Publication Date: July 2019
Revision Date: October 2019July 2022
This paper is concerned with possible model misspeciﬁcation in moment inequality models. Two issues are addressed. First, standard tests and conﬁdence sets for the true parameter in the moment inequality literature are not robust to model misspeciﬁcation in the sense that they exhibit spurious precision when the identiﬁed set is empty. This paper introduces tests and conﬁdence sets that provide correct asymptotic inference for a pseudo-true parameter in such scenarios, and hence, do not su↵er from spurious precision. Second, speciﬁcation tests have relatively low power against a range of misspeciﬁed models. Thus, failure to reject the null of correct speciﬁcation does not necessarily provide evidence of correct speciﬁcation. That is, model speciﬁcation tests are subject to the problem that absence of evidence is not evidence of absence. This paper develops new diagnostics for model misspeciﬁcation in moment inequality models that do not su↵er from this problem.
Supplement pages: 153
Keywords: Asymptotics, confidence set, diagnostics, identification, inference, misspecification, moment inequalities, robust, spurious precision, test
JEL Classification Codes: C10, C12CFDP 2184CFDP 2184R