Parliamentary Speech and Polarization: Evidence from a New Finnish Dataset

• In Event: Institutions and European Politics

Abstract

Recent work has explored the utility of parliamentary speech in evaluating various topics of interest, such as political polarization. I build on this literature by constructing and analyzing a new dataset consisting of all parliamentary speeches given in the Finnish parliament since 1917, linking speeches to data on members’ biographical characteristics, party membership, cabinet positions, constituency representation, and electoral performance.

Using these data, I provide a descriptive overview of the changes in salient topics over time and within and across political parties. Treating parties as the main group within a supervised learning model framework, I assess changes in parliamentary polarization between electoral terms, measuring polarization as the likelihood of accurately classifying members’ political parties based on words and phrases used in parliamentary speech. I also observe discontinuities correlated with changes in government within a single legislature, as well as the longer-term effect of the introduction of question period on the patterns of rhetoric used in parliament.

Methodologically, I introduce new methods of pre-processing text data for lemmatization and morpheme separation adaptable to languages other than Finnish. I also explore the effectiveness of multiple mechanisms for bridging the calculation of polarization from multilingual text: through direct translation, by bridging separate analyses in Finnish and Swedish, and through the use of sentiment analysis as a parallel method to capture aspects of polarization. I conclude by discussing potential extensions of this work both cross-nationally and to additional text data, such as newspapers and electronic media transcripts.

Author

• Alexander Kerchner, Princeton University