The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Working Paper Series Expectation Traps and Discretion V.V. Chari, Lawrence J. Christiano and Martin Eichenbaum Working Papers Series Macroeconomic Issues Research Department Federal Reserve Bank of Chicago March 1996 (WP-96-5) LIBRARY FEDERAL RESERVE BANK OF CHICAGO MAY 0 9 1996 FEDERAL RESERVE BANK OF CHICAGO Expectation T r a p s a n d Discretion^ V . V . C hari* L a w r e n c e J. C h ristia n o * * M a r tin E ich en b a u m * * * M a rc h 1 9 9 6 A bstract We argue that discretionary monetary policy exposes the economy to welfare-decreasing instability. It does so by creating the potential for private expectations about the response of monetary policy to ex ogenous shocks to be self-fulfilling. Among the many equilibria that are possible, some have good welfare properties. But, others exhibit welfare decreasing volatility in output and employment. We refer to the latter type of equilibria as e x p e c t a t io n tra p s. In effect, our paper presents a new argument for commitment in monetary policy because commitment eliminates these bad equilibria. We show that full com mitment is not necessary to achieve the best outcome, and that more limited forms of commitment suffice. tW e are grateful for com m ents from B radford DeLong, B ennett M cCallum and Michael W oodford. C h ristian o and E ichenbaum th an k the N ational Science Foundation for finan cial su p p o rt. T h e views expressed in this paper are strictly those of the authors. T hey do not necessarily represent the position of the Federal Reserve Bank of Chicago, or the Federal Reserve System . * University of M innesota and Federal Reserve Bank of M inneapolis ** N orthw estern U niversity, Federal Reserve Banks of Chicago and M inneapolis, and N B ER *** N orthw estern U niversity, Federal Reserve Bank of Chicago, and N BER 1. I n t r o d u c t i o n The classic literature on discretionary monetary policy attributes persistent episodes of infla tion to central bankers’ willingness to purposefully initiate and pursue inflationary policies. This willingness is said to reflect the appeal of inflation as a source of revenue and as a tool for exploiting perceived short-run Phillips curve trade-offs. This paper offers a new argument about why.inflationary episodes can arise under discretion. Our argument is that, under discretion, policymakers can also be pushed into pursuing inflationary policies. This can happen when the private sector, for whatever reason, expects high inflation. Under these circumstances, the central banker may find it optimal to accomodate private agents’ expectations if the cost of not doing so is a recession. When the monetary authority does accommodate, private agents’ expectations are self-fulfilling and we say the economy has fallen into an e x p e c t a t i o n t ra p . 1 The expectation traps emphasized in this paper arise from agents’ expectations about how monetary policy will respond to exogenous shocks. We argue that the notion of a policy maker caught in an expectations trap is potentially helpful in accounting for inflationary episodes like the one endured by the U.S. in the mid 1960s and 1970s. The expectation traps that we analyze are of two types. The first arises when agents ex pect that monetary policy will react to shocks which do not affect preferences or technology. If these expectations about monetary policy are self-fulfilling, non-fundamental shocks con g e e W illiam Fellner (1976, p p .116-118) for an early discussion of th e idea th a t exp ectatio n s a b o u t m on etary policy could be self-fulfilling by forcing th e hands of benevolent policy m akers. 2 stitute an additional set of impulses to the economy. The second type of expectation trap can arise when agents’ expectations of how monetary policy will respond to fundamental shocks are self-fulfilling. This can cause actual monetary policy to overreact to fundamental shocks. Here, expectation traps amplify volatility by modifying the impact and propagation mechanisms from fundamental shocks to the economy. They could, for example, cause a real shock to lead to a persistent change in the inflation rate even though it would only produce a change in the level of prices under a non-accommodative monetary policy. In either case, expectation traps are a source of welfare-decreasing instability in the economy. VVe articulate the preceding arguments formally using a general equilibrium monetary model which is similar to that of Ireland (1995). The model economy is populated by a representative household and three types of firms: a competitive producer of final goods, a continuum of monopolists, each of whom produces an intermediate good, and a finan cial intermediary. The household purchases the final consumption good, supplies labor to intermediate good firms, and lends funds to the financial intermediary. The financial inter mediary combines funds received from the households with lump-sum injections of money from the government, and makes loans to firms. Firms need loans because they must pay labor before they sell their output. There is a monetary authority which chooses monetary transfers to the financial intermediary with the objective of maximizing the expected utility of the representative household. We model discretion by assuming that the policy maker acts sequentially through time and cannot commit to any future action. The sequence of events within a period is as follows. First, all exogenous shocks, including both fundamental and non-fundamental shocks, are realized. Then, intermediate goods producers set their prices based on their expectation of the current-period money growth 3 rate. Finally, the monetary authority’s action is realized, and all other model variables are determined, with the output of the intermediate good being demand determined. As in Kydland and Prescott (1979), Barro and Gordon (1983) and Ireland (1995), our model is one in which high expected money growth is socially costly while unexpected money growth confers social benefits. High anticipated money growth is costly because it leads to higher inflation which acts as a tax on employment. Unexpected money growth raises output in the model because intermediate goods producers post prices before the realization of the monetary policy action and output is demand determined. This rise in output is socially beneficial because output is inefficiently low reflecting the existence of monopoly power in the intermediate goods industry. This inefficiency cannot be eliminated, in equilibrium, by monetary policy. To establish that expectation traps can occur without commitment, we characterize the space of sustainable outcomes for our model economy. We do this by adapting a charac terization theorem due to Chari and Kehoe (1990,1993a, 1993b), based on Abreu (1988). Specifically, we establish a set of restrictions on a given outcome for it to be the outcome of some sustainable equilibrium. We then show, by way of two examples, that both types of expectation traps discussed above are elements of the space of sustainable outcomes. In the first example, there are no fundamental shocks. Money growth follows a first order au toregression, with the disturbance term determined by a non-fundamental shock. We find this example interesting for two reasons. First, it demonstrates in a particularly simple way, the possibility of instability under discretion. Second, it offers an alternative perspective on the empirical literature on monetary policy shocks. This literature seeks to identify an exogenous component in monetary policy. The conventional interpretation of these shocks 4 is that they represent exogenous randomness inherent within the policy making process. Our results suggest that this randomness may also originate from outside the policy making process and reflect the sort of self-fulfilling expectations that can arise under discretion. ‘ In our second example of an expectations trap, there are only fundamental shocks. We show that there is a sustainable outcome in which money growth is typically low. Occasion ally, following a decline in the state of technology, money growth and, hence, inflation