Gender representation and neutrality: does she make the difference?

(1)

1

Sophie Heikoop s1164570 Thesis MSc Public Administration: Governing Markets Leiden University Supervisor: prof. dr. S.M. Groeneveld Second reader: dr. M. Beerkens Leiden, June 8, 2017

Gender representation and neutrality: does she make the

difference?

A quantitative study on the effect of women’s representation in local politics and

administration on local government expenditures in the Netherlands.

(2)

2

Abstract

The goal of this research is to give an answer to the research question “What is the effect of women’s representation in local politics and administration on local government expenditures in the Netherlands?”. This was done by looking at the effect of gender representation on local public expenditures per inhabitant per year in five policy areas: total expenditures, expenditures on traffic, transport and water management (TTWM), expenditures on education, expenditures on culture and recreation, and expenditures on social security and social services (SSSS). The share of women in politics is measured by the share of women in the Municipal Council, the share of women in the Municipal Executive, and the gender of the Mayor. The share of women in administration is measured by the gender of the Clerk to Council. Two sets of conflicting hypotheses were formulated. One set assuming that the share of women in politics and administration does not have an effect on local government expenditures, and the other set assuming that the share of women in politics and administration does have an effect on local government expenditures.

The results of the OLS regression analyses show that the share of women in administration does not have an effect on local government expenditures, meaning that the variance in local government expenditures is not explained by the gender of the Clerk to Council. This finding is consistent with the first assumption in this research, namely: that the share of women in administration does not have an effect on local government expenditures. With the acceptance of this hypothesis, the underlying assumption of the neutrality of the public official, as formulated by the politics-administration dichotomy, is accepted as well.

The results of the OLS regression analyses on the effect of the share of women in politics on local government expenditures were more ambiguous. The results show that in politics, gender does not have an effect on total expenditures, and neither on expenditures on SSSS. The same applies to the effect of the share of women in the Municipal Executive in all five policy areas. This is in accordance with the first assumption in this research. With the acceptance of this hypothesis, the underlying assumption of the median voter theorem (Downs, 1957) is accepted as well, indicating that background characteristics such as gender do not have an effect on policy outcomes, and therefore, not on expenditures. There are some exceptions however, as the results show that the share of women in the Municipal Council does have an effect on expenditures on TTWM, education, and culture and recreation. Also, the gender of the Mayor has an effect on expenditures on TTWM, and culture and recreation.

The answer to the research question is that the gender of the Clerk to Council does not affect local public expenditures, and in general, gender in politics does not either, except for the share of women in the Municipal Council and the gender of the Mayor in a few policy areas. The implications of these findings will be discussed in this paper.

(5)

5

1. Introduction

It was November 2016, the night of the Presidential elections in the United States of America. It was going to be a historical night: presidential elect Hillary Clinton had a great chance on becoming the first female president of the US. The polls were in favour of Clinton1_{; she was going to make history.} She had chosen to give her winning speech in a building with a glass ceiling. This choice was very symbolic, since if Clinton would win, she would break the glass ceiling.2 Why is this glass ceiling symbolic?

Women have been fighting for equal rights for a long time. Even though they have accomplished a lot in terms of gender equality, the share of women in high positions is low3_compared to men; it is hard for women to move up the career ladder. This is also called ‘the glass ceiling’.4 Their chances on holding an important managerial function are much lower than for men.5_{Today, women’s} representation in top functions is very low, and they often find themselves in the situation of being a ‘token’: as a representative for their category rather than as individuals (Kanter, 1977: 382-383; Barsch et al., 2012).6

To increase gender equality, countries have set (voluntary) gender quota that require a certain number of women in top functions, in the public sector as well as in the private sector, to ensure women’s representativeness.7_{Besides the fact that this will increase gender equality, it will also be} more representative of a countries’ demographic characteristics, because the share of women in the world population is 49.6%.8 This representativeness is important, since it will reflect the preferences of a countries’ population; this will lead to more responsiveness and social support. Social support and responsiveness are crucial in making stable and effective public policy (Krislov, 1974).

In the Dutch government for example, the share of women in high positions in the public sector is 31 percent. They have met the requirement of the gender quota of 30 percent in 2016, two years before the deadline in 2018.9_{Five out of eleven cabinet ministers are women; the Netherlands} are ranked 4th out of 28 European Union countries.10 However, the share of women in high positions in the private sector is very low, and even dropped from 7.8 percent in 2015 to 7.1 percent in 2016 (Joosten et al., 2016: 14). Also, the Netherlands rank below the average of OECD countries in the Glass Ceiling Index, among others due to a large gender wage gap and a short period of paid leave.11

A lot of research that has been done on women’s representation, was focused on the level of representativeness itself. Little research has been done on the effect of this level of representativeness, both in the public and private sector. The research in this thesis for Public Administration will be focussed on the effect of women’s representation on public decision making, both political and administrative. Furthermore, the little research that has been done on this effect, was done on central levels of government, in particular in the United States and Asia (Chen, 2010; Chen, 2013; Mavisakalyan, 2014; Rehavi, 2007). Therefore, this research will be focused on the lowest level of government: the municipalities, and in a European country: the Netherlands. It is important to extend

(6)

6

this research to the level of local government, as decision making on the level of local government is different from, and done by other institutions, in central governments. By extending this research to local levels of government, it is possible to see if the effect that is found at central levels of government, is present at the local government level as well. The effect of women’s representativeness will be measured in terms of its effect on local government expenditures in a municipality.

