Who should you vote for? Empirical evidence from Portuguese local governments

Abstract

The economic literature considers voters quasi-rational agents that care about maximizing their individual welfare when deciding on who to vote for. Voters believe that, once a politician is elected, his or her characteristics will affect policy outcomes and consequently their private welfare. To assess whether mayors’ characteristics influence municipalities’ financial performance, I use a dataset composed of 278 Portuguese mainland municipalities from 2003 to 2016. I find that mayors’ age, education, occupation, and tenure influence the level of public investment, tax revenues, debt, and budget balances. Although most of the Portuguese voters only consider candidates’ political affiliation when deciding on who to vote for, my estimates do not show any significant impact of this characteristic on the financial indicators analyzed. Therefore, these results question the way Portuguese vote by arguing that, when voting for local government representatives, they should care about other characteristics among candidates besides their political affiliation.

1 Introduction

The economic literature considers voters quasi-rational agents that care about maximizing their individual welfare when deciding on who to vote for (Porto and Porto 2000; Schuknecht 2000). As Glass (1985), Popkin (1994) and Lodge et al. (1995) argue, to make such decision, voters recall their summary affective evaluation of candidates. Each evaluation is based not only on information from past experiences, the media, and political campaigns but also on candidates’ personal characteristics such as age, gender, education, occupation, tenure and political affiliation.

Assuming that electors spend the majority of their time in the same municipality where they vote, I study two channels in which mayors’ decisions can affect the private welfare of voters.¹

Firstly, they can directly influence the level of voters’ disposable income and private consumption through their tax decisions. It is expected that higher taxes constrain private consumption thus reducing voters’ welfare.

Secondly, mayors decide on the level of public goods provided to the population such as street lighting, roads’ maintenance, and many other non-tangible goods. Higher levels of municipalities’ investment in public goods are generally related to higher levels of private welfare (Samuelson 1954).

Moreover, I explore mayors’ managerial ability by studying the impact of their characteristics on two measures: municipalities’ debt unsustainability and realized budget balances, which can be linked to both the provision of public goods and the level of tax rates. The higher the debt to revenues ratio and the higher the budget deficits, the lower the managerial ability.

On the one hand, municipalities with low levels of revenues and high levels of debt (sometimes higher than what is legally allowed) are likely to be investment constrained. On the other hand, if a municipality presents budget deficits on consecutive years by having more expenditures than revenues, it will more likely be investment constrained in the future. Consequently, they will be either less capable of offering the socially optimal amount of public goods, as argued by Monacelli et al. (2016), or they will increase their main source of revenues – the tax rates – to keep the investment level constant. Either way, the voters’ private welfare will be negatively affected by lower managerial ability.

According to a survey by the Portuguese Electoral Behaviour Project after the local elections in 2001 and 2005, around 50% of the Portuguese inquired voted for the party they preferred and did not take into account other characteristics of its head candidate.² Notwithstanding, if voters want to maximize their welfare, party affiliation might not be the only characteristic which they should take into account when deciding on who to vote for.

Hence, this paper aims to assess whether certain mayors’ characteristics impact their choices and policy outcomes. More specifically, whether mayors’ age, gender, education, occupation, tenure and political affiliation affect municipalities’ policy outcomes, namely public investment, tax revenues, debt, and budget balances.

To do so, I use a dataset comparing 278 mainland Portuguese municipalities from 2003 to 2016 and employ a generalized least squares approach. I find that mayors’ age, education, occupation, and tenure influence the policy outcomes studied, while mayors’ political affiliation does not.

These results point out that the way Portuguese electors vote is not optimal. When voting for local government representatives, they should care about other differences among candidates besides their political affiliation.

2 Literature review

2.1 How do people vote?

The traditional voting behavior theory conceives voters as quasi-rational agents who assess whether to vote by weighting the expected subjective benefit of voting, against the expected subjective cost of voting. Rogers et al. (2013) and Shafir (2013) critique this traditional view, arguing that voting is not merely a static event but rather a dynamic constellation of behavior. This new theory allows us to understand the dynamics of elections. The authors argue that voters collect information on the policy outcomes and economic performance of institutions over time. Since these outcomes potentially affect their own welfare, the information will be used to decide on who to vote for.

