The Optimal Degree of Commitment to an Intermediate Monetary Target

Author(s): Kenneth Rogoff

Reviewed work(s):
Source: The Quarterly Journal of Economics, Vol. 100, No. 4 (Nov., 1985), pp. 1169-1189
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/1885679 .
Accessed: 23/09/2012 18:18

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 [email protected].

Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to The Quarterly
Journal of Economics.

http://www.jstor.org
 THE OPTIMAL DEGREE OF COMMITMENT TO AN
INTERMEDIATE MONETARY TARGET*
KENNETH ROGOFF

Society can sometimes make itself better off by appointing a central banker
who does not share the social objective function, but instead places "too large" a
weight on inflation-rate stabilization relative to employment stabilization. Al-
though having such an agent head the central bank reduces the time-consistent
rate of inflation, it suboptimally raises the variance of employment when supply
shocks are large. Using an envelope theorem, we show that the ideal agent places
a large, but finite, weight on inflation. The analysis also provides a new framework
for choosing among alternative intermediate monetary targets.

I. INTRODUCTION
It is now widely recognized that even if a country has a per-
fectly benevolent central bank (one that attempts to maximize
the social welfare function), it may suffer from having an inflation
rate which is systematically too high.' Suppose, for example, that
a distortion (such as income taxation) causes the market rate of
employment to be suboptimal. Then inflation can arise because
wage setters rationally fear that the central bank will try to take
advantage of short-term nominal rigidities to raise employment
systematically. Only by setting high rates of wage inflation can
wage setters discourage the central bank from trying to reduce
the real wage below their target level.
This paper considers some institutional responses to the time-
consistency problem described above. In particular, we examine
the practice of appointing "conservatives" to head the central
bank, or of giving the central bank concrete incentives to achieve
an intermediate monetary target. Our analysis of intermediate
monetary targeting is quite different from conventional analyses
in which the central bank is rigidly constrained to follow a par-
ticular feedback rule. Indeed, an important conclusion is that it
is not generally optimal to legally constrain the central bank to
hit its intermediate target (or follow its rule) exactly, or to choose

*I am indebted to Matthew Canzoneri, David Folkerts-Landau, Maurice Obst-

feld, Michael Parkin, Alessandro Penati, Franco Spinelli, Lawrence Summers,
Clifford Wymer, and to three anonymous referees for helpful comments on an
earlier draft.
1. See, for example, Phelps [1967], Kydland and Prescott [1977], or Barro and
Gordon [1983a,b].

t? 1985 by the President and Fellows of Harvard College. Published by John Wiley & Sons, Inc.
The Quarterly Journal of Economics, November 1985 CCC 0033-5533/85/041169-21$04.00
1170 QUARTERLYJOURNAL OF ECONOMICS

"too" conservative an agent to head the central bank. By appoint-

ing a conservative or by providing the central bank with incen-
tives to hit an intermediate monetary target, it is possible to
induce less inflationary wage bargains. But this comes at the cost
of distorting the central bank's responses to unanticipated dis-
turbances, especially supply shocks. This is a cost because al-
though the central bank cannot systematically raise employment
(since private agents anticipate its incentives to inflate) monetary
policy can still be used to stabilize inflation and employment
around their mean market-determined levels.2 Thus, rigid tar-
geting is appropriate only in certain very special cases. It is im-
portant to stress that, while "flexible" monetary targeting is pref-
erable to either fully discretionary monetary policy or rigid
monetary targeting, it is not necessarily the first-best solution to
the problem of stagflation in this model. That depends on the
source of the underlying labor market distortion which causes
the market-determined level of employment to be too low. If this
distortion can be removed at low social cost, then it would be
possible both to raise employment and to lower inflation. A sec-
ond-best solution, which does nothing to raise the mean level of
employment, would be to legally impose a complete state-contin-
gent money supply rule. As is discussed in Section III, there are
a number of problems inherent in designing such a rule. But it
is only when the first- and second-best solutions are too costly or
unachievable that monetary targeting (or appointing a "conserva-
tive" central banker) should be used as a "third-best" solution to
the problem of stagflation.
Section II of the text describes a stochastic rational expec-
tations macroeconomic model in which, because of wage contract-
ing, there is a well-defined role for central bank stabilization
policy. Section III derives the time-consistent equilibrium under
fully discretionary monetary policy. Section IV shows how society
can make itself better off by appointing as head of the central
bank an agent whose dislike for inflation relative to unemploy-
ment is known to be stronger than average. Section V reinterprets
the formal analysis of Section IV as a model of inflation-rate
targeting, and demonstrates how to extend the framework to en-
compass nominal GNP targeting, money supply targeting, and
nominal interest rate targeting. Section VI discusses comparisons

2. This follows from the assumption that there are nominal wage contracts.
See, for example, Fischer [19771.
 THE OPTIMALDEGREE OF COMMITMENT 1171

across regimes. Which target works best depends, of course, on

the structure of the economy and the nature of the underlying
disturbances. (Though we demonstrate that the interest rate is
generally an unsatisfactory tool for precommitment.) In Section
VII, the Conclusions, we stress the envelope-theorem interpre-
tation of the main result: society wants the central bank to place
"too large" a weight on inflation-rate stabilization relative to em-
ployment stabilization, but the weight should not be infinite.

