Journal of Monetary Economics 12 (1983) 101-121.




University of Chicago, Chicago, IL 606J7, I-:SA
NBER. Cambridge, MA 02138. USA

Unirvrsity vf Rochester. Rochester, II’ Y 14627. L‘S,4

In a discretionary regime the monetary authority can print more money and create more inflation than people expect. But, although these inflation surprises can have some benefits, they cannot arise systematically in equilibrium when people understand the policymaker’s incentives and form their expeziations accordingly. Because the policymaker has the power to create inflation shocks ex pest, the equilibrium growth rates of money and prices turn out to be higher than otherwise. Ther,zfore, enforced commitments (rules) for monetary behavior can improve matters. Given the repeated interaction between the policymaker and the private agents, it is possible that reputational forces can substitute for formal rules. Here, we develop an example of a reputational equilibrium where the outcomes turn out to be weighted averages of those from discretion and those from the ideal rule. In particular, the rates of inflation and monetary growth look more like. those under discretion when the discount rate is high.

1. Introduction In a discretionary regime the monetary authority can print more money and create more inflation than people expect. The benefits from this surprise inflation may include expansions of economic activity and reductions in the real value of the government’s nominal liabilities. However, because people understand the policymaker’s incentives, these types of surprises - and their resulting benefits - cannot arise systematically in equihbrium. People adjust their inflationary expectations in order to eliminate a consistent pattern of surprises. In this case the potential for creating inflation shocks, ex post, means that, in equilibrium, the alverage rates of inflation and ‘monetary growth -- and the corresponding costs of inflation - will be higher than otherwise. Enforced commitment!; on monetary behavior, as embodied in monetary or pril:e rules, eliminate the potential for ex post surprises. Therefore, the equilibrium rates of inflation and monetary growth can be
*We have benelitetl from discussion at the conference and from seminars at Chicago. Northwestern and lov’a. We are particularly grateful for comments from Gary Fethke, Roger Mverson, Jose Scheinkman. and John Taylor. Part of this research IS supported by the National Science Foundation. 0304-3923/83/$3.00 0 1983, Elsevier Science Publishers B.V. (North-Holland)


R.J. Bnrro and LIB. Gordon,



arrd reputation




lowered by shifts from monetary institutions that allow discretion to ones that enforce rules. When monetary rules are in place, the policymaker has the temptation each period to ‘cheat’ in order to secure the benefits from inflation shocks. (Because of existing distortions in the economy, these benefits can accrue generally to private agents, rather than merely to the policymaker.) However, this tendency to cheat threatens th.e viability of the rules equilibrium and tends to move the economy toward the inferior equilibrium under discretion. Because of the repeated interactions between the policymaker and the private agents, it is possible that reputational forces can support the rule. That is, the potential loss of reputation - or credibility -- motivates the policymaker to abide by the rule. Then, the policymaker foregoes the short-term benefits from inflation shocks in order to secure the gain from low average inflation over the long term. We extend the positive theory of monetary policy from our previous paper [Barre and Gordon (1983)] to allow for reputational forces. Some monetary rules, but generally not the ideal one, can be enforced by the policymaker’s potential loss of reputation. We find th!at the resulting equilibrium looks lika a weighted average of that under discretion and that under the ideal rule. Specifically, the outcomes1 are superior to those under discretion - where no commitments are pertinent - but inferior to those under the ideal rule (which cannot be enforced in or’.r model by the potential loss of reputation). The results look more like discretion when the policymaker’s discount rate is high, but more like the ideal rule when the discount rate is low. Otherwise, we generate predictions about the behavior of monetary growth and inflation that resemble those from our previous analysis of discretionary policy. Namely, any change that raises the benefits of inflation shocks - such as a supply shock or a war - leads to a higher growth rate of money and prices.

2. l%e policymaker’s objective As in our earlier analysis, we think of the monetary authority’s objective as reflecting the preferences of the “representative’ private agent. Ultimately, we express this objective as a function of actual and expected rates of inflation. Specifically, benefits derive from positive inflation shocks (at least over some range), but costs attach to higher rates of inflation. 2.1. The benefitsfiom surprise inflation

We assume that some benefits arise when the inflation rate for period C, TI,, exceeds the anticipated amount, qp. One source of benefits - discussed in Barr0 and Gordon (1983) and ia an example from Kydland and Prescott (1977, p. 477) derives from lthe expectational Phillips curve. Here,

