UC Irvine

Previous work with survey data casts doubt on the Rational Expectations Hypothesis. In this paper, we develop a model of expectation formation where agents form their forecasts of inflation by selecting a predictor function from a set of costly alternatives whereby they may rationally choose a method other than the most accurate.  We use this model to test whether survey data exhibit rationally heterogeneous expectations. Maximum likelihood is applied to a new discrete choice setting where the observed variable is continuous and the latent variable is discrete. The results show there is dynamic switching that depends on relative mean squared errors of the predictor.