Strategic Social Media Use in Political Campaigning: An Individual Level Analysis Linking a Candidate Survey with Facebook and Twitter Communication

Communication on social media platforms has become a pivotal part of political campaigning, offering candidates additional and direct channels to communicate with citizen, supporters or journalists. While a number of studies has investigated party and candidate differences in the adoption of social media platforms, most studies are based exclusively on data generated from Facebook and Twitter, such as the number of retweets, followers or likes. We argue that integrating social media data and additional sources, in particular candidate surveys, allows a more in-depth investigation of the effects of candidate characteristics, attitudes and strategic intentions on their social media use. Consequently, this study combines the most recent candidate survey (N = 803) of the 2017 German Longitudinal Election Study (GLES) and Twitter and Facebook data for a period of three months before the 2017 German federal election (N = 66,879 Facebook posts and N = 176,968 tweets). The candidate survey contains an extensive set of survey items which we group into three theoretical explanations of social media use. (1) Resources and campaign capacities; (2) political strategy (e.g., attitudes towards the own party, individualized campaign goals, list vs. direct candidate); (3) media strategy (e.g., attitudes towards the media and goals with regard to media coverage, strategic functions attributed to Facebook and Twitter). These are all novel theoretical explanations that could not be tested before in the political campaigning literature due to a lack of individual-level data. We test the explanatory power of these theories with regard to three dependent variables. (1) Candidates' social media adoption; (2) candidates' social media usage intensity (number of posts); (3) the topics discussed by candidates (e.g., are social media used for (local) mobilization or issue campaigning?). For the analysis the dependent variables are aggregated to the level of the candidate and the theoretical expectations tested in a binary regression and a negative binomial regression model. Considering earlier research, we also expect to find individual differences between social media platforms due to mediation effects. Our findings demonstrate that candidate characteristics and strategic considerations at the individual-level, which are often neglected in the literature, are important determinants of the (digital) campaigning strategy.