Abstract
This article investigates the effect of candidates exhibiting local personal vote-earning attributes (PVEA) on the aggregate party vote share at the district level. Previous research has often assumed that packing ballot lists with localized candidates increases the aggregate party vote and seat shares. We present a strict empirical test of this argument by analysing the relative electoral swing of ballot lists at the district level, a measure of change in party vote shares which controls for the national party trend and previous party results in the district. The analysis is based on data of 7527 candidacies during six Belgian regional and federal election cycles between 2003 and 2014, which is aggregated to an original data set of 223 ballot lists. The ordinary least squares (OLS) regression models do not show a significant effect of candidates exhibiting local PVEA on relative electoral swing of ballot lists. However, the results suggest that ballot lists do benefit electorally if candidates with local PVEA are geographically distributed over different municipalities in the district.
Introduction
This article examines how individual-level vote-earning characteristics promoting the ‘localness’ of candidates impact the party ballot list result at the district level. Previous literature identified a set of candidate traits as local ‘personal vote-earning attributes’ (PVEA): specific personal characteristics which provide voters with substantive cues to a politician’s local knowledge and competence to represent them, such as being a district native or having prior local-level political experience (Shugart et al., 2005). The more burdensome acquiring information on all parties and candidates running for election, the more rational voters will rely on this candidate localness as a cue in their voting decision. Especially in candidate-centred systems, these PVEA are crucial attributes for voters as a shortcut, for candidates to signal their credibility, and for political parties as a nomination criterion (Nemoto and Shugart, 2013).
But how do these objective candidate attributes affect the election outcome? Previous research demonstrates that prior local-level political experience significantly increases a candidate’s personal vote share (e.g. Gschwend and Zittel, 2015; Jankowski, 2016; Put and Maddens, 2015; Tavits, 2010). To our knowledge, however, there has not been a single empirical test to date of how PVEA affect the aggregate party vote share at the district level. 1 It is often conveniently assumed that parties in candidate-centred electoral systems will electorally benefit in terms of aggregate vote and legislative seat share if they pack their ballot lists with local office-holders.
In this contribution, we test the electoral effectiveness of this candidate nomination strategy by analysing data on Flemish candidates for six consecutive regional and federal elections in Belgium between 2003 and 2014. In addition, we present and empirically test a new theoretical argument as to how political parties can optimize nomination strategies by taking into account a maximal geographical distribution of localized candidates on their ballot lists. To this end, we build an original data set of 223 ballot lists from Flemish political parties, aggregating information on 7257 individual candidacies at the ballot list–level. 2
We run ordinary least squares (OLS) regression models with the Relative Electoral Swing (RES), which measures the shift in ballot list vote shares taking into account the general party trend and previous party result in the district, of ballot lists as dependent variable. The results suggest that nominating high numbers of candidates exhibiting local PVEA does not significantly increase the electoral result of the ballot. We do find evidence, however, that the aggregate list result benefits from the presence of candidates holding local office in as many different municipalities as possible. This result provides an empirical underpinning for the electoral importance of balancing ballot lists geographically, and therefore has important implications for party scholars and practitioners alike.
Local PVEA as nomination criteria for parties
The concept of PVEA emerges from the vast literature on the personal vote, originally defined by Cain et al. (1987) as the share of votes received by legislators resulting from personal activities and characteristics helping them to develop an individual electoral support base and reduce electoral vulnerability. In their seminal article, Carey and Shugart (1995) develop a comprehensive theoretical framework as to how a candidate’s incentive to cultivate these personal votes is conditional on the electoral formula. A large number of scholars have subsequently analysed the personal vote-seeking behaviour of legislators (e.g. Bräuninger et al., 2012; Crisp et al., 2004; Shomer, 2009) and candidates (e.g. De Winter and Baudewyns, 2015; Zittel and Gschwend, 2008).
In addition to research on personal vote-seeking behaviour, Shugart et al. (2005) argue that politicians can also instrumentalize certain objective attributes to cultivate personal votes. In the extant literature, these are commonly referred to as PVEA: personal traits which signal a politician’s credibility and knowledge as agent of representation to particular subgroups of voters. In line with the earlier mentioned framework, the extent to which politicians exhibit these traits depends on variations in electoral rules (Shugart et al., 2005). In candidate-centred systems the probability that politicians will exhibit PVEA and voters will rely on PVEA in their vote decision is higher than in party-centred systems.
