Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
The relationship between political fragmentation, measured by the number and size of parties, and municipal indebtedness, as well as its effect on voter turnout. The document also discusses the empirical evidence from various studies and provides examples using hypothetical election scenarios.
What you will learn
Typology: Study notes
1 / 195
This page cannot be seen from the preview
Don't miss anything!
Acknowledgements
“I am like a piece of land that of itself is dry and barren, but if you scatter manureover it and cultivate it, it will bear good fruit. By this I mean to say that your Grace’s conversation is the manure that has been cast upon the barren land of mydry wit; the time I spend in your service (…) does the cultivating; and as a result of it all, I hope to bring forth blessed fruits.” (Don Quichote, part II, Ch. XII)
In the course of the past four years, I have built up a lot of debts. Consequently, I gladly take this opportunity to express my sincere gratitude to those who provided answers to my questions or, equally important, confronted me with questions about what I thought were answers. Hopefully, I’ll be able to return the favour(s) some day.
First and foremost, I wish to thank my advisor Bruno Heyndels for his incessant attempts to turn the rough land into a garden-ground. Throughout my period as a doctoral student, his ever-open door (literally and figuratively) as well as the inspiration and opportunities he provided have been a great benefit to me. The same holds for the other members of my Dissertation Committee: Kris Deschouwer and Bruno De Borger. Building on their own fields of expertise they provided important feedback at various stages of the research, which brought an interesting (and necessary) change of perspective.
This work has also benefited greatly from incisive comments provided at different points in time and space by Matthias Benz, Lars Feld, Norman Gemmell, Marc Hooghe, Patrick Stouthuysen, Erik Soetens, Per Tovmo and my colleagues at the department of microeconomics at the Vrije Universiteit Brussel. Special thanks are due to John Ashworth and Filip Heremans for their help at untangling econometrical problems and An Buysschaert and Ilse Verschueren for meticulously going through (parts of) previous drafts of this dissertation.
Access to data is indispensable for empirical research. Hence, I am very grateful to Jo Buelens, Patrick Deboosere, Pascal De Ceuleners, Wendy Goossens, Lydia Putseys, Thibaut Stevens, Stefaan Swaels, Broos Van Buggenhout, Paul Van Herck, Magali Verdonck and staff-members of the Belgian police-courts, local and provincial governments and provincial archives for their assistance in procuring the necessary data.
Table of Contents
CHAPTER 1: Belgian Elections and the Level of Political Fragmentation
CHAPTER 2: Voter Turnout
CHAPTER 3: Ballot Layout Effects
4.2. Expressive voting? 4.3. Ballot scanning
CHAPTER 4: Tax Innovation
CHAPTER 5: Municipal Debt Development
Appendices
References
General Introduction
“Whateverbe insignificant, but it is you do will very important that youdo it.” Mahatma Ghandi
Elections in different countries take different forms. Besides a number of differences in practical arrangements (e.g. registration formalities), dissimilarities also exist with respect to the more technical elements in the election. Especially with reference to the way in which the seats in parliament are distributed after the election, numerous methods are employed. In fact, the electoral system – i.e. “the means by which votes are translated into seats in the process of electing politicians into office” – is never the same in two countries (Farrell, 2001, 4). Still, given the abundance of different systems that are currently in use, two main categories can be discerned: non-proportional and proportional systems. Non-proportional systems aim to achieve a clear majority for one of the parties. As such, one hopes to create strong and stable government. Proportional systems on the other hand allocate seats more or less in line with the electoral result (in terms of votes) obtained by each party. This has the advantage of lowering voter alienation and politically motivated violence, but the ensuing multiparty systems tend to be less stable (Mueller, 2003).
A key difference between both types of electoral systems is the number of parties they allow to develop and/or sustain. Non-proportional systems tend to support two-party regimes while proportional systems tend to lead to multiparty systems (e.g. Duverger, 1954/1972; Riker, 1982 and Lijphart, 1994). Importantly, the proportionality of the electoral system is a necessary but not a sufficient condition to obtain more fragmented political systems. There should also be sufficient socio-economic diversity within the population (Taagepera and Shugart, 1989). This diversity tends to create opposing views about “optimal” policies, which can give rise to the emergence of (and competition between) various political parties. The electoral system functions as a crucial intermediary factor in the rise (and continuation) of these political parties. In general, it holds that a more proportional electoral system and a more heterogeneous population will tend to be associated with a more fragmented the political system (Powell, 1982;
Ordeshook and Shvetsova, 1994; Amorim-Neto and Cox, 1997; Cox, 1997; Mozaffar et al ., 2003).
