Testing and Identifying Structural Change in a Cointegration Regression Testing and Identifying Structural Change in a Cointegration Regression

Testing and Identifying Structural Change in a Cointegration Regression Jae-Young Kim 1 Department of Economics SUNY-Albany Albany, NY 12222 January 1996 Abstract This paper studies how to detect structural change in a cointegrated system under the situation of the change period being unknown. A general type of structural change is considered that causes the failure of an initial cointegration relation or the failure of an initial model speci cation. We derive a statistic and investigate asymptotic behavior of it for testing the occurrence of such a structural change. Our procedure is useful for identifying the nature of a structural change in a cointegration relation as well as for testing the occurrence of a structural change. We show that our test procedure is consistent under fairly general alternatives. Also, a Monte Carlo study shows that our test has reasonably good size and power properties in nite samples. We apply our test to two examples of cointegrated models, the Phillips curve and a model for interest rate determination. For U.S. postwar time series we have found strong evidence of structural change with some interesting features.