In his book, The Reasoning Voter, Popkin (1994) first introduces the term ‘low-information rationality’, popularly known as gut reasoning, to explain how voters make rational choices among candidates. Low-information rationality is a method of, in an economical way, combining learning and information from past experiences, daily life, the media, and political campaigns. In fact, voters are likely to focus on personal attributes of candidates to decide on who to vote for, because they believe that those features matter for policy outcomes and economic performance once a politician gets into office (Porto and Porto 2000; Schuknecht 2000).

Even though individuals forget most of the campaign information they are exposed to, they are still able to recall their candidates’ summary affective evaluation, which they use to build their preferences and vote choice (Lodge et al. 1995).

Accordingly, McDermott (2005) uses experimental survey data collected by the Los Angeles Times Poll during the 1994 statewide elections in California, to show that voters are more likely to support the most qualified candidate in races where they infer that one candidate has higher qualifications than the others. In line with these findings, Besley and Reynal-Querol (2011) argue that democracies are around 20% more likely to select highly educated leaders.

Based on empirical evidence for the US presidential elections held between 1952 and 1984, Glass (1985) argues that both highly educated and less educated individuals express interest in both trivial and important aspects of candidates’ attributes such as tenure, occupation and age. Moreover, the author shows that these characteristics have a large impact on the voting decision of both groups in the US. Such findings clearly contrast with the Portuguese case.

2.2 Do certain characteristics of politicians affect their choices?

Some empirical studies find evidence that mayor’s characteristics do not affect local governments’ financial performance indicators (Fiva, Folke, and Sørensen 2016; Ferreira and Gyourko 2009, 2014). Yet, this view of the political process has been challenged by several other empirical papers that showed a divergence in policies undertaken along partisan lines (Pettersson-Lidbom 2008; Migueis 2012; Müller et al. 2016), or along gender (Chattopadhyay and Duflo 2004; Gagliarducci and Paserman 2011).

Building upon the work of Jones and Olken (2005), who argue that political leaders enhance economic growth, Besley et al. (2011) provide evidence supporting the view that heterogeneity among leaders’ educational attainment is important. The authors suggest that economic growth is higher when leaders are more educated and longer tenured. This finding was later corroborated by Congleton and Zhang (2013).

Using a 17-year panel of Italian municipalities, Alesina et al. (2015) point out that younger mayors are more likely to strategically increase expenditures and attract more government transfers right before the election. Nonetheless, municipalities of both young and old mayors present, on average, a similar level of expenditures and revenues during their mandates.

Using a similar dataset for Italy and in the context of an anti-tax evasion program study, Casaburi and Troiano (2015) show a negative correlation between the age of mayors and tax enforcement following the program. Despite that, the tax enforcement was positively correlated when the mayor was male or born in the same municipality. These correlations provide empirical evidence that younger and male mayors collect more tax revenues than older and female peers, respectively.

More recently, based on the results of mayoral elections in all 2031 municipalities in Germany’s federal state of Bavaria from 1950 through 2009, Freier and Thomasius (2016) find that both education and tenure matter greatly in the electoral success of mayoral candidates. However, mayors’ education does not affect policy outcomes and only mayors with prior experience in office (longer tenure) are more likely to reduce the level of local public debt, enforce lower total municipal expenditures, and decrease local taxes. In contrast, Aragón and Pique (2015) use a dataset of Peruvian municipalities and find that mayors’ tenure has no significant effect on local taxes and public investment.

