One can easily identify four general models of political business cycles: office‐motivated models (both forward and backward looking) and partisan models (again, both forward and backward looking). Each model makes different assumptions about the direction and timing of causal links between the economy and polity. This paper uses Granger causality tests to investigate the causal links between presidential popularity and different measures of aggregate economic performance and aggregate economic policy. The paper's aim is to investigate whether any existing theories receive substantive support and, if not, to suggest the properties that any new theories should display. The results indicate no overwhelming support for any existing theories, though partisan models receive more support than do office‐motivated models. The data appear to be most consistent with Ellis and Thoma's reputational partisan mo