Version for EGPA 2018, 5-7 September, Lausanne

Local Government Efficiency, a Dynamic Double Bootstrapped Analysis

Stefanie Vanneste

Stijn Goeminne

Ghent University

General Economics Department

1. Introduction

Local governments are in charge of a significant amount of public provisions like amongst others waste collection, social support, road maintenance and public education. For this reason it is necessary to evaluate how efficient the local governments are in the provision of their public services. Doing research on the efficiency of public entities allows policy makers to direct actions to the most or the less efficient municipalities. It is possible to provide more or less subsidies to the inefficient units. The second objective of this research is to understand the causes of inefficiency and to be able to improve the overall efficiency of local governments. Improved efficiency in public service production facilitates more services without increasing government expenditures (Borge, 2005). The European Commission has in a recent report emphasized the importance of efficiency. Countries with high levels of public debt need to show that they are committed to reduce these debts and show progress in the right direction. The member states need to improve the quality of their expenses by rising their expenditure efficiency (European Commission, 2014).

Flemish existing work until now, only focused on cross sectional research regarding the efficiency of local municipalities. The efficiency has more specifically only been examined in the year 1989 (De Borger and Kerstens, 1996a; De Borger et al., 1994)and the year 2000 (Ashworth et al., 2014; Geys, 2006; Geys and Moesen, 2009a; 2009b). In the international existing literature one finds next to cross sectional papers (Benito et al., 2010; Bönisch et al., 2011; Borge et al., 2008) also papers that analyze the local government efficiency with panel data (Bruns and Himmler, 2011; Doumpos and Cohen, 2014; Marques et al., 2015). However, the amount of papers doing single year analysis outweighs the panel data research.

Our paper measures and studies the technical efficiency of the Flemish local governments in the years 2005-2016. We employ the dynamic double bootstrapped DEA analysis to measure the efficiency (Du et al., 2018). We introduce political and financial variables (with some control variables) to try and explain the local government efficiency. We find that the fragmentation of the government, the involvement of the voters, the ratio of the collected taxes to the total revenues, the level of debt and the percentage of unemployment have a significant effect on the efficiency of Flemish local governments.

Our research complements other existing work in the literature. In particular it is the first research on the local government efficiency that uses the dynamic double bootstrap DEA method. We do not pool our data but use the data of all the different years separately and combine this with the double bootstrap method. Next to that we introduce 3 political variables (ideology, fragmentation and power) and we look at an array of fiscal indicators.

Our paper is organized as follows. The next section briefly reviews the literature analysis. In section 3 the Data Envelopment Analysis is explained. Section 4 gives an overview of the political system regarding local municipalities in Flanders. Section 5 describes the inputs and outputs used in this paper. Section 6 presents the explanatory variables that are used in the bootstrapped truncated regression to explain the efficiency of the Flemish municipalities. Section 7 answers for the use of the truncated regression. Section 8 presents the efficiency measurements and the results of the bootstrapped truncated regression. Finally, section 9 concludes the research.

2. Existing literature

Our literature review focusses on research that studies the global efficiency of local governments. When measuring the global efficiency one looks at the overall efficiency of the local government in the entire service provision. Research that also looks at the global efficiency are amongst others Boetti et al. (2012), Balaguer-Coll et al. (2007), Cordero et al. (2017) and Kalb et al. (2012). Studies that have researched the efficiency of the local government in providing a particular service are beyond the scope of this paper. Research focusing on a particular local service are amongst others Worthington and Dollery (2001) and Benito et al. (2014) for waste collection, Woodbury and Dollery (2004) and García‐Sánchez (2006) for water provision and Kalb (2014) for road maintenance.

From the literature analysis it is clear that the literature uses different methodological approaches to analyze the efficiency of local governments around the world. The four most commonly used methods are; data envelopment analysis (DEA), stochastic frontier analysis (SFA), full disposal hull (FDH), and the use of an efficiency index (e.g. Borge et al. (2008)). Here we are going to give a short explanation of these methods ranking them from more to less important in the existing literature. First, DEA is the most widely used method to measure the efficiency in the public context. It is a linear programming method which constructs the efficiency frontier that envelops all data points. Of the papers that use DEA as a method the large majority does an input-oriented analysis of the local governments. The choice between constant returns to scale (CRS) and variable returns to scale (VRS) is less straightforward but there is a tendency of preferring VRS to CRS in the existing literature (see Infra section 3). Second, SFA is a parametric method to analyze efficiency which uses more a priori assumptions than DEA. When using SFA it is assumed that the production possibility set and the data generating process are known a priori (Bogetoft and Otto, 2011). Third, FDH is a non-parametric technique which relaxes the convexity assumption of DEA and makes use of a mixed integer programming technique instead of linear programming (Lim et al., 2016). Finally, the efficiency index method is only used by a couple of authors. This method calculated the efficiency as a ratio of aggregate output to revenues (Borge et al., 2008).

