We introduce test statistics based on generalized empirical likelihood methods that can be used to test simple hypotheses involving the unknown parameter vector in moment condition time series models. The test statistics generalize those in Guggenberger and Smith (2005) from the i.i.d. to thetime series context and are alternatives to those in Kleibergen (2001) and Otsu (2003). The mainfeature of these tests is that their empirical null rejection probabilities are not affected much by thestrength or weakness of identification. More precisely, we show that the statistics are asymptoticallydistributed as chisquare under both classical asymptotic theory and weak instrument asymptoticsof Stock and Wright (2000). A Monte Carlo study reveals that the finitesample performance of thesuggested tests is very competitive.
Generalized empirical likelihood tests in time series models with potential identification failure
7 April 2005
Working Paper (CWP01/05)