Real Exchange Rates in the Long and Short Run: A Panel Co-Integration Approach

The empirical literature on long-run real exchange rate behavior has shown mixed evidence due to problems involving the lack of long time series data and the low power of time-series unit root tests in small samples. The main objective of the present paper is to tackle these empirical issues by applying the recently developed panel cointegration techniques to the long-run real exchange rate equation implied by our model. Using annual data for 67 countries over the 1966-97, we find that the cointegrating relationship between the real exchange rate, the ratio of net foreign assets to GDP, the relative Home to Foreign productivity of the traded and non-traded sector, and the terms of trade is valid in the long run. This result holds for all sub-sample of countries (whether they are classified by income per capita or capital controls). Furthermore, our coefficient estimates are consistent with the theoretical values implied by the calibrated parameters of preferences and technology in Stockman and Tesar (1995). Robustness checks reveal that: (i) “pooling” the data to obtain a common long-run equilibrium relationship across countries is valid for the samples of countries with high income and low capital controls, (ii) the oil shock crisis in 1973 represents a structural change for these sub-samples. Finally, deviations from the equilibrium are large and persistent with half-life estimates (between 2.8 and 5) consistent with the consensus interval of 2.5-5 found in the literature (Murray and Papell, 2002).