The fact that women are underrepresented in top functions has been given a lot of attention and there has been done a lot of research on this topic. On the one hand, theories state that an individual’s background does or should not have an effect on their job performance (Downs, 1957; Wilson, 1887; Goodnow, 1900; Weber, 2004). On the other hand, theories state that an individual’s background does have an effect and that this effect is good, as it in sum represents societies’ preferences (Krislov, 1974; Dolan, 2000; 2002). But why is this important and what is the effect of the presence of women in top functions? More specific: what is the effect of the presence of women in local politics and administration on public expenditures? To find this out, the research question of this study is: ‘What is the effect of women’s representation in local politics and administration on local

government expenditures in the Netherlands?’. To examine this effect, the study will include several

typical social welfare policies, such as education, culture and recreation, and social security and social services, and other not-so-typical social welfare policies, such as traffic, transport and water management. The distinction between social welfare policies and other policies is made, because research shows that women and men have different preferences with regard to redistribution policies, such as social welfare (Edlund and Pande, 2002). Moreover, empirical research shows that a higher share of women in politics and administration has a positive effect on public expenditures on social welfare policies (for example, Chen, 2010; Mavisakalyan, 2014; Dolan, 2002). Total public expenditures and expenditures on traffic, transport and water management are included, to see if gender has an effect on the expenditures in these policy areas as well. The study will focus on decision making in the political as well as the administrative domain. Whatever the outcome of this study may be, i.e. whether women do have an effect on public expenditures or not, it brings us further in our knowledge regarding the effects of gender representation and gender neutrality in politics and administration on local levels of government. Hopefully, this study contributes to shifting the attention from research on gender representation in general, towards research on the effects of gender representation in politics and administration.

The quantitative analysis will include fifty Dutch municipalities that will be studied over a period of seven years, from 2009 to 2015. By studying this long period, data shifts due to changes in local government compositions before and after elections are included as well. The local government expenditures consist of five dependent variables, and is measured by the local government expenditures per inhabitant per year in a municipality, divided into the five policy areas. Women’s representation in local politics and administration is measured by four independent variables. For politics, these are the share of women in the Municipal Council, the share of women in the Municipal

(7)

7

Executive, and the gender of the Mayor. For administration, this is the gender of the Clerk to Council. The variables are measured at the interval-ratio level, or have been computed into dummy variables. The quantitative analysis was done through a multiple regression, consisting of three OLS regression models. This research provides a new dataset with new information and insights for political and administrative decision making on the level of local government, in specific the effect of women’s representation in municipalities on local public expenditures. The dataset can be used and expanded for future research.

To be able to give an answer to the research question “What is the effect of women’s representation in local politics and administration on local government expenditures in the Netherlands?”, the next chapter will provide a theoretical framework on gender representation in political and administrative decision making; hypotheses are formulated as well. In Chapter 3, the research context on the Dutch local government will be given. Chapter 4 will provide the methodological framework of this research. In Chapter 5, the results of this research are presented, and in Chapter 6, an answer will be given to the research question. The discussion consists of theoretical implications, methodological implications and limitations, and further research suggestions.

2. Theoretical framework

To be able to give an answer to the research question “What is the effect of women’s representation in local politics and administration on local government expenditures in the Netherlands?” this chapter will provide a theoretical framework. The study includes decision making in the political as well as in the administrative domain. Therefore, in the first section, the median voter theorem will be outlined, and thereafter the theory on political decision making. The first part will be followed up by the formulation of hypotheses on political decision making. The second part of the theoretical framework will outline administrative decision making, starting with the politics-administration dichotomy and followed up by theory on representative bureaucracy. Thereafter, hypotheses on administrative decision making will be formulated.

2.1 Gender representation in politics

The first part of this theoretical framework on gender representation in politics will examine the median voter theorem of Downs (1957). Thereafter, empirical research on political decision making will be outlined.

According to the median voter theorem (Downs, 1957), the gender of a politician does not affect policy outcomes. Assumed that women can vote and that a politician is fully committed to a policy, it should not matter whether the politician is a male or a female, since the voter’s preferences

(8)

8

will be represented by the elected politician. This stems from the fact that in a democracy, where political parties form coalitions after elections, the political parties should create policies that satisfy the broad public that has elected them, to appeal for a later re-election. Therefore, politicians should create policies that reflect the preferences of the broad public, leaving little room for the politician to exert its individual preferences (Downs, 1957: 143-144). Even though these parties each had a distinguished policy idea during the elections, in the end, while they have to govern, their policy is broad and contains a consensus between the coalition parties (Downs, 1957: 143-144). Following this theoretical model implies that gender does not have an influence on policy outcomes. A politician’s background is not decisive in policy making and execution. Their backgrounds are toned down both in their goal to serve the public to win another election, and in the consensus of the coalition. This would lead to the H0, implying that there is no relationship between the share of women in local politics and local government expenditures.

Edlund and Pande (2002) on the other hand, argue that men and women have different preferences with regard to redistribution policies. In their research, Edlund and Pande sought for an explanation for the increased political gender gap in the US. Women tend to have become more supportive of the Democratic party, which is considered to be more left-winged, and men traditionally are more supportive of the Republican party, which is considered to be more right-winged. The results show that the tendency for women to have become more supportive of redistribution policies, comes from their more traditional role within the family. In the traditional family, marriage formed a contract where women promised to take care of the household and their children, and men promised to be the bread-winner of the family, to be able to provide for their wife and children. In this sense, married women are less likely to support the Democratic party. However, recently women have become more left-winged (i.e. supportive of the Democratic party). Edlund and Pande (2002) show that the traditional role of the woman in the family has changed as a result of the increased divorce-rate. Moreover, it has become more common to marry later, or to cohabit. These circumstances have changed the traditional role of women in the family, in the sense that men are no longer the (only) breadwinner, but women have become breadwinners too. Because women now have to provide for their family, they favour redistribution policies, such as social welfare, education, and health. Thus, the study showed that women have become more likely to support the Democratic party.

Empirical research on political decision making supports these findings. Empirical research shows that men and women have different preferences, and this has an effect on political decision making and policy outcomes, especially in the social welfare area. Clots-Figueras (2007) in her study on Indian female politicians for example, found evidence that the political representation of women has influence on an individual’s education. Clots-Figueras controlled for political party preference, to estimate the effect of gender on education expenditures. The results of this study show that the gender of a politician does affect policy outcomes. Female politicians prioritize spending on education more than male politicians, leading to higher public expenditures on education. Furthermore, this effect is

(9)

9

larger when women are in power for a longer period, and when the share of women increases. However, this effect of the influence of each individual woman decreases as the number of women increases, but for the share of women in politics this result is significant. These findings are supported by a study of Chen (2009), where she also finds that an increase of female legislators leads to an increase in educational public expenditures. She, however, finds that political party preference does not necessarily influence public expenditures on education. Even though left-winged governments tend to spend more money on education policy, this effect is not statistically significant (Chen, 2009: 10-11).

