Original Article

Are parties stimulating candidate-centred voting? The case of the Belgian district council elections 2000–2006 Peter Thijssen Department of Political Science, University of Antwerp, Sint-Jacobstraat 2, Antwerpen 2000, Belgium. E-mail: [email protected]

Abstract In recent years, list systems of proportional representation around the world have undergone important electoral reforms. Changes have included the opening of closed lists, guarantees for more demographic variation on ballot lists, the organization of new sub-local elections and the initiation of common lists. Traditional political parties have tended to endorse these reforms, thinking presumably that these reforms would strengthen their positions in the electoral struggle against new contending parties. Meanwhile, scholars have noted that these plans might backfire if instead of strengthening party unity these party-supported reforms enhance the position of individual candidates to the detriment of a given party as a whole. Yet the idea that these types of electoral reforms increase candidate-centred voting has not been tested. Specifically, it is unclear to what extent citizens’ preferential votes are motivated by individual candidate attributes as opposed to party allegiance. Do electoral reforms enhance the relative influence of candidate-based attributes such as gender on voter preferences, thus diminishing the relative influence of party-based attributes such as ballot position? Our analysis of the Belgian district council elections of 2000 and 2006 reveals that although voters may be more inclined to vote for individual candidates they are no less relying on partisan clues. Acta Politica (2013) 48, 144–166. doi:10.1057/ap.2012.24; published online 14 September 2012 Keywords: personalization; electoral change; preferential vote; party dealignment; Belgium

Introduction Many studies have pointed to the growing partisan dealignment in advanced industrial democracies (for example, Dalton and Wattenberg, 2002). In a r 2013 Macmillan Publishers Ltd. 0001-6810 Acta Politica www.palgrave-journals.com/ap/

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

narrow sense, partisan dealignment may manifest itself in declining electoral support for traditional parties. In most proportional systems, the traditional parties are indeed experiencing severe competition from green parties on the left and nationalist or neo-liberal parties on the right (Dalton and Wattenberg, 2002). However, partisan dealignment can also be defined in a broader sense as a tendency toward more candidate-centred voting because parties in general may be replaced by persons as primary criterion of choice (Van Holsteyn and Andeweg, 2010). In this article, we investigate the extent to which remedies against partisan dealignment in the narrow sense may be promoting partisan dealignment in the broader sense. In other words, we will evaluate the extent to which institutional personalization is actually stimulating personalization in the electoral behavior of the public (Rahat and Sheafer, 2007). In many countries, especially in the plural societies of Western Europe, a critical anti-party stance stimulated a number of changes in the electoral opportunity structure (Schlesinger, 1966), such as the opening of (quasi-) closed-list systems, the striving for more demographic variation on ballot lists, the organization of new sub-local elections and the introduction of common lists between traditional parties and their newer contenders. Although these measures that can be cataloged as institutional personalization may result in more preferential votes for individual candidates, they might also lead to more personal vote-seeking behavior (Bra¨uninger et al, 2012) and ultimately a further loosening of the grip parties have on their representatives (Tavits, 2009). In the effort to preserve their electoral appeal, traditional parties may in fact endanger their party unity – unity that is essential in coalition building and roll call voting. The underlying assumption of this line of reasoning is that the reforms are enhancing the role of individual-based vote-earning attributes (IBVEAs) such as gender, ethnicity and political experience, while at the same time diminishing the impact of party-based vote-earning attributes (PBVEAs)1 such as ballot position. However, maybe there is no trade-off between IBVEAs and PBVEAs as voters might be stimulated to vote for individual candidates but maybe only because they are on top of the ballot list. Yet, surprisingly, whether or not electoral reforms of this kind lead to increased candidate-centred voting and decreased party-centred voting has rarely been verified, especially not in a longitudinal context (Marsh, 2007). We will use a case study that enables us to compare the behavioral importance of IBVEAs and PBVEAs in a dynamic context. We focus on the Belgian district council elections of 2000 and 2006 because this case literally constitutes a baseline study. The district council elections of 2000 were actually the first direct elections ever held at this level in Belgium. In this respect, the next and most recent district council elections of 2006 offer a unique quasi-experimental setting in which to dynamically compare IBVEAs and PBVEAs. Generally, it is extremely difficult to disentangle IBVEAs and PBVEAs because selectorates take prior individual r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

145

Thijssen

election results into account in the rank ordering of ballot lists. However, in this particular Belgian setting, we can control for these learning effects by separating new from recurring candidates and incumbents, enabling us to assess the net effects of the aforementioned electoral reforms on the percentages of preferential votes for each of the individual candidates. Moreover, contrary to many local elections none of the specific local lists, lists unrelated to any national party, were successful enough to pass the electoral threshold. Consequently, PBVEAs could be operationalized uniformly in all districts. Finally, the particular Belgian context is very useful to assess Rahat and Sheafer’s (2007) suggestion that media personalization might be a necessary go-between for institutional personalization to have an effect on personalization in the electoral behavior of the public. Individual political campaigns and media attention for individual candidates were almost absent in the Belgian district council elections.

Party Dealignment and Changes in the Electoral Opportunity Structure Certainly in a number of Western European countries, where the traditional political parties are functioning as the main pacifiers of plural societies, party dealignment is often considered to be a crucial challenge to the political system, partly because it might harm the chances of reelecting the pacifying elites from the traditional parties (Benoit, 2004). In some political systems, the critical anti-party stance of the public thus stimulated a number of reforms in the electoral opportunity structure (Karvonen, 2010). Especially in the particracies of Western Europe such as the Netherlands, Belgium and Austria, traditional parties have endorsed reforms in order to decrease the risk of losing vote share to oppositional parties that are often more eager to experiment with electoral rules (Pilet and Bol, 2011). Consequently, electoral choice options have become both larger and deeper (Deschouwer, 2004; De Winter, 2005). Hence, we will focus on four kinds of reform in the electoral opportunity structure that expanded the freedom of choice of the voters. First, because the electoral interests of some traditional party candidates might be damaged by making the link with their own parties overtly explicit, the importance of the linkage was reduced. In practice, this reduction was realized by augmenting the degrees of freedom of the voters and giving them more influence in the seat allocation or the selection of the individual candidates who will actually represent them (Cain et al, 2003; Marsh, 2007). This strategy could be successful because, contrary to their contenders on the left and the right, traditional parties are often considered better equipped to mobilize strong individual candidates because of the elaborateness of their party organization. Second, new (sub-)local elections were introduced to create a new incumbent elite that can stop the success of the populist and extreme rightist parties who 146