II. THE MACROECONOMICMODEL

Here we develop a stochastic rational expectations IS-LM
model. Monetary policy can have short-term real effects in this
model because nominal wage contracts are set a period in advance.
Due to high administrative and negotiation costs, these contracts
are not indexed fully against all possible disturbances.3
1. Aggregate Supply
Each of the large number of identical firms in the economy
has a Cobb-Douglas production function. In the aggregate,
(1) yt co + otk + (I - t)n, + zt,
where y is output, k is the fixed capital stock, n is labor, co is a
constant term, and z is an aggregate productivity disturbance;
z - N(O,oz'). Throughout, lowercase letters denote natural loga-
rithms and subscript t denotes time. All coefficients are nonnega-
tive. Firms hire labor until the marginal product of labor equals
the real wage:
(2) co + log(1 - a) + ak - antd+z =wt-Pt.
where w is the nominal wage, p is the price level, and nid is labor
demand.
Labor supply ns is an upward-sloping function of the real
wage:
(3) nt n + w(wt- pt.
To simplify algebra without loss of generality, _nis set equal to
k + (1/o&)[log(l - ox) + co]. As we shall later discuss, the above
labor supply curve (3) is assumed to embody a distortion that

3. The aggregate demand specification is the same as in Canzoneri, Hender-

son, and Rogoff [1983]. The aggregate supply specification is based on Gray [1976].
1172 QUARTERLY JOURNAL OF ECONOMICS

raises the real wage required to induce a given level of labor

supply.
The nominal wage rate for period t is negotiated (on a firm-
by-firm basis) at the end of period t - 1. The nature of the em-
ployment contract is that laborers agree to supply whatever amount
of labor is demanded by firms in period t, provided that firms pay
the negotiated wage rate wt. The level of employment in period
t is thus found by substituting wt into equation (2):
(4) nt = n + (Pt - wt)/o + ztIa.
In choosing Zwt,wage setters seek to minimize Et 1(n, - nt
where Et-1 denotes expectations based on period t - 1 informa-
tion and nHis the level of employment that would arise if contracts
could be negotiated after observing the productivity disturbance
zt and all other period t information. Hnis found using the labor
supply and demand equations (2) and (3):
(5) nt= n + wzt/(l + aw).
From equations (4) and (5),
(6) nt - nt= ztrq + (Pt - wt)aot,
where -q a(1 + aw). It is clear from equation (6) that
Et-1(nt - nt)2 is minimized by setting wt = Et (pt)i' (The pos-
sibility of indexing wages to the price level will be discussed later.)
With equations (1) and (4), together with the analytically
convenient normalization that - co = otk + (1 - oa)n so that
Et i(Yt) = 0, one can write the aggregate supply equation as
(7) Yt = (1 - 0&(pt - -wt)/oL+ Zt/a,
It is very important to note that output and employment stabi-
lization are not equivalent to price prediction error minimization
in the presence of a productivity shock (z).

2. Aggregate Demand
Demand for the good that firms produce is a decreasing func-
tion of the real interest rate:
(8) d= - {r - [Et(pt+1) - Pt]} + Ut,

4. This is the first of many times throughout the paper where use is made of
the fact that certainty equivalence holds when the loss function is quadratic; see
Sargent [1979].
 THE OPTIMAL DEGREE OF COMMITMENT 1173

where r is the level of the nominal interest rate and E,(p,, 1) - Pt

represents the rate of inflation expected by investors, based on
complete period t information. The serially uncorrelated goods
market demand disturbance is ut - N(0,a2); u may be viewed as
a transitory shift in intertemporal consumption preferences.
The demand for real money balances is a decreasing function
of the nominal interest rate and an increasing function of output:
(9) mt - Pt - Xrt + 'PYt+ Vt,
where m is the logarithm of the nominal money supply and v is
a shift in portfolio preferences between money and bonds;
v N(O, U). To simplify exposition, the disturbances, v, u, and z,
are assumed to be independent and serially uncorrelated.
3. The Social Loss Function
The principal differences between the present paper and pre-
vious rational expectations cum wage contracting analyses of
monetary stabilization policy derive from the specification of the
social objective function. Because most models embody the nat-
ural rate hypothesis, the issue of whether or not the central bank
wishes it could lower the average level of employment is com-
monly ignored. But this potential source of tension is fundamental
to the conduct of stabilization policy. Indeed, if there does not
exist any temptation for the monetary authorities to inflate sys-
tematically, then there is no reason to consider any regime other
than fully discretionary monetary policy. (Note that the central
bank's incentives to inflate need not be motivated by employment
considerations, but can also arise due to the presence of nominal
government debt or short-term rigidities in the tax system.)
Here we shall assume that some factor such as income taxa-
tion or unemployment insurance distorts the labor-leisure deci-
sion and causes the market-determined level of employment nt
to lie below the socially optimal level of employment h' (see Barro
and Gordon [1983a]). We shall further assume that h' - ni is
constant and equal to h - n.
The social loss function A depends on deviations of employ-
ment and inflation from their optimal (socially desired) levels:5

5. A similar social objective function is employed in Kydland and Prescott

[1977] and Barro and Gordon [1983a]. Although the analysis below would have
to be modified substantially if the central bank had a multiperiod objective func-
tion, the main points would still obtain.
1174 QUARTERLYJOURNAL OF ECONOMICS

(10) A, = (n, - t')2 + X(TIrt-

where Irt Pt - Pt- 1, r is the socially desired trend inflation rate,

and X is the relative weight society places on inflation stabili-
zation versus employment stabilization.
It is somewhat difficult, in the context of a rational expec-
tations model, to argue that the level of the inflation rate has
much direct weight in the social loss function.6 (However, the
analysis below does not depend on X being particularly large.)
The costs of inflation include the administrative costs of posting
new prices and the costs of adjusting the tax system to be fully
neutral with respect to inflation. And, of course, high rates of
inflation force agents to economize on their holdings of non-in-
terest-bearing money-the so-called "shoe leather cost of infla-
tion." Despite the foregoing considerations, fr may be nonzero if
alternative taxes to seignorage also generate deadweight costs
through distortions (see Phelps [1973]).

