Hostname: page-component-7b7567c8c-kwt9d Total loading time: 0 Render date: 2026-11-12T11:45:38.090Z Has data issue: false hasContentIssue false
  • English
  • Français

Public Employees as Politicians: Evidence from Close Elections

Published online by Cambridge University Press:  09 August 2017

ARI HYYTINEN,
JAAKKO MERILÄINEN,
TUUKKA SAARIMAA,
OTTO TOIVANEN and
JANNE TUKIAINEN
ARI HYYTINEN*
Affiliation:
University of Jyväskylä, School of Business and Economics (JSBE)
JAAKKO MERILÄINEN*
Affiliation:
Stockholm University
TUUKKA SAARIMAA*
Affiliation:
VATT Institute for Economic Research
OTTO TOIVANEN*
Affiliation:
Aalto University School of Business and University of Leuven
JANNE TUKIAINEN*
Affiliation:
London School of Economics and Political Science and VATT Institute for Economic Research
*
Ari Hyytinen is a Professor, University of Jyväskylä, School of Business and Economics (JSBE), P.O. Box 35, 40014 University of Jyväskylä, Finland (ari.hyytinen@jyu.fi).
Jaakko Meriläinen is a PhD Candidate, Institute for International Economic Studies, Stockholm University, IIES, Stockholm University, SE-10691, Stockholm, Sweden (jaakko.merilainen@iies.su.se).
Tuukka Saarimaa is a Senior Researcher, VATT Institute for Economic Research, Arkadiankatu 7, Helsinki, FI-00101 (tuukka.saarimaa@vatt.fi).
Otto Toivanen is a Professor, Department of Economics, Aalto University School of Business and a Professor, Department of Managerial Economics, University of Leuven, Arkadiankatu 7, 00100 Helsinki, Finland; Naamsestraat 69, B-3000 Leuven, Belgium (otto.toivanen@aalto.fi).
Janne Tukiainen (corresponding author) is a Visiting Associate Professor, Department of Government, London School of Economics and Political Science; and Senior Researcher, VATT Institute for Economic Research, Houghton Street, London WC2A 2AE, United Kingdom; Arkadiankatu 7, Helsinki, FI-00101, Finland (janne.tukiainen@vatt.fi).
Rights & Permissions [Opens in a new window]

Abstract

We analyze the effect of municipal employees’ political representation in municipal councils on local public spending. We use within-party, as-good-as-random variation in close elections in the Finnish open-list proportional election system to quantify the effect. One more councilor employed by the public sector increases spending by about 1%. The effect comes largely through the largest party and is specific to the employment sector of the municipal employee. The results are consistent with public employees having an information advantage over other politicians, and thus, being able to influence policy.

Information

Type
Research Article
Information
American Political Science Review , Volume 112 , Issue 1 , February 2018, pp. 68 - 81
Copyright
Copyright © American Political Science Association 2017 

INTRODUCTION

In 2013, public sector employees accounted on average for 21% of total employment in the OECD countries (OECD 2015). While heterogeneous, they are a large group that share an interest in sustaining public employment and that can influence politics in various ways. In addition to a direct voting channel (Garand Reference Garand1988; Blais, Blake, and Dion Reference Blais, Blake and Dion1990; Bhatti and Hansen Reference Bhatti and Hansen2012), recent research has emphasized the role of public sector unions and their effects on the cost of government, either directly through collective bargaining or indirectly through politics (Sieg and Wang Reference Sieg and Wang2013; Anzia and Moe Reference Anzia and Moe2015).

Quite often public sector employees are also politicians themselves.Footnote 1 This dual role of public sector employees has raised the concern that, when elected, they may be in a better position to extract rents from holding the office than otherwise similar politicians. A concrete example would be a teacher in a municipal council that decides whether the teacher's school should be closed or not, or a public sector nurse participating in deciding on budget cuts in the local public health care sector. In both cases, the public sector employees can possibly exert disproportionate influence in the council due to their information advantage over the other councilors on the true costs and benefits of providing public services in their sector of employment (Niskanen Reference Niskanen1971; Romer and Rosenthal Reference Romer and Rosenthal1979). This disproportionate influence may compromise the political neutrality of public service and also undermine the separation powers more generally (Braendle and Stutzer Reference Braendle and Stutzer2016).

Consistent with such concerns, countries often impose incompatibility and to a lesser extent ineligibility rules on the political involvement of public sector employees.Footnote 2 The former force public employees to give up public service if elected and the latter require giving up public service if they run (Braendle and Stutzer Reference Braendle and Stutzer2016). Imposing such restrictions involves a trade-off by limiting the political participation of a group with possibly ample opportunities for rent-seeking at the cost of discriminating against a large citizen group and excluding informed candidates.Footnote 3 There is surprisingly little evidence on whether public-sector employees would act differently from the other politicians when elected. We start to fill this important gap in the literature by providing causal evidence on how municipal employee representation in a municipal council affects local public spending and on the mechanisms at work.Footnote 4

We use data from Finland. Finland provides a particularly interesting context for our analysis for two reasons. First, almost 30% of employment in Finland is in the public sector and more than 20% of employment is in the local public sector.Footnote 5 An important feature of Finnish local politics, and common in other countries as well (e.g., the UK), is that being a municipal councilor is not a full-time job. About one quarter of the Finnish local politicians work for a municipality.Footnote 6 The distribution of power between private- and public-sector employees in the municipal councils may therefore have a large impact on the size and efficiency of the local public sector. Reflecting this tension and its topicality, the Finnish media has expressed concerns that, when elected, municipal employees can make decisions on their own jobs in municipal councils.Footnote 7

The second reason why Finland provides a suitable context for our study is that the Finnish open-list local elections provide us with plausibly exogenous variation in municipal employee representation. The source of this variation is candidate-level close contests within party lists. We use these contests to construct a municipality-level instrument variable for municipal employee representation. Our instrument captures the extent to which the seat share of municipal employees exceeds or falls short of their expected share due to randomness in the outcomes of the close elections. The identifying assumption is that when measured at the candidate level and sufficiently close to the within party election thresholds, the seat allocation between municipal employees and other candidates can be considered to be as-good-as random. This assumption can be tested indirectly by covariate balance tests. We define candidate-level closeness within the party lists to make sure that differences in party representation (party effects) are not driving the results.

Our main result is that electing one additional municipal employee to a council as opposed to a candidate from the same party, but not employed by the public sector, increases local public spending. Our estimates suggest that in a municipality with a median-sized council (27 seats), the spending increases on average by about 1% over the 4-year council term. The effect is surprisingly large considering two features: First, we are probably looking at a relatively unimportant margin, i.e., the last elected candidate within a party to a council that typically consists of tens of councilors. Second, as we explain later, there are explicit restrictions on the types of political positions that Finnish municipal employees can take in their home municipality. Our result is nevertheless in line with previous findings which show that smaller parties and even individual councilors have an effect on policy in proportional representation systems (Folke Reference Folke2014; Freier and Odendahl Reference Freier and Odendahl2015; Fiva and Halse Reference Fiva and Halse2016).Footnote 8

We also provide evidence on the mechanisms at work. First, we show that the effect varies by the type of municipal employee and the type of spending: electing one more employee who works in health care leads to a significant increase in health expenditures, but not in the other (non-health) municipal expenditures. Similarly, when a non–health care employee gets elected, expenditures unrelated to health care increase.Footnote 9 This evidence is consistent with Niskanen's (Reference Niskanen1971) classic bureaucracy model which predicts that bureaucrats can convince politicians to increase public spending due to their information advantage over politicians. The analogy we draw is that municipal employee politicians have both different incentives than and information advantage over the other politicians. Moreover, we find that the positive effect on local public spending arises in particular in close elections that involve the largest party in the municipality and in smaller councils. This evidence suggests that municipal employee councilors influence outcomes through intraparty decision making (Laver and Shepsle Reference Laver and Shepsle1990).

The increased sector-specific spending cannot be automatically attributed to rent-seeking. A reason for this is that municipal employees are experts in their area of employment and can therefore provide useful information to other councilors and improve decision making.Footnote 10 Even though we cannot conclusively differentiate between the competing hypotheses about the efficiency of the increased spending, it is definitely noteworthy − and somewhat puzzling—that the Finnish municipal councilors employed by the public sector want to increase public expenditures in a country that in 2014 had, at 59% (OECD 2015), the highest public sector ratio to GDP among all OECD countries (between 1996–2012 Finland's position varied between 2nd and 8th). The uniform increase in spending is puzzling because our as-good-as-random design guarantees that the citizens’ needs are identical on average in the treated and other municipalities. Viewed from this angle, Niskanen's (Reference Niskanen1971) concerns about bureaucrats’ information advantage leading to excessive spending seem warranted.