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

seem to flourish in anonymous and impersonal urban contexts (Lowndes and Sullivan, 2008). Because the public generally has more confidence in local political elites, new candidates are recruited who are better embedded in the local neighborhood context and who can profit from a virginal political blazon (Levi and Stoker, 2000). Moreover, once these new recruits get elected, they could become a stable elite because they are able to profit from the incumbency effect. An extensive number of previous studies have indeed demonstrated that incumbency is one of the main, if not the most important, factors shaping people’s ballot decisions (Krebs, 1998; Desposato and Petrocik, 2003; Carson et al, 2007). Third, the traditional parties offered increasing commitments to demographic variation in the selection of the candidates on the ballot. By doing so, they countered the unique selling proposition of leftist oppositional parties, who are usually stronger advocates of descriptive representation. This commitment is necessary because more and more voters are in favor of better representation of traditionally underrepresented socio-demographic groups, such as women and ethnic minorities (Matland and Studlar, 1996; Norris, 2004). With regard to the underrepresentation of minority groups, reference can be made to legal candidate quota and voluntary party quota. Since the mid-1990s, countries hoping to increase the percentage of women legislators have, for instance, adopted quota laws stipulating that ballot lists must contain a minimum percentage of female candidates. Obviously, this strategy will be more successful if minority candidates are presented as such in the political campaign, but even in the absence of this information the names of a candidate as indicated on the ballot list can often be used successfully as a shortcut for identifying the candidate’s gender and ethnicity. Fourth, to fight the proliferation of the party system and to reduce the severe competition with new(er) parties, it can be beneficial to field candidates on a common election list.2 The thinking here is that if you cannot beat them, then join them (Golder, 2006). The introduction of a common list can be an effective strategy for new parties who might not pass the electoral threshold, but it is clear that the established parties can profit too. Nonetheless, little or no research has been done to evaluate the electoral return of common lists by different parties. An interesting recent exception is the experimental study by Gschwend and Hooghe (2008), which concluded surprisingly that many voters who are attracted by a common list would not consider voting for either of the ‘coalition partners’.

Changes in the Electoral Opportunity Structure and Voting Behavior Many studies have given attention to the effects of electoral reforms on the behavior of representatives with respect to their constituencies after being elected (Carey and Shugart, 1995; Shomer, 2009; Karvonen, 2010). The underlying r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

147

Thijssen

assumption of this research is that citizens vote differently as a result of electoral reforms. Yet this assumption has not been tested thoroughly. Although Farrell and McAllister (2006) have demonstrated that voters are more satisfied with the way democracy works in countries with a more preferential system, we do not know how voters use their augmented discretionary power. Voters may be more inclined to vote for individual candidates but maybe only based on partisan clues. As demonstrated by Hibbing and Theiss-Morse (2002), voters may be interested in more electoral impact in theory, but less in practice. Some scholars have tried to tackle this research question by surveying the electorate on the motivations for their own voting behavior (Marsh, 2007). Unfortunately, asking the public to motivate their own voting behavior may often be too demanding and produce questionable post-factum rationalizations (Verba et al, 1995, p. 106). Voters are unlikely to admit having requested more electoral impact that they end up barely using. ‘Although many voters may vote a party ticket they will rationalise this to themselves in terms of candidate qualities’ (Marsh, 2007, p. 520). Thus, a purely subjective demand-side voter-reported approach may not be the best solution to study the attributes that motivate voting behavior. Because demand-side voter motivations are manifested in election results, it seems a better strategy to link these results to characteristics of the supply side (for example, attributes of parties and their candidates). The cross-national comparison of Shugart et al (2005) has informed the way we will tackle this research question. Shugart and colleagues assume that electoral rules, such as district magnitude (number of candidates on a list) and list type (open or closed), determine which kind of shortcuts voters will use for cutting through all the messages being provided by individual candidates in list systems of proportional representation. Indeed, the importance of ascriptive attributes of legislators increases with district magnitude when the list is open, and decreases with magnitude when the list is closed. The elected representatives in open-list systems with long lists appear to have a higher probability to be native to their constituency or have more local political experience. However, as Shugart and colleagues rightfully conclude, we are only beginning to understand the effects of electoral rules on the shortcuts voters employ. Some important questions still remain. First, because the study of Shugart and colleagues does not involve a time dimension, we do not know what will happen if closed-list systems are opened. Second, Shugart and colleagues do not take into account so-called ‘flexible list systems’ (Marsh, 1985; Katz, 1986), ‘who are worthy of analysis in their own right’ (Shugart et al, 2005, p. 442). Flexible list PR systems provide a ranked list, the order of which may be overridden if a candidate surpasses a certain preferential-vote quota. These systems exist in a number of European countries such as Austria, Belgium, the Czech Republic, the Netherlands, Slovenia and Sweden. Third, most research has hitherto used a very narrow conception of electoral system characteristics. 148

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

However, some electoral strategies that are related to a broader electoral opportunity structure, such as the formation of pre-electoral coalitions, can be equally or even more important. Fourth, the study of Shugart and colleagues focuses on the attributes of legislators and does not incorporate information on the candidates who are not elected. Although they are right that data on defeated candidates are generally not available, this data problem may not be completely inconsequential. Certainly, if personalization means the attention for some at the cost of others it seems essential to consider all the candidates on a list. Fifth, whereas Shugart and colleagues mainly focus on factors related to the local embedding of the candidates, we will use a slightly broader conception of individually based attributes. We believe that it is very hard to get reliable indicators for social embedding in the voting district without a candidate survey and acknowledging interaction effects with district characteristics. Sixth, most of the existing research is limited to national elections. At first sight, this focus on the first-order electoral arena seems logical, but from a purely analytical point of view it may not be the best choice. On the one hand, the local level is often used as an experimental ground for electoral engineering (Kersting and Vetter, 2003). On the other hand, the local context also functions as a kind of baseline environment in which to verify the importance of IBVEAs. As argued by Oliver and Ha (2007), local governments are service providers rather than lawmakers, meaning that local politics are more allocational in nature than they are policy driven. Hence, it can be argued that the party label and ballot position are less crucial in local elections than they are in higher-level elections. Ceteris paribus, if IBVEAs do not demonstrate a significant influence at this level, they are unlikely to do so at a higher level either.

Case Selection and Hypotheses In an attempt to tie up these loose ends, we will focus on the Belgian district council elections of 2000 and 2006. These elections enable us to study the longitudinal impact of broad electoral changes on the propensity of voting based on IBVEAs in local elections, taking into account all candidates in a flexible list system of proportional representation. The Belgian district council elections are the result of an intra-municipal territorial decentralization process set up to decrease the geographical and social distance between city governments and its citizens. The new article 41 in the Belgian constitution allows municipalities with more than 100 000 inhabitants to establish directly elected, intra-municipal, territorial bodies that are authorized to regulate matters of municipal interest.3 However, only Antwerp, which at about 450 000 inhabitants is the largest city in the country, has actually implemented this legal option. Presumably, because the extreme rightist party Vlaams Blok was particularly r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

