This paper reformulates the monetary approach after ascertaining that a money demand function with a partial adjustment mechanism had more empirical support than a money demand function which assumed instantaneous stock adjustment. The resulting exchange rate equation is estimated by two rational expectations techniques. The parameter estimates are reasonable, and are robust to the estimation technique used, to different specifications of the driving processes, and to changes in the estimation period. The structural model outperforms the random walk model and its own unconstrained equivalent, a pure time series equation, in outof-sample forecasts.