Woodford Fiscal Requirements of Price Stability

Fiscal Requirements for Price Stability Author(s): Michael Woodford Reviewed work(s): Source: Journal of Money, Credit and
Banking, Vol. 33, No. 3 (Aug., 2001), pp. 669-728 Published by: Ohio State University Press Stable URL: http://www.jstor.org/stable/2673890 . Accessed: 01/01/2012 17:02
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Ohio State University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Money, Credit and Banking.
http://www.jstor.org
MICHAELWOODFORD
Fiscal Requirements for Price Stability

Maintainingprice stabilityrequiresnot only commitmentto an appropriate monetarypolicy rule, but an appropriate fiscal policy rule as well. Ricardianequivalencedoes not imply that fiscal policy is izaelevant, except in the case of a certainclass of policies ("Ricardian" policies). The role of fiscal developmentsin inflationdetermination under a non-Ricardianregime is illustratedthrough an analysis of the bond-price supportregime of the 1940s. A monetary-fiscalregime with attractive propertieswould combine a "Taylor rule" for monetaly policy with nominal-deficittargeting as a fiscal policy commitment. "Proposalsfor a 7Y1071.etS77 rule requirea supplementary proposalof a fiscal KarlBrunner(1986), p. 54
rule.?
RECENTYEARS have seen a worldwide movement toward greateremphasis upon the achievementof inflationtargetsas the plimaly clîtelîon for judging the success of central banks' conduct of monetarypolicy. At the same time, the independenceof centralbanks in thelr choice of the means with which to pursuethis goal has also increased.An implicationwould seem to be that it is now widely acceptedthatthe choice of monetarypolicy to achieve a tal^get path for inflation is a problemthatcan be, and indeed ought to be, separatedfiôm othel^ aspects of governmentpolicy, such as the choice of fiscal policy.1But is this really so clear? Or do the agencies lêsponsible for inflationstabilizationproperlyneed to concernthemselves with fiscal policy choices as well, while the agencies concerned with fiscal policy have a corresponding need to coordinatetheir actions with those of the monetary authority? The argument for separation of decision making about these two aspects of macroeconomicpolicy necessarily relies upon two theses: first, that fiscal policy is
This Jollsnr71 of Moszevy, C7edits7Z?d salzlkislg lecture was presentedat Ohio State Universityon May 1, 2000. The author thanks Michael Bordo, Matt Canzonezi, Steve Cecchetti, Larry Christiano, John Cochrane,Paul Evans, EduardoLoyo, Bennett McCallu1n,Helene Rey, Stephanie Schmitt-Grohe,and Chris Sims for helpful discussions, Gauti Eggertsson for researchassistance, and the National Science Foundationfor researchSUppOlt througha grantto the National Bureauof Economic Research. 1. A particularly strikingexample of an attemptto separatethe two types of policy decisions is the Europeanmonetaryunion, in which monetarypolicy is the responsibilityof a supranational EuropeallCentralBank, while fiscal policies continueto be the prerogativesof individual1lational governments.
MICHAEL WOODFORD is proJessor of eco7107nicls aIt P7i71ceto71 U7ziversity.E-nlail: woodford@phoenix.princeton.edu Jolar71al of Mo7ey, Credit,co7zd Bcl7zki7zg, Vol. 33, No. 3 (August 2001) Copyright2001 by The Ohio State University
670
: MONEY,CREDIT,AND BANKING
of little consequence as far as inflationdetermination is concerned,and second, that monetarypolicy has little effect upon the governmentbudget.I shall arguehere that neitherpropositionis true,for reasonsthatare related.The fiscal effects of monetary policy are often thoughtto be an insignificantconsiderationin the choice of monetary policy by the majorindustrialnations, because seignioragerevenues are such a small fractionof total governmentrevenuesin these countries.But such a calculation neglects a more importantchannel for fiscal effects of monetarypolicy, namely the effects of monetary policy upon the real value of outstanding government debt, throughits effects upon the price level (given that much of the public debt is nominal) and upon bond prices, and upon the real debt scrvice requiredby such debt (insofar as monetarypolicy can affect real as well as nominal interestrates).2 Fiscal policy is often thought to be unimportant for inflation determination at least when, as in countrieslike the United States andthe United Kingdom,a desireto obtain seigniorage revenues plays no apparentrole in the choice of monetarypolicy on two different,though complementary, grounds.On one hand, it is often argued that inflationis purely a monetaryphenomenon,and hence that only the choice of monetary policy matters for what level of inflation one will have. And on the other, the celebrated "Ricardianequivalence" proposition implies that insofar as consumershave rationalexpectations,fiscal policy should have no effect upon aggregatedemand,and hence no effect upon inflation. I shall argue that neither propositionis of such general validity as is often supposed.As a considerablerecentliterature has stressed,3fiscal shocks affect aggregate demand,and the specificationof fiscal policy mattersfor the consequencesof monetarypolicy as well, in rationalexpectationsequilibriaassociatedwith policy regimes of the kind that I shall call "non-Ricardian" (Woodford1995, 1996), even when the monetarypolicy rule involves no explicit dependence upon fiscal variables of any sort.This happens,essentially,throughthe effects of fiscal disturbances upon privatesector budget constraintsand hence upon aggregatedemand. Such effects are neutralized by the existence of rationalexpectations and frictionless financial markets only if it is understood thatthe governmentbudgetitself will always be subsequently adjustedto neutralizethe effects, in presentvalue, of any currentfiscal disturbance. A "non-Ricardian" fiscal policy is one thatdoes not have this property;we show that non-Ricardian policies may easily be consistent with the existence of a rationalexpectations equilibrium,which means that the expectationthat the governmentwill follow such a rule need never be disconfirmed.
2. See King (1995) for discussion of this point, with some quantitative evidence. 3. The discussion of price-leveldetermination undera non-Ricardian policy regimein section 1 below recapitulatesresults from Woodford(1994, 1995, 1996, 1998c), drawingalso upon the importantcontributions of Leeper (1991), Sims (1994), and Cochrane(1999). Importantprecursorsof this literatureinclude Sargent (1982), Begg and Haque (1984), Shim (1984), d'Autume and Michel (1987), and Auernheimerand Contreras(1990, 1993). Otherrecent discussions and extensions of this work include Bassetto (2000), Benhabib, Schmitt-Grohe,and Uribe (2001a, 2001c), Benassy (2000), Bergin (1996), Buiter (1998, 1999), Canzoneriand Diba (1996), Canzoneri,Cumby,and Diba (1998, 1999), Carlstrom and Fuerst (2000), Christianoand Fitzgerald (2000), Cochrane (1998, 2000), Cushing (1999), Daniel (1999), Dupor (2000), Gordonand Leeper (1999), Kenc, Perraudin, and Vitale (1997), Kocherlakota and Phelan (1999), Leith and Wren-Lewis (1998), Loyo (1997, 1999, 2000), McCallum (1998, 1999), Schmitt-Grohe and Uribe (2000), and Sims (1997, 1998, 1999).
MICHAELWOODFORD : 671
This possibility, however, means that a central bank charged with maintaining To be conprice stabilitycannot be indifferentas to how fiscal policy is determined. crete, I shall arguethat the meleecommitmentof a centralbank to conductmonetary policy accordingto a rule such as the "Taylorrule" (Taylor 1993) is insufficientto fisensure a stable, low-equilibriumrate of inflation.On one hand, (non-Ricardian) cal expectations inconsistent with a stable price level may frustratethis outcome, even when monetarypolicy is itself consistent with price stability.Indeed, the combinationof a Taylorrule with certainkinds of fiscal policy may resultin an inflationary or deflationary spiral. And on the other hand, even when fiscal policy is with stable prices, the policy regime (includingthe commitmentto a Tayconsistent expectations equilibria, lor rule) may not preclude otherequally possible lâtional Alternativefiscal polspirals.4 such as equilibriainvolving self-fulfilling deflationary icy commitments may instead exclude these undesired deflationaryequilibria [as discussed by Woodford(1999a)], and thus in this way help to ensure stable prices. both of these issues, I shall suggest that a TayAs a practicalproposalthat addleesses lor rule for monetarypolicy should be accompaniedby targetsfor the size of governmentbudgetdeficits.
REGIME UNDER A BOND PRICE-SUPPORT DETERMINATION 1. PRICE-LEVEL
Before turningto a discussionof Taylorrules, it will be useful to take up the more of the equilibrium generalquestionof how fiscal policy can affect the determination to the plice plice level. The role of fiscal developmentsas a source of disturbances level can be seen most clearlyin policy regimessometimessaidto involve "fiscaldomThese are policy regimes,often associatedwith the special fiscal pressuresof inance." to the goal of warfinance,in which othergoals of centralbankpolicy aresubordinated to note assistingin the financingof the governmentbudget.However,it is important thatthis does notnecessarilymean thatfiscal developmentsaffect the price level only becausethe centralbankadjustsmonetarypolicy in responseto them. A familiartextbook account of fiscally dominantregimes runs as follows: fiscal exigencies determinethe size of a real governmentbudget deficit that must be financed; this budget shortfall is then assigned to the central bank as a level of seignioragerevenuethatit must generatethroughmoney creation;the monetarybase is increasedby whateveramountsuffices to generatethe requiredrevenues;and finally, the rate of money growthdeterminesthe equilibriumrate of inflation,thleough mechanism.Under this account,fiscal developmentsafthe usual quantity-theoretic fect the rateof inflation,but onlybecause they affect monetarypolicy, underthis particular sort of monetary policy rule; inflation is still a "purely monetary" phenomenon.Such an account is still perfectly consistent with the view that commonetarypolicy is sufficientto ensureprice stability. mitmentto an anti-inflationary
4. Benhabib, Schmitt-Grohe,and Uribe (2001b) criticize regimes involving a Taylor rule on this to note thatthe problemthatthey identify is in no way special to the Taylor ground,thoughit is important rule.
672
the model just sketched Illightseem to apply only to a few less-develFurthermore, oped economies, not to advancedeconomies such as the United States or the European Union. For it would seem not to apply in the case of an independentcentral nor would it bank,that need not accept seignioragetargetsdictatedby the Treasury; seem likely to apply to an economy with sophisticatedfinancialmarkets,in which it is difficult for the governmentto raise large seigniorage revenues, because of peoassets. Thus the partof the ple's ability to substituteaway from non-interest-earning world in which such a regime would even be a potentialoutcome might seem to be rapidlyshrinking. Instead,I shall arguethatfiscal policy can affect the price level even when the cenmonetarypolicy, by which I mean a rule for setting tralbankpursuesan autonon1.0lzs its instrument(in practice,a nominal interestrate) that is iMdepeMdeMt offiscal variables.Thus it will not be enough, to avoid price-levelinstabilityresultingfrom fiscal under which the central disturbances,to simply adopt an institutionalarrangement no directivesfiom the Treasurydictatingchanges in policy; nor will it bank leeceives be enough thatthe centralbank commits itself to an interestraterule, like the Taylor rule, thatinvolves no directfeedbackfrom variablessuch as the the governmentbuddescribedhere will conthe potentialeffects of fiscal disturbances get. Furthermore, tinue to exist even in what I shall the "cashless limit" (Woodford 1998a) the hypothetical limiting case of an economy in which financial innovation has proceeded to the extent that available seigniorage revenues are negligible. This is because these effects in no way depend upon attempts to use monetary policy to generate seigniorage revenues. Thus the possibility that fiscal policy may interfere with the achievementof price stability cannot be so easily dismissed, even for advanced economies. regimes often do not involve any direct assignmentof In fact, "fiscallydominant" a seignioragetargetto the centralbank, as in the textbook analysis. Instead, "fiscal dominance''manifests itself throughpressureon the central bank to use monetaley debt.A classic example is provalueof governncent the market policy to maintain vided by U.S. monetarypolicy from 1942 up until the Treasury-Fed"Accord"of Marchl951.5 Beginning in April 1942, the Fed and the Treasuryagreed to an interestrate control program,the declaredaim of which was to maintain"relativelystable prices and yields for government securities."6The yield on ninety-day Treasury bills was pegged at 3/8 of a percent;this peg was maintainedthroughJune 1947, and as shown in Figure la, until that point the price of bills was completely fixed, as the Treasury offered both to buy and sell bills at that price. An intentionwas also announcedof supportingone-year Treasurycertificatesat a price correspondingto a 7/8 percent annualyield; this policy continuedafter 1947, though at a slightly higher yield. Finally, the prices of twenty-five-yearTreasurybonds were supportedat a price correalldGarber (1991);TimandSchwartz (1963,chap.10);Eichengreen Friedman 5. See,forexample, (1997,chap.8). berlake (1993,chap.20);andToma by Timberlake (1993,p. 304). 6. Eccles(1951,p. 350);quoted
674
to Tleasurypolicy, this is this periodis typically describedas thoroughlysubordinate actuallyan example of an autonomousmonetaly policy, in the sense defined above. purchasesand sales so as to stabilize the plices A policy of conductingopen-market of Treasurysecuritiesis one thatrequiresno centralbankmonitoringof fiscal developmentsfor its implementation,nor any directivesfrom the Treasuryabout how to respondto fiscal developments.It is in fact an especially simple example of an intel-rat.e peg.Any effect of fiscal shocks est raterule, essentially equivalentto an interest upon the growthof the monetarybase underthis leegimewas purely a genel-al-equilibriumphenomenon,and not a consequence of any direct dependenceof the Fed's interestrate targetsupon such shocks. Yet fiscal developmentsclearly have a majorimpact upon the course of inflation undelO such regimes. For example, in the case of the United States in the 1940s, the duringthe war period, though wage and price controls supregime was inflationaley pressed much of this inflation until their relaxationin several stages during 1946. to any (The burstof inflationin 1946-47 seen in Figure lb should not be attributed sulege in aggregatedemandat thattime, butratherto the allowanceof prices to finally regime resulted rise to their equilibriumlevel.) On the otherhand, the price-support over the period 1948-50. This correspondsto a periodin which the large in deghation wartimedeficits had ended, and the U.S. governmentbudgetwas insteadchronically in surplus.With the outbreakof the Korean war in June 1950, inflation suddenly regime came to be began again. It was only at this time that the bond price-support and was for thatreason suspended.8 denouncedas "anengine of inflation," How is one to explain these effects upon the generallevel of prices of variationin the fiscal situation?It cannot be through any directeffect of fiscal developments upon monetaly policy, understoodto refer to the Fed's rule for setting interestrates. Rather,such effects indicatethatthe governmentbudgetcan play a role in price-level in additionto the specificationof monetarypolicy. determination view of inflation determiMight one still salvage a traditionalquantity-theoretic nation by saying that in such a regime, the money supply depends upon the governit is true that it does; ment budget, as well as the interestrate rule? In equilibrium, fiscal disturbancesaffect the equilibriumgrowth rate of the money supply. But the causality is notfrom the governmentbudgetto the growthof the money supply,and then only from the change in the money supply to prices. Rather,the government priceschange does it also because budgetaffects the generallevel of prices, and onl.y affect the money supply (as higherprices lesult in highermoney demand,which the Fed passively accommodates under such a regime). Thus one cannot explain the change in the price level as being due to the increasein the money supply. Upon first thought, one might suppose that under a bond price-supportregime, there is a directconnectionbetween the governmentbudget and growthin the monetarybase. One might reasonthata commitmentby the Fed to act as the residualpurchaser of governmentdebt will require the Fed to increase the monetarybase, in
8. See Brunnerand Meltzer (1966) for an importantdiscussion of this period, stressingthat the inflacharacterof the regime dependedupon fiscal policy. tionaryor deflationary
order to increase its holdings of government debt, whenever the Treasuryissues more debt, which is to say, whenever(and to the extent that) the governmentruns a budget deficit. But this superficialanalysis implicitly assumes that the public's demand for governmentbonds is fixed, so that (in the absence of a price change) the Fed will have to acquirethe additionalissues, while it assumes at the same time that there is no obstacle to increasing the public's money holdings by an arbitrary amount,without any change in the relativeyield on money and bonds. Instead, economic theory implies that if anything,the oppositerelations should obtain.Thereare good reasons why it may not be possible for the Fed to increasethe monetarybase withouthaving to accept a change in the yields on Treasurysecurities. A money demand relation of the conventional sort [for example, equation (16) below] implies that the public's desired money balances will be a function of the price level, of the quantityof real transactions,and of the interest differentialbetween money and bonds, but notof fiscal variablessuch as the stock of public debt. Thus it is generally supposedthatthe Fed cannot change the monetarybase without accepting a change in the level of interestrates, something that is precludedunder the bondprice-support regime.At the same time, thereare equally good reasonswhy an increasein governmentborrowingmight well increasethe public's willingness to hold governmentbonds, even in the absence of any change in bond yields. Indeed, the doctrineof RicardianEquivalenceasserts that governmentborrowingautomatically creates an increase in desired privatebond holdings of exactly the same size (due to an increasein expected futuretax obligations), so that bond yields need not change at all to maintainequilibriumin the bond market. The analysis that I shall propose here will not imply that RicardianEquivalence obtains(in thatcase, therewould be no inflationary impactof an expectationof budget deficits,either).But it will assume a conventionalmoney demandrelation,so that the quantityof money thatmust be suppliedin orderto maintainbond prices at their targetlevels is a functionsolely of prices andreal activity.Thus the governmentbudget will be able to affect the money supply onlybecause it is able to affect equilibrium prices throughanotherchannel;prices will not be affected only because of the change in the money supply.
1.1A Sisnple Model