149

Thijssen

successful in this city. The first directly elected district council elections of 2000 were therefore confined to the nine Antwerp districts: Antwerpen, Berchem, Berendrecht-Zandvliet-Lillo, Borgerhout, Deurne, Ekeren, Hoboken, Merksem and Wilrijk (Van Assche and Dierickx, 2007). The electorates of these districts had to evaluate a fairly big number of new candidates in each of the districts, varying from 15 candidates in the smallest district to 33 in the largest. According to Shugart et al (2005, p. 442), it is exactly in this context that voters will use individual-based shortcuts to determine their preferred candidates. These elections are not only interesting because they were an institutional novelty in 2000, but also because each of the aforementioned electoral innovations were implemented in the period between the first elections in 2000 and the second elections in 2006 (Table 1). Moreover, although the district council elections are low-information elections with very limited individual campaigning and media attention, the changes in the electoral opportunity structure received ample (media) attention because they would be simultaneously implemented in the city and provincial council elections, which took place at the same time (Thijssen, 2007). Hence, these changes seem to create an actual context whereby voters are stimulated to think in terms of IBVEAs rather than PBVEAs.4 We formulate hypotheses positing an increased importance of IBVEAs or a decreased importance of the PBVEAs. Table 1: Electoral innovations in the Belgian district council elections A. Making the electoral system more preferential In casu: Dismantlement of the devolutive effect of the party vote K 2000: Half of the party votes used in the seat allocation K 2006: A third of the party votes used in the seat allocation B. Bridging the gap between representative and electorate In casu: Organization of new Belgian district council elections in 2000 Electing a new local political elite at district level K 2000: No retrospective voting and no incumbents K 2006: Retrospective voting is possible and incumbents are available for the first time C. Greater descriptive variation in the selection of candidates In casu: Increasing the number of female candidates on the list in accordance with the gender parity law K 2000: At least one-third of the candidates is of the underrepresented sex K 2006: Half of the candidates of each sex D. Collaboration with contending new(er) parties In casu: Introduction of common lists K 2000: No common lists K 2006: Common list of social democratic and green party and common list of Christian democratic and regionalist party 150

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

Our first hypothesis is related to the opening of electoral lists. Belgium uses a flexible list system of proportional representation (Deschouwer, 2004). Voters may either cast a party vote by ticking the relevant box under one party name or alternatively cast a preferential vote for one or more candidates on the list ranked by the party (by ticking the relevant box(es) next to the candidate(s)). Only the preferential votes can result in a change of the rank ordering of the candidates as proposed by the party. Any candidates with more preferential votes than the eligibility quota computed by dividing their lists’ total constituency vote by one more than the number of seats it won are declared elected. However, owing to the devolutive effect of the party list, preferential votes for individual candidates used to be of limited influence in the allocation of seats. The devolutive effect implies that party votes can be used to top up the individual votes of the candidates placed high on the party list. However, the proportion of party votes that is used in the allocation of the seats has recently been decreased. For the district council elections of 2000, only half of the stock of party votes was used. In 2006, the devolutive effect was further reduced to one-third, thereby making it negligibly small. Because of the further reduction of the influence of the party votes on the individual seat allocation, voters may be more inclined to cast a preferential vote for a candidate occupying a lower ballot position. Moreover, the indirect effect of the electoral change may even be more important. Owing to a third-person effect, the selectorates might overestimate the electoral changes’ impact on the voter. Hence, they might put less popular but highly valuable candidates on the list in order to ascertain their election via the smaller stock of devolved list votes. Hypothesis 1:

The rank ordering of the candidates on the list (‘ballot position’) will have less impact on the individual electoral success (‘district percentage of preferential votes’) in the district council elections of 2006 compared to those of 2000 owing to the further opening of the list system.

The Belgian district council elections indirectly constituted an electoral innovation implemented to combat party dealignment because the purpose of the new elections was to establish a new, locally embedded ‘incumbent’ elite. Obviously, when the district council elections were organized for the first time in 2000, there were no incumbents. Only in the second district council elections of 2006 were there incumbents up for reelection. These members of the district councils or alderman had gained 6 years’ worth of legislating in which to gain personal reputation. Especially in the near absence of other sources of information on individual candidates, we expect to find a substantial incumbency effect. Hence, preferential voting in the elections of 2006 will be more determined by IBVEAs. However, because incumbents have a ‘proven’ electoral potential, those responsible for the compilation of the ballot list are r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

151

Thijssen

often tempted to give them a good rank. As we are dealing with ‘new’ elections, the incumbency effect could be a mere artifact of the ‘initial’ ballot position in 2000. If the ‘incumbency effect’ disappears when controlling for ‘ballot position’, it only confirms the importance of the PBVEAs. Hence, it is essential that Hypothesis 2 provides a control for ballot position. Hypothesis 2:

Incumbents will be significantly more successful in the district council elections of 2006 than other candidates, taking into account their ballot position.

Because some of the smaller leftist parties are successful in attracting votes by presenting lists with a significant number of candidates belonging to politically underrepresented groups, traditional parties also give attention to demographic descriptive representation when compiling their electoral lists. Whereas the leftist parties try to consolidate particular ethnic niches in an increasingly ethnically diverse electorate by augmenting the number of ‘ethnic’ candidates, other parties try to gain access to these niches by mimicking this strategy (Teney et al, 2010). Moreover, in addition to a contagion effect, legal measures oblige parties to strive for more socio-demographically diverse electoral lists. Here the Belgian gender quota laws should be mentioned. The first Belgian gender quota law was introduced in 1994, stipulating that the electoral lists may not comprise more than two-thirds of candidates of the same sex. In practice, this means that at least one-third of the candidates on the lists of the district council elections of 2000 had to be women. In 2002, a new gender quota law was introduced, the so-called ‘parity law’, stipulating that electoral lists should contain an equal number of male and female candidates, although the first three positions on the list may not be occupied by candidates of the same sex. This new, more stringent parity legislation was in effect at the time of the district council elections of 2006. In short, owing to both legal constraints and informal arrangements, party elites are giving more attention to demographic characteristics in the selection of their candidates. As a result, voters might become more sensitive to these characteristics and vote for minority candidates regardless of their particular ballot position. Hypothesis 3: Women candidates and candidates belonging to an ethnic minority are more successful in the district council elections of 2006 than in those of 2000, taking into account their ballot position. Finally, some traditional parties have decided to join forces with a new ideologically related party. In the Belgian district council elections, four parties that presented themselves separately to voters in 2000 have joined forces, introducing two common lists in 2006. The Christian Democratic Party joined forces with the Flemish Regionalist Party, whereas in some districts5 the Social 152

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

Democratic Party fielded their candidates on a common list with the Green Party. Hence, the voters of these common lists have the opportunity to give preferential votes to candidates belonging to different parties, although this socalled panachage is normally not possible in Belgium. It is therefore plausible that many preferential voters of these common lists will not limit themselves to candidates of one party, which will deflate the effect of ‘party label’. Hypothesis 4: Because in the district council elections of 2006 candidates of different parties are fielded on common lists, although this was not the case in the district council elections of 2000, the importance of party label as a predictor of individual electoral success will decrease for the parties involved.

Data: Methodology and Variables To test these hypotheses, we constructed a database that contains pooled unbalanced panel data based both on the 2000 and the 2006 district council elections. Given that we investigate what determines the electoral success of individual candidates, the dependent variable is the logarithmically transformed individual percentage of preferential votes obtained in the district. We use percentages instead of absolute numbers of preferential votes because we want to compare individual results in districts with divergent district magnitudes. We divide by the total number of preferential votes because dividing by the total number of valid votes would put candidates in a disadvantage on lists with many list votes. This does not seem fair as list votes often imply an implicit endorsement of the rank order of the candidates proposed by the party. The log-transformation remediates the extreme skew in the percentages. Moreover, given that the percentages are censored (vote percentages cannot fall below 0), Tobit regressions are used. Because a candidate on a list of a winning party usually obtains a larger share of preferential votes regardless of his or her personal popularity, it is useful to include the party label as an independent variable in the analysis. This is only possible when using ‘district percentages’ instead of ‘list percentages’ as the dependent variable. Moreover, we only focus on the candidates of the political parties that submitted a ballot list in each of the nine Antwerp districts. Concretely, this means that the analysis is based on the electoral results of the candidates who were affiliated with one of the following six political parties: the Green Party (Agalev/Green!), the Social Democratic Party (SP/sp.a), the Christian Democratic Party (CVP/CD&V), the Liberal Party (VLD), the Flemish Regionalist Party (VU/N-VA) and the Radical Right Party (Vlaams Blok/Vlaams Belang). Thus, the dependent variable should be r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

