We estimate habit formation in voting--the effect of past on current turnout--by exploiting transitory voting cost shocks. Using county-level data on U.S. presidential elections from 1952-2012, we find that precipitation on current and past election days reduces voter turnout. Our estimates...