By looking at the preferences of women and men in American politics, Carroll (2013) showed that women and men have different preferences. Female politicians tend to give priority to social welfare policy areas such as health, education, and children more often than male politicians. Since female politicians prioritize higher spending on social welfare policy areas, an increase in the representation of women will lead to higher public spending on these policies. Chen (2010) has studied the effect of gender quotas on women’s representation and policy decisions. Her study shows that gender quotas do have an influence on women’s representation, in the sense that the level of representativeness of women in politics has increased after countries have set gender quotas (Chen, 2010: 21). She also finds that gender quotas lead to an increase in government expenditures on policy areas such as health, education, and social welfare. This suggests that the increased level of women’s representation, because of gender quotas, leads to higher public expenditures on socio-economic policy areas, and this is confirmed by the results in her study. Furthermore, Chen (2010) studied the effect of gender quotas on policy outcomes. The results show that the increased level of representativeness of women in politics have a positive influence on health, education, and social welfare. This is supported by later research of Chen (2013), where the influence of female politicians on public spending in Taiwan is investigated. The study showed that there is a statistically significant positive correlation between the number of female politicians and expenditures on social welfare. However, Chen (2010, 22-23) also finds that the opposite is true for non-social welfare policies, such as defense. As expenditures have to remain within the scope of the budget, and because women prefer higher spending on social welfare policies, this comes at the expense of the fiscal space of non-social welfare policies (Chen, 2010: 19-20). Thus, a higher share of women in politics leads to lower local government expenditures on non-social welfare policies.

Mavisakalyan (2014) has examined the relationship between women’s representation in cabinet and public health spending, in a study of 80 countries in the year 2000. The study showed that there is a positive relationship between the two. The higher the share of women in cabinets, the higher the share of public expenditures on health in GDP. Other studies (for example Rehavi, 2007) have shown that a higher share of women in parliament has an influence on public health expenditures, but this study has shown that this effect is even more present if women have a position in the cabinet, such as a minister. This study indicated that a larger share of women in cabinet leads to higher public

(10)

10

expenditures on health. Thus, research shows that there seems to be a positive relationship between the share of women in politics and public expenditures on social welfare policies.

2.1.1 Hypotheses

The political institution of a municipality consists of the Municipal Council, the Municipal Executive, and the Mayor, whom is part of the Municipal Executive. The Municipal Council creates policy and is elected every four years. The Municipal Executive consists of the Mayor and Aldermen. The Municipal Executive executes the policy of the Municipal Council and is appointed by the Municipal Council every four years.

The abovementioned theories on gender representation in politics lead to conflicting hypotheses. On the one hand, according to the median voter theorem, someone’s background should not affect policy outcomes (Downs, 1957) and therefore, it is assumed that gender does not influence public expenditures. This theory leads to the H0, stating that there is no relationship between the share of women in politics and local government expenditures per inhabitant per year. On the other hand however, according to Edlund and Pande (2002), women and men do have different preferences with regard to redistribution policies. The traditional role of women in the family has changed, and therefore, women have become more left-winged, and supportive of policies in the social welfare area. Empirical research on political decision making supports this gender gap in preferences, in the sense that background characteristics such as gender, do affect policy outcomes. The effect of the representativeness of women in politics on policy outcomes is especially present in social welfare policies, such as social services and education (Clots-Figueras, 2007; Chen, 2009; Carroll, 2013; Mavisakalyan, 2014). This leads to the formulation of H1, H1b, H1c, and H1d, indicating that as the share of women in politics increases, local government expenditures on social welfare policy increase as well. The opposite is true for expenditures in other, non-social welfare policy areas, such as defense (Chen, 2010). This leads to the formulation of H1a. The political decision making theory leads to the following hypotheses:

Table 1: Hypotheses on gender representation in politics

H0

There is no relationship between the share of women in local politics and administration and local government expenditures per inhabitant per year.

H1 The higher the share of women in local politics, the higher the total local government expenditures per _{inhabitant per year.}

H1a

The higher the share of women in local politics, the lower the local government expenditures on traffic, transport and water management per inhabitant per year.

H1b

The higher the share of women in local politics, the higher the local government expenditures on education per inhabitant per year.

(11)

11

H1c

The higher the share of women in local politics, the higher the local government expenditures on culture and recreation per inhabitant per year.

H1d

The higher the share of women in local politics, the higher the local government expenditures on social security and social services per inhabitant per year.

2.2 Gender representation in administration

The first part of the theoretical framework focused on the effect of gender representation in politics on public expenditures. The second part of the theoretical framework will examine administrative decision making, with the politics-administration dichotomy and representative bureaucracy theory. The normative concept of representative bureaucracy assumes that, in contrast to the politics-administration dichotomy, background characteristics are reflected in people’s decision making. In the empirical scientific concept, it is assumed that there is representation, because there is no such thing as pure neutrality. The neutral bureaucracy is only an ideal-type.

Wilson (1887: 22-23) in ‘The Study of Administration’, made a strict distinction between politics and administration. This call for a distinction is a consequence of the problems with the spoils system in the US. After every election, the whole public service was replaced by new politicians and administrators. In general, it were the ones who were most loyal to the winning candidate who got a position in the public service. The result was that the public service was highly inefficient and corrupt (Bovens et al., 2007: 304). A distinction between politics and administration had to solve this. Politics sets the tasks for administration; it formulates the preferences and goals of the government into policy. Administration is the factual execution of this policy. Politics should not engage in the way the administration executes policy. For the administration to be efficient and neutral, it should reflect a business-like model.