rise for an extended period of time before returning to their typically low levels. We find this example interesting because it illustrates another sense in which discretion makes possible volatile equilibria. Also, the example may form the basis for a simple explanation of the prolonged rise in inflation starting in the mid-1960’s. Specifically, it articulates the notion that the US fell into an expectation trap in which a benevolent, sequentially optimizing monetary authority pursued a policy which exacerbated the loss in output caused by various fundamental shocks. Standard economic theory is at a loss to explain how a real shock can trigger a long term rise in inflation. It is also at a loss to explain why the monetary authority would have responded to such shocks by shifting for a period of years to a higher money growth regime which most contemporary observers deplored as being destructive to the economy. The literature does contain discussions about the link between real shocks and long lived episodes of inflation. Many of these discussions involve ideas related to those lying at the heart of our second example of an expectation trap. For example, these ideas play an important role in discussions of the wage price spiral, as well as the notion that an ‘overheating’ economy, i.e. one with high sustained rates of capacity utilization, can trigger a rise in the core inflation rate. They are also implicit in Blinder’s (1982) discussion of how 5 supply shocks could have contributed to the persistent inflation of 1970s: ‘Inflation from special factors can “get into” the baseline rate if it causes an accel eration of wage growth. At this point policymakers face an agonizing choice—the so-called accommodation issue. To the extent that aggregate nominal demand is n o t expanded to accommodate the higher wages and prices, unemployment and slack capacity will result. There will be a recession. On the other hand, to the extent that aggregate demand is expanded (say, by raising the growth rate of money above previous targets), inflation from the special factor will get built into the baseline rate.’ (Blinder (1982, p. 264)) These ideas are not confined to academics. Our reading of Arthur Burns’ speeches sug gests that his views about the genesis of the inflation of the mid 1960’s and 1970’s have much in common with the notion that the economy was caught in an expectation trap, triggered by transitory real shocks. The principal real shock, in his view, was the expansionary fiscal policy (‘the forces of excess demand’) associated with the Great Society programs of the 1960s and the Vietnam war.2 As Burns put it around the time he became Chairman of the Federal Reserve in 1970: ‘The forces of excess demand that originally led to price inflation disappeared well over a year ago. Nevertheless, strong and stubborn inflationary forces, emanating from rising costs, linger on...’ (Burns (1978, p.124).) A key factor fueling rising costs, according to Burns, was the widespread expectation that the inflation which started in the mid 1960s, would continue. He described his vision of how inflationary expectations drive up costs, and hence prices, in a statement before the Joint Economic Committee of the U.S. Congress in 1971: 2 B urns believed th a t crop failures and oil shocks were o th er real d isturbances th a t co n trib u ted to th e inflation of th e 1970s 6 ‘Consumer prices have been rising steadily since 1965 - much of the time at an accelerating rate. Continued substantial increases are now widely anticipated over the months and years ahead...in this environment, workers naturally seek wage increases sufficiently large...to get some protection against future price ad vances... thoughtful employers...reckon, as they now generally do, that cost in creases probably can be passed on to buyers grown accustomed to inflation.’ (Burns (1978, p.126.) Like the central banker in our model, Burns clearly understood that this upward pressure on prices could not be transformed into persistent, high inflation without monetary accom modation. As he put it in a speech in 1977: ‘Neither I nor, I believe, any of my associates would quarrel with the proposition that money creation and inflation are closely linked and that serious inflation could not long proceed without monetary nourishment. We well know-as do many others-that if the Federal Reserve stopped creating new money, or if this activity were slowed drastically, inflation would soon either come to an end or be substantially checked.’ (Burns (1978, p. 417).) Burns chose to accommodate agents’ inflationary expectations for the same reason that the monetary authority in our second example does so: to do otherwise would generate a costly recession. As he put in testimony before the Committee on Banking and Currency of the House of Representatives on July 30, 1974: ‘...an effort to use harsh policies of monetary restraint to offset the exceptionally powerful inflationary forces in recent years would have caused serious financial disorder and economic dislocation. That would not have been a sensible course for monetary policy.’ (Burns (1978, p. 171).) Burns clearly understood the accommodation dilemna alluded to by Blinder, and judged that the best response was to accommodate. In our terminology, he acted like a benevolent 7 central banker caught in an expectation trap. What Burns did not understand was the critical role the institution of discretion played in setting that trap. After our formal analysis of discretionary monetary policy, we investigate alternative in stitutional arrangements that can eliminate the possibility of expectation traps. One solution is full commitment on the part of the monetary authority. An obvious practical problem with this solution is that it is hard to imagine any monetary policy authority committing to a sequence of policy actions infinitely far into the future. More limited forms of commitment are practical. We consider a situation in which the monetary policy authority commits at time t to its time t 4- 1 state contingent growth rate of money. We establish that expecta tion traps cannot occur in this regime and that the equilibrium quantities in this economy coincide with those in the Ramsey equilibrium. The result that one period ahead limited commitment reduces the set of sustainable equilibria to a singleton - the Ramsey equilibrium - depends on particular features of our model economy. The intuition underlying the proof gives some insight into how it could be extended to accommodate different environments, say, those with multiperiod wage and price contracts. The next section presents a formal description of our model. Section 3 discusses equilib rium under full commitment. Section 4 characterizes the set of sustainable outcomes under discretion, and section 5 presents our two examples. In section 6 we discuss equilibrium under limited commitment. Finally, we offer some concluding remarks. 8 2. T h e M o d e l We consider an infinite horizon, monetary economy with uncertainty. In each period, t = 0,1,2,..., the economy experiences an exogenously determined event, s{, drawn from a finite set. Let t. s l = ( s 0) s i , ..., s t) denote the history of exogenous events up to and including time The probability of s l is given by /x(s‘). We denote the probability of s‘ conditional on sr by fi(sl | s r ). The monetary authority chooses an action, financial intermediaries) in period of exogenous events up to date that private agents view At t, t = 0,1,2,.... Let = t. denote the history Throughout we assume X t { h t- \ , s t ). h t + 1 , h t+ 2.... ( h t , s t + \, X t ( h t , s t+ i ) ) , h t+2 — is induced by ATt(*, •) from X ( h t~1, s t) on t, and X(/ir_ 1 , s r ) , r ^ h t ~i ht = st , A t) (/it+i,St+2,X t+i(/i(+i,St+2)), ... . To conserve on notation, from here on we delete the subscript, the number of elements in h t = ( s l , A o , ..., A t ) as being generated according to the policy rule We say that a set of histories that the functions (a lump sum monetary transfer to and of policy actions up to time At if h t+! = At t X. This should not cause confusion: are different is evident from the fact that