Is Twitter representative for German parliamentary networks? An inter-parliamentary, multi-level analysis

Twitter is a prominent and well frequented social network for societal actors across different dimensions and origins. In addition, it is often used as data source for different analyses in social sciences, with parliamentary and party affairs being no exception. The open exchange of different actors, the supposedly semantic structure of the network data by retweets and mentions and the promise of being a suitable and accessible source of data for advanced methods such as natural language processing (NLP) suggests, that social networks data will only grow as data bases for future analyses. Yet, there are some fundamental questions unanswered when it comes to the use of Twitter for such enterprises, especially in the realm of parties und parliaments: Is Twitter a representative sample of a given parliament? Do we find variation across macro and micro level effects, and if so, what are the driving factors? German parliaments often lack representativeness in regard to their constituency and the question whether the representation on Twitter moderates or amplifies this circumstance is important not only normatively, but also to the scholar of parliamentary representation. The aforementioned growing body of work depending on Twitter as data source would also benefit from additional quantitative evidence. The presented work aims to supply future analyses with an answer to this question for the case of German national and state parliaments and their respective MPs. It presents a new data set of 1632 Twitter accounts by 3102 members of all German state and nationals parliaments. The analysis is presented as mixed effects model, incurring different micro predictors as control variables and displaying multi level predictors such as party membership and the state. It can be shown that several configurations of party and state effects, as well as several micro level variables are significant predictors of MP activity on Twitter. The presented final model is a suitable base model for further research endeavors. The advances made in this analysis are twofold. First, the analysis informs the scholar on representation in social media using a yet to formulate inter-parliamentary and multi level perspective. Second, the research informs subsequent work on fundamental effects any Twitter analysis on parliamentarism and parties should consider. It presents a base model for the German case and discusses a possible application of the analytical framework for a cross national and European perspective.