Abstract

In this paper, we investigate monitoring tests for an explosive and collapsing behavior of financial time series. In general, it is often the case that the usual demeaning or the quasi-GLS type demeaning have been implemented in the literature to control the false alert rate of the monitoring procedure and we develop asymptotic theory under the local alternative as well as the mildly explosive or collapsing alternative. When testing for a bubble, the performance of these two methods are comparable, whereas the quasi_GLS type demeaning is found to work poorly in view of power.