Detecting Regime Switching: Analytic Power

This paper is concerned with the power of tests for regime switching. Asymptotic results are currently available only for the null distribution, and this distribution is dependent on the underlying parameter space. This paper addresses the lack of asymptotic results. The asymptotic behavior of the test statistic is determined under a full range of drifting sequences of true distributions. The results are based on a construction of the test statistic in terms of Hermite polynomials. The number of terms in the polynomial construction depends on the size of the parameter space. Using this relation, a collection of power curves, indexed by the sample space, is obtained. The …nite sample properties of the test statistic are analyzed via Monte Carlo simulation. JEL Classi…cation: C12