It’s not only about the leader: Oligarchized personalization and preference voting in Belgium

Abstract

Research on the electoral personalization of politics has stressed a trend towards a greater role of top prominent political figures (party leaders and ministers). This trend was described as centralized electoral personalization. Yet, this trend is merely one side of a more complex story. No leader attracts all voters’ support, and other candidates manage to stand out despite lower resources and visibility. Using a unique dataset of 47,239 actual ballot papers cast for the 2018 Belgian local elections, we show that candidate-level, list-level and district-level factors result in distinct preference voting behaviour. While these factors lead to unmistakable forms of (de-)centralized personalized forms of elections, we furthermore, show that intermediary situations distinctively emerge. A significant number of ‘subtop’ candidates stand out among candidates, by attracting support from voters who do not support the mere leader of the list. This ‘oligarchized personalization’ would deserve greater attention in the literature.

Keywords

Personalization intra-party competition leadership elections candidates

Introduction

One of the prominent trends in contemporary democracies has been the growing personalization of politics. This evolution – defined as ‘the notion that individual political actors have become more prominent at the expense of parties and collective identities’ (Karvonen, 2010: 4) – has been affecting various facets of politics (Cross et al., 2018; Rahat and Kenig, 2018). Yet, some authors have argued that personalization has not affected all politicians identically. Party leaders were the great winners of this trend, resulting in a ‘centralized personalization’ of politics – as opposed to ‘decentralized personalization’ (Balmas et al., 2014). It has translated in a rich literature looking at the importance of leaders, especially in elections and within political parties (Aarts et al., 2011; Bittner, 2011; Clarke et al., 2004; Clarke et al., 2009; Kriesi, 2012; Garzia 2012; Lobo and Curtice, 2014; Pilet and Cross, 2014; Poguntke and Webb, 2005; Wauters et al., 2018).

Another line of research has shown that the centralized personalization of politics, and of elections in particular, is merely one side of a more complex story. Party leaders may have grown in importance, but they do not fully dominate election processes. Several studies have shown that some voters simply do not care about the party brand or its leader (Marsh 2007; Van Holsteyn and Andeweg 2010, Mattes and Milazzo 2014). For those voters, the main driver behind electoral choice is the personality of local candidates. Beyond the few political leaders, other politicians still find their way and attract voters’ attention (30; Holtz-Bacha,et al., 2014: 164).

In this wake, a few studies have shown that the type of electoral systems, and especially the type of preferential-voting list PR systems, critically affect the concentration of votes on a few top candidates and/or the spread of preference votes across a wider spectrum of less prominent candidates (Dodeigne and Pilet 2021). In particular, it appears that “subtop” politicians – defined as politicians who are not the main leader of their party but who still enjoy a good visibility among voters (like less senior minister at national level, or deputy-mayor at local level) – remain crucial actors in the electoral personalization of politics. Those “subtop” politicians contribute to a form of personalization that is situated between “centralized” and “decentralized” electoral personalization. Dodeigne and Pilet (2021) defined this trend as “oligarchized” electoral personalization. In electoral systems where voters are allowed to cast multiple preference votes, “subtop” politicians can counter the predominance of leaders (i.e. centralized personalization). While a few studies have sought to identify factors explaining (de)concentration of preference votes (Wauters, 2021), a systematic analysis is still lacking. In this study, we analyze how preference votes are spread (or not) across multiple candidates at local Belgian elections where multiple preference voting is allowed. We test how candidates-level, lists-level and districts-level factors affect the dispersion of preference votes across candidates. In other words, we seek to identify the factors that resist the overarching trend towards centralized electoral personalization. Within the symposium in which this article is published, our study looks at the electoral stage of intraparty competition, and examines the distribution of intraparty preference votes as the results of the strategies of both candidates and voters.

For that purpose, we rely on a unique dataset from the 2018 Belgian local elections (Wallonia). Local elections in Wallonia are organized under a PR system with open lists. It means that voters first decide which party list they want to support, and then they can mark their preference for one or several candidate(s) within the list. Seats are then allocated between lists, and then within lists solely on the basis of the preference votes received by each candidate. Ballot position has no effect on the allocation of seats within list, even if parties can still decide which candidates occupy which position on the list. For these elections, we have been – for the first time ever – authorized by the regional administration to directly record official ballot papers cast by voters. The direct access to real paper ballots is extremely rare for researchers. A few earlier studies used mock ballot in survey research (Jacobs et al., 2014; Erzeel and Caluwaerts, 2015; Marien et al., 2017; Janssen, 2020), but with all difficulties related to correct reporting and limited number of observations for systematic analysis. In this study, we have access to more reliable data from real electoral ballots cast by 47,239 actual voters, covering 4,906 candidates on 188 lists of candidates across 49 municipalities. Each of those municipalities constitutes a distinct electoral race, with a few lists running and one leader for each list.

The article is structured as follows. We start by a review of past research in order to identify the main factors explaining the candidates’ electoral performance thanks to a concentration of votes on their single name or rather multiple preference votes cast in combination with other co-partisan candidates. Second, we present in detail the Belgian case and its electoral system that is covered in this article. We then turn to the presentation of our data and method. The fourth section presents and discusses our main findings. And we then conclude by discussing their implications for the scholarly debates on the personalization of elections.