and hr_ 1 are different. We adopt this notational convention for all functions of histories. The commodity space in this economy consists of history contingent functions. That is, allocations, prices and policy actions are expressed as functions of history. In standard Arrow-Debreu economies, rules of this type are a function only of the history of exogenous events. The extension considered here is necessary to accommodate the fact that the gov ernment in our model optimizes sequentially and that, being a ‘large’ agent, its actions have 9 a non-trivial impact on private allocations and prices. At every date the government con siders the consequences of all the current and future state-contingent actions that it could possibly take, and selects the set of actions that it prefers. Since its objective is a function of current and future allocations, these must be well-defined for every possible history for the government problem to be well-posed. Histories, ht, can exclude past household actions because, though individual households optimize sequentially like the government, they are small and have no impact on aggregate allocations and prices (see Chari and Kehoe (1990) for a further discussion.) To reiterate from the introduction, the sequence of events within a period is as follows. First, the exogenous event, st, is realized. Then, intermediate goods producers set the prices of their goods. Finally, the monetary authority’s action is realized and all other model variables are determined. We now describe the problems of the agents in our economy in detail. 2.1. F irm s F in a l G ood F i r m s The final good, c ( h t ), is produced by a perfectly competitive firm that combines a con tinuum of intermediate goods, indexed by '(ht) = Here 0 < A < 1 and denote the time t yi(ht) i € (0,1), using the following technology: (2.1) [jT (y i ( h t ) ) Xd i denotes the time t price of intermediate good input of intermediate good i. 10 i. Let P i(h t- \ , s t) This price is not a function of the time t policy action because intermediate goods producers set their prices before A t is realized. The final good producer’s problem is: max P { h t- i , s t ) c ( h t ) c(M,{y<(/n)} - f Jo subject to (2.1). Here, P(/it_ i,s t) denotes the time below why this price is not a function of A t. (2.2) P i ( h t - i , s t ) y i ( h t )di, t price of the final good. We explain Problem (2.2) gives rise to the following input demand functions: P ( h t- i , s t) \ Vi(ht) = c ( h t ) (2.3) P i { h t- l t s t ) ) In conjunction with (2.1), this implies: 1-A X $t) — | which expresses the time t (2.4) P %{ h t - U s t y ^ d i price of the final good as a function of the time t prices of the intermediate goods. I n t e r m e d i a t e G ood F i r m Intermediate good i is produced by a monopolist using the following technology: Vi{h t) = (2.5) 9 ( s t ) n i ( h t ). Here U i ( h t ) denotes time t labor used to produce the i th intermediate good and the stochastic economy wide level of technology at time before the date t government policy action, A t, 11 t. Recall that 0 (sl) P i(h t- i , s t) denotes is chosen is realized. We assume that the producer must supply all of the goods demanded by the final goods producers as determined by (2.3). Thus, at time t, producer i’s problem, which we refer to as max P r o b l e m F, is to maximize profits: S t)yi (ht ) — W ( h t ) R ( h t ) n i ( h t ), (2.6) Pi(ht-i,St) subject to (2.3), (2.5) and taking as given the time t wage rate, terest rate, R ( h t ), as well as P (h t_i, s t ), and c ( h t ). W ( h t ), the gross nominal in The firm’s unit labor costs are W ( h t) R (h t) because the firms need to pay workers before production, and must borrow these funds from the financial intermediary at interest rate R ( h t ). Note that from the firm’s perspective (2.6) is a function of the information set (h£_ i,s (). This is because they view ( h t - u s t , X ( h t- At as being generated according to At = St) and ht = 1 , s t )). Financial Interm ediary A perfectly competitive financial intermediary receives deposits, and a transfer, A t, I ( h t ), from households, from the government. These funds are lent to intermediate good producers at the gross interest rate R ( h t ). At the end of the period, the intermediary pays to households in return for deposits, and distributes R (h t) A t R ( h t)I(h t) to households in the form of profits. 2.2. Household At time r the representative household’s expected present discounted utility is given by: OO I s T) U { c { h t ) , n ( h t ) ) . t —r sl 12 (2.7) Here n (h t) denotes total hours time U (c,n ) where tp > 0 and a t hours worked, and = [c(l - n )*](1 ' * / ( ! - a > —1 . The household faces the following cash constraint on its purchases: P ( h t- U s t)c{ht) < W ( h t) n (h t) Here (2.8) ), + A/ (h t_1) - denotes the household’s end-of-period t (2.9) I ( h t ). —1 holdings of cash. The household’s money holdings evolves according to: M ( h t) = —I ( h t ) W ( h t)n(ht) + (2.10) — P ( h t- i , s t )c(h t ) (2.11) + R { h t ) [ I ( h t ) + A t } + D ( h t ), where D ( h t) denotes total time t profits from the intermediate goods producers. Let Z ( h t ) = [c(ht ) , n ( h t ) , I ( h t ), M ( h t )}. The household’s problem, which we refer to as contingency plan for M ( h t- i ) . so that hr+i = is then to choose a non-negative to maximize (2.7) subject to (2.8)-(2.10), In solving its problem, household views X ( h t - i , s t ), Z ( h t) P r o b l e m H, ( h r , sr+1, X ( h r , s r + i ) ) 13 At n ( h t) < 1 and I ( h t) < as being generated by the policy rule for all r. 2 .3 . G o v e r n m e n t At time t, given history s t ), the government chooses a current action, At, and a se quence of state contingent policy rules for each future period, to maximize the representative household’s utility, (2.7). It takes as given, M(/ir-i) ries will be induced by its future policy rules from takes as given the household’s contingency plan, + A r > 0 , r > t, and that future histo h t = ( h t- i , s t , A t ). The government also { Z ( h T) ; r > t}. We refer to this problem as P r o b l e m G. 2.4. S u stainable E quilibrium In order to characterize the set of equilibrium outcomes, it is useful to define a sustainable equilibrium for our model economy. D efinition 2 . 1 . A s u s t a i n a b l e equilibrium is a col lect io n o f h i s t o r y - c o n t i n g e n t a ll o c a ti o n rules Z ( h t ), p r i c i n g f u n ct io n s, s t ), P ( h t_ i , s t), W ( h t ), R ( h t ), a n d a g o v e r n m e n t p o l i c y rule, X ( h t - i , s t ), s uc h t h a t : 1 . for all histories, h t , the allocation rules Z ( h t ), solve P r o b l e m H, 2 . for all histories, ( h t - i , s t ), the pricing function P i ( h t _ i , s t ) solves P r o b l e m F, 3. for all histories, (/it_i, s t ), producers of the final good make zero profits so that P(/it_ j , s t ) satisfies (2.4),45 4. for all histories, h t , the goods market clears, i.e., c ( h t ) = 6 ( s t ) n ( h t ), the loan market clears, i.e., W ( h t ) n ( h t ) = /(/it) + At, and the money market clears, i.e., M ( h t ) = M (h t-i) + A t. 5. for all histories, (/i£_i,St), the policy rule, X(/it_ i,s t) solves To proceed we find it convenient to scale all time t —1 t P r o b l e m G. nominal variables by the end of time stock of money, M(/it_i). Let Pt (/it - 1; &t) R j ( h t - i , St) M ( h t_,) p(/it-i,St) P ( h t- i , s t) M(/it-x) ’ 14 d(h t) D ( h t) M(/it_i), W ( h t) w ( h t) I{h t) M ( h t^ y [ t] In addition, we define the variable x ( h t- i , s t ) M ( h t- i ) m ( h t_i) of money to the aggregate stock of time X ( h t. u s t ) At M ( h t_i) M (h ,. as the ratio of the household’s time t — 1 money. Of course in equilibrium O’ t —1 stock m ( h t ^ i) = 1. Finally, let z ( h t) = (2.12) [ c ( h t ) , n ( h t ) , i ( h t ), m ( h t )}. In what follows, we proceed in terms of these lower case scaled variables. 