Revisiting Faction Detection: Operationalizing Factionalism Using Cosponsorship Data

Parliamentary Party Groups are often factionalized. Different groups inside the PPG are striving for influence and control, at the same time, party unity and stability seem to be closely related critical factors for legislative success and re-election. Yet, factionalism seems to be a highly heterogenous phenomenon: For example we observe a huge variance in the explicitness of factions and their degree of institutionalisation across different parliaments and likewise in the explicitness of membership of MPs in such groups. Consequently, research of factionalism could benefit from methodological advances that are more or less agnostic to this variance. This paper proposes a method for automated faction detection, using established algorithms of community finding, based on relational data available in nearly all modern parliaments (Briatte, 2016). Traditionally, the quantitative detection of factionalism has largely relied on voting data. Only recently, scholars turn to other relational parliamentary data sources. Using a relational data set of cosponsorship in parties of the 17th German Bundestag, the sponsorhsip of motions is used to deploy a modularity based cluster detection algorithm (Pons and Latapy, 2005). The time series cluster information is then used to visualize the membership of MPs in the respective factions. As second confirmatory step, a dataset of informal membership to intraparty groups is tested for further validation. First results indeed show both stable factions over time and some movement of MPs between factions depending on time and party. This is in line with other findings from the very same data set, where cosponsorship in the Bundestag has been shown to be quite stable over time, yet also to be predicted by other variables. The paper also discusses the use of the accompanying software, as a R library is provided. As relational instruments are present in most modern assemblies, a rigid and automated method of faction detection seems benefical for further comparative analyses. This paper hopes to contribute not only methodological guidance and a first empirical application faction detection based on cosponsorship data in the German case, but also tries to be a stepping stone to establish comparative research designs.