The rest of the article is organized as follows: In the following section, we discuss relevant theory and outline testable hypotheses. Then, we describe the institutional setting and data. We present our econometric identification strategy in the fourth section and the results in the fifth section. The final section concludes the study. Auxiliary results are available in the Online Appendices.

THEORY AND HYPOTHESES

We are interested in (i) whether, (ii) through which mechanisms, and (iii) why public employee representation in municipal councils has an effect on municipal spending. We discuss each of these in turn.

Effect on total expenditures: At least two distinct theoretical debates bear directly on whether public employee representation has an effect on local government expenditures. The first view is that public employees have both the economic incentives and the means to maximize the municipal budget to their own benefit (Niskanen Reference Niskanen1971; Courant, Gramlich, and Rubinfield Reference Courant, Gramlich and Rubinfield1979; Dahlberg and Mörk Reference Dahlberg and Mörk2006). This is likely to obfuscate the separation of powers between the executive and the democratically chosen political branches of the local government (Braendle and Stutzer Reference Braendle and Stutzer2016). A public sector employee politician may also have a variety of ways to target public spending to certain voters, such as her own political constituency or interest group (Alesina, Baqir, and Easterly Reference Alesina, Baqir and Easterly2000). The second view posits that the preferences of public sector employees differ systematically from the individuals employed by the private sector. One reason for the difference is that professional background determines socioeconomic conditions and may thus shape identity (Braendle and Stutzer Reference Braendle and Stutzer2016). Consistent with this, public sector employees seem to be politically more active (Bhatti and Hansen Reference Bhatti and Hansen2012) and lean more to the left ideologically (Knutsen Reference Knutsen2005; Jensen, Sum, and Flynn Reference Jensen, Sum and Flynn2009; Rattsø and Sørensen Reference Rattsø and Sørensen2016). They may also be relatively unwilling to support market-oriented solutions, and thus, a smaller public sector.

Taken together, these views suggest that public sector employees have a tendency to favor a larger public sector. We therefore formulate the following hypothesis:

Hypothesis 1: The council seat share of municipal employees increases municipal spending.

We acknowledge the possibility that municipal employees may be more pro-social and better motivated for public service than other candidates from the same party (Francois Reference Francois2000; Besley and Ghatak Reference Besley and Ghatak2003, Reference Besley and Ghatak2006). If that was the case, spending could also decrease (remain unchanged) if the status quo level of per capita spending is too high (optimal) from the social point of view. Municipal employees may also be relatively immune to specific business interests and lobbying (Braendle and Stutzer Reference Braendle and Stutzer2016), which may reduce inefficient spending to public procurement.

Mechanisms at work: How and through which mechanisms could the spending effect come about? This question is of interest, because there are institutional restrictions on the political representation of the municipal employees (see the next section) and because our empirical close-elections approach identifies by design the effect of allocating the last marginal seats to the council. Two key mechanisms suggest themselves: First, holding other things constant, an individual councilor is likely to exert a greater impact in smaller councils because the likelihood of him (or his party) being pivotal is higher. In a smaller council, a single councilor can also pivot informal within-council discussions to his own advantage and influence which issues the council tackles. Second, the literature on coalitional bargaining (Laver and Shepsle Reference Laver and Shepsle1990) suggests that councilors can influence decision making either between-parties or within-parties. In the former case, the municipal employees would vote in the council as if they had a coalition of their own, independent of the formal parties and the municipal employees’ party affiliation. If, on the other hand, the channel of influence is within-parties, the party lines hold, but municipal employee councilors affect the policy position of their own party. This is a plausible channel of influence in our context because public employees may be a relatively loose and heterogeneous interest group. Moreover, the within-party channel matters for policy outcomes only if the public employees’ party is large and powerful enough in the council. In sum, we have the following two hypotheses about the mechanisms of influence:

Hypothesis 2a: Municipal employees of smaller councils have a greater effect on spending than those of larger councils.

Hypothesis 2b: Municipal employees of larger parties have a greater effect on spending than those of smaller parties.

Rent-seeking vs. efficient provision of public services: Finally, we study whether the effect of public employee representation on public spending mirrors rent-seeking, or is more consistent with efficient provision of public services.

Applied to our context, Niskanen's (Reference Niskanen1971) classic model of bureaucracy predicts that municipal employee councilors have an information advantage over the other municipal councilors about the provision of public services in their own employment sector and that the municipal employees are less likely to have such an advantage over the other public services. We therefore formulate:

Hypothesis 3a: Municipal employees never decrease spending and they increase spending especially in their own sector of employment.

This hypothesis would not get support from the data, if municipal employees lean more to the left and generally favor a larger public sector. In this case, municipal employees ought to increase spending also in sectors other than their own sector of employment. Moreover, while the expertise of municipal employees can also be useful for the efficient provision of public services (Braendle and Stutzer Reference Braendle and Stutzer2016), it is unlikely that, holding citizens’ needs constant, such efficient provision systematically calls for greater spending, especially only in their own sector.

Finally, we also look for more direct signs of rent-seeking (Svaleryd and Vlachos Reference Svaleryd and Vlachos2009). Inefficiencies may arise also through clientelistic behavior (i.e., explicit or implicit quid-pro-quo for political support; see Alesina, Baqir, and Easterly Reference Alesina, Baqir and Easterly2000), which are at least partly captured through better re-election prospects. We therefore formulate:

Hypothesis 3b: Municipal employees enjoy higher returns to office in terms of larger salary and smaller unemployment risk, and/or enjoy from a larger incumbency advantage in subsequent elections than the other candidates.

INSTITUTIONAL SETTING AND DATA

Finnish Local Governments

Tasks and revenue sources of municipalities: Finland has a two-tier system of government consisting of the central government and municipalities as the local level (see Saarimaa and Tukiainen Reference Saarimaa and Tukiainen2015). Finnish municipalities have extensive tasks. In addition to the usual local public goods and services, municipalities are responsible for providing most of social and health care services and primary and secondary education. The GDP share of municipality spending is large (roughly 18%) and the municipalities employ around 20% of the total workforce.

Municipalities have extensive fiscal autonomy. The most important revenue source is the flat local income tax, determined by the municipalities. There are, however, large regional tax base and cost disparities. They are offset by a central government grant system, which the municipalities cannot effectively manipulate to their advantage.

Decision making and elections in municipalities: Municipalities are governed by a municipality council which is the most important political actor.Footnote 11 For example, mayors or city managers are public officials chosen by the councils and have only executive power and no political power. Moreover, municipal boards (i.e., cabinets) have only a preparatory role and the representation in the boards follows the political distribution of the council.

Municipal elections are held simultaneously in all municipalities and each municipality has one electoral district. The elections in our data were held on the fourth Sunday of October in 1996, 2000, 2004, and 2008. The council term starts in January of the year following the election year. The term lasts 4 years.

Within each municipality, the allocation of seats is based on proportional representation, as determined by the open-list D'Hondt election rule. In an election, each candidate has an affiliation with a party list and each voter gives a single vote to a single candidate. The voters cannot vote for a party without specifying a candidate. The total number of votes for the candidates in a given party list determines the votes for each party. The party votes determine how many seats each party gets according to the D'Hondt rule. Given these party seats, the competition for the seats within parties is simply an n-past-the-post rule. The rank of a candidate within the party list is determined by his votes, implying that voters—as opposed to parties—decide which candidates are elected from a given party list.

There are nationwide restrictions on the political roles of municipal employees.Footnote 12 First, a municipal employee who is in an executive position of a branch of public service cannot be a council member. For example, the director of a municipality's school authority cannot be a member of the municipal council. Second, a municipal employee cannot be a member of the subcommittee of his own specific sector. For example, a teacher cannot be a member of the subcommittee for education. Third, a municipal employee working in administrative duties directly under the municipal board cannot be a member of the board. Fourth, a municipal employee who is the presenting official for matters dealt by the municipal board cannot be a member of the board. Fifth, the majority of the municipal board cannot consist of municipal employees. Finally, a municipal employee councilor can participate in the decision making in the council meeting even if the matter relates to her own employment, unless she has been directly involved in preparing or presenting the matter as a bureaucrat for the council.

