A Theory of Ambiguity, Credil Asymmetric Information lity, and Inflation under Discretion and Alex Cukierman; Allan H. Meltzer Econometrica, Vol. 54, No. 5. (Sep., 1986), pp. 1099-1128. Stable URL: hitp://lnks,[Link]/sic?sici=0012-9682% 28 198609%2954%3A5%3C 1099%3AATOACA%3E2.0,CO%3B2-%23 Econometrica is currently published by The Beonometrie Society. ‘Your use of the ISTOR archive indicates your acceptance of JSTOR’s Terms and Conditions of Use, available at hhup:/[Link]/about/[Link]. JSTOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at hup:/www jstor-org/[Link] Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the sereen or printed page of such transmission, JSTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact support @[Link]. hupulwww [Link]/ Fri Jun 9 [Link] 2006Economeirica, Vol. $4, No.5 (September, 1986), 1099-1128, A THEORY OF AMBIGUITY, CREDIBILITY, AND INFLATION UNDER DISCRETION AND ASYMMETRIC INFORMATION By ALEX CUKIERMAN AND ALLAN H. Metter! This paper develops a positive theory of credibility, ambiguity, and inflation under discretion and asymmetic information. The monetary policymaker maximizes hit own (politically motivated) objecve function that is positively related 1 economic stimulation ‘through monetary surprises and negatively related to monetary growth. The relative impor lance he assigns to each target shifts stochastically through time. His current preference ‘tadeoflis known to him bat not othe public. When choosing the (state contingent) path ‘of money growth forthe present and the future, the policymaker compares the beneis from curren wtimultion withthe core associated with higher future inflation expectations ‘Current monetary growth conveys information tothe public about future money growth ‘because there is persistence in the policymakers objectives. Although expectations are ‘ational, information is imperfect because monetary control procedures are imprecise. As ‘result the public cannot corectly distinguish persistent changes of emphasis diffrent policy objectives trom transitory monetary control errors. The public becomes aware of changes gradually by observing past monetary growth. Credibility is defined in terms of | recognizes changes in the objectives of the policymaker. lower the noisier monetary control and the more stable the objectives ofthe policymaker. Looser monetary control and a higher degree of time preference on the part ‘ofthe policymaker induce him to produce higher and more variable monetary growth. ‘When the polieymaker i fee to determine the accuracy of monetary contrl he does not always choose the most effective contol available in spite ofthe fact that monetary surprises always have an expecied value of zero. The reason is that ambiguous control procedures enable the policymaker to generate positive surprises when he cares more than fn average about economic stimulation. He leaves the inevitable negative surprises for periods in which he eares more about inflation prevention. This result provides an explana- tion fr the Fed's preference for ambiguity, recently documented by Goodfriend (1986) ‘The policymaker is more likely to pick more ambiguous contol procedures the more ‘uncertain his objectives andthe higher his time preference. "The paper algo provides a theoretical underpinning for the well documented cross: 0 and the predictor tends to i”. However, we have nat been able to show thatthe weight piven to 8” fsa decreasing function ofp in ll the range between zero and fone although such a result seems plausible.‘THEORY OF AMBIGUITY 107 Substituting (11b) into (11a), substituting the resulting expression into (3), substituting this result into equation (2), using (1), and rearranging, the policy- maker's problem can be rewritten as (12) max Bon So [( mew, it 1 PB, <9 E Mt Vis) a The policymaker chooses the actual value of mf and a contingency plan for ‘mf, i= 1. Recognizing that in each period in the future the policymaker faces @ problem that has the same structure as period zer0's problem, the stochastic Euler equations necessary for an internal maximum of this problem are, following Sargent (1979, Chapter XIV), (13) x= (B—AYBBoi X11 BAX 42+ (BAY iy) m= 0 (i=0,1,...) Equations (13) yield the actual choice of mf and the contingency plan for all future rates of money growth for i> 1.'* Although the policymaker knows x, in period i (and the public does not), he is uncertain about values of x beyond period i. Based on the information available to him in period i, he computes a conditional expected value for x1.) 1. In view of (4) and (5) this expected value is (4) Boxy At Eoin +p! =px+(1=p!)A, 720. Substituting (14) into (13) using (4) and the formulas for infinite geometric progressions, and rearranging, (13) mp= 1 Be 1p? inert apa Rationality of expectations implies that the coefficients of A and of p, should be the same across equations (15) and (6) respectively, so (160) B= Bay 1=Bo* Bor(B) (166) li Boo ~(9-A)Eo FAY Siem] 0 is satisfied for any <1 since the ferm inside the brackets following Eg is finite. This condition ie Sfcient for an internal maximum,1108 ALEX CUKIERMAN AND ALLAN H, MELTZER ‘The dependence of A on B through equation (10) is stressed by writing A as a function of B. Equation (16b) determines B uniquely as an implicit function of Bp, 03, and a3. Uniqueness is demonstrated by noting that the right-hand side of (16b) is increasing in A and that A is decreasing in .