The effect of municipal amalgamations on municipal cost-efficiency and local public finance, a difference in differences approach
Student: Simon Bastiaan van Loo Student Number: 1122347 Date: June 11th 2018 Leiden University
Thesis Supervisor:
Abstract
This thesis investigates how municipal amalgamations influence social and spatial planning expenditure, taxation per capita, financial management and social service provision by local governments. Municipal budgetary and financial data for the period of 2005-2014 are used for this purpose. Using a difference in differences approach, I compare amalgamated municipalities with a control group of municipalities that amalgamated in a different year, and municipalities in which amalgamations were aborted. I find no amalgamation effects for both categories of spending per capita, taxation, financial management or service levels. Exceptions for these findings are a short-term negative effect on social expenditure per capita, and a negative pre-amalgamation effect on taxation. The former effect provides weak evidence that the central governments’ decentralisation of social tasks in 2015 could exploit some efficiencies gained through amalgamation. The main results persist through several robustness checks. I further analyse whether these findings hide opposite direction amalgamation effects for large and small municipalities. This is not the case. My results do not indicate that further amalgamations are inadvisable per se, as there are many legitimate reasons for consolidation, but they do indicate that advocates of amalgamation should not take efficiency gains and increases in management quality due to amalgamation for granted.
Contents
Abstract ... 1
1. Introduction ... 5
1.1 Research Question ... 6
1.2 Approach ... 6
1.3 Structure of the Thesis ... 7
2. Institutional Context... 9
2.1 Municipalities in the Netherlands ... 9
2.2 Municipal finance ... 10
2.3 Amalgamations ... 10
3. Literature Review... 13
3.1 Optimal population size ... 13
3.2 Service levels, taxes and citizens’ views ... 15
3.3 Temporary Effects ... 15
3.4 Expertise and Governing Capability ... 16
3.5 Financial Indicators ... 17
3.6 Hypotheses ... 19
4. Data ... 20
4.1 Dependent variables ... 23
4.1.2 Spending on spatial planning ... 24
4.1.3 Property Tax ... 25
4.1.4 Average Satisfaction Score of Social Support Services ... 26
4.1.5 Operational Result ... 26
4.1.6 Net Debt to Income Ratio ... 27
4.1.7 Solvability Ratio ... 28
4.1.8 Property Tax Exhaustion ... 29
4.2 Control variables ... 30
5. Research Setup ... 32
5.1 Main model ... 33
5.2 Model for Satisfaction with Social Support Services ... 34
5.3 Notes on Fixed Effects ... 35
6. Main Results ... 36
6.1 Social Expenditure ... 36
6.2 Spatial planning Expenditure ... 37
6.3 Property tax ... 39
6.4 Net Debt to Income Ratio and Solvability Ratio ... 41
6.5 Property Tax Exhaustion ... 43
6.6 Operational Result ... 45
7. Robustness Checks... 47
7.1 Different amalgamation-period dummies ... 47
7.2 Excluding general grant ... 49
7.3 Interaction with Population before Amalgamation ... 51
7.4 Alternative Control Groups... 54
7.5 Alternative Financial Indicators ... 57
7.6 Long-term pre-treatment effect ... 58
7.7 Confirmation and Rejection of Hypotheses ... 59
8. Limitations ... 60
8.1 Dynamic Models ... 60
8.2 Spatial Effects ... 60
8.3 Limitations related to the Dataset used ... 61
9. Conclusion and Recommendations ... 63
9.1 Policy Recommendations... 64
References ... 65
10. Appendix ... 69
The effect of municipal amalgamations on municipal cost-efficiency and local public finance, a difference in differences approach
1. Introduction
The Netherlands has recently seen a combination of further decentralisation of government tasks to the municipal level, and pressure exerted by the central government to consolidate municipalities into larger jurisdictions (Allers & de Kam 2010). In 2015, the central government delegated three major social policies to municipalities, the Social Provision and Support act (Wet Maatschappelijke Ondersteuning in Dutch), Youth Care (Jeugdwet) and the Participation Act (Participatiewet). In these acts a broad selection of tasks, including welfare and youth care were delegated to the municipal level, doubling the transfers from central government to municipalities while cutting about 20% of the available budget (Neijendaal 2014, 86). Additionally, decentralisations of more spatial planning and environmental tasks are being prepared in the Environment act (Omgevingswet in Dutch), which will be implemented in 2021 (Rijksoverheid, n.d.). The latter act constitutes another vast expansion of municipal responsibility, for which it is all but certain that municipalities will be sufficiently prepared upon implementation (ICTU 2017). For these decentralisations, a former Minister of the Interior, Ronald Plasterk (2013), has previously asserted that all municipalities should ideally have a population of approximately 100,000 inhabitants (Ministerie van Binnenlandse Zaken en Koninkrijksrelaties 2013a). The rationale behind further municipal consolidation primarily revolves around increasing “governing capability” and acquiring efficiency gains through economies of scale. Accordingly, the number of municipalities in the Netherlands has been declining steadily in the past decades from nearly a thousand in 1960 to 380 at the start of 2018 (CBS, 2004; CBS, n.d.:a).
The optimal size of governmental units has been the subject of debate. Oates decentralisation theorem (1973, 11-12) states that the smaller the unit size, the better the provision of services can be tuned to regional preferences and costs. Consequently, where preferences differ geographically to a great extent, smaller unit sizes might be preferred. This stands in contrast to other arguments relating to economies of scale, increases in professionalism and reduction of spill-over effects, that argue larger unit sizes are more efficient (Ibid 4-7; Shapiro 1963, 180). Because of this debate, and due to the significant economic and social costs that are associated with municipal consolidation, the question arises whether the projected efficiency gains are achieved and whether professionalism is increased or not. Previous studies (Allers en Geertsema, 2016) have shown that municipal amalgamations do not lead to gains in overall cost-efficiency or service levels nor reduction in overhead costs. This thesis, however, focuses on disaggregated categories of spending, as well as taxes and indicators of municipal financial health to look for gains in efficiency or professionalism. Furthermore, citizen satisfaction of social support services are used to test whether social service levels are higher in amalgamated municipalities, even if social expenditure may not decrease.
