Recently there has been a revival of interest in the determinants of long-run exchange rates. This interest has been generated in large part by developments in the time-series literature, particularly those relating to unit root and cointegration testing. The form of the long-run exchange rate that has received most attention is based on the doctrine of purchasing power parity (PPP). This paper presents an overview of the large number of contemporary tests of PPP.
Recent tests of PPP have been conducted in one of two ways. One approach involves examining whether nominal exchange rates are cointegrated with relative prices, while the other (which is complementary) seeks to determine if real exchange rates contain a unit root. This paper demonstrates that each of these approaches may be derived from a particular account of the balance of payments: generally speaking, the former kind of test stems from current account transactions, while the latter emanates from the capital account. Focusing on one or the other account of the balance of payments, however, may result in a misspecified relationship, especially when a researcher is using data from the recent floating experience. It is suggested that when data for this period are being used to test PPP, it would be better to consider the total balance of payments.
The paper identifies a general trend in recent empirical work on long-run exchange rate modeling, which is that PPP does seem to have some long-run validity. In particular, many currencies are found to have a unique cointegrating relationship between an exchange rate and relative prices, and real exchange rates display mean reverting behavior (two pieces of evidence that are complementary and supportive of a traditional form of PPP). Using a new data base and a variety of estimation techniques, this paper confirms these findings. However, the form of the long-run exchange rate relationship unearthed by recent work does not conform exactly to what many would understand as “traditional PPP.” Specifically, there appear to be extremely long-lived deviations from PPP, and the restrictions of symmetry and homogeneity of degree one often associated with PPP are usually rejected. The paper offers some explanations for the apparent discrepancy between the empirical and traditional versions of PPP.