In line with the existing literature, this research will use DEA to measure the efficiency of Flemish local governments. DEA is the better method in the public context because of a couple of reasons. First, there is the ability of DEA to better handle multiple outputs. Instead of estimating the efficiency of one output relative to another output, DEA allows for estimating the efficiency of all the output categories at once (Varabyova and Schreyögg, 2013). Next to that, DEA is a data-driven approach and does not require as many a priory assumptions as SFA1(Bogetoft and Otto, 2011).

3. DEA

In the data envelopment analysis a best practice production frontier is constructed using a linear programming method which envelops all the observations, it is a data-driven approach for the

1 Assumptions about the specification of production function, the distribution of the noise term and the distribution of the inefficiency term.

evaluation of the performance of different units (Bogetoft and Otto, 2011). A unit is fully efficient if there is no other unit that has a lower input use and the same level of output or more output with the same level of inputs. The most efficient unit is the benchmark to analyze the relative efficiency of the other units (Behr, 2015). This analysis can more easily handle multiple outputs than the SFA analysis and in addition it requires almost no a priori assumptions about the data (Cooper et al., 2011). An inefficient unit can reduce the inefficiency and become efficient by projecting on the frontier. This can be reducing the inputs (input-oriented) or increasing the outputs (output-oriented).

Computing the technical efficiency is a linear programming problem, more specifically a minimization problem2. One minimizes the technical efficiency subject to a number of constraints3.

Minθ

s . t .− y ki+Y λ i≥ 0 (1)

θ x ji−X λ i≥ 0 (2)

λ i≥ 0 (3)4

This model is known as the Charnes, Cooper and Rhodes (CCR) model (1978) or the constant returns to scale model. The CCR model gives an efficiency frontier as in figure 1.

Figure 1: CCR frontier

Banker, Charnes and Cooper (1984) extended the CCR model to the variable returns to scale case by adding one more constraint.

2 It is a minimization problem because one wants to downscale the inputs by minimizing the efficiency.3 An output constraint, an input constrained and the constraint concerning the weights of the units.4 Where x is the input and y is the output for each observation. X and Y are the matrices of all inputs and outputs respectively with k the number of outputs, j the number of inputs and i the number of units.

Minθ

s . t .− y ki+Y λ i≥ 0 (1)

θ x ji−X λ i≥ 0 (2)

λ i≥ 0 (3)

λ ' 1=1 (4)5

The BCC model gives an efficiency frontier that looks like the one in figure 2.Figure 2: BCC frontier

The efficiency scores in these two different model (CRS and VRS) can be calculated in two different ways. One possible way of calculating it is using the input-oriented approach. The input-oriented approach is justified when the units that are being studied have more control over the inputs that are being used than over the outputs that are being produced. Another possible way is the output-oriented approach. This approach is justified when the studied units have a larger control over the outputs than over the inputs. In this research, as in most of the papers investigating the efficiency in the public sector, we choose to implement the input-oriented approach. The choice for an input oriented approach is coherent with the nature of public organizations. As public organizations have more control over the inputs they use than over the outputs they have to produce, their objective is more likely to be based on input minimizations rather than output maximization (Doumpos and Cohen, 2014; Worthington and Dollery, 2001).

4. The political system in Flanders

In this section we elaborate upon the political system in Flanders. 308 of the 589 Belgian municipalities are under the supervision of the Flemish (regional) government (De Borger et al., 1994). There are two different types of bodies in the Flemish municipalities, the local council (the legislative body) and the College of Mayor and Aldermen (the executive body). The members of

5 Where x is the input and y is the output for each observation. X and Y are the matrices of all inputs and outputs respectively with k the number of outputs, j the number of inputs and i the number of units.

the local council are elected every six years on the second Sunday of October during the municipal elections (Warnez, 2016). The last municipal elections occurred on the 14th of October 2012. The total amount of members in the local council ranges from 7 to 55 members, depending on the population size of the local government. Seats in the local council are divided according to a proportional representation system. Once the majority is formed, they must decide on the composition of the College of Mayor and Aldermen. Depending on the number of inhabitants, the College consists of 2 to 10 Aldermen in addition to the Mayor and the chairman of the Public Centre of Social Welfare (PCSW). In theory, the local council is the organ which has the most decision power, it is in charge of the chosen policy and can impose specific rules to implement the policy. The theoretical competences of the College of Mayor and Aldermen are proposing specific strategies and implementing them after approval by the local council. In reality however, the College has more power than the local council. The local council that has to approve of the policy proposals made by the college are in part composed of the members of the College.

5. Inputs and outputs

Generally, when talking about efficiency one looks at how inputs are translated into outputs. In the manufacturing industry it is quite straightforward to determine the inputs and the outputs of the unit because there is actual output that is being produced with actual input. In a public context this means that the key in- and outputs have to be determined. In the local governments’ context we have to look at the services the municipalities have to provide even when they do not seem like outputs.