2.3 The Portuguese case

Portugal is a democratic country located in the southwest of Europe. It has 278 municipalities subject to the same regulatory environment.³ These municipalities are the highest-ranking authorities below the national government. Led by its mayor, they intervene at the local level by promoting social and economic development, territory organization, and by supplying public goods such as housing, healthcare, education, culture, sports, civil protection, among other responsibilities.⁴

Public investment is a shared responsibility across levels of government and according to an OECD (2018) report, local governments undertook 13% of the total public investment in Portugal. So, by planning and investing in the above mentioned key areas, they are responsible for improving the welfare of the population living in their territories.

Local governments include both municipalities and parishes (freguesias).⁵ Municipalities are divided into two branches: (i) the Town Council, which has the executive power and is responsible for the elaboration and implementation of the local policies, and (ii) the Municipal Assembly, which has the deliberative power on those same policies.⁶

Each year, the Town Council submits a report encompassing a plan of activities and investments to be executed according to each municipality’s budget. Even though the Municipal Assembly has the power to reject this plan, it can not introduce amendments to it.

Every four years, the voters registered in the municipality vote for a political party or an independent ticket in order to elect the members of the Town Council.⁷ Using the d’Hondt method, votes are then transformed into mandates and the candidate who heads the list that received the most votes is elected as the president of the Town Council, becoming the mayor. Part of the Municipal Assembly is elected directly by voters while the remaining members are the presidents of the parishes’ councils belonging to the municipality, who are also elected by voters living in their territory.

The mayor has a leading role in the executive board. He is seen as the decision-maker at the local government level. He has full managerial autonomy regarding human resources and allocates responsibilities to each member of the Town Council. Additionally, he also chooses which investment projects to pursue, how to fund and when to implement them according to the plan of activities.

Subsequently, once a politician is elected mayor, voters believe that his or her personal characteristics will affect the economic performance of the municipality. Ultimately, they expect that better performance will bring a welfare improvement, either through an increase in investment in public goods or by a decrease in debt levels and tax rates and thus, they will vote accordingly.

There are several empirical studies using data on Portuguese municipalities such as Veiga and Veiga (2005), Aidt et al., (2011), Veiga and Veiga (2014), Ribeiro and Jorge (2015), and Lopes and Peralta (2016). Interestingly, Veiga and Veiga (2007) provide empirical evidence of rational opportunistic behavior by mayors, which is expressed in their ambition to signal greater competence before elections through a reduction in taxes and an increase in expenditure items highly visible to the electorate. The authors argue that this behavior seems to be influenced by mayors’ political affiliation, namely left-wing mayors tend to behave more opportunistically than right-wing ones.

Most researchers use panel data where all the individuals are subject to the same regulatory environment to find a causal link between political leaders’ characteristics and policy outcomes (Veiga and Veiga 2007; Lopes and Peralta 2016). This feature makes this kind of panel preferable to one composed of several states where mayors have different roles and regions have sharp differences regarding their level of financial and political autonomy.

To the best of my knowledge, there is still a gap in the literature with respect to Portuguese local governments, more precisely, regarding the impact of mayors’ characteristics on local governments’ financial performance indicators. This paper tries to fill such gap.⁸

3 Empirical analysis

3.1 The model

I study the influence of mayors’ characteristics on the 278 mainland Portuguese municipalities’ financial performance between 2003 and 2016. To do so, I follow a panel data approach, since it allows for a simultaneous analysis of inter-municipalities variations (sectional dimension) and within-municipalities variations (temporal dimension). The empirical model I implement can be summarized as follows:

$$ y_{it}=\alpha+{X}_{it}^{\prime}\phi_{1}+{Z}_{it}^{\prime}\phi_{2}+{E}_{it}^{\prime}\phi_{3}+\epsilon_{it} \ \ \ \ i = 1,...,N \ \ \ t = 1,...,T $$

(1)

where y_it is one of the five dependent variables used to evaluate municipalities’ economic performance (public investment, tax revenues, debt, debt unsustainability, and budget balances. ${X}_{it}^{\prime }$ is a vector of the main time-varying explanatory variables, i.e. the mayors’ individual characteristics. ${Z}_{it}^{\prime }$ is a vector of four explanatory time-varying control variables. ${E}_{it}^{\prime }$ is a vector with dummies for the coastline characteristic, the crisis period, and the election year and its lag. Finally, ϕ₁, ϕ₂ and ϕ₃ are vectors of parameters to be estimated, and 𝜖_it is the error term. The letter i indexes municipalities and t indexes years.

