1. The case for monetary stabilization of the real economy

At least since the work of Keynes (1936), macroeconomists have tended to be divided into two broad groups: those who favor active policy intervention to attempt to stabilize the real economy on a more or less continuous basis; and those who generally oppose such intervention except perhaps in limited instances. In general, the more interventionist group may be a little more tolerant of a higher average rate of inflation or, put slightly differently, may be a little more anxious to avoid the short-term costs of reducing inflation. The noninterventionist group, on the other hand, may generally be somewhat more concerned about the economic costs of a higher average inflation rate. The main difference between the two groups concerns, however, the role and usefulness of monetary policy as counter-cyclical policy tool.

In the most recent phase of this debate the groups have also tended to be separated by technical issues of economic modeling. The noninterventionists have focused to a considerable extent on “real business cycles,” that is on viewing business cycles as the natural result of real disturbances to tastes and technology affecting a competitive economic system with flexible prices and wages. From the perspective of equilibrium-business-cycle models, there is little that monetary policy can do to affect the behavior of the real economy; and, accordingly, there is little or no useful role for interventionist monetary policy to attempt to stabilize the real economy. ^12/ In contrast, the more interventionist group has focused more on “neo-Keynesian” models of business cycles that feature various market imperfections and frictions, especially sticky wages and prices, assorted liquidity phenomena or missing markets for product or labor market information. ^13/ The essential idea is that these imperfections and frictions give monetary policy some significant bite in influencing the real economy. Accordingly, many who adopt the neo-Keynesian perspective argue that there is a good deal that monetary policy can and should do to stabilize the real economy in the face of both nominal and real shocks that would otherwise cause significant and undesirable departures from medium-term macroeconomic equilibrium.

Because our purpose is to study a variety of important issues concerning the use of monetary policy for stabilization purposes that have been developed in the macroeconomics literature, in this discussion we adopt important elements of the neo-Keynesian perspective—at least to the extent that we assume that there are mechanisms through which monetary policy might have a meaningful influence on the behavior of the real economy. At the outset, however, we emphasize that the assumption of a capacity for such influence does not automatically imply that monetary policy can be used successfully for economic stabilization. Indeed, many of the more traditional monetary economists who have opposed activist monetary policies have done so not because of any doubt of the real influence of monetary policy, but rather because of considerable doubt about the ability to use this influence systematically in a constructive manner.

This critically important point can be illustrated with a variety of models that allow a meaningful role for monetary policy in influencing the real economy. Most of these models share essential common elements and end up providing broadly similar rationales for stabilization policy. The key ingredient is usually some form of (temporary) nominal price or wage rigidity that allows quantities to adjust to purely nominal disturbances. These quantity adjustments may be the best responses available for individual decision makers, but are distinctly inappropriate from the economy-wide perspective. The idea is that the best response to a nominal disturbance would be a price response, but with prices harder to move than quantities, individuals make socially-undesirable quantity responses. In these models, monetary policy clearly can play a useful role in stabilizing real output by offsetting other nominal disturbances that would otherwise cause undesirable real output fluctuations. More generally, if there are real shocks to aggregate demand that arise from nonmonetary sources, but would cause undesirable fluctuations in output, then monetary policy can act to offset (at least partially) the undesirable real effects of these shocks.

Shocks can also occur, however, on the supply side of the economy, for which the economically desirable response is an adjustment of quantities as well as of prices. If monetary policy is used to offset such shocks, the result may be a reduction, rather than an improvement, in economic welfare. Of course, if it is possible to know the nature of the shock affecting the economy and respond in a timely and appropriate manner, then interventionist monetary policy should be a force for good. However, the information requirements for this situation to prevail in practice may not usually be met. If, because of lack of adequate information, monetary policy focuses invariably on the stabilization of output, then the results are likely to be beneficial (on average) only if the shocks affecting output are primarily on the demand side. Also, as a practical matter of great importance, monetary policy appears to affect the real economy only with a considerable and somewhat unpredictable lag. This significantly complicates and limits the potential usefulness of monetary policy in attempting to correct even demand-side disturbances of limited magnitude and duration.

