Since the foundational work of Keynes (1936), macroeconomists have emphasized the importance of agents' expectations in determining macroeconomic outcomes. Yet in recent decades macroeconomists have devoted almost no effort to modeling actual empirical expectations data, instead assuming all agents' expectations are 'rational.' This paper takes up the challenge of modeling empirical household expectations data, and shows that a simple, standard model from epidemiology does a remarkably good job of explaining the deviations of household inflation and unemployment expectations from the `rational expectations' benchmark. Furthermore, a microfoundations or 'agent-based' version of the model may be able to explain, in a way that still permits aggregation, stark rejections of the pure rational expectations framework like Souleles's (2002) finding that members of different demographic groups have sharply different predictions for macroeconomic aggregates like the inflation rate.