Proportional Representation (PR) is not only conducive to fragmented party systems. Casual observation also shows that the number of candidates that a party proposes to the voter in a given electoral district is larger under PR. For example, no party proposed more than one candidate in any of the 659 constituencies during the 2001 UK general elections, where a non-proportional system (i.c. the plurality rule) is used. Proportional systems on the other hand usually witness more than one candidate per party. An evident reason for this difference is that the number of representatives to be elected in a given district (the so- called “district magnitude”) is generally higher under PR relative to plurality systems (Lijphart, 1994; Farrell, 2001). This district magnitude is most often equal to 1 in plurality systems. There is only one seat to be gained and in such a context it clearly makes little sense for a party to bring forward two, three or more candidates. If the voter is only allowed to cast 1 vote – as is the case in all single-member plurality systems (Farrell, 2001, 22) – each vote gained by a given candidate is a vote lost by all other candidates. Hence, parties presenting more than one candidate split “their” votes over several candidates, which lowers their chance of winning the seat in the given district. 1 Under (list) PR, a larger number of seats is to be divided among the competing parties in each district. Hence, each party has an incentive to propose more than one candidate to the voter. This incentive is reinforced by the fact that all candidates in list PR-systems obtain votes for the party as well as for themselves.^2
The electoral system thus contributes to the existence and level of political fragmentation in the elections and in parliament and/or government. The effects of this level of fragmentation on decision-making at several stages in the political process are the focal point of this work. More precisely, we analyse the effect of political fragmentation in two phases of the political process: the election and the policy-making of the government.
(^1) The same principle holds under limited voting where each voter can cast c votes but where s seats (with s
c) are to be filled. The s candidates receiving most votes obtain a seat (Farrell, 2001). Each party then faces a strategic problem in that it may spread votes too thinly over its candidates to obtain many seats if it proposes s candidates. However, if it proposes “only” c candidates, it a priori declines the chance of 2 obtaining all s seats (Mueller, 2003, 270-271). Indeed, the distribution of seats in such electoral systems is based on the vote total obtained by the party as a whole (through list votes) and its candidates (through preferential votes). This means that the candidates of the same party can be regarded more as a team than as competitors (as is the case under plurality elections).
the local political landscape, the lower the share of the (registered) population that turns out to cast its vote.
Chapter 3 analyses the voter’s decision process inside the poll booth, viz. which candidate to vote for. It has been frequently shown that presenting long lists of candidates may affect the in-booth voting process. Candidates on the top (or bottom) of the list tend to receive more votes ceteris paribus , creating so-called Ballot Position Effects (for a review, see Darcy and McAllister, 1990 and Miller and Krosnick, 1998). In Belgium, long candidate lists have necessitated an important organizational change when computerized voting was introduced. Instead of presenting the voter with one long list of candidates per party, these have to be spread over separate columns in order to fit on the computer screen. Using preferential vote data taken from the 1995 Brussels’ Regional Elections, we show that such differences in layout affect the preferential voting behaviour of the electorate. In fact, we find that candidates on the top (or bottom) of each column tend to obtain more votes, ceteris paribus. This finding is not due to the use of computerised voting as the same (layout) effects are observed between different computerised ballots. Important from a normative point of view, we also find that these layout effects are stronger when the level of “cognitive sophistication” of the electorate (proxied by its education level) is lower. The analysis in this chapter is taken from Geys and Heyndels (2003a, b).
Chapters 4 and 5 investigate the effect of political fragmentation on a government’s policy choices. More specifically, we address two issues. The first – brought forward in Chapter 4 – concerns the effect of political fragmentation on the decision of a government to introduce a new tax. Research into this relation has thus far been very limited. Still, a number of studies on the fiscal behaviour of US state governments have shown that “divided” governments (where the governorship and the political majority in the parliament are held by different parties) are slower at introducing new taxes. We take advantage of the substantial fiscal autonomy of local governments in Belgium to re- evaluate these findings in a different context. Specifically, we analyse the introduction of “milieubelastingen” (“green taxes”) in Flemish municipalities as of 1990. It is found that the level of fragmentation of the government affects the timing of introducing this new tax. While coalitions are an sich more likely to set the tax than single-party governments,
it is also shown that the greater the fragmentation of the municipal government, the lower the likelihood that a new tax will be set.