With this general background, it is useful now to turn to two specific developments that have helped shape monetary-policy literature in recent years.

2. The importance of expectations

It has long been recognized that expectations about future economic developments exert an important influence on current economic behavior.

With the shift away from nominal anchors that provided long-run stability of the price level, to nominal stability measured in terms of positive long-run inflation rates, the issue of inflationary expectations gained increasing importance in discussions of the effects of monetary policy. Through the 1960s, much of the analysis of monetary policy and its role in macroeconomic stabilization used a variety of econometric models that generally assumed some form of “adaptive expectations.” In this approach, it was assumed that expectations about future inflation (and about some other relevant variables) were formed by taking backward-looking weighted average of past experience and adjusting expectations partially in response to deviations of new data from this backward-looking weighted average. Such a backward-looking approach to the modelling of expectations was appropriate during times of policy stability, when the evolution of variables like the inflation rate was reasonably well described as an evolutionary and adaptive process.

During the 1970s, however, it came to be recognized that this backward-looking approach to expectations formation could prove seriously misleading, especially in evaluating different possible regimes for the conduct of monetary policy for purposes of macroeconomic stabilization. Specifically, if expectations of inflation were inevitably backward-looking, then it appeared, from many standard macroeconomic models, that a policy of continually accelerating the rate of money creation and hence the rate of price inflation could induce a permanent increase in the level of output above its normal equilibrium level. This was the so-called accelerationist controversy, which was a major focus of Milton Friedman’s (1968) Presidential Address to the American Economic Association.

In that address Friedman argued in favor of the idea that there was no permanent trade-off between higher inflation and lower unemployment and that any effort to continually accelerate the inflation rate to exploit a possible short-run trade-off would be defeated by the forward-looking adjustment of expectations to take account of such acceleration. An essential idea that was introduced in Friedman’s thinking and in other macroeconomic theorizing at that time was the “expectations-augmented Phillips curve,” which we represent formally as:

where y is the level of output, π is the inflation rate, π^x is the anticipated rate of inflation and u is a residual that captures all aspects of aggregate supply not captured elsewhere. ^14/

The accelerationist controversy may be characterized in terms of modeling π^x and setting γ. In Friedman’s original work, γ = 1 and expectations are forward looking; implying that there is no way to induce, on average over the longer-term, a divergence between the actual and the expected inflation rate. If inflationary expectations were necessarily adaptive and backward looking or if γ < 1, however, then by pushing the inflation rate to progressively higher levels, it should be possible to induce a permanent divergence between actual and expected inflation. This would imply the possibility that accelerating inflation could produce a long-term output gain.

Solow (1969) and Gordon (1969), assuming adaptive expectations, produced econometric estimates of γ below unity and thus it appeared initially that steady inflation could result in a long-term output gain. Later work by Sargent (1971), who assumed rational expectations, found γ to be insignificantly different from one. Operationally, rational expectations means that the individuals being modeled form their anticipations as mathematical expectations based on information concerning private and government behavior in the model at hand.

Introduction of the assumption of rational expectations into many standard macroeconomic models turned out to have dramatic implications. Specifically, returning to equation (1), replace π^x by Eπ, where E is the mathematical expectation operator. Now, since γ = 1, the term involving current magnitudes affecting y involves only π = Eπ. With expectations formed rationally, this difference must correspond exclusively to the unexpected part of (or surprise in) the inflation rate. This implies that monetary policy can influence output only by creating a surprise in the inflation rate. As a minimal requirement of rationality, given a well-understood regime for the conduct of monetary policy, such surprises should average out to be zero.

