EUParlspeech: A Dataset of Over 1 Million References to European Integration in Parliamentary Speeches

As part of my PhD dissertation I developed EUParlspeech, a dataset of over 1 million references to European integration made in the plenary debates of ten national parliaments between 1989 and 2019. It is built from existing datasets of parliamentary speeches, most notably Parlspeech (Rauh and Schwalbach 2020). The dataset has applications for scholars of EU integration, party competition, political communication, and international relations. This chapter in my dissertation explains the construction of the dataset, describes its features, and demonstrates its face, convergent, and predictive validity. Automated analysis of parties' EU statements in parliament yield meaningful and well-known cross party differences, with challenger parties more likely to send clearer, more sceptical cues on integration than mainstream parties. Moreover, these automated measures correlate highly with expert assessments (CHES) and - in the case of the UK's Conservative Party - individual MPs' ideal point estimates based on EU statements in plenary debates can predict their subsequent vote and position at the 2016 referendum. I conclude that EUParlspeech data provide a promising new approach to studying party contestation over European integration.