Title: Level shift estimation in the presence of non-stationary volatility with an application to the unit root testing problem
Authors: David Harris - University of Melbourne (Australia) [presenting]
Abstract: The aim is to investigate the properties of the standard residual sum of squares (RSS) based estimators for the location of a level break in cases where the driving innovations are heteroskedastic, displaying non-stationary volatility (permanent changes in unconditional volatility) and/or conditional heteroskedasticity. Although designed for homoskedastic innovations, the RSS estimator retains its usual rates of consistency under such forms of heteroskedasticity. However, we present simulation evidence which highlights the potential for a very serious decline in the finite sample performance of the RSS estimator, relative to the homoskedastic case, when heteroskedasticity is present. As a consequence, we explore a weighted version of the estimator based around an adaptive estimate of the volatility path of the innovations. The consistency of the weighted estimator is demonstrated and their finite sample behaviour explored. Where the level break is located in a low volatility regime this is shown to deliver very significant improvements.