Are real exchange rates mean reverting? Evidence from a panel of OECD countries

Abstract

In our article we employ some contemporaneous panel unit root tests (Maddala and Wu, 1999; Im et al., 2003) to examine whether the real exchange rates are mean reverting. Considering a panel of 26 OECD countries from 1987 to 2006 both using monthly and quarterly observations, we find that assuming a panel framework significantly increases the power of unit root tests. As a result, we find that the nonstationarity of the real exchange rate has strongly been rejected in favour of giving support to the purchasing power parity.

Notes

¹ Among many others, Froot and Rogoff (1995), Rogoff (1996), Taylor (1996), Taylor (2000), Sarno and Taylor (2002), Taylor and Taylor (2004), Taylor (2006) and a recent paper by Saatcioglu et al. (2007), of which the latter is constructed upon the Turkish economy, can be considered highly illuminating papers for the main theoretical issues related to the PPP relationship. Besides, Lothian and Taylor (1996) using a long period of two centuries data for dollar- and franc-sterling exchange rates considering also sub-divisions as to regime changes yield a strong support to the mean-reverting characteristic of real exchange rates thus to the role of PPP as a long-run equilibrium condition.

² See Oh (1996), MacDonald (1996), Wu (1996), Lothian (1997), Papell (1997), Flôres et al. (1999), Heimonen (1999), Wu and Wu (2001), Chiu (2002), MacDonald et al. (2002), Alba and Park (2003) and a recent paper by Alba and Papell (2007) for the applications of panel unit root tests.

³ However, some essential criticism dealing with considering the power of panel unit root tests come from Taylor and Sarno (1998). They emphasize that results from panel unit root tests should be appreciated by researchers in a cautious way, for they found through some Monte Carlo experiments that rejection of the null hypothesis of joint nonstationarity of a group of real exchange rates may be due to as few as one of the exchange rate series of interest tend to be generated by a stationary process. On the other hand, many other papers emphasize the importance of possible bias in panel tests due to the cross-sectional dependence as well. For instance, O'Connell (1998) reveals that standard practice of calculating all real exchange rates relative to the US dollar leads to cross-sectional dependence in panel data. Kuo and Mikkola (2001) also emphasize the possibility that real exchange rates may be highly dependent, since there exists economic dependence between the countries' price levels and exchange rates. Besides, dependence may be due to the construction of the real exchange rates with respect to some benchmark currencies so that any independent variation in the benchmark country's price level or the value of its currency shows up in all the real exchange rates. Following Koedijk et al. (1998), adjusting for this problem makes it much more difficult to reject the random walk in real exchange rates. We leave the investigation of such issues of theory to future papers and thus the rest of the panel tests considered here are built under the assumption of cross-sectional independence.

⁴ As suggested by most empirical studies and following Chiu (2002), time trend in real exchange rates is not consistent with the long-run PPP, and adding a trend in the specification may deteriorate the testing power due to a loss in degrees of freedom. Therefore, we exclude the time trend in our model specification.

⁵ The countries considered are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, UK and the US.

⁶ For a more detailed exposition see OECD Main Economic Indicators.

⁷ All the results are estimated in EViews 5.1.

Table 1.  IPS unit root test null hypothesis: unit root

Table 2.  Intermediate ADF test results

⁸ All these estimation results not reported here to save space are available from the authors upon request.