The asymptotically efficient version of the information matrix test in binary choice models. A study of size and power

Abstract

As Newey (1985) and Orme (1988) argue in the context of discrete binary choice models, the test of the information matrix (IM) is sensitive to heteroscedasticity and the incorrect distribution of the error term, with both these problems leading to inconsistency of the estimators obtained. This paper uses simulation experiments to analyse the size and power of the asymptotically efficient version of this test, with the aim of obtaining evidence on its capacity to detect such specification errors, considering different alternatives.