This paper considers structural change tests under regression misspecificat
ions. The asymptotic distortions of regression misspecifications on the Sup
-Wald statistic are quantified for a class of drifting data generating proc
esses. It is found that there is no asymptotic distortion caused by regress
ion misspecifications when the misspecification term possesses no shifts. (
C) 2001 Elsevier Science B.V. All rights reserved.