If voters use information about the economy to assess the competence of incumbents, a connection between economic conditions and incumbent success should only be discernible in settings where public policy might plausibly affect the economy, and where the assignment of government responsibility is relatively straightforward. Applying this logic to gubernatorial elections in the United States, we test the following hypothesis: the connection between economic conditions and incumbents' vote shares is mediated by the structure of the state economy. This hypothesis is premised on the idea that voters understand that raw macroeconomic aggregates – when driven by factors like weather, commodity prices and federal policy – are poor signals of incumbent performance. Using data from gubernatorial elections held between 1950 and 1998, we show that the connection between macroeconomic indicators and incumbent success is weak in states dominated by natural resources and farming but quite strong elsewhere. This finding helps explain why earlier studies found no connection between state-level economic conditions and gubernatorial elections.