III. TIME-CONSISTENT EQUILIBRIUMUNDER FULLY

DISCRETIONARYMONETARY POLICY

Here, stochastic equilibrium is derived under the assumption

that the monetary authorities attempt to minimize the social loss
function A, given by equation (10) above.
Expectations about the future path of the money supply are
not exogenously given in this model, but depend endogenously on
agents' expectations about the monetary authorities' future short-
run stabilization objectives. Wage setters will not believe prom-
ised future paths for the money supply that are not time-con-
sistent. Instead, equilibrium nominal wage increases are set at a
sufficiently high level so that, in the absence of disturbances, the
central bank will not choose to inflate the money supply beyond
the point consistent with wage setters' desired real wage. At this
high level of inflation, the central bank finds that the marginal
gain from trying to raise employment above the natural rate is
fully offset by the marginal cost of still higher inflation. Note also,
that no individual group of wage setters has any incentive to
change their wage bargain in the time-consistent equilibrium.
Even though individual wage setters are concerned about infla-

6. Unanticipated inflation enters indirectly into the social loss function (10)
through its effect on employment. Fischer and Modigliani [1978] catalog the eco-
nomic costs of both anticipated and unanticipated inflation.
 THE OPTIMALDEGREE OF COMMITMENT 1175

tion, the contract at their firm has only a small impact on the
aggregate inflation rate.
By substituting equation (6) into equation (10), and recalling
that h' - n' = h - n, the central bank's objective function under
fully discretionary monetary policy may be written as
(11) Dt = At = [zWbq+ (Pt - )/o- (nf - n
+ X[Pt - Pt 1 -*],

where superscript D stand for "fully discretionary regime." The

central bank maximizes social welfare by choosing a level of the
money supply consistent with pr, the period t price level that
minimizes At:7
(12)
h - n -
pD [t zt/ -*]/[+(1 2

Recall that (the logarithm of) wage setters' target real wage is
zero. Thus, wage setters select u' by taking expectations across
(12) and setting e =-Et -ID):8
I
(13) O = Et-1(pD) = Pt-i + * + (h - n)/Xot = Pt-i + ir1D.
By choosing t according to (13), wage setters assure themselves
that the monetary authorities will not systematically drive down
the real wage. Thus, as Kydland and Prescott [1977] point out,
the time-consistent rate of inflation is too high when h > n.
We are now prepared to evaluate social welfare under fully
discretionary monetary policy. But to facilitate exposition in later
sections, we shall first develop a notation for evaluating the ex-
pected value of the social welfare function under any arbitrary
monetary policy regime "A", AA:

7. pt is found by setting 8DtlIpt = 0. The second-order conditions for a min-

imum are met; given the quadratic form of D, the minimum is global.
8. Investors can apply the same algorithm repeatedly to derive a time-consis-
tent path for all future prices:
(8.1) Et-1 (pP?8) = (h - n)(s + l)/xo + (s + 1)f* + Pt+ , S _ 0.

(For simplicity, we treat the monetary authorities' objective function as con-

stant.) Note that we are treating the price level as if the monetary authorities
controlled it directly, ignoring the fact that the central bank directly controls only
the money supply. The anticipated future path of the money supply consistent
with (8.1) may be found using the macro model of equations (7)-(9), together with
the assumption of saddlepath stability. (For macroeconomic justification of the
saddlepath assumption in monetary models, see Obstfeld and Rogoff [1983].)
1176 QUARTERLY JOURNAL OF ECONOMICS

(14) It =( - n)2+xHA+FA

where HA (A - X)2 and

fA Et_1{Lzt/r + (p:4 - Et-1(p:))kx]2 + XAp: - Et-1(pt)]21
(we have made use of the fact that Et i(Pt) = at). The first com-
ponent of AA is nonstochastic and invariant across monetary re-
gimes. It represents the deadweight loss due to the labor market
distortion. This loss cannot be reduced through monetary policy
in a time-consistent rational expectations equilibrium. The sec-
ond term depends on the difference between the expected rate of
inflation and society's target rate. This term is also nonstochastic
but does depend on the choice of monetary policy regime. The
final term, ['A, represents the "stabilization" component of the
loss function. It measures how successfully the central bank off-
sets disturbances to stabilize employment and inflation around
their mean market-determined values.
We have already solved for the mean level of inflation under
fully discretionary monetary policy iP; see equation (13). To de-
rive FD, first multiply both sides of equation (12) by (Et - Et_ 1),
noting that e = Et - (p). This yields
(15) [Pt - Et-(pt)] d - Zt/,q[OX + (I/a)] pDZt.

Note that u and v do not enter the expression for the price pre-
diction error that the central bank allows to occur. The central
bank offsets the price level effects of aggregate demand shocks to
the best of its ability (here perfect, because of the complete in-
formation assumption), because offsetting these shocks is con-
sistent with both employment stabilization and inflation-rate sta-
bilization. By substituting (15) into (14), and simplifying, one can
obtain

(16) rD = (or1/12)[x/(x + (1/(x)2)].

