Does a non-linear mean reverting process characterize real GDP movements?

This paper uses non-linear models to investigate non-stationarity of real GDP per capita for seven OECD countries over the period 1900-2000. Unit root tests based on non-linear models are more powerful than traditional ADF statistics in rejecting the null unit root hypothesis. Empirical results show that, contrary to what the linear ADF statistics suggest, stationarity characterizes five out of the seven countries. This finding stands at variance with other recent studies which conclude that movements in real GDP per capita can be characterized as a nonstationary process.