The second policy choice – presented in Chapter 5 – refers to the local governments’ level of indebtedness. The analysis here presents a test of the “Weak Government Hypothesis” (Roubini and Sachs, 1989a, b) on local government data. This hypothesis states that weak (or more fragmented) governments tend to have higher budget deficits and debts because of coordination problems within the government. Using data on Flemish municipal long- term local debt over the period 1977-2000, we find evidence in support of the Weak Government Hypothesis. More precisely, the various measures of political fragmentation introduced in the model always have the expected sign and often are statistically significant at conventional levels. They thus indicate that the level of fragmentation indeed affects local indebtedness. Crucially, the effect is non-linear and reveals that two- party coalitions lead to higher increases (or lower decreases) in the municipalities’ indebtedness (compared to other levels of political fragmentation).
The analyses referred to are predominantly empirical in nature. They are performed on an extensive dataset providing a wide range of information on lower levels of government (municipal and regional elections). Many elements in this dataset have been brought together in the course of writing this work and have never been analysed before. One example is the dataset on municipal debts, analysed in Chapter 5. We must also mention that we keep our analyses in Chapters 2, 4 and 5 restricted to the Flemish municipalities and make no attempt to expand the studies to Brussels and Wallonia (Chapter 3 is restricted to Brussels due to the nature of the data). Though this would in itself be an interesting undertaking, two important factors made us refrain from this enterprise in the current work. Firstly, we lack the data on a number of crucial variables for Walloon municipalities, including those relating to the dependent variables of Chapters 4 and 5. Secondly, the intense political differences between Flanders, Brussels and Wallonia would introduce numerous additional complexities to the analyses. Focussing on Flanders avoids these auxiliary issues and allows a clearer analysis and presentation of the main ideas.
Introduction
The focal point of this work is the effect of political fragmentation on political decision- making by the electorate and the government. However, before one can study this, it is necessary to clarify two important issues. First of all, given that the various rules and procedures that guide the formal organisation of elections constitute the framework in which one examines political decisions, knowledge about the institutional context of the elections is warranted. Secondly, a clear idea about what the term “political fragmentation” entails is indispensable. Both issues are given proper attention in this first chapter.
A number of questions naturally arise concerning the formal organisation of the electoral process. Who is eligible to vote and how does one cast a valid vote? How are votes translated into seats? Is voting compulsory? The first part of this chapter provides an answer to such questions. Hence, the purpose is to familiarize the unacquainted reader with the workings of Belgian elections. We concentrate mainly on those issues that are of immediate relevance for understanding the empirical analyses in this work and do not attempt to give a complete description of the institutional setting of Belgian elections. Obviously, as the empirical analyses deal with the sub-federal political level, most attention is given to the procedures applied in municipal (and regional) elections. We thereby point out if and how these differ from those applied in federal elections.
Whereas the first part of this chapter describes the rules and procedures in Belgian elections, the second part focuses on the Belgian municipal party system. Here we concentrate on the central variable in the remainder of this work, namely “political fragmentation”. We provide a definition of this concept and describe the level of political fragmentation in Belgian municipalities and any trends therein over the period 1982-2000. An explanation of these trends is also provided.
1. Belgian Elections
Belgian municipalities are governed by the local council (parliament) and the College of Mayor and Aldermen (government). The number of seats in these bodies depends on the size of the municipality. In Herstappe and Mesen, the smallest Belgian municipalities, the municipal College has 3 members (2 Aldermen and the Mayor) and the local parliament consists of 7 councillors. Antwerp, the largest city, has 10 Aldermen, one Mayor and 55 councillors. For all other municipalities, the number of seats in the local government and parliament lies between these two extremes. The Aldermen are selected from among the councillors, which are themselves nominated through democratic elections. These elections take place once every six years on the second Sunday of October.