3. P r i v a t e S e c t o r E q u i l i b r i u m U n d e r C o m m i t m e n t A key objective of our paper is to characterize the set of allocations that are the outcome of some sustainable equilibrium in the model economy. To do this it is useful to first analyze the behavior of the economy when the government can commit to a particular policy rule. Specifically we consider the case of commitment, by which we mean that the government never deviates from the policy rule, possible date t x ( h t _ i , s t ). histories can be indexed by sl Under these circumstances, the set of all alone. This is because, under commitment, ht can be expressed as a function of s 4 using the following recursion: h-i = 0, ho = (s°,x(/i_ 1 ,s0)), and similarly for h i = ( s \ x ( h - i , s 0) , x { x ( h - i , s 0) , s i ) ) , h 2 , ^ 3 ,..., h t . Let z*(se) = [c*^4), n * ( s l ), m*(st)] denote the state contingent allocations under commitment. In addition, let p?(sf), p*(sf) ], 15 x*(sl) and d * ( s f) denote the pricing functions, government policy rule, and intermediate good firm profit function under commitment, respectively. D efinition 3.1. A p r i v a t e s e c t o r e q u i l i b r i u m is a c o ll ec t io n o f s t a t e c o n t i n g e n t a l l o ca ti o n rules, z * ( s t ), p r i c i n g f u nc t io n s , [^ ‘(s4), R * ( s l ), p*(s4), p*(s4)], a n d a g o v e r n m e n t p o l i c y rule, x*(s4), s uc h t h a t ( 1) t h e a l l o c a t i o n rules z * ( s t ) s o l v e t h e h o u s e h o l d ’s p r o b l e m , (2) the p r i c i n g f u n c t i o n p ' ( s l ) s o l v e s t h e i n t e r m e d i a t e g o o d p r o d u c e r s ’ p r o b l e m , (3) p r o d u c e r s o f t h e final g o o d m a k e z e r o p r o f it s, ( 4) t h e g o o d s m a r k e t clears, i.e., c*(s4) = 0(s4)n*(s4), t h e l o an m a r k e t clears, i.e., tu*(s4)n*(s4) = i*(s4) + x*(s4), a n d t h e m o n e y m a r k e t clears, i.e., m * ( s t ) = 1. Note that a private sector equilibrium satisfies all the conditions for a sustainable equilib rium, with two exceptions. First, it does not require optimality by the government. Second, private sector allocation rules and pricing functions solve the private sector optimization problems and satisfy market clearing only for histories induced by x*(s4). We now characterize the private sector equilibrium. The first order condition of the intermediary goods firm’s problem is: A0 { s l ) In symmetric equilibria, p*(s4) = iu*(s4) c V )V ~ A y*(s£)y c*(s 4) for all A0(s4) i, R - ( S !). p*(sl) so that, w ' j s 1) R ' i s 1). P m( s l ) The Lagrangian representation of the household’s problem at time 0 is: t= 0 £ M stl5°) W CV ) ’ * V ) ) 5* -t-i^s4) —i*(s4) 4- u>*(s4)n*(s4) —p*(s4)c*(s4)j + r ] (s4)[(u>*(s4)n*(s4) 4- m*(s4-1) —i*(s4) —p*(s4)c*(s4)) 16 (3.1) ■ft*(s£)(i*(s£) + where i'(st),r j(s t) x ( s t)) + d*(s£) —(1 4- x(s£))m*(s£)]} are the non-negative Lagrange multipliers on the cash constraint and the budget constraints, respectively. In addition to the budget and cash constraints, the first order conditions for c*(s£), n*(s£), m*(s£) and i*(s£) are: uc(s£) = (i/(5£) + r/(s£))p*(s£), «n(s£) = - M s‘) + ) H s l) = 0 (3.2) v i s ^ w ' i s 1), £ M st+V )*?(s£+1)fl’(s£+l)> 5‘+lis‘ i/(s£) + rj(s£) = J?(s£)f?*(s£), (3.3) (3.4) (3.5) where 5£+1|s£denotes the set of time £+1 histories of exogenous events s£+1 that are consistent with s£. Also, uc(s£) and u n(s l) denote the marginal utility of consumption and labor at date £. The transversality condition for the household problem is Jim = 0. r_,°° yr (3.6) P KS1 ) These conditions can be simplified as follows. From (3.2) and (3.3), we obtain: ^ ( gt) = i p c ' j s 1) p * ( s t) 1 —n*(st) ’ 17 (3?) according to which the household equates the real time t wage rate to the marginal rate of substitution between consumption and leisure. Non-negativity of the Lagrange multipliers and (3.5) imply > 1. (3.8) Relations (3.2), (3.4) and (3.5) imply u c(s t ) ■n*(ot\ n R (3*) v -" _l_ W e t \ ^ / t+ i | t \ Uc(gt+1) I 'r^ot+lA ^ (3.9) According to (3.9), the household equates the marginal utility of spending an additional dol lar on consumption to the expected marginal utility associated with investing an additional dollar with the financial intermediary. Recall that loan market clearing requires iu*(s£)n*(s£) = i*{sl) + x*(s£). (3.10) Substituting (3.10) into the scaled version of the household’s cash constraint, (2.9) yields: p’(s£)c*(s£) = 1 -I-x*(s£). (3-11) c*(s£) = 0(s£)n*(s£). (3.12) Goods market clearing implies: Using our assumed functional form for the utility function, equations (3.1), (3.9)-(3.12) 18 reduce to: Y 5t+1|s* M 5t+1lst)v (n *(st) .n *(st+1) . ^ ) . ^ ( s ‘+1).^*(st+1)) = 0, (3.13) where v(n>n'l0,0'>x') = Y Z ^ N 1 - n) 1 1_<T - (3-14) Finally, substituting (3.11) and (3.12) into the transversality condition, (3.6), we obtain: [0(sTK ( s T)(i - n ’(s7'))v’] 1 " = o. (3.15) ST We are now in a position to characterize a private sector equilibrium. P ro p o sitio n 3.2. T h e a l l o c a t i o n r ul e n*(s£) is p a r t o f a p r i v a t e s e c t o r e q u i l i b r i u m i f a n d o n l y i f t h e f o l l o w i n g fi ve c o n d i t i o n s a r e s atis fie d: n* (s£) > 0, A 1 - n * ( s £) > l i/> n*(s£) ~ (3.13 ), (3.15), (3.16) and n*(s£) x*(s£) 1 —n*(s£) — l+ x * ( s £)' (3.17) P ro o f. Suppose n*(s£) satisfies the five conditions. Then we can construct the remaining objects in a private sector equilibrium as follows. The consumption allocation rule is c*(s£) = 9 ( s t ) n ' ( s t ), the price rule is obtained from the cash constraint, (3.11). The nominal wage is obtained from (3.7), the nominal interest rate from (3.1) and, finally, i*(s£) is obtained from (3.10). The only conditions that remain to be checked are (3.8) and the constraint on the household’s problem that 0 < i*(s£) < 1. The first is verified by noting that the expression to the left of the inequality in (3.16) is the equilibrium rate of interest. To verify the constraints on i*(s£) notice that i*(s£) = (1 + x * (s£)) —x * ( s l ). The requirement that i*(s£) < 1 is guaranteed by (3.16) and the fact, A < 1. Inequality (3.17) guarantees that 0 < Suppose we have an equilibrium. The proof follows immediately from necessity of (3.13), (3.15), i?*(s£) > 1, and 0 < i * ( s l ) < 1. ■ In general, the set of private sector equilibria under commitment in our economy is not 19 unique. The expectation traps that we focus on in this paper have nothing to do with this form of multiplicity. Consequently, it is useful to proceed under assumptions which guarantee the uniqueness of a private sector equilibrium under commitment. One such assumption is a — 1, in which case, equation (3.13) reduces to n*(st) 1 = / ? X > ( st+V ) 1 + £*(st+1) A 1 —n*(s‘) S*+1 xp (3.18) Consequently, if an equilibrium exists, then it is unique. This equilibrium has standard properties: employment is increasing in A, i.e., decreasing in the degree of market power; employment is decreasing in money growth; and it is decreasing in xp, which governs the marginal utility of leisure. To see the potential for non-uniqueness when £*(s£) = x. a ^ 1, consider the case 9 (sl) = 1 and In this case, there exists an equilibrium in which employment is given by the deterministic analog of (3.18). However, there may be other equilibria as well, even sunspot equilibria (see Mathene (1995) and Woodford (1986).) To eliminate these possibilities, hence forth we restrict ourselves to the case a — l.3 3T o see how su n sp o t equilibria may arise w ith (J / 1, consider th e case in which th e m onetary grow th ra te is a c o n sta n t, equal to x, a n d 0 = 1 . T hen (3.13) requires v = 0. T his im plicitly defines n ' as a function (possibly a correspondence) from n. T h e im plicit function theorem guarantees th a t th ere exists a differentiable function relatin g n ' to n in a neighborhood of stead y -state: n = n' = \ 3 /[xp{\ + x)], with: dn' dn (1 - a) Here, dn'/dn = 0.77 w hen a = 7, A = 0.87, 0 = 3, 0 = 1/1.03. T h e fact th a t th is derivative is less th a n one m akes it possible to c o n stru c t m ultiple determ inistic equilibria an d su n sp o t equilibria. 