153

Thijssen

interpreted as: the electoral success of a candidate measured against not just the other candidates on his/her list, but against all the other candidates from the six main political parties in the district. Given that we have full population data, statistical inference is conditional on the nine districts. In this context, a fixed effects approach using dummy variables is appropriate (Gujarati, 2004). The district dummies control for all time-invariant contextual explanations of variation in preferential vote percentages across districts. To assess whether the effects of IBVEAs and PBVEAs have changed over time, we include interactions with the dummy ‘time’ (election year 2000 ¼ 0 and 2006 ¼ 1). Because it is important not to interpret the effects of interactions as unconditional effects, we will report conditional effects for both values of the dummy ‘time’ with their corresponding standard errors6 (Brambor et al, 2005). Last but not least, in order to control for the unbalanced format of the election data and the learning effect of selectors who take into account the candidate’s previous electoral performance, we include a dummy for candidates that participated in both elections. We subdivide the indepedent variables into two categories; on one hand PBVEAs: ‘party label’ and ‘ballot position’ and on the other hand IBVEAs: ‘minority status’ (ethnicity and gender) and ‘incumbency’. The data for these variables are gathered from the actual ballot lists and official records. Political parties in Belgium are obliged to deposit a list with demographic information on each of their candidates. In order to evaluate our four hypotheses, we estimate a series of fixed effects Tobit models consisting of a series of nested sub-models based on two categories of PBVEAs and two categories of IBVEAs (Table 2). By introducing these categories sequentially, we get a better understanding of the interrelation between the IBVEAs and the PBVEAs. Obviously, the sequence of the submodels is not without relevance. Because our hypotheses assume the IBVEAs to have an increasing impact on the voting behavior, the models based on the IBVEAs are included first. However, we make an exception for the PBVEA ‘party label’ because in list systems of proportional representation this variable should be controlled for from the beginning.7 Like most list systems of proportional representation, the Belgian electoral system does not allow voters to give preferential votes to candidates on more than one party list. Consequently, even a voter whose vote choice is inspired by IBVEAs must first pick a party list in order to select the preferred candidate(s). All in all, this means that we will verify our four hypotheses in reverse order. The effect of the introduction of common lists (Hypothesis 4) is analyzed in Models 1, 2 and 3. The effect of stressing the importance of demographic descriptive representation (Hypothesis 3) is covered in Models 4 and 5. The incumbency effect (Hypothesis 2) and the effect of the opening of the list (Hypothesis 1) are verified, respectively, in Model 6 and in Models 7 and 8. 154

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Party label (Hypothesis 4)

N=2157 N(2000)=1201; N(2006)=956

Constant Time 2006 (Ref. 2000)

Model 1

Model 2

Model 3

Model 4

b

b

b

b

Acta Politica Vol. 48, 2, 144–166

Common list 2006 Social Democrat on common list 2006

155

Minority status Female Ethnic

SE

SE

SE

SE

Incumbency (Hypothesis 2)

Model 5

Model 6

b

b

SE

1.87*** 0.07 1.62*** 0.08 1.57*** 0.08 1.71*** 0.08 1.75*** 0.08 1.72*** 0.02 0.04 0.52*** 0.04 0.52*** 0.09 0.57*** 0.09 0.47*** 0.09 .56***

Party label (Ref.=Liberal Party) Green 0.17* Social Democratic Party 0.71*** Flemish Regionalist Party 0.61*** Christian Democratic Party 0.23** Far-Right Party 0.92*** Party label * Time 2006 Green 2006 Social Democratic 2006 Flemish Regionalist 2006 Christian Democratic 2006 Far-Right 2006

Demographic representation (Hypothesis 3)

0.07 0.09 0.09 0.08 0.09 0.04 0.09 0.02 0.09 0.01 0.07 0.22** 0.09 0.22* 0.09 0.19* 0.09 0.18* 0.09 0.18* 0.08 1.02*** 0.10 1.03*** 0.09 1.03*** 0.09 1.03*** 0.09 1.05*** 0.07 0.02 0.09 0.02 0.09 0.00 0.09 0.00 0.09 .00 0.06 0.60*** 0.09 0.60 0.09 0.61*** 0.09 0.61*** 0.09 0.61***

0.10 1.03*** 1.43 0.42** 0.65***

0.13 0.21 0.13 0.85*** 0.19 0.13 0.27 1.15*** 0.13 0.65***

— — — — —

— — — — —

— —

— —

— —

— —

— —

— —

— —

— —

0.14 0.19 0.14 0.85*** 0.23 0.08 0.27 1.14*** 0.12 0.67***

0.13 0.16 0.14 0.87*** 0.23 0.08 0.27 1.13*** 0.12 0.66***

0.13 0.13 0.13 0.88*** 0.23 0.04 0.27 1.08*** 0.12 0.60***

1.57*** 0.20 1.54*** 0.19 1.54*** 0.19 1.50*** 1.10*** 0.24 1.11*** 0.23 1.11*** 0.23 1.11***

— —

— —

0.25*** 0.04 0.62*** 0.10

0.35*** 0.05 0.92*** 0.16

0.35*** 0.91***

SE

Ballot position (Hypothesis 1) Model 7

Model 8

b

b

SE

SE

0.08 0.91*** 0.09 0.94*** 0.09 0.09 0.55*** 0.09 0.48*** 0.11

0.09 0.01 0.09 0.19* 0.09 1.16*** 0.09 0.00 0.09 0.62***

0.09 0.01 0.09 0.19* 0.09 1.15*** 0.09 0.00 0.09 0.62***

0.08 0.08 0.09 0.08 0.08

0.13 0.12 0.13 0.88*** 0.23 0.06 0.26 1.09*** 0.12 0.63***

0.12 0.12 0.13 0.88*** 0.24 0.05 0.21 1.09*** 0.11 0.63***

0.12 0.12 0.25 0.21 0.11

0.19 1.53*** 0.18 1.53*** 0.18 0.23 1.12*** 0.21 1.12*** 0.21

0.05 0.16

0.34*** 0.05 0.84*** 0.15

0.34*** 0.05 0.84*** 0.15

Are parties stimulating candidate-centred voting?

r 2013 Macmillan Publishers Ltd. 0001-6810

Table 2: Which candidates obtain a lower/higher log percentage of the preferential votes in their constituency?