Theoretical framework and hypotheses

The literature on preference voting under list PR systems is already well consolidated, especially in Netherlands and Belgium (see Wauters et al., 2020 for a review), but also in some other countries (Katz and Bardi, 1980; Wildgen, 1985; Holli and Wass, 2010; Christensen et al., 2021; Söderlund et al., 2021). In a systematic comparative study using candidates’ electoral results, Dodeigne and Pilet (2021) examined electoral intraparty competition in four countries using list PR systems allowing for intraparty choice, namely Belgium, the Czech Republic, Finland and Luxembourg. One of their core findings is that the electoral system, and especially the number of preference votes that the electorate can cast, is a key factor in broadening the scope of intraparty competition. When voters are allowed to cast more preference, they tend to use this opportunity to spread their preference votes. It contrasts with systems that allow to cast a single preference vote in which incentives to concentrate that vote on the party leader are greater. In the same vein, Wauters and his colleagues (2015) have shown that voters could be divided between those casting a vote for the leader only, those casting multiple votes for the leader a few candidates, voters casting preference votes for one or several less visible candidates. The latter two behaviours concur to resisting the full centralized electoral personalization under multiple preference voting.

Yet, we still know little about what factors other than the electoral system are associated to single or rather to multiple preference voting. In one of the rare existing studies, Wauters, and colleagues (2021) have shown that the age of the party is a key factor has voters cast more often single preference vote for the leader when the party is newer, and has therefore less well-known candidates on the list. We will therefore build on earlier studies on preference voting to develop a set of hypotheses on factors concurring to single versus multiple preference voting. We present our hypotheses at three levels of analysis: (1) candidate-level, (2) list-level and (3) district/municipality-level.

For the candidate-level factors, we hypothesize three types of candidates’ characteristics affecting preference voting behavior, but in diverging directions. First, some characteristics seem to contribute to single preference voting, i.e. specific candidates manage to perform at elections thanks to their capacity to attract votes under their own candidacy only. It is especially the case for leadership characteristics such as party leaders and cabinet members. In particular, when a prominent leader – like a minister or the party leader – is on the list, it leads to more voters voting for that candidate only (Wauters et al., 2018), and to a stronger concentration of preference votes within the list (Dodeigne and Pilet 2021). We might therefore expect that being a prominent politician would lead to attract more votes from single preference voters. In the context of local elections, it would apply to two sorts of prominent politicians: the incumbent local mayor as well as national or regional politicians (MPs, ministers, and party leaders) running for local elections.

Mayors attract more single preference votes than multiple preference votes.

National and regional politicians attract more single preference votes than multiple preference votes.

Complementary to these highly prominent figures, we, furthermore, expect that lower prominent characteristics (as incumbent local councillors or aldermen) would still play out but less markedly. There is indeed a wide literature showing that incumbents tend to attract more preference votes (Van Holsteyn and Andeweg, 2010; Thijssen, 2013; Maddens and Put 2013; Górecki and Kukołowicz 2014). Hence, Aldermen (deputy-mayors) and local councillors should also be able to play out their incumbency card. However, their incumbency advantage would be less strong than for mayor or national and regional politicians. They could therefore also attract preference votes but potentially less often single preference votes. Voters who just want to vote for such candidates and do not consider other candidates like upper levels politicians should be rarer.

Incumbent aldermen and incumbent local councillors attract less single preference votes.

Second, we consider as another influential candidate’s characteristic that affects preference voting the position of candidates on the list as attributed by the party. Several studies have shown that candidates that are positioned higher on the list benefit from a competitive advantage over other candidates placed at lower positions on the ballot (Van Erkel and Van Aelst 2016; Söderlund et al., 2021). Many voters use list position as a shortcut to identify more competent candidates (Devroe and Wauters, 2020). Futhermore, candidates positioned higher on the list tend to enjoy greater visibility in the media (Van Erkel et al., 2017). Therefore, we expect that these candidates positioned higher on the ballot will benefit from a greater competitive advantage which will result in a greater capacity to attract single preference votes. By contrast, candidates positioned lower on the list tend to stand out only via multiple preference voting.

Candidates positioned higher on the list attract more single preference votes.

Third, research has identified that candidates from some specific social groups may attract more preference votes. It is especially the case for female candidates and for candidates from ethnic minorities (Teney et al., 2010; Marien et al., 2017). Such candidates may gain electoral support from voters who seek to boost the election of candidates from groups that are often under-represented in representative institutions. Yet, such voting behavior would rather be associated to group voting, i.e. supporting multiple female candidates or ethnic minority candidates. There is no reason for voters with such motivations to cast a vote for one candidate only. And indeed, research has shown some forms of block voting for female or ethnic minority candidates (Jacobs et al., 2014; Marien et al., 2017; Janssen, 2020.¹). Therefore, we expect that female candidates and ethnic minority candidates will be more often associated to multiple preference voting, and therefore to concur to deconcentrating intraparty competition.

Female candidates attract more multiple preference votes.

Candidates from ethnic minorities attract multiple preference votes.

Next to candidates’ characteristics, we expect that list-level factors are influential for the casting of single or multiple preference votes. Wauters and colleagues (2018) have demonstrated that established parties were associated with a lower concentration of preference votes on their lists. It could be due to their greater number of incumbents (see H3), but also to their party organization. In comparison to established parties, newer parties present organisations that are less structured. This tends to favour single preference voters, especially towards prominent candidates as the latter benefit from a greater margin of actions during the electoral campaign. Furthermore, we hypothesize that a list party magnitude also impacts voting behaviour in terms of strategic considerations (Thijssen, 2013). In comparison to smaller party lists, larger party lists favor the electoral emergence of a greater number of candidates, which provide greater incentives for voters to cast multiple preference votes for a wider share of potentially successful candidates (see also Dodeigne and Pilet, 2021). On the opposite, voters are strategically incited to support a limited number of candidates on smaller party lists (if not only their preferred candidate), as hardly a few candidates will get access to office.

Candidates running on established party lists attract fewer single preference votes than multiple preference votes.

Candidates running on smaller party lists (in terms of seats gained) attract fewer single preference votes than multiple preference votes.