(M. pnvately-held comrwnent of the funded . Hence.71.R. Suppose that people held last period the real amount of government bonds. Rules. interest-bearing public debt. Other sources of benefits from surprise inflation involve governmental this channel produces an effective hmp amount of revenue of about $10 billion for each extra I”. whil:h lowers the government’s future real expenditures for interest and repa:/ment of principal.tanding public debt for the U. determines people’s holdings of real czlsh.entail distortions.rrp.B. M. if the distortions from income taxation. a:. we mean here the value that would be ground out by the private sector in the absence of monetary disturbances. . _ 1. This natural rate can shift over time because of supply shocks. . Hence. shifts in governmental tax and transfer progrzlms. the out \.S. By the natural rate. depreciates part of the real value of these bonds. B. which involves revenue from printing high-powered money. of surprise inflation. _ r. reflected in positive values for II.>mina] shocks lower the unemployment rate below the natural rate. The expectation of infliction (formed the previous period). the entire annual flow of revenue through the Federal ‘For national this purpose we should actually look at the debt. _ t)/P. _ 1 These bonds carry the nominal yield. per sr. which is satisfactory given people’s inflationary expectations over the pertinent horizon. in 1981 is around $1 trillion. lead to increases in real economic activity. the government (and the private agents) would value stimulative pohcy actions that lower the unemployment rate below its natural value. Thus. unemployment compensation. _ . Equivalently. the benefit from surprise inflation depends again on some existing externality. _ r/P. Calvo (1978) discusses the necessity of existing distortions in this type of model. for example. and the like make the average level of privately-chosen work and production too low. The policymaker values this inflationary finance if alternative methods of raising revenue such as an income tax . . Barro (1!%3) focuses on the proceeds from inflationary finance. surprise inflation is again a source of revenu: to the government./P. depreciates the real value of these holdings. Surprise inflation. The natural rate also need not be optimal.. rr.. these n. which ii ahout $1”00 billion in 19x1. Quantitatively. Because of the externalities from these distortions.M. In effect. the same argument applies to nominally-denominated. Surprise inflation. . this channel from public debt is likely to be more significant than the usually discussed mechanism. which allows the government to issue more new money in real terms. By contrast. For example. a surprise inflation of I”. IS:. R. Gordon.J. This can occur. Barro and D. R. as a replacement. the benefits from surprise inflation arise when the poiicymaker views the natural rate as excessive. The revenue incentive for surprise inflation relates to governmental liabilities that are fixed in nominal terms. In fact. discretion and reputation in monetnry p(>jir*y 103 unanticipated monetary expansions.’ Therefore.. and so on. rather than to money. demographic changes. lowers the real value of this debt by about $iO billion.

The second term. although the quantitative role of these costs is doubtful. ex where a.when people anticipate. Costs are assumed to rise. to surprise inflation. Gordon. (a/2)(rt)2.I these costs. the distortions arise . with the realized inflation rate.. economists have not presented very convincing arguments to explai .. rather than to the variance of inflation. rr. Rules.=(a/2)(71.for capital. The attractions of generating revenue from surprise inflation are clear if we view the depreciation of real cash or real bonds as an unexpected capital levy. Further. Along the way. money or bonds . rules and reputation. Of course. Although people generaIly regard inflation as very costly. Then. W The first term. ex post. and at an increasing rate. Direct costs of changirng prices fit reasonably well into the model. The policymaker’s objective involves a cost for each period. the possibility of these capital levies. bt(rr.+’ ir$ation The second major element in our model is the cost of inflation. z..B.)2-b~(TI. 2. $13 billion in 1980).J. which is given by z. But. the model predicts a lot of inflation! 3. >O. the present type of cost refers to th: actual amount cf inflation for the period. In any event the analysis has some interesting conclusions for the case where the actual amount of inflation for each period is not perceived as costly. Burro and D. for the moment. . the government can extract revenue without disincentive effects. The costs r?.llustrates the main points about discretion. is the benefit from inflation .IO4 R. Notice that our use of a quadratic form means that these costs rise at an increasing rate with the rate of inflation. is the cost of inflation. b. ex post.2. we are just listing the benefits that attach. discretion and reputation in monetary policy Reserve from the creation of high-powered money is about the same magnitude ($8 billion in 1981. which i. II. That’s why these forms of raising revenue will not end up being so desirable in a full equilibrium where people form expectations rationally. As with a tax on existing capital. surprise inflasion provides for a method of raising funds that is essentially non-distorting. -nr). ex post. which co~ld more easily be seen as costly. we indicate how the results generalize beyond this example. The setup of our example We focus our discussion on the simplest possible example. Once people havle built up the capital or held the real cash or real bonds.

we neglect serial correlation in the natural unemployment rate. is the discount rate that applies between periods r and t + 1. which enables him tc) select the rate of inflation. the coefficient b. Most of our results would not be affected if we allowed for this type of dependence. Barre artd D.. Hence.R. an increase in unexpected inflation. depends on the forecast variance of inflation. We assume that rl is generated from a stationary probability distribution. II. For the first period ahead.~ (Hence. In our example. we could modif!: the cost of inflation tl. We denote the mean and variance for q. The policym. a supply shock . Alternatively. BUI. Then. implies a distribution for the discount factor.’ Given that the benefit parameter. this element matters when we compare across regimes that hale different forecast v. n:‘. is the optimal inflation tax on cash balances.[I. >O depends on the natural unemployment rate for the prlod. % some mod 4s. discretion arrd reputation in monetary polir!.3mly with a fixed mean. h. For example. . reduces costs. of the policymaker’s choice without knowing these parameters.which raises the natural rate of unemployment . However.driances for inflation. We allow the benefit parameter. qr. n. Here. the discount rate is generated independently of the benefit parameter. the dist:ibution of r. Gordon. by CJ and di.only over some range. where ri. government expenditures. (Therefore. at. respectively.. Rules. z. etc.that is. We begin with a symmetric case where no one knows the bend3 parameter../. people form their expectations. to move around over time. the @IC)rnaker bitlucs mflation shocks . Also... We can think of these benefits as reflecting reductions in unemployment or increases in governmental revenue.) Also. the policymaker chooses the inflation rate.lpCr [Barre ~md Gordon ( 198.E. b.iker controls a monetary instrument. For example. b.n I]‘. qr = 1/( 1 +rr). ] where 4. the general nature of the results does not change if we substitute this more complicared form. we use a linear form for convenience. o:. -. we again neglect any serial dependence. depend on (r. when they act for period t. w-ithoul observing eithr:r b. b. such as Lucas ( 1973) and Barro ( 1976).J. 105 shocks.. (2) where r. ‘OilI’previous p. . the policymaker has no incentive to randomize choices of inflation in the model. 6.may increase the value of reducing unemployment through aggressive monetary policy.) The policymaker’s objkctive at date t entails minimization of the expected present value of costs. > n. or qt.. Similarly. a sharp rise in government spending increases the incentives to raise revenue via inflationary finance. The main points of i)ur analysis do not change materially if we introduce random discrepancies between inflation and changes in the monetary instrument. in each period. or the discount factor for the next period.TC~.ii#. we could have shifts in velocity or control errors for the money supply. Also. and variance. Later on we modif) this informatienal structure. LO. . is :>ositive. .1) uwh ii term of the form. !I.. is distributed rand.