156

Table 2: Continued

Minority status * Time 2006 Female 2006 Ethnic 2006 Incumbents 2006 Ballot position Ballot position * Time 2006

Acta Politica Vol. 48, 2, 144–166

Fixed effects ‘controls’ District (Ref.=Antwerpen) Berchem Bezali Borgerhout Deurne Ekeren Hoboken Merksem Wilrijk Panel candidate (2000 and 2006) Pseudo R2

Party label (Hypothesis 4)

Demographic representation (Hypothesis 3)

Model 1

Model 2

Model 3

Model 4

Model 5

b

SE

b

SE

b

SE

b

SE

b

— — — — —

— — — — —

— — — — —

— — — — —

— — — — —

— — — — —

— — — — —

— — — — —

0.21** 0.45* — — —

0.60*** 1.09*** 0.55*** 0.35*** 0.73*** 0.43*** 0.50*** 0.49***

0.08 0.09 0.08 0.08 0.08 0.08 0.08 0.08

0.61*** 1.08*** 0.54*** 0.34** 0.73*** 0.43*** 0.51*** 0.50***

0.08 0.09 0.08 0.08 0.08 0.08 0.08 0.07

0.61*** 1.00*** 0.43*** 0.23** 0.73*** 0.43*** 0.41*** 0.50***

0.07 0.09 0.08 0.08 0.08 0.08 0.08 0.07

0.63*** 1.03*** 0.43*** 0.27** 0.78*** 0.45*** 0.45*** 0.52***

0.07 0.09 0.08 0.07 0.08 0.08 0.08 0.07

0.63*** 1.04*** 0.44*** 0.28** 0.79*** 0.46*** 0.45*** 0.52***

Incumbency (Hypothesis 2) Model 6

SE

b

b

SE

b

SE

0.15* 0.29 0.35*** 0.04*** 0.01

0.07 0.18 0.08 0.00 0.00

0.07 0.09 0.08 0.07 0.08 0.08 0.08 0.07

0.63*** 1.05*** 0.43*** 0.28** 0.79*** 0.45*** 0.44*** 0.51***

0.07 0.09 0.08 0.07 0.08 0.07 0.07 0.07

0.42*** 0.64*** 0.23** 0.131 0.47*** 0.19** 0.27*** 0.31***

0.07 0.09 0.07 0.07 0.07 0.07 0.07 0.07

0.42*** 0.64*** 0.23** 0.131 0.47*** 0.19** 0.27*** 0.31***

0.07 0.09 0.07 0.07 0.07 0.07 0.07 0.07

0.05

0.39*** 0.04

0.40*** 0.04

0.21

0.21

0.63*** 0.04

0.62*** 0.04

0.50***

0.12

0.13

0.14

0.16

0.16

0.16

—

Model 8

0.07 0.18 0.08 0.00 —

0.59*** 0.04 —

Model 7

0.08 0.15* 0.20 0.30 0.08 0.38*** — 0.05*** — —

0.60*** 0.04 —

Ballot position (Hypothesis 1)

0.08 0.18** 0.20 0.43* — 0.56*** — — — —

0.62*** 0.04 —

SE

Thijssen

r 2013 Macmillan Publishers Ltd. 0001-6810

N=2157 N(2000)=1201; N(2006)=956

—

—

Note: Nested Tobit regressions based on pooled unbalanced panel data; 1Po0.1; *Po0.05; **Po0.01; ***Po0.001.

—

—

Are parties stimulating candidate-centred voting?

Results Before discussing the results of the Tobit regressions, we first look at changes in the overall number of preferential votes cast in the two elections in question. In 2000, 63.5 per cent of the voters cast a preferential vote, whereas in 2006 this number increased to 71.7 per cent. This increase, which is similar for other Belgian elections (Wauters et al, 2010), underpins our research question. It only makes sense to assess whether electoral reforms lead to an increased importance of individual attributes of the candidates if the voters cast more preferential votes in the first place. We see here that this necessary condition is fulfilled.

Common list effect (Hypothesis 4) In the first model, we introduce five dummies indicating candidate party affiliation: the Green Party, the Social Democratic Party, the Flemish Regionalist Party, the Christian Democratic Party, the Far-Right Party (reference category: Liberal Party), a dummy-variable for ‘time’, which equals 1 for the elections of 2006 and fixed effects ‘controls’ for district and panel candidate. By comparing the coefficients of the party dummies, we learn, for instance, that candidates of the Far-Right Party and the Social Democratic Party are generally more successful than those of the Liberal Party. The b-coefficient of 0.71*** for the Social Democratic Party indicates that the percentage of preferential votes will increase with about 103 per cent8 when a candidate runs for that party instead of the Liberal Party. However, given that the pseudo R2 for Model 1 equals 0.13, a large part of the variance in individual electoral success is not attributable to distinctive party popularity. In the second model, we use the same independents as in Model 1, but this time we also include party label dummies for the candidates of the 2006 election. The results of this second model show that the conditional effect of being a candidate on the list of the Social Democratic Party is much stronger in 2006 (byx.2006 ¼ 0.22 þ 1.03 ¼ 1.25; SE ¼ 0.09) than in 2000 (byx.2000 ¼ 0.22; SE ¼ 0.09).9 However, this strong ‘party label’ effect for the candidates of the social democratic list in 2006 might decrease when we take into account that in several districts these candidates were fielded on a common list with the Green Party. To estimate the effect of the introduction of common lists that combine a bigger party with a smaller one in the 2006 election, we introduce an additional variable ‘common list’ in Model 3. This variable indicates whether or not a candidate belongs to one of the pre-electoral coalitions, which in all districts is the case for the candidates of the Christian Democratic Party and the Flemish Regionalist Party (rightist pre-electoral coalition), as well as in some districts for the candidates of the Social Democratic Party and the Green r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

157

Thijssen

Party (leftist pre-electoral coalition). The results of this analysis indicate that figuring on a common list has a significant positive effect on the percentage of preferential votes a candidate obtains (b ¼ 1.57***). With the exception of the Social Democratic Party, we find smaller party label effects for candidates of parties involved in pre-electoral coalitions in 2006. Although the effect of the Christian Democratic Party label equalled 0.02 in 2000, this effect decreased to 1.65 (0.020.52–1.15) in 2006, taking into account that this party then introduced a common list. However, in order to test Hypothesis 4 more genuinely, we need to focus on the lists of the greens and the social democrats, as these parties introduced common lists in some districts, whereas in others they did not. Hence, in Model 3 we also introduce an interaction variable of ‘common list * Social Democratic Party’. The corresponding effect parameter equals 1.10***, and the negative sign clearly contradicts Hypothesis 4. Candidates of the Social Democratic Party on a common list are not more successful than candidates on a separate list in the same 2006 elections.10 The conditional effect of a social democrat on a separate list equals 1.07 (byx.2006 ¼ 0.22 þ 0.85 ¼ 1.07; SE ¼ 0.14), whereas the corresponding effect for a social democrat on a common list equals 0.03 (byx.common list in 2006 ¼ 0.22 þ 0.85–1.10 ¼ 0.03; SE ¼ 0.35). Apparently, the introduction of a common list is not advantageous to the candidates of the larger party in this pre-electoral coalition. Things look different for the smaller Green Party. If we would have chosen to include an interaction variable ‘common list * Green Party’, the corresponding effect parameter would be analogous, although with an opposite sign.11 Here the common list effect (1.10) would outstrip the party label effect in 2006 (b ¼ 0.21). We notice that the conditional effect of a green candidate figuring on a separate list in 2006 equals 0.10 (byx.2006 ¼ 0.080.21 ¼ 0.13; SE ¼ 0.14), whereas the corresponding effect for a green candidate figuring on a common list equals 0.97 (byx.common list in 2006 ¼ 0.080.21 þ 1.10 ¼ 0.97; SE ¼ 0.17). Overall, common lists only seem to be beneficial for the candidates of the smaller party in the pre-electoral coalition and not for those of the larger traditional party. To the extent that common lists are initiated by the latter parties, we do not find support for Hypothesis 4.