Finally, we expect that the municipality context matter. Dodeigne and Pilet (2021) have shown that races in smaller districts tend to show higher levels of concentration of preference votes. In the specific context of local elections, this should be reinforced the diversity of political dynamics between large urban centres and small rural municipal contexts. The latter, which are characterised by less densified territorial environment, the seminal “friends and neighbor” effects are the strongest, which results in a larger voters’ familiarity with candidates. This results in greater preference voting in smaller municipalities (André et al., 2012; Put and Maddens, 2015). On the opposite, large urban municipalities tend to favor electoral subcommunities, splitting electoral linkages between candidates and voters into specific neighborhoods and urban communities (Dodeigne et al., 2021). In this context, we expect that voters will specifically target certain candidates, resulting in a greater share of single preference votes.

Candidates running in larger municipalities attract greater single preference votes than multiple preference votes, and vice versa for candidates in smaller municipalities.

Case study: the 2018 local elections in Wallonia (Belgium)

Local elections in Wallonia (Belgium) are organized using list PR systems allowing voters to cast preference votes in 253 municipalities. In each of them, elections are held every six years to elect members of the local council. Each municipality constitutes a single electoral district. The number of local councilors to be elected ranges between 7 in the smallest towns up to 47 in the largest ones such as Liège. On the day of elections, voters must first decide the party lists they want to support. Then, within the list of their choice, voters may either cast a list vote by marking a vote in the box situated on top of the list (i.e. no preference votes towards any candidates of the list), or they can cast a preference vote for one or multiple candidates (up to the number of seats to be allocated).

The allocation of seats to the local council proceeds in two steps. First, seats are allocated between party lists. Each ballot marked by a vote for a list (either with a list vote or with one or several preference votes) is counted as one vote for the list. On the basis of the number of votes received by each list, seats are allocated via the Imperiali method of seat allocation (with successive divisors starting with 1, 1.5, 2, 2.5;…). This method is one of the variants of proportional representation and tends to slightly favor larger lists (Pilet et al., 2020). Then, seats are allocated to candidates within lists according to the number of seats obtained by each of these lists. Since the 2018 elections, the Walloon local electoral system uses an open list system. List votes do not count for the allocation of seats within lists. Parties can determine the position candidates occupy on the ballot but this position does not affect the allocation of seats within lists. And the seats are attributed to the candidates with most preference votes by decreasing order until all seats are allocated.

The structure of party systems and party competition greatly varies across the 253 municipalities. In larger urban municipalities, the local party systems tend to mirror the national party system, with four to five parties running under their national party brand, leading to local executive coalitions. In smaller rural municipalities, party competition tends to be restricted to 2 or 3 lists, with fewer lists running under national party labels and more ‘pure’ local lists (Dodeigne et al., 2019), and often with one list gaining a majority of seats. Overall, the large diversity of local party systems and electoral competition offer a fruitful empirical ground to test the effects of our varying factors presented above.

Data and method

For this study, we use a unique dataset of official electoral ballots that were cast on the 2018 local elections. For the first time ever in Belgium, authorities have granted access to ballots. We were allowed to extract a sample of 47,239 ballots for which preference votes have been cast. The sampling procedure was organized in four steps. First, we created a representative sample of 49 Walloon municipalities of the 253 municipalities (see Table 1) based on two main criteria (population size, degree of urbanization based on Eurostat classification) while controlling for the structure of local party systems (according to the comprehensive analysis of Close et al. 2020). Secondly, within each municipality, we randomly drew several vote-counting stations. Votes counting stations are the bureaus where electoral votes from three polling stations are stored, merged, and counted by election officials. Within each municipality, the number of votes counting stations we selected varies from 2 to 9, depending on the population size of a municipality. Thirdly, we randomly selected large bags in which ballot papers are stored. And, within those bags, we draw a sample of ballots. Finally, we completed the dataset of coded ballots with variables at the candidate, list and the municipality levels (e.g. type of parties, size of the municipality, profile of candidates, see below).

Table 1.

Distribution of the sample of municipalities, by province and level of urbanization.

Provinces level of urbanization	Target population		Sample of municipalities
Provinces level of urbanization	N	%	N	%
Hainaut	69	27	11	22
Urban municipalities	2	1	1	2
Mix municipalities	40	16	4	8
Rural municipalities	27	11	6	12
Liege	75	30	16	33
Urban municipalities	7	3	1	2
Mix municipalities	31	12	7	14
Rural municipalities	37	15	8	16
Luxembourg	44	17	8	16
Mix municipalities	5	2	2	4
Rural municipalities	39	15	6	12
Namur	38	15	8	16
Urban municipalities	1	0	1	2
Mix municipalities	6	2	2	4
Rural municipalities	31	12	5	10
Brabant wallon	27	11	6	12
Mix municipalities	18	7	4	8
Rural municipalities	9	4	2	4
Total	253	100	49	19

This dataset allows us to build our dependent variable. In order to account for the factors that impact the (de)concentration of preference votes on certain candidates, we operationalize our dependent variable as the percentage of preference votes obtained by candidates alone and the percentage of votes they receive with other co-partisan candidates. Table 2 shows that a majority of voters decided to cast a single preference vote (52.2%). The percentage of voters casting multiple preference decreased as to the number of preferences votes increases: two preference votes (15.6%), three preference votes (10.4%), four and five votes (9.3%), and voters casting six or more preference votes (12.6%). These numbers tend to confirm the main trends observed in former electoral surveys (but conducted at upper levels of government): such studies observed that between one-quarter and one third of voters cast only a single preference votes, while very few candidates cast more than 5 preference votes (see André et al., 2017).

Table 2.

Distribution of preference votes cast according to different voting categories.

	1 PV	2 PV	3 PV	4 PV	5 PV	6+ PV	Total
%	52.2	15.6	10.4	5.7	3.6	12.6	100
N	24,654	7352	4895	2689	1690	5,959	47,239

Key: PV = preference votes.