A. the policymaker treats the current inflationary expectation. Therefore.J.106 R. The solution from minimizing Ez... (4) Since innation shocks are zero in equilibrium .. only on variables that are known also to the private agents. . as given when choosing the current inflation rate.) Therefore. everyone knows all previous values of tihese parameters.&a. where z. Barr0 and D. the policymaker effectively chooses rc. (3) We use carets to denote the solution under discretion. Then. (1) ends up depeeding only on 5.the cost from i:q. Since future costs and expectations are independent of the policymaker’s current actions. xl” =0 (rule). and rzf together. In particular. (3) . =z. the cost 1s 2. as well as follows immediately that the best rule prescribes zero inflation at all dates. the discount factor does not enter into the results. b.namely. (Q) . rr. the policymaker can condition the inflation rate.and all future expectations.B.. the expectations are e n. (l).=( 1/2)(Q2/a 5. and qr. Gordon. Discretbnary policy Our previous paper [Barre and Gordon (1983)] discusses discretionary policy in the present context as a non-cooperative game between the policymaker and the private agents.‘.rcr= 0 .that is. rcf+i for i> 0. RF. This rule can relate n. In particular.)’ . P&y under 8 rule (discretion). Ez.hence. tit . to variables that the policymaker knows at date t. But.. Therefore. K. is (discretion). at date t... In the present case no one knows the parameters. (5) Suppose now that the poligrmaker can commit himself in advance to a rule for determining inflation. Given the way that we modeled the costs of inflation .) Given rational expectations. R.. people predict inflation by solving out the pobcymaker’s optimization problem and forecasting the solrrtion for 5. (1). is given in eq. (The policymaker could randomize his choices. discretion and reputation in monetary policy 4. while treating 71: and all future costs as fixed. In-the present case they can calculate exactly the choice of inflation from eq. drops out of the cost function in eq. as (a/2)(7r. but he turns out not to have this incentive. the term that involves the inflation shock. subject to the condition that rr:=~. (With other cost functions. would depend also on rc:. is chosen to minimize the expected cost for the current period.. x.

Rules. where the rate of growth happens to be zero. The expected cost follows frorn eq. (7) The general point is that the costs under the rule. How much can the policymaker gain in period t by cheating‘? Assume that people: have the inflationary expectation.g. the policymaker is tempted to renege on commitments.. the result for 5. =:=O. 6.) = ( 12)( N2 ‘Q> 0. ‘With a different cat function. = 6/a (cheating). Barre awd D. that minimizes z. are lower than those under discretion.but.that is. 5. til ~0 .*-. contractual agreements between the policymaker and the private agents. we can calculate the costs under a rule from eq. from following the rule.then the policymaker would like to implement a positive inflation rate in order to secure some benefits from an inflation shock. x:=0. The lower cost reflects the vale of being able to make commitments . is the one that we found before under discretion4 . (9) The general point is that this expected cost is lower than that. B.2. Cheating and temptatioml As noted by others liIe. We refer to the difference between these expected costs as the temptation to renege on the rule --~-or simply as the temptation. the choice of a. Finally. $. Gordon. from eq. (5).r that under disrwtm.J.R. = -- ( U91i)2/a (cheating). In the pfesent case we have temptation = E( z. this desire does not stem from a peculiarity in the policymaker’s tastes. Ey. 5. (6) amounts to a constant-growth-rate-rule. which they formed at the start of period t. Further. i. ‘. (8) We use tildes to denote values associated with cheating. Rather. inflation ends up being excessive . no benefits from higher inflation result. If the policymaker treats this expectation as a given. (1) as 2: =o (rule). (1) as 42 I.namely. it reflects the distortions that make inflation shocks desirable in the first place. generally differs fro. In occurs under the rule . . discretion and reputatinn in mmetary policy 107 We use an asterisk to denote the results from a rule. Taylor ( 1975). Friedman ( 1979)]. Without these commitments.B. if people expect zero inflation .specifically.

then everyone raises their expectations of future inflation in some manner. Rules. In our subsequent analysis this cannot happen in equilibrium. We consider here only the enforcement that arises from the potential loss of reputation or credibility. Discretion is worse than the rule because first. the rule can specify some dependence of 7~: on the realizations of all variables through date t.utcomes. a credible rule comes with some enforcement power that at least kalances the temptation to cheat. But. the cheating outcome .that is. the cheating outcome is feasible only when people can be systematically deceived into maintaining low inflationary expectations. That’s because the inflation shock eliminates part of the existing distortion in the economy (which is worth the extra inflation). if the policymaker engineers today a higher rate of inflation than people expect. (3) discretion. This ‘mechanism can apply here because of the repeated interaction between the poticymaker and the private agents.I.ting . E.*. these are (1) cheating (with people expecting the rule). Gordon.&= -( l/2)(h2/a. the commitment under the rule avoids excessive inflation. Hence.. However._~=R~_. generate results that are poorer than the second best (rules) and closer to the third best (discretion). no inflation shocks arise in either case. There is a tendency for the pursuit of the first best -. hut second. (2) rule.that is. and ‘Thi% type of repeated game is discussed in . the incentive to cheat determines which rules are sustainable without legal or institutional mechanisms to enforce them. . discretiorl and reputation in monetary policy At the present stage we have three types of c. Friedman (1971). Burro and D. as before. & =( 1/2)(h2/a.?’ if TE.108 R. Generally. the values for date I are still not observed when ~7 is set. However.&en peopie anticipate the rule delivers better results.B. n. Ranging from low costs to high. the cost of cheating today involves the increase in inflationary expectations for the future.1 . or it might dictate some non-zero rate of inflation. The rule might prescribe ~7 =0.J. for period t. zf = 0. We postulate the following form of expectations mechanism. the rule is only a second-best solution. 5 Specifically. Consider a rule that specifies the inflation rate. in a general way. Enforcement of rules Generally. Chea. which we eventually show to be rational: 11) nf=z:.

