Testing linear restrictions on cointegrating vectors Sizes and powers of Wald tests in finite samples

Abstract

The Wald test for linear restrictions on cointegrating vectors is compared infinite samples using the Monte Carlo method. The Wald test within the vector error-correction based methods of Bewley et al. (1994) and of Johansen (1991), the canonical cointegration method of Park (1992), the dynamic ordinary least squares method of Phillips and Loretan (1991), Saikkonen (1991) and Stock and Watson (1993), the fully modified ordinary least squares method of Phillips and Hansen (1990), and the band spectral techniques of Phillips (1991) are considered. In terms of test size, Johansen's method seems to be preferred, and in terms of test power it is Park's and Phillips'. However, the relatively poor results in the context of cointegrating regressions suggest that improvements on the performance of the Wald tests considered here are needed.