Under this definition, many different individual attributes can be labelled as PVEA and emphasized by candidates in their personalized campaign strategies to make themselves known among voters (Gschwend and Zittel, 2015). Related literature has, for instance, mentioned name recognition through parliamentary activity (Däubler et al., 2016), celebrity status as television personality (Arter, 2014) or sports career (Nemoto and Shugart, 2013), physical attractiveness (Rosar et al., 2008), or family ties to a previous politician (Horiuchi et al., 2018).
In addition, many studies have examined the effects of attributes which particularly demonstrate the localness of politicians. Being born locally, living in the electoral district, having previous local-level experience, or even speaking the local dialect have all been mentioned or investigated as examples of ‘local PVEA’ (e.g. Bräuninger et al., 2012; Jankowski, 2016; Put and Maddens, 2015; Tavits, 2010). As Shugart et al. (2005: 441) correctly argue, these different indicators disseminate mildly different messages to voters: whereas being a district native signals candidate knowledge of what are the primary local concerns, local political experience cues voters that candidates have proven track record of catering to local interests.
From the voter perspective, local roots in the district signal candidates’ knowledge of the local needs and therefore deliver an important argument to vote for them. Candidates themselves exhibit their local roots during the electoral campaign to claim credibility as a reliable representative for local interests.
The positive effect of localness on candidates’ electoral success has been extensively demonstrated in the literature. In a survey experiment, Campbell and Cowley (2014) show that the residency of candidates has the strongest impact on the way voters perceive candidates. Other studies demonstrate the substantial effect of the geographical distance between candidates and voters in different electoral contexts (Arzheimer and Evans, 2012; Górecki and Marsh, 2014). However, the existence of such a local candidate effect requires voters to be aware of candidates’ local ties. Recent empirical work has examined which voters use these local characteristics as cues in their voting decisions. Roy and Alcantara (2015) show that the magnitude of candidate effects varies depending on the levels of campaign attention and partisan attachment by voters. Collignon-Delmar and Sajuria (2018) build on this literature and find that voters who feel strongly about their regional identity attach greater importance to candidate localness.
From the party selectorate perspective, these local candidates are electorally valuable which makes it a profitable strategy to nominate them on the ballot list. A limited number of studies has discussed party nomination strategies regarding local candidates in more detail. Arter (2013) states that a ballot list full of ‘quintessentially local candidates’ can be perceived as a level-playing-field nomination strategy where vote shares are maximized by encouraging fierce intra-party competition (see also von Schoultz, 2018). In a study on European Parliament elections, Arter (2015) links these strategic nomination decisions to differences in ballot structures between list proportional representation (PR) systems: party selectorates in closed-list systems will be more inclined to follow the ‘lead candidate’ strategy where an experienced candidate with high vote-earning capacity (e.g. incumbent) is supplemented by candidates with far less vote-earning potential. Furthermore, Nemoto and Shugart (2013) investigate how party nomination strategies regarding localized candidates vary over different electoral formulas. They bring evidence that parties tend to have more localized candidates in systems with preference voting and vote pooling (i.e. votes cast for candidates contribute to the number of seats won by their party in the district). This is not fully in line with Michael Gallagher’s statement that local roots are invariably sought by party selectorates (Gallagher and Marsh, 1988: 251). Instead, the importance of local roots for voters, candidates and parties alike is conditional on mechanical aspects of the electoral system.
Specifically, in systems with preference voting and vote pooling, parties are incentivized to nominate as many localized candidates as possible and to let them engage in fierce intra-party competition with one another (Nemoto and Shugart, 2013: 5). Each of these candidates appeal to their own narrow local subgroup of voters from which the party benefits at the district level in a system with vote pooling. The example of such an electoral system in the Nemoto and Shugart contribution is the Japanese open-list PR system (2001-present), but these mechanical preconditions apply to a wider set of electoral systems. The subcategory of flexible-list systems also allows for both preference votes and vote pooling (Renwick and Pilet, 2016).
As concerns elections under closed-list proportional representation rules, common wisdom states that the localness of candidates – or other PVEA – is of minor importance since votes are cast for parties instead of candidates (Carey and Shugart, 1995; Shugart et al., 2005). Some empirical contributions partially contradict this claim. Riera (2011) shows that vote shares slightly increase in Spanish districts where candidates on the top of the list are high-quality politicians, that is, with some level of experience in parliamentary representation. More importantly, in light of this contribution, localness of the candidate on the top of the list does not determine the party result. Another recent study on the Portuguese case provides somewhat mixed conclusions for the different indicators of localness (Espìrito-Santo and Rodrigues Sanches, 2018): whereas resident candidates are fairly strongly represented on the ballot lists, being native and having experience in local politics do not seem to be valuable candidate attributes in the eyes of Portuguese party selectorates.