If the policymaker cheats during period t. x.*.4= fi. loss from transgressioUs.1 + IT:__1.. The cost of this violation is that discretion. If the government follows its rule in every period. @ = A. [Note that ~9. if the government cheats during period C. + 1. then his best choice of inflation is li.J. expectations are n. from eq.. although at the discretionary levels. . I( 1 -+-V. credibility is restored as of period t +2 .a from eq.hence.$+ 1==O. (10). 5. Rather. the mechanism in eq. =li. is again the discretionary outcome.R. then it also validates expectations each period.from eq. Since costs for period t _t I are discounted by the factor 11.] Then. ( 1 I) specifies only ale period’s worth of punishment for each ‘crime’. Then. z. if in period t + 1 the government chooses the discretionary inflation rate. then the second part of eq. I. the policymaker realizes next period the cost. the actual and expected inflation rates coincide.I in eq.” Other equilibria exist that have punishment intervals of different length. the first part of eq. then people do not expect the government to follow its rule this period .lal given that expectations are ti.. (3. (S). Accordingly. Then. (11) says that pcoplc anticipate the rules outcome. Barto and DA Gordon. t 1 = (1/2)(@‘/a. [ 11) says that the size and length of the punishment do not depend on the size of the crime. things carry on as of date t + 2 as though no previou violation had occurred. discretion and reputarion in numerary pnlic~ 109 In other words if the previous inflation rate. to refer to the expected present value of the . nf #@.that is. Consider our previous rule where rr.*=O. Therefore. for the followmg period. But. Hence. are the ones associated with discretion. R. rather than that. from eq. where A. @_I. In other words the ‘punishment’ from violating the rule during period t is that the discretionary (non-cooperative) solution obtains during period t + 1. private agents anticipate that the policymaker will optimize subject to given expectations. (11) says that the government maintains its reputation (or credibility) in each period. the policymaker gains the temptation.*.I = ( 1/2)(Q2/a. if the actual value departs from expectations last period. applies for period r + I. (11) says that the next period’s expectation!. Hence. rather than the rule. as we discuss later on. I). On the other hand. But. so that nf = 0. (71. enforcement. Suppose that the policymaker has credibility in period t. which defmes a discretionary situation. _ I. the first part of eq.that is. then people trust the government to perform in line with its announced rule for period t . fit + 1 (which is optir. Rules. the = expected present value of Ihc loss is We use the term. E($--‘. (8). accords with expectations.~.

Iknown. at least given the expectations mechanism from eq. In the present setting.110 R.that is. X. rrf. The rules that can apply in equilibrium are those that have enough enforcement to motivate the policymaker to abide by them..” -. @=$=O. he opts for the cheating solution. Hence. rather than discretion. each individual’s projection. rather than cheating. given the expectations mechanism in eq.and Milgrom and Roberts (198(E). Gordon.mnot be rational ‘if the gan~e has . people know whether the policymaker will fmd it worthwhile to cheat. Burro and D. 4.(i. (and suffers the consequences next period).that is. Our framework assumes no known !erminati. -The expectations mechanism from eq. Therefore. if the enforcement is at least as great as the temptation. the solution unravels period by period. n. Generally. (I). a higher probability of termination shows up as a higher orscount rate . given the way the policymaker behaves and given the way others form their expectations. equality would arise for a value of ITthat is either above or below one. the ideal rule.( 1/2)(Q2/a.. the enforcement is 4 .= n. (11).2. In particular. (12).] 8. the policymaker’s choice n.. if the cheating solution is preferable to the rule. the expectations are rational.B. . For some related game-theory literature. But. are not greater than the expected value of having the cooperative (rules) outcome next period. no punishment arises for crimes in the last period. maximizes his objective. Then. where the parameters. rather than eq. n: =0.that is. Then. as a lower mean discount factor. discretion and reputatiorl in monetary policy P’he policymaker abides by the rule during period t . [With a different form of cost function. temptation = E(z. Working backward. see Selten (2978) Kreps and Wilson (13&O). Second.not stick with the expectation mechanism from eq. zero inflation is not an equilibrium in our model.J.*. Therefore. (IO): the temptation is (1/2)(5)‘/a. people would .) 5 enforcement = E[q. when forming expectations for period t. sets rc. the ideal rule may or may not be enforceable. (11). Rules. First. 8 Since 4~ 1.. satisfies the enforceability restriction. Then. the equilibrium satisfies two properties. given the way people form their expectations. subject to the constraint that the enforcement be at least as great as the temptation.’ In equilibrium rules satisfy the enforceability restriction. + 1 -z. $‘. Hence. Consider now whether the proposed rule. is irrational. (l). (I I) c. is the best possible forecast of the policymaker’s actual choice. rrr. finite endpoint. is not enforceable. Otherwise.e one that minimizes expected costs.*+ 1)]. The best enforceable rule We look here for the best enforceable rule . *The two terms are equal when 4= 1 only because of the specific cost functicn from eq. the temptation is strictly greater than the enforcement.n date for the game. then the expectation.* =0. (13) This condition says that the costs incurred today by following the r!lle. th. although the @me may end probabilistically. (11). From eq.-ii. while from eq.