We distinguish, therefore, three profiles of preference voters: single-preference voters, limited multiple preference voters (between 2 and 5 preference votes), and extensive multiple preference voters (supporting 6 and more candidates). Our dependent variable is operationalized accordingly for the 4,906 candidates that are part of our representative sample: (1) percentage of preference votes obtained by a candidate individually, (2) percentage of preference votes obtained with two, three, four or five other candidates, (3) percentage of preference votes obtained with six and more other candidates. Table 3 shows that candidates can rarely claim that their electoral success is due to their mere individual merit: on average, hardly 12 percent of their total number of preference votes is based on single preference votes, and 25 percent of their votes are obtained in combination with up to five co-partisan candidates. The electoral performance of most candidates’ reflects extensive multiple preference voting: 63 percent of their overall preference votes are obtained with six and more co-partisans. Interestingly, there is a significant variance between candidates. Hence, the standard deviation (13 percent) for the percentage of preference votes obtained individually is higher than the mean (12 percent), with minimum and maximum percentages covering the full score (from 0 percent to 100 percent). This is precisely this variance that we seek to explain in terms of candidates-level, list-level and district-level factors in the multivariate analysis in the next section.

Table 3.

Descriptive statistics for the dependent and independent variables.

Variables	Mean	St. Dev	Min	Max
Dependent variable
% pref. Votes obtained individually	0.12	0.13	0.00	1.00
% pref. Votes obtained with up to five other candidates	0.25	0.17	0.00	1.00
% pref. Votes obtained with six and more other candidates	0.63	0.22	0.00	1.00
Independent variables
Candidates' characteristics
Relative position of the candidates on the list (0–1)	0.51	0.29	0.02	1.00
Age of the candidate (continuous variable)	48.12	14.03	18.00	93.00
Sex of the candidates (1 = female candidates)	0.50
Nationality of the candidates (1 = foreign candidates)	0.01
Political experience (ref. category = no XP)	0.80
% mayors	0.01
% deputy mayors	0.05
% councillors	0.12
% cabinet & parliamentary XP	0.01
Lists' characteristics
Political experience (ref. category = Established parties)	0.45
% emerging statewide parties	0.31
% local parties	0.24
List of the opposition (1 = list in the opposition)	0.61
Party magnitude	6.72	4.60	0	23
Municipalities' characteristics
Log. Of population size	10.05	1.07	7.90	12.19
Effective number of parties	6.17	2.71	2.00	12.00

Regarding our independent variables (see Table 3), we gathered official electoral information published by public authorities with previous studies about candidates’ characteristics and party lists in competition at the 2018 Walloon elections (Close et al. 2020; Dodeigne, 2018; Dodeigne et al., 2020). The first six hypotheses are candidates-level factors. For officeholders (H1, H2, H3), we operationalized candidates’ political background as a categorical variable, where no experience at all is the reference category (80% of all candidates in our sample). Candidates that are serving at upper levels of government include members of regional and national parliaments and cabinet ministers. The list position (H4) is operationalized as a relative position according to the absolute number of candidates. This allows us to provide a unique standardized indicator of positions, irrespective of the list length (which greatly varies across municipalities). A higher score on this continuous variable (0-100 percent) indicates that candidates are positioned by their party at the bottom of the list (100 being the latest position), and vice versa for candidates positioned higher on the list². For sex (H5), we use official information to distinguish male and female candidates. Table 3 indicates a mean of 50 percent which perfectly reflects the electoral regulation that imposes gender parity since 2018 (see Pilet et al., 2020). As non-Belgian foreigners can be candidates at local elections, we use the candidates’ nationality as proxy for candidates’ ethnicity (H6)³.

The second set of hypotheses are related to party lists-level factors. Following H7, we first distinguish the local branches of established national party lists from emerging parties (respectively 45 and 31 percent). We also, account for local lists which cover 24 percent of all lists. Furthermore, party magnitude is operationalized as the total number of seats obtained by the party lists at the elections. As a control, we also identify list of the opposition which indicate that candidates were running on a list that was not part of the local majority before the 2018 elections.

The third and final set of variables are related to municipalities’ characteristics. On the one hand, size significantly differ across the 49 municipalities, with the largest municipality (Liège) being 73 times bigger than the smallest one (Tinlot), creating a strong skew to the left towards smaller municipalities (the median value being 18,552 inhabitants and the mean 42,150). We thus log transformed the variable. We also control for the structure of the local party system following Laakso and Taagepera (1979)’s seminal concept of effective number of parties in each of the 49 municipalities.

Findings

Considering the structure of the data (4,906 candidates are nested in 188 lists across 49 municipalities), we specified a multilevel linear regression where level-I covers candidates and level-II are lists nested in municipalities. Our model includes a varying intercept for our three dependent variables, i.e. the average percentage of preference votes obtained individually, with two or four co-partisans and five and more other co-partisans. In this regression equation, γ₀₀ is the intercept while γ_01, γ_02, γ_03, γ_04, γ₀₅ are the regression slopes for explanatory factors varying across electoral lists in municipalities (Incumbent majorities, ENP, Party types and magnitude as well as population size), while γ_10, γ_20, γ _30, γ _40, γ_50, γ₆₀ cover presence of specific candidates’ profile varying across electoral lists (nationality, sex, age, political capital and list position). Finally, ε_ij and δ_oj are the residual error terms. The subscript j is for the lists within municipalities (j = 1…J) and the subscript i is for candidates running on lists (i = 1…n_j)⁴.

