ESTIMATION AND PREDICTION OF TIME-VARYING PARAMETER ECONOMIC MODELS

Abstract

Methods commonly employed to estimate the structural coefficients in an economic model assume that those coeffients are fixed constants. However, econometricians have argued that at least some of the coefficients in an econometric model are not constants, but instead, vary stochastically over time. Of the numereous time-varying parameter models that have been discussed in the literature on model building, one that is of particular interest in the context of econometric models is the Kalman filter model. The Kalman filter model was first used in stochastic control theory, but with some suitable interpretations this model can be written in the form of a simultaneous-equation model in which the coefficient vector follows a first-order Markoff process. This paper suggests procedures for estimation of a regression model with non-stationary output where the coefficient vector of the regressors is a weekly stationary stochastic process. Once the regression model is estimated by this method, the predictive ability of this model will be compared with other standard econometric estimation methods, such as Ordinary Least Squares (OLS) and Three Stage Least Squares (3SLS) methods.