That’s because the cheating solution and the rule prescribe the same inflation rate. at this point (As 71increases further.) The enforcement equals 4(@‘/2a when x = 0. ( 12). Then. That’s because the cost from losing reputation becotnes smaller when the rule prescribes a higher rate of inflation. discwtion and repuiatiorz in monetmy poli~. we focus on the value of II: that delivers the best results in the sense of minimizing the expected costs from eq.the policymaker prefers the discretionary inflation rate.J. when the policymaker cheats. (2).~. to the maintenance of the rule. the enforceability restriction turns out to hold with equality for the best enforceable rule.) = (a/2)( 6/a . the temptation equals zero. Finally. as long as this value is . when 71equals the discretionary value. the temptation increases. 6/a. the temptation follows from eq.2. is not enforceab!:. (When rr increases further. (Note that rr. We can rationalize this focus on one of the possible equilibria by allowing the policymaker to choose which value of rr to announce in some initial period. at date r.hence. i. are unobservable X. to higher rates of inflation. Specifically. the best rule has the simple form.namely. &/a. 1.rc)2. rr. n? = rr. and (15) We graph the temptation and enforcement versus the inflation rate. where people anticipate the inflation rate h/u.R.for given expectations . as II rises.E =rt also applies here. the enforcement becomes negative . we can calculate the temptation and enforcement associated with the rule. Rules. Barre and D. and qt.) The results are temptation = E($ . ($. the enforcement is at least as large as the terilptation for values of rr in the interval. Amcjng these. from eq. Here. the discretionary policymaker abides by the rule. the enforcement equals zero when x equals the discretionsrq value. TV. Then.*=7i. rr=O.) At at=O. where people expect an inflation rate that exceeds b/n. But. we already know that the ideal rule.u)( l--(7):( 1 +q) 5 TIN 6:~. as x rises. Then. peopie expect the same rate of amount b/n as when the inflation -. because . (14) That is. the enforcement declines. As with the temptation. the policymaker prefers the punishment.) Notice that fig. the rule specifies constant inflation (a ‘constant-growth-rate rule’). there is no enforcement. Consequently. Gordon. (10) as (6)2/2a. the gain from cheating diminishes .~ 111 h. (This figure was suggested to us by John Taylor. Usmg the procedures described before. Given this. the temptation falls.that is. in fig. be equilibria.B. I determines a range of announced inflation rates that c.

~r=l$u. Rules. we as!sume that the private agems foBlow along.tion and enforcement intersect in the interior of fig. of the policymaker’s acCons: while taking as given the way the policymaker aves and the way other agents form their expectations. each agent calculates besl forecast. Gordon.) ‘But. . In the enforceable vange. recall that the equilibrium is itself n’on-cooperative.112 R.B. 1. The best of the enforceable rules occurs where the curves for tempti.70 enforceable discretion Fig. In particular. Temptation and enforcement. 1. n:.’ Within this setup.J. announce the value of n that leads to a minimum of expected costs. Burro and D. (ll)]. (The curves also intersect at the discretionary value. but expected costs are higher at this point. the policymaker will. they all use the announced value of 71as the basis for assessing the policymaker’s perftirmance [in accordance with eq. That is. in fact. discretion and reputation in monetary policy = C(a/23 [(F/a) 2 _ ideal / rule best .

implies a relatively high weight on discretion . Gordon. A relatively small valu: of 4. the announced inflntion rate is 7r*7 z(6/u)(1 . 6/n. In fact. 4. leads to higher inflation (but. and that for discretion.R. Rules. so that the parameter a is small. which depends on the parameter u. Also. which triggers high inflation (via high monetary growth). 7c*. reduces rr*. with the weights depending on the mean discount factor. if inflation is not very costly. but which is still lower than the third best (discretion). x*. (16)]. Remember that the ideal rule is itself a seccnd-best solution. relative to the costs of inflation. a. during a war we anticipate that 4’ will be low. Generally.that is. which depends on 5.#/( 1 + 4) (best enforceable rule). (A zero discount factor means zero enforcement. which means a high rate of discount on future costs. 0. and discretion. 4.3)(6)2. which is (1 C?)(6)2 ‘a. has a corresponding effect on inflation. which is inferior to cheating when people anticipate the ideal rule.B.) Notice that any force that influences the mean discount factor.” On the other hand. the inflation rate. raises the temptation. which makes the ideal rule non-enforceable. the lure of the better outcome from cheating creates the temptation. brro and D. on the ratio of cost parameters. anything that raises the mean benefit attached to an inflation shock. rr* tends to 6/u . 6/a.J. We end up with a cost that exceeds the second best (the ideal rule). relative to the enforcement. But. the attraction of the first best makes the second best unattainable. That’s because a decrease in 4 weakens the enforcement [eq. an increase in the mean discount 4 tends to zero. An Increase in the ratio. with TE* tending toward zero (the ideal rule) as 4 tends to one. (18) Notice that. discretion and reputation in monetary policy 113 Hence. Rather.the discretionary result . which requires rr* to increase. not to . In particular. is intermediate between the ideal ule. which requires rr* to increase in order to maintain the equality between the enforcement and the temptation. Hence. with O<g< 1. cheating cannot occur systematically when people understand the policymaker’s incentives and form their expectations accordingly. 6. which is zero. the best enforceable rule is a weighted average of the ideal rule and discretion.aJ g1 -ij)/(l +ij)J2. (17) for which the expected cost in each period is Ezi’=[(1. then we end up with a lot of inflation. 6:a. The expected cost from eq. This ratio pertains to the benefit from inflation shocks. so that on!y discretion is credible. a high value of x *. For example. (18) is also intermediate between that from the ideal rule. The other feature of our results is the dependence of the inflation rate.