{P e r c e n t a g e o f p r e f e r e n c e s v o t e s o b t a i n e d i n d i v i d u a l l y (o r w i t h o t h e r c a n d i d a t e s)}_{i j} = γ_{00} + γ_{01} {I n c u m b e n t m a j o r i t y}_{j} + γ_{02} {E N P}_{j} + γ_{03} {P a r t y t y p e s}_{j} + γ_{04} {P a r t y m a g n i t u d e}_{j} + γ_{05} Log . {p o p u l a t i o n}_{j} + γ_{10} N {a t i o n a l i t y o f c a n d i d a t e s}_{i j} + γ_{20} {S e x o f c a n d i d a t e s}_{i j} + γ_{30} {A g e}_{i j} + γ_{40} A g e^2_{i j} + γ_{50} {P o l i t i c a l m a n d a t e}_{i j} + γ_{60} {R e l a t i v e p o s i t i o n o n t h e l i s t}_{i j} + δ_{0 j} + ε_{i j}

The results of the multilevel linear models are presented in Table 4. It assesses the impact of candidates’, lists’ and municipalities’ characteristics upon the share of candidates’ preference votes obtained individually and with other co-partisans. First, we observe that candidates-level factors critically matter. As predicted by our first two hypotheses (H1 and H2), a candidate’s political background has a statistically significant enhancing effect upon the concentration of preference votes on his/her unique candidacy. Mayors, MPs and Cabinet members tend to attract substantially much more preference votes on an individual basis than other candidates. All other things being equal, these “high profile” candidates attract almost one third of all their preference votes thanks to their own names only. This is three times more than candidates without political experience (11.3 percent). In between these two extremes, we found the candidates with intermediary offices: deputy-mayors and councillors attract respectively 17.0 and 15.5% of single preference votes. These three profiles of candidates are also observed for the “extensive” multiple preference, with an almost perfectly reversed picture (see Figure 1(b)). In other words, lower political profiles entail larger share of preference votes obtained collectively and vice versa for the top candidates. Finally, the category “limited multiple preference votes” seems to offer a transition mode (Figure 1(c)). While the signs of the coefficient for the variable political background remain positive as for the single preference votes (i.e. enhancing effect of political experience on the percentage of preference votes obtained with a limited number of candidates), they substantially shift: the positive effects of mayoral background notoriously decrease while those for incumbent aldermen and councillors increase.

Table 4.

Multilevel linear model predicting share of candidates’ preference votes obtained individually and with other co-partisans.

	% Pref. votes obtained individually	% Pref. votes obtained with two, three or four co-partisans	% Pref. votes obtained with five and more co-partisans
Candidates' characteristics
Relative position of the candidates	−0.07^*** (0.01)	−0.10^*** (0.01)	0.16^*** (0.01)
Candidates' age	−0.001^*** (0.0001)	−0.0002 (0.0001)	0.001^*** (0.0002)
Female candidates	−0.03^*** (0.003)	−0.03^*** (0.004)	0.06^*** (0.005)
Candidates with a foreign nationality	−0.02^* (0.01)	−0.06^*** (0.01)	0.09^*** (0.02)
Candidates’ political background (reference category: no XP)
Mayors	0.20^*** (0.02)	0.12^*** (0.02)	−0.32^*** (0.02)
MPs and ministers	0.19^*** (0.02)	0.18^*** (0.02)	−0.37^*** (0.03)
Deputy mayors	0.06^*** (0.01)	0.14^*** (0.01)	−0.20^*** (0.01)
Councillors	0.04^*** (0.01)	0.10^*** (0.01)	−0.14^*** (0.03)
Lists’ characteristics
List of the opposition	−0.003 (0.01)	−0.03 (0.02)	0.03 (0.03)
Party type (ref. = established parties)
Emerging party lists	0.02 (0.01)	−0.01 (0.02)	−0.01 (0.03)
Local party lists	−0.01 (0.01)	−0.02 (0.02)	0.03 (0.03)
Party magnitude	−0.01^*** (0.001)	0.004^* (0.002)	0.004 (0.003)
Municipalities’ characteristics
Municipality size (log. Population)	0.04^*** (0.01)	−0.03^** (0.01)	−0.01 (0.02)
Effective number of parties	−0.005^* (0.003)	0.01 (0.01)	−0.003 (0.01)
Constant	−0.14^* (0.06)	0.58^*** (0.11)	0.63^*** (0.16)
N candidates	4909	4909	4909
N lists	188	188	188
Log likelihood	3573	2961	1740
Akaike inf. Crit	−7112	−5887	−3446
Bayesian inf. Crit	−7001	−5777	−3335

Key: *p < 0.1;^**p < 0.05;^***p < 0.01.

Figure 1.

(a), (b) and (c). Marginal effects of political background on candidates’ concentration of preference votes according to single, limited and extensive preference voting categories.

Those “sub-top” politicians attract substantially more votes than inexperienced candidates and concentrate a fair share of preference votes cast on their unique candidacies or with a restricted number of co-partisans. Yet, they do not manage to concentrate most of their preference votes on their mere names as do mayors, MPs and ministers. By contrast, candidates without any political capital are those who obtained the greatest shares of preference votes when voters supported five or more other co-partisans. These lay candidates concur to deconcentrate intraparty competition and to a more decentralized form of personalization. In other words, the ‘forces resisting centralized personalization’ can be explicitly identified in the candidates’ political background confirming H1, H2 and H3. Yet, the results cannot be summarized as a mere dichotomous phenomenon between centralization or decentralization of personalized electoral competition. Complex and intermediary situations are clearly observable with middle-range candidates emerging in between these two ‘extreme’ personalized forms.

The same kind of logic is observed with the candidates’ relative position on the list (H4). We observe that this variable has a negative effect on the proportion of preference votes received from ballot with one or a limited multiple preference votes. By contrast, the relative position on the list has a positive effect on the candidates’ percentage of extensive multiple preference votes. In other words, candidates with lower positions on the lists concentrate more votes because of group voting. On the opposite, candidates with better positions on the list tend to secure a higher percentage of votes by their own name only. H4 is, therefore, confirmed.

