[Replication] A Local Shock, Locally Identified: Reproducing and Stress-Testing the Xylella–Far-Right Effect

Abstract. Cremaschi, Bariletto, and De Vries (2025) report that the Xylella fastidiosa olive epidemic raised far-right vote share by 2.2 percentage points in Puglia, via a two-way fixed-effects difference-in-differences over six national elections. The headline reproduces exactly (ATT 2.197, SE 0.666, p≈0.001), all five Table-1 outcomes reproduce, and the design is internally clean: flat pre-trends (p=0.81), a null in-time placebo, single-onset structure, and agreeing cohort-aware estimators. Two features scope it. The effect is confined to the Salento epicenter: Lecce holds 96 of 139 treated municipalities. And the cross-province binary contrast is marginal at the level treatment is assigned—Conley spatial-HAC returns p≈0.07–0.08 at sensible bandwidths, few-cluster bootstraps p≈0.27–0.32. The within-epicenter intensity gradient (+0.087pp per municipality-month) is the robust part: it survives assignment-level clustering (p=0.007), a Salento-specific time control, and a declaration-date placebo. The effect is genuine and point-identified; its generalizable form is the local dose-response, not the cross-province average.