I expect each municipality to have different characteristics, both observable and non-observable, and thus not taking them into account in this model will likely bias the results. By performing the Breusch and Pagan test, I conclude that there are statistically significant differences across municipalities ($Prob > \bar {\chi }^{2} = 0.0001$). So, the model should be estimated assuming those effects as fixed or random by using a generalized least squares (GLS) approach instead.

Notwithstanding, to control for those differences between municipalities that may also explain changes in the financial variables studied, I include four socioeconomic control variables (the monthly average earnings per capita, the unemployment rate, the share of the enterprises whose activities fall into the tertiary sector, and the share of the population who is more than 65 years old) and two dummy variables to account for the crisis period and if the municipality is in the coastline.

Moreover, as mayors do care about their reelection, it is reasonable to assume that throughout their mandate they will behave differently. To account for this behaviour, I also include a set of dummy variables for the election year and its lag.

Some of the main explanatory variables have no variation over time within municipalities. For instance, 151 out of the 278 municipalities studied had always elected the same political party, and 237 out of the same 278 municipalities never had female mayors. So, believing that my estimates should be mostly driven by the variation across municipalities at a moment in time (rather than by the variation over time within each municipality), I use the random effects model.⁹

Regardless of the dependent variable studied, heterogeneity can be originated from two different sources. On the one hand, it can arise from an unobserved municipality effect, where residuals are correlated across years for a given municipality. On the other hand, it can arise from a time effect, where residuals are correlated across municipalities for a given year.

In this dataset, both scenarios are likely to occur. In order to account for that, clustering on those two dimensions would be tempting. However, as the time dimension is rather short, clustering by the more frequent cluster, in this case municipalities, yields consistent results.¹⁰ Therefore, I cluster the standard errors by municipality in model 1. This clustering technique also allows for arbitrary correlation within municipalities, which corrects for the autocorrelation in our persistent dependent variables and consequently, in the error term.

3.2 Data

This study uses a panel dataset composed of annual data on economic, political, fiscal and social variables for all 278 Portuguese municipalities, from 2003 to 2016.¹¹ Data on municipalities’ local accounts were obtained from DGAL (Direção Geral das Autarquias Locais) at the municipalities’ online website (Portal Autárquico).¹²

The data on municipal balance sheets regarding debt, revenues and investments are reported by the executive board of the Town Council to DGAL after the approval of the Municipal Assembly, usually before the end of April of the year after which they refer to. These reports are very detailed, allowing for a clear analysis of each municipality’s financial information. The studied variables are reported in Table 1.¹³¹⁴

Table 1 Dependent variables

Data on mayors’ characteristics were obtained from the Ministry of Internal Affairs (Direção Geral da Administração Interna).¹⁵ These data are available in the form of information bulletins which were filled in by the executive board once they were elected and was hand-collected for each municipality at each of the four mandates analyzed.¹⁶

In each bulletin, there was information about several characteristics of the mayor and the remaining executive board members such as age, gender, education, occupation, political party, civil status, birthplace, and residency. Moreover, using Almeida (2013), I construct the dummy variables for mandates as the number of complete mandates an individual has undertaken as mayor, which will be used as a proxy for tenure (Table 2).

Table 2 Main independent variables. The source for this data is DGAI

Table 3 has data on socioeconomic control variables were obtained online from the Portuguese Institute of Statistics (Instituto Nacional de Estatística), the Portuguese Institute of Employment and Training (Instituto do Emprego e Formação Profissional), and the National Electoral Commission (Comissão Nacional de Eleições). To control for differences among municipalities and for the election cycles, I include regressors that account for their labor market, population and industry structures, as well as dummies for the coastline characteristic, the crisis period, and the election year and its lag.

