Detecting Unbalanced Regressions Using the Durbin-Watson Test.

The aim of this paper is to provide a simulation-based analysis on the bahavior of the Durbin-Watson statistic as a misspecification test againstthe presence of unbalanced regressions among nonstationary fractionally integrated process. Theoretically, it is well-known that this statistic converges in probability to zero under this setup. In finite samples, however, from our Monte Carlo experiments it shows up a rather surprising behaviour with respect to the expected asymptotic one.