The Effects of Autocorrelated Errors on the Statistical Estimation of Distributed Lag Models

Abstract

RECENT economic literature has been invaded by the rapidly increasing use of distributed lag models as a tool for economic analysis.l These models have been used to estimate economic parameters such as the price and income elasticities of demand, the elasticity of supply, and the elasticity of adjustment for various commodities. Agricultural economists utilize these estimates in the evaluation of agricultural policies and programs. Hence, the properties of these estimates are of more than a little importance. This paper considers the statistical estimation of such models and, in particular, the effect of autocorrelated errors upon the estimates. With the aid of empirical examples we shall (1) illustrate the nature of the bias in ordinary least squares estimates arising from autocorrelated errors, (2) present a method of estimation yielding consistent estimates for the parameters of an equation containing autocorrelated errors, (3) demonstrate the low power of the Durbin-Watson statistic when applied to the residuals of an equation containing the lagged endogenous variable, and (4) present a method for computing confidence limits for the estimated long-run elasticities.