Goodnow plead for a distinction between politics and administration as well. According to Goodnow (1900: 29-30), there are three kinds of authorities: judicial, executive, and administrative authorities. And there are two functions of government: politics and administration. Politics has to do with the expression of the will of the state, and administration has to do with the execution of that will. Wilson and Goodnow both wanted to take ‘politics out of administration’ (Overeem, 2005: 316).

Another supporter of the politics-administration dichotomy was Weber. However, his reasons for a politics-administration dichotomy were different from Wilson and Goodnow’s. Weber states that politics is about leadership, preferences, and decision making. Administration, on the other hand, is about passivity, objectivity, and being able to efface oneself (Bovens et al., 2007: 305-307). Therefore, politics as a vocation is essentially different from administration. The genuine official should abstain himself from party politics. He should not be politically active, but administer impartial. In doing its job, an official must carry out the orders of the politician in accordance with the politician’s

(12)

12

convictions. Even though the official does not agree with this conviction, he must act as if he were. This self-denial and discipline requires a high degree of ethics, and the official must feel honoured to do this. A politician’s honour lies in carrying and taking responsibility for his own and his official’s actions (Weber, 2004: 53-54). Since the administration is highly efficient, Weber warns for the political primacy to be turned into an administrative primacy if the administration gets too involved with political decision-making (Bovens et al., 2007: 305-307). Thus, in contrast to Wilson and Goodnow, Weber wants ‘administration to stay out of politics’ (Overeem, 2005: 316).

Wilson, Goodnow and Weber’s perceptions about the politics-administration dichotomy do have something in common: their perception on the political neutrality of the public service. They wanted to keep partisan politics apart from administration, not politics in general. Administration had to be impartial and stay outside the political arena (Overeem, 2005: 317).

The representative bureaucracy concept was at first used to get rid of the ‘elite’-image of government. Nowadays, society tends to care less about class, and more about issues of race, ethnicity, and gender (Krislov, 1974: 20). To make governments and its policies more effective, governance must gain societal support. One way to accomplish this support, is to reflect a wide segment of society into policy, i.e. increasing representativeness (Krislov, 1974: 4-5). By increasing the bureaucracies’ representativeness, it becomes more socially stable and more responsive (Krislov: 81). Representativeness leads to more responsiveness (Krislov: 126) and this representational participation should lead to functional effectiveness (Krislov: 129).

In contrast to the politics-administration dichotomy, the normative concept of representative bureaucracy theory does not think of a strict distinction between politics and administration. It assumes that individuals are influenced by their own preferences and experiences in making decisions. This means that individuals are not neutral in their decisions. They are affected by their own values and interpretations (Dolan, 2002: 354-355). A bureaucracy will be representative if it is constituted of individuals that reflect the same diversity of race, gender, etc. as society. Furthermore, these individuals should share the same values and preferences as the public. Public policy should reflect this diversity (Dolan, 2000: 515). The diversity within public service matters, since representative bureaucracy theory assumes that because men and women have different life experiences, they make decisions differently, and in a way that favours their category. It is assumed therefore, that if more women are present in administrative positions, there will be produced policy that reflects women’s needs and preferences more (Dolan, 2000: 514).

It is assumed therefore, that administrative decision making is influenced by gender. The assumption is twofold. On the one hand, public administrators will press for the interests and preferences of the social group they represent, if the issue is of great importance to that group. On the other hand, the possibility of this active representation is influenced by the level of concentration of group members in the organization (Dolan, 2000: 517). The lower this concentration is, the less these public administrators feel supported in their preferences, and the more likely that they will adapt to the

(13)

13

dominating opinion. In this sense, they will behave as ‘tokens’ (Kanter, 1977). The opposite is true if there is a higher concentration of group members in the organization, which will lead public administrators to more actively represent their social group (Dolan, 2000: 517). Furthermore, active representation is more likely when the environment is politically supportive, and when the values of the public administrator and the mission of their organization correspond (Dolan, 2000: 518). Dolan finds that there is a linear relationship between the proportion of a social group and active representation. That is, as the percentage of women in a certain department increases, the support for women-related issues increases as well (Dolan, 2000: 523).

Representative bureaucracy theory makes a distinction between active and passive representation to explain how bureaucratic diversity shapes policy (Dolan, 2000: 515-516). Passive representation means that individuals mirror the total society. This has to do with who they are; their origins. Active representation is related to what individuals do. It has to do with the assumption that individuals actively press their preferences during decision-making, on behalf of their social groups whom they represent (Dolan, 2000: 516). As with everyone, when bureaucrats are faced with choices, they tend to make the decision that is closest to their values. Therefore, if the bureaucracy reflects the same diversity as society, it will create policy that reflects societies’ preferences (Dolan, 2000: 516).

In another study, Julie Dolan (2002) examines the influence of women’s representation in administrative decision making in terms of budgeting. In general, she does this by comparing representative bureaucracy theory with organisational socialization theory. Representative bureaucracy theory assumes that individuals are influenced by their own perceptions and experiences in taking (policy) decisions. According to representative bureaucracy theory therefore, it is assumed that women make different decisions than men, since their preferences differ (Dolan, 2002: 354). Organisational socialization theory however, assumes that the decisions of men and women do not diverge that much, since common working experience make workers act upon the organizational values and goals and therefore, differences between sexes are minimal (Dolan, 2002: 354).

Even though her study finds some supporting evidence for organizational socialization theory, these effects are altered by the fact that if organizational socialization were at work, the policy preferences of male and female officials would be more similar. However, the policy preferences of the male and female officials are as divergent as in the general public (Dolan, 2002: 367). Overall, women whom hold an administrative function in office, tend to prefer larger spending on social welfare policies, such as childcare, health care, and welfare, than do their male colleagues (Dolan, 2002: 367). This corresponds to the spending preferences of the general public. However, there is an exception for women in the Defense department. These women appear to have the same spending pattern as their male colleagues, which is lower than the spending preferences of the general public (Dolan, 2002: 368-370). Even though organisational socialization processes can be accountable for this outcome, it appears to have more to do with critical mass theory. Women in the Defense department have not reached a critical mass, that is, they are what Kanter (1977) calls ‘tokens’.