° The definition of r in terms of B from equation (10c) implies therefore that both A and the right-hand side of (16b) are monotonically decreasing in B. Hence this equation yields a ‘unique solution for B. Given this solution equation (16a) determines By. {As long as the optimal predictor of money growth is linear in the information set of the public, this solution is also unique. This can be shown by writing the ‘optimal predictor as a general linear function of all past rates of money growth and by allowing m! to be a general function mi = FP Pits Pe-as-) of the entire history of governmental objectives. Note that this formulation is similar to that of Green and Porter (1984) in which the current action of a representative firm depends on the entire history of the industry price. ‘The proof of uniqueness proceeds by substituting the general linear predictor into government's objective function in (2), deriving the Euler equations and showing that they imply the function F above to be a linear function of p, only, as postulated in equation (6). A proof is available upon request. Note that even if the costs of inflation to the policymaker had been expressed in terms of actual money growth rather than in terms of planned money growth, the resulting equilibrium would be the same. The reason is that for any 1=0 (my ‘The second term on the right-hand side of this equation does not depend on m! and the third term has an expected value of zero, given the policymaker's information in period 0. Hence the Euler equations for this reformulation are still given by (13) leading to the equilibrium solution in (15). = (mt P+ yi 2mbh. 4, CREDIBILITY AND THE DETERMINANTS OF THE DISTRIBUTION OF MONETARY GROWTH Credibility is defined as the absolute value of the difference between the policymaker's plans and the publie’s beliefs about those plans. The smaller this difference, the higher the credibility of planned monetary policy. Credibility is relatively low when governmental objectives undergo large changes. In addition it is lower on average the longer it takes the public to recognize a change in governmental objectives. The weight A in equation (10b) measures the degree of sluggishness in expectations. The higher A, the longer is the “memory” of the public and the less important are recent developments for the formation of current expectations. With a low A past policies are quickly forgotten. It can be shown "Let bm p-(142/p. Then from (10b) 2A/ab = (1/2)/B/4~1(VBF/4—1~b) whichis negative since (3/4)64+10, Since is increasing in %, is decreasing in r,THEORY OF AMBIGUITY 1109 that A is a decreasing function of o3 and an increasing function of o3;" the effects of past choices of monetary growth on current expectations are smaller in comparison to more recent choices the larger o? and the lower 3. People give less weight to the more distant past the larger the variance of the innovation to governmental objectives and the lower the variance of the control error. The worse the control of the money stock (high 3), the longer will past policies affect future expectations Since A is related to the speed with which the public recognizes shifts in governmental objectives iti a prime determinant of the credibility accorded to new objectives of the government. Suppose that after remaining above its mean value m! decreases below the mean. This more conservative attitude towards inflation will take longer to be recognized by the public the larger is A. Therefore, the worse the control of the money stock the lower the credibility of shifts to rates of monetary growth that differ from those previously experienced. A can therefore be taken as a measure of credibility. The higher A, the longer it takes the public to recognize a change in governmental objectives and the lower, therefore, the government's credibility. ‘We turn now to the investigation ofthe distribution of money growth, Substitut- ing (15) into (7), actual money growth can be rewritten 1-60, 1-Bo* =x" pps with unconditional mean and variance that are given respectively by +H 1-Bo 7) Bm = Be 7 bey (8) ¥(m) (Ee i= Since A>0,0= A, p= 1, the average rate of monetary expansion is positive pro- vided policymakers have some degree of time preference (8 <1). Equation (17) suggests that mean monetary expansion is systematically related to the underlying parameters of the model. The following proposition summarizes the effects of some of those parameters on average monetary growth, Propostrion 1: For any <1 average monetary growth is larger (a) the higher A, and (b) the higher 03. ‘Let av 03/3. The total effect of a change in a 00 A, taking the dependence of B on a through (169) into consideration, ie «dada = (2/0). B°(1~BpA)/p(1~Bpd ~2aB9B*0A/an)} hich is negative since 94/36 and 64/ar are both negative as implied by footnote 15, The result in the text follows by noting that a is positively related to ai and negatively related to 03.110 ALEX CUKIERMAN AND ALLAN H, MELTZER Part (a) follows immediately from (17) by noting that 6 <1, 0p, A <1. Part (b) follows by noting that Em, is an increasing function of A which is in turn an increasing function of a} (footnote 16). Part (a) of Proposition 1 says that average monetary growth is higher when the policymaker is more