We confront a central assertion of the Kramer (1983) critique of using survey data to study economic voting: namely, that individuals' reported economic perceptions contain information that is largely unrelated to the economy. We show theoretically that individuals who are trying to understand their own economic situation will often be better off relying on local-rather than personal or national-economic information, and that this produces behavior similar to sociotropic voting. Examining data from a novel survey instrument that asks respondents their numerical assessment of the unemployment rate confirms this hypothesis. Specifically, individuals' economic perceptions respond to local conditions. Moreover, these perceptions associate with individuals' vote choices. Extending our analysis to a time series of polls from 1980-2008, we verify that patterns in individual data also exist in aggregate economic perceptions and political support. * We thank