(14)

14

Women are in such a minority, that they stand out in the crowd. To divert attention from themselves and by trying to fit in, women adapt their preferences to the general standard, that is, to male preferences. Thus, the fact that gender differences are small in the Defense department can be explained by critical mass theory (Dolan, 2002: 370).

Furthermore, it is stated that bureaucrats have become more involved in the policy making process. Technological developments have resulted in the increasing complexity of public problems and therefore, solving these problems has become more technical and time-consuming (Aberbach et al., 1981: 111). The technical skills and expertise of bureaucrats have become necessary in solving these problems and providing input in creating public policy. Since bureaucrats have become more involved in the process of policy making, they have become more political. Therefore, besides merely executing policy, bureaucrats do influence the process of decision making. Politicians provide the general guidelines for policy to appeal to the broad public, and it is the task of bureaucrats to give a concrete expression to these generalities and to focus on the technical and administrative aspects of the implementation of policies (Aberbach et al., 1981: 112-113).

2.2.1 Hypotheses

The administrative board of a municipality is measured by the Clerk to Council. The Clerk to Council is the general manager of the municipality and the head of the Civil Service. The Clerk to Council is appointed by the Municipal Executive.

The abovementioned theories on gender representation in administration lead to conflicting hypotheses. On the one hand, according to Wilson (1887), Goodnow (1900), and Weber (2004), there should be a strict distinction between politics and administration. Even though Wilson and Goodnow wanted politics out of administration, and Weber wanted administration to stay out of politics, their perceptions about the politics-administration dichotomy do have something in common: their perception on the political neutrality of the public service. They wanted to keep partisan politics apart from administration, not politics in general. Administration had to be impartial and stay outside the political arena (Overeem, 2005: 317). With the politics-administration dichotomy it is assumed that the share of women in administration does not have an effect on the level of public expenditures in a municipality. This leads to the H0 for administrative decision making.

On the other hand, however, according to representative bureaucracy theory, there is no strict distinction between political and administrative decision making. Public officials are becoming more involved in the process of creating policy and problem-solving (Aberbach et al., 1981). According to representative bureaucracy theory, public officials will try to actively press their preferences, on behalf of the social group they represent (Dolan, 2000). A representative bureaucracy will therefore reflect the same preferences as public preferences (Krislov, 1974). Female spending preferences are different from male spending preferences, in the sense that women prefer higher spending on social welfare

(15)

15

policies (Dolan, 2002). On defense policy however, women prefer less spending (Dolan, 2002). Since female spending preferences are different from male spending preferences, both for public officials as in society, it is assumed that gender does affect government expenditures. Also, this effect is more present as the share of women increases (Dolan, 2002). Therefore, the representative bureaucracy theory leads to the formulation of H2-H2d:

Table 2: Hypotheses on gender representation in administration

H0

There is no relationship between the share of women in local politics and administration and local government expenditures per inhabitant per year.

H2

The higher the share of women in local administration, the higher the total local government expenditures per inhabitant per year.

H2a The higher the share of women in local administration, the lower the local government expenditures on _{traffic, transport and water management per inhabitant per year.}

H2b

The higher the share of women in local administration, the higher the local government expenditures on education per inhabitant per year.

H2c

The higher the share of women in local administration, the higher the local government expenditures on culture and recreation per inhabitant per year.

H2d

The higher the share of women in local administration, the higher the local government expenditures on social security and social services per inhabitant per year.

3 Research context: Dutch local government

The Netherlands consist of multiple administrative layers. From top to bottom there is the European Union, then the central government, then the provinces, the regional water authorities, and finally the municipalities. The Dutch state has a ‘trias politica’, which means that the legislative, executive and jurisprudential powers are separated. On the level of central government, the legislative power consists of the cabinet and the States General.12_{On the level of local government, these are the Provincial} Council and the Municipal Council.13 On the level of central government, the executive powers consists of the cabinet (the King of the Netherlands and the ministers)14_{, and on the local level of the} Province of the County Council and the Royal Commissioner15, and in the municipality of the Municipal Executive and the Clerk to Council.16 The jurisprudential power consists of independent judges.17

The Netherlands are a decentralized Unitarian State. The concept of a decentralized Unitarian State is based on the fact that on the one hand, the Netherlands are decentralized, since certain administrative tasks can be given or assigned from the central government to lower levels of

(16)

16

government, such as the Provinces or the Municipality (Breeman et al., 2008: 17). On the other hand, the Netherlands are a Unitarian State, since legislation and regulation are related to each other, and because the legislation and regulation that are created by lower levels of government, cannot conflict with national legislation and regulation (Breeman et al., 2008: 27).

In 2017, the Netherlands consist of 388 municipalities, spread over twelve Provinces.18_This study is focused on fifty municipalities in the Netherlands. Regarding the decision making in municipalities, this study focuses on two ways of decision making: political decision making and administrative decision making. In a municipality, the political institution consists of the Municipal Council and the Municipal Executive.19_{The Municipal Council and the Municipal Executive have the} task of political decision making. The Municipal Council’s primary tasks are defining the framework for policies and controlling the executive committee, the Municipal Executive. The tasks of the Municipal Executive include the specific filling in of the policies and the implementation and execution thereof. The Municipal Council must monitor the Municipal Executive to control if the execution of this policy is in accordance with their defined framework.20_{The administrative institution} consists of the Clerk to Council and the public officials. The Clerk to Council is the head of the administration in a municipality21_{and is considered neutral, i.e. this person does not have a political} party affiliation. In this research, only the Clerk to Council is included in administrative decision making, regarding the time and scope that is available for this study.