Moreover, the appearance of this term clarifies the possible role for monetary policy. In particular, monetary policy is effective in influencing output only to the extent that such policy can influence inflation surprises. Policy can stabilize the real economy, therefore, only if policymakers can engineer appropriate private-sector inflation surprises. Government monetary policy actions that are understood and fully expected by the private sector are, by definition, not a surprise in relation to the information used to form Eπ and therefore do not influence output. This is the famous policy ineffectiveness result first proposed by Sargent and Wallace (1975).

This result, however, does not mean that monetary policy can have no systematic effect on the behavior of output; it just means that to have an effect the government must have an information advantage. ^15/ In particular, it is consistent with the Sargent and Wallace result that policies based on current information that was unavailable to private decision makers when they formed their expectations can influence the amplitude of business-cycle fluctuations. ^16/

In our consideration below of various nominal-anchor policies we will adopt rational expectations and we will study models that are consistent with the Sargent and Wallace result. The policies that we study influence the volatility of shocks to output due to a government information advantage. ^17/

3. Policy rules

How do we go about modeling and evaluating hypothetical economic policy? The type of formal analysis of economic policy that is currently popular was introduced to macro-analysis by Poole (1970) who studied how to set monetary policy optimally in a simple IS-LM model. At the time Poole wrote, the IS-LM structure was the standard for studying output and interest rate outcomes of particular policy actions. Poole’s innovation was to shift attention from particular outcomes of policy actions to the properties of the distributions of the variables of interest to policymakers, such as mean and variance of inflation, output or unemployment. In particular, Poole studied how the joint distribution of output and the nominal interest rate were influenced by the stance of monetary policy. In Poole’s work, a policy was not a particular movement of a policy-controlled variable it was a systematic rule for setting that variable. ^18/

Poole’s methods have become popular because they allow researchers to model policy from outside the actual policymaking environment. In particular, these methods give policymakers a specific set of rules for policy and then allow the researcher to study how well, on average, a model economy adopting these rules would perform. While Poole’s work established a role and a method for the study of monetary rules it did not build a case for why actual policy should be conducted in accord with predetermined rules rather than on a discretionary basis in response to actually realized events and its to this problem that we turn next.

Is it better for policymakers to respond to events according to a preset list of instructions as in Poole (1970) or should they maintain the freedom to react to events as they are realized? There is an extensive literature on this issue, which demonstrates, in theory, the potential gains to an economy that come from having the monetary authority adopt a set of policy rules rather than act at their own discretion to stabilize the economy.

The basic ideas in the rules-versus-discretion literature were used first in the macroeconomic setting by Kydland and Prescott (1977) and Calvo (1978) to help understand inflation in post-WWII industrial countries. The ideas were subsequently popularized by Barro and Gordon (1983a, b).

Crucial to the results in these models is the notion that distortions in the economy may lead private and public goals to diverge and thus the public finance and political economy considerations we mentioned in Section II may need to be supplemented. In the models, the government prefers a higher level of real activity than does the private sector. This may result, for example, if the private sector is distorted by an income tax to over-consume leisure activities. In the models, the real distortion is immutable and the only method available to the monetary authority to induce additional real activity is to surprise the private sector with unexpectedly high inflation, as in equation (1). Even though the government also dislikes inflation, a large enough divergence in basic goals will lead the government to tolerate some inflation in order to induce additional real activity. Understanding the monetary authority’s motives and options, the private sector is aware that the government will attempt a surprise inflation unless the inflation rate is driven so high that it no longer is worthwhile for the government to attempt such a surprise.

If the private sector and the government could get together and cooperate they could arrange a noninflationary equilibrium that would improve social welfare. But the way the models are setup the two sides cannot get together and the only way the private sector can be confident of their expectations of government behavior is when the inflation rate is driven so high, on average, that the government’s valuation of the activity gains from an inflation surprise exactly balance the costs associated with the additional inflation. The equilibrium that results, therefore, is an inflationary one with the inflation perfectly anticipated and producing no activity gains.

