Cointegrating Tests of Purchasing Power Parity Using Znz Patterned Vecm

This paper consists of two parts. In the first part the sensitivity of PPP testing to the nature of the unit root tests is demonstrated. Three unit root tests are employed for fourteen real bilateral exchange rates. The first two, the augmented Dickey-Fuller test and the Phillips-Perron test are procedures using the null hypothesis of a unit root. The third, the Kwiatkowski et al (1992) test, is a procedure which employs the null hypothesis of stationarity. The results are contradictory. PPP is consistently accepted or rejected for only four exchange rates. For the remaining ten, PPP is accepted using the Kwiatkowski et al procedure, but rejected by the augmented Dickey-Fuller test and the Phillips-Perron test. In the second part of this paper a new test procedure for PPP is introduced. In this procedure PPP is tested in the framework of zero-non-zero (ZNZ) patterned vector error-correction modelling (VECM). Since VECM modelling can accommodate both long-term and dynamic responses, this approach is different from those based on unit root tests. This new procedure adds to the available methods of testing for PPP. The results of this procedure are promising. Both the necessary and the necessary and sufficient conditions for PPP are sequentially tested. Of the fourteen exchange rates investigated, the necessary condition is found acceptable for seven and the necessary and sufficient condition cannot be rejected for three. These results support the existence of PPP, at least for some exchange rates. In analysing these results, this thesis also finds that the short-term dynamic structure may be an important influence in testing the necessary and sufficient condition for PPP. This suggests that a more sophisticated specification could be useful in improving the test for PPP