Socio-demographic voting shortcuts (Hypothesis 3) As far as the effect of sex and the ethnic origin of a candidate are concerned, the regression coefficients are significant and positive when the variables are entered in the equation (Model 4: b ¼ 0.25*** and b ¼ 0.62***, respectively). In order to identify the ethnic minority candidates, we relied on a name recognition method. Three coders independently screened the names that 158

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

appeared on the ballot lists. Only those candidates that were assigned to an ethic minority by at least two of the coders were identified as such (intercoder reliability exceeded 0.99 per cent). In the 2000 elections, 31 ethnic minority candidates were identified in a total of 1201 candidates (2.6 per cent). In the 2006 elections, 65 ethnic minority candidates were identified in a total of 956 candidates (6.8 per cent). At first sight, this may seem a very crude proxy, but voters who want to support a candidate of an ethnic minority will probably use the same proxy in low-information elections. Women and ethnic minority candidates attract a significantly larger percentage of preferential votes. However, in contrast with the strongly positive zero-order effects of the ‘ethnicity’ variable, the interaction effects ‘sex * time 2006’ (b ¼ 0.21**) and ‘ethnic minority * time 2006’ (b ¼ 0.45*) are significantly negative. Hence, the effects of sex and belonging to an ethnic minority have definitely not become stronger between the two elections. Consequently, the positive effect of ‘sex’ is much weaker for the elections of 2006 (byx.2006 ¼ 0.350.21 ¼ 0.14*; SE ¼ 0.06) than for the elections of 2000 (byx.2000 ¼ 0.35***; SE ¼ 0.05). Similarly, the conditional effect of ‘ethic minority status’ is clearly much smaller in 2006 (byx.2006 ¼ 0.920.45 ¼ 0.47***; SE ¼ 0.12) than in 2000 (byx.2000 ¼ 0.92***; SE ¼ 0.16). Although we expect woman and ethnic minority candidates to be more successful in 2006 owing to a higher sensitivity for descriptive representation, the opposite is true. Although the selectorates have, for example, legally committed themselves to placing an equal number of male and female candidates on the lists of 2006, both sexes do not seem to be equally popular with the voters. However, the lower electoral appeal of woman candidates in 2006 could be attributable to the fact that many of these women did not have any political experience and received a low place on the list? Indeed, whereas the negative marginal effect of ‘sex * time 2006’ equals 0.21** in Model 5, it is only 0.15* in Model 7 when controlling for ‘incumbency’ and especially ‘ballot position’. Similarly, the negative marginal effect of ‘ethnic minority * time 2006’ evolves from 0.45* to 0.30. All in all, the previous findings constitute a powerful argument against Hypothesis 3: as the number of candidates belonging to a politically underrepresented socio-demographic group increases, voters do not seem to vote for all these minority candidates. Ethnic and gender voters become more selective in this context and, for example, tend to vote more for those minority candidates who occupy the best ballot position. In other words, formally or informally, stimulating minority representation on the lists seems to deflate the effect of minority status on individual electoral success (IBVEA), this at least partially because voters increasingly rely on PBVEAs such as ballot position to deal with the augmented degrees of freedom in terms of minority candidate supply. Interestingly, however, even when the number of minority candidates r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

159

Thijssen

reaches population parity, as for the woman candidates in 2006, the conditional effects of minority status remain positive. The conditional effect of ‘sex’ in 2006 (byx.2006 ¼ 0.141) even expands when controlling for ‘incumbency’ in Model 6 (byx.2006 ¼ 0.350.18 ¼ 0.17**; SE ¼ 0.06) and when additionally controlling for ‘ballot position’ in Model 7 (byx.2006 ¼ 0.340.15 ¼ 0.19***; SE ¼ 0.05). Similarly, the conditional effect of belonging to an ‘ethnic minority’ in 2006 (byx.2006 ¼ 0.47) substantially increases when taking into account ‘incumbency’ (byx.2006 ¼ 0.910.43 ¼ 0.48***; SE ¼ 0.11) and additionally ‘ballot position’ (byx.2006 ¼ 0.840.30 ¼ 0.54***; SE ¼ 0.11). Voting for incumbents (Hypothesis 2) When the ‘incumbency’ variable is introduced in Model 6, it exerts a significant positive effect on the percentage of preferential votes (b ¼ 0.56***). Moreover, when ‘ballot position’ is controlled for, the effect remains strong, although the regression coefficient drops considerably (b ¼ 0.38***). Incumbents receive a larger percentage of preferential votes as compared with non-incumbents on comparable ballot positions. Apparently, we have some support then for Hypothesis 2. However, we have to admit that the ‘baseline’ incumbency effect (in the first elections with incumbents) appears to be much smaller than the incumbency effect in long-standing elections (Krebs, 1998). The observations indeed suggest that the political activities of the district council members were not that visible to the voter. In the context of local politics, where media attention is limited, the role of incumbency as an IBVEA may be more limited. Ballot position (Hypothesis 1) In Model 7, when the variable ‘ballot position’ is introduced, its effect is strongly negative (b ¼ 0.05***), indicating that a lower rank results in an increased percentage of preferential votes. If a candidate goes up one place on the list, he gets rank k-1 instead of k, his percentage of preferential votes increases with 5 per cent. Moreover, the interaction term ‘ballot position * time’ also has a negative sign (full Model 8; b ¼ 0.01). Contrary to our expectations, ballot position does not have less influence on the percentage of preferential votes in the elections of 2006. The further opening of the list system does not reduce the ballot position effect; on the contrary, so it seems. Occupying favorable ballot positions seems even slightly more important than it already was. We have to reject Hypothesis 1. However, maybe voters still willing to support their preferred party will be inclined to vote exclusively for those candidate(s) with the best ballot position’s, for example the first and the last candidate on the list. In Model 160

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

8bis in Table 3, we see that the candidates occupying the first (b ¼ 1.71***) and the last ballot position (b ¼ 1.52***) receive a huge electoral bonus. We also observe a rather strong effect of being the first woman on the list (b ¼ 0.44***), which corroborates our earlier findings regarding Hypothesis 3. An additional argument in favor of the overwhelming importance of these prominent ballot positions can be found in the explanatory power of the models. When the prominent ballot positions are added in Model 8bis, the explanatory power increases drastically (pseudo R2 increases from 0.21 to 0.34). Interestingly, however, the results of Model 8ter point out that none of the interactions of the prominent positions with ‘time 2006’ exerts a significant positive marginal effect. In other words, although the importance of occupying a prominent ballot position cannot be underestimated, it has not become more important over time. Furthermore, the incumbency effect further decreases (b ¼ 0.21) when the most visible ballot positions are added in Model 8bis. The electoral bonus for incumbents stems only very partially from their incumbency status (PBVEA) as such, but is strongly related to the fact that these candidates occupy prominent ballot positions. Indeed, more than 60 per cent of electoral lists are headed by an incumbent.