20 3.1. T h e R am sey E quilibrium As a benchmark, it is useful to consider the equilibrium of the model under the assumption that the government had access to a commitment technology at t = 0. To this end we now define a Ramsey equilibrium. D efinition 3.3. A R a m s e y e q u i l i b r i u m is a c ol le ct io n o f a l l o c a t i o n r ul es f o r p r i v a t e a g e n t s , z ' i s 1), p r i c i n g f u n c t i o n s [to*(s£), R*(s£), p*(s£), p*(s£)], and a g o v e r n m e n t p o l i c y rule, x * ( s l ), t h a t y i e l d s t h e h i g h e s t u t i l i t y f o r h o u s e h o l d s o v e r t h e s e t o f p r i v a t e s e c t o r equ il ib r ia . P ro p o sitio n 3.4. T h e R a m s e y e q u i l i b r i u m s atis fie s w , (st) P*(s£) = A0(s£), n V ) X > ( s £+V ) 1 1 + £*(st+1) A V>+ A’ * V ) = 1, I A> /?’ — 1 + £*(s£) (3.19) (3.20) P ro o f. From Proposition 1, the Ramsey equilibrium maximizes discounted utility subject to n*(s£) > 0, (3.13), (3.15), (3.16), and (3.17). With o — 1, equation (3.13) reduces to (3.18). It follows from this that solving the Ramsey problem is equivalent to solving a sequence of static problems subject to the constraints (3.15) - (3.17). The first expression in (3.19) follows from the first order condition of intermediate good firms. Condition (3.16), which requires a positive interest rate, implies rc V ) < , A 1 —n * ( s l ) ~ $ (3.21) But as long as (3.21) is satisfied, the period utility function is strictly increasing in n*(s£). Because the left side of (3.18) is increasing in n*(s£), the solution to the Ramsey problem is to set n*(s£) so that (3.21) holds with equality. This establishes that R * ( s l ) = 1, n * ( s l ) = and that the right hand side in (3.18) equals one. This last result implies the first expression in (3.20). Since n*(s{) is constant, the transversality condition is satisfied. Finally the last expression in (3.20) follows by substituting (3.16) with equality into (3.17). ■ Notice that even absent uncertainty in technology, there are many rules for x ' ( s t) that satisfy (3.20), which only requires that the time t conditional expectation of 1/ [1 + x*(s£+1)] be equal to 1/j3. So a constant growth rate of money equal to 21 0 — 1 is sufficient but not necessary to implement the Friedman rule. Finally, note that even in the Ramsey equilibrium, as long as there is monopoly (A < 1), employment is less than the social optimum, 1/(1 + tp)- The monetary authority would like to increase employment beyond the Ramsey, but this is not achievable by monetary policy because of the non-negativity constraint on R ( s l ). 4. C h a r a c t e r i z i n g S u s t a i n a b l e O u t c o m e s We now turn to the task of characterizing the set of sustainable outcomes for our model economy. To do this we must allow for allocation rules, pricing functions and government policy rules to be contingent on both the history of exogenous events and the history of policy actions. Throughout we impose the restriction that x(h(_ i,st) < x for all (/it_ i, 5*), i.e. we impose an upper bound on the growth rate of the money supply. See the conclusion for an extended discussion of this assumption. We begin by characterizing the worst sustainable equilibrium which we refer to as the ‘high inflation’ equilibrium. Consider the following candidate sustainable equilibrium in which the government policy rule is given by xt denote the actual time t , s t) = policy action. For all histories x for all possible (/ie_ i, s4). Let ( h t _ i , s l ), the candidate pricing functions are, P i ( h t - i St) = p { h t - \ , s t) = 1+ x 0(5*) ip( 1 -I- x ) 4* XP (4.1) XP These candidate pricing functions reflect our assumptions that (i) firms set prices in period t prior to the realization of the period t policy action, and (ii) they do so under the expectation that the policy action will be x ( h t , s t+ 1 ) = x. 22 The candidate allocation rules, interest rate and wage functions are c(ht) (1 + x t) c(h t) n{ht) *(«*) ’ P ( h t- i , s t) ’ w(ht) p {h t. l , s t)i}9{st)n {h t) = 1- n (h t) , R (h t) = 1 - (1 + i ( h t) = A (1 ip x t) n ( h t )) n (h t) 1- ’ ipnjht) 1- n (h t) (4.2) (4.3) respectively. By construction, these candidate rules satisfy private agent optimality and market clear ing for all possible histories, ht. So to establish that the candidate equilibrium is sustainable, we need only establish that it is consistent with optimization on the part of the government. Consider a one-shot deviation at period t , given some history, ( h t _ i ,s 4). Given our assumed upper bound on money growth, the only feasible one-shot deviation is for the government to set money growth at xt < x , and then to return to x thereafter. From (4.2), this reduces cur rent consumption and employment. Since utility is strictly increasing in levels of equilibrium employment below 1 /( 1 + ^ ) , this reduces current utility. Future outcomes are unaffected. Thus, the government has no incentive to pursue a one-shot deviation for any history. Since this includes histories in which there have been any number of deviations, it follows that no deviation raises welfare (see Abreu (1988) and Whittle (1983).) This establishes the sustainability of the candidate equilibrium. To see that the candidate equilibrium is the worst sustainable equilibrium, note that in any other equilibrium, employment at some date is necessarily higher because of (3.18). Thus, utility must be higher too. This establishes the following proposition: P ro p o sitio n 4.1. T h e h i g h in fl at io n e q u i l i b r i u m is t h e w o r s t s u s t a i n a b l e e q u i l i b r i u m . We denote by ud(s£) the highest one period utility level associated with a deviation by 23 the government: l + x 1+ x \ - . _________ ] ryiov 1/ If ..... -i<i<s \^p*(s£) ’ 6 ( s t ) p * ( s t ) J ' Let x d( s l ) denote the value of x that achieves the optimum. In addition, let u(s£) denote the one period utility level in the high inflation equilibrium. We are now ready to state our main result. The following proposition establishes a set of restrictions on allocations and prices which are necessary and sufficient for them to be the outcomes of some sustainable equilibrium. In what follows, it is useful to define a class of policy rules characterized by a particular trigger strategy, a grim trigger. Such a rule specifies that in all histories in which there has been some deviation, x(/it_ i,s t) = x. such a policy rule as a g r i m t ri g g e r equilibrium. P ro p o sitio n 4.2. z*(s£), u;*(s£), Let R " ( s t ), We refer to an equilibrium associated with p*(s£), p’(s£) be an a rb itra ry set o f allocations a n d p r ices . L e t x , ( s t) b e an a r b i t r a r y s e q u e n c e o f g o v e r n m e n t pol ic ie s. T h e n , f?*(s£), P i ( s l ), p*(s£), x*(s£)} is t h e o u t c o m e o f s o m e s u s t a i n a b l e e q u i l i b r i u m {z*(s£), w * ( s l ), i f and o n l y if: 1. ( z m( s t ) , w * ( s t ), R * ( s t ) , p * ( s t ) , p * ( s t ) , x * ( s t )) is a p r i v a t e s e c t o r e q u i l i b r iu m ; 2. (z , ( s t ) , w m( s t ) , R * ( s t ) , p ' ( s t ) , p , ( s t ) , x * ( s t)) OO E r ' E ftrM r —t for all s£, all t s atis fie s OO s ‘) U ( c - ( s ' ) , n ‘ ( s ' ) ) > «V ) + I s') fl(sr) Y. r=t + 1 (4.4) s=r = 0 ,1 ,2 ...... P roof. We first consider sufficiency. Suppose the allocations and prices satisfy conditions 1 and 2. We construct a particular grim trigger sustainable equilibrium which supports these outcomes. Consider the following candidate equilibrium. For all histories with no deviation, let x(/it_ !,st) = x*(s£), let p*(/i£_x,st) =P,*(st), p i h ^ ^ S t ) = p * ( s l ), w ( h t ) = w*(s£), R { h t ) = R , ( s t ) and let z ( h t ) = z * ( s t ). For all other histories, let x(/it_ i,s t) = x , and let Pi(ht_ i,s t), p(/it_ i,s t) be defined by (4.1). Also, let w ( h t ), R ( h t) and z ( h t ) be defined by (2.12) and (4.1)-(4.2) with x t replaced by x . For all histories with no deviation, these constitute a private sector equilibrium by assumption. For all other histories, they constitute a private sector equilibrium by the discussion leading up to Proposition 4.1. To show government optimality, note that since the right hand side of (4.4) is the discounted utility associated 24 with a deviation by the government, no deviation can raise welfare. Thus, the candidate equilibrium is a sustainable equilibrium, which establishes sufficiency. We now consider necessity. Suppose (z ( h t ), w ( h t ), R ( h t ) , p i ( h t ^ u s t ) , p ( h t ^ 1, s t ), x ( h t ^ u s t )) is a sustainable equilibrium with outcomes (z * ( s t ), w * ( s t ), R * ( s l ), pt*(s£), p * ( s l ), x * ( s t )). Condition 1 is satisfied by the definition of a sustainable equilibrium. We establish condi tion 2 by contradiction. Suppose 2 is violated for some s l . Consider the following devia tion: the government sets x t = x d ( s t ), and specifies future policies according to x ( h r _ 1, s r ), r > t. The expected present discounted value of utility under this deviation is u d ( s l ) + /?r-t 12s=r At(5’’ I s t ) u ( s r ), where u ( s r) denotes utility in state s r . From the discussion preceding Proposition 4.1, u ( s r ) is a lower bound on period utility when policy is set ac cording to x ( h r - i , s r ). It follows that u ( s r ) > u(sr). Therefore, the return associated with the deviation is greater than, or equal to, the right hand side of equation (4.4). Given our supposition that condition 2 is violated, this means that the proposed deviation will be imple mented at s l . Consequently, the equilibrium is not consistent with government optimization. This establishes the contradiction. ■ We conclude this section with a brief discussion of the proposition. The proposition says that a particular outcome is sustainable if, and only if, it is the outcome of the associated grim trigger equilibrium. The proposition does only not say that a particular outcome is sustainable by a grim trigger equilibrium. In general, an outcome may be sustainable by a variety of equilibria. For example, the equilibria could involve less extreme trigger strategies, or no trigger strategy at all. The proposition is silent on the nature of the equilibria that support a given outcome. It only provides conditions under which a particular outcome is sustainable by some equilibrium. In the next section, we make these observations concrete through a series of examples. 5. E x a m p l e s This section reports two examples which illustrate the types of expectation traps discussed in the introduction. In both cases, we establish that the expectation trap outcome is the out come of some sustainable equilibrium. Our first example illustrates the type of expectation 25 trap that can arise when agents expect that monetary policy will react to non-fundamental shocks. In addition, we discuss three equilibria that can sustain this outcome. In the sec ond example, agents expect monetary policy to react to technology shocks in a particular way. The example is constructed to articulate in a stylized way an interpretation of the US inflation experience since the mid-1960s. E x a m p l e 1: N o n - F u n d a m e n t a l Sh ocks The parameter values for the model economy are given by xb — 3, A = 0.95, P = 1/1.03, x = 0.30. In each period there is a shock, s 9t = 1, a = 1, € (s(l), 5(2)}, which is drawn from a highly persistent, symmetric, two state Markov chain, where the probability of switching states is 0.10. The shock is non-fundamental, in that it has no impact on preferences or technology. Still, there is a sustainable outcome, in which money growth responds to When s = s(l) the money growth rate is 0 percent, and when money is 3 percent. The Wold representation for Xt where et xt s s. = 5(2), the growth rate of is given by —0.15 4- .9xt_i + €t, (5.1) is uncorrelated with past variables, and has standard deviation 0.0092. Several features of this example are worth emphasizing. First, the model is observationally equivalent to one in which policy is set according to (5.1). In this respect, the model is formally identical to a standard, monetized real business cycle model (see, for example, the ‘cash-in-advance’ model in Christiano and Eichenbaum (1996)). The interpretation of the policy shock, however, is very different. In the literature, e( is assumed to reflect the effects of fundamental disturbances to the policy making process, e.g., shocks to preferences 26 of policy makers. In our environment, the shocks originate outside the policy making process and reflect the non-fundamental disturbances that can generate an expectation trap. Second, consistent with the previous observations, the dynamic response of real variables to changes in xt resemble those in standard monetary real business cycle models. For exam ple, Table 1 indicates that employment, consumption and beginning-of-period real balances (M /P ) are low when xt is high, while Rt and inflation are high. Third, Table 1 contains information relevant for verifying that the outcome in this exam ple is sustainable. The table reports, for both states of the world, the current period utility, it, of the household and the expected present value of its utility from tomorrow on, v, along the equilibrium path. In addition, the table reports, for both states of the world, the current period utility, u d, and expected value of utility from tomorrow on, one-period deviation from the equilibrium path. In the example, ud vd, associated with a turns out to be the level of utility associated with the socially efficient level of employment, 1/(1 + 1/>) = 0.25. The deviation growth rate of money, xd, needed to achieve this level of employment is larger in the s = s(2) state of the world. To see why, note that achieving the real value of end-of-period money balances, (1 + x )/p , n = 0.25 requires setting equal to 0.25. But, in the high state of the world, intermediate goods producers anticipate a high (3 percent) growth rate of money. Consequently, the price level is high. With p high, x must be high too. Since the deviation induces a higher level of consumption and employment, nevertheless sustainable because u + v > ud + vd 27 u < ud. The outcome is in both states of the world. M /P x (h t-i , s t) = 0.0 II o oO C variable ?>T eC *o Table 1: Non-Fundamental Expectation Trap c 1 0.9709 71 0.2346 0.2304 0.2346 0.2237 W /P 0.9196 0.8962 u -2.2519 -2.2536 V -75.0892 -75.0950 ud -2.2413 -2.2413 ua -2.2910 -2.2910 va -76.3650 -76.3650 (beg. of period) Fourth, we briefly consider what sort of equilibria can sustain this outcome. One such equilibrium is the grim trigger equilibrium used in the proof to Proposition 4.2. In results hot reported here, we verified that a milder trigger strategy equilibrium that also sustains the outcome is one in which a deviation triggers a shift to the maximum growth rate of money, x, for one period. Finally, we discuss a particular non-trigger equilibrium which sustains the expectations trap outcome. Loosely speaking, in this equilibrium the monetary policy rule specifies that deviations in which money growth is high in a particular state persist forever, but deviations down have no impact on future policy. Formally, for histories, (/it_l t s(), in 28 which there has never been a deviation, x ( h t- u s t) 0.0, if { s = ( h t - i , s t ), where = s(l) (5.2) 0.03, if s = Next, consider histories, v ( s t ), s ( 2) in which there was: (i) one deviation in the past, (ii) the deviation money growth rate was a, (iii) the growth rate called for by the policy rule was x < a,and (iv) the state was x ( h t- i , s t) = where s = s(k). i>{s(k)) = a In histories like this, the policy rule specifies that and P(s(j)) = i'(s(j)) for j ^ k. That is, the rule specifies that if the policy authority deviates up in a particular state, the deviation action will be followed whenever that state recurs in the future. Histories in which there has been more than one deviation are handled in the same way. Deviations down have no impact on the policy rule. We verified, using numerical methods, that the outcome in example 1 is sustainable by this equilibrium. Taken together, these three equilibria make clear the technical role played by the grim strategy equilibrium in the proof of the characterization result. Expectation trap outcomes can be supported by variety of equilibria. E x a m p l e 2: F u n d a m e n t a l Shocks. Apart from the stochastic process governing the state of the world, s, and the state of technology, 6t , the parameter values used in this example coincide with those in Example 1. The state of the world can take on values numbered 1 through 7. In state