Additionally, the financial position of many Dutch municipalities deteriorated during the financial crisis in the years following 2008. The main cause was the fact that municipalities had bought private land in order to make it ready for development and subsequently sell it for a higher price. These lands were registered as assets on municipal balance sheets, but during the crisis, demand in the construction sector plummeted, and the price of land fell correspondingly, leaving many municipalities with problems. Some municipalities have had to devalue unexploited municipal lands on their balance sheets to such extent that they are negative. In other cases, municipalities have written off these lands to such extent that equity is below zero on their balance sheets, meaning that they have been illiquid. In 2012 even more losses due to forced writing off of lands bought by municipalities for exploitation was predicted for the period 2013-2015 (Deloitte 2013; Trouw 2012). These problems have prompted several Dutch organisations, among which the Dutch Association for Municipalities (VNG), to investigate anew the fiscal and financial1 management of municipalities and indicators for financial resilience (VNG 2016).
1.1Research Question
Regarding the nature of both the tasks already decentralised in 2015 and those to be decentralised in 2021, it is interesting to investigate whether amalgamations do lead to efficiency gains in the area of social expenditure or spatial planning. Taking into account these decentralisations, the possible lack of efficiency gains, and recent developments in the financial situation of municipalities, it is relevant to investigate whether amalgamations lead to better financial management of local government.
This leads to the following research question:
“What is the effect of municipal amalgamation on efficiency of social expenditure and spatial
planning, taxation and the quality of financial management of local government?”
With this research question, the thesis aims to explore what is the sense and the nonsense of the current calls of the central government for further municipal consolidation in light of decentralisation. To explore this, an empirical assessment of the effects of amalgamation on various aspects of municipal finance is employed.
1.2Approach
To investigate the effect of amalgamation on different aspects of municipal finance, following Allers and Geertsema (2016), I employ a difference-in-difference approach. I compare a treatment
1 The terms fiscal and financial management are used interchangeably in this thesis. They refer to
all decisions municipalities make with regard to revenue raising and expenditure, borrowing and loaning and other decisions that influence municipal balance sheets.
group of amalgamated municipalities with three control groups which are the group of all municipalities, a group of municipalities that were amalgamated in the research period, but in different years and a group that includes municipalities that were almost amalgamated. The latter group consists out of all municipalities that were amalgamated within the research period, as well as municipalities in which amalgamations have been cancelled and municipalities that have been or will be amalgamated in 2018 and 2019.
Because factors that drive selection for amalgamation are also likely to influence municipal finance, the exogeneity condition for a difference-in-difference approach is violated for the first control group. Therefore, the latter control group is preferred. First, the effect of amalgamation on spending on social services and spatial planning per capita is tested. Because gains in efficiency might not be translated into lower spending, but into better services instead. Consequently, I also test the effect of amalgamation on citizen satisfaction with social support services. Because efficiency gains might also be converted in lower taxes, amalgamation effects are also tested on the revenue for the most important tax for municipalities. Additionally, because amalgamation might not only lead to increased efficiency, but also to improved governing capability or civil service expertise, I test several financial indicators that are related to the quality of financial management. I find no significant effects of amalgamation on expenditure on social services and spatial planning. An exception to this is that for social expenditure, we I find a short-term decrease in per capita spending after amalgamation. For property tax, I find robust results indicating that in the years prior to amalgamation, municipalities lower their taxes, or at least increase their taxes less in anticipation of amalgamation. Not one of the financial indicators shows significant amalgamation effects, and no effect on citizen satisfaction with social services is found. I check the robustness of the main results by comparing the treatment group to each of the different control groups, and checking whether there are opposing amalgamation effects for small and large municipalities. Furthermore, I check the results by excluding control variables that may downwardly bias the amalgamation effect, because they are or might be related to the treatment. Finally, regressions on alternative financial indicators are performed to check the validity of the results for the initially used financial indicators.
1.3Structure of the Thesis
The layout of this thesis is as follows. Section 2 describes the particular financial, policy and political context in which decentralisation of governmental tasks and municipal consolidation in the Netherlands takes place. Section 3 will offer theoretical considerations on the organisation of decentral governments, municipal spending and the effects of municipal amalgamation as well as a review of empirical studies in the area of municipal government in general and effects municipal consolidation specifically. Section 4 describes the dataset and explains the dependent and control variables used in the analysis. Section 5 explains the research setup, empirical model and identification strategy. Section 6 will show the results of the difference-in-difference analysis. Section 7 will explain the rationale behind some robustness checks of the results and show the
results of these checks. Section 8 discusses the limitations of the scope of the results due to the data and methods used. Section 9 concludes and provides brief policy recommendations.
2. Institutional Context
This chapter will explain the institutional context in which municipalities operate and the manner in which amalgamations are planned and executed. Furthermore, the nature of, and rationale behind the decentralisations of 2015 is explained.
2.1Municipalities in the Netherlands
The Netherlands has 3 levels of government. Apart from national government, the country consists of 12 provinces and, as of January 1st 2018, 380 municipalities. The character of Dutch public administration is best described by the term ‘decentralised unitary state’. The state is decentralised because there are several categories of subnational governments with differing geographical units of jurisdiction, tasks, and geographical areas of competence. The primary responsibilities of municipalities are in the areas of spatial planning, education, social services, culture, and recreation and healthcare. Beside municipalities, the provincial governments have tasks relating to regional development, quality of local government and regional environmental and agricultural policy. Provincial governments also initiate the amalgamation process and submit a proposal for amalgamation to the national parliament (Schaap 2012 1-4).
As a rule, activities of local governments cannot transgress rules set by higher authorities. That is, municipal activities have to be in line with rules set by both the provincial government as direct supervisor, as well as rules set by central government. The unitary character of the state is primarily observable in the uniformity with which public services are implemented throughout the country. For example, central government makes rules about the welfare state by which local government has to abide (Schaap 2012, 5-6).
Local public administration is further characterised by the concepts of autonomy and co-governance. Autonomy refers to the freedom of municipalities to make local policy tailored to local issues and circumstances. Historically, autonomy has been on the decline as the modern welfare state increasingly demands redistribution and uniformity across geographical distance. Corresponding to declining autonomy, the extent to which local government engages in co-governance, the execution of policy made by a higher level of government, has risen. However, even when local government engages in co-governance, municipalities have some degree of policy discretion which varies greatly between policies (Ibid., 7). Moreover, recent reforms in co-governance tasks such as the execution of the work and social assistance act, and the decentralisation of the social support act, incentivise municipalities to use their policy discretion. The relationship between central and local government is not defined strictly in the constitution, which means tasks, competences, and funding shifts over time. These shifts are the result of considerations about the design of Dutch public administration, but also of negotiations and competence struggles between different levels of government (Schaap 2012, 7). It is unclear whether the decentralisations will increase or decrease policy discretion for municipalities. On the
one hand, devolution of tasks to municipalities implies more discretion, but because there have been efficiency-related cuts in funding for the decentralised tasks, policy discretion has come under pressure (Ibid., 15).