It should be observed that an individual group of wage setters
has little incentive to index positively to the price level under the
fully discretionary regime. The monetary authorities fully offset
the effects of demand shocks, and because inflation as well as
employment enters the social objective function, they do not allow
the price level to move enough to optimally offset the employment
effects of aggregate supply shocks.
Obviously, a first-best solution to the stagflation problem de-
scribed above would be to remove the labor market distortion. If
 THE OPTIMAL DEGREE OF COMMITMENT 1177

this cannot be achieved at low social cost, a second-best solution

would be to design a permanent constitutional reform that ab-
solutely ruled out systematic inflation, and yet left the central
bank scope to respond to disturbances. However, there are some
practical drawbacks with constitutionally instituting a state-con-
tingent money supply rule. To be fully effective, a rule must be
set in place in such a way that it is very difficult to change. This,
in turn, raises the danger that the rule will be difficult to alter
after it becomes outmoded. Such problems might well arise be-
cause it is very difficult to predict the qualitative nature of the
shocks buffeting the economy decades in advance. In the sixties,
for example, it might have been difficult to anticipate the supply
shocks of the seventies. Other factors such as innovations in trans-
actions technology, regulatory changes, and the evolution of fi-
nancial intermediaries all complicate the problem of designing a
permanent monetary rule. (We are not suggesting that these prob-
lems are necessarily insurmountable.)

IV. SOCIALWELFAREUNDERA "CONSERVATIVE"

CENTRALBANKER
Here we consider an alternative, less drastic, response to the
stagflation problem posed above. We demonstrate that society can
make itself better off by selecting an agent to head the indepen-
dent central bank who is known to place a greater weight on
inflation stabilization (relative to unemployment stabilization)
than is embodied in the social loss function A. The term of the
agent need last only one period, though in a multiperiod setting,
reputational considerations will further help ameliorate the cen-
tral bank's time-consistency problems.9 However, in choosing among
potential candidates, it is never optimal to choose an individual
who is known to care "too little" about unemployment, in a sense
that will be made precise below.
Suppose, for example, that in period t - 1 society selects an
agent to head the central bank in period t. The reputation of this
individual is such that it is known that if he is appointed to head
the central bank, he will maximize the following objective func-
tion (henceforth, time t subscripts are omitted where the meaning
is obvious):
(17) I= (n - h')2 + (X +E) Or _*)2, X + E>.

9. See Barro and Gordon [1983b], for example.

 1178 QUARTERLYJOURNAL OF ECONOMICS

When E is strictly greater than zero, then this agent places a

greater relative weight on inflation stabilization than society does.
The algorithm for deriving the time-consistent equilibrium
is exactly the same as in the previous section. Equations (18),
(19), and (20) are the "I" regime counterparts of equations (13),
(15), and (16), respectively:
(18) H' =&:I - )2 = [(h - n)Ia(x + e)]2,

(19) [Pt - E (pt)]= dp =- zt/q[(l/o) + cX + ae],

(20) F' = (oy/ii2)[(x + E)2 + (xI/x2)]/[(l/t)2 + X + e]2.

Note that F' is obtained by plugging dpt into fA, as defined in

equation (14). The regime should be evaluated on the basis of the
expected value of the social loss function, not the expected value
of the central banker's loss function. The reader can confirm that
equations (18)-(20) are the same as (13), (15), and (16) when
E = 0. The expected value of the social loss function under regime
"I"is
(21) A' = (t- n)2+ H'+F
To solve for the value of E that minimizes A', differentiate
(21) with respect to e:

(22a) cA X= (H) +

(22b) -= 2 (w-2 + X + ]
(3)

(22c) a1' 2[(h -n

Define Emin as the value of E that minimizes A'.10 We are now

ready to prove:
THEOREM. For h > n, 0 < Emin < oc

Proof. Note that E > - X by assumption. Thus, by inspection

of (22c), (3HIh3Eis strictly negative. Note also, by inspection of
(22b), that 8F1/8Eis strictly negative for - X < E < 0, zero when
E = 0, and positive for E > 0. Therefore, aA'/(3 is strictly negative
for E ' 0. dA'/8l must change from negative to positive at some
10. Although it is extremely difficult to write down a closed-form solution for
E , one can prove that Emmn is the unique positive real root of WA1/dE= 0, and
therefore that A, is concave in e.
 THE OPTIMAL DEGREE OF COMMITMENT 1179

sufficiently large value of u, since as E approaches positive infinity,

WarF'/converges to zero at rate _2, whereas aH'1/Th converges to
zero at rate E-3. Therefore, um'in <x
Q.E.D.
It follows immediately that for nf = n, Fmin = 0. The Theorem
states that in the presence of a labor market distortion, it is
optimal to choose an agent to head the central bank who places
a greater, but not infinitely greater, weight on inflation than
society does. To interpret the Theorem intuitively, consider the
effects of raising E from zero. By increasing the central bank's
commitment to fighting inflation, the time-consistent average rate
of wage inflation is reduced. But the relative weight the central
bank places on inflation versus employment stabilization is al-
tered, and this distorts the monetary authorities' responses to
unanticipated shocks. To see why the benefit outweighs the cost
at E 0, it is suggestive to expand (22a) as follows:
A~~i r aAI 1 rats
(I - 1rIf
)I aF
(23) K,1' a)'1 +