during a recession.y the informational structure in ways that motivate the policymaker to employ a contingent rule.114 R. This view mounts for’ ’ a rise in the mean inflation rate along with a rise in the natural unemployment rate (as in the U. For example. If people also condition their expectations. ‘-when the deadweight losses from conventional taxes are high.when the natural unemployment rate is high. and q. In each case we predict that the high value of & triggers a high value of n* _. we discussed some changes in the economy that can affect the benefits from inflation shocks. and q.J.-_ inflationary effect from the outstanding real stock of public debt. --during a war or other period where government expenditures rise sharply. h. So. a high rate of monetary expansion by the which focused on the results under discretion. q. when setting demands for real money balancas or of real government bonds. Barre und D. their real wealth nt inflation shocks occur. on 6. an x- We get some new results when we mr3di. that is. I+.. . although thL svernment and privat agents may have the same information at any point ‘ “Stmr of Ihew results can UISO be sxpl. over the last lO--15yea. this olka is ptobably ~rn~i?~~~~~l rn~~~~~rygrowlh during wartime irnd in less developedcountries. for . discretion ond reputution in mcrnetary policy more benefits from inflation shocks).B. the parameter li tends to be high in the following cases: -. In our previous paper [Barro and Gordon ( 1983)J. R. For example. One possibility is that the policymaker receives information more quickly than private agents. Then. Rules. For example. and ~ when the outstanding real stock of nominally-denominated public debt is large. the values of the benefit parameter. and “. and the discount factor. R:. then the results change ttte from those already presented. Suppose that the policymaker knows. However. our setup does not require this informational asymmetry. we focus on the case where RI* still is enerated without knowledge of the contemporuneous variables. “_” high rates of monetary growth in some less developed countries. when choosing the inflation rate.ained by chungcs in the optimal lux rate cn cash ~~~~~~~~~ which applies IO the systcmrttic port of inflation.S. ogle hold the government’s nominal liabilitias... Gordon. “___” countercyclical response of monetary policy. Therefore. PDple have to forecast rates of inflation.. the inflation rats varies each period in accordance with the state of the economy. h rates of monetary expansion during wartime.

the situation resembles that from above where people choose their holdmgs of money and bonds based on forecasts of inflation. . some models with long-term nominal contracting [Gray (1976). Looking OIW period ahcad. we can think of RF as not being conditioned on the realizations. Specifically. in Lucas (1972. I+. Therefore. However. The situation is less clear for the example of the Phillips curve. Therefore. we get Ii. Gordon. Burro and D. However. production. generate departures of inflation from ias expectation. . ~6.g” ( I /2)(!#/a . B. However. Taylor (1980)] suggest that inflationary expectations formed at earlier dates will matter for today’s choices of employment. (20) The results correspond to those from before [eq. people’s expectations not conditioned on h. (1’)) Now. we can culculute ‘The new term is the variance of the benefit parameter. rather than to its mean. 5. -I. discretion and reptrturion in ntone~ury poiic! 115 in time.J. these realizations can influence the actual inflation rate. Fischer (1977). cri.are RF =6/a. z.R. h.6). [5)] if b. the agents’ decisions (on how much real money and bonds to hold) depend on expectations of inflation that were formed earlier. Therefore. In models where only unperceived nominal disturbances matter for real variables --.(b. the rationality of the Gray-Fischer-Taylor contracts has been questioned [Barre ( 1977)J.. 1973) and Barro (1976) ./a)(h. The costs under discretion are now 2. the policymakcr reacts to the actual value of the benefit parameter. the inflation are sometimes shocks --.the pertinent value for @ is the one based on contemporaneously available information. and q. the realizations for I). Then. = b&l.. although n: = Eti. etc. Rules.. and . We find the rasults under discretion in the same way as before.he corresponding benefits from them positive and sometimes negative.

e. bite the bullet through uncxpecrediy iuw iafia. benefits in the high states. realized inflation exce.that IS. The costs associated with the ideal rule turn out to be . the expectation of these costs is (24) e ~Qllcymaker can match the variations in h.when b.*.the realizations for the benefit parameter. rather than negative. inflation is below its expectation . 6. b. is high. the negative inflation shocks may appear pointless. Conversely. The present example is simple enough to write out the ideal contingent rule in closed form.. inflation for the low states means that the prior expectation of inflation is higher than otherwise. when on this investment when it is most important . the ideal rule says that the policymaker should ‘bite the bullet . when 6. ex post.namely.under some circumstances. -6).* ‘I = --( 1 2Ll)[(h.So.)~3. But. Note that inflationary expectations are always zero. is below its mean.ion in order to ‘buy’ the unexpectedly high inflation and the corresponding. Then. (22) As before. =fi. is low . rather than negative. it would cteariy be preferable to have zero. is re:latively high.)2-(&'I- (23) Agam. Yet.. deflation) when the benefits parameter takes on relatively low values. That’s because the surprisingly low rule of inflation when the benefit parameter.the best rule turns out (after a targe amount of algebra) to be ?r: =( l/Q)(h. These realizations may show up as a recession or as other costs from a negative inflation shock. cause a recession through coptractionary monetary policy . Yet.when h.’ = En. but the policymaker creates surprisingly low inflation (i. Then. we get our previous results [eq. inflation. the prior mean of inflation. Looking ahead one that costs from unexpectedly low inflation occur . it is worthwhile to incur some costs in the low states . exceeds its mean. the policymaker achieves lower benefits in the states where b. n. Specifically -abstracting from enforcement problems . (711 if b. is zero.eds its expectation . with appropriate . In effect the policymaker invests in credibility when it is relatively cheap to do so namely.and benefit from inflation shocks arise . we avoid the negative inflation shock and also have less costs due to inflation [which are (a/2)(7~. is low is the counterpart of the surprisingly high rate of inflation when the benefit parameter is high. In fact.that order to cash . Choosing zero.

