Rational Expectations and Distributed Lag Expectations Proxies

Abstract

exogenous variables which actually generate the variable to be predicted. Empirical researchers, however, have employed a modified (and weaker) form of Muth's hypothesis. Expectations are defined as rational if they fully incorporate all of the information contained in a specified information set that is available at the time that the forecast is made.3 When using distributed lag proxies, researchers implicitly assume that the information set used to generate forecasts consists solely of the information contained in the history of the variable which is being forecast. By invoking the hypothesis that expectations are rationally formed, one might hope to provide a more rigorous framework in which to guide the construction and evaluation of these distributed lag expectations proxies. The immediate purpose of this article is twofold: first, to review the properties of rational expectations in the context of autoregressive forecasting, the expectationsgenerating mechanism implicit in the use of distributed lag proxies; and second, to test the rational expectations hypothesis on an as yet unexplored set of directly observed expectations data. The broader objective, as just noted, is to provide insights into the usefulness of rationality as a criterion with which to analyze the performance of distributed lag expectations proxies. In so