The Municipal Council is elected every four years by the inhabitants of the municipality. In the Netherlands, the electoral system is based on proportional representation. The seats in the political institution are allocated based on the number of votes that have been obtained during the elections. This means that multiple parties have to cooperate and join a coalition to govern. Because of this cooperation, compromises have to be made regarding policy decisions.22_{The Municipal Council is} elected every four years by the inhabitants of the municipality. Because of the electoral system, the number of votes is decisive for the allocation of the Council Members. Based on the electoral results, the Municipal Council appoints the Municipal Executive, whom form a coalition for four years.23_In these political institutions, the allocation of seats is determinant for the policy of the municipality, regarding political party preference. The ratio of left-winged and right-winged parties is determinant for the direction of the policy, as it is assumed that a left-winged coalition prioritises higher spending on social welfare policies more than right-winged coalitions. Also, left-winged parties are supportive of additional government intervention, which leads to higher expenditures, whereas right-winged parties are supportive of relying on market mechanisms (Kelly-Gagnon and Geloso, 2013: 1). Therefore, political party preference is included in this research as a control variable, to be able to find the effect of the share of women in politics and administration on public expenditures per inhabitant in a municipality.

(17)

17

4. Methods

The goal of this research is to see if there is a difference in local public expenditures in the Netherlands, and whether this difference can be explained by the share of women (i.e. gender), or by other causal inferences, such as a change in political party preference. Therefore, fifty municipalities of the Netherlands will be studied over a period of seven years. In this chapter, the case selection will be outlined in paragraph 4.1. In paragraph 4.2 the data collection will be described. In paragraph 4.3, the dependent, independent, and control variables will be operationalized. In the last paragraph, 4.4, attention will be given to the validity and reliability of this research.

4.1 Case selection

In 2017, the Netherlands consist of 388 municipalities, spread over twelve Provinces.24_{Regarding the} time and scope of this research, it is impossible to study all 388 municipalities. Instead of randomly selecting fifty municipalities, the fifty largest municipalities are selected. This was done because it is expected that the fifty largest municipalities are most similar regarding demographic characteristics, and therefore better in terms of comparability. Even though the fifty largest municipalities are not representative for all 388 municipalities in the Netherlands, the geographical distribution of the selected cases in Figure 1 shows that every part of the Netherlands is represented. However, carefulness is presumed when making generalizations.

Figure 1: Distribution of the selected cases

(18)

18

Source: CBS (2014)

The fifty largest municipalities are selected from the list of 388 municipalities on the size of the municipality (CBS, 2014). The selected cases are presented in Table 3. The size of the municipality is selected on the number of inhabitants in a year according to the Central Bureau of Statistics in the Netherlands. The number of inhabitants is determined by the number of inhabitants that are registered in the municipal administration (CBS, 2014). The number of inhabitants is also being used to determine the relative local government expenditures i.e. the expenditures per inhabitant per year, and at the same time to control for a change in expenditures because of population growth.

Table 3: Case selection – the fifty largest municipalities of the Netherlands in 2009

Nr. Municipality Inhabitants Nr. Municipality Inhabitants

1 Amsterdam 755605 26 Westland 99436

2 Rotterdam 587134 27 Deventer 97892

3 's-Gravenhage (gemeente) 481864 28 Delft 96517

4 Utrecht (gemeente) 299891 29 Sittard-Geleen 95327

5 Eindhoven 212269 30 Leeuwarden 93498

6 Tilburg 203464 31 Alkmaar 93416

7 Almere 185746 32 Venlo 91467

8 Groningen (gemeente) 184227 33 Heerlen 89356

9 Breda 171916 34 Helmond 87757

10 Nijmegen 161817 35 Hilversum 84422

11 Enschede 156071 36 Hengelo (O.) 80925

12 Apeldoorn 155332 37 Amstelveen 79768 13 Haarlem 148191 38 Purmerend 78862 14 Arnhem 145574 39 Roosendaal 77482 15 Zaanstad 144055 40 Oss 77097 16 Amersfoort 143212 41 Schiedam 75326 17 Haarlemmermeer 142042 42 Lelystad 73848 18 's-Hertogenbosch 137775 43 Leidschendam-Voorburg 72697 19 Zoetermeer 120881 44 Almelo 72428

20 Dordrecht 118408 45 Alphen aan den Rijn 72178

21 Maastricht 118286 46 Gouda 70828

22 Zwolle 117703 47 Vlaardingen 70433

(19)

19

24 Emmen 109441 49 Velsen* 67528 25 Ede 107623 50 Assen** 66369 Source: CBS (2014)

The fifty municipalities in Table 3 are the municipalities that will be studied in the period of 2009-2015. From 2009 to 2015, 48 of the 50 selected cases belonged to the fifty largest municipalities (CBS, 2014). As from 2011, the municipalities Súdwest-Fryslân and Spijkenisse were one of the fifty largest municipalities, and the municipalities Velsen* and Assen** no longer were (CBS, 2014). This is the result of the recent redivision that has taken place in multiple local governments in the Netherlands.25_{However, following this selection would cause comparability problems and therefore,} the chance on bias would increase. Therefore, the choice was made to exclude Súdwest-Fryslân and Spijkenisse from the selection, and keep Velsen and Assen.

The period of 2009-2015 was chosen for several reasons. First, every four years, elections are being held for the Municipal Council. The period of 2009 to 2015 contains one full electoral period, from March 2010 to March 2014. Second, the years 2009 and 2015 are respectively one year before and after the electoral period. By including these years, changes in data due to electoral outcomes (regarding the share of women in politics) can be detected. By this, causality can be explained better, as the differences between years can be taken into account. This is important, because of potential causality to be observed. Third, the period of 2009-2015 is the most recent period regarding data availability. Finally, by including fifty municipalities over seven years, it is possible to study the variance between the municipalities, as well as over time.

To be able to examine the association between the share of women in politics and administration and local public expenditures, control variables are included in this research as well. For this study, control variables such as political party preference, the income in a policy area, and the year are taken into account. These control variables serve as alternative explanations26, to prevent observing a non-existent causality. For example, municipality X has a large increase in total public expenditures in Year X, and the number of women in politics has increased. This indicates that H1 is true, namely that a higher share of women in politics leads to higher total expenditures. However, the coalition in municipality X has changed from right-winged to left-winged. Therefore, the increase in expenditures may be due to a change in political party preferences instead of gender (women’s representation in politics). By controlling for political party preferences, it is prevented that wrong conclusions are drawn, as described in the example above.