The remainder of this section takes a look at the institutional features of Belgian (local) elections. We concentrate on 4 issues that are of immediate relevance for the empirical analyses later in this work. Questions concerning who is allowed to vote, how they can cast a valid vote and other practical issues are considered in section 1.1. The Belgian electoral system – the method used to transform votes into seats – is discussed in detail in section 1.2. Finally, two key characteristics of Belgian elections – compulsory voting and electronic voting – are discussed in sections 1.3 and 1.4 respectively.
We must remark here that the regional governments are currently (2003-2004) deciding upon new institutional rules surrounding the provincial and municipal elections. Formerly the federal government decided upon these rules, but during the latest step in the reform of the Belgian state (the so-called Lambermont-agreements) the authority for the organisation of local elections was transferred from the federal government to the regional governments (Van Grembergen, 2003). Nonetheless, we provide a description of the contemporary electoral procedures. The reason is that our empirical work analyses election and policy outcomes derived under the current rules and procedures. Still, where deemed necessary, reference is made to proposals of the Flemish ministry of internal affairs concerning the impending new legislation.
before the actual election date, it is possible that there are people still on the list at the time of the election where they should have been removed (or vice versa). This implies that the officially reported number of non-voters diverges to some extent from the real number of people not presenting themselves at the polls. Nevertheless, in Belgium, the period between the official make-up of the list and the election is since 1977 limited to only two months. Furthermore, the municipal council is held to keep the voter list updated until the day of the election (VVSG, 2000).
Once in the poll booth, the Belgian voter has two options to cast a legal vote. He can choose to support a party list (by casting a “list vote”) or he can vote for one or more particular candidates from the same list (a “preferential vote”). Panachage, which is the combination of votes for candidates of different party lists, ceased to be a legal way of casting one’s vote as of the municipal elections of 1976 (Dewachter, 1982, 457). It is important to bear in mind that with the advent of computerised voting (cfr. infra), casting invalid votes has become virtually impossible – at least in districts where the computer is used to record votes. More precisely, the voting software renders impossible all ways of making the ballot invalid such as panachage or writing on the ballot form. Blank votes are obviously still possible.
Finally, we note that the type of vote cast (i.e. list or preferential vote) affects the allocation of seats within the party. The reason is that, in Belgian elections, those candidates that achieve a personal vote total above a quota receive a seat. 4 Usually, at most the top candidate manages to reach this quota on the basis of his/her preferential votes alone. To assign the other seats, the party’s list votes are distributed over the various candidates in order of their appearance on the list until all seats are filled. 5 If the party runs out of list votes before all its seats are filled, the remaining representatives for that party are designated according to the amount of preferential votes they obtained. Therefore, by casting preferential votes instead of list votes, the amount of list votes up for
(^4) This quota differs for local and federal elections. In municipal and provincial elections, it equals the quotient of (a) the party’s total vote (V (^) i) multiplied by the number of seats it obtained (Si) and (b) the number of seats it obtained plus one (or [Vi*Si]/[Si+1]). For federal and regional elections the quota is equal to the party’s total vote divided by number of seats it obtained plus one (or Vi/[Si+1]) (see also 5 Janssens and Dewachter, 1995, 18). Because the quota is higher in local elections, the amount of list votes to be distributed over the party’s candidates is also multiplied by the number of seats the party obtains. Note, however, that recent legislation halved the value of the list vote (Law of 27 December 2000). That is, the stock of list votes to be distributed is reduced to 50% of the list votes a party obtains.
distribution reduces and the number of preferential votes obtained by a candidate becomes more important in awarding the party’s seats. As a direct consequence the list order – as presented by the party – is more likely to be overthrown if more preferential votes are cast by the electorate.
This is supported by the data. In the 1995 and 1999 elections at the federal level about 60% of the ballots contained votes for a particular candidate (Smits and Thomas, 1998 and Smits and Wauters, 2000). While no politician managed to get elected outside the list order in the elections for the federal Chamber, 1 Senator achieved this in 1995 and 2 in 1999 (Smits and Thomas, 1998 and Smits and Wauters, 2000). Provincial elections also witness about 60% preferential votes (1994 data) and 22% of the representatives are elected without the use of list votes (Janssens and Dewachter, 1995). Finally, at the municipal level, looking at data from 1988 and 1994, some 80% of the ballots contain preferential votes and approximately 65% of the politicians are elected without the use of list votes (Janssens and Dewachter, 1994, 8; Wauters, 2000).
The large difference between the share of elected politicians overthrowing the list order in federal and local elections has three main explanations.