one, while 6t = 0.90 in the other six states. The Markov chain governing s 9t = 1, has the following properties. If the economy is in state 1, it stays with probability 0.99, and it moves to state 29 2 with probability 0.01. If the economy is in states 2 through 6, the economy moves to the next higher state with probability 0.95. It stays in the current state with the complementary probability. If the economy is in state 7 it moves to state 1 with probability 0.95. In this example, there is a sustainable expectation trap outcome in which money growth is 0.0, 0.05, 0.10, 0.15, 0.20, 0.15, and 0.10 in each of states 1 through 7. This outcome has the property that money growth is typically low. Occasionally, following a decline in the state of technology, money growth and, hence, inflation rise for aft extended period of time before returning to their typically low levels. Figure 1 illustrates a typical realization in which the economy is in state 1 for 5 periods, then switches to state 2 for 8 periods, and finally returns to state 1 for 2 periods. Presumably, it is easy to construct similar examples in which expectation traps like this are triggered by other fundamental shocks, like changes in government purchases. We find Example 2 interesting because it illustrates another sense in which discretion makes possible volatile equilibria. Also, this type of example may form the basis for a simple explanation of the prolonged rise in inflation during the mid-1960s and 1970s. Authors, such as Fellner (1976) have argued that this rise in inflation was triggered by the simulta neous increase in expenditures due to the expansion of the Vietnam war and Great Society Programs. Other authors such as Blinder (1982) argue that the inflation of the 1970s was triggered by the supply shocks of that decade. Our analysis raises the possibility that the long lived inflation is consistent with sequential rationality on the part of private agents and a benevolent monetary authority. It simply reflects the type of expectation trap that can arise under discretion. 30 6. L i m i t e d C o m m i t m e n t We now show that a simple institutional change in the way monetary policy is conducted eliminates the expectation traps associated with discretionary monetary policy. The change is that we impose a form of limited commitment upon the monetary authority, in which it is required to commit to a state-contingent action one period in advance. We formally model this by specifying the time t sequence of events as follows. First, the exogenous event, realized. Second, intermediate goods producers set the time the monetary authorities’ date other date t of date ht t is t t st, is prices of their goods. Then, 4- 1 action, contingent upon St+i, is realized. Finally, the model variables are determined. In this modified version, the history at the end = (h t - i , S t , x ( s t + l )), where x(st+1) denotes the monetary authority’s time t + 1 state-contingent action. P ro p o sitio n 6.1. U nder lim ite d co m m itm e n t, th e set o f sustainable o u tc o m e s coincides w i t h t h o s e in t h e R a m s e y e q u i l i b r i u m . R em ark 1. Recall t h a t AJ a n d t h e g r o w t h .rate in a R a m s e y e q u i l i b r iu m , e m p l o y m e n t , n R , is g i v e n b y n R = A/(t/>+ o f t h e m o n e y s a t j s h e s ^^t+i M st+1 I st) i+x(7~q = h T h u s > in a R a m s e y e q u i l i b r i u m , a l l o c a t i o n s and p r i c e s ar e u n i q u e l y d e t e r m i n e d a n d m o n e t a r y p o l i c i e s are d e t e r m i n e d o n l y in an e x p e c t e d val ue sense. P roof. In the Ramsey equilibrium, employment is independent of the state of nature. The first step in the proof is to show that in any sustainable equilibrium, the date t consumption and employment allocations depend only upon the date t + 1 state-contingent monetary policy actions. To see this, first recall that, in a sustainable equilibrium, for all histories, the continuation outcome induced by the sustainable equilibrium must solve the consumer’s optimization problem and satisfy market clearing. The analog of (3.18) is: n(ht) A1 - n(ht) \ sl ) 5t+l 1 1 + x(st+1) (6.1) It follows that, for all histories, the employment allocation at date t depends only upon the state-contingent actions of the monetary authority at date t 4-1. The second step in the proof is by contradiction. Consider a sustainable outcome such that n ( h t ) ^ n R for some h t . 31 Consider the following sequence of deviations by the government: x(sr+1) = ^ —1 for r > t . Under this deviation, n ( h r ) = n R for r > t. But, equations (3.16) and (6.1) imply that n ( h t) is strictly less than n R. Since utility is strictly increasing for levels of employment less than 1/(1 + xp), this deviation raises utility. This establishes the contradiction. ■ To understand this result, it is useful to recall why expectation traps are possible when there is no commitment. Two conditions that are necessary for expectation traps to arise are: (i) private agents’ actions depend upon their expectations about future monetary policy, and (ii) private agents suffer a loss if their expectations are not validated. When these conditions are satisfied and the monetary authority is benevolent, then it is possible for expectations to drive monetary policy. Many equilibria can arise, because there is nothing in the environment to uniquely pin down agents’ expectations. Limited commitment eliminates expectation traps by forcing the monetary authorities to commit before private agents make their decisions. In this way, expectations are uniquely pinned down. We conclude this section by discussing how the previous proposition would have to be modified to accommodate different environments. For example, suppose agents committed to nominal wages or prices K periods in advance. To eliminate expectation traps in this environment, we anticipate that monetary policy would have to commit K 4-1 periods in advance. In environments in which decisions depend on the entire future history of monetary policy, for example, when agents have the option to invest in physical capital, then it may not be possible to eliminate all expectation traps with limited commitment. 7. C o n c l u d i n g R e m a r k s This paper studied the operating characteristics of monetary policy when the monetary au thority cannot commit to future policies. Our main finding is that discretion exposes the 32 economy to expectation traps. In addition, we argue that alternative institutional arrange ments for the conduct of monetary policy which impose limited forms of commitment on policy makers can eliminate the possibility of expectations traps. The existing literature emphasizing the benefits of commitment has not recognized that expectation traps of the sort analyzed in this paper can occur can occur under discretion.4 Perhaps the paper closest in spirit to ours is Cole and Kehoe (1995), who construct a model of sovereign debt and show that, under discretion, default can occur in response to a non fun damental shock. In the debt default literature, Calvo (1988) and Chari and Kehoe (1993a,b) show that, under discretion there can be multiple equilibria, even with a finite horizon. Neither of these papers discusses the possibility of excessive volatility. The sunspot literature has discussed extensively the possibility of excessive volatility.5 This literature shows that there are economies in which private agents respond to nonfun damental shocks and also shows that private agents can respond excessively to fundamental shocks (see Beaudry and Devereux, 1994). In this literature the excessive volatility arises from the structure of preferences and technology of private agents. In our environment, it arises from the structure of policy making. We conclude by discussing a shortcoming of our model, and then arguing that the basic findings of this paper are likely to be robust to a variety of plausible remedies. In our model we assumed there is an exogenously given upper bound, x, on the money growth 4See for example, Kydland and Prescott (1977), Calvo (1978), Fischer (1980), Barro and Gordon (1983), Stokey (1991), Chari and Kehoe (1990,1993a,b) and Ireland (1995). A reference to the possibility of ex pectations traps in which monetary policy responds to nonfundamental shocks appears in Rogoff (1989, p. 245-46). However, Rogoff refers to this in the context of trigger-strategy equilibria, while we show that these equilibria can be supported without trigger strategies. 