Powerful Tests of Structural Change That are Robust to Strong Serial Correlation

摘要: This paper proposes powerful and serial correlation robust test statistics that can be used to for the presence of structural change in trend function a univariate time series. Four models are analyzed, each model corresponding different way which break might occur. Given model, proposed tests designed detect single at an unknown date. The do not require knowledge form data, they made highly persistent unit root errors by using more comprehensive version scaling factor approach Vogelsang (1998b). utilize popular nonparametric kernel variance estimators. fixed-bandwidth asymptotic framework, Kiefer (2003), is approximate effects estimators on statistics. framework makes possible choice bandwidth deliver with maximal power within specific class tests. For tests, concrete recommendations practice. recommended shown have good finite sample size properties.