In this literature the noncooperative interaction that develops between the private and public sectors is known as a game and this game’s equilibrium results in the government attempting to trick the private sector into producing more, but in the end only damaging the economy with excessive inflation. One answer to this type of problem is for the private sector to closely proscribe government actions so that the government is effectively prohibited from playing policy games. What is often recommended in this literature is that the government behave according to simple easy-to-monitor rules and in the next subsection we will review some simulations of such rules.

We recognize that our attention to fixed rules is very far from the immediately-relevant type of question that policymakers must ask themselves on a day-to-day basis, which is essentially: “What should be done today?” At first we will proceed as if the answer is always “Follow the rule.” After the next subsection we will expand our discussion to recognize the possibility that rules may be abandoned temporarily during exceptional circumstances.

4. Specific policy rules evaluated in multi-country models

Our aim here is to try to find out how many of the major economies would behave if their monetary authorities followed some specific moving nominal anchor policies. Of course, in economics there is never really an opportunity to conduct controlled experiments on actual economies in order to sort out the performance of policy alternatives. Instead economists construct econometric models from historical data, build the policy alternatives into the models and then hope that the altered models behave as history would have behaved had the policies actually been followed. Such experiments are called policy simulations. If economists had great confidence in their models then they might also have confidence in the results. Unfortunately such is not the case.

In fact there are a large number of carefully-constructed models built by different groups, using the same historical data that are really quite different from each other. While each modeling group may have confidence in its own work, it is likely that each group is equally confident that other models are lacking in various ways. In view of such doubts, rather than report the results of policy simulations in one particular economic model we use the approach recommended by McCallum (1988), which is to report the results of policy simulations across a wide range of state-of-the art models. ^19/ The work that we draw on here is that reported by Bryant, Hooper and Mann (1993), BHM, where nine multi-country econometric models were asked to simulate the effects of adopting a set of monetary policy rules. ^20/ The rules studied are: (1) a money targeting rule; (2) a nominal income targeting rule; (2a) a combination inflation and real output targeting rule; and (3) an exchange rate targeting rule. The precise rules that the various modelling groups were asked to study are:

where

• RS = the short-term nominal interest rate measured in percentage points per year divided by 100

• M = the monetary base (or some other narrow monetary aggregate)

• P = the price level (GNP or GDP deflator)

• Y = real GNP or GDP

• it = the rate of inflation in the price level P (expressed as percent per year)

• S = the nominal exchange rate, measured in U.S. dollars per unit of local currency

and where an asterisk superscript (*) denotes a target (baseline) value. ^21/ The baselines are country-specific and are discussed in BHM.

The specific rules that are studied were chosen because of their relevance to actually-used or proposed rules. The money targeting rule (Regime A) is similar in form to the type of monetary targeting reported in Isard and Rojas-Suarez (1986), which studies monetary targeting in major industrial countries in the period 1975–85. The nominal income targeting rule (Regime 2A) is similar to the types of rules studied by McCallum (1986) and by Hall and Mankiw (1993) for the United States. Regime 2B, which targets a combination of inflation and the level of real output is pursued as a possibly relevant variant of Regime 2A. In regime 2B equal weight is given to deviations of inflation from its baseline and deviations of output-growth from its baseline. Regime 3 is designed to smooth exchange rate fluctuations and mimics exchange rate smoothing by major countries. ^22/

The simulations work as follows. First the models are estimated country-by-country and behavioral parameters are taken to be invariant to changes in policy. The properties of the residuals from the estimated equations set the stochastic structure (the variance-covariance matrix) for shocks in the simulations. Second, the regime-specific equations for short-term interest rates replace previously relevant interest rate equations for all countries. In other words, if one country adopts, for example, a nominal income anchor, all countries adopt similar anchors. Third, the model produces a simulated history built around the new policy by drawing new disturbances whose stochastic properties are similar to the historical ones. ^23/

The models are simulated assuming the individuals being modeled have rational expectations and understand government policy sufficiently well so that expectations are correct on average. This implies that what is being simulated is a sort of steady position that the economy would come to if the relevant policy had been in place for some time. No attempt is being made to simulate the transition from one policy to another.

