The Exact Powers of Some Autocorrelation Tests when the Disturbances are Heteroscedastic

This paper considers the exact finite sample powers of five popular tests for AR(1) disturbances when one of several types of heteroscedasticity is also preseut. Severe reductions in power are found, particularly under strong positive autocorrelation. Factors influencing these power reductions are identified analytically and the limiting powers are also considered