5See Azariadis (1981), Benhabib and Farmer (1994,1995), Bryant (1983), Cass and Shell (1983), Cooper and John (1988), Farmer and Guo (1994), Gali (1994a,b), Matsuyama (1991), Shleifer (1986) and Woodford (1986,1991). 33 rate. Our characterization theorem requires the existence of a worst sustainable equilibrium. In our model economy there is no worst equilibrium absent an exogenous upper bound on the money growth rate. This reflects the fact that, in our environment, there is only a contemporaneous benefit, and no contemporaneous cost, from an unexpected increase in money growth. To avoid the need to specify an exogenous upper bound on money growth, we must introduce features which associate contemporaneous costs with high money growth. One such cost that has been widely emphasized is that increases in the inflation rate induce resource misallocation by changing the cross-sectional distribution of relative prices. One simple way to capture this resource misallocation is to introduce staggered price setting along the lines of Taylor (1980). To see this, consider the case where each period, half the intermediate goods producers set a price for their goods which will prevail for two periods. Under these circumstances, the higher is inflation, the greater the difference is between the two sets of prices. This results in an inefficient distribution of production among the intermediate goods producers. If we assume that the firms setting prices in any particular period do so after the current period realization of the monetary action, then this obviously creates a contemporaneous cost to increasing the money growth rate, and this cost is greater at higher inflation rates. We expect that expectation traps can arise under discretion in this environment too: suppose the subset of firms setting prices in a particular period expect money growth to be high in the next period. They will find it in their interest to set prices high. Then, in the next period the monetary authority faces a choice: either validate the expectations of the previous periods’ price setters, or incur a recession. As in the environment of this paper, it may choose to validate the expectations, thus ensnaring the economy in an expectations trap. 34 FIG U R E 1 employment output years years References [1] Abreu, D., 1988, ‘On the Theory of Infinitely Repeated Games with Discounting,’ E c o n o m e tr ic a , vol. 56, pp. 383-396. [2] Azariadis, Costas, 1981, ‘Self-Fulfilling Prophesies,’ pp. 380-396. J o u rn a l o f E c o n o m ic T h e o r y , 25, [3] Beaudry, Paul and Michael B. Devereux, 1994, ‘Monopolistic Competition, Price Set ting and the Effects of Real and Monetary Shocks,’ manuscript, University of British Columbia. [4] Barro, Robert J., and David B. Gordon, 1983, ‘Rules, Discretion and Reputation in a Model of Monetary Policy,’ J o u rn a l o f M o n e ta r y E c o n o m ic s , vol. 12, pp. lul-121. [51 Benhabib, Jess, and Roger E. A. Farmer, 1994, ‘Indeterminacy and Growth,’ o f E c o n o m ic T h e o ry 63, pp. 19-41. Jou rn al [6] Benhabib, Jess, and Roger E. A. Farmer, 1995, ‘Indeterminacy and Sector-Specific Externalities,’ manuscript, department of economics, UCLA, May. [7] Blinder, Alan S., 1982, ‘Anatomy of Double-Digit Inflation in the 1970s,’ in Robert Hall, editor, In fla tio n : C a u ses an d E ffects, National Bureau of Economic Research, University of Chicago Press. [8] Bryant, John, 1983, ‘A Simple Rational Expectations Keynes-Type Model,’ J o u rn a l o f E c o n o m ic s, Vol. XCVIII, no. 3. Q u a r te r ly [9] Burns, Arthur F., 1978, ‘Reflections of an Economic Policy Maker, Speeches and Con gressional Statements: 1969-1978’, American Enterprise Institute for Public Policy Re search, Washington, D.C. [10] Calvo, Guillermo A., 1978, ‘On the Time Consistency of Optimal Policy in a Monetary Economy,’ E c o n o m e tric a , vol. 46, pp. 1411-1428. [11] Calvo, Guillermo A., 1988, ‘Servicing the Public Debt: The Role of Expectations,’ A m e r ic a n E c o n o m ic R e v ie w , vol. 78, pp. 647-661. [12] Cass, David, and Karl Shell, 1983, ‘Do Sunspots Matter?’, 91, 193-227. J o u rn a l o f P o litic a l E c o n o m y [13] Chari, V.V., 1988, ‘Time Consistency and Optimal Policy Design,’ Federal Reserve Bank of Minneapolis Q u a r te rly R e v ie w , Fall. [14] Chari, V.V., and Patrick J. Kehoe, 1990, ‘Sustainable Plans,’ J o u rn a l o m y, vol. 98, no. 4, pp. 783-802. 36 o f P o litic a l E c o n [15] Chari, V.V., and Patrick J. Kehoe, 1993a, ‘Sustainable Plans and Debt,’ E c o n o m ic T h e o r y , vol. 61, no. 2, pp. 230-161. Jou rn al o f [16] Chari, V.V., and Patrick J. Kehoe, 1993b, ‘Sustainable Plans and Mutual Default,’ R e v ie w o f E c o n o m ic S tu d ie s, vol. 60, pp. 175-195. [17] Christiano, Lawrence J., Martin Eichenbaum, and Charles Evans, 1996, ‘The Effects of Monetary Policy Shocks: Evidence from the Flow of Funds,’ forthcoming, R e v ie w o f E c o n o m ic s a n d S ta tis tic s . [18] Cole, Harold, and Timothy Kehoe, 1995, ‘Self-Fulfilling Lending Crises,’ manuscript, Federal Reserve Bank of Minneapolis. [19] Cooper, Russell, and Andrew John, 1988, ‘Coordinating Coordination Failures in Key nesian Models,’ Q u a r te rly J o u rn a l o f E c o n o m ic s, 103, August, pp. 441-463. [20] Farmer, Roger E. A., and Guo, J.-T., 1994, ‘Real Business Cycles and the Animal Spirits Hypothesis,’ J o u rn a l o f E c o n o m ic T h e o ry 63, pp. 42-73. [21] Fellner, William J., 1976, T o w a rd s a R e c o n stru c tio n T h e o ry a n d P o lic y , American Enterprise Institute. o f M a c ro e c o n o m ic s, P ro b le m s o f [22] Fisher, Stanley, 1980, ‘Dynamic Inconsistency, Cooperation and the Benevolent Dis sembling Government,’ J o u r n a l o f E c o n o m ic D y n a m ic s a n d C o n tro l, vol. 2, pp. 93-107. [23] Gali, Jordi, 1994a, ‘Monopolistic Competition, Business Cycles, and the Composition of Aggregate Demand,’ J o u rn a l o f E c o n o m ic T h e o ry 63, pp. 73-96. [24] Gali, Jordi, 1994b, ‘Monopolistic Competition, Endogenous Markups and Growth,’ E u ro pean E c o n o m ic R e v ie w 38, pp. 748-756. [25] Ireland, Peter N., 1995, ‘Sustainable Monetary Policies,’ manuscript, Federal Reserve Bank of Richmond. [26] Kydland, Finn, and Edward C. Prescott, 1977, ‘Rules Rather than Discretion: The Inconsistency of Optimal Plans,’ J o u rn a l o f P o litic a l E c o n o m y , vol. 85, pp. 473-491. [27] Mathene, Ken, 1995, ‘Increasing Returns and Monetary Policy,’ manuscript, Depart ment of Economics, Purdue University. [28] Matsuyama, Kiminori, 1991, ‘Increasing Returns, Industrialization, and Indeterminacy of Equilibrium,’ Q u a r te r ly J o u rn a l o f E c o n o m ic s, vol. CVI, no. 2, pp. 617-650. [29] Rogoff, Kenneth, 1989, ‘Reputation, Coordination, and Monetary Policy,’ in Robert Barro, editor, M o d e m B u s in e s s C ycle T h eory, Harvard University Press. [30] Shleifer, Andrei, 1986, ‘Implementation Cycles,’ J o u r n a l pp.1163-1190. 37 o f P o litic a l E c o n o m y , 94, no.6, [31] Stokey, Nancy, 1991, ‘Credible Public Policy,’ J o u rn a l tr o l , vol. 15, pp. 627-656. o f E c o n o m ic D y n a m ic s a n d C o n [32] Taylor, John B., 1980, ‘Aggregate Dynamics and Staggered Contracts,’ J o u r n a l litic a l E c o n o m y , vol. 88. [33] Whittle, P , 1983, O p tim iz a tio n C o n tro l, II, Wiley: New York. O v e r T im e: of Po D y n a m ic P ro g ra m m in g a n d S to c h a s tic [34] Woodford, Michael, 1986, ‘Stationary Sunspot Equilibria in a Finance Constrained Economy,’ J o u r n a l o f E c o n o m ic T h e o ry , vol. 40, no. 1, October. [35] Woodford, Michael, 1991, ‘Self-Fulfilling Expectations and Fluctuations in Aggregate Demand,’ in N. Gregory Mankiw and David Romer, editors, N e w K e y n e s ia n E co n o m ics, vol. 2, C o o rd in a tio n F ailu res a n d R eal R ig id itie s , MIT Press. 38