2.2Municipal finance
In the years 2008-2014, municipal income has approximately accounted for between 50 and 60 percent of grants for the central government (Ministerie van Binnenlandse Zaken en Koninkrijksrelaties n.d.). Although specific grants that match expenses exist, they are small in size and number, and gradually transferred to the general grant. Consequently, revenue streams from central government are largely independent of local spending and taxation, which means that efficiency gains can be used to increase service levels or to cut taxes or spending (Geertsema 2017, 16). During 2008-2014, two thirds of total grants from central government were general non-matching grants and one third consisted of specific grants. After 2015, the percentage financed by the general grant has increased due to the decentralization that took effect. In general, about ten percent is financed through local taxes and charges, of which the property tax is by far the most important as it constitutes on average 75% of local taxes. Municipalities are allowed to levy a limited number of taxes, but they are more or less free to determine the rates. Approximately one third of revenues is generated by market activities such as selling goods and services, renting out public buildings and by selling public lands to private developers (Ministerie van Binnenlandse Zaken en Koninkrijksrelaties n.d.).
Dutch provincial governments are responsible for financially auditing the municipalities that are located within their borders. If municipalities run excessive deficits for extended periods, they are placed under the “preventive” authority of their provincial government. If this does not improve the situation, the central government can opt to give financial aid through the general grant. These measures usually involve reform and austerity measures (de Ridder & Kooij 2013).
2.3Amalgamations
In political discussions in the Netherlands many municipalities are seen as too small to be effective. Decentralisations put a strain on the ability of municipalities to properly execute their co-governance tasks, either because they simply lack the capacity or because some tasks have become so specialised the small municipality relies on one civil servant to execute them. It is often assumed that amalgamation can mitigate these problems, and that larger municipalities will benefit from efficiency gains (Geertsema 2017, 17). At the start of the research period in 2004, there were 483 municipalities and at the beginning of 2017, there were 388 remaining.
Year N Average Population Size 2004 483 33660 2005 467 34915 2006 458 35664 2007 443 36925 2008 443 37033 2009 441 37383 2010 431 38457 2011 418 39846 2012 415 40314 2013 408 41126 2014 403 41760 2015 393 43004 2016 390 43536 2017 388 44025
Table 1 Number of Municipalities and Average Population within Research PeriodW
Although the Dutch central government endorses further consolidation of local government in general, amalgamations do not happen through a top-down process. Amalgamations often take years to complete from start to finish. Both municipalities and provincial governments can submit a proposal for amalgamation to the central government, but in practice provincial governments initiate amalgamations due to their responsibility for the quality of local government. Usually, proposals are only submitted after all stakeholders (citizens, municipal councils, local private organisations, and provincial governments) have been consulted. Municipal councils can make recommendations for changes in amalgamation proposals, but provincial governments may revise them as they see fit. Because amalgamations are unpopular and run the risk of taking many years to complete due to these consultations, provincial governments have been given more power to act unilaterally in recent years. Eventually, the province submits a proposal to central government and parliament makes a definitive decision (Ministerie van Binnenlandse Zaken en Koninkrijksrelaties 2013, 1-3).
Generally speaking, the Dutch government classifies a merger as an official amalgamation if the population of one of the merging parties changes by more than 10%. Otherwise, the government classifies it as an annexation. Annexations are not counted as amalgamations in this thesis, annexed municipalities are dropped from the analysis. In November of the year
preceding amalgamation elections for the municipal council of the newly formed, jurisdictions are organised. As a rule, newly amalgamated municipalities are compensated through the general grant. This compensation is approximately 40% of costs incurred due to the amalgamation in the first year and 20% of costs in the three years after that.
3. Literature Review
This chapter reviews the existing literature regarding the effects of municipal consolidation on different aspects of local public administration. Each section explains one or more mechanism at work, and explains the rationale behind one of the hypotheses, which is summarised in the final section of the chapter. It starts by reviewing literature that explains the change in population size of merged municipalities on spending per capita. Then literature regarding the effects of amalgamation on the level and quality of services provided by local government, including several ways these service levels have been operationalised, is reviewed. In the third section, the different effects of amalgamations through time are explained by reviewing both theoretical and empirical works. The fourth section describes the theoretical effect of amalgamation on municipal governing capability and expertise of civil servants. The fifth section explains why several financial indicators are related to performance of civil servants and politicians in a municipality. At the end of chapter 3, the hypotheses that were formulated due to the literature reviewed will be made explicit for each category of dependent variables.
3.1Optimal population size
The most common argument for municipal consolidation claims that increases in the size of political administrative units lower the costs of local services (Baldersheim & Rose, 2010). Increases in jurisdiction size can result in a decrease in service costs per citizen, or improved quality of services provided, or both. Bigger municipalities achieve these gains in efficiency because their larger size allows for more specialisation, and fixed costs can be spread over a larger population (Ibid.). However, increases in jurisdiction size might also increase costs per capita due to higher information costs. This might primarily arise because of increased requirements for the planning and control cycle of larger organisations. For example, there might be more obligations for monitoring and evaluation activities in larger organisations than in smaller ones.
Furthermore, various researchers have suggested that for each area of policy in local government, there exist economies of scale up to a certain population size, after which diseconomies of scale set in. In essence, there has been discussion on whether there is a U-shaped cost per capita curve related to population size of municipalities (Carey et al, 1996, Soukopovà et al 2014, 153). U-curve-based theories suggests that population size has an optimum in terms of governing capability for each policy area (Carey et al, 1996). If cost curves are u-shaped, amalgamation would be beneficial to small municipalities as they profit from economies of scale, but detrimental to municipalities that are already substantial in terms of population because they suffer from diseconomies of scale. There seems to be little empirical evidence that u-shaped cost curves exist. However, this could also be because scholars often do not test the disaggregated effect of amalgamation on municipalities of different sizes, causing opposite effects to cancel each other out (Allers & Geertsema, 2016, 661). A well-executed empirical study should take into account heterogeneous effects of amalgamation on municipalities of different sizes. Therefore, I will check the robustness of the main results of this thesis for different effects in municipalities of different
sizes.