Conclusion In many countries, especially in the plural societies of Western Europe, party dealignment seems to have stimulated a number of electoral reforms such as the opening of (quasi-) closed-list systems, the striving for more demographic variation on the ballot list, the organization of new sub-local elections and the introduction of common lists by a pre-electoral coalition of traditional parties and their newer contenders. These electoral reforms might serve to strengthen the position of the traditional parties in their electoral struggle with new contending parties both at the left and the right. However, they might also strengthen the position of individual candidates to the detriment of the parties. In many list systems of proportional representation, the number of preferential votes increases, whereas the number of party votes decreases. However, this tendency toward more preferential votes is only detrimental for the parties if they do not control which of their candidates obtain the most preferential votes. We therefore tested whether the implementation of electoral reforms has led to increased voting on the basis of IBVEAs as opposed to PBVEAs. Our analysis of the voting behavior in the Belgian district council elections of 2000 and 2006 indeed revealed an increase in the number of preferential votes of 8.2 per cent points. Yet, at the same time we also established that these reforms have only strengthened the relative importance of PBVEAs, particularly ballot position. Neither the fact that voters have more influence on r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

161

Thijssen

Table 3: Which candidates obtain a lower/higher log percentage of the preferential votes in their constituency? N=2157 N(2000)=1201; N(2006)=956

Ballot position General effect (baseline)

Ballot position Focus on specific positions

Model 8

Constant Time 2006 (Ref. 2000) Minority status Female Ethnic

Model 8bis

b

SE

0.94*** 0.48***

0.09 0.11

0.34*** 0.84***

0.05 0.15

b

Model 8ter

SE

b

SE

1.08*** 0.08 1.08*** 0.08 0.44*** 0.09 0.42*** 0.09

0.40*** 0.04 1.05*** 0.15

0.41*** 0.04 1.04*** 0.19

Minority status * Time 2006 Female 2006 Ethnic 2006

0.15* 0.29

0.07 0.18

0.17** 0.391

Incumbents 2006 Ballot position Ballot position * Time 2006

0.35*** 0.04*** 0.01

0.08 0.00 0.00

0.21** 0.07 0.21** 0.07 0.04*** 0.00 0.04*** 0.00 0.011 0.00 0.011 0.00

Specific ballot positions First on the list (‘puller’) Last on the list (‘pusher’) First woman candidate First ethnic candidate

— — — —

— — — —

1.71*** 1.52*** 0.44*** 0.24

Specific ballot positions * Time 2006 First on the list (‘puller’) 2006 Last on the list (‘pusher’) 2006 First woman candidate 2006 First ethnic candidate 2006

— — — —

— — — —

— — — —

Panel candidate (2000 and 2006)

0.40***

Including controls fixed effects Party, district Pseudo R2

0.21

0.04

—

0.06 0.18** 0.15 0.371

0.08 1.76*** 0.07 1.45*** 0.08 0.41*** 0.14 0.22

— — — —

0.11 0.17 0.08 0.04

0.06 0.22

0.10 0.10 0.10 0.25

0.15 0.15 0.15 0.30

0.22*** 0.04

0.22*** 0.04

0.34

0.34

—

—

Note: Nested Tobit regressions based on pooled unbalanced panel data; 1P o0.1; *P o0.05; **Po0.01; ***Po0.001.

the seat allocation (opening of flexible list system) nor the fact that the list of candidates must be more demographically diverse (implementation of gender parity law), nor the fact there are incumbents (new elections in 2000), nor the fact that there are common lists have led to an increasing importance of IBVEAs. 162

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

These findings make at least three contributions that go well beyond the quasi-experimental setting of the Belgian district council elections. First, our study is a good illustration of the meaningfulness of using revealed electoral preferences (actual individual election results) instead of stated preferences (survey response) as an alternative to study voting motivations. Future research could try to integrate other variables such as individual campaign spending or associational memberships in similar supply-side driven analyses. Second, our findings provide support for the fact that institutional personalization, for example further opening of a flexible list system, may not automatically lead to an increasing importance of IBVEAs (personalization in the voting behavior of the public). Similar research in the context of other ‘opening’ flexible list PR-systems such as Austria or Sweden can point out whether media personalization is a necessary go-between in this respect. Last but not least, our study provides strong support for a conceptual distinction between preferential voting and preferential voting based on individual attributes of the candidates. In this respect, our electoral findings parallel the work of sociologists such as De Beer (2007) who insist that increased freedom of choice not necessarily leads to the heterogenization of individual choices and motives. To the extent that individualization stimulates voters to affirm their individual autonomy, it is understandable that they no longer go for the standardized party vote, but instead choose a preferential vote. However, in the end they often still rely on partisan clues and hence ‘prefer’ the same candidates. Similarly, consumers in a supermarket adore an extensive assortment but nonetheless they often end up buying the same attractively displayed ‘top shelf ’ product.

Acknowledgements The author would like to acknowledge the helpful suggestions by two anonymous reviewers, and to thank Sandra Sliwa for her help with the data collection.

About the Author Peter Thijssen is an Associate Professor in Political Sociology in the Department of Political Science and member of the research group Media, Movements and Politics (M2P) at the University of Antwerp, Belgium. His research focuses on public opinion, political participation and intergenerational solidarity. Recently, he has published in European Journal of Social Theory, British Educational Research Journal and Res Publica. r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

163

Thijssen

Notes 1 Our concept of IBVEAs is inspired by the concept of ‘personal vote-earning attributes’ (PVEAs) as defined by Shugart et al (2005). However, although they use PVEAs as dependent variables and explain in which circumscriptions the PVEAs will be most important, we use IBVEAs as independent variables. Moreover, because we explicitly want to compare individual-based attributes with party-based attributes, we have reserved the acronym PVEA for the party-based attributes. 2 Political commentators in Belgium commonly use the term ‘cartel list’ when referring to one or more parties presenting their candidates on a ‘common list’. However, because this particular pre-electoral coalition strategy is unrelated to the concept of ‘Cartel Party’, we did not use this term. 3 The district councils in Antwerp control the following policy areas: youth, seniors, sport, traffic safety and last but not least infrastructure. 4 Alternatively, due to the novelty of the 2000 elections, voters could have been more attentive to the attributes of the individual candidates. However, this expectation is contradicted by the fact that the proportion of preferential voters (as compared with party voters) has clearly risen in the 2006 elections. Furthermore, we expect this novelty effect to be very limited in a context of concurrent elections. 5 The Social Democratic Party and the Green Party submitted common lists in four of the nine electoral districts (Berendrecht-Zandvliet-Lillo, Borgerhout, Deurne and Merksem). 6 Although we are dealing with population data, standard errors still give some information on the reliability of the regression coefficients because most likely our model does not include all relevant variables. 7 Because the relative size of a party in the previous election is not available in 2000 and the party landscape in 2006 was fundamentally different from that in 2000, we opted for a categorical party label. 8 The b coefficient for a dummy variable in a semi-logarithmic regression corresponds with a 100*[exp(b(Var(b)/2))1] percentage change in the dependent variable. 9 Given that the dichotomous variable ‘time’ (T) equals 1 for the election of 2006, the regression equation Y ¼ b0 þ b1X þ b2T þ b3XT could be rewritten as Y ¼ (b0 þ b2) þ X(b1 þ b3). Consequently, we do not have to take into account the marginal effect b2 of the conditioning variable T in order to assess the conditional effect of being a candidate of the Social Democratic Party (X) in the election of 2006 (T ¼ 1). Similarly, it is not necessary to integrate the variance of the marginal effect of the conditioning variable election time T in the computation of the standard error of the conditional effect b1 þ b3. However, we did take into account the covariance of the 1 2, in casu b ) ¼ [Var(b ) þ Var(b ) þ 2.Cov(b b )] estimated effect parameters b1 and b3; SE(b1 þ 3 1 3 1 3 1

SE byx.t ¼ 1 ¼ [0.008 þ 0.016 þ 2(0.008)]2 ¼ 0.09.

