Testing linear restrictions in linear models with empirical likelihood

In this paper we analyse the higher order asymptotic behaviour of a profiled empirical likelihood ratio which can be used to test a set of linear restrictions in linear regression models. We show that the resulting profiled empirical likelihood ratio admits a Bartlett correction which can be used to improve to third order the accuracy of commonly used tests in applied research without any distributional assumptions about the error process. Copyright Royal Economic Society 2002