The broader institutional context may also limit the opportunities of the municipal employees to influence outcomes and extract rents while in office. For example, Finland is one of the least corrupted countries in the world. Moreover, wages are largely set at the national-level wage bargaining between the municipal employer organization and various labor unions. However, a municipality can pay more than agreed upon nationally.

Data

Our data come from a number of sources:

Candidate and elections data: We have obtained data on municipal elections held between 1996 and 2008 from the Ministry of Justice. These data consist of candidate-level election results, in particular party affiliation, number of votes, and elected status. The election data also include the age and gender of the candidates. Information on municipal employment status comes from KEVA, which manages local government pensions, and we have linked the candidate data also to Statistics Finland data on education, occupation and socioeconomic status and to income data from the Finnish Tax Authority.

Overall we have roughly 160,000 candidate-election observations (see Online Appendix A for descriptive statistics and descriptions of sample restrictions). For our purposes, a candidate is a municipal employee, if she was employed by a municipality at the end of the election year. Compared to other candidates, municipal employees are more often female (nurse is the most common profession among them), classified as high professionals in terms of socioeconomic status, and running for the Social Democratic Party. We return to these observable differences in candidate characteristics when we present our econometric analysis and results.

Municipal data: We use Statistics Finland's data on municipal expenditures and demographics for the years 1996–2012. We have 1,544 municipality-council term observations (see Online Appendix A for the summary statistics): On average, municipalities’ total expenditures are 5,500 euros per capita. The single most important expenditure category is health care (1,700 euros per capita). Municipal employees’ seat share is on average 26.4%.

ECONOMETRIC APPROACH

Identification Strategy

To estimate the effect of political representation of municipal employees on municipal policy, we use the following regression specification:

(1) $$\begin{equation} {Y_{mt}} = {\rm{\delta }}{M_{mt}} + {X'_{mt}}{\rm{\beta }} + {u_{mt}}, \end{equation}$$

where Ymt is the outcome of interest, Mmt is the seat share of municipal employees in the council, X ' mt is a vector of control variables (possibly lagged), and umt is the error term in municipality m at time t. The parameter of interest is δ which measures the effect of a change in the seat share of municipal employees on the outcome.

Our main outcome variable is municipal expenditures. A simple OLS estimation of equation (1) may suffer from both reverse causality and omitted variable bias. This could be the case if, e.g., voters in a municipality demand a high level of municipal services. Such a municipality would have a high number of municipal employees. This calls for greater municipal expenditures and would show up as a greater council seat share of public sector employees as well.

We make use of close elections to estimate the treatment effect of interest (δ). Unlike in much of the recent literature using close elections for identification, the Finnish municipal election system of proportional representation with open party lists does not render itself to a simple regression discontinuity design (RDD) analysis (Lee, Moretti, and Butler Reference Lee, Moretti and Butler2004). The reason is simple: Despite there being an RDD flavor to our close elections approach, we cannot construct a well-defined forcing variable at the municipality level. We therefore build on Clots-Figueras (Reference Clots-Figueras2011, Reference Clots-Figueras2012) who uses the fraction of women winning close elections as an instrument for the share of women in the legislature (see also Folke Reference Folke2014 and Freier and Odendahl Reference Freier and Odendahl2015, who use IV and close elections data). Our IV procedure uses as-good-as-random variation at candidate-level in the close elections and aggregates this variation to get a municipality-level instrumental variable. To properly capture the treatment effect of political representation of municipal employees, we focus on closeness within party lists. This choice means that between-party differences do not confound our results. For example, if municipal employees are more often left- than right-wing, between party comparisons would give us the joint effect of municipal employees and party status.Footnote 13

We construct our instrument in the following steps:

Step 1: For each party list p, we define the pivotal number of votes as the average of the maximum number of votes among the nonelected candidates and the minimum number of votes among the elected candidates. The distance to getting elected for each candidate is the number of votes of the candidate minus the pivotal number of votes of her party list. We normalize this distance by dividing it by the total number of votes of the party list and then multiplying it by 100. We denote the variable thus obtained vipmt .Footnote 14 Closeness of each candidate i in party list p in municipality m in election t, Cipmt , is then defined as

(2) $$\begin{equation} {C_{ipmt}}{\rm{\ }}\;{\rm{ = }}\;{\rm{\ }}\left\{ {\begin{array}{@{}*{1}{l}@{}} {{\rm{1\ if\ |}}{v_{ipmt}}{\rm{|\ }} \le \varepsilon }\\[8pt] {{\rm{0\ if\ |}}{v_{ipmt}}|{\rm{\ > \ }}\varepsilon } \end{array}} \right., \end{equation}$$

where ε is some small bandwidth, expressed in percentages (e.g., ε = 0.4 means “0.4%”; that is, 4 votes out of 1,000). Due to randomness in the outcomes of elections, candidates just above and below the pivotal number do not differ systematically from each other. Indeed, when ε = 0 in our data, there was a tie within a party list between two (or more) candidates at the threshold of getting into the council. In such a case, a lottery decides which of the candidates are elected (see Hyytinen et al. Reference Hyytinen, Meriläinen, Saarimaa, Toivanen and Tukiainen2017 for details). There are 1,351 candidates who end up in these lotteries and 335 of them are municipal employees.

Step 2: Quasi-randomization taking place within each party list influences how many municipal employees get elected from each list. To capture this list-level variation, we calculate the difference between the realized outcome and the expected outcome of the close races within each party. Formally, this can expressed as

(3) $$\begin{eqnarray} T_{pmt} &=& \left( \sum \limits_i^{N_p} C_{ipmt}D_{ipmt}M_{ipmt} \right)\nonumber\\ && -\, \left[ \frac{\sum_{i}^{N_p} C_{ipmt}M_{ipmt}} {\sum_i^{N_p} C_{ipmt}}\sum \limits_i^{N_p}C_{ipmt}D_{ipmt}\right], \end{eqnarray}$$

where Mipmt is equal to one if candidate i is a municipal employee and zero otherwise, Dipmt equals one if candidate i in municipality m was elected in election t and zero otherwise and p refers to a party list and Np to the number of candidates in the list p. The first term is the number of municipal employees that are elected in the close elections. The second term is the expected number of municipal employees who get elected in the close elections. The expected number comes from a hypergeometric distribution, because close elections can be seen as a basic urn problem.Footnote 15 The reason for using Equation (3) is that there may be more than two candidates that are close and thus subject to randomization and any number of the close candidates can be municipal employees.Footnote 16 Moreover, the set of candidates defined as close may compete for more than one seat within the party list. These features are the main difference between our and Clots-Figueras’ (Reference Clots-Figueras2011, Reference Clots-Figueras2012) approach, because she considers only situations where one male and one female candidate compete for one seat.

Step 3: We aggregate the random variation at the party list level to construct a municipal-level instrumental variable, Tmt . This is done by adding up Tpmt over all the party lists within a municipality and by dividing the sum by council size (CS):

(4) $$\begin{equation} T_{mt} = 100^*\left( \sum\limits_p T_{pmt}\right)/CS_{mt} \end{equation}$$

Our instrument, Tmt , captures the extent to which the seat share of municipal employees exceeds (Tmt > 0) or falls short of (Tmt < 0) their expected share due to randomness in the outcomes of the close elections. In other words, the instrument obtains higher values for those municipalities in which the municipal employee candidates were lucky and smaller values for those municipalities in which they were unlucky. If, in a given municipality, municipal employees were lucky within one party list and equally unlucky in another, the instrument at the municipal level would be zero. One can think of Tmt as the part of the variation in Mmt that is as-good-as random. Our IV approach thus assumes that Tmt is a determinant of Mmt , i.e., the (actual) seat share of municipal employees in the council and uncorrelated with umt in (1). This assumption can to an extent be tested using municipality-level covariate balance tests. Moreover, the candidate-level bandwidth can be used to check the robustness of the results to the bandwidth choice.

Empirically, Tmt appears to work as expected (see Online Appendix B for details): First, it is symmetrically distributed around zero. Second, when the seat share of municipal employees increases due to randomness in the outcomes of the close elections (i.e., when Tmt increases by one unit), so does their actual share (i.e., Mmt ). This implies that the coefficient of Tmt in the first stage of the IV should be close to one in a finite sample and equals unity asymptotically. This is indeed empirically the case in our data (Online Appendix B). This feature means that in the reduced form of our IV of equation (1), which means regressing the outcome directly on the instrument (and controls; Wooldridge Reference Wooldridge2002, Ch. 5), the coefficient of the instrument ought to be very close to the IV estimate of δ. This observation is useful and in the subsequent section, we will report both the 2SLS and the reduced form of IV results.