biased towards economic stimulation than to preventing inflation (high A). Part (b) implies that average monetary growth is also higher the less effective is control of monetary growth as measured by a relatively high 3, The reason is that at higher a3 it takes longer for the public to recognize a shift to @ more expansionary policy. The negative eflects of increased monetary expansion on government's objectives are delayed, so the government gains more from trading future inflation for current economic stimulation The following proposition summarizes the effects of the underlying parameters of the model on the variability of monetary growth. Prorosirion 2: (a) For any B <1 the variance of monetary growth is larger the larger 23, (b) Fora given finite value of the variance of the monetary contol error, 3, the variance of monetary growth is larger the lower the discount factor B. Part (a) follows from (18) by noting that V(m,) is an increasing function of 2 both directly and through its dependence on A. Part (b) is proved by noting that V(m) is increasing in B and that B is decreasing in B."" Variability and uncertainty are not identical (Cukierman, 1984, Chapter 4, Section 4). A natural measure of monetary uncertainty is the variance of the money growth forecast error V(e) = Elm, — Elm 1]P Prorosrrion 3: Fora given value ofthe variance of the monetary control error, 7}, monetary uncertainty as measured by V(e) is larger the lower the discount {actor B. The proof is developed in Part 2 of the Appendix Part (a) of Proposition 2 states that the variance of monetary growth is larger the larger the variance of the control error in money growth. There are two reasons. First is the direct effect. For any level of planned growth actual monetary growth is more variable. Second is the effect on A. The public is slower to detect shifts in governmental objectives, so it pays government to induce a higher degree of stimulation by creating more uncertainty. For a similar reason when the discount factor is low, government discounts the costs associated with expectations of future inflation more heavily, and chooses more current stimulus. As suggested. "The last relation follows by diferentaton of (16b) with respect to 6. This yields 38/08 = 28p(4~p)/(1~ Bak}. This expression is nonposie since Ax p and strictly negative fora finite a,‘THEORY oF AMBIGUITY mu by Proposition 3 this is done by creating more uncertainty, part of which takes the form of higher variability. The deterministic component of the policymaker's objective function in (2) is similar to that used by Barro and Gordon (1983b). As a result, our solution exhibits a similar inflationary bias which increases with A, the average relative preference of the policymaker for economic stimulation (Proposition la). A novel element of our framework is that the public's information about the shifting objectives of the policymaker changes overtime.” Shifting objectives and noisy control permit people to supplement the information on average objectives, given by A, by observing past money growth. Our solution specializes to Barro and Gordon's discretionary solution when asymmetric information is removed from the model. Formally, asymmetric information is eliminated when for a given (nonzero) «7 the variability of governmental objectives, 03, is zero. In this case the actual relative preference of the policymaker for stimulation is common knowledge and is given by A. When o=0,r=0 which implies through (10b) that A =p and through (4) and (5) that p,=0 for all i. In this case the solution in (15) specializes to mi =A forall i=0, which is the solution obtained by Barro and Gordon under discretion. In addition it can be seen from (9) that in this case Elm|i)=4. ‘That is, as in Barro and Gordon (1983b, p. 595), individuals do not pay attention to actual rates of money growth in forming expectations. Since they know with certainty the structure of governmental objectives they do not need to use information about observed rates of money growth to forecast future growth. Whatever the realization of money growth, individuals interpret its deviation from A as a transitory control deviation and stick to the preconceived notion that future monetary growth will be A. In this particular case the only way to change expectations is by convincing the public that discretion has been abandoned in favor of a different regime. In the more general case considered here, expectations depend both on the Preconception, *i” in (11a), and on the past history of money growth. This last dependence is induced by asymmetric information between the policymaker and. the public. Correspondingly, as can be seen from (15), planned monetary growth is composed of two components. The first, which depends on A, is common knowledge. The second, which depends on p,, is known with certainty only to 0 and Gordon (1983a and 1983b) limit their analysis of discretionary policy to the casein which the public does not need to learn about changes inthe policymakers objectives because those objectives are fixed. "3 For 0, (106) implies that A=(1/2)(\/0+o)-VOU/AaN7p+0Pm2 ALEX CUKIERMAN AND ALLAN H. MELTZER the policymaker. It is this second component that makes expectations dependent ‘on past rates of money growth and makes it possible to change expectations without necessarily changing the discretionary nature of the policy regime. Note that the relative weights given to the preconception 1” and the past history of monetary growth depend on the degree of persistence, p, in that part of govern- mental objectives about which the policymaker possesses an information advan- tage. When p tends to one the weight given to the preconception ri? becomes negligible (see (11a)). In Barro and Gordon (1983b) and in the example introduced by Kydland and Prescott (1977, 477-480) a rule which binds the policymaker to set m?