The actual simulations changed the suggested rules slightly to fit the structure of the models being used. Table 1, which is taken directly from BHM, reports one aspect of the results of the simulations. The table reports the regime, for each country and each model that provides the best performance of the indicator of interest. The left-hand column of the table lists countries and characteristics of simulated series for the countries in question. The first entry in the column, for example, is for the United States and it is “Loss: Real GNP alone.” What this means is that the model characteristic that is being compared across regimes is the minimization of average surprises in real U.S. GNP. ^24/ In column 2 of the table, which gives the results for the GEM model it is reported that Regime 2BX minimizes this measure of real-output variance. The “X” attached to regime 2B in 2BX means that simulations of the GEM model changed rule 2B to be compatible with the GEM model. BHM report the model-specific rule changes.

Table 1.

Identification of the Regime Minimizing Three Simple Illustrative Loss Functions ^1/

Source: B-1 tables in annex B.n.a.: Results not available.

^1/When the RMSDs for two regimes differ by 1 percent or less, the table identifies the regimes as tied. When an entry in a cell is in boldface type, that regime is associated with a reduction in the value of the loss function by 10 percent or more relative to the second-best type of regime. See the text for further explanation.

Table 1.

Identification of the Regime Minimizing Three Simple Illustrative Loss Functions ^1/

	Model
	Country and Loss Function	GEM	Intermod-A	Intermod-C	Liverpool	MPS	MSG	MULTIMOD	MX3	Taylor
United States
	Loss: Real GNP alone	2BX	2B	2B	3X	2BX	2BX	2BX	2B	2AX
	Loss: Inflation alone	2BX	2A	2AX	3X	2AX	2AX	2BX	2AX	2BX
	Loss: Real GNP + inflation	2AX/2BX	2B	2AX	3X	2BX	2BX	2BX	2B	2AX
Japan
	Loss: Real GNP alone	2BX	2B	2B	2A	n.a.	2AX	1	2B	2BX
	Loss: Inflation alone	2AX	2A	2AX	1X	n.a.	1X	1X/2BX	2AX	3X
	Loss: Real GNP + inflation	2BX	2B	1/2AX	1X/2A	n.a.	2AX	1	2AX	2BX
Germany
	Loss: Real GNP alone	2BX	2B	2B	2AX	n.a.	2BX	2AX	2B	2AX
	Loss: Inflation alone	2AX	2B	2AX	1X	n,a.	ZA	2BX	2AX	1X/2BX
	Loss: Real GNP + inflation	2AX	2B	2AX	1X	n.a.	2BX	2AX	2B	2BX
United Kingdom
	Loss: Real GNP alone	n.a.	2B	2B	2AX	n.a.	n.a.	3X	n.a.	2AX
	Loss: Inflation alone	n.a.	2A	2AX	2A	n.a.	n.a.	2BX	n.a.	3X
	Loss: Real GNP + inflation	n.a.	2B	2AX	2A	n.a.	n.a.	3X	n.a.	2AX
Canada
	Loss: Real GNP alone	n.a.	2B	2B	3X	n.a.	n.a.	1X/2AX	n.a.	2BX
	Loss: Inflation alone	n.a.	ZA/2B	2AX	2A	n.a.	n.a.	2BX	n.a.	2AX/2BX
	Loss: Real GNP + inflation	n.a.	2B	2AX	2A	n.a.	n.a.	1X/2AX	n.a.	2BX

Source: B-1 tables in annex B.n.a.: Results not available.

^1/When the RMSDs for two regimes differ by 1 percent or less, the table identifies the regimes as tied. When an entry in a cell is in boldface type, that regime is associated with a reduction in the value of the loss function by 10 percent or more relative to the second-best type of regime. See the text for further explanation.

