Abstract
Purchasing Power Parity is tested for in post-Bretton Woods real exchange rate data from 20 developed countries using univariate tests and covariate augmented versions of the Augmented Dickey–Fuller (CADF) and feasible point optimal (CPT) unit root tests. The covariates are a combination of stationary variables – inflation, monetary, income, and current account. A cross method comparison of the results is performed. Very strong evidence is found of PPP using the CPT test, rejecting the unit root null for 12 out of the 20 countries at the 5% significance level or better, and six more at the 10% level. Much less evidence is found of PPP with the CADF and univariate tests.
Acknowledgements
We are grateful to Bruce Hansen, Michael Jansson, and Elena Pesavento for helpful comments and discussions. Papell thanks the National Science Foundation for financial support.
Notes
1 Bowman (Citation1999) shows that the tests with heterogeneous speeds of mean reversion have higher power for mixed panels. In the context of testing for PPP, however, higher power to reject the unit root null in favour of the alternative that at least one real exchange rate is stationary is the same as worse size for the alternative that all real exchange rates are stationary.
2 They find rejections for France at the 5% level and Italy, Germany, Netherlands, Norway and the UK at the 10% level.
3 Caporale and Pittis (Citation1999) use CADF to analyse US macroeconomic data as in Nelson and Plosser (Citation1982) and in most cases end up reversing the finding of a unit root.
4 The time period for the data is not extended beyond 1998 when the Euro was adopted and the nominal exchange rates within the Eurozone fixed.
5 Chang et al. (Citation2001) apply CADF to covariates from the extended Nelson-Plosser data set for the post-1929 samples as well as postwar annual CPI based real exchange rates for 14 OECD countries.
6 The growth rate of the current account of the USA relative to the income of the US Δ(caus yus) is not used because of its low variance.