Score-type tests for Markov switching models

Speaker: Prof Enrique Sentana

Affiliation: CEMFI

Location: Room 214, Chamberlain Building (#35), St Lucia Campus

Abstract: Testing the null of a single regime in Markov switching models is complex because some parameters become increasingly underidentified along alternative paths to the null, others are inequality constrained, and several higher-order derivatives are 0, which render the maximum of the alternative model log-likelihood function, and therefore the likelihood ratio, numerically unreliable. We propose score-type tests asymptotically equivalent to the likelihood ratio as the largest of two simple intuitive statistics computed under the null hypothesis treating symmetrically all ways of approaching it without excluding any region of the parameter space. We establish the admissibility of our tests and consequently of the likelihood ratio. Our simulation results confirm the power of our proposals, and our empirical application to exchange rates their ability to distinguish genuine dual regimes from spurious ones.