Firstly , looking at the number of representatives elected without the use of list votes (as in the studies on provincial and municipal elections) is broader than counting the representatives elected outside the list order (as in the studies on federal elections). The former category includes those candidates high upon the list who get elected on the basis of just their preferential votes as well as those candidates lower on the list who obtain seats after all list votes have been distributed. The first are situated within the list order while the second do not necessarily overthrow it.
Secondly , as mentioned in footnote 4, the “quota” is higher for municipal and provincial elections compared to federal elections. Consequently, the stock of list votes is used up faster and more candidates have to rely only on their preferential votes. This increases the possibility that candidates outside the list order are (s)elected in local elections. In line with this argument, Smits and Thomas (1998) calculate that the quota used in local elections would have increased the number of politicians
Proportional systems allocate seats in proportion to the votes obtained by each party. The idea behind such a system is one of fairness and representation of minority groups. Indeed, small parties have a (much) higher chance of gaining representation in proportional systems compared to non-proportional systems. Obviously, to be able to allocate seats proportionately, more than one seat must be awarded per district. Or, in technical terms, the “district magnitude” in proportional systems is by definition larger than 1.
The Belgian electoral system is part of the latter, proportional tradition. More specifically, Belgium has a so-called List System of Proportional Representation (PR). In such a system, voters vote for party lists of candidates instead of voting for individual candidates
The “highest averages”-methods successively divide the vote total of a party by a series of divisors. The party with the “highest average” after each division is awarded a seat. At the same time, its vote total is divided by the next divisor. This continues until all seats are awarded (Lijphart, 1994, 153; Farrell, 2001, 73-74). Methods of this type that are currently made use of include D’Hondt (with divisor series 1, 2, 3, …), Imperiali (2, 3, 4, …) and Sainte Laguë (1, 3, 5, …).
The “largest remainders”-method instead makes use of an electoral quota based on the division of the total number of valid votes cast in the election (V) and the number of seats to be allocated (S). Parties obtaining this quota obtain a seat and the quota is subtracted from their vote total. This continues until no more parties reach the quota. If by then there are still seats to be allocated, these are awarded to the parties with the largest remaining number of votes in order of vote size (Lijphart, 1994, 153; Farrell, 2001, 71-72). Some common rules of this type are Hare (quota = V/S), Hagenbach-
(^7) The technicalities of both these types of electoral systems are explained in Appendix A by means of an example.
Bischof (quota = V/[S+1]), Droop (quota = [V/[S+1]]+1) and Imperiali (quota = V/[S+2].
Belgian elections make use of the “highest averages” methods for the allocation of seats. However, at the various levels of government, different divisor series are used. In federal, regional and provincial elections, highest averages D’Hondt is used whereas municipal elections make use of highest averages Imperiali (for a comparison of these electoral systems, see appendix B). Recently, however, the Flemish regional government is considering applying the same rule for allocating seats (namely highest averages D’Hondt) at all levels of government as of the municipal elections in 2006. The intention of this proposition is to increase the transparency and uniformity of Belgian elections (Van Grembergen, 2003, 5).
Hitherto, we have separated electoral systems in 2 groups: proportional and non- proportional. This obviously is a very crude division. Proportionality should rather be seen as a continuum. Indeed, “all electoral systems tend to be at least somewhat disproportional” (Lijphart, 1994, 75). This idea is captured in indices measuring the degree of disproportionality of electoral systems. Such indices are based on the differences between the vote shares (vi) and the seat shares (si) of each party that participated in the election. Especially the index proposed by Loosemore and Hanby (1971) has become very widely used (for a review of other indices, see Lijphart, 1994, 58- 62). For an election with n parties, this measure calculates:
Higher values indicate a less proportional electoral system. By way of example, let us refer back to the example used above to compare highest averages D’Hondt and highest averages Imperiali (see appendix B). The four parties in this example obtain 36.0%, 31.5%, 21.0% and 11.5% of the votes respectively. The seat shares under highest averages D’Hondt are 42.9%, 28.6%, 14.3% and 14.3%. This leads to a degree of disproportionality Dev = 0.096. Under highest averages Imperiali the same vote shares translate into the following seat distribution: 42.9%, 42.9%, 14.3% and 0%. With an index value equal to 0.182, this implies that the latter electoral system is less proportional
