THEORY OF RULES VERSUS DISCRETION tive expenses for changing prices and the transaction costs associated with economising on cash holdings. The positive analysis of monetary policy implies that a downward shift in the costs of inflation will lead to more inflation If people think that inflation is not a serious problem, then the economy will end p with a lot of inflation! The analysis implies also that each flicker in the benefits from inflation surprises or in the costs of inflation will be reflected in variations in inflation. In contrast to an environment in which the government stabilises prices, there will be substantial random fluctuations of inflation and monetary growth. Further, the variances of prices and money will be larger the greater the random fluctuations in the variables that influence the benefits from inflation shocks For example, if there are frequent supply shocks(which alter the natural rate of output), then inflation and monetary growth will be volatile III MONETARY RULES The results under discretion contrast with those under rules, which are regimes where the policymaker can and does make commitments about future monetary growth and inflation. Under discretion, the equilibrium involved high inflation, but no tendency toward surprisingly high inflation. Hence, the economy suffered the costs from high inflation, but secured none of the benefits from inflation surprises. The policymaker can improve on this outcome if he can commit himself ex ante to low inflation. If this commitment is credible which means that some mechanism prevents violations ex post-then people also anticipate low inflation. Therefore, the equilibrium would exhibit low and stable inflation, with the same average amount of surprise inflation(zero)as before. These results support a form of'constant-growth-rate rule, although applied to prices rather than to the quantity of money, per se There is a tension in this type of rules equilibrium because the policymaker may retain the capacity to produce large social gains at any point in time b cheating-that is, by generating surprisingly high inflation. Then there may be a temporary economic boom or at least a substantial amount of government revenue obtained via a distortion-free tax. But, if such cheating were feasible and desirable, then people would understand the situation beforehand. In this case the low-inflation equilibrium would be untenable. (Sometimes people say that this equilibrium is ' time inconsistent, although it is actually not an equilibrium at all. )Rather, there would be a high-inflation, discretionary equilibrium, as described earlier. That is why the enforcement power behind the low-inflation rule is crucial. There must be a mechanism for binding the policymaker's hands in advance, so that(surprisingly) high inflation cannot be generated later, even if such a choice looks good to everyone ex post. Note that the rationale for this " binding of hands'applies even though(or actually lly if)the policymaker is well-meaning. This type of commitment is
THE ECONOMIC JOURNAL necessary in order for low inflation to be incentive-compatible and hence Although the low-inflation, rules equilibrium is superior to the high-inflation discretionary equilibrium, the e rules Librium is still benefits from inflation surprises-for example, from lower unemployment or external effects that have not been eliminated. It is the desire to approach the first-best solution via inflation surprises that threatens the viability of the low inflation equilibrium. The pursuit of the first-best tends to push the econom away from the second best of a rule with low inflation, and toward the third best of discretionary policy with high inflation. Again, this perspective highlights the importance of the enforcement power that makes a rule sustainable IV CONTINGENT RULES More generally, the optimal rule may set money or prices contingent on exogenous events, rather than being non-contingent. In some models, such as those where the monetary authority has superior information about the economy, a contingent reaction to business-cycle variables may help to smooth It business fluctuations. However the direct communication of the government's information may be a substitute for the feedback response of Another example of contingent response is the association of wars with high growth rates of money and prices. High wartime inflation constitutes rprisingly high inflation from the standpoint of earlier times at which low inflation during peacetime. This type of contingent rule may be desirable because it generates lots of easy revenue via the capital levy from unexpected inflation during emergencies. In particular, it is possible to hold down distortions from the income tax at the most important times, such as wars. 1 Although the necessary accompaniment is a loss of revenue during the non mergencies effect of this contingent policy is likely to be beneficial Under the gold standard, governments did in fact tend to go off gold during wars, as in the case of Britain during the Napoleonic period and around World War I, and for the United States during the Civil War. This procedure enables a government to pursue the type of contingent policy for inflation that I sketched above. In this sense a movement off gold during wars is not necessarily a violation of the rules. However the subsequent return to gold at the previous lessens this incentive, but does not el able to trigger the distortion-fre Svensson(1984). These models feature public debt with a contingent real payo, - posited by Persson and during peacetime and low in wartime If government bonds are nominally denominated and no sh contingent(for reasons that escape me), the contingent behaviour of inflation achieves the same end