The Thermostatic Voter: Why Local Policy Success Fails to Build National Support

Abstract

Does experience with local policy implementation affect citizens' preferences for national policy? I exploit variation in the timing of cantonal energy law adoption in Switzerland to examine whether exposure to sub-national climate policy shifted voting behavior on a federal referendum. Using municipality-level (Gemeinde) data from the May 2017 Energy Strategy 2050 referendum, I find no robust evidence that prior cantonal energy law exposure increased federal policy support—inconsistent with the policy feedback hypothesis. A spatial regression discontinuity design restricted to same-language (German–German) canton borders—the cleanest specification, free of R\"{o}stigraben confounding—yields an estimate of $-1.6$ pp (SE = 1.18, $p = 0.17$). The pooled border estimate is $-1.2$ pp (SE = 1.10, $p = 0.29$); OLS with language controls gives $-1.8$ pp (SE = 1.93, $p = 0.35$; wild cluster bootstrap $p = 0.42$). Stratified randomization inference, permuting treatment within German-speaking cantons only, yields $p = 0.53$. Panel difference-in-differences with time-varying treatment coding across four energy referendums (2000–2017) yields a larger estimate of $-5.2$ pp (SE = 1.55, $p = 0.002$), with parallel pre-trends. Placebo referendums on non-energy issues show positive discontinuities at the same borders, suggesting that the null energy result is not an artifact of design. These findings provide no support for the policy feedback hypothesis and are consistent with policy satiation ("thermostatic" preferences), cost salience from implementation experience, or resistance to federal overreach.