Replication of Frey (2024) "Larger Legislatures and the Cost of Political Brokerage: Evidence from Brazil"

Frey (2024) studies the effect of council size on the electoral performance of local incumbent parties in mayoral, gubernatorial, and presidential races in Brazil. Using a fuzzy regression discontinuity (FRD) design, the author exploits exogenous variation in the number of council seats induced by population thresholds. In this replication report, we assess the computational reproducibility and robustness of the article's main finding that an increase in council size reduces electoral support for incumbent parties in the executive. We successfully reproduce the paper's results using the provided replication materials. We then check for robustness in three ways: first, we test for manipulation of the running variable. Results from a local polynomial density test indicate that municipalities did not systematically sort around the population thresholds, thereby supporting the validity of the identification strategy. Second, bandwidth sensitivity analyses show that statistical significance weakens as wider bandwidths are considered, particularly when estimates are based on individual election outcomes. Third, when using bias-corrected point estimates and robust confidence intervals, FDR estimates from most model specifications no longer reach conventional levels of statistical significance, whereas estimates based on aggregated information across elections remain largely robust. Overall, while the substantive conclusions of Frey (2024) are partly supported, the precision of the estimated effects is sensitive to modelling and inference choices.