In this article, we empirically investigate the claim that nominating as many localized candidates as possible is an effective nomination strategy to increase aggregate vote shares at the district level. Several scholars have demonstrated that candidates with local PVEA are electorally more successful individually (Jankowski, 2016; Put and Maddens, 2015; Tavits, 2010). Yet no study has tested whether the electoral benefits of local PVEA hold at the aggregate level of the electoral district. Are ballot lists with a high concentration of localized candidates comparatively more successful? In the extant literature, it is generally assumed that parties benefit from nominating as many candidates with local PVEA as possible. Therefore, the first hypothesis states:
H1: Ballot lists with high numbers of candidates with local PVEA are electorally more successful.
We argue, however, that an important element is missing from the theoretical mechanism as to what extent parties can benefit from localized candidates at the district level. As mentioned, candidates with local PVEA appeal to a narrow local subgroup of voters, geographically concentrated within the candidate’s own municipality or neighbourhood within the city. Party selectorates can therefore optimize this strategy by nominating localized candidates from different municipalities or areas within the district (Thomsen and Suiter, 2015). This would ideally result in a geographically balanced ballot list with localized candidates each appealing to voters of their own separate municipality.
We argue that covering as many municipalities as possible with PVEA-exhibiting candidates is electorally more rewarding for parties, since a proportion of voters might be inclined to favour and to vote for a local candidate even if such a candidate is not running for their preferred party (Blais and Daoust, 2017; Blais et al., 2003). If parties nominate localized candidates from all over the electoral district on the ballot list, the likelihood of losing voters who are inclined to support someone from their own area will decrease. Therefore, we expect that parties will benefit electorally if they geographically distribute their localized candidates. This brings us to the second hypothesis:
H2: Ballot lists with strong geographical distribution of candidates with local PVEA are electorally more successful.
The case of Flanders
To test how candidates with local PVEA affect the aggregate ballot list result, we use data on election candidates of Flemish political parties for the Belgian federal Lower House elections and the Flemish regional elections between 2003 and 2014. Specifically, these data cover information on 7527 candidacies on 223 ballot lists from seven different parties. 3 For both elections, a flexible-list PR system is in place where voters are able to either cast one or multiple preference vote(s) for candidates on a list, or to support the pre-electorally determined candidate rank order by casting a list vote. If individual candidates are able to reach the eligibility threshold of preference votes on their own, they get elected automatically. In practice, only a few well-known politicians have the vote-earning capacity to do so (André et al., 2012). For the other seats, list votes are transferred to the remaining candidates according to their rank on the ballot list. Renwick and Pilet (2016) categorize this subtype of flexible-list system as a transfer type, referring to the mechanical aspect of list votes being transferred to the highest ranked candidates not reaching the eligibility threshold independently.
There are strong similarities between elections at the federal and regional level in Flanders. André et al. (2012) report comparable shares of voters casting preference votes in the two election types. Belgium also has the peculiar situation that there are no national parties: all parties are split across the language divide, which leads to a distinct Flemish regional party system with the same parties competing in elections at both levels. In addition, these elections are organized over identical provincial districts during the period under investigation (2003–2014). As there has been a considerable number of MPs and candidates jumping from the regional level to the federal and vice versa since 2003, it seems that Flemish politicians also consider the federal and regional levels to be relatively equal (Vanlangenakker et al., 2013). Therefore, it makes sense to perform a pooled analysis of federal and regional ballot lists by Flemish political parties.
To be sure, in this study, we examine the electoral success of ballot lists. Their vote shares are determined by the number of voters casting one or multiple preference votes plus the number of voters casting a list vote (i.e. vote pooling). As a result, the Flemish case corresponds to the electoral context described by Nemoto and Shugart (2013) where parties’ strategies are aimed at nominating as many localized candidates as possible, and it is in the parties’ interest to let them engage in ‘laissez faire’ intra-party competition.