It is obvious that a municipality does not provide just one service, but a vast array of services in different policy areas. The provision of these services requires different conditions that determine whether amalgamation or increase in size of a municipality can expect to generate economies of scale (Hirsch 1959). The implication is that every municipality has to try to find an average optimal size depending on the combination of services it provides. Indeed, Bikker and van der Linde (2015) have done research into the optimal population size with regards to economies of scale for spending on different policy areas by Dutch municipalities, and have found different optima for different areas. To make their estimations, Bikker and van der Linde use several cost models with different underlying assumptions to estimate the cost elasticity of additional citizens for different policy areas. One of Bikker and van der Linde’s additional findings was that the optimal size for Dutch municipalities in terms of population size had risen to 57,000 from 49,000 in recent years. Consequently, they recommended more municipal consolidation in the case of the Netherlands in the light of further decentralisation of government tasks, although maybe not to the point that every municipality would have around 100,000 inhabitants (Bikker & van der Linde 2015, 22).
However, municipalities provide services for a larger population than that of their own jurisdiction in several ways. First, it might be possible to contract certain services to private companies that serve multiple municipalities at once. However, contracting out of government services can come with costs such as market transactions costs or costs incurred due to principal-agent links (Savas 2002, 4). Second, municipalities can engage in inter-municipal cooperation, either by founding a new public body that produces the service, or by one municipality providing the service to other municipalities for a fee (Geertsema 2017, 7). However, control of municipal councils of these governmental bodies or inter-municipal cooperation agreements have been limited, and have come under scrutiny in recent years. Municipal councils have had varying success in exercising democratic control on these inter-municipal governing bodies (Nehmelman 2015, 5-7). Consequently, inter-municipal cooperation is not an optimal solution due to risks of creating democratic deficits.
Given the scale and importance of the tasks that have been decentralised in 2015 and will be decentralised in 2021, it would be convenient if the amalgamation policy executed in the Netherlands brought municipalities closer to the optimal size for the provision of services in spatial planning, environment and social policy. If this were the case, amalgamations would reduce the difficulties that come along with inter-municipal cooperation and contracting out of services to private parties, and will likely improve cost-efficiency.
3.2Service levels, taxes and citizens’ views
In their review of literature in the field of municipal amalgamation, Eythórsson & Karlsson (2016, 6) state that results from empirical studies about efficiency gains due to consolidation point in different directions. When results do not suggest efficiency gains in terms of average costs per citizens, these gains might be detected in better or more services. This is the case because decreases in spending can be unpopular, so policy makers opt to decrease taxes or increase service levels instead. Rises in service levels however, are usually harder to quantify, evidenced by the fact that different indirect measures such as birth, migration rate and housing prices have been used in various studies (Allers & Geertsema, 2016, 662). Moreover, improved service levels might have a delayed onset or only be distributed to a part of the population, further complicating proper measurement.
To measure increased service levels Eythòrsson & Karlsson (2016, 11-14) use survey data from citizens living in amalgamated municipalities. The survey used asks citizens about their views regarding the general level of services in their municipalities, as well as their views regarding services in all the different fields their municipality is responsible for. Specifically, the questions ask whether citizens think service levels have improved since amalgamation. The authors compare the results between municipalities using the percentage of respondents that answered ‘agree’ or ‘totally agree’.
3.3Temporary Effects
I can expect amalgamations to have different effects through time as well. First, spending might be higher in municipalities that know they will merge in the near future. Politicians and civil servants may act as ‘amoral calculators’ because after amalgamation a common pool for funds is created (Lodge & Wegrich 2012, 91-94). Political scientists have long been aware of these so-called common pool problems, which arise in a variety of different situations when public goods are provided (Ostrom, 1990). Unprincipled municipalities can, for example, choose to increase their deficits and debts in order to finance expensive projects in their municipality. They will do so knowing that the costs of these projects will be shared between citizens of the amalgamated municipality in the future, while the benefits will fall to citizens in the former municipality (Jordahl & Liang 2009, 157-158). Allers & Geertsema (2016, 663) mention several authors who have found evidence that this behaviour exists, although not all of them consider all relevant ways spending might rise in the years before amalgamation. Other authors, however, link rising spending levels to the size of the common pool relative to the initial pool of funds available to a municipality, and provide convincing support to the amoral calculator hypothesis (Saarima & Tukkainen 2014, 2-4).
Both in the years leading up to amalgamation and in the first years afterwards, I can expect municipalities to incur various types of reorganisation costs. Town halls might have to relocate, computer systems have to be combined, two bureaucracies have to be integrated and local rules have to be set in line with each other. Amalgamations are also likely to have a disruptive effect on
the execution of regular government tasks, causing efficiency and effectiveness to decrease for a period of time (Ibid.). This is why increases in costs per capita are expected in various policy fields in the years immediately after amalgamation. Reorganisation costs can be expected to be higher in cases where multiple municipalities amalgamate, and in cases where amalgamating entities are more equal in size. For example, when two municipalities of equal size amalgamate, it is more likely the new municipality adopts new and more office space than when a small municipality merges with a large municipality. Lastly, amalgamations are theoretically expected to have incidental positive effects as well. Organisations tend to have entrenched processes in place to execute tasks that do not necessarily update when needed. Consolidation drives municipalities to examine these processes which can result in more efficient management (Hansen et al 2014, 201).
3.4Expertise and Governing Capability
Another consolidationist argument is that amalgamations might lead to increased effectiveness in redistributing economic and organizational resources across the unit (Baldersheim & Rose, 2010). This argument is corroborated by several sources. Oates (1973) states that larger municipalities may attract better civil servants, council members and aldermen and consequently improve service levels or professionalism. Bikker and van der Linde (2015a, 347) also attest that an increase in expertise and professionalism in the local civil service, usually comes with larger jurisdiction sizes.
Boogers et al (2007) discuss several practical and theoretical conceptions of ‘governing capability’ (‘bestuurskracht’ in Dutch). The first conception is that of quality of local public administration as a democratic political arena, a service provider and a unit of legal administration. Quality then, is the extent to which municipalities effectively play these roles. Another conception perceives governing capability as the extent it is characterised by responsivity, integrity and reliability. Yet another conception defines governing capability as the extent to which municipalities are able to execute both the tasks mandated by central government as well as any tasks that are part of their own policy space. Because cooperation with other municipalities is often necessary to execute these tasks well, capability to be a good cooperating partner is also seen as part of governing capability in this conception.
Eventually, Boogers et al (2007, 7) synthesise these conceptions in a conception within which governing capability is measured as the extent to which a municipality can play each of the five following roles:
• Governing Body of the Local Community: To what extent is local government responsive
to the particular wishes of the local community, and is it able to commute these wishes to actual policies or actions?
• Partner in Public Administration: Is the municipality able to be a capable governing
partner for other governments at different levels, and is it able to tackle issues that supersede its territorial boundaries in cooperation with provincial or central government?
• Upholder of Rules: To what extent is the municipality able to uphold and maintain both
the laws of central government as well as the rules it sets itself?
