Bringing the voter back in: Candidates and voters in the intra-party electoral competition

According to many political commentators and scholars (Mcallister, 2009), we live in an age of personalization. This manifests itself in different ways, such as the growing importance of leaders to explain the electoral success of political parties. It is also argued that next to the important role of political leaders, the importance of political candidates in general increased. In a number of countries (Belgium, Bulgaria, Denmark), electoral reforms were introduced, giving more weight to so-called ‘personalized’ or preferential votes. Also at the voter level we see increased use of preferential votes, when comparing with the 1970s and 1980s. All these developments foster the importance of intra-party competition; the competition between candidates for preferential votes within a political party (Katz, 2003). Consequently, academic interest in this topic has increased(for a selection see Andre et al, 2012; Carey and Shugart, 1995; Marsh 1985). An important question in this field is which factors are important for attracting preferential votes. The literature has identified a number of key explanations, such as ballot list position (Marcinkiewicz, 2013) socio-demographic factor (McElroy and Marsh, 2010), and media attention (Van Aelst et al., 2008). Yet, all these studies suffer from the problem that they only study preferential voting from the perspective of candidates, using the percentage of preferential votes as dependent variable. The problem with this design, however, is that it omits voters from the model and does not allow to focus on the underlying mechanisms. In order to solve this problem, we propose a new method to study preferential voting, bringing the voter back in. We do this by investigating voters and candidates in dyadic relationships in order to more validly examine which factors matter for citizens when casting a preference vote for a certain candidate. This approach is often used in marketing literature to better understand the decisions of consumers, and was recently introduced to study voting behaviour with regard to political parties (van der Eijk et al., 2006), but has not been applied to political candidates yet. Using a mock ballot immediately after the Belgian 2014 elections, Belgian respondents had to indicate how they voted in the ballot box. This data is then linked to data about all Belgian candidates which were gathered in the same period. Once linked, we use conditional logit models in order to study when the dyadic pair between the voter and the candidate is a match (i.e., a vote was cast) and when it is not. The benefit of this model is that we can include characteristics of candidates, as well as congruence measures (e.g., gender or age similarity, same ethnic background, etc.) between voters and candidates at the same time in the model. With regard to the results we expect that, ceteris paribus, citizens are more likely to give their preferential vote to candidates that are more congruent. However, we expect that when interacting these congruence measures with characteristics of voters they have a stronger effect for less represented groups; women, candidates from ethnic minority background, and the younger.