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...