Multiple structural change tests by Bei and Perron (1998) are applied to the regression by Demetrescu, Kuzin and Hassler (2008)
Non-technical summaryOur paper contributes to the existing literature by introducing and applying a new test for determining the timing, number and significance of breaks in the persistence of time series. Here, persistence is modeled by the length of the time series' memory, measured by the degree of fractional integration. To the best of our knowledge, there exists no other method which allows to test for multiple unknown break points in the degree of fractional integration. One advantage of our method is its robustness against variations in volatility. Furthermore, the test may find breaks relatively close to the boarder of the observed sample. In a sample of 500 observations, for example, the test only needs 150 observations before or after the break in order to reliably determine the break if the difference in persistence is large enough.Knowledge about the (change of) persistence of economic time series is important in many policy and research areas. In the context of inflation, the degree of persistence provides information on the effectiveness of monetary policy. The better inflation expectations are anchored, the shorter the effect of a price shock will last less time a price shock will influence inflation and the lower inflation persistence will be. A break point test can thus, in retrospect, indicate if and how a change in monetary policy affects inflation persistence.In the literature, there is still disagreement about whether and when the U.S. inflation persistence has changed over the past five decades. Opposing empirical results can in part be attributed to differences in the measurement of inflation persistence. Based on the results of the test proposed in this paper, we come to the conclusion that there has been a break towards higher persistence in 1973. The break date roughly coincides with the collapse of the Bretton Woods system, the beginning of a period of high inflation and some oil price shocks. By contrast, the test does not confirm that a break took place in the 1980s as suggested by some previous studies.
Nichttechnische Zusammenfassung