Abstract
The performance of the Phillips and Perron unit root tests is examined using the optimal bandwidth selection procedure of Andrews and the pre-whitening procedure of Andrews and Monahan. It has been shown that they do not make significant improvements over the tests using fixed truncation lags.