Finally, the multivariate regressions confirm that female and ethnic minority candidates receive mostly preference votes from voters casting extensive multiple preference votes (H5 and H6). By contrast, these candidates receive relatively fewer votes from ballot with single preference votes. Because they struggle to concentre preference votes on their mere candidacies, female and ethnic minority candidates face more challenges to emerge vis-à-vis their co-partisan candidates. And contrary to some expectations, the higher proportion of multiple preference votes is not necessarily an electoral advantage for those female and ethnic minority candidates: they hardly benefit from “bloc voting” (i.e. voting only for female candidates or ethnic minority candidates). In our sample, we even observe that there are slightly more preference votes for ‘men only’ (2.8 percent of the voters), than for ‘women only’ (2.3% of the voters).

Our second set of hypotheses are related to the nature of the list on which candidates are running. We expect candidates campaigning on lists from establishing and emerging party lists (H7), and from smaller lists (H8) to present a relatively larger share of their preference votes from a single or a few preference votes. Here, our findings are less neat. While the sign of the coefficients for emerging parties are positive, showing that candidates from emerging party lists receive relatively more votes from single-preference voters, they are not statistically significant. It is also not significant for the negative sign of multiple preference votes. However, party magnitude shows clear and significant effects confirming H8: as party magnitude increases, the percentage of single preference obtained by candidates substantially decreases (Figure 2(a)). On the opposite, greater party magnitude favours extensive multiple preference votes, although the confidence intervals are significantly larger. Finally, our expectations vis-à-vis municipal-level factors are confirmed (H9). In larger municipalities, candidates receive a greater share of their preference votes thanks to single-preference votes whereas smaller municipalities favour the opposite. Results are significant in all models but for extensive multiple preference votes.

Figure 2.

(a) and (b). Marginal effects of party magnitude on candidates’ concentration of preference votes according to single and extensive preference voting categories

Discussion and conclusion

The empirical goal of this study was to determine the effects of candidates-level, lists-level and municipalities-level factors upon candidates’ capacity to earn votes on their own name only, or in support with other co-partisan candidates. By doing so, we seek to theoretically contribute to ongoing debates regarding centralized and decentralized electoral personalization.

Our findings have first shown a stark difference between candidates attracting mostly single-preference votes and those receiving mostly multiple preference votes. In line with the centralized personalization thesis, prominent political figures (mayors, MPs, ministers) stand up among other candidates and attract preference votes on their mere names. Candidates attracting mostly single preference votes are also favored by the higher position they occupy on their list. And they come from smaller parties with fewer other prominent politicians who cannot contest their leadership. By contrast, candidates that are supported by voters making an extensive use of preference voting have a very different profile. They are lower on their party list. They come from larger parties. And they can mobilize identity-based block voting based upon their gender and their nationality (other than Belgian). These findings indicate that a first set of factors contributing to resisting the overarching trend of centralized personalization by fostering multiple preference votes are related to identity politics, and to the electoral mobilization of sociodemographic traits related to traditionally underrepresented groups (Holli and Wass, 2010; Teney et al., 2010).

But the most original contribution of our study is probably related to the analysis of a third profile of candidates. In a context where Belgian voters tend to cast a limited number of preference votes (35.3 percent of all ballots examined), an intermediate profile of candidates stand up vis-à-vis top and lower political figures discussed above. Candidates attracting these voters are often “subtop” politicians. They can be ministers, MPs or mayors, but they are also very often deputy-mayors and local councilors. They are also placed in the highest positions on the list, albeit not necessarily on the very top position. And they come from a party who obtain more seats and have, therefore, a greater range of candidates with some political capital. We can therefore conclude that those candidates contribute to limiting a full centralized personalization of electoral competition, but with a rather different logic. They confirm the idea that there was something in between centralized and decentralized personalization (Balmas et al., 2014), what earlier studies have defined as an “oligarchization” or an “elitization” of politics (Lindqvist, 2018; Dodeigne and Pilet, 2021). Next to the main leaders, there are “subtop” politicians that have some visibility and that may survive in multiple preference voting systems by attracting support from voters who would not vote for them only, but for them and a few additional candidates, often the main leaders. They can make a career in the shadow of the most prominent politicians from their party, and contribute to a total concentration of electoral politics around (local) party leaders only. Such findings concur to debates in other settings, and especially to the necessity to pay more attention to those mid-level or “subtop” politicians in order to better understand how they build up their electoral and political capital, and how they might also contribute to resisting the centralized personalization of politics.

Finally, future research could further investigate the implications of our findings about differences in lists and parties. Parties develop specific electoral strategies during candidate selection processes in PR systems (Hazan and Rahat, 2010; Vandeleene et al., 2016). They try to reach an ideal equilibrium between different types of candidates who could appeal to different subsets of the electorate. Different profiles of candidates – with different levels of political seniority, with specific sociodemographic traits – result in different patterns of preference voting behaviour. Some attracted single-preference voters, other multiple preference voters. We have also seen that those patterns might different according to the type of party, or to the local context. The next logical step would be to go for analyses at list-level, revealing how the combination of different types of candidates within a single list might lead to different patterns of preference votes at list-level. This would contribute to unveiling the dynamics of multiple preference votes in relationship to the debate about the personalization of elections.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Fonds De La Recherche Scientifique - FNRS grant no 40003171.

ORCID iDs

J. Dodeigne

J.B. Pilet

Notes

Author biographies

Jeremy Dodeigne is an associate professor in political science at the University of Namur. His research areas cover the study of political representation in European multi-level democracies, personalization of politics and legislative behaviour. His work appears in journals such as West European Politics, Party Politics, American Behavioral Scientist, Regional & Federal Studies, Government & Opposition, and Representation.

Jean-Benoit Pilet is professor of political science at Université libre de Bruxelles (ULB). He works on elections, political parties, and democratic reforms. He has co-authored Faces on the Ballot. The Personalization of Electoral Systems in Europe (OUP, 2016, with Alan Renwick) and The Politics of Party Leadership (OUP, 2016, with William Cross).