View Table

Table 1.

Identification of the Regime Minimizing Three Simple Illustrative Loss Functions ^1/

	Model
	Country and Loss Function	GEM	Intermod-A	Intermod-C	Liverpool	MPS	MSG	MULTIMOD	MX3	Taylor
United States
	Loss: Real GNP alone	2BX	2B	2B	3X	2BX	2BX	2BX	2B	2AX
	Loss: Inflation alone	2BX	2A	2AX	3X	2AX	2AX	2BX	2AX	2BX
	Loss: Real GNP + inflation	2AX/2BX	2B	2AX	3X	2BX	2BX	2BX	2B	2AX
Japan
	Loss: Real GNP alone	2BX	2B	2B	2A	n.a.	2AX	1	2B	2BX
	Loss: Inflation alone	2AX	2A	2AX	1X	n.a.	1X	1X/2BX	2AX	3X
	Loss: Real GNP + inflation	2BX	2B	1/2AX	1X/2A	n.a.	2AX	1	2AX	2BX
Germany
	Loss: Real GNP alone	2BX	2B	2B	2AX	n.a.	2BX	2AX	2B	2AX
	Loss: Inflation alone	2AX	2B	2AX	1X	n,a.	ZA	2BX	2AX	1X/2BX
	Loss: Real GNP + inflation	2AX	2B	2AX	1X	n.a.	2BX	2AX	2B	2BX
United Kingdom
	Loss: Real GNP alone	n.a.	2B	2B	2AX	n.a.	n.a.	3X	n.a.	2AX
	Loss: Inflation alone	n.a.	2A	2AX	2A	n.a.	n.a.	2BX	n.a.	3X
	Loss: Real GNP + inflation	n.a.	2B	2AX	2A	n.a.	n.a.	3X	n.a.	2AX
Canada
	Loss: Real GNP alone	n.a.	2B	2B	3X	n.a.	n.a.	1X/2AX	n.a.	2BX
	Loss: Inflation alone	n.a.	ZA/2B	2AX	2A	n.a.	n.a.	2BX	n.a.	2AX/2BX
	Loss: Real GNP + inflation	n.a.	2B	2AX	2A	n.a.	n.a.	1X/2AX	n.a.	2BX

Source: B-1 tables in annex B.n.a.: Results not available.

^1/When the RMSDs for two regimes differ by 1 percent or less, the table identifies the regimes as tied. When an entry in a cell is in boldface type, that regime is associated with a reduction in the value of the loss function by 10 percent or more relative to the second-best type of regime. See the text for further explanation.

Rules 2A and 2B, the nominal income targeting rule and the variant on the nominal income rule, which targets inflation plus real output out-perform money targeting and exchange rate targeting by a wide margin, with the two nominal income variants producing a better loss-function value around 80 percent of the time. From the reported results there is little to choose from between the nominal-income-based rules, 2A and 2B, which perform best about an equal number of times.

The results of this section are clear. In a wide variety empirically based open-economy models, monetary policy rules that are directed toward some version of nominal income stabilization perform better in terms of output stability and inflation stability in realistic simulations than do rules based on stabilizing exchange rates or money supplies. While confidence in any single simulation model may be low, it is noteworthy that the models have similar results and that these results match those obtained by McCallum (1988) and Hall and Mankiw (1993) in single-economy simulations.

5. Specific monetary policy rules are temporary

The previous subsection evaluated specific rules based on the assumption that the rules would last indefinitely. In practice, price-fixing regimes and some other nominal anchors for monetary policy are temporary as demonstrated, for example, by the ending of past gold standards, the abandonment of fixed Bretton Woods parities and the frequent central-parity changes in the EMU. Recent research has begun to trace out the implications of such impermanence for the behavior of private individuals and governments. Most of this work starts from the basic idea that when a government targets a specific market price as a nominal anchor for monetary policy, stabilizing that nominal price is not the government’s primary policy goal. If a government adopts a fixed exchange rate, for example, then it is understood that if the fixed exchange rate gets in the way of other, more important, policy goals then the fixed rate will be abandoned. This literature also shows that it may be desirable, even in light of the rules-versus-discretion issues studied above, for the government to abandon a policy rule in some situations.

