The Empirical Study On The Purchasing Power Parity Of Chinese RMB's Real Exchange Rate

The Purchasing Power Parity (PPP) theory is the basis of the Exchange Rate decision theory, so it's important to give an empirical study on it, especially under the global economy background. Because of the advantages of straight and easy to understand of PPP, people always wish to do empirical study on it, although the hypothesis of it is very strict. Almost every new econometric method appears, it will be used to the empirical study on PPP, because PPP theory is very important to the real economy, just like forecasting finance crisis, using as exchange rate normal anchor and judging exchange rate mismatch. With the developing of Chinese economy, it's important too to give an empirical study on the PPP of Chinese RMB's real exchange rate.Since PPP theory had been taken out, there were many papers about testing it, but the results always were blow hot and cold. This paper will give an empirical study on the PPP of Chinese RMB's real exchange rate through tow ways as follow: first, we will calculate the half-live of unit shock (HLS) of RMB's real exchange rate based on the linear AR model, using the method of median unbias estimation to test if the Rogoff Puzzle of Chinese RMB exist or not; after that we will test if the revert to PPP of the real exchange rate of Chinese RMB exist or not based on the nonlinear model.To AR model, the OLS method has an under estimation, the median unbias estimation is a very good way to correct the bias. Based on the exchange rate of Chinese RMB from January 1994 to May 2004, we calculate the HLS of Chinese real exchange rate, using the median unbias estimation method and the Boostrap simulation, and the result is that Chinese real exchange rate's HLS is 2.1 years, and the rate that it converges to PPP is 24% per year which indicates that the problem of Rogoff Puzzle maybe not so sierious as Rogoff(1996) had said. Another result is the probability that the Chinese real exchange rate is stationary is 95%, which indicates the probability that the PPP of Chinese real exchange rate is stand is 95% in the sample period based on the linear AR model.But there are many shortages to characterize the behaviors of the real exchange rate based on the linear AR model. Because of the existence of the trade cost, it will be more acceptable to characterize the behaviors of real exchange rate based on the nonlinear model. Granger and Terasirta(1993) brought forward a nonlinear STAR model which has been used widely in the exchange rate research. This paper will do the empirical study on the PPP of Chinese exchange rate on the STAR model, and give an improvement to the transition function of it. The result of my study based on the exchange rate of Chinese RMB from January 1980 to October 2005 indicates that the behaviors of Chinese RMB's real exchange rate obeys to the LSTAR model, which indicates that Chinese RMB's real exchange rate is globlely stationary and converges to the PPP, and the dynamic behavior of Chinese RMB's real exchange rate is asymmetry, the behavior that Chinese RMB's real exchange rate converges to PPP not only dependence on the size of the departure to the PPP, but also on the direction of the departure. Through the compare on simulation, the nonlinear LSTAR model is more excellent than the linear AR model to characterize the behaviors of the real exchange rate of Chinese RMB, and the improved LSTAR model is better than the unimproved model. During the sample period, the real exchange rate of Chinese RMB is in the upper regime in most time, in which Chinese RMB's real exchange rate is a stationary process of AR(1) model, and is in the lower regime in some time, in which Chinese RMB's real exchange rate is a expand process of AR(1) model, this means Chinese RMB's real exchange rate is stationary in the long run and unstationary in the short run. The transition function of Chinese RMB's LSTAR model is very steep, this means the transition between lower regime and upper regime is very quick, which close to the TAR model. The impulse response function of the estimated LSTAR model describes that with the stock becoming larger the impulse response function becomes steeper, this means with the departure becoming larger the reverting of the real exchange rate of Chinese RMB to PPP becomes sooner. When the direction of the stock is not the same, the function of the positive stock is larger than that of the negative shock, this means that the adjustment to PPP when the real exchange rate of Chinese RMB is overate is sooner than when it is underrate. At the same time, the HLS of the nonlinear model is not only related to the size of the shock, but also to the history and the direction of the shock. Using January 1980 as the start time, we calculate the HLS of the real exchange rate of Chinese RMB, which is 18 months, while when the shock is 1% the HLS of the real exchange rate of Chinese RMB is 48 months, indicates that the problem of Rogoff Puzzle of Chinese RMB exist partially. Using the data from October 1995 to October 2005, we do a twelve step ahead forecasting and the result shows that the effect of the forecasting of the nonlinear LSTAR model is better than that of the linear AR model, and the improved LSTAR model is better than the model hasn't been improved.In the end, according to the study result of this paper, we put forward some advices about Chinese exchange rate policies, and give a prospect to the future study of our work.