In this paper we re-examine whether purchasing power parity holds in the long run in China from a two-steps procedure correcting outliers and testing unit roots. Thus, the efficient unit root tests developed by Elliott, Rothenberg and Stock (1996) and Ng and Perron (2001) are applied on the Renminbi bilateral (to the US dollar) real exchange rate, corrected from outliers, over the period 1970 to 2006 (in monthly frequency). We underlined the effects of large, but infrequent shocks due to changes of Chinese exchange policy undertaken since the China’s foreign exchange reform on the real exchange rate, in particular several devaluations between 1984-1994. We also show that there is no tendency to the purchasing power parity in China to hold in the long run during this period.

In this paper we use a Threshold AutoRegressive (TAR) model to capture the nonlinear dynamics of monthly real effective exchange rate data for the G7 countries. The novelty of our approach relates to the use of the real interest differential as the switching variable. This choice allows us to consider jointly the nonlinearity and nonstationarity issues using recent advances in asymptotic theory. We find that the null of linearity is easily rejected against the nonlinear model for all currencies considered. Further, for five out of the seven countries, where the null of unit root is rejected, we report evidence of quite rapid mean reversion.