4.2 Data collection

The fifty municipalities have been selected on the number of inhabitants in a year. The data on the number of inhabitants was already collected by the CBS, the Central Bureau of Statistics. These

(20)

20

numbers have been used for the case selection (CBS, 2014). The same was done for the dependent variables. The CBS created a dataset in which the expenditures and income of a municipality in a year were presented (CBS, 2015). It included the total income and expenditures as well as for the four policy areas in this study (traffic, transport and water management, education, culture and recreation, and social security and social services) (Rijksoverheid, 2014). Due to reasons of uniformity, comparability and limited time, these existing data are being used in this research. To prevent selection bias, these data on expenditures have been controlled by randomly picking a year, an expenditure, and a municipality, and comparing the data of the CBS with the annual financial statement of that municipality in that year. To find the missing data, another dataset of CBS (2016) was used to find corresponding results, and municipalities were contacted by phone or email to provide for the data.

Municipalities only report the composition of the Municipal Council and the Municipal Executive for the current electoral period of 2014-2018, and not for the periods of 2006-2010 and 2010-2014. To acquire this data, the annual financial statements of a municipality in a year were consulted. However, most of the times the composition of these boards were not reported either. Therefore, the website of the Overheid in Nederland was consulted.27_{It reports for most individual} Aldermen in what municipality they were Aldermen and during which period. Especially for 2009 there was a lack of data availability. But when two or three names were given, putting these names in Google search and adding the name of the municipality most of the times gave a hit with the full composition in a document of the municipality, or an article of a local newspaper. This data gave insight in both the gender composition and the political party preference of the Municipal Executive.

The allocation of seats of the Municipal Council in a year was found in the database of the Kiesraad.28_{After collecting the number of seats per political party, these numbers were grouped into} ‘left-winged’, ‘right-winged’, or ‘unknown’ (see §4.3 Operationalization variables). Unfortunately, the websites of the municipalities only report the compositions of the Municipal Council of the current electoral period 2014-2018. To find the missing data for electoral period 2006-2010 (2009) and 2010-2014, the web archive was used or the municipalities were contacted by phone or email to provide for the data.29

The central government reported the current name, gender, and political party of the mayor in a municipality in 2016.30 Also, it reported the date of appointment. Most of the mayors were appointed before 2009, so the dataset provided most of the data needed. The missing data was found by searching on Google for its predecessors or on http://overheidinnederland.nl.

The gender of the Clerk to Council was found by looking at each municipalities’ website. Most of the times it reported the date of appointment as well. If not, this was found by searching on Google, the annual financial statement of a municipality in a year, or on http://overheidinnederland.nl. The Clerk to Council is a public official, whose party preference should not play a role in his work.

(21)

21

However, if their political party preference is known, it is mentioned in the data. If known, it was found on their personal LinkedIn-page or via Google.

The dataset includes 350 cases, as fifty municipalities have been studied over a period of seven years (N=350). For every case (i.e. a municipality in a year), the dataset contains the number of inhabitants, the total expenditures and income, and the expenditures and income on traffic, transport and water management, on education, on culture and recreation, and on social security and social services. Expenditures and income per inhabitant in these policy areas are included as well, by dividing the expenditures and income with the number of inhabitants. The valid N is 349, since Alkmaar did not provide for the missing data on the expenditures and income in the five policy areas in 2015. Furthermore, the dataset contains the number of women in the MC, and the ME, and the number of men in the MC, and the ME. This lead to the computation of the independent variables for the share of women in the MC, and the ME. The gender of the Mayor and the Clerk to Council are included as well. Also, the political party preference (left, right, other, or unknown) of the MC, the ME, the Mayor, and the CtC is included. This lead to the computation of the control variables on the share of left political party in the MC, and the ME, and the dummy variables on the political party of the Mayor, and the CtC.

4.3 Operationalization

To be able to answer the research question ‘What is the effect of women’s representation in local politics and administration on local government expenditures in the Netherlands?’, the concepts must be operationalized and defined clearly. In this paragraph, the independent, dependent, and control variables are operationalized. Table 4 shows the descriptive statistics of the independent (1-4), dependent (5-9), and control variables (10-19).

Independent variables: share of women in politics

• The share of women in local politics: Is measured by the share of women in the Municipal Council, the share of women in the Municipal Executive, and the gender of the Mayor.

• (1) The share of women in the Municipal Council: The Municipal Council is, together with the Municipal Executive, the political institution of a municipality in the Netherlands. Its job is to translate their political goals into policy and to control the Municipal Executive on the proper execution of their policy. Its members are elected every four years. The independent variable ‘the share of women in the Municipal Council’ was created by dividing the number of women in the Municipal Council by the total number of members in the Municipal Council. This is an interval-ratio level of measurement and the share is expressed in a interval-ratio.

(22)

22

• (2) The share of women in the Municipal Executive: The Municipal Executive consists of the Mayor and Aldermen. The Mayor is appointed by the Municipal Council for a period of six years. The Aldermen are being appointed by the Municipal Council every four year. Their tasks are to execute the policy and decisions of the Municipal Council. The Mayor and Aldermen each have their own policy domain, and they must work in the interest of their domain. The independent variable ‘the share of women in the Municipal Executive’ was created by dividing the number of women in the Municipal Executive by the total number of members in the Municipal Executive. This is an interval-ratio level of measurement and the share is expressed in a ratio.

• (3) Gender of the Mayor: The Mayor is appointed by the Municipal Council for a period of six years. The Mayor is the chairman of the Municipal Council and the Municipal Executive, and must guard the internal and external quality and safety of the municipality. The share of women as a mayor did not have to be computed separately, since a municipality only has one mayor. The gender of the mayor was reported as ‘0’ for a man, and ‘1’ for a woman. As table 4 reports, on average, 14 out of 100 mayors are female.