Let us consider price-level determinationunder such a regime using a simple monetaryframework,namely, a representative-household model of the kind introduced by Sidrauski(1967) and Brock (1974, 1975). I shall suppose that the representativehousehold seeks to maximize a discountedsum of utilities of the form
r 00
Eoj , |3tU(ct + gt,MtI

t=0
Pt)J
1
(l)
where U(c, m) is an increasing, concave function of both arguments,and the discount factor satisfies O < 13< 1. The second argumentof U indicates the liquidity
676
services providedby end-of-periodmoney balances Mt;these depend upon the real purchasingpower of those balances, so thatMtis deflatedby the price level Pt. In the specification(1), I assume that (real) governmentpurchasesgt areperfect substitutes for (real) privateconsumptionexpenditurect. This simplificationallows us to focus solely upon the effects of fiscal policy upon privatebudget constraints;government purchaseshave exactly the same effect on the economy as transfersto householdsof funds sufficientto finance privateconsumptionof exactly the same amount.(I shall assumethattaxes are lump sum for the same reason;a tax increasewill then have the same effect as a reductionin transfersthat reduces household budgets in the same amount.) The representative householdis subjecteach period to a flow budgetconstraintof the form Mt + Et[Rtt+l(Wt+l-Mt)] ' Wt+ Ptyt-Tt-Ptct, (2)
statingthat end-of-periodfinancialwealth (money balancesMt plus bonds) must be no greaterin value than financialwealth Wtat the beginning of the period, plus income from the sale of period t productionYt' net of tax paymellts and consumption expenditure.The variableTtrepresents(nominal)tax obligations net of any government transfers; the two componentsneed not be distinguished,as taxes are assumed to be lump sum. The difference Wt+ 1-Mt representsthe (nominal)value in period t + 1 of the household'sbond portfolio at the end of period t; as I assume complete financialmarkets,this portfolio may include state-contingentclaims of many sorts. The (nominal)marketvalue of such a bundleof state-contingent claims in periodt is given by Et[Rtt+l(Wt+l-Mt)], where the randomvariableRtt+l is a stochastic discount factor for pricing arbitrary(nonmonetary)financial claims.9 Note that the household,as a price takerin financialmarkets(as well as goods markets),takes the evolution of the stochasticdiscountfactor as being independentof its own portfolio decisions (indicatedby the evolution of Mt and Wt). The nominal interest rate it on a one-period riskless claim purchasedin period t must satisfy
1 + it = Et[Rt,t+l]
Using this, we may rewrite(2) in the form
Poct +
Mo + + Er[Rt r+lWt+t] < Wo + [Ptyo-Tr],
(4)
9. The existence of such a pricingkernelfollows from the absence of arbitrage opportunities; the pricing relationapplies, of course, only to financialassets that (unlike money) do not yield additionalnonpecuniarybenefits.Under our assumptionof complete markets,R, ,+ g is uniquelydefined.
form. in which it/(l + it) appearsas the effective cost of holding wealth in rllonetary limit each period, according to which the houseLet us also assume a borwowing hold's portfolio (including any shortpositions) must satisfy
00
Wt+1 -
E
T=t+1
Et+1[Rt+1,T(PTYT
TT)]
(5)
in each possible state in period t + 1; this states thatthe household must never have debts greaterthanthe presentvalue of all futureafter-taxincome.10The sequence of flow budgetconstraints(4) combined with (5) is thell equivalentto the intertemporal budgetconstraintll
00 1 00
E
T=t
+ EtRt,TPTCT
1 + tT
. MT C
_
Wt +
E
T=t
ElRtvT[PTYT
TT]
(6)
We may thus state the household'sproblem,looking forwardfrom any date t, as the choice of a consumptionplan and plannedmoney holdings to maximize (1) subject to (6), given financialwealth Wt. Necessary and sufficientconditionsfor householdoptimizationl2 arethen thatthe conditionsl3 first-order
Usal(Ct + gt ' }}>t) =
UC (Ct + gt ' }}1't )
tt
1 + it
UC(ct + gt}nt)
(8)
Uc(ct+l + gt+l,mt+l)
Rt,t+l Pt+l
budgetconstraint, hold at all times, and thatthe householdexhaustits intertemporal that is, that
00 * 00
E
T=t
+ 1 . MT = EtRtT PTCT
+ tT _
Wt +
E
T=t
EtRt,T[PT)T
TT]
( )
This last conditionstates both thatthe left- and right-handsides of (6) are equal, and
for discountingincome in period Tback to period t + 1 is defined 10. FIerethe discountfactorR{+1,T as tlle productof factorsRss+l for s runninghom t + l tllroughT-l; it is equal to one when T = t + 1. 1l . See Woodford( l 999a) for details. 12. For simplicity,we ignore the possibility of cornersolutions. 13. In writingthese, I use the notationm,-M,lP, for real money balances.
EtRt,T
PTCT
MT
<
(10)
678
that both infinite sums converge.14 This condition for optimalitycould equivalently be replacedby the stipulationthat the household's plannedexpenditurehas a finite presentvalue,
T=t
+ tT
togetherwith a transversality conditionon wealth accumulation,15 lim Et[Rt TWT ] (11)
A rationalexpectationsequilibriumis then a collection of state-contingentpaths for the variousendogenous variablesthat satisfy these conditions for household optimization,togetherwith the market-clearing conditions ct + gt = Yt, Mt= Mts, Wt+1= Wt+1 (12) (13) (14)
at all dates andin all possible states.16Here the aggregatesupply of goods Ytis an exogenously specified stochasticprocess, whereas the money supply Mtsand the market value of total beginning-of-period government liabilities Wt+1 evolve in accordance with the specification of monetaryandfiscal policy (to be clarifiedbelow). Substituting(12)-(13) into (7), we obtainthe equilibriumcondition u,7](yt,Mt IPt)=
UC(yt,MtslEt)
ir .
l+it
(1.5)
Under standardassumptions on preferences,17this equation can be solved for a uniqueequilibriumlevel of real money balances,
14. The latterstipulationis necessary,as both left- and right-handside being infinite would not imply thatthe householdcould not affordto consum more. Indeed,in such a case, (S) would impose no limit on borrowing,and "Ponzischemes"would be possible, allowing unboundedconsumptionat all dates. 15. Again, see Woodford(1999a) for details. 16. Equilibriumfrom some date t onwardrequiresthat (12)-(14) be expected to hold at all dates t ' T. The fact that WT= WTwould follow from the specificationof the initial portfolio of the representative household,ratherthanbeing a market-clearing condition. 17. In additionto those noted earlier,we assume thatboth consumptionand liquidityservices are normal goods, and also assume boundaryconditions guaranteeingan interiorsolution to (15).
WOODFORD : 679 MICHAEL

Ms
' = L(yt,it) t
(16)
t
where the "liquiditypreferencefunction"L is increasingin its firstargumentand deconditionstating an equilibrium creasingin the second. Thus our model incorporates that the price level is at all times such that the implied real value of the money supply is equal to desired real balances;but as we shall see, this need not mean that the by the evolutionof the money supply. evolutionof the price level is best explained A similar substitutionof (12)-(13) into (8) allows us to solve for the stochastic discountfactor,obtaining
Rtt+1=
pt / Pt+l) Pt+ D Uc(Yt+lMt+l Uc(yt,MtIPt)
(17)
Substitutionof this into (3) then yields
tE, Uc(Yt+lM,+l / P+l) Pt 1 + i, = 13-1
(18)
linking nominalinterestrates relationis a sort of "Fisherequation," This equilibrium to expected inflation,but also involving the real factors that determinethe equilibium real rate of interest.In the familiartextbook case of a utility function U that is additivelyseparablebetween consumptionand liquidity services (or in the "cashless limit" discussed below), (18) reduces to
l+i,=:
IjE,
,('+)
p'
(19)
where u(ct + gt) is the palt of U thatdependsupon consumption(or the value of U in the "cashlesslimit").In this special case, the expectedrate of inflationis the only endogenous variableon the right-handside of the equation. Under similar substitutions,the remainingrequirementsfor optimality,(10) and (1 1), become
00
ZnT )CT , ,B Et[Uc(YT, T=t
+ UE71(YT
MT )MT ] Z
(20)
(21)
]= )WTI PT Tlim: Et[Uc(yT,Mt I PT
Et
F(YT
MT
PT
<
(22)
680
Here I have substituted(17) to eliminatethe stochastic discount factors, and in (20) thatthe have also substituted(15) for the factoril(1 + i). Let us suppose furthermore shareof governmentpurchasesin the total nationalproductis bounded,that is, that ' gt ' Bt at all times, for some bound0 < Y < 1. Then we must have CT ' YT' (1 ny) CT
at all times, so that (20) is equivalentto the condition

00
T=t
where
F(Y,
m)-Uc(y,
m)y +
Us7Z(y,
m)m .
Thusboth of the remainingequilibriumconditions,(21) and (22), place boundsupon to the nomihow far the price level can diverge asymptoticallyfrom proportionality nal asset supplies Mtsand Wts. be condition for optimal wealth accumulationcan alternatively The transversality expressedby the equality in (9). A similar substitutionof conditions (12)-(14) and into this equationyields
_ _
D]T-tEUC(YTH7HT)
TT (23)
as a substitutefor (21). One notes that the presentvalue of the YT-gT terms on the left-hand side must be finite, as a consequence of (22) and the assumed bound on one these termsfrom both sides andrearranging, governmentpurchases.Subtracting obtainsthe equilibriumcondition WT= Pt T-tE E [D]
T=t
(YT t T) s + UC Uc (Yt,m, )
T
1+
iT
T PT
(24)
where st denotes the real primarygovernmentbudget surplus T

S Pt
t _
This condition states that the real value of net governmentliabilities must equal the presentvalue of expected futureprimarybudgetsurpluses,correctedto take account of the government'sinterestsaved on the partof its liabilities that the public is willing to hold in monetaryform. Note, however,thatthis relationnecessarilyobtainsin a rationalexpectationsequilibrium,not because we have assumed it as a constraint upon the government'sfiscal policy, but ratherbecause it follows from privatesector optimization,togetherwith marketclearing. (This point will be of considerableimportancefor the discussion below.) To sum up, a rational expectations equilibrium is a collection of stochastic processes {Pt, it, Mts,Wts}that satisfy (16), (18), and (22), as well as either (21) or (24), along with the equations specifying monetaryand fiscal policy. These equations suffice to determineequilibriumill the case that both the monetarypolicy rule and the law of motion for governmentliabilities given the fiscal policy rule can be specified without referenceto asset prices other than it. (An example of such a case is presentedin the next subsection.) Once an equilibrium(that is, solution to these equations)is found, the implied equilibriumprocesses for all other asset prices are then given by (17). If insteadmonetaryand/orfiscal policy cannotbe specified without referenceto longer-term bond prices, the necessarybond pricingequationsmust be adjoinedto the system of equationslisted above, and the bond prices in question addedto the list of endogenousvariablesthat arejointly determined. 1.2 A Treasury-Bill Peg Let us now consider the equilibrium price level under a bond price-support regime. As a first simple example, suppose that monetarypolicy pegs the price of a one-period Treasury bill; thus it is equivalent to specification of an exogenous process {it} for the short-term nominalinterestrate.We shall assume that it > 0 at all times.l8 Let us suppose furthermore that fiscal policy is describedby an exogenous primary-surplus process {st}. Since Ytis assumed to be exogenous, such a fiscal specification might correspondto an exogenous process {gt} for real government purchases,togetherwith an exogenous process for a proportional tax rate {t}, with aggregatetax collections then evolving as Tt = ttPtyt.Such a specificationof fiscal expectationsis particularly likely to apply in wartime,when governmentpurchases varyfor reasonslargely independentof the state of the economy or the government's budget, and when the government'sability to furtherincrease tax rates may also be tightly constrained. Suppose also, for simplicity,thatthe public debt consists entirelyof (riskless nominal) one-periodTreasurybills. Then total governmentliabilities at the beginning of any period t are equal to
Wts = Mts_l + (1 + it_l)Bts_l, 18. The theory extends directly to the case of a zero yield, as long as preferencesinvolve satiationin money balances at some finite level. In that case, the equilibriumpath of the price level would still be uniquely defined, but the equilibriummoney supply would be indeterminate(it could take any value greaterthanor equal to the satiationlevel) in all periods with i, = O.
682
where Btsdenotes the supply of Treasurybills at the end of period t (measuredby their marketvalue at the time of issuance). The flow budget constraintfor the governmentimplies thatthe supply of bills must satisfy
Bts = Wt -P4t-Mt .
It follows thatunderthis fiscal regime, total governmentliabilities evolve according to the law of motion
Wt+l=(l+ir)
Wts_P,st-1
'
Mts
(25)
OUrproblemis now to solve for rationalexpectationsequilibriumpleocesses{Pt, Mts, Wtssatisfying (16), (18), (22), (24), and (25), given exogenous processes {Yt,it, st} and an initial quantityof nominalgovernmentliabilities. The equilibriumconditions may be solved sequentially,as follows. We first note that (16) determinesthe equilibriumevolution of real balances, given the exogenous processes {Yt,it}. Substitutingthis solution for real balancesinto (24), we obtain
P,
T=t
()),,
i, )
+ iT
(26)
where
k(, i)-Uc(J, L(,i))