The choice of the input and the output variable can be very arbitrary that is why we take a look at the services the local governments have to provide and the resources they can address for this purpose. The inputs and the outputs should underline the commitments and aims the local governments have (Cullinane et al., 2006). As the outputs in the public context are not always straightforward to measure and data is often scarce it is sometimes necessary to use proxies for these outputs in the analysis (Narbón‐Perpiñá and De Witte, 2018).

5.1. Input

In line with the existing literature we use the current expenditures of the local government as the input variable in the efficiency analysis (Balaguer-Coll et al. (2010), Geys et al. (2010), Lampe et al.(2015)). The expenditures are the resources the local governments address to provide in their local service provision. The current expenditures are calculated based on four financial categories: staff expenditures, production costs, transfers and capital payments on loans.

5.2. Outputs

The outputs of the local governments are the services that are being provided with the current expenditures (Geys and Moesen, 2009b). The outputs can thus be found by studying the list of services the local governments are in charge of providing (Cullinane et al., 2006). Although the Flemish municipalities are the most decentralized level of government in Flanders, they have a great deal of autonomy. They are free to adopt their own policies and to collect revenues (Ashworth et al., 2006). There are three categories of government powers the local governments

in Flanders are in charge of. The first category are the local governmental powers. The local governments are in charge of the public primary and kindergarten schools and of the public childcare. As D'Inverno et al. (2018), Kalb (2010) and Revelli (2010) we use the number of students in kindergarten and primary education as percentage of the total population. Kindergarten and primary education places are used as a proxy for the provision of educational services from the local governments (De Sousa and Stošić, 2005). The number of childcare places as a percentage of the total population are also used as a third output. The second category the local governments in Flanders are in charge of are the tasks the federal government imposes on the local governments. These tasks consist of the administration activities. Following Marques et al. (2015) we include the number of building applications per capita that are being processed by the administration department of the municipality. This can be seen as a proxy for the administrative tasks the local government carries out (Narbón‐Perpiñá and De Witte, 2018). The third category are the intergovernmental collaborations. The municipalities can work together to provide in some public services like amongst others garbage collection and social assistance. In line with Barone and Mocetti (2011), De Sousa and Stošić (2005) and Šťastná and Gregor (2015) we use the waste collection per capita in the municipality as an output variable. This variable is measured as the amount of garbage (in kg) the municipality collects per capita. Like Coffé and Geys (2005), Eeckaut et al. (1993) and De Sousa and Stošić (2005) we use the number of people who get support from the government as a last output variable More specifically this support includes the number of people entitled to a subsistence minimum, employment measures or a reinstallation grant. This variable is expressed in percentage of the population.

6. Explanatory variables

We will try to explain the efficiency scores of the Flemish local governments by a number of political, financial and control variables. The first political characteristic we introduce is the government fragmentation. There are three possible reasons the fragmentation of the government could have an impact on the efficiency of the government. First, tax money is paid by the entire adult population, while the expenditures are often targeted at a specific group (Weingast et al., 1981). Therefore, politicians might compete with each other to target these benefits to their electorate (Eslava, 2011). This could result in a more inefficient allocation of the resources. Second, if the coalition is more fragmented, different parties are less sure about being in the government in the next legislature. To make governing in the next legislature more difficult for their successors with different political preferences, parties ensure that there are high levels of debt (Alesina and Tabellini, 1990; Persson and Svensson, 1989). This could mean that they do less well considered expenditures leading to more inefficiency. Third it is harder to reach a decision when a lot of parties have to agree upon the decision, this may have a detrimental effect on the way a government manages its budget and can lead to more inefficiency.

The second political variable we introduce is the ideology of the local government. The ideology can have a potential impact on the efficiency of the local government due to a number of reasons. First of all a left wing majority is stated to have a stronger bond with the less wealthy part of the population, they are more interested in redistributive spending and higher levels of public goods than right-wing parties (Hibbs, 1992). Borge et al. (2008) claim that it is harder for leftist governments to impose budget constraint due to the larger public sector. Next to that left wing

governments have strong ties with the public sector unions and are believed to increase the public sector wages more than right wing governments (Borge et al., 2008; Revelli and Tovmo, 2007). Existing literature thus suggests that left wing governments are expected to have lower degree of efficiency.

Next, we introduce the power of the government. The first explanation for the possible impact of the power is that if the government has a lot of power than it is easier to implement the chosen policies and to withstand the challenges presented by different interest groups which renders the high powered government more efficient (Doumpos and Cohen, 2014; Ricciuti, 2004). On the other hand it is also possible that the power of the government has a negative impact on the efficiency of the local government. If there is a high majority it is easier to influence the choice of the expenditures and due to rent extracting efficiency may be of subordinate importance (Balaguer-Coll et al., 2007).