This literature’s application to price-fixing regimes began with the work of Salant and Henderson (1978) and Krugman (1979) who showed how the collapse of a government price-fixing regime can be a natural economic event based on economic fundamentals and devoid of expected profit opportunities. Krugman’s example involved a crisis in a fixed exchange rate regime, which is developed as a privately-driven transition from a fixed exchange rate regime to a flexible one. In Krugman’s work the timing of a crisis for a fixed exchange rate rule and its collapse is determined by speculators pursuing profit opportunities. It is the competition among speculators that determines the timing of a crisis and removes expected profit opportunities from the crisis. In this sort of model, as speculators see the crisis approaching they attempt to borrow large quantities of the currency whose devaluation is suspected, convert the borrowed funds into foreign exchange and then repay in terms of a devalued currency. Even without a government interest rate defense, short-term market interest rates can be driven to levels unheard of in normal times. Krugman-style modeling of an exchange rate crisis is based on speculators having unlimited access to short-term credit markets, which describes well the available markets for industrial countries but may overstate market access in some controlled developing-country financial markets.

Following Krugman’s work, it came to be recognized that a crisis precipitated by economic fundamentals was not the only crisis that could be explained in the context of the Krugman-style model. It was shown that if the crisis caused a fundamental change in government policy then the crisis could be triggered by a speculative whim and validated by the very policy change it triggered. In other words, instead of a fundamentals-driven event, the crisis and the attack could be just one of a multiciplicity of possible outcomes with the exchange rate regime surviving, or not, as dictated by speculative fancy. ^25/

Models of crises in government price-fixing rules are all based on the idea that price fixing is not the government’s primary objective and that the resources the government is willing to commit to price fixing are limited. If a government’s commitment to a nominal price of any asset or commodity in units of its own currency is absolute then it is unquestioned that the price target can be preserved and speculation against the fixed price would be futile. If the commitment is limited, however, then well-financed speculators can provoke a crisis in the system and have the incentive to do so when economic fundamentals are inconsistent with the policy or when the crisis itself provokes a change in the basic stance of government policy.

The implications of temporariness of other policies is explored in models of policy rules with an escape clause, where it is recognized that inflexible monetary-policy rules, such as those studied in the last section, do not have the ring-of-truth in the sense that it is almost unimaginable that central bankers would consult a mathematical rule in a time of crisis in the financial system. ^26/ The point developed here is that following a monetary rule may not be ideal all of the time, even accounting for the rules-versus-discretion issues as outlined above. In particular, while following a rule may be ideal, on average, during “normal” times, society may benefit from having the policymaker abandon the rule in exceptional circumstances and confront events as warranted by the situation at hand. ^27/ The lesson from this work is that following a rule can be valuable even if the rule must be abandoned temporarily during extraordinary situations.

The formal study of policy temporariness is in its infancy but two of its lessons for policy design are clear already. First, less than complete government commitment to a price-fixing target can bring powerful speculative forces to bear that may provoke a crisis and end or modify the price-fixing policy. The real-world relevance of this theoretical point is illustrated by the European Currency Crisis of 1992 and the changes in European Monetary Union’s exchange rate arrangements provoked by that crisis. ^28/ Second, while it is hard to imagine actual central bankers seriously consulting a mathematical rule before setting policy, it is easier to think of a monetary policy rule as being modeling shorthand for policymakers systematically checking on a set of widely available indicators of the state of the economy with different rules then being different weighting schemes for actions responding to the various indicators. With this broad interpretation of a rule, the rules-versus-discretion literature simply recommends keeping the weighting schemes constant and making them public during normal times responding differently only during times of crisis.