Earlier research points out that local roots constitute an important element in selection processes of Flemish political parties. Put and Maddens (2013) show that local office-holders are more likely to be selected in realistic positions on the list, that is, positions in which candidates have a realistic chance to get elected. Other scholars which analysed Belgium in its entirety show that the importance of local roots also holds for other Belgian regions. Together with France, Belgium has the highest share of MPs holding local office simultaneously, which shows that local roots are considered a crucial asset for political careers in the entire country (Van de Voorde and de Vet, 2017). At the same time, this can hardly be perceived as Belgian exceptionalism, as multiple office-holders in the parliaments of Spain, Portugal, Switzerland, and Austria also range between 20% and 60% (Van de Voorde and de Vet, 2017). Moreover, practically all election candidates in Belgium run for parliamentary office in the district where they live (De Winter, 1997). Belgian politicians generally remain faithful to the electoral district where they started their careers. Finally, Dodeigne (2015) shows that Walloon regional MPs with municipal office focus on getting support for local projects to secure re-election.
Figure 1 shows the relative number of local office-holders on the ballot lists of the seven Flemish parties included in this study, averaged over all elections and split up according to the type of local office. In Belgium, there are three types of local office: mayor, alderman, and local councillor. We observe strong inter-party variation as to how many local office-holders are, on average, nominated on the ballot lists. Specifically the three oldest or ‘traditional’ parties in the Flemish party landscape (Christian-democrats, liberals, and social-democrats) exhibit the strongest concentration of localized candidates (i.e. 60%–75%), followed by the far right party (53.0%) and the Flemish-nationalist party (49.4%). These numbers split per party show that Flemish party selectorates generally follow the strategy of packing their ballot lists with local PVEA candidates. Only the green party (37.2%) and LDD (12.1%) follow at a distance. 4
Mean share of local office-holders on ballot lists over all elections (2003–2014).
Self-evidently, this variation is to a large extent related to party strength and party institutionalization. Political parties with high vote shares in parliamentary elections can capitalize on their electoral popularity during local elections, which leads them to pile up the number of local office-holders. In addition, parties with a more institutionalized position in the party landscape have been able to participate in local elections on many occasions, which allowed them to develop local party strongholds both in terms of party organization and local government participation. As a result, a strong correlation between relative vote share and number of local office-holders on the ballot list might reflect two possible underlying causal links.
Data, operationalization, and method
The electoral result of ballot lists primarily depends on the aggregate result of the party itself. Especially in highly nationalized party systems (Caramani, 2004; Deschouwer, 2009), list results are highly comparable across different electoral districts. Although it is theoretically possible that the result of a ballot list contradicts with the general party trend, in practice this is rather exceptional. As a result, the electoral result of a ballot list, expressed as a raw percentage in the electoral district, is not an accurate dependent variable for this study. It is highly unlikely that, for instance, one of the lists of a small party would completely overturn the inter-party balances of power in a district by nominating a higher number of local office-holders.
Taking this into account, we use the Relative Electoral Swing (RES) as the dependent variable:
How did we construct this variable? First, we expect that a ‘strong’ list in a given electoral district will score relatively better than the aggregate party result. For instance, the Flemish green party Agalev faced a major electoral defeat in the federal elections of 2003. But presumably, the party could have been able to limit the electoral damage in districts with a high number of local office-holders on the ballot list. Therefore, we should calculate the ratio between the vote share of the ballot list in the electoral district and the vote share of the party at the aggregate level (i.e. over all electoral districts). In other words, the overall electoral result of the party is the base level, which we then use to examine to what extent a ballot list in a given district scored better or worse. 5 Again using the example of Agalev in 2003, this party received a vote share of 3.9% at the aggregate level, that is, across all Flemish electoral districts. In the district of Leuven, however, the party was able to achieve better results, with the ballot list receiving 5.2% of the votes. The ratio between both results then amounts to 1.33. In the district of Limburg, however, the ballot list only received 2.4% which leads to a ratio of only 0.62.
If we use this ratio as dependent variable and find a significant association with the number of local office-holders, this would not yet imply the existence of a causal relationship. We know that, for all parties, cross-district differences in ballot list results are generally fixed, since parties have developed electoral strongholds in certain areas over time. The Flemish Christian-democratic party CD&V, for instance, traditionally wins the largest vote share in the district of West Flanders. It is plausible that, mainly as a result of these traditional strongholds, ballot lists in these areas have higher numbers of local office-holders than elsewhere. To take this into account, we use a relative measure comparing the result of election t with that of the previous election t – 1. We calculate the electoral swing between two elections by dividing the result of election t by the result of election t – 1. 6 As explained above, we also control for the aggregate result of the party. Therefore, we calculate the RES as the dependent variable: the electoral swing of a party list in an electoral district divided by the electoral swing of the party at the aggregate level.
