Testing unstable econometric models for stability. An empirical study

Abstract

Econometric models may be tested for stability using the asymptotic distribution of the dominant characteristic root of the system as derived by Theil and Boot (1962) or by Oberhofer and Kmenta (1973). This paper points out that this distribution is known for stable models only which implies that the null hypothesis must always be stability and the power of the test is in question. A Monte Carlo study is performed to investigate the power of the test and the distribution of the test statistic for unstable and stable models in the small sample case. © 1978.