Abstract
This paper derives panel unit root tests, based on the unit root tests in single time series by He and Sandberg (Oxf Bull Econ Stat 68:835–861, 2006), in a nonlinear dynamic heterogeneous panel accommodating smooth structural changes in the intercepts, in the dynamics, and in the trends. The finite sample size properties of the tests are satisfactory, and the tests are conspicuously more powerful than the panel data unit root tests by Im et al. (J Econom 115:53–74, 2003; Bull Econ Stat 67:393–419, 2005), and Pesaran (J Appl Econom 22:265–312, 2007) in panels subject to structural changes. To accommodate panels with cross-section-dependent idiosyncratic errors, a bootstrap version of the tests is considered. The tests are applied to unemployment rates in Scandinavia, and strong evidence against the hysteresis hypothesis is found. This is not the case if panel unit root tests based on a linear panel or a nonlinear panel with a break in the intercept are applied.