To illustrate this approach, we refer to the case of the Flemish-nationalist party N-VA in the district of Antwerp. In the 2014 federal elections, the ballot list in this district won 39.4% of the votes, while this was only 30.7% in the previous elections. This major progress is reflected in a ratio of 1.28. The aggregate party result increased less strongly, from 27.8% to 32.4%, leading to a ratio of only 1.17. Put differently, while the party substantially increased its vote share at the aggregate level and won the election, the ballot list in the Antwerp district was able to do even better. The fact that the Antwerp ballot list scored substantially better than the general party trend is reflected in a RES-score of 1.09.
Calculating the RES-score, however, is not always straightforward. First, comparing the 2003 federal elections with those of 1999 causes some difficulties, as electoral districts were reshaped in 2002 by merging existing districts. Therefore, we have to compare the ballot list results of larger electoral districts in 2003 with an aggregate score of all the smaller districts of 1999 comprising the same area. Next, some parties formed electoral alliances between 2003 and 2007. As a result, we had to correct the results of the alliance in order to be able to calculate the electoral swing for these lists as well. 7 Third, parties that participate in elections for the first time are automatically dropped from the analysis, as there is no result for election t – 1. This brings our number of observations for the analysis to 223.
As concerns the explanatory variables, the presence of candidates with local PVEA on the ballot list is the main variable of interest. Empirical studies on local PVEA have used three measures: local birthplace in the district, local residence in the district, and local-level political experience through local office. It has been argued that being born in the district is the weakest indicator of local roots (Marsh, 1987; Tavits, 2010): being district native will not be rated important by voters if, for instance, politicians have been active in other areas during the largest part of their careers. As a result, this first measure represents the weakest argument politicians might use to convince voters of their local connection to the area. Regarding the second measure, the description of the Flemish case points out that there is hardly any variation between candidates in terms of residing in the district or not, rendering local residence unsuitable for empirical analysis in the Belgian context.
In this contribution, we therefore consider candidates which hold local office to have local PVEA. For all 7527 candidacies in the data set, we checked whether they were holding local office (i.e. mayor, alderman, local councillor) at the time of election. To investigate the effect of a higher percentage of local office-holders on the ballot list result (H1), we calculate a relative measure at the ballot list–level. Specifically, we consider the ratio between the percentage of local office-holders on the ballot list and the average percentage of local office-holders on ballot lists from the involved party in that specific election. We thus examine whether a ballot list obtains a higher vote share if it has a higher number of local office-holders compared with the party average.
We test the second hypothesis in two distinct ways. On one hand, we calculate the percentage of municipalities that is being represented by at least one local office-holder on the ballot list. As there are always more municipalities in a district than there are ballot list positions, the number of municipalities which can theoretically be represented on a party list equals the number of list positions. Subsequently, this figure is calculated as a percentage of the total number of list positions. In an extreme case, the percentage is 100, namely, when all list positions are held by local office-holders from all different municipalities. In practice, however, this percentage will be lower as not all list positions will be held by office-holders or not all office-holders reside in different municipalities. As argued earlier, it is appropriate for this variable as well to calculate the ratio of the ballot list percentage vis-à-vis the party average.
On the other hand, party selectorates might consider the geographical distribution of their candidates not in terms of municipalities but rather in terms of relative population. Covering the largest municipality in the district with localized candidates might be more valuable than to cover two or three smaller municipalities with limited population shares in the district. Therefore, we alternatively calculate the percentage of the district population represented by at least one local office-holder on the ballot list. These two calculations provide slightly differing interpretations of the nomination strategy under study, that is, the geographical distribution of localized candidates on the ballot list.
In addition to the variables of interest, the analyses include control variables to account for confounding effects. First, we registered whether the party’s campaign figurehead was running on the ballot list or not. 8 Ballot lists with the campaign figurehead generally accomplish a significantly higher vote share as these candidates receive strong media attention during election campaigns.
