Regression-Based Tests for Persistence in Conditional Variances.

This paper considers regression-based test criteria for the hypothesis of conditional variance nonstationary in the logarithmic family of GARCH processes. The tests are based on teh ARMA representations that appropriate nonlinear transformations of GARCH-type processes admit. Simulation experiments investigate the performance of the tests in finite samples, both in the presence and absence of a structural change in the conditional variance process. The methods are also used to test the hypothesis of integration in variance for some economic time series.