What characteristics affect political speech in parliament and on social media? Analyzing German MPs language with regards to ‘migration’ in parliament and on Facebook from 2013-2017

Legislative speech has been at the center of the attention of political text analysis for decades. Prior work underlined the role of individual characteristics, constituency characteristics and institutional foundations of elite political behavior in parliament. However, only recently, scholars turned to investigate the intersection of social media use and legislative speech in parliament. While recent comparative analyses focused on Twitter, Facebook maybe even more crucial for political communication purposes for most parliamentarians (especially backbenchers). Therefore, this study focuses on the language used by German parliamentarians concerning migration policy and investigates factors for differences between individuals and differences between language on Facebook and in parliament for individuals. This study links parliamentary speech data from the ParlSpeechV2 dataset from 2013-2017, with data on candidates from the German federal election officer (Bundeswahlleiter) and the GLES Candidate Study and social media data from Facebook and Twitter. The multivariate regression analysis assesses what individual, constituency, or party-level factors are associated with higher (Euclidean) distances to party leadership, party average, and parliament average. Moreover, the author suspects a semantic shift during the migration crisis in 2015 which is tested with an embedding regression including said individual-level characteristics as regressors.