Second, we control for the effect of incumbent MPs and ministers as these offices usually generate high visibility and media attention as well. Moreover, incumbents can also rely on previous political experience which fosters a stronger perception of reliability among voters. Analogous to the proposed operationalization for local office-holders, we calculate a relative measure to check whether a higher percentage of incumbents leads to an increased vote share at the aggregate level. Again, for every ballot list the ratio is calculated between the percentage of incumbents on the ballot vis-à-vis the party average. Finally, district magnitude (DM) is added to the models, as it provides an additional check that the effectiveness of the described nomination strategies is not conditional on the number of seats in the district.
We ran OLS regression models to test the two hypotheses on the relationship between local PVEA and the ballot list result. In addition, we added election type, year, and party-fixed effects to capture unobserved changes across time and to control for unobserved differences between political parties and elections.
Results
We begin our analysis by inspecting some descriptive statistics on the dependent variable, RES, and its correlation with the independent variables of interest in this article. Figure 2 shows that RES has a slightly right-skewed distribution with a median of 1.0000, mean of 1.0073, and standard deviation of 0.1114. The distribution shows that there are a few relatively extreme values on the far right-hand side (i.e. parties enjoying a strong upwards swing in certain electoral districts).
Distribution of the Relative Electoral Swing (RES)-scores.
If we look more closely at these extreme observations, we find that the non-traditional parties such as the Flemish-nationalist N-VA, the far right Vlaams Blok/Belang, the green party Agalev/Groen, and the libertarian LDD are somewhat overrepresented. Table 1 presents the number of observations per party in the data set and the standard deviation for each party, which indeed points to a stronger variability in RES-scores among non-traditional versus traditional parties. In the OLS models, we account for this inter-party heterogeneity by adding party-fixed effects. In addition, we use election type and year fixed effects to control for variation between regional and federal elections on one hand, and electoral volatility on the other, as high volatility would inflate the RES-scores.
Number of observations and standard deviations for Relative Electoral Swing of ballot lists, per party.
Moving on to the OLS models with fixed effects, Table 2 shows the result for three separate models. Model 1 regresses the RES on the percentage of local office-holders, taking into account the control variables and mentioned fixed effects. Model 2 uses the percentage of municipalities represented by local office-holders as independent variable of interest. Finally, in Model 3, we test the effect of the percentage of the population represented by local office-holders. In all three models, the presence of party figureheads clearly has the strongest effect on the outcome variable with comparable coefficient effects. The amount of incumbents and district magnitude, on the contrary, do not significantly affect the RES of ballot lists in any of the models.
OLS regression models with party and election year fixed effects.
Standard errors between parentheses. Bolded coefficients are significant at the .05 level. OLS: ordinary least squares.
More importantly, the results of Model 1 do not show support for the hypothesis that ballot lists with high numbers of candidates with local PVEA (i.e. local office-holders) are electorally more successful (H1). There is no significant association between the percentage of local office-holders and the RES in the multivariate model. 9 Interestingly, Model 2 confirms the hypothesis that ballot lists which contain a high number of candidates with local PVEA from different municipalities score better: we find a positive effect of the independent variable taking into account geographical distribution on RES-scores. Ballot lists that cover a higher share of municipalities with local office-holders than their party average exhibit higher electoral swings than their counterparts in other districts. This does not work similarly if we measure geographical distribution by calculating the population share represented by local office-holders on the ballot list (Model 3).
Figure 3 visualizes the only significant effect to facilitate interpretation, based on the OLS regression model: we see that, if ballot lists contain twice the number of municipalities their party on average covers with localized candidates (i.e. value of 2 on the x-axis), the RES increases to around 1.07. If, on the contrary, the party list contains only half of the average number of municipalities covered with local PVEA candidates, the RES-score decreases below 1.
The simulated effect of percentage of municipalities with local PVEA on Relative Electoral Swing.
An alternative empirical approach would be to analyse the results of ballot lists at the municipal level. This would allow a more fine-grained analysis of the relationship between local PVEA and electoral success of ballot lists. Unfortunately, Belgian parliamentary election results are only available at this disaggregated level for the 2014 elections. As a result, we are not able to similarly calculate the RES for ballot lists in all Flemish municipalities. Nevertheless, we can examine the association between the number of local office-holders on the list from a given municipality and the result of the ballot list at the municipal level. This constitutes a municipal level analysis to present a supplementary test of the relationship between local PVEA and ballot list vote share. We stress, however, that this is not an equally strict test as we cannot compare the 2014 scores with vote shares in previous elections.