Table 3 Control independent variables

In order to assess whether the chosen control variables are able to account for some of the expected differences and also to conjecture the sign of their impact, I estimate model 1 using the GLS approach only with the control variables.

In Table 4, one can see that all the control variables have a plausible impact on each financial variable. For example, if the monthly average earnings on municipality i increases by 1% one would expect the level of public investment to decrease by 1.5% and the level of tax revenues to increase by 1.7% ceteris paribus. Interestingly and corroborating the findings of Veiga and Veiga (2014), I also observe a negative impact of the election year dummy on the budget balance variable amounting to 66.58 euros, on average, ceteris paribus. Finally, one can also confirm that in times of crisis the debt levels increased, most likely to face the decrease in tax revenues.

Table 4 Control variables relation

Here, I find some evidence in favor of the opportunistic behavior of local governments, corroborating most of the studies on election cycles in Portuguese local governments such as Baleiras and Santos (2000) and Veiga and Veiga (2007). In pre-electoral periods, incumbent mayors significantly increase the level of public investment incurring in budget deficits with higher probability. This behavior is consistent with an effort to signal competence while increasing the chances of reelection.¹⁷

All control variables are statistically significant so, for the remaining estimates, I will always include this set of variables.

4 Results

4.1 Estimates

Table 5 summarizes the main results of this study.¹⁸ According to the first column, if a mayor from a municipality has completed at least the Bachelor degree, the yearly public investment is expected to be 12.5% lower when comparing to mayors who completed at most the secondary school, on average, ceteris paribus.¹⁹ Moreover, if a mayor has completed at least one mandate, the yearly public investment is expected to be approximately 8% higher, on average, ceteris paribus.

Regarding the second column, I find statistical evidence that differences in the previous mayor occupation impact the level of tax revenues. A municipality ran by a mayor who is seasoned in the financial and management world, that is, a mayor who has a high proficiency job directly related to the management of enterprises and/or public companies is expected to collect more taxes. Such mayors are expected to collect more 4.3% of taxes when compared to mayors that are either retired or have other kinds of jobs.

Table 5 Main results

The third and fourth columns should be interpreted together because higher debt levels is not a bad thing in itself. However, if its increase is followed by a rise in the ratio of debt to revenues, then the municipality is more likely to become investment constrained. The only significant impact on both of these variables is a quadratic effect of the mayor’s age.

As expected, one can also see that both the gender and the political variables do not impact significantly any of the studied financial variables. Even though, there is some evidence that both center right and center left parties do enjoy higher budget balances, that evidence does not survive the robustness checks discussed in the following section. Hence, one can already see that the Portuguese voters’ belief might be unfounded.²⁰

4.2 Robustness checks

I present several regression specifications for the dependent variables. For the first layer of robustness checks in the Online Appendix, specification (1) reports estimations based on the GLS methodology, with random effects - the same as in Table 5.²¹

Specification (2) reports model 1 with all 308 Portuguese municipalities, including the Portuguese islands. Although these 30 regions are subject to an extra layer of intervention most of the regressors maintain their signs and statistical significance. Table 7 also presents similar descriptive statistics.

In many panel data empirical applications, the residuals from different cross-sectional units are likely to be correlated with one another (De Hoyos and Sarafidis 2006). In this study, this can be seen typically in neighboring municipalities.²² Imagine, for instance, a municipality which constructs from scratch a big leisure park. Even though this investment is targeted at its population, individuals from neighboring municipalities are free to visit it. Therefore, the neighboring local governments do not need to provide a similar public good, and thus their investment, tax, and borrowing decisions will be affected not only in the short but also in the medium run.²³

In this paper, I opt not to capture this neighborhood effect but rather to perform two final robustness checks. Specification (3) is based on the GLS methodology with fixed-effects which relies on the within-variation to exclude time-invariant municipalities characteristics. However, Driscoll and Kraay (1998) claim that when there are arbitrary cross-sectional correlations in the residuals, time dummies will not mitigate the problem of spatial dependence so, specification (4) reports the estimates using the xtscc program presented in Hoechle (2007) which produces Driscoll and Kraay’s (DK’s) standard errors. The later are robust to the autocorrelation, heteroskedasticity, and cross-sectional dependence existing in my dataset. Finally, specification (5) performs the same estimation without the variable for the education level which has approximately 700 missing observations.