Next to the political variables we also introduce a number of financial variables. There is literature that shows evidence for the presence of grant illusion in local governments (Goeminne et al., 2017). The grant illusion literature claims that the increase in expenditures followed by a grant will be higher than the increase in expenditures followed by an increase in the municipality’s tax income (Courant et al., 1978). When dealing with grant illusion the government expenditures are thus more responsive to increases in grant money than to increases in tax income money (Goeminne et al., 2017). Extending this line of research one could wonder if grants will also have an impact on the efficiency of local governments. Efficiency could decrease if there is a bigger amount of the revenue that comes from state grants, as the costs of this inefficiency are now spread over a larger group of people (the entire state instead of the municipalities population) (Balaguer-Coll et al., 2007). Next, larger grants increase the potential for local governments to behave in a less efficient way (De Borger et al., 1994). If there is a larger amount of subsidies received from the state government it is possible that the local governments are less interested in minimizing their input usage (De Borger and Kerstens, 1996b). Third, incumbents are more wasteful in allocating money that was given to them through grants than they are regarding the self-collected tax money (Geys and Moesen, 2009a). Geys (2006) on the other hand states that grant money allows the government to provide more public goods for the same income level and thus that grants have a positive impact on the local governments’ efficiency. In the existing empirical literature the effect of federal subsidies is measured in various ways. First, part of the research includes the grants per capita the local government receives from the federal government. The majority of the papers using this indicator find a negative impact of the variable on the efficiency of local governments. The second variable that is used in the literature is the grant to revenues ratio. Research using this indicator finds on average a positive impact of the variable on the efficiency of local governments.

Next we want to investigate whether fiscal autonomy can improve the performance of local governments and thus the efficiency. Local governments that collect a significant amount of taxes to provide in their expenditures are believed to be more electoral accountable and thus will have a positive impact on the efficiency (Boetti et al., 2012).

Another variable that could have an impact on the efficiency of a government is the level of debt. Built debt in previous years has to be reimbursed out of the current budget. The interest and the capital repayments could therefore limit the ability of the government to provide public services to the inhabitants (Ashworth et al., 2006). Due to a high level of debt local governments could suffer of general resource mismanagement, reducing their efficiency (Revelli and Tovmo, 2007).

Next we also include the unemployment rate as an explanatory variable. Higher unemployment might imply that there have to be more expenditures towards social benefits that cannot be used in another expenditure category.

The concentration of the population could also play a role in the efficiency. On the one hand there could be scale economies which lead to an increase in efficiency due to the large scale by which the public services can be provided (Afonso et al., 2005). Some services are harder to provide and are more expensive6 if the population is more scarcely concentrated (Ashworth et al., 2006). A disadvantage of a higher population density is that there may be problems of congestion and problems due to the higher complexity of management (Da Cruz and Marques, 2014).

Next to that we also include the disposable income in our analysis. On the one hand, research states that high income inhabitants are generally more educated and they are more likely to demand more efficient public services (Knack, 2002). On the other hand, people with high income are expected to have high opportunity costs of time in monitoring the actions of the incumbents which leaves room for more inefficiency (De Borger and Kerstens, 1996a).

The democratic participation is also of interest to the efficiency of the local governments. . When the voters are more interested in understanding what politics is about then they are more able to demand a good service provision (Geys et al., 2010). On top of that they are also more able to monitor the politicians (Squire et al., 1987). This could have a positive effect on the efficiency in the public service provision (Borge et al., 2008; Šťastná and Gregor, 2015).

Table 1: Descriptive statistics and definitions.

Description Mean St. dev. Min. Max. Sourc

eInput

ExpendituresThe exploitation expenditures are calculated based on four financial categories: staff expenditures, production costs, transfers and capital payments on loans

815.75 238.75 108.03 2254.92 1

Outputs

Waste The amount of household waste and municipality waste collected by the municipality 153.06 51.93 57.51 658.19 5

SubsistenceThe number of people entitled to a subsistence minimum, employment measures or a reinstallation grant in percentage of the population

2.99 2.39 0 24.21 6

Building The number of residential and non-residential building applications 1.98 1.09 0 21.19 4

6 More costly infrastructure services like drinking water provision or garbage disposal collection.

Kindergarten Number of children attending kindergarten 3.78 0.72 0 9.66 7

Primary Number of students in primary education 6.28 1.27 0 14.87 7

Explanatory

Frag The number of parties in the ruling coalition 1.86 0.73 1 5 2

ICGThe ideological score of the coalition. The index can range between 0 and 10, with 0 being the left end of the spectrum and 10 the right end of the spectrum

5.04 0.69 2.50 6.70 2

Power The proportion of local council seats held by the ruling coalition parties 0.63 0.10 0.5 1 2

Fund

Unconditional lump-sum grant which stems from the "Municipality Fund" (Gemeentefonds), a general solidarity fund, provided yearly by the Regional Flemish government. We specified the data in real per capita terms

129.36 67.47 43.82 842.78 1

Fund/revenues

Unconditional lump-sum grant which stems from the "Municipality Fund" (Gemeentefonds), a general solidarity fund, provided yearly by the Regional Flemish government. We specified the data as a ratio to the revenues of the government

0.11 0.05 0.06 0.47 1

Tax/revenues The amount of taxes collected by the municipality divided by the total amount of revenues 0.51 0.25 0.17 0.76