Independent variable: share of women in administration

• (4) The gender of the Clerk to Council: The share of women in local administration in a municipality in a year is measured by the gender of the Clerk to Council. In this study, the administration of a municipality is measured by the Clerk to Council; public officials are excluded from this study. The Clerk to Council is the general manager of the municipality and the head of the civil service. The Clerk to Council is appointed by the Municipal Executive.

The ‘share of women in local administration’ is measured by the gender of the Clerk to Council. Since a municipality only has one Clerk to Council, this variable did not have to be computed separately to report the share of women. The gender of the Clerk to Council was reported as ‘0’ for a man, and as ‘1’ for a woman. As table 4 reports, on average, 27 out of 100 Clerks to Council are female.

Dependent variables: local government expenditures per inhabitant in a municipality in a year

• (5) Total local government expenditures per inhabitant in a municipality in a year: The total of all local costs per inhabitant for all functions, after allocation of the result in a municipality in a year. The allocation of the result is adding or withdrawing amounts from the reserves.

• (6) Local government expenditures on traffic, transport and water management per inhabitant in a

municipality in a year: The local public expenses per inhabitant in a municipality in a year related to roads and traffic, public transport, parking and ports, aviation and water management.

(23)

23

• (7) Local government expenditures on education per inhabitant in a municipality in a year: The local public expenses per inhabitant in a municipality in a year related to education and the accommodation thereof. Education includes primary education, special (secondary) education, secondary education, and adult education.

• (8) Local government expenditures on culture and recreation per inhabitant in a municipality in a

year: The local public expenses per inhabitant in a municipality in a year related to recreation and

nature, sports and arts, libraries and other leisure facilities.

• (9) Local government expenditures on social security and social services per inhabitant in a

municipality in a year: The local public expenses per inhabitant in a municipality in a year related

to social security and social services. This includes income support, employment opportunities, youth and Wmo, and refugees, shelters and child care. Wmo stands for ‘Wet Maatschappelijke Ondersteuning’.

The dependent variables on the expenditures per inhabitant have been computed by dividing the expenditures in a municipality in a year by the number of inhabitants in a municipality in a year. This was done to be able to compare the expenses of each municipality. The expenses per inhabitant are more informative than the total expenses, since a larger municipality in terms of inhabitants has larger expenditures (absolute) than a smaller municipality. By calculating the expenditures per inhabitant, the information on the relative expenses is provided. This resulted in five new dependent variables: ‘total expenditures per inhabitant’, ‘traffic, transport, and water management expenditures per inhabitant’, ‘education expenditures per inhabitant’, ‘culture and recreation expenditures per inhabitant’, and ‘social security and social services expenditures per inhabitant’. These variables are at the interval-ratio level of measurement and the data is expressed in Euro’s.

Control variables: political party preference

A left-winged political party is a political party with a party ideology that is considered left-winged. Left-winged political parties tend to spend more on social welfare policies such as health, since these parties prefer a large welfare state. They are economically left-oriented and social-culturally progressive. Political parties that are considered left-winged include: SP, PvdD, 50-plus, GroenLinks, and PvdA.31 A right-winged political party is a political party with a party ideology that is considered right-winged. Right-winged political parties tend to spend less on social welfare policies than left-winged parties, since these parties prefer a smaller welfare state. They are economically right-oriented and social-culturally conservative. Political parties that are considered right-winged include: VNL, VVD, CDA, SGP, ChristenUnie, D66, and PVV.32 An other political party is a political party other than left- or right-winged. These are local political parties that are not classified in the common ‘left’ or ‘right’ category. An unknown political party means that there is no data available on the political party preference, i.e. ‘neutral’.

(24)

24

• (10) Share of left political party in the Municipal Council: The share of left political party in the Municipal Council is about the political party preference of the members of the Municipal Council, classified as either left- or right-winged, or other. It is measured how left-winged the coalition is, where left-winged is classified as ‘1’, and right-winged or other as ‘0’. This is a continuous variable, as there are multiple seats in the Municipal Council. The closer the value is to 1, the more left-winged a coalition is, and vice versa.

• (11) Share of left political party in the Municipal Executive: The share of left political party in the Municipal Executive is about the political party preference of the members of the Municipal Executive, classified as either left- or right-winged, or other. It is measured how left-winged the coalition is, where left-winged is classified as ‘1’, and right-winged or other as ‘0’. This is a continuous variable, as there are multiple seats in the Municipal Executive. The closer the value is to 1, the more left-winged a coalition is, and vice versa.

• (12) Political party preference of the Mayor: The political party preference of the Mayor in a municipality in a year, classified as either left- or right-winged, or other. Left-winged is classified as ‘1’, and right-winged or other as ‘0’. There is only one Mayor in a municipality. This is a dummy variable.

• (13) Political party preference in Administration: The political party preference of the Clerk to Council in a municipality in a year. This preference is classified as either left- or right-winged, other, or unknown. Unknown is added as a category, since a Clerk to Council is considered not to be affiliated to a political party. Therefore, it can also be read as ‘neutral’. Left-winged is classified as ‘1’, and right-winged, other, or unknown as ‘0’. This is a dummy variable, as there is only one Clerk to Council in a municipality.

Control variables: local government income per inhabitant in a municipality in a year

• (14) Total local government income per inhabitant in a municipality in a year: The total of all benefits per inhabitant for all functions, after allocation of the result in a municipality in a year. The allocation of the result is adding or withdrawing amounts from reserves.

• (15) Local government income on traffic, transport and water management per inhabitant in a

municipality in a year: The local public income per inhabitant in a municipality in a year related to

roads and traffic, public transport, parking and ports, aviation and water management.

• (16) Local government income on education per inhabitant in a municipality in a year: The local public income per inhabitant in a municipality in a year related to education and the accommodation thereof. Education includes primary education, special (secondary) education, secondary education, and adult education.

Gender representation and neutrality: does she make the difference?

1