In the Online Appendix, I also present a second layer of robustness checks. Specification (1) reports estimations based on the GLS methodology, with random effects with an extra control variable - the unconditional government transfers which are formula-determined and thus, exogenous to mayors’ characteristics. These transfers have a significant impact in all variables and thus are included in all further specifications but have no impact on the results presented before.

Specification (2) is a sanity check where I include other mayors’ characteristics which also might play a significant role. The variables are:

* Majorities - a dummy which indicates whether mayors enjoy a majority position in both the Town Council and the Municipal Assembly. It is relevant because it indicates whether a mayor has a bigger margin of maneuver to approve a budget.

* Birth Place and Foreigner - dummies that take the value 1 if mayors were born in the same municipality that they represent or if she was born outside Portugal, respectively. Here, I am looking for a skin in the game effect, where people born in the same municipality might have bigger concerns about financial stability and the long term perspective.

* Residency - a dummy which indicates whether mayors live in the same municipality that they represent. Here, I am also looking for the skin in the game effect.

As expected, mayors with majorities on both the Town Council and Municipal Assembly are able to have significantly higher levels of public investment and debt. Nevertheless, they do not have significantly higher budget deficits nor debt to revenues, meaning that having a majority increases the management ability and freedom in a very positive way.

The skin in the game effect is also an interesting finding. Mayors who were born in the same municipality have significantly lower levels of debt, the debt they have is more sustainable and present more positive budget balances. This probably happens because they have a genuine concern about the long term financial stability of the municipality.

As one can see in Table 6, the means of all dependent variables are bigger than their medians, which is due to the presence of outliers. Most of these outliers occur when the municipalities make large investments, which are usually funded with a mix of mostly debt, own revenues and government transfers.

For example, the municipality of Fornos de Algodres increased public investment spending from 1.6 in 2009 to 15 million euros in 2010 and then decreased it to 2.1 million euros in 2011. In 2010, its Town Council borrowed over 30 million euros not only to pay old debts to suppliers and workers, but mainly to invest in several projects, namely the construction of an elementary school, a municipal library, and a central bus station.

I add an outlier dummy variable to assess the robustness of my estimations in the presence of outliers.²⁴ According to specification (3), all of the outlier dummies are statistically significantly different from zero with the exception of debt unsustainability.

Even though it does not disturb the significance of my estimates for any variable, I estimate the same model without the outliers pinpointed in specification (4). Not surprisingly, the estimates remain unaltered.

Adding to the previous checks, this result shows that my findings are indeed robust and that, by using the proposed model, we can infer whether mayors’ characteristics affect municipalities’ financial performance and hence, voters private welfare.

5 Discussion

The results from this paper add to the literature on how certain characteristics of politicians affect their choices and policy outcomes. It provides evidence that characteristics of local governments’ political leaders such as age, education, professional occupation and tenure matter for policy outcomes and the financial stability of municipalities.

Alesina et al. (2015) argue that younger and older mayors choose a similar level of expenditures on average during their term. In my model, age seems to have a quadratic relationship with debt and debt to revenues ratio. Both the youngest and oldest mayors incur in higher and unsustainable debt levels.²⁵

It is plausible that younger and older politicians differ from middle-aged ones. Younger mayors have potentially longer career horizons, and therefore more opportunities outside politics (Alesina et al. 2015). Besides having shorter career horizons, older mayors are probably less energetic and less productive at work. Thus, both groups have fewer incentives and/or capacities to keep the municipalities’ debt level under control, which corroborates the idea of a quadratic relationship.