• Local Policy Maker: What does the municipality prioritise within both its mandated and
autonomous tasks? What kind of projects are undertaken, and is the municipality able to manage these projects well?
• Service Provider: From garbage collection to doling out of social assistance, to providing
subsidies and permits, municipalities provide a vast array of services. To what extent is a municipality capable of providing these services in an efficient manner that is satisfactory to its citizens?
Furthermore, Boogers et al (2007, 14) state that gains in expertise in civil servants through amalgamations in the Dutch context can be considered an artefact of the financial relationship between central and local government. Smaller municipalities do not receive the financial means from central government to employ better salaried civil servants. Consequently, municipal consolidation, at least in principal, increases the quality of civil servants. However, there could be other, more efficient, ways to do this like inter-municipal cooperation on regional issues, or implementation of standards for human resources for all municipalities. All in all, Boogers et al state that municipal consolidation in the Netherlands has not lead to increases in governing capability or significant cost reductions. According to Boogers et al previous research shows that effects of increases in jurisdiction size on financial efficiency and governing capability of local government are unclear. However, the research cited usually uses qualitative methods for the assessment of governing capability, and where quantitative methods are used, these methods test other measures than this thesis.
In this thesis, the first dimension and latter two dimensions of the conception of Boogers et al of governing capability are tested. The role of ‘local policy maker’ is primarily tested by looking at the effect of amalgamation on several indicators of financial health, and managing of the planning and control cycle such as difference between budgeted and realised shortages and real debt levels. The role of ‘service provider’ is primarily tested by using spending per capita and satisfaction scores of survey data as dependent variables. By looking at citizen satisfaction with social support services, I test the role “Governing Body of the Local Community”.
3.5 Financial Indicators
Apart from spending per capita in different categories and measurement of citizen’s views there are other relevant financial indicators for measuring the fiscal situation of municipalities. Honadle
et al (2003) review various sets of indicators of fiscal health for local governments for various
purposes. They define so-called ‘actual’ fiscal health of local government as the difference between the revenue-raising capacity and the expenditure need. It measures a municipalities ability to deliver services within the restrictions of its economic and social circumstances, the grants it receives and the fiscal framework in which it operates (Ibid, 18). It is important to note that these
restrictions entail that not all possible financial indicators are also indicative of the quality of management in a given municipality. For example, the per capita assessed valuation, which measures the demand for public services and infrastructure relative to the taxable property base, is seen as an important indicator for long-term fiscal health, but policy makers have only marginal influence on it (Ibid, 29). The authors argue that different sets of indicators are appropriate for different purposes and audiences. For example, cash liquidities, solvability ratios and yearly realised deficits indicate more about the short term financial health, while long-term debt and base of taxable property indicate more about the long term. Honadle et al also argue that it is never possible to fully assess the financial health of a municipality by looking at one indicators, and that an appropriate set of indicators can be chosen according to audience and purpose.
Hansen et al (2014, 199) define three concepts of fiscal management: resource management, activity-based management and result-based management. The focal points of these concepts are economy, efficiency and effectiveness and input, output and outcome respectively.
Production Flow
Management Concept Fiscal Management Focus
Input Resource-Based Management Economy Output Activity-Based Management Efficiency
Outcome Result-Based Management Effectiveness Table 2. Dimensions of Fiscal Management by Hansen et al
Viewed through this lens, a focus on spending per capita and service levels as chosen by Allers and Geertsema (2016), is a focus on activity-based management and result-based management. An empirical study of the effects of municipal consolidation should take into account indicators for all three concepts of fiscal management proposed by Hansen et al if it wants to provide a complete presentation of these effects. Hansen et al (2014, 205) choose to focus on resource management of local government by using four financial indicators: Budgeted operating results, realised operating results, long-term debt and liquid assets. These financial data are possible indicators of both financial management and the ‘local policy maker’-role as described by Boogers et al. as they indirectly relate to the performance of civil servants. After all, municipalities that manage projects well, will in theory report lower levels of both budgeted and realised deficits and smaller differences between the two indicators, as well as higher levels of liquid assets and lower levels of debt. Contrary to Allers & Geertsema (2016; 2017) Hansen et al find significant, but statistically unstable, short term results for their financial indicators.
The indicators chosen by Hansen et al, could use refinement to increase their validity. For instance, higher revenue or liquid assets could offset high debt levels, rendering the indicator inaccurate.
Accounting for these in the indicator could more accurately affect the resource-based management they want to measure.
De Ridder and Kooij (2013) evaluated the financial position of Dutch municipalities. Because provincial governments only intervene in emergencies, they analysed how many municipalities were in danger of such an intervention. In order to do this, they developed the financial concepts
Instability and Vulnerability. Instability refers to a situation in which revenues and expenditures
are not balanced. Structural budgetary deficits and suddenly rising debts indicate instability. Structural deficits are operationalised as a negative result in three consecutive years and suddenly rising debts are operationalised as a situation in which (total net debt – all possessions with economic value) is 20% higher than in the previous year. The authors consider municipalities to be vulnerable when financial risks are not sufficiently covered on the municipal balance sheet, or when there is no room to levy extra tax revenues. The former situation arises when the total value of outstanding debt in all forms is larger than 200% of total assets excluding outstanding debts. The latter situation arises when the level of the property tax is equal to or higher than the norm value given by the central government. Although the critical values for these indicators are set rather arbitrarily by the authors, the indicators themselves are especially useful for this thesis as far as they can be influenced by municipal governments in the short-to-medium-term, accurately measure different aspects of financial health so that I can identify well-managed and poorly managed municipalities. It is possible, at least largely, to calculate these indicators from entries of municipal balance sheets which are publicly available through Statistics Netherlands (CBS) (Ibid.).
3.6Hypotheses
In conclusion, 5 hypotheses were formulated in this chapter:
1. In the years preceding amalgamation, municipalities are expected to increase spending per capita or decrease taxes.
2. In the first three years following amalgamation, temporary efficiency losses due to amalgamation are expected to be visible in higher spending per capita and property tax. 3. After three years, efficiency gains due to amalgamation are expected to be visible through
lower spending per capita and/or property tax.
4. Amalgamations are expected to increase municipal service levels, particularly if spending and taxes per capita remain constant.
5. Amalgamations are expected to increase governing capability and professionalism of civil servants, which translates into better financial and resource-based management.
