We are very grateful to Lohmann for identifying an inconsistency in Flood and Isard (1989) and pointing to an important error in our reasoning. Lohmann’s discussion of our methodology, however, provides only a partial description of the premises that motivate our analysis. As this reply will argue, there are more ways than one to correct the analysis. The procedure that we find most attractive for correcting the inconsistency is different from the procedure suggested by Lohmann. In any case, we see the problem as a matter of getting the argument straight rather than a basic difficulty with the type of monetary policy strategy we suggest.

Our main objectives in Flood and Isard (1989) were: (1) to shift the discussion of monetary rules and discretion from the unrealistic setting in which the structure of the economy is assumed to be known completely to the realistic case in which knowledge about the structure of the economy and the nature of economic disturbances is incomplete; and (2) to make normative suggestions for inducing central banks to combine rules and discretion in better ways. We refer to such combined strategies as rules with escape options.¹

We start from the premise that a fully state-contingent rule for monetary policy is simply not a relevant possibility in a world in which knowledge about the macroeconomic structure and the nature of disturbances is incomplete. Since partially state-contingent rules and discretion cannot be unambiguously ranked, it is natural to investigate strategies that optimally combine a partially state-contingent rule and discretion. All of the rules that monetary authorities have employed, or that economists have proposed for monetary policy, amount to partially state-contingent or non-state-contingent rules. The authorities may sometimes have incentives to deviate from such rules, but the typical methodology for evaluating these rules, and for deriving optimal parameter values, is based on the assumption that market participants expect the authorities to adhere to the rules indefinitely. It thus becomes interesting to explore conditions under which the parameters obtained through traditional methodology are also optimal when the policy strategy is recognized to be a rule with an option to exercise occasional discretion.

An important related issue is how to design institutional arrangements to ensure that the central bank uses its escape option appropriately. Because of well-known problems with discretionary policymaking, society wants the central bank to exercise discretion only when there are relatively large payoffs in terms of a social welfare function. For purposes of illustration, Flood and Isard (1989) considered a simple framework with a single stochastic term in which we hypothesized that society wants the central bank to exercise discretion only during periods in which the absolute value of the stochastic shock exceeds some threshold magnitude. We further conjectured that society can induce the central bank to behave in the desired way by imposing a cost whenever the central bank chooses to exercise discretion. We suggested that the level of the cost can be calibrated to make discretion attractive only when the size of the shock exceeds the prespecified threshold.

As Lohmann (1990) emphasizes, we overlooked the fact that the quadratic social loss function we specified, following the mainstream literature on rules versus discretion, is not symmetric in the size of the money-demand shock, the stochastic term in the particular example on which we chose to focus. Accordingly, as Lohmann demonstrates, a rule calling for zero base-money growth, when combined with an escape option calling for symmetric responses to positive and negative money-demand shocks, cannot be optimal. Moreover, since positive money-demand shocks would generate different social losses under the rule than negative money-demand shocks of equal magnitude, a uniform penalty (cost) for exercising discretion would not induce the central bank to exercise its escape option symmetrically in response to positive and negative shocks.

The procedure that we find most attractive for correcting this inconsistency is different from the procedure suggested by Lohmann. Under the assumption that the distribution of money-demand shocks is symmetric around zero, a rule calling for zero base-money growth remains attractive, since this is the optimal rule in the limiting case in which the variance of the money-demand shock shrinks to zero, consistent with the “traditional” assumption that market participants expect the authorities to adhere to the rule indefinitely. Accordingly, the attractive policy is one in which the rule calls for zero base-money growth, while the escape option specifies two separate thresholds with different absolute values for exercising discretion in response to positive and negative shocks. Consistently, the costs imposed to induce the central bank to respect the thresholds must also make a distinction between the responses to positive and negative shocks. Such a distinction would be straightforward to apply in practice, since it would suffice to make the level of the cost depend simply on the range within which the authorities set the rate of base-money growth.²