=0 in all periods is believed, so the expectation of money growth is zero. The rule achieves a better value for the policymaker’s objective function than discretion. However, none of these frameworks incorporates asymmetric information. In the presence of asymmetric information a zero rate of money growth rule does not always achieve a better value for the policymaker’s objective function. The reason is that the public is slow to recognize shifts in governmental objectives. As a result, for sufficiently unstable objectives and slow adjustment of expectations, the positive contribution of a current positive ¢, to the policymaker's objective function in (2) can dominate the negative effects of higher inflation and future negative e,—s by enough to make discretion preferable to a binding zero rate of money growth rule2® In such cases there is no difference between the [Link] solution and the optimal solution. In heir reputational paper Barro and Gordon (1983a) postulate an exogenously given “punishment” period.” They also briefly consider an extension of their basic model in which government has an information advantage. However, even in the extension, the expected rate of monetary growth is fixed and independent of actual changes in monetary growth. This expectation is rational, given the stochastic structure postulated by Barro and Gordon. But their structure lacks descriptive realism, since forecasts of inflation are usually influenced by actual inflation and monetary growth. A dramatic example is the decrease in expected inflation between the end of the seventies and the present. Further, recent empirical work leaves no doubt that expectations regarding future monetary growth and inflation change within a discretionary framework. See inter alia, Hardouvelis (1984) and the many references there. Explicit modeling of the way expectations change is essential for discussing changes in credibility in the absence of a constitutional rule. ‘The present framework links expectations to observed money growth by introducing persistence in the policymaker's objectives and imperfect monetary control. This permits us to discuss different degrees of credibility within a discretionary regime. The evidence presented in Hardouvelis (1984) suggests that in spite of the fact that U.S. monetary policy has remained discretionary, the ® An example for which this isthe case is discussed atthe end of Section 6, 2This term is borrowed from the repeated games literature to describe the relationship between past choices of monetary growth and current expectations. In this Iterator, the public chooses he Expectation so as to induce socially desirable behavior by government in supergameTHEORY OF AMBIGUITY 13 1979 change in the Fed's operating procedures changed the public's perception of the objectives of the Fed. The model of Barro and Gordon does not handle ‘a change in perceptions of this type. An additional advantage of our formulation is that it relates the speed with which expectations adjust to the quality of monetary control and other parameters of the environment. The determinants of the size and timing of “punishment” are identified rather than postulated exogeneously as in Barro and Gordon. Further, as stressed by Backus and Driffll (1985) a weakness of the Barro and Gordon analysis is that their equilibrium solution critically depends on the form this punishment strategy takes. As a result their model has multiple equilibria with no mechanism for choosing among them. We have shown that at least wi the class of linear minimum square error predictors of money growth our equili- brium is unique. Given linearity, the attempt by each individual to minimize his, error of forecast determines a unique pattern of learning that acts as a deterrent to the policymaker. This differs conceptually from Barro and Gordon's (1983a) iment mechanism. In Barro and Gordon, the public is viewed as a single player who picks his expectation strategically in order to induce “better” behavior on the part of the policymaker. Here the structure of deterrence is induced, as. a by-product, of each individual's attempt to minimize his forecast error. As stressed by Bull (1982) casual empiricism suggests that a positive amount of resources is devoted to monitoring the central bank. Our framework is con- sistent with this observation; itis rational to study central bank behavior. Further, because expectations are influenced by past monetary growth, our concept of discretionary policy permits the policymaker to consider the effects of current policy on future expectations. In contrast, Barro and Gordon (1983a, p. 106) restrict discretion to mean that “. the policymaker treats the current inflationary expectation, and all future expectations, as given when choosing the current inflation rate.” Actual