Modelling Probabilities of Devaluations By GABRIELA MUNDACA University of Oslo Final version received 20 January 2003. I show why, when the realized rates of depreciation within the exchange rate band are regressed on a given information set and conditioned on (ex post) actual no realignment (a ` la drift adjustment), a ‘peso problem’ is still encountered. The reason is that the frequency of realignments in the data need not be the same as the frequency of the (even small) subjective probabilities that a realignment may take place. I suggest an alternative approach to solve the peso problem and provide consistent estimates. My estimates of the expected realignment rates are greater than the ones obtained using the drift adjustment method. INTRODUCTION In the period 1979–93, countries in the ERM and the Nordic countries essentially had exchange rate systems with fixed but adjustable parities as their monetary policy regime. 1 It can reasonably be assumed that market partici- pants assigned subjective probabilities to possible realignments. Studies of monetary regimes, even past ones, will always be important for assessing the degree of credibility of these regimes. This can enable policy-makers and the public to establish certain criteria for future monetary arrangements. By now, most countries have adopted other monetary policy regimes (e.g. inflation targeting), but studies measuring realignment/devaluation expectations of past regimes are still of interest and are still being conducted. For example, Hallwood et al. (2000) estimate the dollar–sterling exchange rate expectations for the years 1890–1908. One of the purposes of this paper is to suggest a methodology for estimating the expected exchange rate depreciation for exchange-rate-band regimes. I draw attention to a possible drawback of the drift adjustment method, which is widely used to obtain estimates of realignment/devaluation expectations and depreciation within the band in exchange-rate-band regimes. I show that this method cannot yield consistent estimates, 2 and suggest an alternative procedure that provides consistent estimates. My own estimates of the expected realignment (devaluation) rates prove to be always greater than the ones obtained using the drift adjustment method. The ‘drift adjustment’ method suggested by Bertola and Svensson (1993) has been used widely, for example by Chen and Giovannini (1993), Lindberg et al. (1993), Svensson (1993), Rose and Svensson (1991, 1994) and Holden and Vikøren (1994). Svensson (1992) has suggested that all important regimes in the international monetary system in which exchange rate fixity is the modus operandi can be analysed using a target zone framework. 3 This argument was then adopted by Hallwood et al. (2000), who used the drift adjustment method to estimate realignment or devaluation expectations of exchange rates during the classical Gold Standard period in the context of the target zone literature. Economica (2004) 71, 13–37 r The London School of Economics and Political Science 2004