References

Aarts

Blais

Schmitt

(eds) (2011) Political Leaders and Democratic Elections. Oxford: Oxford University Press.

André

Wauters

Pilet

J-B

(2012) It’s not only about lists: explaining preference voting in Belgium. Journal of Elections, Public Opinion & Parties 22: 293–313.

André

Depauw

Pilet

J-B

(2017) The nature of preference voting: Disentangling the party and personal components of candidate choice. In: Deschouwer

(ed) Mind the Gap. Political Participation and Representation in Belgium. London: Rowman & Littlefield, pp. 251–274.

Balmas

Rahat

Sheafer

, et al. (2014) Two routes to personalized politics: centralized and decentralized personalization. Party Politics 20(1): 37–51.

Bittner

(2011) Platform or Personality? The Role of Party Leaders in Elections. Oxford: Oxford University Press.

Christensen

Järvi

Mattila

, et al. (2021) How voters choose one out of many: a conjoint analysis of the effects of endorsements on candidate choice. Political Research Exchange 3(1): 1892456. DOI: 10.1080/2474736X.2021.1892456

Clarke

Sanders

Stewart

, et al. (2004) Political Choice in Britain. Oxford: Oxford University Press.

Clarke

Kornberg

Scotto

(2009) Making Political Choices: Canada and the United States. Toronto: University of Toronto Press.

Cross

Katz

Pruysers

(2018) The Personalization of Democratic Politics and the Challenge for Political Parties. Oxford: Oxford University Press.

10.

Devroe

Wauters

(2020) Does high on the ballot means highly competent? Explaining the ballot position effect in list-PR systems. Acta Politica 55(3): 454–471.

11.

Dodeigne

(2018) Who governs? The disputed effects of regionalism on legislative career orientation in multilevel systems. West European Politics 41(3): 728–753.

12.

Dodeigne

Jacquet

, et al. (2020a) Le scrutin communal : une affaire nationale ou locale. In: Dodeigne

Jacquet

Matagne

(eds). Les élections Locales du 14 Octobre 2018 en Wallonie et a Bruxelles: Une Offre Politique Renouvelée? Bruges: Vanden Broele, 32–57.

13.

Dodeigne

Teuber

(2021) Nationalisation of local party systems in Belgium (1976-2018): The combined effects of municipality size and parliamentary parties’ dominance. Local Government Studies 47(1): 100–125.

14.

Dodeigne

Pilet

J-B

(2021) Centralized or decentralized personalization? Measuring intra-party competition in open and flexible list PR systems. Party Politics 27(2): 234–245.

15.

Dodeigne

Pilet

J-B

Talukder

(2020b) Personnalisation des élections dans un contexte de fin de l’effet. In: dévolutifDodeigne

Jacquet

, et al. (eds) Les élections Locales du 14 Octobre 2018 en Wallonie et a Bruxelles : Une Offre Politique Renouvelée? Bruges: Vanden Broele, 175–215.

16.

Dodeigne

Put

G.-J.

Teuber

(2023) Is electoral intraparty competition more intense in urban areas? Electoral studies Online first.

17.

Dodeigne

Reuchamps

Jacquet

(2019) The attractiveness of local and national list labels: the role of socio-economic inequalities on the success of electoral lists”. Local Government Studies 46(5): 763–779.

18.

Dodeigne

Teuber

Vandeleene

(2020d) “La professionnalisation du scrutin communal de 2018 en Wallonie : loin d’une arene d’‘amateurs’, mais une professionnalisation en baisse”. In: Dodeigne

Jacquet

, et al. (eds) Les élections Locales du 14 Octobre 2018 en Wallonie et a Bruxelles: Une Offre Politique Renouvelée? Bruges: Vanden Broele, pp. 149–174.

19.

Erzeel

Caluwaerts

(2015) Is it gender, ideology or resources? Individual-level determinants of preferential voting for male or female candidates. Journal of Elections, Public Opinion and Parties 25(3): 265–283.

20.

Garzia

(2012) Party and leader effects in parliamentary elections: towards a reassessment. Politics 32(3): 175–185.

21.

Górecki

Kukołowicz

(2014) Gender quotas, candidate background and the election of women: a paradox of gender quotas in open-list proportional representation systems. Electoral Studies 36: 65–80.

22.

Hazan

Rahat

(2010) Democracy within Parties: Candidate Selection Methods and Their Political Consequences. Oxford: Oxford University Press.

23.

Holli

Wass

(2009) Gender-based voting in the parliamentary elections of 2007 in Finland. European Journal of Political Research 49(5): 598–630.

24.

Holtz–Bacha

Langer

Merkle

(2014) The personalization of politics in comparative perspective: Campaign coverage in Germany and the United Kingdom. European Journal of Communication 29(2): 153–170.

25.

Jacobs

Kelbel

Pilet

(2014) Les préférences politiques des électeurs issus de l’immigration lors des élections communales d’octobre 2012 à Bruxelles. In: Pilet

Dassonneville

Hooghe

, et al. (eds) L’électeur Local. Le Comportement électoral au Scrutin Communal de 2012. Brussels: Editions de l’Université de Bruxelles, pp. 127–147.

26.

Janssen

(2020) Shaping the (dis) advantage: the impact of partisan and demographic factors on ethnic minority candidates’ success in preferential voting systems. Evidence from the Brussels case. Journal of Elections, Public Opinion and Parties 32: 22–45.

27.

Karvonen

(2010) The Personalisation of Politics: A Study of Parliamentary Democracies. Colchester: ECPR Press.

28.

Katz

Bardi

(1980) Preference voting and turnover in italian parliamentary elections. American Journal of Political Science 24(1): 97–114.

29.

Kriesi

(2012) Personalization of national election campaigns. Party Politics 18(6): 825–844.

30.

