Interaction entre racines unitaires et ruptures structurelles

Abstract

In this paper, we review the recent econometric methods related to unit root tests. The central idea is the interaction between structural breaks and unit roots. We consider the standard Dickey-Fuller test and its modifications that allow under the alternative hypothesis one or multiple structural breaks. The break dates are endogenous and the number of breaks may be unknown. We investigate the size and power of these tests. Thus we consider the problem of estimating the number of structural breaks and the problem of estimating the break dates. A second type of test is reviewed, the LM unit root tests that allow under the null and the alter native hypothesis one or two unknown breaks. We also discuss the tests of structural breaks built for a stationary variables. We distinguish two types of tests : tests for a single break and tests for multiple breaks. © Presses de Sciences Po. Tous droits ré servés pour tous pays.