There have been many realignments in the Exchange Rate Mechanism (ERM) of the EMS. In Discussion Paper No. 552, Andrew Rose and Research Fellow Lars Svensson develop an empirical model of exchange rate realignments, which they use to focus on the behaviour of the French franc/Deutschmark exchange rate using daily data for the period of the ERM. Throughout 1979-90, the variation of this rate has been restricted within a band or `target zone' of 2.25% around its central parity.

Assuming uncovered interest parity, Rose and Svensson show that there is probably no significant risk premium separating interest rate differentials from expected changes in the exchange rate. They decompose the exchange rate into the expected rate of change within the band (or `depreciation') and any expected change in the position of the band (or `devaluation'), in order to estimate the latter as the difference between the interest rate differential and an estimate of the expected rate of depreciation.

They develop a number of simple empirical models of the exchange rate to show that when the exchange rate is perturbed from the middle of the band, it tends to revert towards the central parity within a month; and the authors use this mean-reversion to generate useful observable expectations of the exchange rate's depreciation. They then subtract this estimate of the expected rate of depreciation from the interest differential to estimate the expected change of the central parity (i.e. the rate of devaluation). During exchange rate crises which sometimes precede realignments the effect of the interest rate differential swamps the movement in the estimated rate of devaluation. During periods of tranquillity, however, the expected rate of depreciation and the interest rate differential are of the same order of magnitude; the expected rate of devaluation fluctuates considerably and is imperfectly correlated with the interest rate differential. The expected rate of devaluation has settled down in recent years of the ERM, however, which reflects the increased credibility of the ERM parities.

Rose and Svensson use their estimates to show that allowing devaluation risk to fluctuate resolves some of the problems associated with the early target zone exchange rate models with only a single driving force. Moreover, their expected rate of devaluation predicts the actual ERM realignments rather better than the interest rate differential.

Expected and Predicted Realignments: The FF/DM Exchange Rate During the EMS
Andrew K Rose and Lars E O Svensson

Discussion Paper No. 552, July 1991 (IM)