In the neighborhood of E = 0, the monetary authorities are mini-

mizing [V (they are stabilizing optimally), so that is zero.
aris/e

But inflation is not being minimized, so that neither term in

(a - ifr) is zero. We can argue similarly to suggest why Fmin < X.
As E becomes large, i' goes to fr and HI is being minimized. Thus,
for large E, the marginal inflation cost of reducing E is small
relative to the stabilization gain. Of course, when there is no labor
market distortion, so that n = n, then D= fr, and it does not
pay to appoint a central banker who minimizes anything other
than the social loss function.
We have assumed that the preferences of the agent appointed
to head the central bank can be known with certainty. Clearly,
many strategic problems arise when this assumption is relaxed.
However, as long as there is some information on the probable
preferences of alternative candidates, the basic point of the above
analysis is still germane." The model is certainly consistent with
the fact that central bankers are typically chosen from conserva-
tive elements of the financial community. One incentive that the
head of the central bank might have for holding down inflation

11. For an illustration of some issues that arise when the monetary author-
ities' preferences are unknown, see Backus and Driffill's
[i1985 interpretation of
Kreps and Wilson's [1982] analysis.
1180 QUARTERLY JOURNAL OF ECONOMICS

is that he can thereby improve his standing in the financial com-

munity, and thus earn greater remuneration upon returning to
the private sector.

V. INTERMEDIATEMONETARY TARGETING
In the previous section we demonstrated conditions under
which society can make itself better off by appointing an indi-
vidual to head the central bank who is (somewhat) more inflation-
conscious than average. The same model can be employed to ex-
plain many of the measures that countries take to insulate their
central banks from inflationary pressures. For example, central
banks are often endowed with a significant measure of political
and fiscal independence. The analysis also suggests why it would
be desirable to have a central bank's operations financed in such
a way that its expenditures are independent of the government's
seignorage revenues. (It is interesting to observe that during the
high inflation years of the late sixties and the seventies, archi-
tecturally stunning new Federal Reserve buildings sprouted up
all over the United States.)
In some sense, the widespread adoption of intermediate mone-
tary targeting during the seventies may be viewed as an insti-
tutional response to the time-consistency problem. Suppose, for
example, that through a system of rewards and punishments the
central bank's incentives are altered so that it places some direct
weight on achieving a low rate of growth for a nominal variable
such as the price level, nominal GNP, or the money supply. Al-
though these alternative targets have different stabilization prop-
erties, credibly increasing the central bank's commitment to
achieving any of them would reduce the time-consistent rate of
inflation (as can be demonstrated along the lines of the Theorem
in Section IV). A very direct way of making the commitment
credible would be to tie the annual remuneration or budget of the
monetary authorities to their success in hitting their intermediate
monetary targets. This could perhaps be accomplished through a
system of bonuses or by fixing their income in nominal terms.
Other explicit penalties might include requiring the central bank
to devote substantial resources to publicly justifying a deviation
from its targets. An implicit penalty would be if the central bank's
powers and independence are affected by how well it succeeds in
hitting its stated targets. (Of course, if the central bank is already
 THE OPTIMAL DEGREE OF COMMITMENT 1181

controlled by inflation-conscious forces, measures to reduce its

independence could raise the time-consistent inflation rate.)

A. Inflation-Rate (Price Level) Targeting

Using the above interpretation of monetary targeting, we
may view the analysis of Section IV as formally equivalent to an
analysis of inflation-rate (price-level) targeting."2 With this inter-
pretation, the E term in the central bank's objective function (17)
measures the extra incentives (rewards or punishments) the cen-
tral bank has for fulfilling its inflation-rate target. These incen-
tives are additional to the fact that the inflation rate enters di-
rectly into the social objective function A.
In the absence of productivity disturbances, inflation-rate tar-
geting works extremely well, since there is then no tradeoff with
employment stabilization. Indeed, it would then make sense to
make E as large as possible; raising E lowers the time-consistent
inflation rate without placing any constraints on the ability of
the monetary authorities to offset aggregate demand distur-
bances. If money supply changes transmit quickly into price level
changes, then CPL targeting would be relatively easy to imple-
ment. The CPI is published monthly in many countries, and it
would not be extraordinarily expensive to gather data at higher
frequencies. The present model abstracts, however, from problems
that arise if there are long lags in the monetary transmission
mechanism.
If supply shocks are important, then it is natural to ask whether
there are other intermediate targets that would allow the central
bank to bring down the mean inflation rate at lower stabilization
cost. We shall consider, in turn, nominal GNP targeting, money
supply targeting, and interest rate targeting.
B. Nominal GNP Targeting
Because a positive supply shock tends to raise output and
lower the price level, one would expect there to be some circum-
stances in which nominal GNP targeting is more appropriate than
inflation-rate targeting. Suppose then that the central bank's tar-
get is nominal GNP so that its objective function is given by

12. Price-level targeting and inflation-rate targeting are equivalent here,

since Pt-i is known at the time the central bank commits itself to achieving a
target for pt - pt- i.
1182 QUARTERLY JOURNAL OF ECONOMICS

(24) G = (n - h,)2 + X(7 - fT)2

+ T(Yt + Pt - Y -Pt- - 2

where G stands for "nominal GNP targeting regime," and the

index parameter X gives the weight that the central bank places
on achieving its intermediate target relative to the weight it
places on directly maximizing the social objective function A; y
is the level of output corresponding to n = n, and z = 0. The
assumption that the central bank's nominal GNP target is con-
sistent with the natural rate _nrather than the socially optimal
& is nontrivial. It may be politically difficult for the central bank
to commit to a nominal GNP target consistent with a level of
employment lower than the socially optimal rate h. If y is replaced
by 5 in (24), the stabilization properties (F) of nominal GNP tar-
geting would remain unchanged, but the mean inflation rate would
be higher for any level of T. In particular 1T - *r will not converge
to zero but to (1 - &)(h - _n)as T -a oo. This problem is not suf-
ficient to reject nominal GNP targeting when the variance of
supply shocks is large, but is an important drawback.
We can rewrite the objective function G by using equations
(6) and (7):