Debt The stock of debt generated through the years. We specified the data in real per capita terms 880.34 472.45 0.02 6528.98 1

Unemployment The percentage of unemployed inhabitants 1.96 0.69 0.53 5.94 8

Population density Number of inhabitants per squared kilometer 1821.28 2878.77 7.67 47000.39 4

NTIThe average net income of the municipality's inhabitants. We specified the data in real per capita terms. Variable in 1000 euro

14.47 3.01 4.26 44.45 1

Democratic participation The percentage of blank votes 3.46 0.89 0 8.85

Notes: 1: "Administration Internal Affairs" (Administratie Binnenlandse Aangelegenheden), Flemish government. 2: Period 1995-2006: "FPS Home Affairs (FOD Binnenlandse Zaken),"General Directorate Institutions and Population" (Algemene Directie Instellingen en Bevolking), "Department Elections" (Dienst Verkiezingen). Period 2007-2011: Website "Local Government Agency" (Agentschap Binnenlands Bestuur), Period 1995-2006: MICE (Micro-Economics for Profit and Non-Profit Sector) research team of the Faculty of Economic, Social and Political Sciences, “Free University of Brussels” (Vrije Universiteit Brussel) and Solvay Business School. Period 2007-2011: “The municipal notebook Database 2010” (Het gemeentelijke zakboekje Databank 2010) from Wolters Kluwer for 2006-2011. 3: Own Calculations. 4: "General Directorate Statistics" (Algemene Directie Statistiek), "FPS Economy" (FOD Economie), Federal Government.5: OVAM (Flemish Regional Waste Authority).6: “FPS Social Integration” (FOD Sociale Integratie)7: Flemish agency for educational services (Agentschap voor Onderwijsdiensten)

8: public employment service of Flanders (VDAB)7. Double bootstrap truncated regression

Most research papers that implement DEA in a public sector context and try to explain the efficiency scores in a second stage use a tobit model (Afonso and Fernandes, 2008; Boetti et al., 2012; Monkam, 2014). Others use a single bootstrap truncated regression (Cuadrado-Ballesteros et al., 2013; Fogarty and Mugera, 2013). Only Doumpos and Cohen (2014) and Da Cruz and Marques (2014) use a double bootstrapped regression. Simar and Wilson (2007) (SW) show that only this last method is correct because the DEA efficiency scores that are used as dependent variables in the second stage are estimates of the real efficiency scores and are therefore serially correlated by construction (Simar and Wilson, 2011). The double bootstrap method performs the regression in two stages. In the first stage the efficiency estimates are bias corrected and in the second loop the bias-corrected efficiency measures are used in a truncated regression bootstrap. In this paper we use the dynamic version of the Simar and Wilson (2007) approach by Du et al. (2018). The dynamic version extends the SW approach by allowing for efficiency changes over time by estimating the efficiency for each year separately instead of pooling all the observations in the dataset and estimating one single efficiency frontier. The Monte Carlo experiments performed by (Du et al., 2018) show that the dynamic approach performs better than the pooled approach and that this improvement increases if the difference between the frontiers of the different years becomes bigger. Like all parametric and non-parametric methodologies, the (dynamic) two stage DEA has a limitation, it uses the separability assumption. The separabilty assumptions assumes that the estimated frontier is unconditional on the environmental variables. The dynamic double bootstrap method could in the future be extended by introducing the concepts of partial frontiers and/or conditional DEA (Bădin et al., 2012; Daraio and Simar, 2007; Du et al., 2018).

8. Results

The DEA analysis is sensitive to outliers as the outliers’ shape the frontier on which the efficiency of the other units is based (Bogetoft and Otto 2011). If the outliers are of a significant magnitude they could potentially bias all the efficiency scores of the other units as DEA measures relative efficiency to the frontier (Wilson 1995). Consequently, the outliers have been removed before performing DEA. In line with Geylani and Stefanou (2013) and Kapelko, Lansink, and Stefanou (2014), observations are seen as outliers if the ratio of outputs to inputs is larger (or smaller) than the median plus (or minus) two times the standard deviation.

As the outputs of a government are mainly out of control of the governments themselves it is assumed that municipalities minimize the inputs given the outputs rather than maximize the outputs given the inputs. For this reason we estimate an input-oriented model. One looks at how much the input can be reduced without changing the level of outputs (Afonso et al., 2005). We analyze the input oriented technical efficiency of 308 Flemish local governments for the government for the period 2005 to 2016 (see the histogram for every year in the appendix).

In order to know the driving forces of the efficiency of local governments we perform a second stage analysis. We perform the bootstrap truncated regression as suggested by Du et al. (2018) and regress the bias corrected efficiency scores obtained from the DEA analysis on the explanatory variables. The regression was 2000 times bootstrapped and the results are shown in table 2.

