There is considerable interest among students of political science, political communication, and media studies regarding the effect of media coverage on electoral politics. Despite this interest, the comparative study of political communication is often hampered by insufficient data across countries and elections. The paucity in relevant data, moreover, is unsurprising given the substantial amount of work required to content analyze media coverage across multiple countries and different election years. In an effort to overcome several key challenges, this study explores the utility of recent advances in automated text analysis for content coding media coverage surrounding the 2009 European elections. Specifically, we apply a set of supervised and semi-supervised text classification routines to examine roughly 50,000 human coded media stories—newspaper and television transcripts—produced as part of the 2009 European Election Media Study. As we will demonstrate, this dataset provides a unique opportunity to explore the “promises and pitfalls” of using text-mining methods to study comparative political communication.