4. Data
Statistics Netherlands provides municipal budgetary and balance sheet data for the years 2005-2014. Table 3 shows the means and standard deviations for all dependent and control variables for the treatment and control groups. The dataset contains several categories regarding projected and realised revenue and spending per capita in different degrees of disaggregation For example, there is a variable for (projected and realised) total spending for each municipality, variables for spending on social services and spatial planning, and variables for the constituent categories of spending in these categories. The panel is unbalanced because data are missing for some municipalities in some years. Robust standard errors are used to account for bias that could arise from this. Amounts are expressed in euros of 2015 using the consumer price index provided by Statistics Netherlands. In the data for projected spending and revenues, it is possible that some measurement error is present. This is because municipalities have to provide Statistics Netherlands with provisional budgetary projections before November 14th of the year preceding the year that the budget takes effect. Municipalities can still make amendments to their budgets after this date, but they are not obligated to send these amendments to Statistics Netherlands. It is therefore possible that the data does not reflect these amendments (CBS, n.d.:b).
Additionally, I use survey data collected by municipalities and made available centrally by Statistics Netherlands. After the decentralisation of social support tasks to municipalities in 2015, they were obligated to gather both administrative and survey data on the execution of these tasks as part of the municipal monitor for the social domain (“Gemeentelijke Monitor Sociaal Domein” in Dutch). Statistics Netherlands has provided municipalities with a data collection format and manual to ensure that data is collected according to scientific standards. Statistics Netherlands checks the data before publication, on among others: sample size relative to population; correctness of data and whether respondents received the services in the year they were surveyed (VNG 2018). Reasonably complete data for this variable is only available for the years 2016 and 2017. This means that for most amalgamated municipalities there is only data for the period after amalgamation.
Variables All municipalities Amalgamated Not Amalgamated
Amalgamated or almost
amalgamated Min Max
Social expenditure per Capita 262,99 224,14 271,61 234,45 -173,98 846,05
(1,62) (3,39) (1,80) (2,73)
Spatial planning Spending per Capita
(narrow) 33,67 32,80 33,87 33,34 -474 401
(0,39) (0,84) (0,45) (0,63)
Property tax per Capita 156,28 143,12 159,20 145,42 37,11 588,86
(0,82) (1,52) (0,94) (1,41)
Average Social Service Satisfaction 0,77 0,77 0,76 0,77 0,21 0,97
(0,0027) (0,0098) (0,0029) (0,0052)
Operational Result per Capita 25,40 26,96 25,05 29,98 -1580,8 5384,92
(4,07) (9,96) (4,45) (8,24)
Debt to Income Ratio 0,53 0,43 0,55 0,48 -1,49 2,55
(0,0065) (0,013) (0,0074) (0,011)
Solvability Ratio 0,36 0,39 0,36 0,37 -0,32 0,91
(0,0028) (0,0059) (0,0032) (0,0048)
Property Tax Percentage 0,070 0,060 0,072 0,065 0,01 0,41
(0,00047) (0,00079) (0,00055) (0,00088)
General Grant 795,47 735,36 808,80 754,68 234,46 2708,30
Variables All municipalities Amalgamated Not Amalgamated
Amalgamated or almost
amalgamated Min Max
Average House Price 191869,30 239769,70 230767,30 230260,9 72000 702000
(1256,33) (2347,12) (908,15) (1651,98) Population Size 38962 25206,68 41803,06 23891,1 919 844947 (801) (829,80) (946,4) (596,93) Council Concentration 0,20 0,22 0,20 0,21 0,10 0,47 (0,00072) (0,00198) (0,00077) (0,0013) Council Ideology 0,14 0,16 0,13 0,16 -0,62 0,91 (0,0029) (0,0074) (0,0032) (0,0058) Population Density 1028,25 1138,43 1021,36 823 109 6066 (10,16) (32,82) (10,6) (16,87) Number of Municipalities 483 138 345 196
Table 3. Means, standard errors, and lower and upper bounds of dependent and control variables (2005-2017) 2
Composition of “Amalgamated or Almost Amalgamated” is the group “Amalgamated”, plus municipalities in which amalgamations have been aborted due to political reasons and municipalities that are scheduled to amalgamate in 2018 and 2019.
8 Annexed municipalities are counted in the “Not Amalgamated Group”
4.1Dependent variables
Eight dependent variables were/are used to get a complete overview of the effects of
amalgamation on the fiscal management of municipalities. All dependent variables except for the survey and property tax exhaustion variables are expressed in euros per capita. In terms of the categorisation of financial management offered by Hansen et al (2014, 199), the dependent variables can be categorised as such:
Table 3 Classification of dependent variables Production
Flow
Management Concept Fiscal Management Focus Dependent variable Input Resource-Based Management Economy Operational Result Net Debt to Income Ratio Solvability Ratio
Exhaustion of Property Tax Capacity
Output Activity-Based Management
Efficiency
Property Tax
Spending on Social Support Spending on Spatial planning Outcome Result-Based
Management
Effectiveness
Average Satisfaction Score Spatial planning Indicator
The dependent variables are explained briefly in the remainder of this section. The direction in which they are expected to change due to amalgamation will also be mentioned.
4.1.1 Social expenditure
This variable measures spending per capita on social services in general and services provided as part of the social support act. It includes spending on social assistance, retraining programmes, elderly care, and youth and social support services. It is relevant to analyse the effect of amalgamation on this variable because the central government implemented both amalgamations and decentralisation of tasks in this domain.
Figure 1. Social expenditure per capita for different groups, in euros of 2015,,confidence intervals shown for Amalgamated and All Municipalities groups
4.1.2 Spending on spatial planning
This variable measures spending per capita on spatial planning. It is relevant to analyse the effect of amalgamation on this variable, because the central government implemented both amalgamations and decentralisation of tasks in this domain. There are two categories in the financial statements of municipalities regarding spatial planning. One is a broad category of all activities regarding spatial planning and development, which includes revenues from issuing building permits and exploitation of real estate and sale of public lands for development. This category includes a smaller category, which is strictly devoted to spending on spatial planning and any revenue that may exist from specific grants from the central government. I use the smaller category for my analysis as revenues from exploitation of real estate and sales of lands distort the spending figures, because of their large and incidental nature.
0 50 100 150 200 250 300 350 400 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 E u ro s p e r C a p it a
Figure 2. Spatial planning (narrow) spending per capita for different groups, in euros of 2015, confidence intervals shown for Amalgamated and All municipality groups
4.1.3 Property Tax
This dependent variable is a measure of the property tax per capita that is levied per year in each municipality. As I have seen, property tax is by far the most important tax that municipalities are allowed to levy themselves. Municipalities usually set a new property tax rate for each year during the budgeting phase. If amalgamations lead to efficiency gains, per capita spending might not reflect these gains because spending cuts might be unpopular. However, policy makers can opt to invest possible efficiency gains in a decrease of taxes, so I might observe amalgamation effects in property tax per capita.