policymaking in the absence of @ constitutional rule recognizes the effect of current policy actions on future expectations, perhaps with some discounting. Barro and Gordon's definition of discretion seems overly restrictive, since it applies only to a world in which current policy actions do not affect future expectations. ‘The wider notion of discretion we use applies to any arrangement in which there is no constitutional rule to restrict the range of possible actions of the policymaker. The policymaker may follow a decision rule, in the sense of the ‘optimal control literature, that takes account of the effect of present policy actions on future expectations. The use of a decision rule permits the policymaker to maximize his objective function. The policymaker, in our analysis, is free to ® Ourframework differs from tht of BG in several other respects. Our objective Function represents the attempt of the policymaker to elicit support for his policies while BG interpret the objective Function ofthe policymaker as rocal welfare function, Imperfect credibility in BO concerns the socially optimal constitutional rile and occurs because the policy rule isnot dynamically consistent In our framework credibility is imperfect even with a dynamically consistent dicretionary policy because of noisy contol and shifting objectives. Taylor (1983) rises doubts about the relevance of the Barro-Gordon model as a positive theory of inflation. Canzoneri (1985) tres to resolve those doubts by appealing to asymmetric informationns [ALEX CUKIERMAN AND ALLAN Hl. MELTZER ‘choose the weights he places on the arguments of the decision rule. The contingent decision rule that we derive corresponds to discretionary policy, since no a priori restrictions are imposed on the feasible set of policy actions. In contrast, a Friedman-type rule in which policy actions are subject to a priori, binding constraints is a constitutional rut. Recently Backus and Drifill (1985) have formulated credibility as the outcome of @ sequential equilibrium in a repeated game. Our model shares with theirs the asymmetric information about government objectives and the notion that dis- cretionary policy is dynamically consistent. It differs in that government objectives are allowed to change continuously through time and to assume an infinite number of values. In Backus and Driffil there are only two possible types of government, ‘and these types never change. Consequently, inflationary expectations can assume only two possible values. Moreover, once a government inflates it is revealed to be the weak type and asymmetric information is eliminated forever. Backus and Drifill restrict government's discount factor to 1 and do not analyze the effect of the precision of monetary control on the choice of policies. They define credibility in terms of the probability that government is “hard nosed” (has no incentive to inflate) whereas we conceive of credibility as the speed with which the public detects changes in the policymaker's objectives. ‘As in Backus and Driffil imperfect credibility arises here without any dynamic inconsistency of the type defined by Kydland and Prescott (1977, p. 475). The reason dynamic inconsistency does not arse in our framework is that the “action” taken by the public—evaluating Ef m|J,] does not depend on the future settings, ‘mts, j=), of government's choice of instruments. See (11). In cases of this kind, Kydland and Prescott (1977, p. 476) point out that the time consistent solution is also optimal. When period i+ j (j= 1) arrives, the government follows the contingency plan made in period & At the risk of repetition it should be pointed out that this 1g since our solution corresponds conceptually to Kydland and Prescott’s discretionary consistent solution in their example on pp. 477-480. ‘The basis for imperfect credibility here is the policymaker’s advantage over the public that is due to shifting objectives, and noisy control. He knows his stochastically changing objectives, but the public does not. The best the public can do isto form expectations, allowing for this noise, and use all the information available each period to infer current and future money growth. 6. THE POLITICALLY OPTIMAL LEVEL OF AMBIGUITY To this point we considered the level of noise in the control of money as a technologically given parameter. Suppose however that technology only puts a lower bound gon the variance of the control error. The policymaker can choose any o3> 23. We assume for simplicity ¢3=0 In this section the policymaker sets the value of o3 once and for all so as to ‘maximize the long-run expected value of his objective function in (12). The choice of this variance determines the politically optimal level of ambiguity inthe‘THEORY OF AMBIGUITY as conduct of monetary poliey, since a higher choice of «3 conveys more ambiguous signals to the public. For a given level of «3 the optimized value of the objective function of the government is obtained by substituting the optimal choice of monetary growth (equation (15)) into the objective function of the government in (12). Abstracting from the conditional expected value operator in (12), this yields (19) sys to)%)= § | [A+ B+ 8, = (9-4) E(BoA+ Bp. % where {y,}"~ and {p,}"., denote the sequences of w and p. Since the policymaker sets on the basis ofthe long-run