Note that all termson the right-handside are now functions of exogenous variables. Let us suppose that the fiscal expectationsrepresentedby the process {st} are such that the right-handside has a finite positive value.l9 We then also obselve that Wts is a predetersined quantityin period t, undele the fiscal regime specified here. Thus if Wts> 0, thereis a unique equilibriumprice level Pt > 0 that satisfies (26). Once we have solved for Pt, (25) then implies a value for Wt+1,given by
19. If not, and if (as we assume) Wts > 0, then no equilibrium is possible. This would representa monetary-fiscalpolicy mix that is inconsistent;in equilibrium,one policy or the other would have to be expected to deviate from the proposed specification at some point. If one supposes that the the primary surplusprocess is unchangeable,this would mean thatpeople would not be able to expect maintenanceof the bill-ratepeg forever.If the "inflationtax"proceeds iL@,i)l(l + i) are increasingin i, and expectedprimarydeficits are too large to be consistentwith the contemplatedsequence {i,}, an increasein the bill rate at some point might solve the problem.On the otherhand,if projectedprimarydeficits are too larae, there might be slo path of bill yields consistent with the {s,} process, which would then necessarily have to be adjusted.We do not take up such cases here, but insteadconsiderthe effects of fiscal news within the class of processes {s,} thatare consistent with the postulatedbill-ratepeg.
Wts+l=(l+it) Wts-Ptst-
'
1 + tt
PtL(yt,it)
(27)
We may then apply the same reasoningin period t + 1, solving (26) for Pt+1, and so on iteratively. We thus solve for uniqueequilibriumprocesses {Pt, WtS}, given an initial (positive) level of governmentliabilities and expectationsregardingthe exogenous processes. The equilibrium process for the price level then implies an endogenous evolution for the money supply, given by (16), and for any other asset prices that may be of interest,given by (17). It might be thoughtproblematicthat the above constructionof an equilibriumrequires that Wt+l turnout to be positive in all periods. But in fact it suffices that the process {st} satisfy boundsthatimply thatthe right-handside of (26) is positive at all dates.Underthis assumption,one can show thatthe law of motion (27) always yields a positive value for Wt+l, given a positive value for Wts. This allows continuationof the constructionforever.The constructedseries must also satisfy (22) in orderfor it to representan equilibrium.However,this simply requirescertainbounds on the exogenous processes {Yt,it}; in particular, it suffices that F(yt, L(yt, it)) be a bounded process. It may also be notedthatno referenceto equilibrium condition(18) has been made in this construction. This might lead to a suspicion that equilibrium is actually "overdetermined" underthe kind of policy regime that has been postulated.But in fact the equilibrium just constructednecessarilysatisfies (18). Note thatif (26), with all time subscriptsadvancedby one, is expectedto determinethe price level in period t + 1, it follows that in period t the conditionalexpectationshould satisfy
Er[X(yo+l,
it+l )P+1 ] =
Et(YT'
iT) ST + l + i
L(YT) iT)
= Ws (X(Yt,it) p -(Yt,
(Yt,it)
(1 + it )Pt
it) st + 1 + i L(Yt,it) )
where the final line uses (27) to substitutefor Wts+l. Thus (18) holds as well. Note the effects of fiscal disturbancesupon the price level in this equilibrium. News thatreduces the conditionalexpectationat date t of cultent and/orfuturevalues of the primarysurplusST, results (other things being equal) in a lower positive value for the right-hand side of (26). As a result, since Wts is predetermined, the equilibriumprice level Pt must rise. Thus fiscal disturbancesresult in variationsin the rate of inflationundersuch a regime. Furthermore, the natureof the effect is consis-
684
tent with the observationthat the outbreakof war in June 1950 (leading to expectations of lower governmentsurplusesin the nearfuture)resultedin an increasein the U.S. price level. This effect of fiscal developmentson inflation cannot really be explained by the fact that the money supply expandswhen the governmentbudget deteriorates(or is expected to in the future).It is true that the quantityequation (16) is satisfied at all times;but the reasonfor the increasein the price level is suppliedby (26), while (16) simply indicateshow much the money supply must expandgiveM thatthe price level the fact thatthe price level may rise (andthe money supply thererises. Furthermore, fore expand) even the reduced surpluses actually materialize,but simply because makes it clear that a mechanicalconnection between the governthey are expected, ment budgetand the monetarybase is not at work. The principle that most directly explains inflation determinationunder such a regime is instead the following: the price level adjustsas necessary to maintainintheor of thepricelevelmakes governmentbudgetbalance.Such afiscal tertemporal the connection between fiscal developments and price-level instability straightforupon ward.The basic economic mechanismis the wealth effect of fiscal disturbances privateexpenditure.The anticipationof lower primarygovernmentsurplusesmakes householdsfeel wealthier(able to afforda greatersum of privateand governmentexpenditure, given theirexpected after-taxincome and given expected governmentpurchases on theirbehalf), and thus leads them to demandgoods and services in excess of those the economy can supply, except insofar as prices rise. A sufficient rise in prices can restoreequilibriumby reducingthe real value of the nominal assets held by households(which, in aggregate,are simply the nominalliabilities of the government). Equilibriumis restored when prices rise to the point that the real value of those nominalassets no longer exceeds the presentvalue of expected futureprimary surpluses, since at this point the (privateplus public) expenditurethat households can affordis exactly equal in value to what the economy can produce. Note that in this analysis, the inflationaryeffects of fiscal disturbancesdo not relate primarilyto changes in expected seigniorage revenues. The fiscal effect of the change in the real valuationof nominal governmentliabilities is also an important consequence of inflation; and this effect may well be the more importantone for high-debteconomies with sophisticatedfinancialmarkets. just describedremainswell definedin the limiting case of Indeed,the equilibrium frictions a "cashless"economy. By this I mean an economy in which the transactions responsiblefor the demandfor cash balances are negligible.20In this limiting case, seigniorage becomes negligible relative to the size of the governmentbudget, and variationsin real balances (in percentageterms) come to have a negligible effect on the marginalutility of income. This means thatthe marginalutility of income may be expressed simply as S(ct + gt), a decreasing function of total (privateand public) simply to tlle value of (interestpurchases; thattotal nominalliabilitiesWtcorrespond earning)public debt;and thatthe primarybudgetsu1plusneed not be correctedto in20. See Woodford(1998a) for a more formal analysis.
clude interestsavings on the monetarybase in the evolutionequationfor government liabilities.Thus in this limiting case, (26) and (27) reduceto
Wt = E T=t DT
tEt
T ST
Pt and
S(YF )
(28)
Wts+ 1 = (1 + it)[Wt -Ptst]
(29)
respectively.This pair of equationscan be solved recursivelyto obtainunique equilibriumsequences {Pt, Wts}, just as in the discussion above. 1.3 An Extensionto Longer-Term Government Debt A similar analysis is possible of price-support regimes with debt of longer duration, at the price of greateralgebraiccomplexity.Here I consider a single, relatively simple case thatillustratesthe main new element introducedby longel termdebt:the fact that Wts is in generalno longer completelypredetermined, as it will dependupon the marketvalue at t of governmentdebt thathas not yet matured.In this simple case, I shall suppose that all govelalmentdebt consists of perpetuitieswith coupons that decay exponentially.Specifically, I suppose that a bond issued in period t pays p dollarsj + 1 periods later,for each j ' Oand some decay factor 0 c p < D 1. The classic "consol"is a securityof this kind, with p = 1. More generally,in an environment with stable prices, the durationof such a bond is (1 _ p) 1.Thus our simple assumptionallows us to analyze bonds of arbitrary duration.At the same time, we need consider the equilibriumprice at each point in time of only one type of bond because a bond of this type thathas been issued k periodsago is equivalentto pk new bonds. Let Qt be the price in period t of a new bond. (Note that the bond's yield-tomaturityis a monotonicfunction of this, given by Qt 1-(1-p).) Now let us consider a price-support policy underwhich the centralbank fixes the price of this bond each period.To simplify the analysis, let us supposethat {Qt} is an exogenously specified deterministicpositive sequence.2lThen arbitrageconsiderations determinea unique rationalexpectations equilibriumsequence for the shorttermnominalinterestrate it, given by
i = l+PQt+l
-1
Qt
21. This assumption still allows us to considerthe effects of a one-timesurprise changein monetary policy, afterwhich householdsare assumedto have perfectforesightaboutthe economy's path.In the case of small enoughrandomfluctuations in the bond-pricetargets,the effects of randomvariationsin bond prices are approximately the same as in this perfect-foresight analysis,butthe extensionis not takenup here.
686
(I assume thatthe bond-pricetargetssatisfy Qt+l > P l(Qt-1) at all times, so that the implied short-terminterest rate sequence satisfies it > 0.) The policy is thus equivalentto a Treasury-billpeg correspondingto this particularsequence, and we may solve for the equilibriumprice level as above. If the public debt consists solely of this single type of bond, the value of total governmentliabilities at the beginningof any period t is given by
WtS = Mts_ 1 + Bt - 1(1 + PQt),
where now Btsdenotes the quantityof the geometricallydecaying bonds outstanding at the end of period t. When p > 0, the dependenceupon Qt means that Wtsis no longer a predetermined variable.Nonetheless, Wtsdepends only upon the predetermined variablesMts_l,Bts_l and the exogenous variableQt. Given the specification of monetaryand fiscal policy from date t onward,and the predetermined values of Mts_ 1,Bts_ 1,thereis a uniquelydeterminedvalue for Wts. Thereis also a uniquelydeterminedvalue for the right-handside of (26), given the uniquely determinedsequence { iT} just discussed. Thus (26) continues to uniquely determine the equilibriumprice level Pt. The money supply in period t is determinedby money demand given this price level,
Mts = PtL(yt it)
while the supply of bonds is then determinedby the government'sflow budget constraint, Bts =
Q t
1[Mts_ l+ Bt -
1( 1 +
Qt)-Ptst-PtL(yt,
it)]
(3 1)
These equationsthen determinea value for Wts+l in the following period, given the exogenously specifiedvalue for Qt+ l.-One can then use (26) to solve for Pt+1,and so on, iterating on the system of equations comprised by (26), (30), and (31). Once again, we assume monetary/fiscal commitmentssuch thatthe right-handside of (26) is positive and finite at all times. Then if we startfrom initial conditionsthatimply a positive value for Wtsat some initial date, the implied price level and the implied value of total governmentliabilities will also be positive at all later dates. Thus the basic logic of price-level determinationremains the same in this case. The main difference that longer-termdebt makes is in the case of an unexpected cha>ge in the sequence of bond-pricetargetsexpected to be maintainedfrom some date t onward.In the case thatall debt is shortterm, Wts is predetermined, and is thus unaffectedby an unexpected change in monetarypolicy (currentor future interest rateexpectations)at date t. A change in monetarypolicy then cannotaffect the price level immediately,except insofar as it affects the presentvalue of futurebudget surpluses (including the government's interest savings on the monetary base). This means that in the case of a high-debteconomy, in which means for economizing on
cash balances are also well developed, the main effect of an increase in nominal interestratesby the centralbankwill be a fasterrateof growthof nominalgovernment liabilities, resulting in faster inflation. (This can be clearly seen in the case of the "cashless limit'' discussed above.) Yet such a result makes it puzzling that in early 1951, the Fed wished to suspendits commitmentto keep interestrates low, in order to containthe increasein prices underwayat thattime. (It would seem instead,under the presentanalysis, that an increase in nominal interestrates would only make the price level grow even fastel:) Allowing for longer-termgovernmentdebt changes this conclusion.A decision to increase targetbond yields lowers Qt, and so lowers the value of Wtsfor any given values Mt_ l, Bt_ l > 0. In the absence of any change in the value of predetermined the right-handside of (26), the increase in bond yields would thereforerequirea decline in the equilibriumprice level Pt. In fact, the effects of interestrate changes on the presentvalue of futuresurplusesare likely to be small;in the "cashlesslimit,"the right-handside of (28) is completely independentof monetarypolicy. Thus in the case of greatestinterest,an increasein bond yields will be associatedwith deflation, initially,thoughit will also lead to faster subsequentgrowthof nominal government liabilities, and consequentlyto a highereventual price level. (It is this expectationof highergoods prices in the futurethatjustifies the immediatedecline in bond prices.) The theoryjust expoundedhas several appealingfeatures as a model of the U.S. regime of the 1940s. Firstof all, it can explain why a regime that bond price-support sought to fix nominalinterestrates was consistentwith relativelystable prices for so many years. Conventionaltheories of interest rate pegs generally imply that such policies should lead to severe price instability.According to the familiar (Wicksellian) view summarizedby Friedman(1968), an attemptto peg nominal interestrates should lead to either an inflationaryor a deflationaryspiral, requiringthe peg to be abandonedbefore long. According to Sargentand Wallace (1975), instead, it should lead to indeterminacyof the rational expectations equilibriumprice level, so that fluctuationsin inflationmay occur as a pureresultof self-fulfilling expectations.The relative stabilityof prices in the 1940s is a puzzle from either point of view. In particular,it is strikingthatpeople continuedto be willing to hold long-termU.S. Treasury securities at low nominal yields (below 2.5 percent per year) during the high inflation(a 25 percentannualrate) of 1946-47; evidently there was temporary regime would generatechronicinlittle fear thatthis indicatedthatthe price-support process." flation,let alone an explosive Wicksellian"cumulative Friedmanand Schwartz (1963, pp. 583-85) hypothesize that people did not expect inflationto continue because previous postwarperiods (such as that following But it is unclearwhy post-warpeWorldWarI) had been associatedwith deflation.22
22. However,they also discuss a mechanismclosely relatedto the one analyzed here, when they discuss the role of the government'sbudget surpluses."Hadthe federal governmentnot run a surplus,the public, with its accumulatedliquid assets and pent-updemand,would have to spend more in the postwar period than it received, . . . [This] would have tended to raise prices and incomes and so would have reroute.... As it was, the federal surduced the level of liquid assets relativeto income by this inflationary plus enabledsome reductionof liquid assets relativeto income to be achievedwithout inflation"(p. 583). However,Friedmanand Schwartzexpoundtheiridea in termsof effects of the governmentbudgeton "the
688
riods should be expected to bring aboutdeflationin the absence of a commitmentto returnto the gold standardat a prewarparity,which there was no reason to expect following WorldWarII. Eichengreenand Galsber ( 1991) insteadproposethatthe policy regime of the early post-warperiodwas actuallyan "implicittargetzone" for the price level, with the price level maintainedwithin the zone by an expectationof interventionshouldthe boundariesever be reached,even thoughlittle interventionwas observedduringthese years. But such a hypothesisexplains the behaviorof the price level in termsof a purelyhypotheticalcommitmentto interventionsthatwere not actually observed;it is hardto see why the public should have had confidence in such a presumedcommitment.The explanationoffered here, instead, depends only upon credibility of the commitment to interest rate targeting (which commitment was being continuouslydemonstrated by the Fed's actions), and beliefs aboutthe exogenous evolution of primarybudget surpluses(which again requiredonly a simple extrapolation into the futureof the policy thatcould alreadybe observed). The model can also explain the variationsovelstime in the degree to which the regime generated inflation, at least broadly speaking. During World War II, the regime was inflationary, thoughmuch of the inflationwas suppressedby a system of price controls, until their removal in 1946. This correspondedto a period of time in which large governmentdeficits turnedout to be necessary that would not initially have been expected.The transitory burstof inflationin 1946-47 represented delayed price adjustment once the wartimecontrolshad been removed,ratherthana demanddriveninflation.Once this adjustment had occurred,the regime was actuallymoderately deflationary in the early postwar period. The model predicts that pegging nominalinterestratesat a low rate,with expectationsof primarysurplusessufficient to make these interestrates consistent with equilibrium,should lead to steady mild deflationas the nominalliabilities of the governmentcontractover time. Finally, inHationtook off again suddenlyin the second half of 1950, following the outbreakof war in Korea.The model explains why such a suddenchange in expectationsregarding the govelmmentbudget should be inflationary. Furthermore, it can explain why the outbreakof war was able to cause inflationeven before any large budget deficits materialized.23 It is the present value of currentand expected futuresurplusesthat mattersin equation(26); because of the crucialrole of fiscal expectationsin this theory, it is completely understandable that the outbreakof war should affect inflation even before it has significantlychangedthe governmentbudget. The model also offers an explanationfor the abandonmentof the price-support regime after March 1951. As just explained, the model implies that such a regime should lead to inflation when previously unexpected deficits come to be anticipated.24 The Fed's complaint that the regime had become an "engine of inflation"
marketfor loanable funds"ratherthan a general-equilibrium analysis in ternasof its effect upon private budgetconstraints,of the kind presentedhere. 23. Note that the U.S. governmentbudget continuedto be in surplusduringthe second half of 1950 (Timberlake1993, p. 313). 24. In the flexible-pricemodel set out above, the price-level increaseresultingfiom any single change in fiscal expectationsoccurs immediately,as soon as informationchanges. However, in a mole realistic
the model (in was justified, given the returnto wartime fiscal policy. Furthermore, the version with long-term bonds) implies that an increase in bond yields should have been able to mitigatethe degree to which prices needed to rise in the shortrun, following the revision of fiscal expectations. Hence the Fed's interest in allowing bond yields to rise above the 2.5 percent ceiling, in orderto contain inflation.The model implies that,in the absence of any change in fiscal expectations,the change in monetarypolicy actuallyimplied greatereventualincreasesin the price level, though policy allowed them to be delayed in time. the suspensionof the bond price-support it Given the inflationthat did in fact occur in the years subsequentto the "Accord," would be hardto call this an inaccurateprediction.25 Finally, the model offers insight into why a policy regime of this kind would be to war finance.Firstof all, the regime is one which loosens appealingas an approach the constraintupon fiscal policy requiredfor consistency with stable prices. Note thatequilibriumcondition(26) must hold in the case of any monetaryand fiscal policies; the right-handside is simply not always a function of purely exogenous variables. This implies thatonly a certainspecific value for the expected presentvalue of future primarygovernmentbudget surpluseswill be consistent with maintaininga price level Pt+j-Pt_1 for all j ' 0. The bond price-supportregime, as modeled in responseto fiscal shocks, here, insteadallows thatpresentvalue to vary arbitrarily within certain bounds. Such flexibility would obviously be quite valuable during At the same time, the regime is one underwhich inflationvariwartimein particular. even when news of governmentbudgetshortfalls ationsare expectedto be transitory; resultsin inflation,people can be confident(insofaras they expect equilibriumto be determinedas describedhere) thatinflationwill quickly returnto a stable (and quite low, possibly negative) long-run level. Such stable long-run inflation expectations would be valuable to a governmentneeding to issue large quantitiesof long-term bonds exactly at a time when (because it has just been learnedthatthe government's fiscal needs are more dire than previously anticipated)prices are currentlyrising. Fromthis point of view, the Treasury'spressureupon the Fed to cooperatewith such a regime duringWorldWarII would hardlybe surprising.
2. RICARDIANAND NON-RICARDIANFISCALPOLICIES
Before turningto the policy implicationsof this view of the effects of fiscal policy to addresssome questionsthatmay arise aboutthe logic on inflation,it is appropriate of the analysisjust presented.One of the most obvious of these is, why should not Ricardianequivalence imply that fiscal disturbanceshave no effect upon aggregate demand,and hence no effect upon the price level?
sticky-priceextensionof the model [expoundedby Woodford(1996)], the price increaseis predictedto be more gradual,and to be associatedwith high outputduringthe period of adjustment. 25. Toma (1997, pp. 109-110) similarly interpretsthe Treasury-Fed"Accord"of March 1951 as a "default"on the government'sprevious "commitmentto long-runmonetaryconstraint"under the bond regime. See also Grossman(1990). price-support
690
The answer is that the usual argumentfor Ricardianequivalence assumesthat changes in the governmentbudget must involve no change in the presentvalueof currentand futurebudgets.(It is asserted,for example, thata currenttax cut financed by governmentborrowingis necessarily accompaniedby the expectationof tax increases at some laterdate or dates, of equal presentvalue.) If this is so, then equation (26) is satisfiedby the sameprice level Pt as would have been the case in the absence of the fiscal disturbance.We would then indeed find that there should be no effect upon the price level of any such event. We have reacheda differentconclusion above because we have assumedthatwhen war breaksout unexpectedly,this news reduces the present value of expected future budget sulpluses. We have thus considered a type of fiscal disturbance thatRicardiantheoryassunces cannotoccu^ Let us call a fiscal policy commitmentRicardian if it implies thatthe present-value relation (24), or equivalentlythe transversality condition (21), necessarily holds for all possible goods-priceand asset-priceprocesses.26 As an exampleof how this could be so, supposethateach periodthe primarysurplusis set accordingto the lmle
P,st
OlWoS-1
MrS,
(32)
for some coefficient O < oc ' 1. This rule states that the primarybudget surplusis chosen to pay off a certainpositivefractionof existing governmentliabilitieseach period, but thatthe requiredsurplusis adjustedto take accountof the government'sinterestsavings on the monetarybase. Using the government'sflow budgetconstraint,
Mts+ Et[Rt t+ l (Wt+ l-Mt )] = Wt -Ptst,

we observe that (32) implies that
Et[Rtt+l Wts+l] = (1-oc) Wt
(33)
each period, and hence that Et[Rt T WT ] = (1 - oc) Wt

26. This differs slightly from the definition of Ricardianpolicy originally proposed in Woodford (1995). Therepolicy was definedto be Ricardianif it ensuredthatthe value of the interest-earning public debt (as opposed to total governmentliabilities) satisfied a transversalitycondition. My reason for the originalproposalwas thatI wantedto arguethatthe Ricardianpostulatewas implicit in standard quantitytheoretic analyses of price-level determination; the definition was thereforetailored to include a policy regime with an exogenous money supply and zero governmentdebt at all times as an example of "Ricardian"policy. Such a regime is not Ricardianin the sense used here, since for price-level paths involving sufficientdeflation,real balances would grow rapidlyenough to violate the transversality condition (21); and this fact can be used to exclude deflationarypaths that would otherwise satisfy all conditions for rational expectationsequilibria(see, for example, Woodford, l999a, sec. 4.2). The definitionof Ricardian policy used here, and in referencessuch as Benhabib,Schmitt-Grohe,and Uribe (2001a), is conceptually preferable,as it defines the case in which the transversality condition ceases to play any role in equilibrium determination.
for all T > t. But this guaranteesthat lim

