Effects of Applying Linear and Nonlinear Filters on Tests for Unit Roots with Additive Outliers
Conventional univariate Dickey-Fuller tests tend to produce spurious stationarity when there exist additive outlying observations in the time series. Correct critical values are usually obtained by adding dummy variables to the Dickey-Fuller regression. This is a nice theoretical result but not attractive from the empirical point of view since almost any result can be obtained just by a convenient selection of dummy variables. In this paper we suggest a robust procedure based on running Dickey-Fuller tests on the trend component instead of the original series. We provide both finite-sample and large-sample justifications. Practical implementation is illustrated through an empirical example based on the US/Finland real exchange rate series.