Publication: Journal of Economic Education
Volume: Volume 27, No. 1
Issue: Winter 1996
Pages:
Author(s): Asatoshi Maeshiro (University of Pittsburgh)
Address (Principal Author):Asatoshi Maeshiro, Department of Economics, University of Pittsburgh, Pittsburgh, PA 15260, (412) 648-1760
Internet Address (Principal Author): aiueo@vms.cis.pitt.edu
Title: Teaching Regressions with a Lagged Dependent Variable and Autocorrelated Disturbances
Abstract: The article attempts to rectify the unsatisfactory textbook treatment of the finite-sample properties of estimators of regression models with a lagged dependent variable and autocorrelated disturbances. It contents that the bias of the OLS estimator of a regression model with a lagged dependent variable and autocorrelated disturbances is determined by two effects, the dynamic effect and the correlation effect. It then argues that when both the coefficient of the lagged dependent variable and the autocorrelation coefficient of AR(1) disturbance are positive, the two effects have opposite signs, thus making the OLS estimator perform well in terms of bias. The article lists the factors that produce the two effects and then examines the past Monte Carlo studies that produced conflicting results and provides a reason or reasons for the main results.