Finally, even with the smallest possible bandwidth (ε = 0), we have variation in Tmt , as explained above. As we increase the bandwidth, almost all municipalities in our data have a close contest within at least one party list. For example, for bandwidth ε = 0.4, we observe either a positive or a negative instrument in 1,145 municipalities out of 1,544. This does not imply that we would use for identification all the variation in the municipal employee council seat share in the data for these 1,145 municipalities: To estimate δ, we only use the part of exogenous variation in Mmt that the instrument, Tmt , isolates.

Validity Tests

In Table 1, we report balance tests for pretreatment covariates using the largest bandwidth that is employed in the regressions (ε = 0.4).Footnote 17 We divide the data into two groups, based on the seat share of municipal employees exceeding (Tmt > 0) or falling short of (Tmt < 0) its expectation. Table 1 shows that the pretreatment variables are well balanced, including the lagged total expenditures, the lagged municipal employee share in the council, and its lagged instrument.Footnote 18 This means that the municipalities where the municipal employees won, by chance, more seats are very similar to the municipalities where municipal employees lost, by chance, seats to other occupation groups.

TABLE 1.

Pretreatment Covariate Balance at Municipality Level

Notes: The statistical significance of the differences is tested using a t-test adjusted for clustering at the municipality level. ***, **, and * denote statistical significance at 1%, 5%, and 10% level, respectively.

Table 2 reports balance tests on council characteristics for the current election term.Footnote 19 As Panel A shows, the posttreatment council characteristics are well balanced. For example, the municipal employees that by chance won a seat from a candidate from another occupation are of no better or worse quality (see Ferreira and Gyourko Reference Ferreira and Gyourko2014 who argue that, e.g., gender discrimination would imply that candidates with the same number of votes would be of different quality), as measured by their incumbency and education. The only exception to the good balance is the councils’ gender composition. This finding mirrors the strong positive correlation in the data between gender and occupation status at the candidate level.Footnote 20 The imbalance is not a result of failed randomization, but rather an intrinsic feature of municipal employees: When a municipal employee is randomly allocated into a council, a female is more likely to get a seat in the council.

TABLE 2.

Posttreatment Council Covariate Balance

Notes: In Panel A, the treatment groups are based on all municipal employees. In Panel B, the groups are based on health care sector employees. In Panel C, the groups are based on those municipal employees who do not work in the health care sector. The statistical significance of the differences is tested using a t-test adjusted for clustering at the municipality level. ***, **, and * denote statistical significance at 1%, 5%, and 10% level, respectively.

In Panels B and C of Table 2, we divide municipal employees into two categories: those who work in the health care sector and those who work in the remaining (non–health care) sectors.Footnote 21 The division allows us to analyze whether the positive correlation between municipal employment status and gender is driven by the health care sector and, in particular, by nursing being a female-dominated occupation. Panel B and C of Table 2 suggest that, indeed, the gender imbalance is related to the health care sector. We return to the importance of gender for our findings below.

RESULTS

Treatment Effect on Total Expenditures

We start by analyzing the effect of the share of municipal employees in the council on the (log) per capita total expenditures of the local government, measured as the average over the 4-year council term.

Preliminary regression results: To have a point of comparison, we report naïve OLS results with different sets of controls (Panel A of Table 3) along with the IV results (Panel B of Table 3) and the reduced form of IV (Panel C of Table 3), using the narrowest possible bandwidth of ε = 0. The OLS estimations obviously do not correct for the potential endogeneity of the seat share of the municipal employees, while the latter two ought to do that. The difference between the four columns of each panel is that they include successively more controls. We use lags (means over the t–1 election term) of the control variables to avoid the possible problem of introducing bad controls (i.e., alternative outcomes) in the models.

TABLE 3.

Results for Total Expenditures: OLS and IV Analysis with ε = 0

Notes: The unit of observation is a municipality m in election period t. The dependent variable is the logarithm of the mean of per capita total expenditures over the council term. Standard errors are clustered at the municipality level and reported in brackets. Party controls include parties’ lagged seat shares. Municipality controls include lagged population, squared population and shares of young and old citizens. Vote share includes a second order polynomial of the municipal employees’ vote share. ***, **, and * denote 1%, 5%, and 10% statistical significance levels, respectively.

As the first three columns of Panel A of Table 3 show, the OLS estimations suggest a positive and statistically significant association between the political representation of public employees and total expenditures. This association vanishes completely once we include a second-order polynomial of the vote share of municipal employees (see column 4). This is not unexpected, because the municipal employees’ vote and seat shares are highly correlated. While insignificant, the point estimates from the IV (Panel B) and the reduced form of IV (Panel C) estimations provide us with three important empirical insights: First, the IV point estimates are positive and larger in magnitude than the OLS estimates. Second, if our instrument is as-good-as random, the only implication of having more control variables in the model ought to be that they reduce residual variance. The results reported in Panel B and C bear this out: The magnitudes of the IV estimates do not change (much) when the municipal employee vote share is controlled for. This finding indicates that unlike OLS, the IV estimates are not confounded by voter preferences. Moreover, the standard errors of the estimates tend to get smaller when more controls are added. Third, the results reported in Panel B and C suggest that the limited amount of variation in the instrument is a potential problem with using the narrowest possible bandwidth (ε = 0). If so, the first-stage regressions may suffer from low power, especially when fewer controls are included. This is indeed what we observe: The first stage F-tests become larger when we control for the municipal employee vote share (see column (8)).

Main regression results: To explore whether we can estimate the effect of political representation of municipal employees on municipal expenditures more precisely, we use the wider bandwidth of ε = 0.4. The wider bandwidth allows us to bring in more variation from the close elections. These results are reported in Table 4, where Panel A reports our IV estimates and Panel B our reduced form estimates. The estimations that rely on the wider bandwidths can be taken to be more reliable if they produce a point estimate that is similar in magnitude to that produced by the narrowest bandwidth and if the effect can be estimated with greater precision (smaller standard error).

TABLE 4.

Results for Total Expenditures: IV Analysis with ε = 0.4

Notes: The unit of observation is a municipality m in election period t. The dependent variable is the logarithm of the mean of per capita total expenditures over the council term. Standard errors are clustered at the municipality level and reported in brackets. Party controls include parties’ lagged seat shares. Municipality controls include lagged population, squared population and shares of young and old citizens. Vote share includes a second order polynomial of the municipal employees’ vote share. ***, **, and * denote 1%, 5%, and 10% statistical significance levels, respectively.

Starting from the IV estimates in Panel A of Table 4, we find across all specifications a statistically significant treatment effect of 0.0034–0.0046 on municipal spending from having a larger share of municipal employees in the council. The reduced form results in Panel B echo the IV findings: They yield treatment effect estimates that are statistically significant and very similar to those obtained with IV, but somewhat smaller in magnitude. It is especially noteworthy that both estimators deliver point estimates that are very close to those we obtained using the narrowest possible bandwidth (ε = 0.0; see Panel B in Table 3). The fact that the reduced form estimates are a little smaller in absolute value than the IV estimates suggests that the first stage coefficient of the instrument is close to, but somewhat smaller than, one (as it often is; see Online Appendix B). It is comforting to report that we cannot reject the null hypothesis that the 1st stage coefficient of the instrument is unity.

The point estimates of Table 4 show that, consistent with our Hypothesis 1, the council seat share of municipal employees causally increases municipal spending. Increasing municipal employees’ seat share by 1 percentage point increases per capita total expenditures annually by circa 0.4% over one election term. As one seat is on average 3 percentage points of the total number of seats, the overall average effect of an increase of one seat is roughly (at least) 1%. Because the average annual municipal spending is around 5,600 Euros per capita, this effect translates into around 60 euros per capita. The effect is surprisingly large taking three features into account: First, we are identifying the effect at a potentially unimportant of margin of allocating the last seats to the council. Second, the nonelected marginal candidates, to which the elected ones are compared, are often vice-councilors. Our estimate is conservative, because vice-councilors get to attend council meetings if the councilor is absent, may get a council seat if elected councilors step down during the term, and are sometimes given positions in the municipal subcommittees. Third, there are nonnegligible institutional restrictions on the political representation of the municipal employees. The effect could have been larger, had there not been no such rules in place. While a detailed comparison is not straightforward, the magnitude of the effect is nevertheless quite comparable to those reported in the prior papers using data from similar countries and identification based on marginal seats (Freier and Odendahl Reference Freier and Odendahl2015; Fiva and Halse Reference Fiva and Halse2016; Fiva, Folke, and Sørensen Reference Fiva, Folke and Sørensen2017).

