The purpose of this research is to explore a new approach to the economic analysis of nonstationary time series data. Most currently available model specifications assume that optimal forecasts of economic events are linear functions of past observations. By contrast, the model developed in this project is nonlinear in the data. The nonlinearity arises from the hypothesis that the process governing important economic variables are occasionally subject to discrete, dramatic changes in regime. Despite its nonlinear nature, the model is shown to be completely tractable for many theoretical and empirical rational-expectations applications. This research is important because the new model holds promise for a number of interesting applications, among them; (a) an alternative approach to characterizing the business cycle, which may enable us to better forecast business fluctuations; (b) a more flexible tool for analyzing financial data, for example in trying to identify and understand the dramatic fluctuations in the stock market; (c) a better approach for analyzing the effects of dramatic shifts in public policy within a rigorous econometric framework.