Et[Rt,TWT ] '
and hence that the transversality condition (21) holds, regardlessof the paths of any of the endogenousvariables.This condition plus the fact that (33) holds at all times can also be used to show that (24) necessarilyholds. In such a case, neither(21) nor (24) places any additionalrestrictionsupon possible equilibriumpathsfor goods prices or asset prices. Rationalexpectationsequilibrium is then defined simply by satisfactionof conditions (16), (18), and (22), along with the equation specifying monetarypolicy. None of the first three equationsinvolves any fiscal variable(such as the governmentbudget or the size of the public debt).Then if the monetarypolicy rule is autonomousin the sense definedabove, the final equationis independentof all such variablesas well, and the complete system of equations available to determinethe equilibriumpath of the price level is independentof all fiscal variables. We thus obtain the RicardianEquivalenceproposition:if monetarypolicy is autonomous, the set of possible rationalexpectationsequilibriumprocesses for goods and asset prices is the same for all alternativefiscal policy specificationswithin the Ricardianclass. If monetary policy suffices to uniquely determine equilibriumin such a case, then a change in fiscal policy does not change the equilibriumpath of prices. More typically,therewill be a set of possible equilibriumprice processes;but as the set is the same for each possible fiscal policy, one might suppose thatthe same equilibriumshould be selected regardlessof fiscal policy.27 We have obtaineda differentresultin the previoussection because we have instead assumeda non-Ricardian fiscal policy specification.In the case of an exogenous real primarysurplusprocess {St}, most paths for the price level and the nominalinterest rate even most of the pathsthatare consistentwith the otherrequirements for rational expectationsequilibrium will not imply dynamics for total governmentliabilities that satisfy the transversality condition (21). Unlike what is assumedin (32), in the previous section we did not assume that the governmentbudget would be automaticallyadjustedin responseto changes in the level of total liabilities, so as to keep the latter quantityfrom growing explosively. The consequence is that only certain price-levelpathswill resultin the transversality conditionnonethelessbeing satisfied; these are those that satisfy condition (26). Hence the latterbecomes a condition for equilibrium, makingfiscal expectationsrelevantto price-leveldetermination. 2.1 Mustn'tFiscal Policy Satisfyan Intertemporal Budget Constraint? The explanation just offered raises questions of its own. It may be doubted whetherit is in fact possible for fiscal policy to be anythingother than Ricardian;if
27. In fact, I shall arguebelow thattheremay insteadbe good reasonsfor fiscal variablesto affect the equilibriumselection in such a case. See section 3.2.
692
not, Ricardianequivalence should indeed hold (and the qualificationabout "Ricardian"fiscal policies may be omitted). A commonobjectionto the logical possibilityof a non-Ricardian fiscal policy is to assel^t thatcondition(24) is nothingbutthe intertemporal budget constraint of the government; it is al^gued thatgovel^nment policy noust be expectedto satisfythis constraint, regardlessof what prices the governmentfaces, just as in the case of privatehouseholds andfirms.It would thenfollow thatfiscal policy mustnecessarilybe Ricardian. It is true that general equilibriummodels always assume that privatehouseholds and firmsoptimize subjectto a set of budget constraintsthat insply an intel^temporal budget collstraint,though they may be even more stringent(as it may not even be possible to bolaow againstall of a household or firm'sexpected futureincome). But it is not obvious that governmentfiscal policy must be modeled as subjectto a similar constraint,for the situation of a governmentis different from that of a private agent in certainimportant respects. Firstof all, if plîvate agentswere allowed to borrow(by issuing debt thatplômises to pay a marketrate of retul^n) without any limit relatedto the amountthat their expected future income should make it possible for them to eventuallylêpay, then an equilibriumwould be impossible. For no plan involving finite amountsof bolaowing and consumptionat each date will be optimal for such an agent;it would always be preferred to borrowand consume even more, simply rolling over the additionaldebt forever.And if demands are unboundedat any prices, there cannot be any marketclearingprices. But there is no similarproblemwith a general equilibriummodel in which governmentpolicy is assumedto be specified by a rule that does not satisfy a colaespondingintertemporal budget constraint.As the example in the previous section shows, one may specify non-Ricardian policy rules that are nonetheless consistent with the existence of a rationalexpectationsequilibrium. Indeed, it is not even necessary,at the level of general principle,for an intertemporal budget constraintof the form (24) to be satisfied i> equilibriun>. The famous overlappinggenerationsmodel of Diamond (1965) describes a situation in which (because the equilibriumreal rate of returndoes not exceed the economy's growth rate)it is possible for a governmentto financetransfersto an initial old generationby issuing debt thatit then "rollsover"forever,withoutever raising taxes. Of course, in the setting assumed above, condition (24) is a requirement for equilibrium, and the Diamond result that it is possible to violate this condition in equilibriumdepends upon a numberof ratherspecial assumptions(even once one has grantedthatpeople are finite-lived),as explained in Santos and Woodford(1997).28But this is a consequence of optimal wealth accumulationby households, not of any constraintupon government bol^rowing programs other than the requirementthat in equilibrium someone has to choose to holdthe debt thatthe governmentissues.
28. Note thatthe possibility of rolling over governmentdebt foreverimplies the possibility of an equilibriuminvolving an asset pricing "bubble"; the same people who hold the governmentdebt in the debt roll-overexample can hold the "bubbleasset" instead.Thus the Santos-Woodford results on the fragility of examples with "bubbles"also apply to the possibility of rationalexpectationsequilibriain which (24) is violated.
Even if we wish to analyze the behaviorof an optimizinggovernment,the government should not optimize subject to given marketprices and a given budget constraint,as privateagents a1ê assumedto in the theoryof competitiveequilibrium.For the government is a laIge agent, whose actions can certainly change equilibrium prices, and an optimizing governmentsurely should take accountof this in choosing its actions. Such a governmentshould also understand the advantagesof committing itself to a rule (given the way thatexpected futuregovernmentpolicy affects equilibrium), and should consider which rule is most desirableby computingthe equilibria that should result undercommitmentto one sort of policy rule or another. Advice to such a governmentwould then involve computingsuch equilibriaunderthe assumption of one rule or another,as an inputto the government'sdeliberationsaboutoptimal policy. There would be no reason to exclude non-Ricardian regimes from the rules that are considered in such an exercise, in those cases where they are in fact consistentwith an equilibrium.29 Thus far I have addressedonly the questionof whethera commitmentto satisfy an intertemporal governmentbudget constraintis a logical necessity, as suggested by authorssuch as Buiter (1998, 1999). A subtlerquestion is whetherit makes sense to suppose that actual marketinstitutionsdo not actually impose a constraintof this kind upon governments(whetherlogically necessary or not), given that we believe thatthey impose such borrowinglimits upon households and firms.The best answer to this question,I believe, is to note that a governmentthat issues debt denominated in its own currencyis in a differentsituationthan from that of privateborrowers,in thatits debt is a promiseonly to delivermore of its own lirlbilities. (A Treasurybond is simply a promiseto pay dollarsat variousfuturedates, but these dollarsare simply additional governmentliabilities that happen to be non-interest-earning.) There is thus no possible doubtaboutthe government'stechnicalability to deliverwhat it has promised;this is not an implausible1êason for financialmarketsto treatgovernment debt issues in a differentway thanthe issuance of privatedebt obligations. Furthermore, no one would doubtthe ability of a governmentto issue an arbitrary amountof currency,without any commitmentto 1êtiring it from circulation(for example, by runningbudget surpluses)at some later date. Marketparticipantsdo not consider whether newly issued government liabilities of this kind exceed some bound on what it is consideredprudentfor the governmentto issue before deciding whetherto accept them as paymentfor real goods and services; instead, each agent makes an individualdecision about the terms on which to accept such government paper,thatdependupon the expectedrateof returnon the asset in equilibrium. An issuance of furthermonetaryliabilities by the government,withoutany increasein the
29. Woodford(1998c, section 5) gives an example of a case in which a non-Ricardian policy regimeone quite similarto our descriptionabove of a bond price-support regime, in fact providesa simple way of implementingthe Ramsey-optimalallocation of resources. [See also Sims (1999) and Christianoand Fitzgerald(2000).] I do not wish to dwell upon this case here, as I do not wish to suggest that such a regime is likely to be a desirablepolicy commitmentin general.But the example illustratesthe point that the mere wish to hypothesizethat governmentpolicy is optimal,from the point of view of some coherent objectivethatthe governmenthappensto pursue,is not in itself a reasonto exclude the possibility of nollRicardianpolicy.
694
: MONEY.CREDIT,AND BANKING
real money balancesthatthe privatesectorwishes to hold, requiresan increasein the price level (reductionin the exchangevalue of the governmentpaper)in orderfor the marketto clear;butthis is a conditionfor marketequilibriumgiven the government's policy, and not a preconditionthatmust happento hold for otherreasonsin orderfor the governmentto be able to create additionalmoney. All this is a familiar way of thinking about monetaryfinancing of the governmentbudget. But what is fundamentally differentaboutthe issuance of interest-earning debt, when this is simply a promiseof futuredeliveryof money? A useful analogy is suggestedby Cochrane(2000) and Sims (1999). Considerthe equilibriumvaluationof the stock of a companythatpays no dividends,and instead distributesits earningsto its shareholdersentirely throughshare repurchases.(The example may seem fanciful, but in fact share repurchaseshave become a more importantsource of distributions to shareholdersin the case of some U.S. stocks, such as Microsoft, and the tax code favors this development.)The correct(beginning-ofperiod) equilibriumshare price qt for such a stock would generally be agreed to be given by the present-valuerelation
qtSt = E E T=t
Et
(Yt )
eT T
(34)
where St is the numberof shares outstandingat the beginning of period t, et is the total earningsof the companyused to financesharerepurchases in period t, and (Yt) is againthe marginalutility of (real) income in period t. (I here assume the "cashless limit,"as is standard in financialeconomics.) The argumentfor this valuation equation is simple. If the earnings stream {et} were insteadpaid out in the form of dividends,the valuationformula(34) would follow from standardtheory. But suppose instead that each period, the stock were to "split"at a rate given by
(7t=
qtSt-et
et
(35)
meaning that duringperiod t, each owner of a share of the stock receives a distribution of (St additionalunits, followed by a repurchase of ( 1 + cyt)etlqt sharesof the outstandingstock. Underthis alternative policy, the total payoutby the companyduring period t is again equal to et (because the price per share after the split is qt/l + (St), and the value of the distribution per shareis again etlSt. The same prices and portfolio allocationsthen continueto describean equilibrium; it should not matterwhether the distribution is called a distributionof stock followed by a repurchaseof exactly the numberof new sharesjust issued, or a cash dividend. On the otherhand,the fact thatthe stock splits at exactly the rate (35) shouldbe irrelevantto its valuation.Whateverthe process {(ST} describingthe expected rate of splittingfor dates T ' t, the equilibriumevolutionof the total value of the company's stock qtSt should be the same, as long as the process {eT} describing the total re-
sources used to financerepurchasesremainsthe same. Thus (34) should continueto apply,regardlessof the splittingpolicy. Note thatfor any given process {eT} specifying the funds availablefor repurchases,and any given process {(5T}describingthe rate of splitting,the evolution of the numberof outstandingshares {ST} is given by the accountingidentity St+l = (1 + cyt)[St-etlqt] (36)
Equations(34) and (36) thenjointly describethe evolutionof the endogenousvariables {qt, St} undera rationalexpectationsequilibrium. Note thatthe equilibrium valuation qt impliedby (34) is necessarilysuch that(36) implies a positive numberof outstanding sharesat the beginningof the following period,so thatthese two equationscan be solved recursivelyforever,yielding a positiveequilibliumshareprice at all dates. One may now observe a formal analogy between equations(34) and (36) for the valuationof the zero-dividendstock and equations(28) and (29) for the equilibrium valuationof nominalgovernmentliabilities(also in the "cashlesslimit").To the variable Stin the stock exampletherecorrespondsWts (the numberof dollarclaims on the government outstanding at the beginningof periodt);to qttherecorrespondsl /Pt (the exchange value of each dollar'sworthof public debt);to et therecorrespondsst (the streamof "earnings" used to retirepublic debt);and to st therecorrespondsit (the rate at which additionaldollarclaims are distributed to the holdersof existing claims). The advantageof consideringthis analogy is that it is clear in the stock case that (34) is an equilibrium condition that determinesthe share price, given earningsexpectations(thatmay well be causally independentof the evolution of the company's stock value), and not a constraint upon possible corporatepolicies. There is no requirement,enforced by the financial markets,that the company generate earnings that validate whatevermarketvaluationof its stock may happen to exist. Indeed, if the company'searningswereto be determined by such a requirement, its equilibrium shareprice would come to be indeterminate, just as the equilibriumprice level is indeterminateunder a bond-pricesupportregime, if the governmentbudget is determined by a Ricardianrule such as (32). The analogyis also deeperthana mere similarityof algebraicform.The economic mechanismthatensuresthat (34) must hold in equilibriumis in fact the requirement that households must exhaust their intertemporal budget constraintsif they are behaving optimally;a stock valuationqtSt in excess of the presentvalue of futurecorporate earnings would imply that households should believe themselves wealthy enough to purchasea streamof goods with greatervalue thanthe economy's product (the source of corporateearnings), which (if households exhaust their budget constraints)will be inconsistentwith goods marketclearing.30 Finally,the stock analogy providesan answerto a common questionaboutthe fiscal theory of the price level: What is special about the government,that its budget
30. Above, I have insteadpresenteda heuristicargumentfor (34) startingfrom the standardpresentvalue theory in the case that all distributionsare cash dividends.But that lattertheory relies upon the requirement that households exhaust their intertemporal budget constraints in order to exclude the possibility of an equilibriumpricing "bubble"; see, for example, Santos and Woodford(1997).
696
shouldbe able to determinethe equilibriumprice level (underpolicy regimes like the bond price-support regiIneanalyzed above), and not that of any other person or organization?Tlle answer is not simply that nationalgovernmentsroutinelyissue liabilities that entitle the holders only to the receipt of furthersimilar liabilities in the future;as we have seen, a privateorganizationsuch as Microsoft could do this as well, in principle. (It would call its liabilities "stock"ratherthan "debt"in such a case.) The other crucial special feature of a national governmentis that prices are commonly quoted in units of its liabilities, that is, in terms of the nationalcurrency. If it happenedthat prices of goods and services were routinely quoted in units of Microsoft stock, say, then it would indeed be Microsoft's budget that would determine the price level, and not that of the federal government. 2.2 Consequencesof a Govers>nent Bolfrowing Linit Finally, even if one supposes that marketsdo impose a limit on how much a governmentcan borrow,it is not clear thatthis invalidatesthe analysis given above of the way in which fiscal developmentsdeterminethe equilibriumprice level undera bond price-support regime. Suppose that there is a finite level of real public debt (the determinationof which we shall not model here) beyond which new debt issues will simply not be purchased.This would mean thatit is not possible for a governmentto refuseto adjustits budgetwhen its debts grow too large;thus a purelyexogenous primarysurplusprocess, as assumedin section 1, would be precluded. For the sake of concreteness,suppose thatthereis an upperbound on the possible end-of-periodvalue of outstandinggovernmentliabilities, so that governmentborrowing must remainwithin the bound Mts+ QtBts d
Pt
at all times, for some finite positive bound d. (For simplicity, I shall assume in the discussion to follow that governmentdebt consists entirely of single type of bond with geometric coupons, the price of which is Qt.) This implies a lower bound st ' wt-d upon possible levels of the real primarysurplus,where wt-WtslPt is the real value of beginning-of-periodgovernmentliabilities. We may similarly imagine that there should be a lower bound on the value of end-of-periodliabilities as well (not so much because the privatesector will not allow unlimited governmentlending, but because we may suppose that governmentswill never actually be so generous); for simplicity,let us suppose thatthis is zero. This would imply an upperbound upon feasible primarysurplusesas well, st ' wt, the value that would leave the governmentwith no net liabilities at the end of the period.3lThen the fact that total governmentliabilities must satisfy the bounds
31. As noted earliel; this bound would alreadyimply some governmentlending, but the government would discountprivateobligationsonly in the a1nount that the centralbank could hold as backing for the monetarybase.
oc Mt +QtBt d
(37)
implies that the primarysurplusprocess {st} would have at all times to satisfy the bounds
Wt-d C St C Wt * (38)
Condition(37) implies that in any equilibrium
-Et
U171(YT, n>T)nT
CE
Et
A(YT, iT) 1 + i
T(
i )
MT + QTBT
<
Et A(YTHiT) T
QTBT
' i A(YT, iT )d
Furthermore, the thirdterm in this series is equal to
Et A(YT+ 1, iT+ 1)
'
(39)
T+1
so that expression (39) must be bounded above and below by the initial and final terms in the previous series. But equilibriumcondition (22) implies that the initial term must convergeto zero as t is made unboundedlylarge; and for all interestrate paths {iT} satisfying bounds sufficient to imply that {X(YT iT)} is uniformly bounded,32 the final term must convergeto zero as well. Thus (39) must convergeto zero as Tbecomes large as well, and the transversality condition (21) is necessarily satisfied.It follows that any fiscal policy that satisfies the bounds (38) at all times is essentiallyRicardian;33 and condition (24) places no restrictionsupon possible equilibriumpathsof the price level.
32. This assumptionis certainlynot problematicin the case of a bond price-support regime which implies boundson interestrates that are independentof the evolution of goods prices. 33. The qualificationis that the transversalitycondition has not been shown to hold for allpossible price-leveland interestratepaths, but only those that satisfy certainbounds on interestrates and the equilibriumcondition(22). But this near-Ricardian propertysuffices to imply thatthe transversality condition, or alternatively condition(24), places no additionalrestrictionsupon the possible equilibriu1n pathsof the price level, given a monetarypolicy that maintainsinterestrates within the assumedbounds.
698
Nonetheless, fiscal disturbancesmight well affect the price level. To see why, let us consider a modified version of our previous analysis of the bond price-support regime. Suppose that the government's"desired"real primarysurplus evolves according to some exogenous stochastic process {st} as assumed earlier,and that the actualbudgetsurplusequalsthis, except when one of the boundsin (38) would be violated; in the lattercase, the real primarysurplusis equal to the bound. Thus under this fiscal regime,
( q
St = S(Wt<SF)
wt if Wt < SF St if st ' wt _ st + d d if
WF > SF
lwt
Otherwise, policy is as assumed earlier: there is a single type of interest-earning governmentdebt (with geometric coupons), and monetarypolicy is specified by an exogenous bond-pricesequence { Qt}. Note that as long as the fluctuationsin {st} are small enough [or the bounds (38) are loose enough], the equilibriumpresentedearlier is still a rationalexpectations equilibriumunder the modified specification of fiscal policy. This is because the equilibriumdescribed earlier does not involve explosive growth of the real public debt. If the bounds (38) would not have been violated in equilibriumin any event, stipulatingthat fiscal policy must respect those bounds does nothing to exclude an equilibriumof thatkind. Thus one still can have an equilibriumin which fiscal disturbances affect the price level; this is simply no longer the onlyequilibrium.However,it continuesto be a locallyunique equilibrium,in the following sense: it is the only equilibriumin which the state variable {(Mts + QfltS)lPt} remains forever in the interiorof the interval (37). Thus there is no other equilibrium"near"this one in the sense of involving nearbyvalues for all endogenousvariables(includingthis one) at all times. This concept of local uniquenessor determinacyof equilibriumsuffices, for example, for the usual sorts of "comparative statics" analyses of the effects of perturbations of the model (see, for example,Woodford1998a, 1999a). If we are simply interestedin the existence of a determinateequilibriumin which fiscal disturbancesaffect the price level, then the global specification of the fiscal policy rule does not matter:all thatmattersis how the government'sbudgetwould be differentin the case of price-level and asset-pricepaths that are close to the equilibriumpaths.It thus only mattersthatfiscal policy be locallynon-Ricardian:34 thatthe dynamicsof the public debt be locallyexplosive, for most price-level paths near the
34. This is what Leeper (1991) calls "active"fiscal policy; his definition is for a specific parametric family of fiscal policy rules, but his more generalintent seems to be the one identifiedhere. Note that his analysis of equilibriumis purelylocal, in thathe relies upon linearapproximations to the equilibriumconditions, and considers only nonexplosive solutions. See Woodford(1998a) for furtherdiscussion of "locally Ricardian" and "locally non-Ricardian" policies.
equilibriumpath.The existence of debt limits such as (37), that eventuallyconstrain the growth of the public debt in the case of paths far from the equilibriumin question, do nothingto interferewith this. Despite this conclusion, under fiscal policy of the type just described, Ricardian Equivalenceholds, when properlyunderstood.The set of possible state-contingent equilibriumpaths is the same regardlessof the evolution of the desiredprimarysurplus process {st}. But under this policy regime, the set of equilibrium price processes is quite large;among the possibilities is a solution in which the price level happensto fluctuateat the same time as unexpectedchanges in the desired primary surplus. In orderto illustratethe multiplicity of possible equilibria,it is useful to further specialize our example, and considerthe case in which s t = s > Oforever,Qt = Q < (1 _ p) l forever,andYt= Y > Oforever.Let z > Obe the constantnominalinterest rateimpliedby the bond-pricetarget,and m-L(y, z ) be the implied stationaryequilibriumlevel of real money balances.We assume that > O, (40)
fl > 1
D S
+-m
so thatan equilibriumof the kind describedearlierexists, andremainsforeverwithin the interiorof the bounds (38). To simplify, let us consider simply the set of perfect foresight (deterministic)equilibria(p.f.e.) consistentwith such a regime. Given an initial condition WOs-Ms l + QBs l > O,any path for {wt} satisfying the differenceequation
Wo+l = t
Wt-S(Wf,
S)-m
( 1+ )
(41)
for all t ' O,and the transversality condition lim twt = O,