Robustness checks: We have explored the robustness of our main findings and their internal and external validity in a number of ways (see Online Appendix B).

First, electing public employees has the documented spending effect irrespectively of their attributes (e.g., gender, age, education). However, a consequence of electing a public employee is greater female participation in the council. This increase may in itself increase public spending (Chattopadhyay and Duflo Reference Chattopadhyay and Duflo2004; Clots-Figueras Reference Clots-Figueras2011; Svaleryd Reference Svaleryd2009; Braendle and Colombier Reference Braendle and Colombier2016). We therefore explore whether the council seat share of municipal employees increases municipal spending also when the gender composition of the marginal seats, i.e., the seat share of females, is accounted for. We instrument this potentially endogenous share by the share of females who were randomly elected in the close contests. This instrument is calculated using the same procedure that produced the instrument for the share of municipal employees. When female seat share is included in the model, we get at the effect of electing a municipal employee while keeping the gender composition of the council constant. Adding the seat share of females to the estimations of Table 4 has only a minor impact on our results, suggesting that there is a municipal employee effect on spending independent of gender; for example, the IV estimates are still statistically significant and vary from 0.0032 to 0.0035.

Second, the choice of bandwidth ε = 0.4 for our main analysis is somewhat ad hoc. The point estimates of the municipal employee effect are quite stable across a wide range of bandwidths and statistically significant for the bandwidths from ε = 0.24 upwards.

Third, we have analyzed the expenditure effects separately for each year instead of the mean over the whole council term. These by-year estimates are all significant, similar in magnitude to what we reported earlier and stable over the council term (no within-term trend). We have also run by-year placebo regressions (4 years prior to the council term of interest), and the estimates are insignificant as they should. However, the expenditure effect is somewhat persistent, as it is different from zero and significant for two years after the council term ends. The effect becomes insignificant by the third postterm year.

Fourth, we have also constructed the instrument using placebo thresholds of getting elected within the party lists. Reassuringly, neither the first nor the second stage IV estimates are significantly different from zero when we use the placebo thresholds.

Finally, our main results are based on the entire sample of 1,544 municipality-election period observations, even though the instrument can be different from zero only within the chosen bandwidth. This choice may lead to a selection bias if the municipalities implicitly selected by the bandwidth are different from the rest of the municipalities. For example, in the close sample defined by the choice of bandwidth of ε = 0, the covariates balance perfectly. On the other hand, for ε = 0.4, the close sample is different from the other municipalities, because larger municipalities are selected into the close sample.Footnote 22 However, it is unlikely that this selection compromises the validity of our findings, because our point estimates are robust to changing the bandwidth. We have also replicated the results of Table 4 using only those observations in which close elections take place: The point estimates remain unchanged, standard errors are slightly larger, but the estimates are mostly statistically significant nonetheless.

Mechanisms at Work

Our results show that when elected, municipal employees influence per capita local public spending (Hypothesis 1). We now turn to our Hypothesis 2a and test whether the influence of an additional municipal employee depends on council size. In the two leftmost columns of Table 5, we present the results for which we have divided the sample into two based on the median council size (27 councilors). Consistent with Hypothesis 2a, we find that the effect is larger and significant in the municipalities with a smaller council size. The difference between the smaller and larger councils is also statistically significant.Footnote 23

TABLE 5.

Heterogeneity in the Total Expenditures Effect by Council and Party Size

Notes: The unit of observation is a municipality m in election period t. The dependent variable is the logarithm of the mean of per capita total expenditures over the council term. Standard errors are clustered at the municipality level and reported in brackets. Party controls include parties’ lagged seat shares. Municipality controls include lagged population, squared population and shares of young and old citizens. ***, **, and * denote 1%, 5%, and 10% statistical significance levels, respectively.

We then test Hypothesis 2b and specifically the possibility that the municipal employees have a disproportionate effect within and via their party. Table 5 reports results for the largest and second largest parties. Unlike in our earlier regressions, here the endogenous explanatory variable and the instrument refer to the shares of municipal employees within the respective party, not in the entire council. We find a significant effect for the largest party, whereas the estimates are smaller and insignificant for the second largest party (see also Online Appendix C). However, the effects are not statistically significantly different from each other. Thus, while not conclusive, the evidence is consistent with Hypothesis 2b.Footnote 24 This result suggests that municipal employees may be a nonpartisan interest group that is able to influence decision making especially within the party. If the party is large, they have a disproportionate effect on policy.Footnote 25

Evidence on Rent-Seeking

To shed light on whether the effect of public employee representation on public spending reflects rent-seeking, or whether it is more consistent with efficient provision of public services, we test Hypothesis 3a and Hypothesis 3b.

To start with, we explore whether the link between municipal employees in the council and municipal spending is occupation specific. It is plausible that municipal employees have more information on their own employment sector. However, there is no reason why, for example, a teacher would have better information about the appropriate level of health care spending than an otherwise similar councilor from the private sector. In columns (1) and (3) of Table 6, the outcome variable is municipal expenditures that are not related to health care, whereas in columns (2) and (4) of the panels the outcome variable is health care expenditures.Footnote 26 In these models, the interpretation for the coefficient for municipal health care employees is that it mirrors the effect of increasing their seat share relative to any nonmunicipal employee occupation.Footnote 27

TABLE 6.

Results According to Occupation and Spending Category

Notes: The unit of observation is a municipality m in election period t. The dependent variables are the logarithms of the means of per capita expenditures over the council term. Standard errors are clustered at the municipality level and reported in brackets. Party controls include parties’ lagged seat shares. Municipality controls include lagged population, squared population and shares of young and old citizens. ***, **, and * denote 1%, 5%, and 10% statistical significance levels, respectively.

The results suggest that health care municipal employees increase health care expenditures, but non–health care municipal employees have no effect on them. Similarly, health care employees do not affect non–health care expenditures, but municipal employees in the sectors other than health increase the other (nonhealth) municipal expenditures. Thus, consistent with Hypothesis 3a, spending increases seem to be confined to the sectors that have, by chance, more representation through municipal employees in the municipal council.Footnote 28

While not entirely conclusive, the evidence is consistent with the information advantage of municipal employees (Niskanen Reference Niskanen1971; Romer and Rosenthal Reference Romer and Rosenthal1979). Of course, one has to bear in mind that the context of those models is somewhat different from ours as these models focus on how bureaucrats can convince politicians to overspend. Moreover, the evidence speaks—at least mildly—for inefficient spending, because our as-good-as-random instrument ensures that variation in the needs of citizens is not driving the results. This raises the obvious question of why municipal employees’ information advantage leads to increased—and not to decreased—spending and only in their own sector of employment. These results also speak against the interpretation that municipal employees increase spending because they generally prefer a larger public sector (Knutsen Reference Knutsen2005; Jensen, Sum, and Flynn Reference Jensen, Sum and Flynn2009; Rattsø and Sørensen Reference Rattsø and Sørensen2016). Finally, we would like to point out that intraparty bargaining—for which we already provided support earlier—is an example of an indirect mechanism that could generate the observed sector specific effects: Given that councilors with municipal employment cannot be members of the subcommittee of their own sector, they have to influence sector-specific spending indirectly.

The results reported so far are consistent with our Hypothesis 3a, i.e., with municipal employees having an information advantage over politicians and being able to convince politicians to spend more on public services. To test Hypothesis 3b, which postulates that the extra spending is excessive and related to rent-seeking, we analyze whether municipal employees enjoy larger returns to office in terms of receiving larger salary increases and/or facing smaller unemployment risk, and whether they enjoy from a larger incumbency advantage than the other candidates. When we use candidate-level data (either lottery outcomes that make the election status truly random or RDD), we find no systematic evidence that that the municipal employees would get higher salaries, be more likely to be employed subsequently, or that they would be more likely to get re-elected or get more votes (in the next election at t + 1) than the other candidates due to getting elected at time t (see the Online Appendix E for details of these results).Footnote 29

These null results do not support Hypothesis 3b. Using auxiliary survey data from the Finnish Broadcasting Company (YLE), we have, however, confirmed that municipal employees who run for a council differ from the other candidates in two intriguing ways: First, they oppose more strongly firing of municipal employees in connection with municipal mergers. Second, they oppose more strongly restrictions on nomination of municipal employees in municipal boards (see Online Appendix E for details and further media references). One could argue that these stated views, as well as the concerns expressed in the Finnish media, are harder to reconcile with pro-social behavior than with rent-seeking.