The first political variable that we introduced is the fragmentation of the government, this is measured as the number of parties in the coalition. We see that the fragmentation has a significant negative effect on the efficiency. The more political parties in the coalition the lower the efficiency of the local government. This finding is in line with the theoretical expectation formulated in section five of this research. The earlier cross-sectional study of Ashworth et al. (2006) for Flanders has come to the same conclusion. In column 3 of table 2 we replaced the variable fragmentation by two dummy variables. One party has a value of one if there is only one coalition party and is zero otherwise. More parties has a value of one when there are three or more coalition parties and zero otherwise. We see that one party governments have on average a higher efficiency than two party governments. Governments with three or more coalition parties have on average a lower efficiency score than two party governments.

Second, we introduce the ideology of the government. The ideology is measured as the Ideological Complexion of Government (ICG) index as introduced by Perotti and Kontopoulos (2002) to test the impact of ideology on the budget balance. For this index, local politicians were asked to score their political party on a left-right scale, with 0 being the left end of the spectrum and 10 being the right end of the spectrum. Averages for all political parties are calculated. The ICG index of every local government is calculated as the average of all ICG indices of every political party in the College weighted by the number of Aldermen each political party has in the College. From table 2 we see that the ideology of the government has a positive significant effect on the efficiency of local governments in Flanders in the first column. This means that if the government is more right winged, the efficiency will be larger. If we replace the icg index with a dummy variable that has the value one when the government is right winged and zero otherwise we come to the same conclusion. This result however is not robust in the second column of table 2.

The power of the coalition is measured as the share of seats the coalition parties of the College of Mayor and Aldermen have in the local council. The power of the local governments seems to have no impact on the technical efficiency. This could be due to the fact that there are two possible forces that work in different directions. On the one hand, high powered coalition are more capable of withstanding the challenges presented by different interest groups. And on the other hand, if there is a high majority it is easier to influence the choice of the expenditures and due to rent extracting efficiency may be of subordinate importance (Balaguer-Coll et al., 2007).

We see that the percentage of blank votes has a positive impact on the efficiency this could indicate that only people who are really interested in politics vote and are more able to control politics afterwards.

The fund the local governments receive from the Flemish government as well as the ratio of funds to all revenues seems to have no significant effect on the efficiency. We can consequently

not proof that the flypaper effect that can be found in Flemish municipalities has an impact on the efficiency of the local governments.

Fiscal autonomy seems to have a positive impact on the efficiency of the local governments. If the government herself collects a more significant amount of the resources, the efficiency of the government is higher. This is in line with the previously stated theoretical expectations.

The level of outstanding debt has a positive robust effect on the efficiency of Flemish local governments. This is not what is expected from the literature review but could be explained by the fact that local governments with a higher level of debt have larger interest and capital repayments and thus could be more inclined to use the remaining resources in a more efficient manner (Balaguer-Coll and Prior, 2009).

The net taxable income of the population seems to have no significant impact on the efficiency. This could be explained because there are two possible sources working against one another.

Next we introduce the unemployment percentage. If the unemployment percentage is higher than the municipality will be less efficient. This is in accordance with the theoretical expectations depicted in section four.

Lastly, the population density seems to have no significant robust impact on the efficiency of Flemish local governments.

Table 2: Dynamic bootstrapped truncated regression.

VRS(1) VRS(2) VRS(3)

Fragmentation -0.45*** -0.45 ***

One party 0.05***

More parties -0.45***

ICG 0.04*** 0.006

Right 0.04 ***Power -0.009 -0.0007 -0.0007 Democratic participation 0.05

*** 0.04

*** 0.01 ***

Fund -0.00009 -0.0009 -0.00009 Fund/rev -0.004 -0.0006 -0.004

Bel/rev 2.92*** 2.82

*** 2.96 ***

Debt 0.01*** 0.01

*** 0.01 ***

NTI -0.002 -0.0004 -0.002

Unemployment -0.0001*** -0.0001

*** -0.0001 *

Population density 0.005 0.005 0.009 *Year dummies yes yes yes

9. Conclusion

In this research we investigated the efficiency of the Flemish local governments for the period 2005-2016. This is the first research in the public context that uses the dynamic extension of the Simar and Wilson (2007) approach of Du et al. (2018). In the first step of the efficiency analysis we utilized the input-oriented bootstrapped DEA. In the second step we try to explain the bias corrected efficiency scores from by a number of political, financial and control variables. From the double bootstrapped truncated regression we learn that the fragmentation of the local government, the voter involvement, the ratio of taxes to revenues, the level of debt and the percentage of unemployment in the municipalities have a significant impact on the efficiency score. This draft version of the research still needs to have a couple of extension. We want to test the robustness of the results even further by doing next to the VRS analysis also a CRS analysis. We want to include some variables for the height of the tax rates and we will also include an overview of the exiting papers in the literature in the appendix. Next to that we also aspire to do some policy recommendations based on the results from our empirical analysis.

Appendix: Histogram of the calculated efficiency scores (2005-2016).

Sources

Afonso, A., & Fernandes, S. (2008). Assessing and explaining the relative efficiency of local government. The Journal of Socio-Economics, 37(5), 1946-1979.

