As mentioned in chapter two, since the relative literature using unit root and cointegration technique to test the real exchange rate were mostly published in the 1980s, the sample period is just 15 years (that’s because only after 1973 the floating exchange rate is implemented). Even though the data can be 10 years more, the test accuracy only increases only slightly. Due to the limitation of statistical software and restriction of data source, the specific data is limited to 10 years in this dissertation. Hence the sample size is not big enough. The Monte Carlo experiments also show that we cannot avoid the inefficiencies in traditional unit root tests by increasing the observation frequency. Although expanding the observation point from the annual or quarterly data to monthly data increases the number of available sample data, it cannot fundamentally solve the problem. From the perspective of the spectrum, we examine the behavior of the low-frequency components in the real exchange rate, which necessarily requires a long period of test data to improve inefficiencies. Basically, if data is generated from longer time period, the test outcome maybe more convictive.In the verification process of purchasing power parity hypothesis, the main price indexes are the following: consumer price index (CPI), producer price index (PPI), wholesale price index (WPI) and GNP deflator. Choosing different price indexes can lead to different test result of purchasing power parity. As discussed in the previous chapter, one common view is that the exchange rate is the relative price of tradable goods so the price index should only include the tradable goods.
Since the statistical methods vary greatly and price indexes published by national statistical offices are inconsistent in different Asian countries, in this paper we can only select the most universally applicable price index¬—CPI rather than wholesale price index which includes less non-tradable goods. This has a certain impact on the empirical test results of this paper.