CONCLUSIONS

We have produced three novel findings in this paper. First, the political representation of municipal employees has a positive causal effect on overall local public spending. Second, the effect is sector specific: Having more health care sector employees in the council increases health care spending and having more non–health sector employees increases non–health care spending, but there are no significant cross-sector effects. Third, the effect appears to be related to the interest group influencing the policy from within the parties.

Our findings hold two lessons for contemporary research in political economics and political science. The first is that politicians’ identities matter in local political decision making characterized by proportional representation and open-list D'Hondt election rule. The citizen-candidate model (Osborne and Slivinski Reference Osborne1996; Besley and Coate Reference Besley and Coate1997) is therefore more in line with our evidence than the median voter model or Tiebout (Reference Tiebout1956) competition. The second lesson is that the marginally elected candidates are able to influence local policy. This influence may explain why in the very same Finnish elections that we have studied in this article, a greater likelihood of being the pivotal voter increases turnout (Lyytikäinen and Tukiainen Reference Lyytikäinen and Tukiainen2016).

It is important to interpret our results in the context to which they apply. Our findings refer to a country that has a large public sector and that has traditionally given the local municipalities a major role in the allocation of public resources and production of public services. While we do not find systematic evidence of rent-seeking, our results show that the Finnish municipal councilors employed by the public sector want—by revealed preference—to increase public expenditures in a country that in 2014 had the highest public sector ratio to GDP and whose local governments were among the most indebted among all OECD countries. This is puzzling because our as-good-as-random design guarantees that the citizens’ needs are identical on average in the analyzed municipalities across all sectors. One can therefore raise the question why, in this context, would informed and benevolent municipal employee councilors increase rather than decrease their own sector's public spending? Moreover, can it be desirable—in this context and more generally—that municipal spending is strongly affected by one particular interest group?

Making precise statements about the external validity of a close elections analysis is challenging. On one hand, there are about 40 countries in the world using an open-list PR, similar in spirit to what we have studied. Moreover, the Finnish rules governing the political representation of municipal employees have the same broad goal as many other countries’ corresponding rules: They have been written in order to prevent public employees from having undue influence on political decision making. It thus seems possible that our results generalize at least to countries with a similar political system at the local level. On the other hand, details of political processes tend to matter: We should not extrapolate too much, as there is quite a bit cross-country variation in both the precise functioning of the open-list PR systems as well as in the design of ineligibility and incompatibility rules (Braendle and Stutzer Reference Braendle and Stutzer2016).

Rather than offering detailed policy recommendations, we conclude with a call for more research. There are three reasons to this call. First, while our findings support the argument that some regulation of public employees’ political involvement is warranted, they do not provide guidance on the optimal design of ineligibility and incompatibility rules. We can only conjecture how large the estimated effect would have been, had there not been any restrictions on political participation of public sector employees in Finland. Moreover, we would need to understand better what the interests of other groups are to optimally design policy. Second, when, how, and why ineligibility and incompatibility rules prevent public employees from having undue influence on political decision making in general and spending in particular is likely to be context dependent. This calls for replicating our analysis in other institutional contexts. Finally, the empirical procedure presented here can be applied to a wider range of electoral systems than just the open-list PR. For example, one can use similar aggregation of close races to look at effects of council composition in plurality systems, where the council is composed of politicians elected from many (single or multimember) districts. Subsequent work can thus make use of our procedure to provide more analyses of the desirability to restrict public employees’ political participation in different environments.

SUPPLEMENTARY MATERIAL

To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055417000284.

Replication Files can be found on Dataverse at http://doi.org/10.7910/DVN/V7SQRW.

Footnotes

We thank Manuel Bagues, Thushyanthan Baskaran, Matz Dahlberg, Dominik Hangartner, Robert Inman, Tatiana Komarova, Benjamin Lauderdale, Eva Mörk, Oskar Nordström Skans, Torsten Persson, Heikki Pursiainen, Marko Terviö, and Guy Whitten for helpful comments. Seminar audiences at AEA San Francisco, BI Oslo, FEEAM Pori, EPSA Vienna, HECER Helsinki, IIES, IIPF Dublin, Uppsala and VATT Helsinki provided great feedback. We are also grateful to three anonymous reviewers and the editor, Ingo Rohlfing, for their thoughtful comments on the manuscript. Financial support from the Emil Aaltonen Foundation is gratefully acknowledged.

1 For example, Braendle and Stutzer (Reference Braendle and Stutzer2016) report that in their sample of 76 countries the average fraction of politicians in national parliaments with a public sector background is 31.3%.

2 Prominent examples include the Hatch Act of 1939 in the US and the House of Commons Disqualification Act of 1975 in the UK. The Local Government Act of 1972 and the Local Government and Housing Act 1989 include similar restrictions for local government employees in the UK. See Braendle and Stutzer (Reference Braendle and Stutzer2016) for examples from other countries.

3 Braendle and Stutzer (Reference Braendle and Stutzer2010, Reference Braendle and Stutzer2016) show using German and cross-country data, respectively, that stricter ineligibility and incompatibility rules decrease the share of public servants in parliaments. Rosenson (Reference Rosenson2006) finds a connection between various ethics laws and representation of occupations. Braendle (Reference Braendle2016) reviews the effects of institutions and eligibility rules on political selection.

4 Prior analyses closest to ours are Braendle and Stutzer (Reference Braendle and Stutzer2013, Reference Braendle and Stutzer2016), but neither focuses on estimating causal effects. For example, Braendle and Stutzer (Reference Braendle and Stutzer2016) find using cross-country data a positive association between government size and the share of public servants in parliament.

5 Source: Statistics Finland Labor Force Survey 2015.

6 In Finland, municipal employees are eligible to run for a council seat and can hold on to their municipal job if elected. There are, however, other restrictions (see section Institutional Setting and Data).

7 For example, the Finnish National Broadcasting company YLE expressed at the time of the 2012 municipal council elections the concern that municipal employees can decide on their own jobs in municipal councils.

8 For studies on party effects in the U.S. context, see Ferreira and Gyourko (Reference Ferreira and Gyourko2009), Gerber and Hopkins (Reference Gerber and Hopkins2011) and de Benedictis-Kessner and Warshaw (Reference de Benedictis-Kessner and Warshaw2016). The effects of the political representation of other non-partisan interests groups, such as women, minorities and occupation groups, on policy outcomes are studied by, for instance, Pande (Reference Pande2003), Chattopadhyay and Duflo (Reference Chattopadhyay and Duflo2004), Gehlbach, Sonin, and Zhuravskaya (Reference Gehlbach, Sonin and Zhuravskaya2010), Ferreira and Gyourko (Reference Ferreira and Gyourko2014), Matter and Stutzer (Reference Matter and Stutzer2015) and Bagues and Campa (Reference Bagues and Campa2017). Gagliarducci and Nannicini (Reference Gagliarducci and Nannicini2013) and Freier and Thomasius (Reference Freier and Thomasius2016) study the effects of politicians’ qualifications on fiscal outcomes.

9 Data limitations prevent us from analyzing occupation groups in more detail.

10 We are unable to find systematic evidence for the extra spending being related to rents that the politicians employed by the municipalities get from holding the office (through better employment opportunities, or greater wages; see Dahlberg and Mörk Reference Dahlberg and Mörk2006 and Brueckner and Neumark Reference Brueckner and Neumark2014). Neither do we find evidence that the increased spending reflects pro-social behavior or competence of public sector employees (Best and Cotta Reference Best and Cotta2000; Francois Reference Francois2000; Besley and Ghatak Reference Besley and Ghatak2003, Reference Besley and Ghatak2006).

11 The Finnish law dictates that council size is a step function of population: 13, 15, or 17 for municipal population of 2,000 or less; 21 for 2,001–4,000; 27 for 4,001–8,000; 35 for 8,001–15,000; 43 for 15,001–30,000; 51 for 30,001–60,000; 59 for 60,001–120,000; 67 for 120,001–250,000; 75 for 250,001–400,000; and 85 for over 400,000.

