RDP 2001-04: Measuring the Real Exchange Rate: Pitfalls and Practicalities 5. Conclusion

This paper sets out some issues that must be considered when constructing real exchange rate indices for analytical or econometric work. We point out that there is no one single measure of the real exchange rate (or the nominal exchange rate, for that matter) that is ideal for all purposes. There are many feasible choices of weighting schemes and frequencies, price deflators, index number formulae and sets of countries to include. The appropriate choices depend on the task at hand.

Beyond these choices, there are some basic issues for which there is a preferred approach. Some of these are arithmetic considerations: geometric averages are preferable to arithmetic averages; period-average exchange rate indices should be constructed from period-average bilateral rates. Other issues relate to index number theory: if the quantities underlying the weights used in the index's construction are moving, it makes sense to allow the weights to change. If the weights change, however, it is essential that the index be spliced together as described in Section 2.

It may sometimes be necessary to make compromises between what is theoretically most appropriate and what is feasible given data availability and quality. In some cases, measurement issues might indicate one approach for econometric work, but this will be infeasible for day-to-day analysis and policy work. Accordingly, there is no guarantee that the appropriate exchange rate index for one task will be a good match for the data needs of another.