The power and robustness properties of tests for heteroskedasticity when the regressors are trended

Abstract

Even though it is widely known c hat heteroskedasticity reduces the efficiency of t,he OLS estimator, very little is known about what the impact trends have upon the power and robustness (to different trends in the data) of tests for heteroskedasticity. Research in this area fails to build the degree of severity of heteroskedasticity into tests designed to detect its presence. This paper evaluates the impacts of various trends in the regressors on the power and robustness of tests for heteroskedasticity through Monte Carlo studies and suggests that the power of each test increases when the degree of severity of'heteroskedasticity is built into the test statislic. 927