0 5 10 15 20 25 30 35 40 45 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 E u ro s p e r C a p it a
Figure 4. Average property tax (OZB) in different groups, in euros of 2015, confidence intervals shown for Amalgamated and All municipality groups
4.1.4 Average Satisfaction Score of Social Support Services
The survey data asks citizens who have used social support services whether they agree that the provided services were adequate in several ways. Although self-reported data are not fully reliable due to their subjectivity, they are an appropriate source of data in this case. Other measures for service levels that have been used in the literature are: housing prices; life expectancy; migration; birth rate and other aggregate indicators of general wellbeing. These measures are not sufficiently accurate if one wants to measure service levels of services in specific policy area. This variable is the average percentage of respondents that answered ‘totally agree’ or ‘agree’ on nine questions regarding their satisfaction with services rendered. Unfortunately, only data on the percentage of respondents that gave these answers are available, not the percentages for all possible answers. Survey data is only available for 2016 and 2017. I include this variable because if amalgamation effects are not visible through lower spending or taxation levels, this could be caused by policy makers opting to increase service levels instead. Accordingly, if amalgamation effects on social expenditure are non-existent, I might expect citizen satisfaction to be higher in amalgamated jurisdictions. Alternatively, when amalgamation effects are visible through decreases in social expenditure per capita, this might also be caused by a decrease in service quality. Therefore, it is relevant to check whether this is the case.
4.1.5 Operational Result
In the data set, several variables relating to the projected and realised result of the statement of accounts exist in the dataset. Municipalities usually project their budgets to be neutral. That is: their results are zero. However, there exists an account “deposits to or withdrawals from reserves” in the budget that functions in essence as a closing account for the budget which ensures the result
0 50 100 150 200 250 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 E u ro P e r C a p it a
is closer to zero. Municipalities, in theory, mostly deposit to or withdraw from reserves to postpone payments for costs that will be incurred in later years for which the revenue is already raised. For example, when the central government has already included earmarked funds for the construction of a road in year X, but payments do not start until year X + 1, a municipality will add those funds to the reserves for that particular goal at the end of year X. In this way municipalities try to use reservations to present their results more favourably, avoiding presenting large surpluses or deficits (Van der Schaaf 2010, 30-31). A good indicator, then is the operational result, which is the result of the statement of accounts before these reservations are made. Large negative values for this variable indicate bad financial years for the municipality. Because the projected result is usually zero. I subtract the projected result from the operational result for the variable I use in the analysis. The spike in operational result in 2009 and the decline in subsequent years for virtually all municipalities is as of yet unexplained, but is likely related to aforementioned purchases of lands for building development and the economic crisis.
Figure 5. Operational deficit or surplus per capita for different groups, in euros of 2015, confidence intervals shown for Amalgamated and All municipality groups
4.1.6 Net Debt to Income Ratio
This variable gives an indication of the size of the pressure of interest payments on a municipality’s budget. A high net debt ratio could spell higher interest rates in the future. De Ridder and Kooij (2013) use the net debt ratio as a metric for financial stability. Moreover, policy makers are instructed by the Dutch Association of Municipalities (VNG) to use this metric for the assessment of the financial stability of their municipalities. Consequently, I expect well-managed municipalities to have a lower ratio than poorly managed ones (VNG 2017; VNG n.d.). This variable is computed as: (debt + provisions) – (liquid assets + claims) per capita, divided by the total revenue per capita.
-200 -100 0 100 200 300 400 500 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 E u ro P e r C a p it a
Figure 6. Net debt to income ratio for different groups, in percentages, confidence intervals shown for Amalgamated and All municipality groups
4.1.7 Solvability Ratio
The solvability ratio represents the extent to which the total balance sheet value is not burdened by debt. It displays in a percentage which part of the asset side of the balance sheet theoretically is ‘free’ from debt if all assets were sold immediately. I expect well-managed municipalities to have a higher solvability ratio than poorly-managed ones. The Dutch Association (VNG) for municipalities encourages policy makers to use the solvability ratio as a secondary indicator for municipal financial resilience, alongside the net debt to income ratio (VNG 2017, 10). The solvability ratio is computed as the total of reserves divided by the balance sheet total.
0% 10% 20% 30% 40% 50% 60% 70% 80% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 N e t D e b t a s p e rc e n ta g e o f in co m e
Figure 7. Solvability ratio for different groups, percentages, confidence intervals shown for Amalgamated and All municipality groups
4.1.8 Property Tax Exhaustion
Municipalities may increase taxes in times of financial hardship. Property tax is by far the most important instrument municipalities have to do this. Whether citizens accept a tax increase is likely dependent on the height of the tax percentage relative to the percentage in other municipalities. The Dutch Association for Municipalities (VNG) instructs policy makers to use the property tax percentage relative to the mean percentage, as a metric for estimating the available bandwidth for increasing the property tax in times of financial hardship (VNG n.d.). I calculate the tax percentage by taking the projected property tax revenue per capita in a given year and dividing it by the municipal average property price. In my analysis, I compare treatment and control groups to test whether amalgamation leads to lower percentages.
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 S o lv a b il it y r a ti o
Figure 8. Property tax per capita as percentage of average house price for different groups, confidence intervals shown for Amalgamated and All municipality groups
4.2Control variables
I control for several municipal characteristics. All control variables are available for all municipalities for the period 2004-2017. Therefore, the control variables can also be used in the regressions for the survey data. I largely follow Allers & Geertsema (2016) in the variables I control for. All variables have been retrieved from Statistics Netherlands:
• Population size. As explained in chapter two, population size is related to the decision to
amalgamate or not. Additionally, spending per capita might not grow proportionally to the number of citizens.
• Average house prices. This variable controls for prevailing service levels and for
background variables related to social expenditure. That is, in municipalities with higher average house prices, I expect services to be higher. Because amalgamations could lead to better services and consequently higher housing prices, I also show results that do not include this control variable as a robustness check.
• Amount received from central government for the general grant per capita. As the general grant for each municipality is calculated based on more than 60 municipal characteristics, this control variable controls for a multitude of relevant variables simultaneously (Ministerie van Binnenlandse Zaken en Koninkrijksrelaties 2017). Moreover, as the general grant is a significant portion of the total income of municipalities, it is likely to influence spending per capita in many areas. As stated earlier, the general grant corrects 0,00% 0,02% 0,04% 0,06% 0,08% 0,10% 0,12% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p ro p e rt y t a x p e rc e n ta g e
for 20-40% of the amalgamation costs municipalities are expected to incur, so I also show results that do not include this control variable as a robustness check.
