Title: Detecting end-of-sample explosive behaviour using persistence change tests
Authors: Sam Astill - University of Essex (United Kingdom) [presenting]
Abstract: The purpose is to explore the possibility of utilising persistence change tests to detect a change in persistence in an economic time series from being an I(1) process to an explosive process at the end of the sample. The first differences of a series that is subject to a change in persistence from I(1) to explosive behaviour at the end of the sample is itself a series that follows an I(0) process with a shift to explosive behaviour at the end of the sample. As such, we examine the ability of tests for a change in persistence of a series from being I(0) to I(1) to detect explosive end-of-sample bubbles when applied to the first differences of a financial time series. We find, unsurprisingly, that persistence change tests reject for explosive alternatives in addition to the I(1) alternative for which they are designed. We also find that persistence change tests have good power to detect short lived end of sample explosive episodes as well as ongoing episodes of a longer duration. A test with such power properties will be desirable for practitioners, as existing tests are usually best suited to detecting explosive episodes of a particular duration.