On the origins of high persistence in GARCH-models

Abstract

The estimated persistence in various types of GARCH-models is known to be too large when the parameters of the model undergo structural changes somewhere in the sample. The present paper adds further insights into this phenomenon for the Baillie and Chung (2001) minimum distance estimates of the model parameters. While previous research has focused on the effects of changes in the GARCH-parameters, we investigate here the consequences of a changing mean.