Testing Markov switching models

Abstract

In this article, we propose a new test for Markov switching models. Unlike the tests in the existing literature (e.g. Hansen, 1992; Garcia, 1998; Cho and White, 2007), we focus on testing the null of two regimes, instead of one single regime, in a switching framework. To implement our test, we propose a Markov switching model with absorbing states and examine whether the absorption probabilities are close to the boundary of the parameter space. We exploit recent advances by Andrews (2001) and conduct inference in the proposed model.

Keywords:

• absorbing state
• boundary
• Markov switching model

JEL Classification:

• C22
• C51

Acknowledgements

We thank Chung-Ming Kuan and Shu-Ling Chen for valuable suggestions on early drafts of this article. All remaining errors are ours.

Funding

This work was supported by the National Science Council of the Republic of China [NSC-103-2410-H-007-007].

Notes

¹ Note that the initial value of was assumed to be either 2 or 3 and for .

² Andrews (2001) specified a set of high-level and primitive sufficient conditions under which the asymptotic null distributions of QLR, Wald and score tests were determined. However, verifying these conditions in the proposed model is beyond the scope of this article.

³ The dataset, named GDP4795.PRN, is taken from the web site of C. J. Kim and C. R. Nelson: http://www.econ.washington.edu/user/cnelson/ssmarkov.htm