F4
(42)
representsa p.f.e. The equilibriumpath of the price level correspondingto any such solution is given by
Pt =
[t(1 + z )]t .
wO
A graphof the right-handside of (41) is shown in Figure 2. Under assumptions (40), there are threepossible steady-statesolutions to (41), given by
W* w _
......................................
........................ ....
"
.................................. / F / / /
'
..'
.....
X,
700
Wt+
A,
45
/e
pv
/
! j; t
,''.
,
fS'
1/
.
q
'
/
*
,
.
/
/ /
//f
/ / ' . b 6 . / / s / / ' . 4 .
/ / /
' * ::
'
* ; : :
' q
'
',
'
>
Wt
WOW1 W2
W3
W4
FIG.2. PerfectForesightDebt Dynamics
w---
m O,
< 1+1
wt-1-
s+t-8m
\1+1 J
>O
w-8
d-t-8m
\1+1 J
>w8.
In additionto these three solutions with a constantlevel of real governmentliabilities, there exists a continuumof nonstationarysolutions to the difference equation. In particular, for any choice of wO> O,the sequence can be continuedforever.(An example of a nonstationary solution with wX < wO< w is shown in the figure.) Because the implied sequence {wt} is necessarilybounded,the transversality condition (42) is necessarilysatisfied,and as long as wO> O,the implied price sequence is forever positive. Thus this entire continuumof solutions representsalternativepossible p.f.e. undersuch a regime.
One of these solutions, the one with WF = W I forever,is the fiscalist equilibrium discussed earlier.Previously,when we assumedthat st = s at all times, this was the only possible equilibrium;in that case, the correspondinggraphwould continue the steep center segment off indefinitelyin both directio1ls, and all solutions to the difference equationstartingfrom w0 7& w* would be explosive, and would violate (42). Withthe assumedboundson governmentliabilities, this is no longer true.For example, anotherequilibriumis the one shown in the figure,in which the price level is initially (andforeverafter)lower thanin the fiscalist equilibrium.This lower price level is sustainedas an equilibriumby people's (correct)expectationsthat the exploding public debt will eventually lead to a fiscal consolidation,following which primary surplusesare increased.Anticipationof this leads to feel less wealthy, so that lower prices are requiredfor goods marketsto clear. But the fiscalist equilibriumis still the only equilibriumin which neitherbound in (37) is ever binding; thus this equilibriumis locally isolated, as mentioned above. One may also observe that none of the otherp.f.e. are locally unique;corresponding to any equilibriumlike the one with an exploding public debt shown in Figure 2, there exist an infinite numberof other equilibriaarbitrarily close to it (in the sense thatthe price level and othervariablesare nearlythe same, but not exactly the same, at all times as in this equilibrium).Slightly differentexpectationsaboutthe size and timing of the eventualfiscal consolidationare equally consistentwith equilibriu1n. Given the existence of a multiplicityof possible self-fulfillingexpectations,an obvious questionis whetherthe fiscalistequilibrium remainsa plausibleeqleilibrium selection, undera regimeof the kindjust described. This is presumably whatMcCallum (1998) means to challenge, in arguingthat the fiscalist equilibriumis a bubbleequilibrium." Howeve1;what constitutesa "bubbleequilibrium" is often in the eye of the beholder;one might at least as easily say thatthe equilibriumshown in Figure 2 is a "bubbleequilibrium," as the higher value of the public debt is sustainedby self-fulfilling expectationsof a future fiscal consolidation.Here I present an argume1lt for why expectationsmight naturallycoordinateupon the fiscalist equilibrium.35 Essentially,I would argue that the fiscalist equilibriumis an especially plausible focal point for households' expectations,because it involves simple^ fiscal expectations than the otherpossible equilibria.More formally,one may note that these particular expectations are ones that households could converge upon through an Evans-Ramey(1998) process of "expectationcalculation." Evans and Ramey define an intuitivelyplausibleform of iterativerefinementof expectations,in which people make use of their knowledge of a true model of the economy to decide what to expect should happenin equilibrium.The process is one which converges if it converges to a rationalexpectationsequilibrium(r.e.e.);Evans and Ramey are instead interestedin the kind of approximate r.e.e. thatmay resultfrozniteratingonly a finite numberof times, owing to "calculationcosts."One might alternativelyuse the same
35. It is worth noting that, in the presentexample, there is no "monetarist equilibrium" (the selection principlethatMcCallumwould insteadprefer)because of the endogeneityof the money supply.Rationalexpectations monetaristanalyses generally argue that the price level should be indeterminateunder a monetarypolicy of the kind consideredhere.
702
analysis to considerthe stabilityof alternativer.e.e. underthe calculationdynamics, and use this judge whethera given equilibriumis likely to be reachedin practice. Such an analysis begins with a mapping of expectations into equilibrium outcomes. To simplify,I shall suppose thatthereis no uncertaintyaboutthe deterministic paths of real and nominal interest rates, as these are the same in each of the continuumof perfectforesightequilibriajust described;I shall only considerthe coordinationof expectationsupon one expected path or anotherfor the primarybudget surplus {st}. In the model just considered, a household's optimal plan for the sequences {CT + gT, mT} at dates T ' t depends solely upon its estimate of its "total wealth" W1/PF + Ht, where "humanwealth"is defined as
00
Ht - , 13
T=t
[Y-ST]
(Here we have included governmentpurchasesas partof the household's "income" net of taxes, as they substitutefor privateexpenditures.) Specifically,in the contextof a constantreal rate of returnon (nonmonetary) savings equal to ,B-1-1, and a constantnominalinterestrateI, the optimalplaninvolves a constantlevel of total (private plus public) purchases,CT + gT = yd, and a constantlevel of real money balances, mT= L(yd,z ), for each T ' t, whereyd is the highest level of total purchasesconsistent with the household'sintertemporal budgetconstraint, thatis, the solutionto
1
yd
L(yd
l)
= _
HT .
In the case of arbitraryfiscal expectations, I shall assume that the representative household chooses consumptionct = ytd-gt and money balancesMt = PtL(ytd, l) where the values f gt and Pt are observed by the household at the time that it decides, and ytdis the solution to (43), given the observedvalues of Wtand Pt, and the household's subjectiveestimateHte of its humanwealth HF. If each householdhas a common (thoughnot necessarilyrational)expectationHFe, then goods marketclearing requiresthatytd= y. This occurs if and only if the price level satisfies
-+
H,e = x-
y +-L(y,
I)
(44)
Corresponding to any sequence36 of expectations {Hte}, the correspondingsequence

36. As usual (see, for example, Grandmont1983), certainbounds upon expectations are requiredin order for a temporaryequilibrium to exist. In the present context, equations (45)-(46) below imply uniquely defined temporaryequilibriumsequences with P, > Oand W,s+l> Oeach period, startingfrom
of temporary competitive equilibrium (t.c.e.) paths for the price level is given by (44), where Wts evolves accordingto
Wt+l = Pt[(1+ 1)(x-Hte _ SF)-I-L(y,l)],

and the primarybudget surpluseach period is given by st = s(WtIPt, s )
(45)
(46)
In this way, an arbitrary evolutionover time of beliefs aboutfuturegovernmentbudgets (summarizedby the evolution of {Hte}) gives rise to a sequence of actual governmentbudgets [the t.c.e. sequence {st} given by (46)]. We now consider what expectationsa household might have which understands the above model, and hence the mappingjust described.Evans and Ramey propose thata household should begin by conjecturinga pathfor the economy (for example, the t.c.e. dynamics generated by some simple expectational hypothesis such as "adaptive expectations"),and then refinethis conjectureby iteratingon the mapping just derived.That is, they propose that a household should form its ownforecast of the economy's evolution by assuming that the beliefs of otherswill evolve in the conjecturedway, and then calculatingthe t.c.e. dynamics that should actually occur given thatothershave beliefs of the simplersort. (The process might then be iterated again, if the householdinsteadassumes that otherhouseholdswill form theirbeliefs by iteratingonce, and so on.) Any of the sequences {st} associatedwith one of the continuumof p.f.e. identified earlieris a fixed point of this mapping;thus the proposedmethodof mappingexpectations into t.c.e. outcomes does not exclude any of these possibilities by itself. And the beliefs obtained(even if one imagines iterationan infinite numberof times) obviously dependupon what the initial conjectureis. But we may still regardbeliefs as more likely to be convergedupon if they have a larger"basinof attraction" underthe expectation-calculation dynamics(thatis, iterationof the above map). It is thus noteworthythatthe fiscalist equilibrium(corresponding to the expectationthatST = S- for all T ' t) is the onlyp.f.e. with the propertythatexpectationsconvergeto it underthe calculationdynamics,from any nearenough initial beliefs. Specifically,supposethat one initially conjecturesan evolutionof fiscal beliefs such thatthe sequence {HTe}is expected to satisfy the bounds
H< HTe < H