Additionally, I use two political control variables. The first is a Herfindahl-Hirschmann index of power concentration of political parties within the municipal council, to account for heterogeneity of preferences within the council, divided by 10000. The idea is that the more concentrated the power within the council, the easier spending is adjusted or services or financial management policy is improved on a political level. The second variable is an indicator of political ideology of the council. The indicator is computed by coding -1 for each seat that belongs to a left party, 0 for parties without clear ideology (for example local parties without a clear ideological stance) and 1 for each seat belonging to right leaning parties, and dividing the total by the number of seats in the council. This operation results in a continuous scale that ranges from -1 for an entirely left council and 1 for an entirely right leaning council. I include these variables because several authors in the area of comparative political economy have shown that both more concentration of political power, as well as a higher percentage of left wing parties in representative bodies, lead to more government spending (Allan & Scruggs 2004; Armingeon 2012).
To create a variable for political ideology of the municipal council, I coded the political parties as follows:
• VVD, CDA, PVV, SGP, CU/SGP unions, local parties with ‘Leefbaar’ in their name and other right leaning local parties scored ‘1’.
• PvdA, Groenlinks, SP, local parties with ‘Progressief’ or equivalent in their name nad other left leaning local parties scored ‘-1’.
• All other parties scored ‘0’.
Additionally, dummy variables are used for election years as well as pre- and post-election years, as elections might influence politicians and civil servants to spend more, change financial indicators where possible, or increase service levels (Allers & Geertsema 2016, 672). The inclusion of these dummies might interfere, however, with the pre-treatment effect variable included in the model. In order to check whether this is the case, models will be fitted and shown with and without these election variables.
I retrieved both the election data used for the power concentration and political ideology variables from the website of the Dutch electoral council. The election dummy variables were derived from these data as well. The electoral council is an advisory council responsible for executive tasks in organising Dutch elections. Furthermore, the council announces and retains all election results (Kiesraad, n.d.).
5. Research Setup
I use difference-in-difference estimation to compare changes in the different dependent variables for amalgamated municipalities, with changes in a control group of municipalities that have not been amalgamated. I use amalgamation dummies to estimate the effect of amalgamation. Difference-in-difference estimation has been used in the field of consolidation of local government by, for example, Jordahl and Liang (2010), Hansen et al (2012) and Reingewertz (2012). I follow these authors by using the fact that some municipalities have been amalgamated while others were not. Additionally, I use the fact that amalgamations in the Netherlands are spread through time to compare amalgamated municipalities with municipalities that also amalgamated in the research period, but in a different year (Allers & Geertsema 2016, 666).
I have organised the data set so that every municipality that is amalgamated within the research period contains a dummy-variable that is ‘1’ for each year. For all observations in years following the amalgamation of a municipality, data of the new amalgamated entity is entered in the old entity or entities. For example, if municipality A and B amalgamate into the newly formed municipality C in 2007, data for municipality C are used for municipality A and B from observations in 2007 and onwards and municipality C is dropped from the dataset. This happens 30 times within the research period. If municipality A merges in 2007 into municipality B that is thereafter still called “B”, the data from municipality B for all observations in 2007 and onwards are used for both municipalities. This setup allows me to also use data from amalgamated municipalities before they were amalgamated, instead of only using the number of municipalities that existed on January 1st 2017.
It is unlikely that selection for amalgamation happens at random, and the variables that drive amalgamation are likely to be related to the various variables of interest. Thus, the exogeneity condition of the difference-in-difference approach is violated (Lechner 2010, 177). To counter this, I try to control for selection effects by using municipality fixed effects and year fixed effects. Difference-in-difference estimation requires the counterfactual time trend of treatment and control groups to be the same. It is unlikely that this condition is met for the current research, which could lead to omitted variable bias (Angrist & Pischke 2008, 165). To correct for this, and for time-invariant variables affecting treatment status, I use unit fixed effects. The inclusion of these controls allows me to relax the common time trend condition for difference-in-difference estimation (Angrist & Pischke 2014, 197).
Although I include these extra controls, some bias may still exist. This might exist because the common pool effect mentioned earlier causes municipalities to anticipate amalgamation, causing budgets to deteriorate and service levels to improve in the years before amalgamation. These phenomena are called Ashenfelter dips or peaks (Angrist & Pischke 2008, 182). Moreover, if municipalities are managed poorly, various stakeholders might push for amalgamation. If these
municipalities perform as well as non-amalgamated municipalities after amalgamation, these effects might not be found in our results (Allers & Geertsema 2016, 667). Because of these risks, I include a pre-amalgamation dummy to check for differences before amalgamation.
Aside from the treatment group, which consists of amalgamated municipalities in the years after amalgamation, following Allers & Geertsema (2016), I use three different control groups. The first control group consists of all municipalities that existed on January 1st 2005. This control group is not sufficient. As can be seen from the table and graphs from chapter 4, this group is too different from the treatment group. Difference-in-difference estimation does not require treatment and control groups to be similar; it only requires a common time trend between groups. However, the control variables listed in Table 3 are variables that influence selection for amalgamation. If these differ significantly between treatment and control groups, selection bias might be present. The second control group consists of 145 municipalities that were amalgamated in the research period, but in a different year. In analyses using this control group, only a single difference is measured effectively. I measure the difference between amalgamated municipalities in different periods before and after amalgamation between those municipalities in other years.
The third control group is the same as the second, but following Allers & Geertsema (2016) and Reingewertz (2012), I add 21 municipalities in which amalgamations were prepared but were aborted due to political reasons. Additionally, municipalities which have been amalgamated in 2018 and which are scheduled to amalgamate in 2019 have been added to this control group. This is the preferred control group because the graphs and table of Chapter 4 show it to be more similar to the treatment group than the group of all municipalities. Initial results will be presented using this control group. Ultimately, this control group consists of 196 municipalities. A list of non-amalgamated municipalities added to the control group and the reason for their inclusion is shown in Table A1 in the appendix.
5.1Main model
I start my analysis with the following model: (1) yit = βi + γt + δ1P1 + δ2P2 +δ3P3 + ηAit + εit Where:
• yit is one of the dependent variables reviewed in chapter 4.
• γt describes the year effect and contains a dummy variable for each year except for the first.
• βi is a matrix that contains dummy variables for each municipality except for a reference municipality.
• δ1P1,δ2P2, and δ3P3 describe the effect dummy variables that are ‘1’ for amalgamated municipalities in different periods before and after amalgamation. This will be explained more thoroughly below.