value of his objective function rather than on the basis of particular recent realizations of x,, the relevant objective function for the choice of «3 is the unconditional expected value of J(e). It is shown in Part 3 of the Appendix that this expected value is (20) Go) EX if ae ‘LG Bpa)(1=pd) (16° 20= Bo 20 gery) eel where A A vow ‘The first term in brackets on the right-hand side of (20) represents the mean (positive) contribution of economic stimulation to governmental objectives. The mean value of economic stimulation through surprise creation is zero since negative and positive surprises cancel each other on average. But the contribution of monetary surprises to governmental objectives is positive on average. The reason is that the rate of money growth is positively related to the marginal benefit of surprise creation to the government. As can be seen from (15), when the marginal benefit of a surprise is higher than average (x, > A+> p, > 0) govern- ment chooses a higher than average rate of monetary growth and when the marginal benefit of a surprise is lower than average (x, p, <0) the govern- ‘ment chooses a lower than average rate of monetary growth. Consequently, when government cares more than on average about economic stimulation, surprises are positive on average, and when it cares less than on average about economic stimulation surprises are negative on average, making the unconditional expected value of the benefits from surprise creation positive. The government derives a positive gain, on average, from the ability to create surprises because it can allocate large positive surprises to periods in which x; is relatively high and leave the inevitable negative surprises for periods with relatively low values of x,. More116 ALEX CUKIERMAN AND ALLAN H, MELTZER formally, by using (7) and (11a), it ean be shown that & lex e2— pop, —pt en BE Bem = 725 epl-pt) where pr=(o-n) 3 Wp, Obviously E(p,~p?) = 0. But Ep,(p,~ pt) > 0 since there is a positive correlation between p, and p,—p'. This positive correlation is created by the government's attempt to maximize its objective function which leads to the contingent behavior summarized in (15). By contrast when the government does not possess an information advantage, o?=0 and p, is identically zero. This implies that the expression in (21) is zero as well. Thus the existence of asymmetric information ‘makes it possible for government to attain a higher value for its objectives through surprise creation in a time consistent equilibrium, The last two terms on the right-hand side of (20) represent the mean (negative) contribution of inflation to governmental objectives. The effect of 23 on G'(-) comes from its effect on A. Since A is monotonically increasing in 3 and all the ‘other elements that affect A are fixed parameters, the choice of a3 is equivalent to a choice of A. Hence the choice of the tightness of control procedures by the ‘government can be expressed formally as (22) max G(A)= max O(@3) where G'(°) is given in (20). If the policymaker chooses perfect control, o3->0 and from (106) 0. At the other extreme when 3c, Ap from below.” Hence the range of choice open for A is from a minimum of zero (which corresponds to perfect control) to a maximum of p (which corresponds to the minimum possible amount of control) A sufficient condition for a positive (politically) optimal level of ambiguity (o> 0) is that G(A) be an increasing function of A at A=0. Using (20) and diflerentiating (22) with respect to A, 2 [e(1=Bp*(1+ 8-28) __Bo(1~ Bp")? BL (1=BpaY—pay? (=p) Bay (=f)? ry-ayrl By manipulating the monetary control parameter, 3, the government affects the speed with which the public becomes aware of changes in governmental objectives and therefore the average value of benefits from surprise creation. In particular an increase in ambiguity increases the mean value of benefits from (3) FA)= pK? For a!-+c0,7->0 which implies A». See alto fotnote 19.THEORY OF AMBIGUITY 7 surprise creation” but it also increases the average rate of bad from government's point of view. The level of monetary ambiguity is chosen by weighting the effects of those two conflicting elements on governmental objectives. ALA the expression in (23) reduces to To ppa+p) Bf E[oa Bo?V(1+8) af ‘The expression in (24) is positive if p>0 and 8 is sufficiently small. Hence a strong degree of time preference on the part of government leads to a choice of institutions which produce loose control of the money supply even if perfect, control is technologically feasible. For a sufficiently high value of 8, G is decreas- ing at A =0, so the politically optimal level of control may be perfect. In other ‘words a negative value of (24) (which obtains for sufficiently high B) isa necessary ‘but not sufficient condition for the optimal value of 73 to be at zero. (24) FCO) At the other extreme when A =p, (23) becomes. p(1+B~26p") [ 2 2, J] Fp) =e B= 280g 8s x . "Gaye La=BoV =p) PLi= aa" “Tap This expression is negative fora sufficiently large P and p bounded away from cone, For this case a sufficiently low degree of time preference on the part of government is necessary for a finite optimal level of ambiguity, o? 0, equation (23) implies that necessary conditions for A*= p,00, F(A*) =0, and F(0)=0 respectively. The following proposi- tion provides a characterization of the three types of solution in terms of the parameters (p, B, and K?) by providing a sufficient condition for each type of solution, PROPOSITION 4: Let ai-v'8)| 1 [Z BBY LI-—* e198 B (1-8) {1 p*(1+6(0-B))], That is, the fist ferm inside the brackets on the righthand side of (23) i positive since 1+ ~28ph">0. The intuition again Periods with high x, ~ s the positive effec of postponing ‘adjustment of expectations on governmental objectives is larger than the negative effect of such 8 postponement when x, is below its mein valve11g ALEX CUKIERMAN AND ALLAN H, MELTZER eB IB RE(~p)0~ p84 8) Bs then for K2= A*/o3, © axes Gi) Rey at (iii) if K?