Determinants of Twitter interactions between candidates in multilevel electoral campaigns

Parties and their candidates increasingly evolve in multi-layered contexts where supranationaland sub-national levels play a big role in the political competition. A growingliterature focuses on the behaviour of parties and candidates in this new institutional context(Deschouwer 2003; Fabre 2008, 2011). Yet, these studies mainly consider how it impactsintra-party processes during elections (e.g. candidate selection), affects individual careerpatterns within parties (Stolz 2003; Dodeigne 2014), or influences voting behaviour (Hough& Jeffery 2006). We take the first step to exploring how parties and candidates interact duringelectoral campaigns in multi-layered contexts. We investigate whether these interactions arestructured horizontally, reflecting a certain degree of autonomy between layers (stratarchy –see Carty 2004; Bolleyer 2012), or if they follow vertical structures across layers, based on ahierarchical partisan logic or other cleavage dimensions (left-right, centre-periphery, etc.).The case of the May 2014 Belgian elections - where parties and candidates competed at theregional, federal, and European levels simultaneously - offers a unique opportunity toinvestigate campaign behaviours in a multi-layered context. We explore interactions betweencandidates on Twitter in the run up to the vote. We define interactions based on the flow ofretweets and conversations (@) initiated by candidates. In so doing we capture dynamicinteractions or networks, as opposed to a more static definition based on followers. Moreoverthe Belgian case enables us to explore whether Twitter based networks go beyond thelanguage divide (French/Flemish). Interactions are analysed using Social Network Analysisand tools like Ucinet (Borgatti, Everett and Freeman, 2002). Ultimately, we seek to unveil anddisentangle whether patterns of vertical or horizontal structure prevail.