International evidence on shock persistence: structural change, nonlinearities and subsample robustness

Testing for unit roots and the related issue of measuring shock persistence has attracted considerable theoretical and applied econometric interest. The issue of the size of the random walk component raised by Cochrane (Journal of Political Economy, 96, 893-920, 1988) has been extended in the work of other authors. Leung (Economics Letters, 40, 435-44, 1992), in particular, considers the effects of structural breaks on measures of persistence. In this paper we consider new results for the effects of structural change, nonlinearities and subsample robustness on spectral-based measures of persistence illustrating the potential problems via an updated and extended version of the Cogley (Journal of Political Economy, 98, 501-18, 1990) data set. We find that significant structural breaks exist in the majority of the series investigated. Furthermore, measured persistence differs markedly across distinct periods with the assumption of common growth rates over very long periods leading on occasions to potentially erroneous conclusions on the degree of persistence. Nonlinearities measured by the significance of time squared (TSQ), seem important in many cases, particularly post-World War Two (WW2), potentially explaining the apparently high levels of post-WW2 persistence. The paper concludes with a warning on the use of measures of persistence when the data include structural change and/or nonlinearities, highlighting the importance of the correct choice for breakpoints and second sub-sample start points.