Summary / Abstract

Empirical work in macroeconomics almost universally relies on the hypothesis of rational expectations. This paper departs from the literature by considering a variety of alternative expectations formation models. We study the econometric properties of a popular New Keynesian monetary DSGE model under different expectational assumptions: the benchmark case of rational expectations, rational expectations extended to allow for `news' about future shocks, nearrational expectations and learning, and observed subjective expectations from surveys. The results show that the econometric evaluation of the model is extremely sensitive to how expectations are modeled. The posterior distributions for the structural parameters significantly shift when the assumption of rational expectations is modified. Estimates of the structural disturbances under different expectation processes are often dissimilar. The modeling of expectations has important effects on the ability of the model to fit macroeconomic time series. The model achieves its worse fit under rational expectations. The introduction of news improves fit. The bestfitting specifications, however, are those that assume learning. Expectations also have large effects on forecasting. Survey expectations, news, and learning all work to improve the model's onestepahead forecasting accuracy. Rational expectations, however, dominate over longer horizons, such as oneyear ahead or beyond. 