Comparing cointegrating regression estimators: Some additional Monte Carlo results

This paper compares the finite sample performance of the canonical correlation regression estimator (CCR) and Stock and Watsonʹs (A simple estimator of cointegration vectors in higher order integrated systems, Econometrica, 1993, 61(4), 783–820) dynamic ordinary least squares estimator (DOLS) using the models proposed by Inder (Journal of Econometrics, 1993, 57, 53–68). The CCR estimator shows smaller bias than the OLS and the fully modified. The DOLS estimator performs systematically better than the CCR estimator.