A good deal of attention has rece tly been devoted by monetary conomis s to the socalled instrument problem.l Briefly, this problem arises s a result of the inv rse lationship that exists etween the money supply nd the rate of in erest. Because of thi ela ionffiip, th monetary authoriti s are u able to use b th th money upply a d the rate of int rest simul aneously a indep ndent policy instruments (or more precisely as roxima targ ts), but instead are forced t choos be w n them. T e soluti to the instrument problem de r ines h w th s c oice sh uld be ade. A pointed ut by previous authors (see for exampl Poole [10]), th problem re lly becom s interest g only in a st cha tic nvironment. Under p rely ter inistic c nditions, the ch ice of monetary instrume t does ot matter-t at is, as long as policy makers do n t ascribe a y costs to the adjustment of heir instru ents. This atter ssumption, which und rlie almost all of th exis i g work his area (al hough often it is not stated xplicitly), is in fact cr cial. If di ferent positive co ts could b as igned t he vario s mon tary i struments, then ev tu der certai y a wel -defin d i strumen p oblem would exist. With one exception, the existing treatments of the problem deal with the case where all the parameters of the system are known and deterministic and the only stochastic influences are introduced through additive disturbances. These solutions