The results also show that gender does not play a significant role. This finding is in line with Gagliarducci and Paserman (2011) and Ferreira and Gyourko (2014) who found that the mayor’s gender has essentially no effect on investment, tax revenues and their managerial ability. Nonetheless, we must be careful when doing such inference since female mayors account for only 7% of the sample.

Mayors with prior experience in office, i.e. with at least one completed mandate, are expected to invest more. On the one hand, this relationship may be driven by the opportunistic behavior of mayors as they seek to increase the probability of reelection (Veiga and Veiga 2007). On the other hand, we know that one of the key drivers of the investment at the municipal level is the gathering of public funds, especially European funds as Veiga and Pinho (2007) show.²⁶ The trend of increasing available public funds for municipalities started in 1986 when Portugal joined the European Economic Community, and it is growing for more recent programs. The current one is Portugal 2020 where, from 2016 to 2020, municipalities will receive up to 2 billion euros of European funds, which corresponds to two times the municipalities’ investment level in 2015 (around 1.2 billion euros).

In order to successfully engage with such funds, mayors lead a team mainly composed of Town Council executive board members. More experienced mayors that are used to lead such teams are more likely to receive European and other community funds, which are vital to boost the level of public investment and substitute tax collection and external borrowing.

According to Aragón and Pique (2015), the lack of effect of tenure on other policy outcomes may reflect quick “learning-by-doing” when in office. It can also arise since there may be effective substitutes for mayors’ on-the-job experience, such as education or previous occupations as discussed before.

A survey implemented by the Portuguese Electoral Behaviour Project right after the local elections in 2001 and 2005 considered around 50% of the Portuguese inquired as “staunch electors”. Bourdain (2008) explains that these electors voted for the party they preferred and did not take into account other characteristics of the head candidate because they believed that being a representative of a given political party is the most important characteristic for policy outcomes. Such information alone made me expectant for some effect but, surprisingly, I do not find statistical evidence supporting this common belief, as mayors from different parties do not behave differently regarding the level of investment, tax revenues, nor managerial ability.

This finding is in line with Ferreira and Gyourko (2009) and Fiva et al. (2016) who use evidence from US cities and Norwegian municipalities to show that the mayor’s political affiliation does not affect the allocation of local public spending. They argue that most of the empirical models discussed are likely to predict less partisanship the more alike the residents of any jurisdiction are. Reasoning inspired in the theoretical framework set by Tiebout (1956) on local expenditures.

6 Conclusion

In local elections, voters choose between candidates by focusing on their personal attributes. They believe that, once a mayor is elected, its characteristics will affect the economic performance of the municipality and consequently their private welfare. I use a GLS approach to study the impact of mayors’ age, gender, education, occupation, tenure and political affiliation on municipalities’ policy outcomes, namely public investment, tax revenues, debt, and budget balances.

Portuguese mainland municipalities are all subject to the same regulatory environment and enjoy some political freedom. Moreover, mayors have substantial autonomy to make decisions and are made accountable for them. Thus, the data used in this study present a good setup to study such relationships.

I find that mayors’ age displays a “quasi-quadratic” relationship with debt levels, that mayors’ tenure and higher education positively affect the level of public investment, and that mayors’ occupation, when directly related to managing enterprises, has a positive impact on the amount of collected taxes. Contrarily to the electors’ belief, I do not find that mayors’ political affiliation influences any of the financial indicators studied.

These results make us question the way Portuguese electors vote by showing that, when voting for local government representatives, they should care about other characteristics among candidates besides their political affiliation.

Appendix A: Descriptive statistics

Table 6 Descriptive statistics - main sample

Table 7 Descriptive statistics with Azores and Madeira’s municipalities

1.1 Appendix B: Estimates

Table 8 Dependent variable: log(public investment)

Table 9 Dependent variable: log(tax revenues)

Table 10 Dependent variable: log(debt)

Table 11 Dependent variable: debt unsustainability

Table 12 Dependent variable: budget balance

1.2 Appendix C: Information bulletin