Content area
Abstract
Taking the Lucas critique as a point of departure, these essays propose a variety of estimation strategies for time series data that cannot be assumed a priori to be drawn from a single regime. We consider first hypothesis testing for structural change, trying to formalize statistically the concept of change in regime. In an application to U.S. monetary data, a case is made to classify the Volcker monetary aggregate targeting and subsequent relaxation as changes is regime. Interest rates are shown to respond differently to money supply announcements in three sub-periods.
The first estimation strategy is a Bayesian one. We argue that in the presence of heteroscedasticity that deviations from normality cannot be classified between changes in the conditional mean and variance. A Bayesian learning model is proposed where posterior densities are a weighted average of past forecasts and current forecast errors. The model is shown to contribute to excess volatility in securities' prices.
An alternative approach to the Bayesian procedure is offered in the fourth chapter. Classical statistical techniques suggest using time series information to produce combined forecasts when there are insufficient degrees of freedom available in a new regime. The money demand equation is first shown to be unstable using the procedures of Chapter 2. Combined forecasts from three previous regimes are shown to produce superior forecasts for money demand in the 1980s.
The last approach tries to capture the invariant parameters of a decision problem. A model of a representative firm is used to generate a reduced form where the parameters of the cost function can be recovered from the point estimates. As applied to the inventory problem, I show that previous researchers have made errors in both estimation and specification of the intertemporal dynamics. In particular, I show that the "forward" solution to the rational expectations difference equation can be consistently estimated and that serial correlation must be forward filtered as well.
A closing chapter summarizes the main results and assesses the impact of the Lucas critique on econometrics.