Afonso, A., Schuknecht, L., & Tanzi, V. (2005). Public sector efficiency: an international comparison. Public Choice, 123(3-4), 321-347.

Alesina, A., & Tabellini, G. (1990). A positive theory of fiscal deficits and government debt. The Review of Economic Studies, 57(3), 403-414.

Ashworth, J., Geys, B., & Heyndels, B. (2006). Determinants of tax innovation: The case of environmental taxes in Flemish municipalities. European Journal of Political Economy, 22(1), 223-247.

Ashworth, J., Geys, B., Heyndels, B., & Wille, F. (2014). Competition in the political arena and local government performance. Applied Economics, 46(19), 2264-2276.

Bădin, L., Daraio, C., & Simar, L. (2012). How to measure the impact of environmental factors in a nonparametric production model. European journal of operational research, 223(3), 818-833.

Balaguer-Coll, M. T., & Prior, D. (2009). Short-and long-term evaluation of efficiency and quality. An application to Spanish municipalities. Applied Economics, 41(23), 2991-3002.

Balaguer-Coll, M. T., Prior, D., & Tortosa-Ausina, E. (2007). On the determinants of local government performance: A two-stage nonparametric approach. European Economic Review, 51(2), 425-451.

Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scale inefficiencies in data envelopment analysis. Management science, 30(9), 1078-1092.

Behr, A. (2015). Production and Efficiency Analysis with R: Springer.Benito, B., Bastida, F., & García, J. A. (2010). Explaining differences in efficiency: an application to Spanish

municipalities. Applied Economics, 42(4), 515-528. Benito, B., Solana, J., & Moreno, M.-R. (2014). Explaining efficiency in municipal services providers.

Journal of Productivity Analysis, 42(3), 225-239. Boetti, L., Piacenza, M., & Turati, G. (2012). Decentralization and local governments' performance: how

does fiscal autonomy affect spending efficiency? FinanzArchiv: Public Finance Analysis, 68(3), 269-302.

Bogetoft, P., & Otto, L. (2011). Benchmarking with DEA, SFA, and R, vol. 157 of International Series in Operations Research & Management Science: Springer.

Bönisch, P., Haug, P., Illy, A., & Schreier, L. (2011). Municipality Size and Efficiency of Local Public Services: Does Size Matter? Retrieved from

Borge, L.-E. (2005). Strong politicians, small deficits: evidence from Norwegian local governments. European Journal of Political Economy, 21(2), 325-344.

Borge, L.-E., Falch, T., & Tovmo, P. (2008). Public sector efficiency: the roles of political and budgetary institutions, fiscal capacity, and democratic participation. Public Choice, 136(3), 475-495.

Bruns, C., & Himmler, O. (2011). Newspaper circulation and local government efficiency. Scandinavian Journal of Economics, 113(2), 470-492.

Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European journal of operational research, 2(6), 429-444.

Cooper, W. W., Seiford, L. M., & Zhu, J. (2011). Handbook on data envelopment analysis (Vol. 164): Springer Science & Business Media.

Cordero, J. M., Pedraja-Chaparro, F., Pisaflores, E. C., & Polo, C. (2017). Efficiency assessment of Portuguese municipalities using a conditional nonparametric approach. Journal of Productivity Analysis, 48(1), 1-24.

Courant, P. N., Gramlich, E. M., & Rubinfeld, D. L. (1978). The stimulative effects of intergovernmental grants: Or why money sticks where it hits: University of Michigan, Institute of Public Policy Studies.

Cuadrado-Ballesteros, B., García-Sánchez, I.-M., & Prado-Lorenzo, J.-M. (2013). Effect of modes of public services delivery on the efficiency of local governments: A two-stage approach. Utilities Policy, 26, 23-35.

Cullinane, K., Wang, T.-F., Song, D.-W., & Ji, P. (2006). The technical efficiency of container ports: comparing data envelopment analysis and stochastic frontier analysis. Transportation Research Part A: Policy and Practice, 40(4), 354-374.

Da Cruz, N. F., & Marques, R. C. (2014). Revisiting the determinants of local government performance. Omega, 44, 91-103.

Daraio, C., & Simar, L. (2007). Conditional nonparametric frontier models for convex and nonconvex technologies: a unifying approach. Journal of Productivity Analysis, 28(1-2), 13-32.

De Borger, B., & Kerstens, K. (1996a). Cost efficiency of Belgian local governments: A comparative analysis of FDH, DEA, and econometric approaches. Regional science and urban economics, 26(2), 145-170.

De Borger, B., & Kerstens, K. (1996b). Radial and nonradial measures of technical efficiency: an empirical illustration for Belgian local governments using an FDH reference technology. Journal of Productivity Analysis, 7(1), 41-62.

De Borger, B., Kerstens, K., Moesen, W., & Vanneste, J. (1994). Explaining differences in productive efficiency: An application to Belgian municipalities. Public Choice, 80(3-4), 339-358.

