The common-trend and transitory dynamics in real exchange rate fluctuations

Abstract

This study examines the behaviour of both Common Trend (CT) and transitory components of Real Exchange Rate (RER) fluctuations under the current float. The CT component is in most cases found to be sizeable, albeit its relative importance can vary considerably across major currencies and its estimate can be sensitive to whether or not long-run Purchasing Power Parity (PPP) is imposed on the data. Further analysis suggests that both CT and transitory innovations are linked much more to interest rate changes than to productivity changes. Accordingly, it is interest rate, not productivity, disturbances that drive the highly persistent RER.

Acknowledgements

This is a substantially revised version of the discussion paper ‘Productivity shocks, monetary shocks and the short- and long-run dynamics of exchange rates and relative prices’. We have received valuable comments from Niels Arne Dam, Søren Johansen, Katarina Juselius, Ronald MacDonald, and seminar participants at the University of Copenhagen and at the 2005 AEA Conference on Exchange Rate Econometrics. Michael Bergman gratefully acknowledges financial support from the Swedish Research Council.

Notes

¹ Engel (2000) observes that standard tests for stationarity cannot determine the actual relevance of permanent dynamics in RERs when transitory dynamics are more volatile. Hence, even when finding stationarity in RERs, real shocks may still contribute to the persistence of RERs. Indeed, Rogoff (1996) points out that the empirical rate of RER reversion seems too sluggish to be explained by purely nominal shocks.

² A popular approach used in previous studies of RER dynamics is the structural VAR method (see, e.g. Clarida and Gali, 1994; Eichenbaum and Evans, 1995; Rogers, 1999). This approach studies the determinants of the RER itself. It does not investigate the possibly different determinants of the CT and transitory components for the RER.

³ If cointegration exists but PPP does not hold, then CT innovations can have long-run effects on the RER.

⁴ Exchange rate data for Germany are denominated in euros instead of deutsche marks from January 1999 onwards. Data conversions are made using the official conversion rate of 1.95583 deutsche marks per euro.

⁵ In addition to just using the SIC-determined lag order, we also checked and tried out with different lag parameter values, including k = 1, 2, 3 and 4. Largely, similar test results were obtained with these different lag specifications.

⁶ We allow for a linear trend in the cointegration vector for both the UK and the German cases. In the Japanese case, a linear trend is permitted in the levels.

⁷ The notably different estimate of the cointegration vector for the UK data are puzzling. It may suggest the possibility of some structural change in the data. If a structural break exists, it could seriously distort the data and confound the estimation procedure. While the relevance of the structural-break explanation is beyond the scope of this analysis, we may still keep this possible explanation in mind when interpreting the UK results.

⁸ To further examine the robustness of our results with respect to the lag choice, we experimented with alternative lag specifications, different from that suggested by the SIC. These various specifications were found to produce qualitatively similar results to those reported here.