On the Formulation of Wald Tests on Long-Run Parameters.

In a single-equation error correction model, two alternative formulations of a linear hypothesis on the long-run parameters and associated Wald test statistics are shown to arise from the covariance matrix estimator. A choice between the statistics is based on invariance properties and on lack of moments considerations. As a consequence, the use of asymptotic standard errors of long-run parameters is questioned. Copyright 1993 by Blackwell Publishing Ltd