>c, A*=0; (iv) if. cy, the optimal level of ambiguity is zero. This a ssa" ge Fount t 2% However, even when yy and o, are seh that cc, £0 that condition (i) of Proposition 4 not satised, A* may be internal if €)=K*<¢,. Condition (v) is sufficient but not necessary for an Fnternal solution. Fora given nonzero A this is equivalent tothe case 0 (and }=0) discussed in Section 8.‘THEORY OF AMBIGUITY 119 ‘monotonic relationship between the level of instability in governmental objectives land the degree of ambiguity embodied in monetary institutions holds also for internal solutions. Proposition 5 summarizes the effect of larger uncertainty in relative governmental objectives on the precision of monetary institutions chosen by government. Proposition 5: For given values of B and p, i) if 00. Hence the larger of the larger the value of ‘needed to atain the optimal value of r* and A*, This interpretation was suggested by Robert Hetzl1120 ALEX CUKIERMAN AND ALLAN H. MELTZER Proposition 6: For given values of p, B, and K?, the politically optimal level of 3 is a nondecreasing function of «7. Equation (23) implies that two countries with identical values ofthe parameters Band K° will have identical levels of credibility since Ais the same in both. But Proposition 6 implies that the country with the higher will achieve this ‘common credibility level by choosing a higher level of noise in the control of money. ‘A large body of empirical evidence (Okun, 1971; Logue and Willet, 1976; Jaffe and Kleiman, 1977) suggests that the level and the variability of inflation are positively related across countries. Our framework links this positive relation- ship to differences in the relative instability of governmental objectives across countries. More precisely, for a given A, a country with a higher 9 will have a lower K* and, by Proposition 5, a higher a3. Propositions 1 and 3 imply that, ceteris paribus, a higher 07 implies both higher average monetary growth as well as higher monetary variability.”* Hence if political instability varies across coun- tries, this variation will produce a positive relation between average money growth and the variability of money growth. This, in turn, induces a positive relationship ‘between the average level and the variability of inflation across countries. We now consider the effect of the policymaker's discount factor 8 on the optimal level of ambiguity. Since for given 3, p, and A the optimal level of credibility, A*, and the optimal degree of ambiguity as embodied in o are monotonically related, it is enough to find the effect of 8 on A%. For intemal solutions the direction ofthis eflect depends on the sign of the partial derivative of F(A*) (from (23)) with respect to B. The sign is ambiguous in general. For corner solutions at 0 or p, the sign depends on the effect of on F(A) in the range 0 A= p. This sign is also ambiguous. In spite of the ambiguities, however, ‘numerical simulations for a wide range of values of p, B, and K” suggest that in more than 99 per cent of the cases examined A* is weakly decreasing in .” Moreover the few cases in which A* turned out to be strictly increasing in fall occurred for B>0.8. We conclude on the basis of these simulations that, except fora relatively small subset of parameters, the optimal level of ambiguity increases, or at least does not decrease, when the degree of time preference increases ( 0 forall 0p. The Lemma implies (490) F\Q)= Ap(0)~Boe)~ColP)K*= Hs, = A=p, Hence a sufiient condition for AT=> is H,>0 which, using (A30), is equivalent to 3 Keen We prove that e,> by contradiction. Suppose c; c (AD1) implies that A* "0. But since K*0 and Filp) <0. ‘Using (A2S) in (A32) and rearranging, this sulicient condition can be reformulated as (033) < K*¢_ both of those numbers must be betwoen c, and ¢, by contradiction Suppose (A3S) is satisfied but that ei lower than cy Then there exist a K* which satisfies both (ASI) and(A33, implying hat * sinteral and also equal to, whichisa contradiction. Alternatively suppose (ASS) holds and ¢y > e,Then there exist aK tha catisties both (429) and (A33), implying ‘that A* is Both internal and equal tO, which is contradiction, I follows that when cy =e, Both cand cy, must be bounded between ¢, and c. This establishes part(s) of the Proposition, REFERENCES BACKUS, D, AND J. DRIFFILL (1985): “Inflation and Reputation,” American Economic Review, 75, 530-538, BARRO, RJ “Rules, Diseretion and Reputation ina Model of “Monetary Policy." Journal af Monetary Economies, 12, 101-123, (19836): “A Postive Theory of Monetary Policy n'a Natur Rate Model,” Journal ofPlical Economy, 91, 889-610. Bommorr, E. J. (1982): “Predicting the Price Level in a World that Changes