The Treatment Of Expectations Effects In Large-Scale Models

This paper builds on and summarises earlier work by Hall (1987), Currie and Hall (1989), Hall (1993), and Garratt and Hall (1992) which introduced data coherent models of the exchange rate which focuses on the important role of expectations in this market and which introduces these models into full-scale econometric models. Recent developments have abandoned the simple assumption of rational expectations for a more subtle treatment of expectations based on learning mechanisms. This papers summarizes the experience of introducing this assumption into a range of models including the LBS domestic model of the UK economy, the Global Econometric Model (GEM) and the International Monetary Fund model (MULTIMOD). The introduction of explicit learning allows us to assess the advantages and disadvantages of the strong rational expectations assumption more effectively and to analyze questions of credibility more effectively.We show that the introduction of model consistent learning to a large empirical model is an important innovation which can model properties that are quite different from either ignoring expectations effects or the dramatically oversimplified assumption of rational expectations.