G Zt Pt - - (ft- h)] + X[Pt- Pt-i -

+T[+(1 () (hPt +) t- Pt 1 - j *]

(25)

To find the time-consistent path of the economy under nom-

inal GNP targeting, we again follow the algorithm of Section III:
(26) HlG = ('TG - *)2 = (h - n)2/[ + XaL]2
(27) dpG - Zt[(ct/ri) + T]/[1 + -T + a2X] = pGZt,

( -0o= {[(Xa2/Xr) (T-tW/-q)]2 + X[a/-q + T]2}

(28) JIG
~ +T+O2
As in the case of inflation-rate targeting, one can prove that
0 < Tmin < o. 13 Nominal GNP targeting and fully discretionary
monetary policy always imply the same response to aggregate

13. The fact that 0 < rmin < cc may be demonstrated analogously to the Theo-
rem in Section IV. A simpler proof makes use of the fact that NA is minimized at
p =D, and that FA is strictly increasing in JpA
 THE OPTIMAL DEGREE OF COMMITMENT 1183

demand disturbances; in both cases the central bank tries to sta-

bilize the price level. The response to supply shocks is in general
different. However, if wages are indexed to unanticipated move-
ments in the price level (so that Wt iW+ P(pt - iwt), 0 - - 1)
then there does exist one special parameter configuration such
that fixing nominal GNP is exactly what the monetary authorities
would do if unconstrained. Obviously, Tmin = 00 in this case."4
C. Money-Supply Targeting
Because the formal analysis of money-supply targeting is
quite similar to that of the preceding sections, we shall focus only
on the special features of this case. The central bank's objective
function under money-supply targeting is assumed to be given by
(29) M =(n -h') + X _T*f)2+ (mt - lt)2 p.?-O

where th = + Pt- 1 + X rrM;superscript M denotes the money-

*

supply targeting regimes. Employing the macro model of equa-

tions (7)-(9), one can demonstrate that the target level of the
money supply uit is the level of mt that would be consistent with
society's desired inflation rate * provided that (a) there are no
disturbances in period t, and (b) nM is the expected inflation rate
between periods t and t + 1.
The fact that the demand for money is interest elastic creates
some problems in selecting an appropriate level for mtt. In setting
mtt, the central bank needs to form an estimate at time t - 1 of
Irt+ 1, which depends in turn on the policy regime expected to
prevail in period t + 1. Here we have assumed that everyone
expects the same money-supply targeting regime to remain in
place. Because the nominal contract rigidities last only one period,
this assumption is not crucial. It is crucial, however, that the
central bank's money supply target be set at a level consistent
with market expectations of rrt+,. If the central bank were to set
t optimistically under the assumption thatEt(wrt+?) = *, it would
overestimate the demand for real balances and choose too high a
level of it. Choosing a target level for mi that is too high would
not affect the stabilization properties of money-supply targeting,
but would raise the time-consistent inflation rate for any level of
[t. This problem with medium-term money-supply targeting is

14. Tmin = w if (1
- p)2 + aL2X= (1 -
p)(1 + 43 - p)/(otw+ 1). Note that if
L = 0, then rigid nominal GNP targeting is equivalent to employment tar-
geting.
1184 QUARTERLYJOURNAL OF ECONOMICS

quite similar to the problem with medium-term nominal GNP

targeting discussed earlier.
Equation (29) can be rewritten with Pt as the central bank's
control variable by using equations (7)-(9) to substitute out for
mt:
(30) mt = - XIM + aPt + (1 - ;)Wm - (X/8) ut
+ (e - 1)zt/(1 - () + Vt,
where e-1 + [(1 - 0)/od](/ib+ A). By substituting into equation
(29) using equations (30) and (6), one obtains

M = L-t +Ptwt _(f t-h)1

(31) + X[Pt - Pt-i + *]2 + [tLePt + (1 - t)w4 - (X/i)ut

+ (; - 1)(1 t ) + Vt - Pt-1 -

As before, one can demonstrate that 0 < Lmin < o except in

the special case where there are no aggregate demand shocks
(of = = 0), and one special parameter configuration obtains.15
In this special case pm= pD, and Rmin= o. (An infinite p is
equivalent to a k percent rule.) Monetary targeting then can lower
the mean rate of inflation to its socially optimal level without any
stabilization cost. The condition that the aggregate demand shocks
be zero can be relaxed when the monetary authorities have in-
complete contemporaneous information.16
D. Nominal Interest Rate Targeting
The monetary authorities cannot systematically raise or lower
the mean value of the real interest rate in the rational expecta-
tions model employed in this paper. But it might seem reasonable
for the central bank to try to bring down the inflation rate by
committing itself to achieving a low nominal interest rate.17 Here

15. The condition for umin = o is that

(1 + ca2X)(e2 - ()(1 + axe) = (2(1 - t).

