Testing for Factor Loading Structural Change under Common Breaks

摘要

This paper proposes a new test for factor loading structural change in dynamic factor models. The proposed test is robust to the nonmonotonic power problem that occurs if the factor loadings exhibit structural changes at common dates over cross-sections. To illustrate the usefulness of our test, we first show that the leading test proposed by Breitung and Eickmeier (2011) exhibits nonmonotonic power, essentially because the breaks are considered as spurious factors with stable factor loadings. We use both local and non-local asymptotic frameworks to investigate the power of their test. The new test eliminates the effects of the spurious factors by maximizing the test statistic over possible numbers of the original factors. This approach is effective because the original factors are not identified under the alternative hypothesis. Monte Carlo simulations and an empirical example using U.S. Treasury yield curve data clearly illustrate the validity of the asymptotic power analysis and usefulness of the proposed test.