Blame Attribution in State Duma: Natural Language Inference in the Analysis of Parliamentary Speeches

• In Event: Text-as-Data Approaches to Studying Exclusion & Inclusion Strategies in Authoritarian Regimes

Abstract

Are controlled opposition parties in authoritarian parliaments allowed to criticize the government? We argue that they frequently do. Building on recent studies of non-democratic legislatures, we challenge the view that these parties merely serve as "window dressing" and propose that they strategically engage in government criticism. Using a dataset of parliamentary speeches delivered in the Russian parliament, the State Duma, between 1999 and 2024, we classify these speeches as either neutral, critical, or supportive of the government using the Natural Language Inference (NLI) approach. The results support the study's central hypothesis: controlled opposition parties are more likely to deliver speeches critical of the government than the regime-sanctioned, executive-aligned party, United Russia. However, we find that controlled opposition parties in the later convocations of the parliament are less likely to deliver critical speeches compared to earlier convocations, a trend that corresponds with the increasing consolidation of the authoritarian regime.

Authors

• David Avagian, U of Mannheim (Germany)

• Artur Baranov, Northwestern U