然而,泰勒& Sarno(1998)指出,面板数据可能是误导,因为虚假设调查的共同non-stationarity实际汇率的使用购买力平价的面板数据测试,只要有一个系列是静止不动的,可以拒绝零假设。因此,如果测试是在一群汇率和拒绝零假设,结果不能非常重要。因此不能的结果这一理论存在的长期购买力平价条件。因为泰勒& Sarno(1998)提出了批评的面板数据方法,有许多改进的测试面板数据方法,如高的(1999)面板数据协整测试基于残差,Hadri(2000)LM测试使用不均匀的面板数据,崔的(2001)单位根和协整测试基于联合P值,Chang(2002)的SN测试,等。尽管面板数据测试方法不断改进,结论仍不能一致的PPP测试。因此,尽管面板数据测试被广泛使用,它是发展和完善。
But long-time period data test is also criticised. Since the data period is very long, so it will involve a variety of different exchange rate regimes. Moreover, the real external shocks may cause the structural break of the real exchange rate or the equilibrium exchange rate shifting, and the use of long-term data is difficult to test the structural break. Also, because the long-term historical data is only from the industrial countries, the test results may be exaggerated. These test results caused by the large amount of samples has been confirmed in some studies, such as Engel, Kim (1999), Engel (2000). To provide a convincing test to the stationarity of real exchange rate in latter Breton Woods’s system, it is necessary to design a test only using data from this period of time, so panel data is applied in empirical tests of PPP.
However, Taylor & Sarno (1998) pointed out that the panel data may be misleading, because the null hypothesis investigates the joint non-stationarity of real exchange rate in the use of panel data tests of purchasing power parity, as long as there is one series which is stationary, the null hypothesis can be rejected. Therefore, if the test is under a group of exchange rates and rejects the null hypothesis, the result cannot be very significant. Hence this theory cannot be the result of the existences of all long-term PPP conditions. Because Taylor & Sarno (1998) proposed a paper which criticized the panel data method, there had been many improvements to tests of panel data methods, such as Kao’s (1999) panel data co-integration test based on residuals, Hadri’s (2000) LM test using non-homogeneous panel data, Choi’s (2001) unit root and cointegration tests based on joint P values, Chang’s (2002) SN test, etc. Although the panel data test method has been continuously improved, the conclusions still cannot be consistent in terms of PPP test. For that reason, although the panel data test is widely used, it is developing and consummating.