we determine the value of the c-coefficients in order to equate the temptation to the enforcement for all values of (h. b.. . (25) where the c’s are constants. Since eq. and q. ISThe enfon:emc*nt is linear in q. cz=l:‘a.irr linear m q. q. the tempt&m discount factors.+c3&. “With other cost functions. l’j= -2(F. n:. Then. the trmpt. ’ ’ As before. (25). c’~. Further. (25) has the correct form. contingent rule. Thcreforc. The present example is sufficiently simple to work out the results in closed form. as a stationary function of the state. Hence.. l4 The temptation cannot exceed the enforcement for arv of these realizations in order for the rule to be credible.. if the en *cement exceeds the temptation in some state. (aI.. the expected costs fall with an increase in the variance of the benefit parameter. Cam.13 Therefore... and of J.rtion m\olves the square <If the mflation r::It’.. and the realizations for b. if we conjecture that the rule for inflation takes the form of eq. the best enforceable rule in our model turns out to equate the temptation to the enforcement {sr all realizations of b. But.). In particular.. 9. if 77 14Lncar 111 T. the enforcement maI r~~txcl the temptation for ~omc rcaiizations.. we can show that the ideal rule is not enforceable in our model. we then find that thr inflation rat. Further. 0:. the best enforceable contingent “When co:Adering the 1de4 rule.q. thrn the trmptatlon itIs<) mvol\rs terms that . then we can do better by changing the inflation rate for that state. which specifies the values of the two variables.. which have to be determined. we go on now to examine the best enforceable. L . that are less than one.responses in II. (26) rule wh’:re V.” *t *=cI +c. thr temptation and rnforccmcrnt turn ijut IO lx rxeeds the enfonxment for all independent of the realization for h.that is.jt rract to vdnations in q in wmz regions. and 4. does L’S. and qt.:U)l~. Given that the ideal rule is unattainable. That is. The solution for inflation turns out to be a linear function of 6. then we can work out the temptation and enforcement as functions of the parameters. The results are “I =Q. is the mean of \ (jr.3. this operation turns out to be feasible.b.:(I +#.q. we bear more costs than necessary by having excessive enforcement. Enforceable contingent rules We look at rules that express the inflation rate.

or high v&tes of qr . ilt is worthwhile to suffer negative inflation shocks in some cases . (22). 7rt*.. the ideal rule does not depend on q. in eq. which is 6/a. large gaim are attained in the cases v~here the benefit parameter.both on the parameters of the probability distribution for q1 and on the realized value. (The last interval ’93s solution reduces to the previous one in cq.and discretion . So far. and the realized value of &. (27). 4 =: var( d/7. These last cases are likely to be emergencies such as wars or other times where economic activity or government revenues are valued especially highly.. ConverstAy. Then. and_vq== V.d to unexpectedly high inflation. In effect.B. That is. Therefore. conditional on cheatin).eq. ~:alw lowers n: Thus follows by using the formuia. q. creates negative inflation !hocks . the best enforceable rule does depend cn q1 in eq. Ruins. the mean rate of inflation is positive. 4. b.ed with discretion. relatively high reahzations for h.when the benefit parameter is lower than normal or the discount factor is higher than normal. bz = 6. or in the discount factor. IO. yI. Given the parameters of the distribution. In pa*:ticular. the length of time for which the discretionary outcome obtains.Jlet’ . (22) .) +(.eq.that is. for low realizations of ql. equals the length of time over which the policymaker can enjoy the results of his cheating. (17) if there is no rar.” ote that the realization of the discount factor does not affect current benefits and costs from inflation. low inflation rates are not credible. a higher realbzation for (I~ means a lower int?ation rate.-. T&e length of the punishment interval interval is fixed at one period. the policymaker bites the bullet during the non-emergencies in order to invest in credibility . = 4. because the temptation would exceed the enforcement. The relative weights depend on the discount factor . and relatively low ones for qr 1eP...raker ‘bites the investment that yields returns during the emergencies. The reasoning here is similar to that from before. is t &OW. md qr..that is. The inflation rate now moves around with flcctJations in the benefit parameter. the policyr8. discretion and reputarion in monetqv pky for inflation is “’ (27) The enforceable: rule can again be viewed as a weighted average of the ideal rule . is high or the discount factor.dorn variation in b.I-$. Then. but does influence the amount of order to sustain low prior expectations of inflation. (19). !br low values of 6. but lower than that associa. In partil:ular. b. a higher value of .z ‘-Given the variance for .J. Gordon. Thus. y.. our results apply when the length of the punishment . But. Bum. and D.118 R.

