W OULD a movement from fixed to flexible exchange rates increase or decrease the stability of economic activity? This paper presents calculations of how increased responsiveness of exchange rates to balance-of-payments pressures would have affected macro-economic stability for most of the world's developed nations in the recent past. It also tries to assess how openness affects the optimum responsiveness of the exchange rate to these pressures from the standpoint of macro-economic stability, an important issue in the debate over what constitutes an optimum currency area.' Our point of departure is the model which J. L. Stein (1963) built over a decade ago in order to relate macro-economic stability to the exchange-rate system. He argued that a balanceof-payments deficit under fixed exchange rates would cause currency depreciation if the exchange rate were flexible and depreciation stimulates output. Also, noting that the reverse is true for a surplus under fixed rates, he observed that in order to foster the stability of real absorption in a Keynesian economy,2 one would want the domestic currency to depreciate (appreciate) during periods when the level of output is below (above) normal. Thus, he concluded, Keynesian economies should opt for flexible or fixed exchange rates according to whether or not fluctuations in output under fixed rates (Y) and fluctuations in the excess supply of foreign exchange under fixed rates (i.e., the balance of payments surplus) (s) tend to have the same sign. In other words, Stein's rule prescribes a flexible rate for any Keynesian economy in which cov (Y, s) > 0. However, as this paper shows, if maximization of the stability of output is the goal, a better rule prescribes a flexible rate for only those economies in which the variance of what output would be under flexible rates is less than what the variance of output would have been under fixed rates. This latter rule is superior to Stein's version because there may be only a weak correlation between what output would have been under fixed rates and the state of the foreignexchange market; when the correlation is weak, the effect of induced changes in the exchange rate will be just like that of any other random disturbance: namely, to increase the variance of output. Thus, even if the covariance is positive, if the correlation is less than one, adoption of a flexible exchange rate may still increase the variance of output.3' 4