There is an extensive statistics and econometrics literature on statistical tests for structural changes at an unknown time. The research on these tests reflect their important role in getting accurate, unbiased estimates of economic relationships from time series data. The contribution of this project comes from providing a better understanding of the properties of existing tests for structural change and from developing new tests with better properties. These tests are then used to improve empirical studies in macroeconomics and financial economics. This project analyzes the behavior of tests for structural change in a time series context that ultimately allows for series having a general correlation structure and possible nonstationarity, either of the form of a polynomial time trend or of a stochastic nature induced by the presence of a unit root. The components of the investigation consist of the following: 1) Analysis of the finite sample properties of current available tests for structural change which allow for the presence of serial correlation, e.g., CUSUM, Chow test, LR test, and others; 2) Explanation of the finite sample behavior using non-standard asymptotic distribution theory, e.g., continuous-time asymptotic and small-sigma asymptotic; 3) Development of alternative statistical procedures that have better finite sample properties; 4) Extensions of currently available procedures to the case where the time series is characterized by either a deterministic or stochastic trend; 5) Derivation of the asymptotic distributions and the analysis of the finite sample properties for tests for structural change with trending; 6) Empirical application to issues related to macroeconomic variables with special emphasis on testing for a change in the trend function of a univariate time series.