12 Most of Finnish local politicians have a normal day job. The task of being a municipal councilor typically takes a few hours a week and the monetary compensation involved is not nearly enough to live on. The same applies, by and large, e.g., to the UK (Local Government Association 2012).

13 Using similar Finnish close elections data as we do, Kotakorpi, Poutvaara, and Terviö (Reference Kotakorpi, Poutvaara and Terviö2017) study returns to office and Hyytinen et al. (Reference Hyytinen, Meriläinen, Saarimaa, Toivanen and Tukiainen2017) study incumbency advantage and the performance of close elections RDD. Unlike these articles, we are interested in municipal level outcomes.

14 Because vipmt cannot be defined for party lists where none of the candidates or all of the candidates get elected, approximately 4,800 candidate-election observations are left out.

15 In an urn problem, the expected value is n(K/N) with and without replacement, where n is the number of available close seats, K the number of close municipal employees and N the number of close candidates.

16 Simply “adding up” candidate level realized outcomes would not be appropriate. To see why, consider three municipal employees who are close and compete for one seat. In this case, a municipal employee is always elected.

17 We face the standard trade-off that smaller bandwidths lead to less-precise estimates: The narrower the bandwidth, the less there is variation in Tmt , but the more plausible the assumption of “as-good-as-random assignment.” The results for the narrowest possible bandwidth (ε = 0) and party affiliation balance tests echo the results reported in Table 1 (see Online Appendix B).

18 The number of observations varies because some of the pre-treatment variables for the 1996 election term are not available.

19 The posttreatment seat shares are by definition balanced, because our instrument is based on within party closed contests (see Online Appendix B).

20 Municipal employees are more often female and have higher socioeconomic status than the candidates that have other employment status (see Table A1 in Online Appendix A).

21 We do not disaggregate the latter group into more specific categories, because the data get sparse: First, candidates at finer level occupations are involved in close elections infrequently; second, detailed sector specific spending data are not always available; and third, most occupational groups are small overall (e.g., even education employees have only 3.5% seat share on average).

22 The reason for this is that we define the bandwidth within parties in vote shares. This means that even the bandwidth of 0.4 (4 votes out of 1,000) is very narrow. For example, a party list needs get more than 500 votes for a candidate with a two-vote distance to the threshold to be within the bandwidth. Larger municipalities have such narrow bandwidths more often.

23 The effect for the larger councils is not significantly different from zero. This does not imply that municipal employees could not affect spending in some types of larger councils, but studying such heterogeneity in detail would call for larger datasets and an alternative identification strategy.

24 We should note that the Centre Party is most often the largest party in the Finnish municipalities, due to its considerable support in the smaller rural municipalities (which constitute the bulk of municipalities). Therefore, the effect captured in Table 5 may be a Centre Party phenomenon rather than a more general party size effect.

25 We have also considered a number of other explanations. First, the marginally elected municipal employees do not lead to municipal employees having a majority in the council or to their party becoming dominated by municipal employees: such instances are very rare in the data. Second, the effect is not larger in the municipalities where the marginally elected councilor was the only elected municipal employee from his/her party (not reported). Moreover, instances where there would be only one municipal employee in the entire council are very rare in the data. Finally, the increase in the municipal employee representation apparently does not increase the probability that a political leader (chairman of the council board or chairman of the council) would be a municipal employee.

26 When there are more than one endogenous variable, we report the Angrist-Pischke first-stage F-statistics of individual endogenous regressor produced by the ivreg2 STATA command.

27 The results for pre-treatment covariate balance tests and the first stage estimations of the IV suggest that the instrument works as expected (Online Appendix D).

28 The effects are not statistically significantly different from each other. These results are similar also if we run the analysis by party size or if we add the seat share of females to the models (see Online Appendix D). The results for the non–health care expenditures are also robust to using other bandwidth choices. However, the effect of the seat share of municipal health care employees on health spending is less robust in this regard.

29 We have also analyzed whether the political representation of municipal employees shows up in house prices, because high levels of government rent extraction might be capitalized in them (Gyourko and Tracy Reference Gyourko and Tracy1991). Using municipal-level data on real estate transactions, we find no effect on house prices.

