Michael Imre, Alejandro Ecker
Measuring policy positions using 280 characters (or less) – Introducing a new data set on intra-party preference heterogeneity

10th Annual Conference of the European Political Science Association (virtual conference), Prague, June 18th to June 19th, 2020

The assumption of parties as unitary actors is generally considered an oversimplification, appeals to abolish it have been repeated in the literature for decades. While research opening up the ‘black box’ party has started to pick up pace in recent years, the vast majority of empirical studies in research areas like party competition, coalition governments and voting behaviour still treats parties as unitary actors. Data on preferences of party factions or even single politicians are still scarce and usually only include single countries or even individual parties, making comparative work challenging. This paper presents the first version of a novel data set on intra-party preference heterogeneity. Our data provide valuable insights into the heterogeneity of preferences within parties, future research will be able to use them in a broad range of contexts concerning the determinants and effects of intra-party preference heterogeneity. We use textual data from Twitter, a social media platform popular with politicians where they can publish direct and unsolicited political statements with little party control, to estimate the ideological positions of members of parliament using quantitative text analysis. As data from social networks are very noisy, this step is preceded by exhaustive data cleaning and sorting out of non-political content using a machine learning algorithm. Covering a total of more than 20 European countries, we estimate the positions of all members of parliament with an active Twitter account on a latent policy scale.