Likelihood-based tests for parameter constancy in I(2) CVAR models with an application to fixed-term deposit data

摘要

This paper explores likelihood-based tests for parameter constancy in I(2) cointegrated vector autoregressive (CVAR) models. A new class of test statistics for parameter stability is introduced in the I(2) CVAR framework. This study proves that their asymptotic distributions take non-standard forms involving the integrals of Brownian motions, but they are free of any nuisance parameters. It is thus feasible to approximate the non-standard distributions by simulation. Selected quantiles of the approximate distributions are presented as statistical tables for applied use. Monte Carlo experiments are also conducted to investigate finite-sample properties of the test statistics. Finally, an empirical study of Japan's fixed-term deposit data is performed to demonstrate the practicality of the proposed tests in applied research. (C) 2020 Elsevier Inc. All rights reserved.