Stationarity tests with unattended nonlinearity

The paper develops a test with the null of stationarity that allows for the possibility of an unknown number of structural breaks, or other nonlinearities, in the data-generating process. The test is based on the fact that the behavior of a breaking process can often be captured using a single frequency component of a Fourier approximation. Hence, instead of selecting specific break dates, the number of breaks, and the form of any nonlinearities, the specification problem is transformed into selecting a low frequency component to include in the estimating equation. Our proposed test does not exhibit any serious size distortions, and shows reasonable power. The appropriate use of the test is illustrated using real exchange rates in the post-Bretton Woods period