Laakso

Taagepera

(1979) ‘Effective’ number of parties. A measure with application to West Europe. Comparative Political Studies 12(1): 3–27.

31.

Lindqvist

(2018) Personalization and elitization – preference voting in the Netherlands In: ECPR Joint Sessions of Workshops, Nicosia, 10–14 April 2018.

32.

Lobo

Curtice

(eds) (2014) Personality Politics? The Role of Leader Evaluations in Democratic Elections. Oxford: OUP.

33.

Maddens

Put

(2013) Office effects and campaign spending in a semi-open list PR system: the Belgian/Flemish federal and regional elections 1999–2010. Electoral Studies 32: 852–863.

34.

Marien

Schouteden

Wauters

(2017) Voting for women in Belgium’s flexible list system. Politics & Gender 13(2): 305–335.

35.

Marsh

(2007) Candidates or parties? objects of electoral choice in Ireland. Party Politics 13(4): 500–527.

36.

Mattes

Milazzo

(2014) Pretty faces, marginal races: predicting election outcomes using trait assessments of British parliamentary candidates. Electoral Studies 34(1): 177–189.

37.

Ohr

Dieter

Henrik

Oscarsson

(2001) Leader Traits, Leader Image, and Vote Choice. In Kees Aarts, André Blais, and Hermann Schmitt (eds), Political Leaders and Democratic Elections, Oxford: Oxford University Press, pp. 187–219.

38.

Pilet

J-B

Cross

(eds) (2014) The Selection of Political Party Leaders in Contemporary Parliamentary Democracies: A Comparative Study. London: Routledge.

39.

Pilet

J-B

Talukder

Sanhueza

, et al. (2020) La représentation politique des femmes au niveau communal en Wallonie apres les élections de 2018. In: Dodeigne

Jacquet

, et al. (eds) Les élections Locales du 14 Octobre 2018 en Wallonie et a Bruxelles : Une Offre Politique Renouvelée? Bruges: Vanden Broele pp. 193–216.

40.

Poguntke

Webb

(2005) The Presidentialization of Politics in Parliamentary Democracies. Oxford: Oxford University Press.

41.

Put

Maddens

(2015) The effect of municipality size and local office on the electoral success of Belgian/Flemish election candidates: a multilevel analysis. Government and Opposition 50(4): 607–628.

42.

Rahat

Kenig

(2018) From Party Politics to Personalized Politics? Party Change and Political Personalization in Democracies. Oxford: Oxford University Press.

43.

Söderlund

von Schoultz

Papageorgiou

(2021) Coping with complexity: ballot position effects in the Finnish open-list proportional representation system. Electoral Studies 71: 102330.

44.

Teney

Jacobs

Rea

, et al. (2010) Ethnic voting in Brussels: voting patterns among ethnic minorities in Brussels (Belgium) during the 2006 local elections. Acta Politica 45(3): 273–297.

45.

Thijssen

(2013) Are parties stimulating candidate-centred voting? The case of the Belgian district council elections 2000–2006. Acta Politica 48(2): 144–166.

46.

Thijssen

Jacobs

(2004) Determinanten van voorkeurstemproporties bij (sub-)lokale verkiezingen: de Antwerpse districtraadsverkiezingen van 8 oktober 2000. Res Publica : Tijdschrift Voor Politieke Wetenschappen 46(4): 460–485.

47.

Van Aelst

Maddens

Noppe

, et al. (2008) Politicians in the News: media or party logic? media attention and electoral success in the Belgian election campaign of 2003. European Journal of Communication 23(23): 193–210.

48.

Vandeleene

Dodeigne

De Winter

(2016) What do selectorates seek? A comparative analysis of Belgian federal and regional candidate selection processes in 2014. American Behavioral Scientist 60(7): 889–908.

49.

Van Erkel

Thijssen

Van Aelst

(2017) One for all or all for one: the electoral effects of personalized campaign strategies. Acta Politica 52(3): 384–405.

50.

van Erkel

Thijssen

(2016) The first one wins: distilling the primacy effect. Electoral Studies 44(44): 245–254.

51.

Van Holsteyn

Andeweg

(2010) Demoted leaders and exiled candidates: disentangling party and person in the voter’s mind. Electoral Studies 29(4): 628–635.

52.

Wauters

(2021) Intraparty competition and preference voting in local elections. Personalization at different levels? In: ECPR Virtual Joint Sessions, University of Namur (Namur, Belgium), May 2021.

53.

Wauters

Weekers

(2008) Het gebruik van de voorkeurstem bij de federale parlementsverkiezingen van 10 juni 2007. Res Publica 50(2): 49–88.

54.

Wauters

Weekers

Pilet

J-B

(2004) Het gebruik van de voorkeurstem bij de regionale en Europese parlementsverkiezingen van 13 juni 2004. Res Publica 46(2–3): 377–412.

55.

Wauters

Weekers

Maddens

(2010) Explaining the number of preferential votes for women in an open-list PR system: an investigation of the 2003 federal elections in Flanders (Belgium). Acta Politica 45(4): 468–490.

56.

Wauters

Van Aelst

Thijssen

, et al. (2015) Présidentialisation versus personnalisation? Clés de lecture à la baisse des votes préférentielles In: Décrypter l’électeur: le comportement électoral et les motivations du vote du 25 mai 2014. Lannoo campus, pp. 76–98.

57.

Wauters

Thijssen

Van Aelst

, et al. (2018) Centralized personalization at the expense of decentralized personalization. The decline of preferential voting in Belgium (2003–2014). Party Politics 24: 511–523.

58.

Wauters

Thijssen

Erkel

(2020) Preference voting in the low countries. Politics of the Low Countries 2(1): 77–106.

59.

Wildgen

(1985) Preference voting and intra-party competition in Italy: some new evidence on the communist-Christian democrat stalemate. The Journal of Politics 47(3): 947–957.