References

REFERENCES

Alesina, Alberto, Baqir, Reza, and Easterly, William. 2000. “Redistributive Public Employment.” Journal of Urban Economics 48, 219–41.CrossRefGoogle Scholar
Anzia, Sarah F., and Moe, Terry M.. 2015. “Public Sector Unions and the Costs of Government.” The Journal of Politics 77, 114–27.CrossRefGoogle Scholar
Bagues, Manuel, and Campa, Pamela. 2017. “Can Gender Quotas in Candidate Lists Empower Women? Evidence from a Regression Discontinuity Design.” Manuscript. http://www.manuelbagues.com/gender%20quotas%20-%20bagues%20and%20campa%20-%20wp.pdf Google Scholar
Besley, Tim, and Coate, Stephen. 1997. “An Economic Model of Representative Democracy.” Quarterly Journal of Economics 112, 85–114.CrossRefGoogle Scholar
Besley, Tim, and Ghatak, Maitreesh. 2003. “Incentives, Choice, and Accountability in the Provision of Public Services.” Oxford Review of Economic Policy 19 (2), 235–49.CrossRefGoogle Scholar
Besley, Tim, and Ghatak, Maitreesh. 2006. “Competition and Incentives with Motivated Agents.” American Economic Review 95 (3), 616–36.CrossRefGoogle Scholar
Best, Heinrich, and Cotta, Maurizio. 2000. Parliamentary Representatives in Europe, 1848–2000. Legislative Recruitment and Careers in Eleven European Countries. Oxford: Oxford University Press.Google Scholar
Bhatti, Yosef, and Hansen, Kasper M.. 2012. “Public Employees Lining Up at the Polls—The Conditional Effect of Living and Working in the Same Municipality.” Public Choice 156 (3–4), 611–29.CrossRefGoogle Scholar
Blais, André, Blake, Donald, and Dion, Stéphane. 1990. “The Public/ Private Cleavage in North America: The Political Behavior and Attitudes of Public Sector Employees.” Comparative Political Studies 23, 381–403.CrossRefGoogle Scholar
Braendle, Thomas. 2016. “Do Institutions Affect Citizens' Selection into Politics.” Journal of Economic Surveys 30 (2), 205–27.CrossRefGoogle Scholar
Braendle, Thomas, and Colombier, Carsten. 2016. “What Drives Public Health Care Expenditure Growth? Evidence from Swiss Cantons, 1970–2012.” Health Policy 120 (9), 1051–60.CrossRefGoogle ScholarPubMed
Braendle, Thomas, and Stutzer, Alois. 2010. “Public Servants in Parliament: Theory and Evidence on Its Determinants in Germany.” Public Choice 145, 223–52.CrossRefGoogle Scholar
Braendle, Thomas, and Stutzer, Alois. 2013. “Political Selection of Public Servants and Parliamentary Oversight.” Economics of Governance 14, 45–76.CrossRefGoogle Scholar
Braendle, Thomas, and Stutzer, Alois. 2016. “Selection of Public Servants into Politics.” Journal of Comparative Economics 44 (3), 696–719.CrossRefGoogle Scholar
Brueckner, Jan K., and Neumark, David. 2014. “Beaches, Sunshine, and Public-Sector Pay: Theory and Evidence on Amenities and Rent Extraction by Government Workers.” American Economic Journal: Economic Policy 6, 198–230.Google Scholar
Chattopadhyay, Raghabendra, and Duflo, Esther. 2004. “Women as Policy Makers: Evidence from a Randomized Policy Experiment in India.” Econometrica 72 (5), 1409–43.CrossRefGoogle Scholar
Clots-Figueras, Irma. 2011. “Women in politics: Evidence from the Indian States.” Journal of Public Economics 95 (7–8), 664–90.CrossRefGoogle Scholar
Clots-Figueras, Irma. 2012. “Are Female Leaders Good for Education? Evidence from India.” American Economic Journal: Applied Economics 4 (1), 212–44.Google Scholar
Courant, Paul N., Gramlich, Edward M., and Rubinfield, Daniel L.. 1979. “Public Employee Market Power and the Level of Government Spending.” American Economic Review 69 (5), 806–17.Google Scholar
Dahlberg, Matz, and Mörk, Eva. 2006. “Public Employment and the Double Role of Bureacrats.” Public Choice 126, 387–404.CrossRefGoogle Scholar
de Benedictis-Kessner, Justin, and Warshaw, Christopher. 2016. “Mayoral Partisanship and Municipal Fiscal Policy.” Journal of Politics 78 (4), 1124–38.CrossRefGoogle Scholar
Ferreira, Fernando, and Gyourko, Joseph. 2009. “Do Political Parties Matter? Evidence from U.S. Cities.” Quarterly Journal of Economics 124 (1), 399–422.CrossRefGoogle Scholar
Ferreira, Fernando, and Gyourko, Joseph. 2014. “Does Gender Matter for Political Leadership? The Case of U.S. Mayors.” Journal of Public Economics 112, 24–39.CrossRefGoogle Scholar
Fiva, Jon H., Folke, Olle, and Sørensen, Rune J.. 2017. “The Power of Parties: Evidence from Close Municipal Elections in Norway.” Forthcoming in Scandinavian Journal of Economics.Google Scholar
Fiva, Jon H., and Halse, Askil H.. 2016. “Local Favoritism in At-Large Proportional Representation Systems.” Journal of Public Economics 143, 15–26.CrossRefGoogle Scholar
Folke, Olle. 2014. “Shades of Brown and Green: Party Effects in Proportional Election Systems.” Journal of the European Economic Association 12 (5), 1361–95.CrossRefGoogle Scholar
Francois, Patrick. 2000. “‘Public Service Motivation’ as an Argument for Government Provision.” Journal of Public Economics 78 (3), 275–99.CrossRefGoogle Scholar
Freier, Ronny, and Odendahl, Christian. 2015. “Do Parties Matter? Estimating the Effect of Political Representation in Multi-Party Systems.” European Economic Review 80, 310–28.CrossRefGoogle Scholar
Freier, Ronny, and Thomasius, Sebastian. 2016. “Voters Prefer More Qualified Mayors, But Does It Matter for Public Finances? Evidence for Germany.” International Tax and Public Finance 23 (5), 875–910.CrossRefGoogle Scholar
Gagliarducci, Stephano, and Nannicini, Tommaso. 2013. “Do Better Paid Politicians Perform Better? Disentangling Incentives from Selection.” Journal of the European Economic Association 11, 369–98.CrossRefGoogle Scholar
Garand, James C. 1988. “Explaining Government Growth in the U.S. States.” American Political Science Review 82, 837–49.CrossRefGoogle Scholar
Gehlbach, Scott, Sonin, Konstantin, and Zhuravskaya, Elena. 2010. “Businessman Candidates.” American Journal of Political Science 54 (3), 718–36.CrossRefGoogle Scholar
Gerber, Elisabeth R., and Hopkins, Daniel J.. 2011. “When Mayors Matter: Estimating the Impact of Mayoral Partisanship on City Policy.” American Journal of Political Science 55 (2), 326–39.CrossRefGoogle Scholar
Gyourko, Joseph, and Tracy, Joseph. 1991. “The Structure of Local Public Finance and the Quality of Life.” Journal of Political Economy 99 (4), 774–806.CrossRefGoogle Scholar
Hyytinen, Ari, Meriläinen, Jaakko, Saarimaa, Tuukka, Toivanen, Otto, and Tukiainen, Janne. 2017. “When Does Regression Discontinuity Design Work? Evidence from Random Election Outcomes.” Manuscript. https://dl.dropboxusercontent.com/u/40759261/IncumbencyRDD_All_webpage.pdf CrossRefGoogle Scholar
Jensen, Jason L., Sum, Paul E., and Flynn, David T.. 2009. “Political Orientation and Behavior of Public Employees: A Cross-National Comparison.” Journal of Public Administration Research and Policy 19, 709–30.Google Scholar
Knutsen, Oddbjørn. 2005. “The Impact of Sector Employment on Party Choice: A Comparative Study of Eight West European Countries.” European Journal of Political Research 44 (4), 593–621.CrossRefGoogle Scholar
Kotakorpi, Kaisa, Poutvaara, Panu, and Terviö, Marko. 2017. “Returns to Office in National and Local Politics: A Bootstrap Method and Evidence from Finland.” Forthcoming in Journal of Law, Economics, and Organization.CrossRefGoogle Scholar
Laver, Michael, and Shepsle, Kenneth A.. 1990. “Government Coalitions and Intraparty Politics.” British Journal of Political Science 20 (4), 489–507.CrossRefGoogle Scholar
Lee, David S., Moretti, Enrico, and Butler, Matthew J.. 2004. “Do Voters Affect or Elect Policies? Evidence from the U.S. House.” Quarterly Journal of Economics 119 (1), 807–59.CrossRefGoogle Scholar
Local Government Association. 2012. “Stand for What You Belief In. Be a Councilor.” http://www.local.gov.uk/c/document_library/get_file?uuid=fa4de86d-1009-4b58-a9e7-3103fe3d9a36 Google Scholar
Lyytikäinen, Teemu, and Tukiainen, Janne. 2016. “Are Voters Rational?” VATT Working Papers 50. http://www.vatt.fi/file/vatt_publication_pdf/wp50.pdf Google Scholar
Matter, Ulrich, and Stutzer, Alois. 2015. “The Role of Lawyer-Legislators in Shaping the Law: Evidence from Voting on Tort Reforms.” Journal of Law & Economics 58 (2), 357–84.CrossRefGoogle Scholar
Niskanen, William A. 1971. Bureaucracy and Representative Government. Chicago: Aldine, Atherton.Google Scholar
OECD. 2015. Government at a Glance 2015. DOI: 10.1787/22214399. http://www.oecd.org/gov/govataglance.htm CrossRefGoogle Scholar
Osborne, Martin J., and Al Slivinski. 1996. “A Model of Political Competition with Citizen-Candidates.” Quarterly Journal of Economics 111, 65–96.CrossRefGoogle Scholar
Pande, Rohini. 2003. “Can Mandated Political Representation Increase Policy Influence for Disadvantaged Minorities? Theory and Evidence from India.” American Economic Review 93 (4), 1132–51.CrossRefGoogle Scholar
Rattsø, Jørn, and Sørensen, Rune J.. 2016. “Public–Private Political Cleavage: What Happens After Retirement?” Public Choice 166 (3), 315–33.CrossRefGoogle Scholar
Romer, Thomas, and Rosenthal, Howard. 1979. “Bureaucrats Versus Voters: On the Political Economy of Resource Allocation by Direct Democracy.” Quarterly Journal of Economics 93 (4), 563–87.CrossRefGoogle Scholar
Rosenson, Beth A. 2006. “The Impact of Ethics Laws on Legislative Recruitment and the Occupational Composition of State Legislatures.” Political Research Quarterly 59 (4), 619–27.CrossRefGoogle Scholar
Saarimaa, Tuukka, and Tukiainen, Janne. 2015. “Common Pool Problems in Voluntary Municipal Mergers.” European Journal of Political Economy 38, 140–52.CrossRefGoogle Scholar
Sieg, Holger, and Wang, Yu. 2013. “The Impact of Unions on Municipal Elections and Fiscal Urban Fiscal Policies.” Journal of Monetary Economics 60 (5), 554–67.CrossRefGoogle Scholar
Svaleryd, Helena. 2009. “Women's Representation and Public Spending.” European Journal of Political Economy 93, 355–72.Google Scholar
Svaleryd, Helena, and Vlachos, Jonas. 2009. “Political Rents in a Non-Corrupt Democracy” Journal of Public Economics 93 (3–4), 355–72.CrossRefGoogle Scholar
Tiebout, Charles M. 1956. “A Pure Theory of Local Expenditures.” Journal of Political Economy 64 (5), 416–24.CrossRefGoogle Scholar
Wooldridge, Jeffrey. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press.Google Scholar
Figure 0

TABLE 1.Pretreatment Covariate Balance at Municipality Level

Figure 1

TABLE 2.Posttreatment Council Covariate Balance

Figure 2

TABLE 3.Results for Total Expenditures: OLS and IV Analysis with ε = 0

Figure 3

TABLE 4.Results for Total Expenditures: IV Analysis with ε = 0.4

Figure 4

TABLE 5.Heterogeneity in the Total Expenditures Effect by Council and Party Size

Figure 5

TABLE 6.Results According to Occupation and Spending Category

Supplementary material: PDF

Hyytinen supplementary material

Appendices

Download Hyytinen supplementary material(PDF)
PDF 476.4 KB
Supplementary material: Link

Hyytinen et al. Dataset

Dataset

http://dx.doi.org/10.7910/DVN/V7SQRW
Link
Submit a response

Comments

No Comments have been published for this article.