we obtained a unique reputational equilibrium by allowing the policymaker to announce the best one. sholj that unique equilibria sometimes obtain.b. If he could.actually occur from time to time. this loss of reputation . which means that the only equilibrium is the discretionary one.R. it is likely that the optimal punishment interval would be finite. which corresponds here to raising the length of the punishment interval.B. Then. it does not necessarily follow that the equilibrium with this punishment interval will be selected. Barrn and D. then we find an array of zputational equilibria. we always do at least as well if we increase enforcement. which amounts to a form of ‘capital punishment’. At this point. Kreps and Wilson (1980).and hence. First.never occur as part of a reputational equilibrium. But possibly. another possibility is to illtroduce uncertainty about the policymaker’s preferences. In particular. assume that people cannot fully sort out these two influences on inflation.‘” For example.i gencr ~1type of model. 119 essentially defines the length of the period. Then.that is. we have some observatiorns. there would be no enforcement. But.J. the policymaker knows that people learn from his actions. Rules. and Milgrom and Roberts (1980). and acts accordingly. Further. But. (Ei~wever. In the present model the punishments . suppose that inflation depends partly on the policymaker’s actions and partly on uncontrollable ea’ents. Thus. from a positive standpoint.fferences in tastes among potentitil policymakers.) Finally. That is. this idea would become meaningful . beccruseit relies on d. people adopt a form of control rule where the policymaker loses reputation if the observed inflation rate exceeds some critical value. we have not yet pursued this route in our context. we have nothing interesting to say about the sources of these differences. Hence. even ex post. the policymaker cannot instantly restore a lost reputation. in contemplating a more severe form of punishment. it always looks desirable in this model to have an infinite interval. However. the punishment . We can calculate the effect of longer punishment intervals on expected costs. Unfortunately. we have no satisfactory way to resolve this problem of multiple equilibria. people try to learn about these preferences by observing behavior. because of the uncontrollable element. Gordon.!t. Further. We can modify the model so that pun*shments take place occasionally. if we look at different punishment intervals (which can be either greater or smaller than one period). who uses tt1l. discretionay outcomes . Thfn.) Given the length of the punishment interval. we have to weigh :he losses when punishments occur against the benefits from greater enforcement. we know that the length of the punishment interval cannot be zero. diwrerion urrd reputation rn monerar~~ po1ic. I’ B.

Review of Economic Studies 38. Thus. and monetary policy. Journal of Monetary Economics 2. From a predictive standpoint for monetary growth and inflation. the policymaker sometimes engineers surprisingly low inflation.120 R. the problem arises in determining how long a loss of reputaticn persists. 1978. 861-8’?4. Optim:. which show up also in the related game-theory literature.A. RJ. Fischelr. Long-t*rm contracts. Jan. We have difIieulties with mul~tiplicity of equilibria. Gray. 191-205. Califorr ia institute of Technology. Green. G. Jan.. R. we can figure out the optimal length for this interval of punishment. Econometrica 46. Journal of Monetary Economics 3. Rules. Friedman. . ‘d. discretion and reputation in monetnry policy if we identified policymakers with shifting interest groups.?maker sustains a reputation that @rmits surprisingly high inflation in oth ?r states of the world. C. where inflation depenids OR the realization of the benefit parameter or tire discount factor. Noncooperative collusion under imperfect price infarmation. under clkcretion and rules. Feb.tlneory of monetary policy in a natural-rate model. 305-316 Barre. In an extended version of the model. J.pril. July. But. given some states of the world. A non-cooperati!-e equilibrium for supergames. CA. which is cos’tly at a point in time. By acting this way.B.. Aug.. References Barre. E.J. 1983. a:ld D. 1977.. Long-term contracts. from a positive standpoint. the polic. Jan.. Porter. Pasadena. Here.. Journal of Monetary Economics 2. 1971. A positive . sticky prices. In some environments the ruies take a contingent form.d extend those that we discussed previously.. Gordorl. macroeconomic approach. S. Jan. each of which were affected differently by variations in infiation. Calvo.. Journal of Po!itiG. Conchiding Our observations analysis provides an example of a reputational equilibrium tar monetary policy. the relative weights attached to the discretionary and rules solutions depend on the policymaker’s discount rate a. Jan. 1979. Barre. RJ. l-32.d expectations and the extreme informatkn a:ssumptions of ‘rational expectarions’ macromodels. Nov. 1976. 2?. Here. Now. Canadian Journal of Economics. J. un pulrhshed.J. Previously.B. 11. Journal of Politicai Economy 91. Wagr: indexation: i.. Burro and D. Rational expectations and the rule of monetary policy. 1983.l Economy 85.W. Inflationary finance. Barre.. 1111-1428.nd some other factors. RJ.1-236. It is unclear which equilibrium will prevail. the results modify an. the monetary authority ‘bites the bullet’ and pursues a contractionary policy. 1981. Gordon. 011 the time consistency of optimal policy in o monetary economy. 19T6. Friedman.J.. we analyzed discretion and rules as distinct possible equilibria.H... The results amount to a combination of the outcomes from discretion with those from the ideal rule. Journal of Monetary Economics 5. 23-42.. 1977. rational expectations and the optimal money supply rule. and R.

Lucas.. CA. 1972. Journal of Political Economy 83. 1980. unpublished. The chain-store paradox. Jan. Journal of Political Economy t(X. 1978.. and EC. Prescott.. Journal of Political Economy 85.E. J. Lucas. P. J. Taylor. Milgrom. Kydland. 1977. Wilson. Northwestern University.J.E.E. IL. Some international evidence on output-inflation tradeoffs. April. Oct..B. Rules rather than discretion: The inconsistency of optimal plans. 1973. July. and 4. June. Roberts. D. Rules. discretion and rt putation in monetary plic! 121 Kreps.R. and J. .. Stanford University. Expectations and the neutrality of money. Gordon. R. June. R.. 1009-1022. Monetary policy during a transition to rational expectations. On tht: chain-store paradox and predation: Reputation for toughness. Theory and Decision 9. 1975. Stanford. 103-124. Selten. American Economic Review 63. Journal of Economic T’r. 326-334. reputation and rntry deterrence. l-23. 127-159. 473-491. F..eory 4. 1980. Evalrston. Aggregate dy:?amics and staggered contracts. R. unpubhshed. 1980. Predation. Barrc~ cmd D.M. Taylor.