Doumpos, M., & Cohen, S. (2014). Applying data envelopment analysis on accounting data to assess and optimize the efficiency of Greek local governments. Omega, 46, 74-85.

Du, K., Worthington, A. C., & Zelenyuk, V. (2018). Data envelopment analysis, truncated regression and double-bootstrap for panel data with application to Chinese banking. European journal of operational research, 265(2), 748-764.

Eslava, M. (2011). The political economy of fiscal deficits: a survey. Journal of Economic Surveys, 25(4), 645-673.

European Commission. (2014). Report on Public finances in EMU. Retrieved from Fogarty, J., & Mugera, A. (2013). Local government efficiency: evidence from Western Australia.

Australian Economic Review, 46(3), 300-311. García Sánchez, I. M. (2006). Efficiency measurement in Spanish local government: the case of municipal ‐

water services. Review of Policy Research, 23(2), 355-372. Geys, B. (2006). Looking across borders: A test of spatial policy interdependence using local government

efficiency ratings. Journal of urban economics, 60(3), 443-462. Geys, B., Heinemann, F., & Kalb, A. (2010). Voter involvement, fiscal autonomy and public sector

efficiency: evidence from German municipalities. European Journal of Political Economy, 26(2), 265-278.

Geys, B., & Moesen, W. (2009a). Exploring sources of local government technical inefficiency: evidence from Flemish municipalities. Public Finance and Management, 9(1), 1.

Geys, B., & Moesen, W. (2009b). Measuring local government technical (in) efficiency: an application and comparison of FDH, DEA, and econometric approaches. Public Performance & Management Review, 32(4), 499-513.

Goeminne, S., Smolders, C., & Vandorpe, E. (2017). The real impact of a one-off fiscal restriction: empirical evidence of a flypaper effect in Flemish municipalities. Public Money & Management, 37(4), 285-292.

Hibbs, D. A. (1992). Partisan theory after fifteen years. European Journal of Political Economy, 8(3), 361-373.

Kalb, A. (2014). What determines local governments' cost-efficiency? The case of road maintenance. Regional Studies, 48(9), 1483-1498.

Kalb, A., Geys, B., & Heinemann, F. (2012). Value for money? German local government efficiency in a comparative perspective. Applied Economics, 44(2), 201-218.

Knack, S. (2002). Social capital and the quality of government: Evidence from the states. American Journal of Political Science, 772-785.

Lim, B., Lee, K., & Lee, C. (2016). Free Disposal Hull (FDH) Analysis for Efficiency Measurement: An update to dea. The Stata Journal.

Marques, R. C., Kortt, M. A., & Dollery, B. (2015). Determining the optimal size of local government: the case of Tasmanian councils. Australian Journal of Public Administration, 74(2), 212-226.

Monkam, N. F. (2014). Local municipality productive efficiency and its determinants in South Africa. Development Southern Africa, 31(2), 275-298.

Perotti, R., & Kontopoulos, Y. (2002). Fragmented fiscal policy. Journal of Public Economics, 86(2), 191-222.

Persson, T., & Svensson, L. E. (1989). Why a stubborn conservative would run a deficit: Policy with time-inconsistent preferences. The Quarterly Journal of Economics, 104(2), 325-345.

Revelli, F., & Tovmo, P. (2007). Revealed yardstick competition: Local government efficiency patterns in Norway. Journal of urban economics, 62(1), 121-134.

Ricciuti, R. (2004). Political fragmentation and fiscal outcomes. Public Choice, 118(3-4), 365-388. Simar, L., & Wilson, P. W. (2007). Estimation and inference in two-stage, semi-parametric models of

production processes. Journal of econometrics, 136(1), 31-64. Simar, L., & Wilson, P. W. (2011). Two-stage DEA: caveat emptor. Journal of Productivity Analysis, 36(2),

205. Squire, P., Wolfinger, R. E., & Glass, D. P. (1987). Residential mobility and voter turnout. American

Political Science Review, 81(1), 45-65. Šťastná, L., & Gregor, M. (2015). Public sector efficiency in transition and beyond: evidence from Czech

local governments. Applied Economics, 47(7), 680-699. Varabyova, Y., & Schreyögg, J. (2013). International comparisons of the technical efficiency of the

hospital sector: panel data analysis of OECD countries using parametric and non-parametric approaches. Health Policy, 112(1), 70-79.

Warnez, B. (2016). Gemeentelijk zakboekje bestuur 2016. Mechelen: Kluwer.Weingast, B. R., Shepsle, K. A., & Johnsen, C. (1981). The political economy of benefits and costs: A

neoclassical approach to distributive politics. The Journal of Political Economy, 89(4), 642-664. Woodbury, K., & Dollery, B. (2004). Efficiency measurement in Australian local government: The case of

New South Wales municipal water services. Review of Policy Research, 21(5), 615-636. Worthington, A. C., & Dollery, B. E. (2001). Measuring efficiency in local government: an analysis of New

South Wales municipalities' domestic waste management function. Policy Studies Journal, 29(2), 232-249.