[Replication] Robust in Sign, Fragile in Magnitude: A Replication of "Sources of Partisan Change: Evidence from the Shale Gas Shock in American Coal Country"

Abstract. Gazmararian (2024) reports that the shale-gas shock raised the two-party Republican presidential vote share in U.S. coal counties by 4.9 percentage points, identified by a matched adjacent-neighbor difference-in-differences with a common 2008 onset and the Liu-Wang-Xu counterfactual estimator. Every headline and supplementary cell reproduces from the deposited data to within 0.05pp (`fect` ATT 4.89, matrix-completion 4.76). The sign survives an aggressive battery: leave-one-state-out (0/23), influence trimming, three small-cluster inference methods, and removal of the Trump elections (3.90pp). The magnitude is fragile. The full-window parallel-trends assumption is decisively rejected (joint Wald χ²=47, p=5×10⁻⁸); coal counties were already drifting Republican through the 1980s-1990s. The pooled estimate falls monotonically as the pre-window shrinks toward treatment, reaching 3.04pp at the cleanest 2004-only counterfactual, and the average post effect is HonestDiD-fragile (M̄\*≈0.25). The 4.9pp figure is the upper end of a baseline-dependent range; the qualitative realignment finding holds throughout.