16. An appendix to an earlier version of this paper discusses how to extend

the present analysis to the case of incomplete contemporaneous information. This
appendix and details of the proofs concerning Rmin and Tmin are available from the
author on request.
17. As Canzoneri et al. [1983] demonstrate, the central bank can only literally
peg the nominal interest rate if it also announces at least one point on a mutually
consistent money supply path; otherwise the price level is indeterminate.
 THE OPTIMAL DEGREE OF COMMITMENT 1185

we argue that this method of intermediate monetary targeting is

counterproductive.
Consider the objective function,
(32) R = (n - h')2 + X(- - *)2 + 0(r - ?)2, 0 ' 0 <c,
where r-is the nominal interest rate target. Using equations (6)-(8),
we may write the objective function (32) as

R + (Pt )- (h - n X(Pt - Pt-i 2

2
(33) 1+o[7i+i + (1 ) Pt ( Zt)

As we shall see, the value of s?t+1 E,(ptA+1) - Pt is crucial to the

analysis; it depends on the monetary policy regime expected to
prevail in period t + 1. Following the algorithm of Section III for
deriving the time consistent equilbrium, one obtains

(34) nR - * = [(ft -) + (0(1 - )/)(r+ - x,

where rt+ 1 has been substituted in for It+ 1. (Given the assump-
tions we have made about the parameters of the macro model,
the mean real interest rate under any monetary regime is zero.)
A comparison of equations (13) and (33) reveals that rR
S D as r + 1. In other words, suppose that an intermediate
targeting regime is put in place for one period, and the central
bank is given incentives to bring interest rates below their trend
rate, so that ? < rt+ 1. Then, instead of falling, the expected infla-
tion rate and expected nominal interest rate rise. They rise be-
cause wage setters recognize that once wages are set, the central
bank can lower interest rates through money growth. While it is
true that the central bank could try to bring down inflation by
setting ? greater than rt+ 1, the fact that this would indeed cause
the market-determined interest rate rt to be less than itA+1 sug-
gests a serious credibility problem. The central bank has to target
high interest rates if it wants low interest rates.
The underlying problem is that given Et- (pt+l - Pt), an-
nouncing a target for rt is, in fact, tantamount to targeting the
real interest rate. For the regimes analyzed earlier, targeting
succeeds in at least temporarily lowering the inflation rate re-
gardless of how long the targeting regime is expected to last. This
is no longer true when the nominal interest rate is used as a
 1186 QUARTERLY JOURNAL OF ECONOMICS

target.18 (It should be clear that a misguided attempt to target a

low real interest rate will produce similar problems.)
The nominal interest rate would not appear to be a suitable
instrument for precommitment. This conclusion, of course, does
not imply that the interest rate should not be used as an infor-
mation variable in setting monetary policy, as in Poole [1970].

VI. ON COMPARINGALTERNATIVETARGETINGREGIMES
In more traditional analyses of intermediate monetary tar-
geting, a conventional result is that the optimal target choice
depends on all the parameters of the model as well as on the
relative sizes of the disturbances. 19 While one must also consider
strategic factors in the more general model developed here, the
standard stabilization considerations are still relevant. For ex-
ample, money-supply targeting makes less sense when the mone-
tary authorities have information about how aggregate demand
shocks are affecting the price level. Inflation-rate targeting works
poorly when supply shocks are significant, etc. Indeed, if we were
to restrict our attention to "rigid" targeting regimes (that is, if
the monetary authorities are required to hit their target exactly),
then (of course) the ranking of regimes depends only on these
conventional stabilization considerations (see the Appendix).
Ranking the optimally flexible regimes of Section V is some-
what more difficult because they do not have tractable closed-
form solutions. (For specific parameter values, numerical com-
parisons are easily obtained.) It is worth noting, however, that
the ranking of optimally flexible regimes is not necessarily the
same as the ranking of "rigid" regimes presented in the Appen-
dix.20 This is not surprising in view of the "second-best" nature
of the rigid targeting regimes; it is almost never optimal to con-
strain the monetary authorities to hit their target exactly.
18. It can be shown that low nominal interest rate targeting is counterproduc-
tive when the regime is expected to last for any finite number of periods, as long
as the expected inflation rate in the final period is consistent with a return to
fully discretionary monetary policy. The regime fails because the central bank
cannot systematically achieve a below-market real interest rate for any future
period.
19. See Poole [1975], Friedman [1975], or Parkin [1978] for analyses of the
stabilization properties of alternative intermediate monetary targets.
20. Numerical examples of rank reversal are presented in an earlier version
of this paper, available on request. Numerical solutions for Tmin are easier to obtain
when AAi/aTiS strictly concave in T. Using Descartes' Law of Signs, one can derive
the sufficient condition 1I(a - (X2) > X. Similarly, dAM/dI is definitely concave in
p. if [4~+ (XIS)]/(ct - at2) > X.
 THE OPTIMAL DEGREE OF COMMITMENT 1187

VII. CONCLUSIONS
It can be entirely rational for society to structure its central
bank in such a way that the monetary authorities have an ob-
jective function very different from the social welfare function.
Whenever a distortion causes the time-consistent rate of inflation
to be too high, then society can be made better off by having the
central bank place "too large" a weight on inflation rate stabili-
zation. The model presented here may help explain why many
countries set up an independent central bank and choose its gov-
ernors from conservative elements of the financial community.
Although society does want the central bank to place a large
weight on inflation rate stabilization relative to employment sta-
bilization, society will not (in general) want the weight to be infi-
nite. By having the central bank place an infinite weight on infla-
tion stabilization, society could succeed in bringing inflation down
to its socially optimal level. But the central bank would also end
up responding very inappropriately to supply shocks, allowing them
to pass entirely through to employment. By lowering the weight
which the central bank places on inflation, society could achieve a
first-order stabilization gain at a second-order inflation cost.
However, the inflation weight should not be so low that the cen-
tral bank is placing the same weight on inflation-stabilization as
society does. For then the central bank would be stabilizing opti-
mally and by raising the central bank's weight on inflation, it would
be possible to achieve a first-order inflation gain at a second-order
stabilization cost (by the envelope theorem).
When supply shocks are important, society may prefer to give
the central bank incentives to focus on a monetary target other than
the inflation rate (though again, it is not optimal to have the weight
on the target be infinite). It might be expected that the best mone-
tary target would be the one most highly correlated with the so-
ciety's ultimate objective function. But while this is a useful rule
of thumb, the situation is actually somewhat more complicated. If
one compares how each of the targets would work if used rigidly,
one does not necessarily get the same ranking as when the cen-
tral bank gives its target an optimal weight relative to direct so-
cial objectives. Thus, it can be misleading to analyze separately the
stabilization and credibility problems of the central bank. The
model also highlights strategic problems that can arise in setting
targets at a non-inflationary level. If the central bank sets its
nominal GNP target consistent with the socially optimal rate of
1188 QUARTERLY JOURNAL OF ECONOMICS