All the Time,” Camegie-Rochester Conference Series on Pubic Poiey, 17, 7-57 BRUNNER, KA. CUKIERMAN, AND A. H. MELTZER (1983): “Money and Economic Activity, Tnventores and Business Cycles," Journal of Monetary Economies, 1, 281-320, BRUNNER, K,, AND A. H. MELTZER (1968): The Federal Reerue Attachment to Pree Reseroes Washington, D.C: House Commitee on Banking and Currency Butt, C:"Ritional Expectations, Monetary Data and Cental Bank Watching.” Giornale degli Economist ¢ Anna di Economia 4, 31-40, BURNS, AF. (1999): The Anguish of Ceniral Banking. Belgrade, Yugoslavia: Per Jacobson Foundation. (CANZONERI, M.B. (1985): “Monetary Policy Games and the Role of Private Information,” Economic Review, 15, 1056-1070. CUKIERMAN, A. (1984): Inflation, Stagflation, Relative Prices and Imperfect Information Cambridge ‘Cambridge University Press. CUKIERMAN, A., AND A. H. MELtzen (1986); “The Credibility of Monetary Announcements,” in ‘Monetary Policy and Uncertainty, ed. by M. J. M. Neumann, Baden-Baden: Nomos Vevlagsgesel. chat FELLER, W. (1976): Towards a Reconstruction of Macroeconomics: Problems of Theory and Poy. Washington, D.C: American Enterprise Institute (1979): “The Credibility Eifect and Rational Expectations: Implications of the Gramlich Study,” Brookingr Papers on Economic Activity, 1, 167-178. (980): “The Vad Core of Rationality Hypotheses Inthe Theory of Expectations” Journal ‘af Money, Credit and Banking, 12, 163-787, FiscHe®, 8. (1977):"“Long Term Contracts, Rational Expectations andthe Optimal Money Supply Rule,” Journal of Poca! Economy, 85, 191-206. (1984): "Contracts, Credibility and Disnflaion,” Working Paper No, 1339, Natio ‘of Economic Research Fisch, S, AnD J. Hu1ZiNoA (1982): “Inflation, Unemployment, and Public Opinion Poll “Journal of Money, Credit dnd Banking. V4, 1-19. Fac, B.S, anp F. ScHveIDER (1978): “An Empitical Study of Politico-Economie Interactions in the US." Review of Economics and Stoitiy 0, 174-198, Fxispsan, M. (1960): A Program for Monetary Stability. New York: Fordham University Press Goopraienp, M, (1986): "Monetary Mystique Secrecy and Central Banking,” Journal of Monetary Economics, 17, 63-92. Gooptant,C.A.E,AND RJ. BHANSALI GREEN, B,J, AND R. H. PORTER (1988) “ information,” Econometrica, $2, #7-100. American 51 Bureau “Political Economy. Political Studie, 18, 43-106 ive Collusion under Imperfext Price1128 ALEX CUKIERMAN AND ALLAN H. MELTZER Hanencen, G. (1980): “Notes on Rational and Inational Expectations,” Reprint No. 11, American Enterpase Institute, Reprinted from Wandlungen in Wirwohaft und Gesellscheft: Die Wirtschaft tnd die Socialwisenschaflen vor neuen Aufgaben, ed. by Emil Kung. Tubingen: .C.B. Mobs. [HaRDOUVELIS, G. (1984): "Market Perceptions of Federal Reserve Policy and the Weekly Monetary "Announcements," Joural of Monetary Economics, 18, 225-240 HerzeL, RL, (1984); “The Formulation of Monetary Policy in a Democracy," unpublished ‘manuscript, Federal Reserve Bank of Richmond, JAH¥E, D.M., AND E. KLEDMAN (1977): "The Welfare Implications of Uneven Inflation” in Inflation Theory and Ant-Inflation Policy, ed. by E- Lundberg. London: Macmillan ‘Kane, EJ. (1982): "External Pressure and the Operation of the Fed,” in Political Economy of “International and Domestic Monetary Relations, ed. by R.E. Lombra and W. E. Witte Ames: Towa Univesity Press, 21-282 KYDLAND, F. E, AND E. C. Parscor® (1977): “Rules Rather than Disretion: The Inconsistency ‘of Optimal Plans," Joural of Polical Ecanamy 85, 473-492. Logue, D. E, Nb T. D, WILLET (1976): "A Note on the Relation Between the Rate and the Variability of Inflation,” Economica, 43, 151-158. Lomita, Ry AND M. MORAN (1980): “Policy Advice and Policymaking atthe Federal Reserve,” ‘Monetary Tnstitutins andthe Pole Process; Carnegie Rochester Conference Series on Public Poy, 15, ed. by K. Brunner and A. H. Meltzer, pp. 9-68 Lucas, RE. Jn, (1973): "Some Intemational Evidence on Output Inflation Tradsots,” American ‘Economie Review, 63, 326-335, Meyer, L. H., aNb C. WensteR, JR. (1982): “Monetary Policy and Rational Expectations: A ‘Comparison of Least Squares and Bayesian Learning," Carnegie Rochester Conference on Public Pole, 17, 67-99. Murai, J. F. (1960): “Optimal Properties of Exponentially Weighted Forecasts," Joural of the “American Statistical Asvociation, 35, 299-306, OKUN, AM, (1971): "The Mirage of Steady Anticipated Inflation,” Brookings Papers on Economic Actos, 2, 485-498, SARGENT, TJ. (1979): Macroeconomic Theory, New York: Academic Press ‘TAYLOR, J.B (1980a): “Aggregate Dynamics and Staggered Contacts," Journal of Political Economy, 88, 1-23, 19800); “Recent Developments inthe Theory of Stabilization Policy,” in Stabilization Policies; “Lessons from the Seventies and Implications for the Eighes, ed. by Lawrence Meyer, Center for the Study of American Business, Conference Volume No, 4 St Louis: Federal Reserve Board of St Louis, 1-40. (1982): “Establishing Cred ‘Review, 72, 81-85 11583}: "Comment on Rules, Discretion and Reputation in a Model of Monetary Policy. Journal of Monetary Economics, 13 123-125. ‘Wenrnava, R. (1978): "Congressional Supervision of Monetary Policy,” Jounal of Monetary ‘Economics, 4, 341-382 WoottEy, JT (1984): The Federal Reserse andthe Poles of Monetary Policy. Cambridge: Cambridge ‘University Press A Rational Expectations Viewpoint," American Economic