of Internal Affairs 5201.46.5 0 1 Unemployment rate (national) INSEE 5201 9.11 1.13 8.01 10.28 26

Table 2: Frequency distribution of the Win Margin Observations Frequency Percentage under -20 92 2-20-00 456 9 00-20 2038 42 20-40 1534 32 40-60 626 13 60-80 83 2 4829 100 Source: Author s calculations. 27

Figure 1: Trend of main variables Source: Author, based on Ministry of Finance data. 28