for all T ' t, where
(47)
any positive initial level of nominal government liabilities, as long as expectations satisfy the bound
Hte< H each period,where H is defined in (48) below.
704
H-X-S-d H-X-S1
, L(Y,I).
+ w
(48)
Then (44) implies thateach period,the t.c.e. price level will satisfy
s <-<s PT
+d
so that neitherbound on governmentliabilities ever binds, and ST= S each period. One thus obtainsconvergenceto the beliefs associatedwith the fiscalist equilibrium, algorithm,startingfrom after even a single iterationof the expectation-calculation any initial conjecturesatisfying the bounds (47).37 an initial conjecturesatisfying (47) would not be implausible. For Furthermore, example, suppose a household generates its initial conjectureby assuming that the beliefs of others will evolve accordingto an "adaptiveexpectations"formula.Since Hteis an estimateof (1 _ )-1 times a certain(discounted)long-runaveragevalue of Y-ST one might assume thatpeople will adjusttheirestimatein responseto the discrepancybetween their currentestimate and the most recently observedvalue of the variablein question.This would suggest a rule of the form HT+1= SHT+ (1 - A)(1 - )
= XHeT + (1 )(1 -
(Y - ST)
S(X -
13) 1 (y -
HT S)).
Combiningthis law of motion for expectationswith (44)-(46), one generatesa t.c.e. in which expectationssatisfy the bounds (47) forever,as long as one conjecturesthat expectationswill initially startfrom a value of Htesatisfying the bounds. (The same would be truefor a large numberof other simple adaptiveschemes.) Thus it is easy to describereasoningthat would lead households who understand the model to convergeupon expectationsthatwould bring aboutthe fiscalist equilibrium,ratherthanany of the others.But if one accepts thatthis should be equally true following a fiscal disturbance(say, a one-time permanentchange in the value of s, correctly understood by the households who form their beliefs according to the should affect the price level in the way deabove algorithm),then such a disturbance scribedby the fiscal theory,and contraryto the doctrineof RicardianEquivalence.
37. On the other hand, no other p.f.e. has the propertythat its basin of attractionincludes any neighborhoodof the equilibriumbeliefs themselves.This is because no otherp.f.e. is locally isolated, and every mapping. p.f.e. is a fixed point of the expectation-calculation
2.3 EmpiricalEvidence on the Characterof Fiscal Policy
Thusfarourprimary concern hasbeenwiththelogicalpossibility of a non-Ricardianfiscalregime. It is worth asking as well whether actual fiscalpoliciesseemto be of thiskind.Thishas beenchallenged in a number of studies, though on the whole there hasbeenlittleempirical analysis relative to theamount of recent theoretical interest in regimes of thiskind.38 It will takeus toofarafieldto address theissuein any detailhere;buta few comments on the relevance of the empirical literature to our discussion maynonetheless be appropriate. Several studies havesought to showthatfiscalpolicyis Ricardian, at leastin particular placesandtimes,suchas the UnitedStatesin the twentieth century. Forexample,Bohn (1998a)regresses the primary U.S. government budgetsurplus (as a shareof GDP)on privately heldU.S. publicdebt(againas a shareof GDP),along withothervariables thatareregarded as plausible arguments of a fiscalpolicyrule, andfindsa significantly positivecoefficient (witha pointestimate on the order of 3 to 5 percent) foreachof thesample periods thathe considers. Bohninterprets his resultsas estimates of thecoefficients of a fiscalpolicyrule,andin particular he interpretsthe regression coefficient on the debt/GDP ratioas indicating the degreeof feedback fromthesizeof thepublicdebtto thesize of theprimary surplus, thatis, as the coefficient otin a feedback ruleof the form(32). If thisis correct, thenourreasoningabovewouldimplythatfiscalpolicyoverthis periodwas Ricardian, as the transversality condition wouldholdforall inflation andasset-price processes.39 However, sucha regression coefficient neednothavea structural interpretation of the proposed sort.In the case of an exogenous primary surplus process a simple example of a non-Ricardian regime the existenceof an equilibrium present-value relation between the valueof thepublicdebtandexpected future primary surpluses wouldimplythatvariations in the valueof the publicdebtrelative to GDPwould tendto forecast subsequent variations in primary surpluses relative to GDP. Thereason is the sameas the reasonthat,according to the present-value theoryof stock prices, theprice-earnings ratioof a stockshould forecast future earnings growth; yet no one wouldinterpret econometric support forthelatter prediction as a demonstrationthathighp/e ratioscause companies' earnings to growfaster. Similar alternative interpretations arepossible in thecaseof themoresophisticated "tests" forRicardian as opposedto non-Ricardian policy presented by Canzoneri, Cumby,and Diba (1999).4 Indeed, as Cochrane (1998)andWoodford (1998b)demonstrate, it is possible to proposenon-Ricardian policy regimes,underwhichthe inflation process evolvesprimarily in response to fiscaldisturbances, whichmatch quitewell thejoint stochastic processes of inflation, interest rates,thepublicdebt,andtheprimary budget surplus for the UnitedStatessince 1960.Thusthereare no aspectsof the co38. Two noteworthyearly attemptswere Shim (1984) and Leeper (1989). 39. Bohn (1998b) suggests exactly this interpretation of the results reportedby Bohn (1998a). Note that the analysis above does not apply precisely to the rule estimatedby Bohn because his estimatedrule does not have a termrespondingto variationsin the interestsaved on the monetarybase; but it is unlikely thatthis mattersmuch for the validity of his conclusion. 40. See Cochrane(1998, 1999) and Woodford(1998b) for detailed discussion.
706
movements of thoseseriesthatareincompatible withRicardian regimes(andfiscal determination of thepricelevel)as such. A general difficulty witheconometric testsof whether policyis Ricardian is thatin the modelexpounded above,the present-value relation(24) musthold in equilibrium,w11ether policyis Ricardian O1 not.Thustheempirical question is notwhether or not sucha relation appears to hold in practice; it is whether the fiscalpolicy is suchthatit wollld alsohaveheldin the case of otherpricepathsthathavenotactually occurred, andperhaps couldnot everoccur. Thisis inherently a difficult questionto tryto answer, andit plainly cannot be answered in theabsence of assumptions about the structural formof thepolicyrulesin place.(Thisdoesnot,however, make it anyless meaningful an empirical question thantheothercounterfactual questions thateconometricians habitually address, whenestimating elasticities of demand and the like.) Furthermore, simple structural specifications like the one proposed by Bohnarenottoopersuasive in thiscontext, owingto thefactthattheprocess thatdetermines thesize of government budgets clearly involves considerable lags.It is thus quiteplausible thatmarket participants will oftenhaveadditional important information aboutlikelyfuture budgets, beyondthatcontained in a smallnumber of variables like those includedin Bohn's regression;but then the possibility that a regressor suchas the current debt/GDP ratioresponds to suchinformation mustbe takenseriously. Bohn(1998b)suggests thathis proposed structural interpretation of his regression coefficient is moresensiblethanthe alternative suggested by Cochrane (1998)becauseit is basedon "acoherent modelof government behavior," namely, a modelof optimal tax smoothing. But whilethe theoretical modelto whichBohnrefersmay justifythe inclusion of certain regressors in his fiscalrule,it does not provide any theoretical justification fortheassumption thatno otherinformation canhelpto forecastfuture budgets, andit is this exclusion restriction thatis crucial to Bohn'sproposedidentification. Nor is it reallytruethatthe modelof optimaltax smoothing provides theoretical justification forassuming a fiscalruleof Bohn'sform(inparticular,one in whichthe primary budgetsurplus is a function of the size of the public debt).Themodelof optimal tax smoothing justifiesa particular pattern of state-contingentgovernment budgets, but thereis no unique formof feedback rule for the government budget thatimplements the desired equilibrium. (Onceagain,different specifications of government policy out of equilibrium may be equallyconsistent with a givenequilibrium.) Forexample, Woodford (1998c)showshow an optimal tax-smoothing policycanbe implemented by a regime thatcombines a non-Ricardian fiscal policy (specifically, one in whichthe realprimary budgetsurplus is exogenous)withaninterest ratepegformonetary policy exactlythetypeof regime analyzedin theintroduction of thispaper. An alternative possiblesourceof information aboutthe likelyformof the fiscal policyrulecan be obtained fromestimates of the monetary policyrule,combined withdeductions about thekindof equilibrium behavior thatshould result fromalternative monetary/fiscal policyrulecombinations. Forexample, John Taylor (1993)arguesthatU.S. monetary policysinceatleastthelate 1980shasbeencharacterized by
an interestratefeedbackrule according to whichincreasesin inflation cause increases in thefederal fundsrateof anevenlarger magnitude, anda largesubsequent empirical literature hasreached similar conclusions. However, thissortof policyrule whichLeeper (1991),calls"active monetary policy," wouldimplyexplosive debtdynamicsin thepresence of an exogenous realprimary surplus process, or othersimilarlocallynon-Ricardian fiscalrule.4l (Wediscussthisfurther in thenextsection, in connection withLoyo's(1999)analysisof Brazil'spolicymix in the early1980s.) Thefactthata Brazilian-style debtexplosion andhyperinflation didnot occurin the UnitedStatesfollowingthe shiftto a monetary policy similar to the "Taylor rule" impliesthatU.S. fiscal policy has been locally Ricardian, at least since the late 1980s.42 Thisalsoconforms withwhatwe knowabout theemphasis uponthegoalof budget balance in discussions of U.S. fiscalpolicysincethelate 1980s. Ontheother hand, Taylor (1999)argues thatFedpolicyin the 1960sand1970sinvolvedan interest rateresponse to inflation increases thatwas substantially weaker, andin particular failedto raisenominal interest ratesby as muchas the increase in inflation.Such a policy, which Leeper(1991) calls "passivemonetary policy") wouldnot implyexplosive debtdynamics whencombined witha locallyRicardian fiscalpolicy; indeed, it is onlyin thecaseof sucha fiscalpolicythatthepolicyregime wouldimplydeterminate equilibrium paths fornominal variables, including thepath of the publicdebt.43 Clarida, Gali,andGertler (2000) similarly findthattheirestimated Fedpolicyrulefortheperiod1960-1979wouldimplyanindeterminate equilibrium, under their(implicit) hypothesis of Ricardian fiscalpolicy;butonce again (andfor similar reasons) theirestimated policywouldbe consistent with a unique nonexplosive equilibrium under a hypothesis of locallynon-Ricardian fiscalpolicy. Theempirical relevance of the hypothesis of "passive" monetary policyon thepart of theFedis evenclearer forthe 1940s,as discussed above.Onceagain,thiswould at leastmakeit possibleto reconcile anapparently stableinflation process witha hypothesis of non-Ricardian fiscalpolicy;andindeed, we haveargued thatthishypothesisprovides anappealing explanation of several aspects of inflation variation during thatperiod. It is important to recognize thattheanalysis presented aboveof equilibrium determination under thebond-price support regimeis notoffered as a universal theory of
41. Leeper(1991) andLoyo (1999) show this in the case of flexible-pricemodels. However,the analysis in Woodford(1996) of determinacyof equilibriumin a model with sticky prices (Calvo-style staggered pricing) shows that the same classification of policy regimes consistent with unique non-explosive debt dynamicscontinuesto apply in thatcontext. The only differenceis thatthe demarcationbetween "active" and "passive"monetarypolicy rules must be modifiedto take accountof the effects of feedbackfrom the level of outputto the centralbank's interestrate instrument,as in the discussion of the "Taylorprinciple" in Woodford(2000). 42. Of course, such a judgment depends upon believing the identifying assumptionsthat are made in estimates of Fed reactionfunctions;but these are arguablygroundedto a greaterextent in knowledge of the factors actually consideredin Fed decision making than is the case with attemptsto estimate a fiscal policy rule. 43. In the case of a model with nominalrigidities, indeterminacy of the paths of nominalvariabilities would imply indeterminacy of the equilibriumpathsof real variablesas well. The determinacyof equilibriumundera policy rule of the kind thatTaylorestimatesfor the 1960s and 1970s, when combinedwith a locally non-Ricardian fiscal policy, is shown by Leeper ( 1991) for a flexible-pricemodel andby Woodford (1996) for a sticky-pricevariantof that model.
708
(24) shouldalwayshold,rerelation Whilethe equilibrium determination. inflation it mayor maynotbe policyregime, of the monetary/fiscal of thecharacter gardless andfiscaldisdetermination, as the key to price-level usefulto view thiscondition effect uponinflation these latter may or may not have an important turbances thefacttllat However, of thepolicyregime. uponthenature will depend propositions fiscalpolithepossiblekindsof systematic areamong policyregimes non-Ricardian is evenif one does notthinkthatU.S. fiscalpolicy(forexample) cies is important, upnext. to be taken of thatkind,forreasons culorently
FOR INFLATIONCONTROL 3. IMPLICATIONS
that forthedesignof publicpolicyof a recognition I nowturnto theimplications Connot,of course,a necessity). arepossible(though fiscalregimes non-Ricardian sorts.In takingthemup, I of several hasconsequences of thispossibility sideration price of as stablea general thata keygoalof policyis themaintenance shallassume be a goalis left thisshould orto whatextent, whether, thequestion levelas possible; occasion.44 foranother to be policyis expected budgetary Firstof all, in the case thatthe government's policyruleis choosinga monetary thata policymaker non-Ricardianfor leeasons be policyrulesshould to change thisfactaffectswhichmonetary notin a position Rulesthatwould of pricestability. degree withthegreatest to be consistent expected fiscalpolicy,suchas a "Tayof a (locally)Ricardian in thecontext be quitedesirable wllencomfor pricestability consequences may insteadhavedisastrous lor rule," fiscalpolicy. binedwithanalternative to an But this very fact impliesthatthe choice of fiscalpolicy is also relevant of policy andso oursecondcategory pricestability, of achieving chances economy's with thechoiceof a fiscalpolicyrulethatwouldbe consistent considers implications pointis thatfiscalpolicyshouldbe locallyRicarHerethe essential pricestability. bank'suse of a suitably the central do not frustrate dian,so thatfiscalexpectations thepricelevel. policyto stabilize monetary "active" can maketo thata suitablefiscal policy commitment Finally,the contribution equilibwitha desirable of failingto interfere is not simplya matter pricestability policy with the centralbank'smonetary be consistent riumthatwouldotherwise maybe fiscalcommitment locallyRicardian) (though non-Ricardian rule.A globally less stableprices, ones involving equilibria, usefulin orderto excludeundesirable policyregime.I takeup each withthe monetary be consistent thatwouldotherwise in sequence. of implications of thesecategories
used above is one in which there 44. One reasonfor doing so is that the simple theoreticalfranaework are no frictions of the sort that imply that any economic distortionsshould result from unexpectedvariation in the equilibriumprice level; in reality,such frictions are important.See Woodford(1999b) for discussion of the welfare consequences of inflationvariabilityin some simple models with sticky prices.
3.1 Monetary PolicyChoiceWhen FiscalPolicyIs Non-Ricardian If fiscalpolicyis expected to be non-Ricardian, thesefiscalexpectations constrain thesetof possibleequiliblium outcomes thatmonetary policycanachieve. Thisconstraint mayfuthermore haveimportant consequences for the ranking of alternative monetary policies.Inthepresence of non-Ricardian fiscalexpectations, thechoiceof a monetary policythatis intended to be anti-inflationary maylead(at leasteventually)to evenmoreinflation. Indeed, it couldevenresultin a hyperinflation, as Loyo (1999)argues occurred in Brazilin the 1980s. Consider, for example, the consequences of a central bankcommitment to set a short-term nominal interest rateinstrument according to a "Taylor rule"45
it= (IIt)'
whererlt-P/Pt_l iS the gross rateof inflation, and@(n)is an increasing, nonnegative function, consistent withanimplicit target rateof inflation Ilt > , thatis, suchthat 1 + @(Tl)= -lI I shallassume thatthereis a unique Ilt > > satisfying (50). Letus againconsider a deterministic environment forsimplicity, andsuppose that output is constant andequalto y > 0 eachperiod. Givena commitment to themonetarypolicyrule(49)perfect foresight equilibrium requires thattheinflation sequence { Ilt} satisfy thenonlinear difference equation
+l ((t+l )' Y) = :(1 ((t + (t )) )' Y)
(51)
obtained by using(49) to substitute for it in (18). In the "cashless limit," or thecase of additively separable preferences, (51) reduces to
Ilt+ 1 = (1 + (t)) *
(52)
Ouranalysis of the qualitative properties of this difference equation is simplified if (following Loyo)we restrict attention to thelatter specialcase. If, in accordance with Taylor's (1993) characterization of U.S. policy since the late 1980s,we assume thate<)(IIt) > 1, where e<)(II) is theelasticity of 1 + @withre45. The rule discussed by Taylor (1993) also involves feedback from deviations of real output from trendand/orpotential.But In the presentflexible-pricemodel, thereare no deviationsof outputfrom potential,and any deviationsfrom trendareexogenous, so thatsuch feedbackwould representat most an exogenous shift term in (49), with consequences identical to those of time variationin the inflationtarget. Such an extension would not change our conclusions about the stability of inflationdynamics underthe policy rule.
710
spectto [I, then the graphof the right-hand side of (52) cuts the diagonalfrom below,as shownin Figure3. As the figureshows,in this case [It is an "unstable" steadystateunderthe dynamics impliedby (52). (Thefigureillustrates a solution starting fromaninitialinflation rateflo > [It: in thiscase,inflation mustgrowwithoutbound overtime.If instead onewereto assume [Io < [It, inflation wouldhaveto be forever declining, eventually leading to permanent deflation.) Thismeansthatthe onlysolution to (52)in whichflt remains withina neighborhood of [It forever is the onein which[It = [It forall t. Thusthetarget steady stateis notonlyconsistent with policyrule(49)but,if (52) is theonlyrestriction uponequilibrium inflation, therule also makesit a determinate (locallyunique) equilibrium.46 Onemightthenbe optimisticthatpeoplewouldin factsucceed in coordinating theirexpectations uponthat equilibrium, so thatthe"Taylor rule" wouldsucceed in stabilizing inflation atthedesiredrate47
l. t+1 A
(1+f(n))
45o
rlF
.........................
t1:t
tIO 1
FIG.
tI2 rI3
3. An Inflationary Spiralundera TaylorRule
46. Here it is important to note thattIt iS not a predetermined statevariable,so thathistoryprovidesno initial conditionfor the differenceequation(52). Instead,the variableis free to "jump"so as to be consistent with the expected futureevolution of inflation. 47. This is the sort of analysis of inflationdetermination undersuch a rule proposed,for example, in Woodford (1999a). See section 4 of thatmanuscript for discussion of otherpossible equilibria.
Butis thisin facttheonlyrequirement fora perfect foresight equilibrium? Theanswerdepends uponthe character of fiscalpolicy.If fiscalpolicyis "locallyRicardian," in the sense introduced above the fiscalruleimpliesthatrealgovernment liabilities will necessarily remain withina bounded interval, in the case of anypath for inflation thatremains forever nearenoughto the target inflation rateIl R then (52) is theonlyrestriction uponinflation pathsthatremain forever nearthetarget inflation rate.Theconclusion in thatcasewouldbe (i) thatthetarget steadystateis indeeda perfectforesight equilibrium, and(ii) thatit is indeeddeterminate, as there will be no othernearby solutions. Thusin sucha casethe"Taylor rule" wouldbe an appealing approach to inflation stabilization. Butsuppose instead thatfiscalexpectations canbe described by anexogenous sequence{st}.Thenwe can show,as in ouranalysisaboveof the bondprice-support regime, thatthereis onlya singleinitialvalueforIlo thatis consistent withthesefiscal expectations. Thisis mosteasilyseenin thecaseof short-term (one-period) nominalgovernment debt.48 ThenW0s is givenas aninitialcondition, whiletheright-hand sideof (28) depends onlyuponthe exogenous sequence of surplus expectations, so thereis clearly a unique valueof P0thatsatisfies (28),andcorrespondingly a unique possibleequilibrium inflation rateIlo, givenan initialcondition forP_ l Thisinitial condition thenpicksouta unique solution fromamong thecontinuum of solutions to thedifference equation (52),andthiswill be theunique p.f.e.inflation sequence consistentwiththe fiscalexpectations in question. Alternatively, the evolution of inflationis determined by recursive solution of thepairof equations (28) and(29),withit in the latterequation beingsubstituted outusing(49); as shownearlier, this generatesan inflation sequence thatalso satisfies(18) andhence(52).Thelatter calculation betterrepresents the causallogic by whicha particular sequenceof inflation ratesis generated in equilibrium; but simplereference to (52) is an easierway to quickly determine whattheequilibrium inflation dynamics arelike. Theunique valueof Ilo thatsolves(28) will, in general, nothappen to equalIl8. If, forexample, theexpected future budget surpluses aretoo small,so thatIlo > Il , thenthe only possiblep.f.e. is one in whichthe inflation rategrowswithout bound overtime,as shownin Figure3. Thisequilibrium is characterized by aninflationary spiral, in whichprogressively higher ratesof inflation leadto higher nominal interest rates,hencehigher ratesof growth of nominal government liabilities, whichin turn leadto stillhigher ratesof inflation. Alternatively, if theexpected future budget surplusesaretoo large,thissortof policyregimewill instead leadto a deflationary spiral,in whichthe logic is the samebutin the opposite direction. Thusthisparticular typeof policycombination almostinevitably leadsto equilibrium inflation farfrom thetarget rate. Ouranalysisof this problem has assumed a globallynon-Ricardian policy;but evenif we instead wereto assume thattheprimary surplus is equal to theexogenously evolving"desired" surplus onlywhenit doesnotviolatethebounds(38), muchthe
48. See Woodford(1998c) for an extension of the analysis to the case of longer-term governmentdebt.
712
same problemarises. Just as before, the impositionof these bounds does nothingto change the character of the fiscalist equilibrium(which here involves an inflationary or deflationary spiral);as long as the fluctuationsin {SF} are small enough, wt never violates the bounds in this equilibrium,and so the same equilibriumcontinuesto be possible. It is truethatthe boundson the evolutionof real government liabilitiesmake possible otherequilibriaas well; in particular, in the presentcase therewill now be a possible equilibriumin which Ilt = Ilt forever.(In the case where the expected sequenceof "desired" primarysurplusesis too small, this equilibrium is associatedwith explosive growthin real governmentliabilities, as shown in Figure2, and the expectation that at a certainfuturedate a fiscal retrenchment will bring about actual surpluses higher than the "desired"surpluses.)But the argumentsgiven in section 1.2 above for selection of the fiscalistequilibrium would continueto applyin this case. In particular, it is still true that for any conjecturedfiscal expectationsclose enough to those consistent with the fiscalist equilibrium,a single iterationof the expectationcalculationmappingdefinedabove would lead to the expectationsthatbringaboutthe fiscalist equilibrium.It thus remainsplausible that people could coordinateupon an equilibrium with unstableinflationundersuch a policy configuration. On the other hand, if the centralbank were to commit itself to a policy rule like (49), but with e<)(IIt)< 1,49 then the graphof the right-handside of (52) cuts the diagonalfrom above, so thatthe dynamicsconvergeto fIt regardlessof the initial condition Ilo. In this case, in the presenceof an exogenous primarysurplusprocess, one has a uniquely determinedp.f.e., in which fIo will in general not exactly equal fIt, but the sequence { Ilt} will converge asymptoticallyto ze in any event. Thus there will be a substantialrange of fiscal expectationsthat are all consistent with an equilibriuminflationrate that remainsforever near the targetrate. (In the case of a stochastic version of this model, the equilibriuminflationrate will vary in response to randomshocks, but with bounded shocks it will fluctuateforever within an interval aroundthe targetinflationrate.)This alternativetype of monetarypolicy rule would accordinglybe more conducive to price stability,in the context of a non-Ricardian fiscal policy of the kind assumed.50 We thus observe thata monetarypolicy rule thatwould conventionallybe thought to be anti-inflationary, such as a Taylor rule with an aggressive response to deviations of inflationfrom the targetrate,may insteadlead to an inflationaryspiralwhen combined with an unsuitablefiscal policy. This is exactly what Loyo (1999) argues occurredin Brazil in the early 1980s. As shown in Figure4a, the Brazilianinflation rate remainedquite stable (though nontrivial,2 to 3 percentper month) throughout
49. This is what Leeper (1991) calls "passive"monetarypolicy, by contrast with the "active"case when the inequalityis reversed. 50. Leeper (1991) obtains a similarconclusion in the context of a purely local analysis. He finds that thereis a determinate r.e.e., involving stationaryfluctuationsin the inflationrate, when an "active"monetary policy is combined with a "passive"(locally Ricardian)fiscal policy, or alternativelywhen a "passive" monetarypolicy is combined with an "active"(locally non-Ricardian) fiscal policy. He regardsthe combinationof an "active"monetarypolicy and an "active"fiscal policy as mutuallyincompatible,as in this case there will generally be no nonexplosive equilibriumat all. But as we have seen, this need not mean that there is actually no equilibriumpossible; instead, the equilibriummay necessarily involve an inflationary or deflationaryspiral.
30
10
total
deficit
</XX
levelsin theearly1980s,degenerathigher thelate 1970s,butgrewto progressively a moreantithisto a shifttoward by 1985.5lLoyoattributes ing intohyperinflation interest in 1980.He showsthatnominal ratepolicybeginning interest inflationary in werequitesteadyin the 1970s,despitefluctuations obligations rateson Treasury in the in inflation with increases while they rose morethanone-for-one inflation, < 1 to thatthe shiftwas froma policyrulewithe<, early1980s,andthusproposes both remained non-Ricardian thatfiscalexpectations > 1. If we suppose onewithe<, thentheabovemodelimpliesthatequipolicychange, beforeandafterthemonetary libriuminflationcould well have been stablebefore 1980 (as it was), while the of thekindshown spiral in aninflationary policycouldresult in themonetary change
* n-
n rlgure
o.
(shownin Figure4b) did not inrevenues seigniorage Loyo also notesthatloeal cannot inflation the increased so that spiral, the inflationary notably during crease in the in an increase needsthatresulted revenue to increased be attlibuted plausibly explowith was associated spiral The inflationary target. bank's seigniorage central
ta)
14 12tO |-
, inflation l
I L interestrate
1975
1976
1977
1978
1979
11380 (b)
1981
1982
1983
19t34
1985
1986
5040-
&
r
/\
-seignorage interest on debt
20-
-10 1975
i 1976
1977
1978
1979
l 1980
l 1981
1982
l 1983
l 1984
1985
1986
nominalinterestrate, in perin Brazil. (a) Inflationrate and short-term FIG.4. Onset of Hyperinflation cent per month. (b) Real value of seigniorage revenues, interestpaymentson public debt, and deficit inclusive of interestpayments.Source:Loyo (1999). 51. The dataplotted in Figure4 are kindly suppliedby EduardoLoyo.
714
sivegrowth of nominal government liabilities, resulting fromincreased conventional deficits(plottedin real termsin Figure4b). As the figureshows, the increased deficitswerelargelydue to rapidgrowthin the interest payments required on the public debt,whichexploded as a resultof thechange in interest ratepolicy. Why,then,didtheshiftin theUnited Statesto a policysimilar to the"Taylor rule" in the 1980snot leadto a similar inflationary spiral?52 A possibleanswer is thatin theUnitedStatesthiskindof monetary policywas accompanied by a diffelent type of fiscalexpectations. Flomthemid-1980s onward, concern withthesize of thepublic debtled to callsforconstraints uponthegovernment budget, suchas thoseincorporated in the Gramm-Rudman-HollingsAct of 1985, whichwouldautomatically adjust annual budgets so as to prevent further growth in the debt.Andat leastsince the 1990budget, thisconcern (implying feedback fromthe size of thepublicdebtto the size of theprimary surplus) hasbeena major determinant of theevolution of the U.S. federalbudget. If thesedevelopments werecorrectly anticipated, thenpeople wouldindeedhaveexpected fiscalpolicyto be Ricardian, allowingexpectations to coordinate uponthedesirable equilibrium in whichinflation fluctuates around its target level. Notethatthe successof theAmerican experiment neednotimplythatpeopleunderstood thatfiscalpolicywouldbe Ricardian fiomthebeginning of vigorously antiinflationary interestrate policy underPaul Volcker.53 Even in the context of unchanged expectations regarding thefuture pathof realprimary budget surpluses, a shifttoward a lowertarget inflation rateanda valueof ,C,> 1 neednot resultin an immediate increase in inflation. Instead, in an economy withlong-term, nonindexed government debt,nominalinterest rateincreasescan lowerthe (nominal) market valueof existinggovernment bonds,andthuslead initiallyto a lower pathfor Wts thanwouldotherwise havebeenfollowed,andcorrespondingly a lowerpathforthe pricelevel.Theinflationary spiral thuswouldnothavehadto manifest itselfimmediatelyin the American context(though it wouldhavebeen expected to in Brazil, owingto the muchshorter maturity of the Brazilian publicdebt).It is thuspossible thattheexpectation of a Ricardian fiscalpolicydeveloped onlylater, butstillin time to headoff inflationary debtdynamics in theUnitedStates. Thesuccessful disinflation in theUnitedStatesduring the 1980sandthe successful maintenance of low inflation sinceindicate thatcommitment to an interest rate policysimilar to the"Taylor rule" canindeed be anappropriate wayof stabilizing inflation around a low level.Butthecontrary example of Brazilsuggests thatcountries plagued by highinflation andseekingto emulate theU.S. recipeshouldnotassume thatmereadoption of a Taylor rulewill be sufficient; instead, it is important also to emulate the constraints uponfiscalpolicycharacteristic of the UnitedStatesduring thisperiod.
52. As noted earliel;Taylor(1999) arguesthat U.S. policy in the 1960s and 1970s could be described a similarinterestratefeedback,but with an inflationelasticity < 1, whereaspolicy since the late 1980s at least can be describedby a rule with > 1. 53. Articles such as Sargentand Wallace (1981) indicate that the existence of a fiscal commitmentof this kind was not clear duringthe early Volckeryealos.
3.2 ConstrainingFiscal Expectations
policyruleso as to minuponthechoiceof a monetary Abovewe havecommented of a non-Ricardian of thepricelevel,in thepresence instability imizetheundesirable to the infiscalpolicythatis takenas given.But it is clearthatsuchan adaptation the optisurpluses is hardly budget in primary fluctuations of exogenous evitability in Forexample, at leastfromthepointof view of pricestability.54 malarrangement, will be inconsissurplus budget processfor theprimary a givenexogenous general, policy of themonetary of thepricelevel,regardless stabilization tentwithcomplete of stabilization hand,complete 1996, 1998c).Ontheother thatis chosen(Woodford maywell be possiblein princithepricelevel,evenin theface of realdisturbances, thatshiftsappropriately intercept rulewitha time-varying ple-say, usinga Taylor l999a)-as long as fiscalexpecta(Woodford to the realdisturbances in response to theachievement Thusthebestapproach withthatequilibrium. tionsareconsistent ruleforfiscalpolicy,in adthechoiceof anappropriate will involve of pricestability policyrule. monetary ditionto thechoiceof a desirable would answer Whatkindof fiscalpolicywouldbestservethisend?Thesimplest withstableprices.But thisdoes not go farenough. be, anypolicythatis consistent to couldhappen sulplus,forexample, budget fortheprimary process An exogenous pathfor the pricelevel.Thusone mightsuggestthatthe witha target be consistent one will haveinfla3 will notoccur,andinstead spiralshownin Figure inflationary surof primary sequence as longas theexpected rateforever, tionequalto thetarget upongettingthe to (28). Butthiswoulddepend plusesmakesflo = flt the solution (in the case exactlyright,whichis simplyincredible surpluses size of the expected stateof the exogenous of theestimated areset simplyas a function thatthesurpluses such as the actual developments, fromendogenous world,andnot with feedback of prices). evolution mightseemto be a regimelikethebondprice-support proposal A moreplausible surplus processis the exogenous in section1, in which,however, regimedescribed withstableprices.(Onewouldbe to be consistent chosento be onethatis calculated in responseto an event surpluses for example,not to reduceprimary committed, in thatconproposal, Thisis a morepractical of warin Korea.) suchas theoutbreak wouldnotlead surplus theprimary attempts to target in thegovernment's trolerrors Still,thistypeof regimeis equilibrium. fromthe desired deviation to anyexplosive It makesthe equilibis likelyto be too stablein practice. notone in whichinflation in the nearterm,butof not merelyof fiscalexpectations riumpricelevel a function value (sinceit is thepresent budgets farin thefuture government about expectations maybe future aboutthedistant thatenters(24). Expectations surpluses of all future policyannouncements. through to "manage" for the government difficult especially makesit unlikely processin a democracy of the legislative the nature Furthermore, as moneto the samedegreeof discipline canbe subjected budgets thatgovernment
