This study develops a method to measure congressional partisanship using social media data. Analyzing over 2.1 million tweets sent during two terms (2015-2018), we classify partisan language and quantify legislators' commitment to their party.
Methodology & Data: We employ computational linguistics on Twitter data, analyzing rhetoric toward Democrats and Republicans across two congressional cycles.
Our findings demonstrate a strong connection between online partisanship indicators and offline legislative behaviors. This measure—based solely on explicitly partisan language—predicts party unity voting even after controlling for ideology. It also explains bipartisanship expressions in Congress more clearly than traditional metrics.
Presidential Politics: We find presidential support ratings are primarily driven by a member's partisanship, not their ideological leanings.
This research provides two key contributions: (1) it operationalizes the difficult-to-measure concept of legislative partisanship using digital footprints; and (2) it shows this novel measure offers clearer insights into how partisan identity influences representation than established approaches.