Table 3 presents the results of such a municipal level OLS regression analysis, solely based on the 2014 ballot lists and election results for the Lower House. The number of observations equals 1848, which comprises the ballot list results of six parties 10 in 308 Flemish municipalities. We examine the effect of the number of local office-holders from the municipality on the ballot list result at the municipal level, controlling for the presence of incumbent MPs or ministers and the party vote share at the district level. The model shows that the presence of one additional local office-holder increases the ballot list vote share with 1.54%. This result needs to be interpreted carefully, however, since more candidates on the list from a given municipality could simply reflect that this particular municipality is an electoral stronghold of the party.
OLS regression model of ballot list vote share at municipal level, 2014.
Standard errors between parentheses. Bolded coefficients are significant at the .05 level. OLS: ordinary least squares.
Conclusion
This article analysed the effects of nominating candidates with local PVEA on the aggregate ballot list vote share in the district. Whereas earlier studies had already pointed out that individual candidates benefit electorally from exhibiting these attributes, we investigated whether this electoral advantage can be extended to the party at district level. The analysis shows no support for the hypothesis that nominating high numbers of localized candidates increases the vote share of the ballot list. Theorizing on the importance of the geographical distribution of the selected candidates, we presented a second hypothesis which states that ballot lists covering as many municipalities with localized candidates as possible in the district would increase the ballot list vote share. The second OLS model confirms this hypothesis, and suggests that ballot lists with double the amount of local office-holders from different municipalities on the list increases their RES (i.e. shift in party vote share from previous election) score to 1.07.
While these findings seem somewhat surprising and partly at odds with scholarly wisdom, we need to make two important remarks in this conclusion. First, the empirical test of the electoral value of local PVEA at the ballot list–level is admittedly a strict one. The dependent variable is a ratio which compares vote shares in election t with election t – 1 and additionally controls for the general party trend over all districts. Moreover, the analysis is based on aggregate percentages of local office-holders for only 223 ballot lists. This already required collecting information on more than 7000 Flemish election candidates. Ideally, we would of course include a larger number of ballot lists from a broader set of elections and countries.
This brings us to our second remark: the external validity of this contribution is limited to preferential-list PR systems with medium to large electoral districts. On one hand, our rich data set on Flemish party candidates does not allow to test our hypotheses within the context of small electoral districts such as is the case in Wallonia. While we know that the nomination strategy under study here is also applied in the southern region of Belgium, where district magnitude can be as low as 2 or 3 and the number of list positions is relatively limited (Dodeigne, 2015), we cannot ascertain whether the reported effects on the ballot list vote share hold up in very small districts, as the smallest district in our data set on regional and federal elections in Flanders between 2003 and 2014 is Leuven (DM = 7). Elections in Wallonia are organized in considerably smaller electoral districts. Therefore, investigating small DM cases constitutes an important venue for further research on the effect of party nomination strategies in list PR systems.
On the other hand, we only claim external validity for preferential-list PR cases which comprises flexible-list PR and open-list PR electoral systems. On the behavioural aspects studied in this article, flexible-list systems incentivize candidates and party selectorates in a similar manner as open-list systems. In both categories, candidates electorally benefit from their local roots which is in turn highly valued by party selectorates in nomination processes. As the literature points out that local roots are, on the contrary, rather ineffective PVEA in closed-list PR cases, it is highly unlikely that the studied nomination strategy affects ballot results similarly in closed-list systems. Another venue for follow-up research would be to detect alternative nomination strategies that are successful in closed-list PR systems. Arter (2015) suggests that the alternative lead candidate strategy might prove effective, where one dominant candidate at the top of the list draws additional voters.
Nevertheless, the added value of this study is that it presents the first empirical test of the argument that local PVEA boosts higher vote shares at the district level, and we make use of an original data set on aggregate ballot list characteristics which presents a relatively novel approach to our knowledge. Future work should not only expand the number of ballot lists and country cases, but more broadly investigate the interaction between individual candidate and aggregate list characteristics. In line with theories of descriptive representation, parties in list PR systems are inclined to balance their ballot lists for a number of candidate characteristics such as sex, age, profession, or location. Empirical research on how exactly these individual characteristics affect the parties’ electoral fortunes remain, however, remarkably scarce.
Footnotes
Acknowledgements
The authors thank the editorial team and anonymous reviewers of Politics for their valuable suggestions and feedback.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research Foundation Flanders (Grant Number: 11F6813N).
Notes
Author biographies
).
).