10 We tested whether the social democrats were more inclined to introduce a common list with the greens in districts where they were relatively less successful in 2000, but this is not the case. 11 Due to perfect collinearity, it is impossible to introduce the interactions ‘common list * Social Democratic Party’ and ‘common list * Green Party’ in the same model.

References Benoit, K. (2004) Models of electoral system change. Electoral Studies 23(3): 363–389. Brambor, T., Clark, W.R. and Golder, M. (2005) Understanding interaction models: Improving empirical analyses. Political Analysis 14(1): 1–20. 164

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

Are parties stimulating candidate-centred voting?

Bra¨uninger, T., Brunner, M. and Da¨ubler, T. (2012) Personal vote-seeking in flexible list systems: How electoral incentives shape Belgian MP’s bill initiation behavior. European Journal of Political Research 51(5): 607–645. Cain, B.E., Dalton, R.J. and Scarrow, S.E. (eds.) (2003) Democracy Transformed? Expanding Political Opportunities in Advanced Industrial Democracies. Oxford: Oxford University Press. Carey, J.M. and Shugart, M.S. (1995) Incentives to cultivate a personal vote: A rank ordering of electoral formulas. Electoral Studies 14(4): 417–439. Carson, J.L., Engstrom, E.J. and Roberts, J.M. (2007) Candidate quality, the personal vote, and the incumbency advantage in congress. American Political Science Review 101(2): 289–301. Dalton, R.J. and Wattenberg, M.P. (eds.) (2002) Unthinkable democracy: Political change in advanced industrial democracies. Parties without Partisans: Political Change in Advanced Industrial Democracies. Oxford: Oxford University Press, pp. 3–16. De Beer, P. (2007) How individualized are the Dutch? Current Sociology 55(3): 389–413. Deschouwer, K. (2004) Political parties and their reactions to the erosion of voter loyalty in Belgium: Caught in a trap. In: P. Mair, W.C. Mu¨ller and F. Plasser (eds.) Political Parties and Electoral Change. London: Sage, pp. 179–206. Desposato, S.W. and Petrocik, J.R. (2003) The variable incumbency advantage: New voters, redistricting and the personal vote. American Journal of Political Science 47(1): 18–32. De Winter, L. (2005) Belgium: Empowering voters or party elites? In: M. Gallagher and P. Mitchell (eds.) The Politics of Electoral Systems. Oxford: Oxford University Press, pp. 417–432. Farrell, D.M. and McAllister, I. (2006) Voter satisfaction and electoral systems: Does preferential voting in candidate-centred systems make a difference? European Journal of Political Research 45(5): 723–749. Golder, S.N. (2006) Pre-electoral coalition formation in parliamentary democracies. British Journal of Political Science 36(2): 193–212. Gschwend, T. and Hooghe, M. (2008) Should I stay or should I go? An experimental study on voter responses to pre-electoral coalitions. European Journal of Political Research 47(5): 556–577. Gujarati, D.N. (2004) Basic Econometrics. New York: McGraw-Hill. Hibbing, J.R. and Theiss-Morse, E. (2002) Stealth Democracy: Americans’ Beliefs about How Government should Work. Cambridge, UK: Cambridge University Press. Karvonen, L. (2010) The Personalization of Politics: A Study of Parliamentary Democracies. Colchester, UK: ECPR Press. Katz, R.S. (1986) Intraparty preference voting. In: B. Grofman and A. Lijphart (eds.) Electoral Laws and their Political Consequences. New York: Agathon Press, pp. 85–103. Kersting, N. and Vetter, A. (2003) Reforming Local Government in Europe: Closing the Gap between Democracy and Efficiency. Opladen, Germany: Leske & Budrich. Krebs, T.B. (1998) The determinants of candidates’ vote share and the advantages of incumbency in city council elections. American Journal of Political Science 42(3): 921–935. Levi, M. and Stoker, L. (2000) Political trust and trustworthiness. Annual Review of Political Science 3(1): 475–507. Lowndes, V. and Sullivan, H. (2008) How low can you go? Rationales and challenges for neighbourhood governance. Public Administration 86(1): 53–74. Marsh, M. (1985) The voters decide? Preferential voting in European list systems. European Journal of Political Research 13(4): 365–378. Marsh, M. (2007) Candidates or parties? Objects of electoral choice in Ireland. Party Politics 13(4): 500–527. Matland, R.E. and Studlar, D.T. (1996) The contagion of women candidates in single-member district and proportional representation systems: Canada and Norway. Journal of Politics 58(3): 707–733. r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166

165

Thijssen

Norris, P. (2004) Electoral Engineering: Voting Rules and Political Behavior. Cambridge, UK: Cambridge University Press. Oliver, E. and Ha, S.E. (2007) Vote choice in suburban elections. American Political Science Review 101(3): 393–408. Pilet, J.-B. and Bol, D. (2011) Party preferences and electoral reform: How time in government affects the likelihood of supporting electoral change. West European Politics 34(3): 568–586. Rahat, G. and Sheafer, T. (2007) The personalization(s) of politics: Israel, 1949–2003. Political Communication 24(1): 65–80. Schlesinger, J.R. (1966) Political Ambition. Chicago, IL: Rand Mcnally. Shomer, Y. (2009) Candidate selection procedures, seniority, and vote-seeking behavior. Comparative Political Studies 42(7): 945–970. Shugart, M.S., Valdini, M.E. and Suominen, K. (2005) Looking for locals: Voter information demands and personal vote-earning attributes of legislators under proportional representation. American Journal of Political Science 49(2): 437–449. Tavits, M. (2009) The making of mavericks: Local loyalties and party defection. Comparative Political Studies 42(6): 793–815. Teney, C., Jacobs, D., Rea, A. and Delwit, P. (2010) Ethnic voting in Brussels: Voting patterns among ethnic minorities in Brussels (Belgium) during the 2006 local elections. Acta Politica 45(3): 273–297. Thijssen, P. (2007) Het IMBY-syndroom: de (on)zin van de Antwerpse districtraadsverkiezingen. Res Publica 49(1): 175–193. Van Assche, D. and Dierickx, G. (2007) The decentralisation of city government and the restoration of political trust. Local Government Studies 33(1): 25–47. Van Holsteyn, J. and Andeweg, R. (2010) Demoted leaders and exiled candidates: Disentangling party and person in the voter’s mind. Electoral Studies 29(4): 628–635. Verba, S., Schlozman, K.L. and Brady, H.E. (1995) Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press. Wauters, B., Weekers, K. and Maddens, B. (2010) Explaining the number of preferential votes for women in an open list-PR system: The 2003 federal elections in Flanders (Belgium). Acta Politica 45(4): 468–490.

166

r 2013 Macmillan Publishers Ltd. 0001-6810

Acta Politica

Vol. 48, 2, 144–166