[Replication] Exact Numbers, Fragile Inference: A Replication of When Do Männerparteien Elect Women?

Abstract. Weeks, Meguid, Kittilson, and Coffé (2023) report that radical-right populist (RRP) parties elect more women when electorally threatened, identified by a negative male/female-voter-ratio × vote-change interaction in a multilevel model of 58 RRP party-elections. Every cell of the headline table reproduces from the deposited data to four decimals; the result is real and contains no coding error. It is also fragile. The interaction collapses when a single observation is removed (France, National Front, 2012, a roughly two-seat delegation: p rises from 0.008 to 0.71), under country fixed effects (p=0.21), under few-cluster-robust CR2 standard errors (p=0.15), and under a stricter outlier trim (sign flip). It is absent in the 48 non-quota party-elections (p=0.70) and before 2008 (p=0.86). A blind analyst given only the abstract reached the same estimand, predicted exactly these failures, and defaulted to the fixed-effects estimator that nullifies the result.