employment, rather than the lower natural rate (perhaps because

of political constraints), then nominal GNP targeting will have an
inflationary bias. Money supply targeting presents similar prob-
lems if the demand for money is interest elastic. The central bank
will overestimate next period's demand for money if it assumes (or
is constrained to assume) that next period's expected inflation rate
will be at the socially desired rate, rather than at the higher time-
consistent rate. (This problem does not arise under a permanent k
percent rule.) We demonstrated that targeting low nominal inter-
est rates is counterproductive and actually raises the time-con-
sistent inflation rate. The problem is that once nominal wages are
set, interest rates fall as money growth rises. Real interest rate
targeting presents similar problems.
In order to make the macroeconomic model rich enough to
do a meaningful analysis of alternative intermediate monetary
targets, it was necessary to choose a rather simple game-theoretic
structure. In particular, we did not allow for reputational factors
that can be important in a multiperiod setting.21 However, al-
though reputational considerations can ameliorate the central
bank's credibility problems, they do not eliminate them except in
certain very special cases. Therefore, the main point of this paper
should extend to a more dynamic setting.

APPENDIX

Equations (Al) through (A4) give the expected value of the

social loss function (10) under fully discretionary monetary policy,
and rigid inflation rate, nominal GNP and money-supply target-
ing (the common term (h - n)2 is omitted):

(Al) AD ()L1 + X l +

(A2) A

(A3) AGITC = 1 + (1) + X 2

21 See Barro and Gordon [1983b], Backus and Driffill [1985], or Canzoneri
[1985].
 THE OPTIMAL DEGREE OF COMMITMENT 1189

2i
(A4) AMI =
CC L(-[1) + (X + X)J2

+ [- + x1[oT + ()1 2,

whereJ-(t - 1)/(1 -
UNIVERSITYOF WISCONSIN-MADISON

REFERENCES
Backus, David, and John Driffill, "Inflation and Reputation," American Economic
Review, LXXV (1985), 530-38.
Barro, Robert J., and David B. Gordon, "A Positive Theory of Monetary Policy in
a Natural-Rate Model," Journal of Political Economy, XCI (Aug. 1983a),
589-610.
,and , "Rules, Discretion and Reputation in a Model of Monetary Policy,"
Journal of Monetary Economics, XII (July 1983b), 101-21.
Canzoneri, Matthew B., Dale W. Henderson, and Kenneth Rogoff, "The Infor-
mation Content of the Interest Rate and Optimal Monetary Policy," this
Journal, XCVIII (Nov. 1983), 545-66.
Canzoneri, Matthew B., "Monetary Policy Games and the Role of Private Infor-
mation," American Economic Review, LXXV (1985), forthcoming.
Fischer, Stanley, "Long-Term Contracts, Rational Expectations, and the Optimal
Money Supply Rule," Journal of Political Economy, LXXXV (Feb. 1977),
191-206.
,and Franco Modigliani, "Towards an Understanding of the Real Effects and
Costs of Inflation," Weltwirtschaftliches Archiv, CXIV (1978), 810-33.
Friedman, Benjamin M., "Targets, Instruments, and Indicators of Monetary Pol-
icy," Journal of Monetary Economics, I (Oct. 1975), 443-73.
Gray, Jo Anna, "Wage Indexation: A Macroeconomic Approach," Journal of Mone-
tary Economics, II (April 1976), 221-35.
Kreps, David, and Robert Wilson, "Reputation and Imperfect Information," Jour-
nal of Economic Theory, XXVII (Aug. 1982), 253-79.
Kydland, Finn E., and Edward C. Prescott, "Rules Rather than Indiscretion: The
Inconsistency of Optimal Plans," Journal of Political Economy, LXXXV (June
1977), 473-92.
Obstfeld, Maurice, and Kenneth Rogoff, "Speculative Hyperinflations in Maxi-
mizing Models: Can We Rule Them Out?" Journal of Political Economy, XCI
(Aug. 1983), 675-87.
Parkin, Michael,. "A Comparison of Alternative Techniques of Monetary Control
under Rational Expectations," The Manchester School, XI (Sept. 1978), 252-87.
Phelps, Edmund S., "Phillips Curves, Expectations of Inflation and Optimal Em-
ployment Over Time," Economica, XXXIV (Aug. 1967), 254-81.
, "Inflation in the Theory of Public Finance," Swedish Journal of Economics,
LXXV (March 1973), 68-82.
Poole, William, "Optimal Choice of Monetary Instruments in a Simple Stochastic
Macro Model," this Journal, LXXXIV (May 1970), 197-216.
Sargent, Thomas J., Macroeconomic Theory (New York: Academic Press, 1979).