54. Whetherit is possible or desirablefor primarysurplusesto follow an alternativepath depends, of course, upon what sources of revenue are available and how pressing the government'sneed for funds may be. These issues are beyond the scope of the analysis here.
716
tarypolicyactions. A nontrivial degreeof random variation in theequilibrium price levelwouldbe inevitable under theprice-support regime, bothas a resultof random disturbances to fiscalpolicythatcouldnotbe prevented, andas a result of inability to adjust fiscalpolicywithsufficient precision to offsetthe consequences of otherreal disturbances (suchas fluctuations in theequilibrium realrateof interest). Controlling inflation through an interest raterule suchas the Taylor rulerepresents a morepractical alternative, both becauseit is morepoliticallyrealisticto imaginemonetary policybeingsubordinated whollyto this task,andbecauseit is technically morefeasibleto "fine-tune" monetary policy actionsas necessaryto maintain consistency withstableprices.Finally, under a Taylor rule(together witha locallyRicardian fiscalpolicy),expectations regarding future monetary policydo affectequilibrium inflation, buttheseexpectations arediscounted morerapidly (asone considers datesfarther in the future) in thiscase thanunder the bondprice-support regime muchmorerapidly, in the case of sufficiently aggressive response to currentinflation (Woodford l999a). It wouldmakesense,then,to choosea fiscalrule thatis compatible withthis kindof approach to controlof inflation. Whatthis requires is choosing a fiscalrulethatis consistent notsimplywithoneparticular target pathforinflation, butwithall pathsinvolving sufficiently moderate deviations of the inflation ratefromits desired path;one wouldthenrelyuponan "active" monetary policyto determine whichof thesepathswouldactually be the equilibrium path. This meanschoosinga locall^Ricardian fiscalpolicy.(Whether it shouldalso be globallyRicardian is another matter; in thenextsection,I shallargue thatit is better thatit notbe.) Because theproperty of beinglocallyRicardian requires onlythatthe pathof the publicdebtsatisfycertain bounds, this is a goal thatremains practical evenin a worldwherethe government budget will inevitably be subject to only imperfect control. What kindof constraint uponfiscalpolicydoesthismean? A merecommitment to "satisfy the transversality condition" is plainlyunsuitable; thiswouldplaceno constraints uponobservable behavior overanyfinitetimeperiod, so thatit is hard to see how the publicshouldbe convinced of the truth of sucha commitment, in the absenceof a commitment to somemorespecificconstraint thathappens to implysatisfactionof the transversality condition. One suchpossibility, discussedabove,is a commitment to keeprealgovernment liabilities withinbounds suchas (37). Thisis exactlythe spiritof therequirement of the Maastricht treaty (andof the subsequent "Stability Pact" binding themembers of theEuropean Monetary Union)constraining each nation'spublicdebt to (eventually) remainno greater than60 pelcentof a year'sGDP.55 However, whilea commitment to sucha constraint wouldsufficeto ensuresatisfactionof the transversality condition (under mildconditions, as discussed above), andso render policyRicardian, it maynotsuffice to eliminate theproblem illustrated
55. Above we have considereda ceiling on the real value of governmentliabilities, but in a model in which outputdoes not grow over time. Similarconclusions can easily be obtainedin a model with growth in the case of debt limit thatgrows in proportion to real GDP.
by the Brazilian case.56 For as we have shownin section1.3 above,fiscalpolicy mightremain locallynon-Ricardian evenwhilerespecting thebounds (37).In sucha case,therewouldremain a determinate (locallyisolated) equilibrium in whichfiscal expectations woulddetermine thepricelevel,andpeoplemightwell coordinate their expectations uponthisparticular equilibrium. In thecasethatmonetary policyis described by a Taylor rulewith,C,> 1, this fiscalistequilibrium almostinevitably involves an inflationary or deflationary spiral.Of course,the commitment to the bounds wouldalsomakepossibleother equilibria (justas in OU1previous discussion, when it was hypothesized that such boundswouldbe inevitably satisfied); these equilibria wouldinclude onein whichinflation is alwaysequalto thetarget rate.But because of thecomplicated nature of the fiscalexpectations required to support that "good" equilibrium, one may not wish to rely uponpeopleto coordinate uponit rather thanone of theothers. In thecaseof a locallyRicardian fiscalpolicy,instead, therewill be no unique inflationpathwith the property thatthe realvalueof government liabilitiesremains within narrow bounds, to provide a natural focalpointforexpectations. (Thiswill insteadbe equally trueof all of the solutions to (52).)It thenmaybe plausible forexpectations to coordinate uponthe locallyuniqueequilibrium in whichinflation and interest ratesnever deviate toofarfromthevaluesfl* and<)(tl k) respectively, which is theone in whichtheTaylor rulesuccessfully stabilizes inflation.57 Onesimpleexample of a locallyRicardian regimewouldbe to maketheprimary budget surplus a linear function of accumulated realgovernment liabilities,
St = 5 + t(Wt-W)
where1-,8 < ot' 1, andw > Ois some"target" levelforrealgovernment liabilities. [Thisis closelyrelated to Leeper's (1991)definition of "passive fiscalpolicy."] Substituting (53) into(31) yieldsa law of motionforrealpublicdebtof theform
bt = t(bt-1 rlt Qt-1 Qt,
mt-l nt) >
(54)
rlt Qt,
wherebt-QtBtSIPt. Evaluated neara stationary equilibrium withconstant valuesof andit, thederivative of thisfunction withrespect to its firstargument is
at
- 1 (1- )
whichis non-negative andstrictly less thanone.Thusthe dynamics of therealpublic debtarestable(in the absenceof perturbations to the otherarguments of 93).It
56. Here I take a differentview of the consequences of such a constraintthan thatpresentedin Woodford (1996). 57. Whethertherecontinueto actuallybe otherequilibriadependsthe specificationof policy far away from the targetinflationrate;see the next section.
718
followsthatforanysmallenough fluctuations in (Tlt,Qt, it} around theconstant valuesjust assumed (whichin turnimplysmallfluctuations in (mt}around a constant value,givensmallenough fluctuations in (Yt}), thepathof (bt}will be restricted to a bounded interval as long as it startsfroman initialcondition withinthatinterval. Thisin turnimpliesbounded fluctuations in wtas well;thussucha policyis locally Ricardian. An alternative possibility wouldbe realconventional deficittargeting. Under such a rule, the conventional (thatis, inclusiveof intereston the publicdebt)budget
deficit,58
tt-(1
PQt-Qt-l)Bt-l
Ptgt-Tt
is set eachperiod according to a ruleof theform AtIPt = 6, (55)
where6 is a constant targetlevel. Equivalently, we may assumea constant target valueforthedeficitas a share of GDP;sucha ruleis in thespirit of thelimitson government budgets imposed by theMaastricht treaty andthe"Stability Pact." Notethat in thespecialcasethat6 = O,thisreduces to a balanced-budget rule.59 Ina steady statewitha constant rateof inflation IIt, andunder theassumption that the officialinterest rateis chosento equalthe steady-state interest rateassociated with the inflation target,z = z lrlt- 1, deficittarget(55) is associated with a steady-state levelof (end-of-period) realgovernment liabilities equalto m+ b =
fI -1
Letus assume that IIt > O,andthat6 ' O,so thatend-of-period government liabilities arenon-negative in the steadystate.(Inpractice, thevalueof 6 wouldneedto be set at a level thatis highenoughto be consistent witha steadystatein whichthe government is not a net creditor, butlow enoughto be consistent witha steadystatein whichthe tax collectionsrequired by the rule are not too burdensome. The exact level of 6 does not matter, however, for ourargument here.)Fiscalrule(55) again impliespublicdebtdynamics of theform(54),butnowthederivative is
58. Here we measure"interest" on the public debt by the realized nominalone-periodholding return, ratherthan by the partof payouts to bondholdersthat is officially designated"interest" as opposed to repaymentof principal;the latterquantityhas no economic significance.However,our conclusion as to the locally Ricardiancharacterof such a policy does not depend upon this particular definitionof the deficit. The main convenience of this definitionis that it establishes a simple equivalencebetween deficit targeting and targetingthe pathof nominalgovernmentliabilities;these would otherwisebe closely related,but slightly distincttypes of fiscal commitment. 59. Schmitt-Grohe and Uribe (2000) analyze the consequencesof a fiscal rule of this kind in a slightly differentmodel, and underthe assumptionthat p = 0.
at =rl-l
ab
< 1.
Hence(aslongas thetarget inflation rateis positive) thedebtdynamics areagainstable, andpolicyis locallyRicardian. Thusadoption of deficittargets of this kindin conjunction witha Taylor ruleformonetary policywouldcreate a regimeconsistent withstable,low inflation, andin whichtherewouldbe no reasonto expectations to coordinate uponanequilibrium other thantheonein whichthisoutcome is achieved. 3.3 FiscalCommitments to Exclude a Detationary Trap Thusfarwe haveconsidered only the problem of choosinga fiscalcommitment thatwill notinterfere withthecentral bank's efforts to stabilize inflation through interest ratepolicy. Wehaveargued thatthiscanbestbe achieved through a fiscalcommitment thatensures thatthe dynamics of thepublicdebtwill be stable,in the case of anymoderate fluctuations in inflation, interest rates, andbondprices,so thatfiscal policywill be equallyconsistent withanyof thesepaths.But a fiscalrulecan also servethe goal of pricestability by excluding unwanted equilibria thatwouldotherwise be consistent withthe monetary policyrule.Thustheremaybe advantages to commitment to a rulethatis notgloballyRicardian: fiscalexpectations of thatkind mayprevent unwanted price-level instability dueto self-fulfilling expectations. Wehavepointed out abovethatcommitment to theTaylor rulein itself does not exclude anyof thepossiblesolutions to thedifference equation (52) frombeingpossible perfectforesightequilibria. (And the completeset of rationalexpectations equilibria wouldincludea largeset of stochastic equilibria as well.) We havesuggestedthatthe locallyuniqueequilibrium withinflation forevelnearthe target rate mightbe a logicalone forpeopleto coordinate upon,buta policyregimethatcould actually exclude theother pathsas genuine equilibria (that is, as outcomes thatcould be expected by peoplewitha correct understanding of thepolicyruleandof theprinciplesof price-level determination) wouldallowgreater confidence thatthe desired equilibrium should actually be reached. It is particularly difficult forthecentral bank to achievethisby itselfin the case of the possibility of a defationary spiralof the kindshownin Figure5. (Thisis another solution to the samedifference equation as in Figure 3, butnowassuming aninitialinflation rateflo < flt ) Inflationary spiralsof the kindshownin Figure3 mightbe excluded, or at least rendered less likelyforpeopleto coordinate upon,through a commitment to raiseinterest ratessharply in thecaseof highratesof inflation; thiswouldmeanthatanyinflationratehigherthanthe targetratecouldbe sustained only by expectations of future inflation thatarequiteextreme, if not impossible.60 But similarly aggressive responses to deflation arenot possible,insofar as the zerolowerboundon nominal interest rateslimitsthe extentto whichinterest ratescanbe reduced. As Benhabib, Schmitt-Grohe, andUribe(2001b)pointout,thisimpliesthattheremustexista sec60. See Woodford(1999a, sec. 4.3) for furtherdiscussion.
...
720
rlt+ 1 ^
:(1++(n))
45
low
/ / / |
/ / / / / J /
: : : :
/: /
i
* : : *
* /
: : :
h
t:
*
: :
s h
: :
. & | .
:
J
rl*+=p
FIG. 5.
rt2
nl
fIo
rlF
A Self-fulfilling Deflation undera TaylorRule
ond, lower steady-stateequilibriuminflation rate in the case of a Taylor rule with ,p(ll8) > 1, at or above the gross inflationrate of , which correspondsto deflatio at the rate of time preference.(Underour assumptions,this second steady state Ilt X involves deHation at exactly thatrate, and a nominalinterestrateof zero.) Therenecessarily exists a continuumof solutions to (51) convergingto n, and the deflationary expectations required to support these solutions are not too extreme, either. Benhabib,Schmitt-Grohe,and Uribe suggest that the possibility of such self-fulfilling deflationarytraps is an importantweakness of the Taylorrule as a prescription for monetarypolicy. However,whetherthese solutionsrepresentgenuine perfectforesightequilibria.or not dependsupon the natureof fiscal policy. Benhabib,Schmitt-Grohe, and Uribe assume a (globally) Ricardianfiscal policy, so that all solutions to (51) representperfect foresight equilibria. But as Woodford (1999a, sec. 4.2) points out, under alternative fiscal policy commitments,this is no longer true.6lIn particular, it is possible for fiscal policy to be locally Ricardian(so that the transversality condition is satisfiedin the case of all inflationpaths thatremainforeverwithin an intervalof inflationrates aroundthe targetrate), and still imply thatthe transversality conditionis
61. The idea is furtherdeveloped in BenhabibSchmitt-Grohe, and Uribe (2001c).
violated in the case of any deflationary path of the kind shown in Figure 5. To achieve the latter end, it suffices that fiscal policy place a floor on the asymptotic growth of nominal governmentliabilities, which implies that their real value will grow too rapidly for consistency with the transversalitycondition if the price COI1tracts asymptoticallyat the rate of time preference.Such a growth rate of nominal governmelltliabilities may insteadbe consistent with the transvelsalitycondition as long as the rate of inflationneverfalls too far below the targetrate. One way of guaranteeingsuch a floor on the growthrate of nominal government liabilities is to directly targetthis variable.A policy of this type that has often been assumedin the theoreticalliteratureis a money growthtarget, MtslMts_l = pl, (56)
for some R ' 1, togetherwith an implicit commitmentto maintainBts ' Oas well. For example, a policy of this kind is shown to exclude deflationaryequilibria in Brock (1975) and in Obstfeld and Rogoff (1983), where it is assumed that there is zero governmentdebt at all times. In these papers, of course, the monetarypolicy rule is specified by (56), ratherthanby a Taylorrule. But the eliminationof the possibility of self-fulfillingdeflationsis no special propertyof monetarytargetingas opposed to commitment to an interest rate rule; rather,the result follows from the assumptionof afiscal rule that puts a floor on the path of nominal goverllmentliabilities. Indeed, the money-growthtargetalone would not avoid this problem,in the absence of a stipulationthatthe governmentbudget will ensurethatBts' 0 forever. A deficit targetof the form (55) also puts a floor on the path of total nominal liabilities of the government.Note thatunderour above definitionof the deficit, end-ofperiodtotal liabilities Dt--Mts + QtBts evolve accordillgto
Dt = Dt-l + At-
Any non-negativevalue for the deficit target6 thereforeimplies that {Dt} will be a non-decreasingsequence. Hence if prices fall at the rate of time preference,as they do asymptoticallyin the case of the deflationarypath shown in Figure 5, the real valueDtlPtgrows as t t, and the transversality conditionwill be violated. In the case of a path in which the deflationis only asynq7totically this great, the condition may or may not be violated; but one can ensure that it is, in the case of any solution to (52), throughan appropriate choice of the function (II). (Interestrates must fall to zero quickly enough as the inflationrate falls.) Alternatively, one can ensure a positive floor on the growthrclteof nominal liabilities, so that the transversality condition is violated in the case of any sustainedrate of deflationthat even approachesthe rate of time preference.If one adopts a deficit targetof the form
YDt-l) '
(57)
tt
= max(6Pt,
722
where O < Y < Ilt-1, 6 > O, then in the steady state with inflation at the target rate,the binding constraintis the one definedby 6, ratherthan the one defined by y. Hence the debt dynamics near the steady state are as defined above, and such a policy is once again locally Ricardian.At the same, such a commitmentestablishes a floor for the growthrate of total nominalliabilities, as it ensuresthat
Dt/Dt-l > (1 + 7)
It follows that any path of prices along which the gross inflationrate eventuallysatisfies llt < 5(1 + 7) foreverwill imply that tDtlPt will be boundedaway from zero, and so cannotconstitutean equilibrium. The crucialaspect of such policies, in orderto exclude the possibility of a self-fulfilling deflation, is that the governmentbe committed to continued growth of its nominalliabilities (or at the very least, to preventingthem from contracting),even if the price level, and hence nominal GDP,does steadily decline. This would mean allowing the ratioof governmentliabilities to GDP to rise, in principlewithoutbound, along such a deflationarypath though if the privatesector believes in the government's commitment,the deflationary path (andhence the explosion of the debt/GDP ratio) should never occur in equilibrium. Admittedly,this requiresa differentway of thinkingaboutthe "soundness" of fiscal policy than is yet common. An example is provided by recent discussion of Japanesefiscal policy. As shown in Figure 6, the ratio of the public debt to GDP has grown sharplyin the l990s, rising from only 49 percentat the end of 1991 to more
800 , GDP . .
700 -
ExponentialTrend - - Pubtic l:)ebt+ MB - Public Debt _
600 -
500-
.
c:
._
,,;/ ./
- 400 -
./
' X
FIG.
6. Public Debt, the MonetaryBase, and Nominal GDP for Japan
than96 percentby the end of 1999. This is widely deploredas an indicationof reckless fiscal policy during the l990s, and has been accompaniedby assurancesfrom the governmentthata futurefiscal retrenchment would plevent furthergrowthof the debt ratio. However,a commitmentto maintaina ceiling on the ratio of government liabilities to GDP is exactly the sort of fiscal commitmentthat makes self-fulfilling deflations possible, under an interest rate rule or a money-gl-owth rule alike.62For such a commitmentimplies that in the event of a self-fulfilling deflation,the government will run surplusesof the size necessaryto contractnominal liabilities as fast as prices fall. But then there is no violation of the transversality condition, or alternatively, no accumulationof wealth by householdsthatmakes them feel in a position to spendmore thanthe shrinkinglevel of nominalGDP,so thatthe ever-smallerlevel of nominalGDP is indeed an equilibrium. Instead,the exclusion of such equilibriarequiresa commitmentto a targetpathfor nominal governmentliabilities that will not decline over time even if nominal GDP does; this shouldbe explainedto the public by referenceto a targetpathfor nominal GDP,to which the governmentis committedto return(throughreflationof the economy) even if actual nominal GDP drops below the target path fol- a time. Interestingly, from this point of view the growth of the Japanese public debt during the l990s does not seem so alarming.Figure 6 also plots an exponentialtrend fitted to nominal GDP for the period through 1991; relative to this trend (which might have been a plausibletargetpath for nominal GDP aftel-1991 as well),63neitherdebt nor the sum of the public debt and the monetarybase has grown to historicallyextreme levels. While both have increasedsome since 1991, each is still well below its typical level, relativeto the nominalGDP trend,in the 1970s and 1980s. (Publicdebt relative to trendGDP had risen to 59 percentby the end of 1999, while it was above 80 percentin the 1970s.) Nonetheless, given the Japanesegovernment'sstatements,the Japanese public may reasonably have been expecting that the government is not committed to continued growth of nominal liabilities at such a rate, and that the growth of the debt now forecasts tight budgets later. These are exactly the sort of Ricardianfiscal expectationsthatshould undercutany stimulusto aggregatedemand from the currentdeficits, as householdsjudge that they must increaseprivatesaving to preparefor the future governmentbudget cuts. They may also have allowed the Japaneseeconomy to fall into a deflationary trapof the kind depictedin Figure 5. 4.
62. They will not be possible in the case of a money-growthrule with R 2 1, if there is also a commitmentto maintaina non-negativedebt in the hands of the public. But in this case, there is no commitment to a boundon the ratioof governmentliabilities to GDP.The ratio of the monetarybase to GDP will grow at the rate of time preference along a deflationarypath, and so will the ratio of total liabilities to GDP.Furthermore, insofaras the monetarybase is backedby holdings of governmentdebt, the ratioof the public debt to GDP (countingthe debt held by the centralbank) will also grow at the rate of time preference undersuch a policy. 63. This trendgrows at approximately7 percentper year; in the period 1975-91, this consisted of 4 percentaveragereal growthand 3 percentaverageinflation.
724
CONCLUSIONS
OU1 resultsimply thata centralbank chargedwith maintainingprice stabilitycannot be indifferentas to how fiscal policy is determined.Commitmentto an anti-inflationary monetary policy rule, such as a Taylor rule with a low implicit inflation target,cannotby itself ensureprice stability.Firstof all, fiscal expectationsinconsistent with a stable price level may preventthat outcome from occurring.This possibility is often discussed (for example,in SargentandWallace 1981) as resultingfrom an inconsistency between the policies of the centralbank and of the fiscal authority, so thatthe outcome dependsupon which must accommodatethe other'spolicy commitment in practice;this sometimes leads to the argumentthat a sufficiently independent or sufficiently credible central bank can eliminate the problem, simply by insisting upon its commitmentto its own rule. We have seen, however,thatthe problem is more subtle;it is possible thatthe policies are not actuallyinconsistent(in the sense thatan equilibriumexists in which both commitmentsare maintainedforever), but thatthe only possible equilibriumwill involve an inflationary or deflationaryspiral. Policymakersconcernedwith price stability will not wish to allow expectations of this kind to develop. In the case that the non-Ricardianfiscal expectations assumed in the Loyo (1999) model are unavoidable,the choice of a differenttype of monetarypolicy would be prudent;but in general, a more desirablesolution will be to constrainfiscal expectationsso that stable prices will not requireexplosive debt dynamics. We have also seen that, even when both fiscal and monetarypolicy are consistent with an equilibriumwith stable prices (as one among many possible outcomes), there may be good reason for people's expectationsto coordinateupon an equilibriumotherthanthis one one in which the price level is determinedby expectations regardingthe governmentbudget.In such a case, commitmentby the centralbankto a Taylorrule would again resultin inflationfar from the targetlevel. To exclude this possibility,one would need a commitmentto a fiscal policy thatis locally Ricardian, and not merely globally Ricardian(as in the case of a primarybudget that evolves exogenously until certaindebt limits are reached). An example of a suitablefiscal commitmentfor this purposewould be a targetfor the real value of the conventionalbudget deficit (inclusive of interest on the public debt). Fortunately, commitmentsto budgetbalance or to deficit limits have achieved new prominencein macroeconomicpolicy in the same periodthathas seen increased emphasis upon central bank independenceand actively anti-inflationary monetary policy, both in the United States and in the EuropeanUnion. We have seen that this type of fiscal commitmentalso has the advantageof placing a floor on the pathof the nominalvalue of total governmentliabilities, which can be useful as a means of excluding self-fulfilling deflationsthat would otherwise be possible equilibriaundera Taylorrule. Thus a fiscal commitmentof this kind, in conjunctionwith a monetary policy commitment such as a Taylor rule, represents a sound approach to the achievementof long-runprice stability.
LITERATURE CITED
Auernheimer, Leonardo,and BenjaminContre1as."Controlof the InterestRate with a Government Budget Constraint:Determinacy of the Price Level, and Other Results."Texas A&M University,February1990. . "ANominal InterestRate Rule in the Open Economy."Paper,TexasA&M University, August 1993. Bassetto, Marco. "A Game-TheoreticView of the Fiscal Theory of the Price Level." Paper, Northwestern University,March2000. Begg, David K.H., and BadrulHaque. "A Nominal InterestRate Rule and Price Level IndeterminacyReconsidered." GreekEconomicReview6 (1984), 31-46. Benassy, Jean-Pascal."Price-LevelDeterminacyunder a Pure InterestRate Peg." Review of EconomicDynamics 3 (2000), 194-211 Benhabib,Jess, StephanieSchmitt-Grohe,a1ldMartinUribe. "Monetary Policy and Multiple Equilibria." AmericanEconomicReview 91 (2001a), 167-86. . "ThePerils of TaylorRules."Journal of Economic Theosy96 (2001b), 40-69. . "AvoidingLiquidityTraps." Paper,New YorkUniversity,February 2001c. Bergin,PaulR. "PriceLevel Deter1nination in a HeterogeneousMonetaryUnion."Paper,University of California,Davis, December 1996. Bohn, Henning. "TheBehavior of U.S. Public Debt and Deficits." Qllczrterly JourMczl of Economics 113 (1998a), 949 -64. . "Commenton John Cochrane 'A FrictionlessView of U.S. Inflation'." Ncztional Bureczu of EconomicReseczrch MczcroeconomicsAnnual 13 (1998b), 384-89. Brock, WilliamA. "Money and Growth:The Case of Long-RunPerfect Foresight."International Econonic Review 15 (1974), 750-77. . "A Simple Perfect Foresight Monetary Rule." Journcllof Moszetczry Economvs 1 (1975), 133-50. Brunner,Karl. "Fiscal Policy in Macro Theory:A Survey and Evaluation." In The Monetczry versus Fiscal Policy Debczte, edited by R. W. Hafer.Totowa,N.J.: Rowmanand Allanheld, 1986. Brunner,Karl, and Allan H. Meltzer. "A CreditMarketTheory of the Money Supply and an Explanationof Two Puzzles in U.S. MonetaryPolicy."In Investigationsin Economic Theosy and Methodology,edited by T. Bagiotti. Padova,Italy:CEDAM, 1966. Buiter,Willem H. "TheYoungPerson'sGuide to Neutrality,Price Level Indeterminacy, Interest Rate Pegs, and Fiscal Theories of the Price Level."National Bureau of Economic ResearchWorkingPaperno. 6396, February1998. . "TheFallacy of the Fiscal Theory of the Price Level."National Bureauof Economic ResearchWorkingPaperno. 7302, August 1999. Canzoneri,MatthewB., Robert E. Cumby, and Behzad T. Diba. "Fiscal Discipline and Exchange Rate Regimes."CEPRDiscussion Paperno. 1899, May 1998. . "Isthe Price Level Determinedby the Needs of Fiscal Solvency?"Paper,Georgetown University,June 1999. Canzoneri,MatthewB., and Behzad Diba. "FiscalConstraints on CentralBank Independence and Price Stability." CEPRDiscussion Paperno. 1463, September1996. Carlstrom,CharlesT., and Timothy Fuerst. "The Fiscal Theory of the Price Level." Federal Reserve Bank of Cleveland,EconomicReview 36:1 (2000), 22-32. Christiano,Lawrence J., and Terry J. Fitzgerald. "Understanding the Fiscal Theory of the Price Level."NationalBureauof Economic ResearchWorkingPaperno. 7668, April 2000.
726
Clarida, Richard, Jordi Gali,andMark Gertler. "Monetary PolicyRulesandMacroeconomic Stability:Evidenceand Some Theory." Quczrterly Journczlof Economics 115 (2000), 147-80. Cochrane, JohnH. "AFrictionless Viewof U.S. Inflation." Ncztional Bureau of Economic ReseclrchMacroeconomicsAnnual 13 (1998),323-84. . "Long-Term DebtandOptimal Policyin theFiscalTheory of thePriceLevel." Paper, Universityof Chicago, June 1999. [Availableat http://gsbwww.uchicago.edu/fac/ john .cochrane/research/Papers .] ."Money as Stock: PriceLevelDetermination withNo MoneyDemand." National Bureauof Economic Research Working Paper no. 7498,January 2000. Cushing, Matthew J. "TheIndeterminacy of PricesunderInterest RatePegging: The NonRicardian Case." Journczl of Monetczry Economics 44 (1999),131-48. Daniel,BettyC. "AFiscalTheory of Currency Crises." Paper, SUNYAlbany, April1999. d'Autume, Antoine,andPhilippe Michel."'Transversality Conditions, BudgetConstraints, andthe Determinacy of a PerfectForesight Equilibrium in a Monetary Growth Model." EuropeanEconomicReview 31 (1987),1343 -67. Diamond, Peter A. "National Debtin a Neoclassical Growth Model." AmericanEconomicReview 55 (1965),1126 -50. Dupor, Bill. "Exchange RatesandtheFiscalTheory of thePriceLevel." Journal of Monetary Economics45 (2000),613-30. Eccles,Marriner. BeckoningFrontiers.NewYork: Alfred A. Knopf,1951. Eichengreen, Barry, andPeterGarber. "Before theAccord: U.S. Monetary-Financial Policy, 1945-51."In Finczncial Markets and FinanciczlCrises, editedby G. Hubbard. Chicago: University of Chicago Press,1991. Evans,GeorgeW.,andGarey Ramey. "Calculation, Adaptation andRational Expectations.'^ Mczcroeconomic Dynamics2 (1998),156-82. Fratianni, Michele,andFranco Spinelli.MonetaryHistory of Italy. Cambridge: Cambridge University Press,1997. Friedman, Milton."TheRole of Monetary Policy." Americczn Economic Review 58 (1968), 1-17. Friedman, Milton, andAnna J. Schwaltz. A Monetczry Historyof the UnitedStates, 1867-1960. Princeton: Princeton University Press,1963. Gordon, DavidB., andEricM. Leeper. "Money, DebtandPriceDetermination: TheQuantity andFiscalTheories." Indiana University, July1999. Grandmont, Jean-Michel. Moneycznd Vczlue. Cambridge: Cambridge University Press,1983. Grossman, Herschel I. "The Political Economy of War DebtandInflation." InMonetczry Policy for a ChangingFinancial Environment, editedby W.Haraf andP.Cagan. Washington, 1990. Kenc,Turalay, William Perraudin, andPaoloVitale. "AFiscalTheory of Inflation withSovereignDefault." Birkbeck College,July1997. King,Mervyn. "Commentary: Monetary PolicyImplications of Greater FiscalDiscipline." In BudgetDeficits cznd Debt: Issues cznd Options,Federal Reserve Bankof Kansas City,1995. Kocherlakota, Narayana, andChristopher Phelan. "Explaining theFiscalTheory of thePrice Level." Quczrterly Review,Federal Reserve Bankof Minneapolis, 23(4) (1999),14-23. Leeper, Eric."Policy Rules,Information andFiscalEffectsin a 'Ricardian' Model." InternationalFinance Discussion Paper no. 360,Federal Reserve Board, August1989. ."Equilibria under'Active'and 'Passive' Monetary andFiscalPolicies." Journczl of Monetczry Economics 27 (1991), 129- 47.
Leith,Campbell, andSimonWren-Lewis. "Interactions between Monetary andFiscalPolicy Rules." University of Exeter, October 1998. Loyo,Eduardo H.M."Going International withtheFiscalTheory of thePriceLevel." PrincetonUniversity, November 1997. . "Tight MoneyParadox on theLoose:A Fiscalist Hyperinflation." Kennedy Schoolof Government, June1999. ."Demand-Pull Stagflation." Kennedy Schoolof Government, April2000. McCallum, Bennett T. "Indeterminacy, Bubbles, andtheFiscalTheory of PriceLevelDetermination." National Bureau of Economic Research Working Paper no. 6456,March 1998. . "Theoretical IssuesPertaining to Monetary Unions." National Bureau of Economic Research Working paper no. 7393,October 1999. Obstfeld, Maurice, andKenneth Rogoff."Speculative Hyperinflations in Maximizing Models: CanWeRuleThemOut?" Journal of Political Economy91 (1983),675-87. Santos, Manuel S., andMichael Woodford. "Rational AssetPricing Bubbles." Econometrica 65 (1997),19-58. Sargent, ThomasJ. "TheEndsof FourBig Inflations." In In;fation, editedby R. E. Hall. Chicago: University of Chicago Press,1982. Sargent, Thomas J., andNeil Wallace. "'Rational' Expectations, the Optimal Monetary Instrument, andtheOptimal MoneySupplyRule." Jolernalof Political Economy 83 (1975), 241-54. . "Some Unpleasant Monetarist Arithmetic." QuarterlyReview, Federal Reserve Bank of Minneapolis, Fall 1981. Schmitt-Grohe, Stephanie, andMartin Uribe."Price LevelDeterminacy andMonetary Policy undera Balanced-Budget Requirement." Journal of Monetaty Economics 45 (2000), 211-46. Shim,SangDal. "Inflation andthe Government BudgetConstraint: International Evidence." Ph.D.Dissertation, Department of Economics, University of Minnesota, August1984. Sidrauski, Miguel."Rational ChoiceandPatterns of Growth in a Monetary Economy." American EconomicReview57(2)(1967),534-44. Sims,Christopher A. "ASimpleModelfortheStudyof theDetermination of thePriceLevel andtheInteraction of Monetary andFiscalPolicy." Economic Theory4 (1994),381-99. . "Fiscal Foundations of PriceStability in OpenEconomies." YaleUniversity, September1997.[Available athttp://www.princeton.edu/{ }^\sim$woodford.] . "Econometric Implications of theGovernment Budget Constraint." Journalof Econometrics 83 (1998),9-19. . "Fiat Debtas Equity: Domestic Currency Denominated Debtas Equity in thePrimary Surplus." Princeton University, August1999.[Available athttp://www.princeton.edu/ sims.] Taylor, JohnB. "Discretion versus PolicyRulesin Practice." Carnegie-Rochester Conference Series on Public Policy 39 (1993),195-214. . "AHistorical Analysisof Monetary PolicyRules." in MonetaryPolicy Rules, edited by J. B. Taylo1: Chicago: University of Chicago Press,1999. Timberlake, Richard H. Monetary Policy in the United States: An Intellectual and Institutional History.Chicago: University of Chicago Press,1993. Toma,Mark.Competitionand Monopoly in the Fedessal Reserve System, 1914-1951. Cambridge: Cambridge University Press,1997. Woodford, Michael."Monetary PolicyandPrice-Level Determinacy in a Cash-in-Advance Economy." Economic Theory4 (1994),345-80.
728
. "Price-level Determinacy without Control of a Monetary Aggregate." CarnegieRochesterConferenceSeries on Public Policy 43 (1995), 1-46. . "Controlof the Public Debt: A Requirementfor Price Stability?"National Bureauof Economic ResearchWorkingPaperno. 5684, July 1996. [Shorterversion publishedin The Debt Burden and Its Consequencesfor Monetary Policy, edited by G. A. Calvo and M. King. London:Macmillan, 1997.] Review of . "Doing without Money: ControllingInflationin a Post-MonetaryWorld." EconomicDynamics 1 (1998a), 173-219. . "Commenton John Cochrane,'A FrictionlessView of U.S. Inflation."'National Bureau of EconomicResearchMacroeconomicsAnnual13 (1998b), 390-418. . "PublicDebt and the Price Level." PrincetonUniversity,June 1998c. [Availableat http://www.princeton.edu/ woodford.] . "Price-LevelDeterminationunder InterestRate Rules." PrincetonUniversity,April woodford.] l999a. [Availableat http://www.princeton.edu/ PrincetonUniversity,August l999b. [Available . "InflationStabilizationand Welfare." at http://www.princeton.edu/ woodford.] for the Analysis of MonetaryPolicy."PrincetonUni. "ANeo-WicksellianFramework woodford.] versity,September2000. [Availableat http://www.princeton.edu/

Woodford Fiscal Requirements of Price Stability

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Woodford Fiscal Requirements of Price Stability

Enviado por

Direitos autorais:

Formatos disponíveis

Fiscal Requirements for Price Stability Author(s): Michael Woodford Reviewed work(s): Source: Journal of Money, Credit and

Fiscal Requirements for Price Stability

1.1A Sisnple Model

Eoj , |3tU(ct + gt,MtI

Using this, we may rewrite(2) in the form

Mo + + Er[Rt r+lWt+t] < Wo + [Ptyo-Tr],

togetherwith a transversality conditionon wealth accumulation,15 lim Et[Rt TWT ] (11)

WOODFORD : 679 MICHAEL

pt / Pt+l) Pt+ D Uc(Yt+lMt+l Uc(yt,MtIPt)

Substitutionof this into (3) then yields

tE, Uc(Yt+lM,+l / P+l) Pt 1 + i, = 13-1

ZnT )CT , ,B Et[Uc(YT, T=t

]= )WTI PT Tlim: Et[Uc(yT,Mt I PT

at all times, so that (20) is equivalentto the condition

where st denotes the real primarygovernmentbudget surplus T

k(, i)-Uc(J, L(,i))

Wts+ 1 = (1 + it)[Wt -Ptst]

Mts+ Et[Rt t+ l (Wt+ l-Mt )] = Wt -Ptst,

each period, and hence that Et[Rt T WT ] = (1 - oc) Wt

for all T > t. But this guaranteesthat lim

Condition(37) implies that in any equilibrium

Furthermore, the thirdterm in this series is equal to

for all t ' O,and the transversality condition lim twt = O,

FIG.2. PerfectForesightDebt Dynamics

Corresponding to any sequence36 of expectations {Hte}, the correspondingsequence

Wt+l = Pt[(1+ 1)(x-Hte _ SF)-I-L(y,l)],

H< HTe < H

Hte< H each period,where H is defined in (48) below.

2.3 EmpiricalEvidence on the Characterof Fiscal Policy

3. An Inflationary Spiralundera TaylorRule

-seignorage interest on debt

3.2 ConstrainingFiscal Expectations

is set eachperiod according to a ruleof theform AtIPt = 6, (55)

A Self-fulfilling Deflation undera TaylorRule

ExponentialTrend - - Pubtic l:)ebt+ MB - Public Debt _

6. Public Debt, the MonetaryBase, and Nominal GDP for Japan

Você também pode gostar