Vous êtes sur la page 1sur 6

American Economic Association

Independent Central Banks: Low Inflation at No Cost?

Author(s): Alberto Alesina and Roberta Gatti
Source: The American Economic Review, Vol. 85, No. 2, Papers and Proceedings of the
Hundredth and Seventh Annual Meeting of the American Economic Association Washington,
DC, January 6-8, 1995 (May, 1995), pp. 196-200
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2117918
Accessed: 22/10/2010 08:09

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

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.

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.


Independent CentralBanks:
Low Inflationat No Cost?


A widely held view suggests that politi- dard" exogenous shocks that monetary pol-
cally independentcentralbanksbringabout icy is supposed to stabilize, for instance,
relativelylow and stable inflation rates.' A money demand shocks or supply shocks.
more debated question is whether one has The second source of variability is "politi-
to "pay" for this good outcome with more cal" or, more generally, policy-induced. This
"real"instability. is the variability introduced in the system by
In his seminal contribution, Kenneth the uncertainty about the future course of
Rogoff (1985) suggeststhat an independent policy. For instance, Alesina (1987) studies
and inflation-aversecentral bank reduces the effect of uncertain electoral outcomes in
average inflation but, as a result, increases a model where the two contending parties
output variability;the "conservative"cen- have different preferences over inflation and
tral banker reduces the inflation bias, due unemployment.
to the time-inconsistencyproblem,but sta- An inflation-averse, independent central
bilizes less. banker does not stabilize as much the "eco-
However,Alesina and Summers(1993)do nomic" variability, in order to keep inflation
not find that, at least within the OECD low and stable. This is Rogoff's point. How-
countries, more independent central banks ever, by insulating monetary policy from
are associated with more variability of political pressures, an independent central
growth or unemployment.Thus, they con- bank can reduce the "political" variability.
clude that independent central banks bring The overall effect of independence on out-
about low inflation at no apparent "real" put variability is, thus, ambiguous. This re-
costs. The point of this paper is to provide sult is consistent, at least prima facie, with
theoretical underpinningsto this finding, the evidence in Alesina and Summers (1993)
which is in contrast to Rogoff (1985).2 on the lack of correlation between central-
The basic idea is that one can isolate two bank independence and output variability.
sources of output variability. One is the In fact, it is possible that when the politi-
"economic" variability induced by "stan- cally induced output variability is predomi-
nant, a more independent central bank re-
duces average inflation and the variance of
tDiscussants:Michael Bruno, World Bank; David
Mullins,Long-TermCapitalManagement,New York; I. RogofifsModel
Donald Kohn,FederalReserveBoard.
*Alesina:Departmentof Economics,HarvardUni- In a nutshell, Rogoff's (1985) model is as
versity, Cambridge,MA 02138, NBER, and CEPR; follows: Output (y) is given by
Gatti:Departmentof Economics,HarvardUniversity.
1For empiricalevidence on this point concerning
OECD countriessee Alesina and LawrenceSummers
(1993). For evidence on OECD and LDC's, see Alex (1) Yt t t +?t
This short paperdescribesthe model and provides
the intuition for the results. For a more complete where 7 is inflation is expected infla-
treatment,see Alesina and Gatti (1995). 7re
tion, and Et is an independently and identi-

cally distributed shock with mean zero and optimal policy rule, which, in this model, is
variance o-a2. Thus, the "natural" level of contingent on the realization of E. Through-
expected output, with Jt = 7t , is normal- out the paper we assume, realistically, that
ized at zero. With no loss of generality, commitment is not an available option to
parameters are set equal to 1. Expectations the policymakers.
(i.e., wage contracts) are set before the shock Rogoff (1985) notes that social welfare
is realized and before the policymaker can be increased if the policymaker dele-
chooses n-t. Thus, the timing is: first, rt gates ex ante the choice of monetary policy
then Et, and finally vt, which is the policy to an independent agent, chosen before ev-
instrument. The policymaker's loss function erything else. Independence implies that the
is agent cannot be dismissed ex post, when he
has to choose policy. The policymaker can
1 2
b 2 select an agent with a preference parameter
(2) L -r+
2 t 2 (tk b different from his own, if he so wishes.
The timing in each period is as follows:
where b > 0 and k > 0. The arguments un- First the policymaker chooses an agent from
derlying this loss function and, in particular, a population with different parameters bs.
the reason why the target output, k, is above Then expectations are formed; next the
the "natural" level, zero, are well known shock E occurs; and finally the agent chooses
since the paper by Robert Barro and David policy. In the following period, the policy-
Gordon (1983). maker can change the agent, but since all
Substituting (1) in (2), taking first-order periods are identical to each other, the opti-
conditions with respect to 7rt, solving for mal choice for the policymaker is as follows:
rational expectations, and dropping time
subscripts one obtains (6) minE(L(b,b))
(3) 7r=bk- 1+E
1+ b =E( (bk - 1+ )+ - l+b - k)

(4) 7Je= bk
where b is the parameter of the loss func-
1 tion of the agent. Equation (6) embodies
(5) Y l+b the knowledge that, once appointed, the
agent will follow rule (3) with b instead of
The policy rule (3) embodies the well-known b. Since the policymaker is acting first, be-
inflation bias (bk) and a stabilization term fore rJe and E are realized, he has to take
([b/l + bkE).From (3)-(5) it follows that expectations over the shock.
The solution of problem (6) delivers
E(7n)=bk E(y)=0 Rogoffs (1985) result that 0 < b < b. The
policymaker improves his own utility (and
/b \2 social welfare if they coincide) by delegating
Var(n-)= ( 2 2 monetary policy to an agent who is more
inflation-averse (i.e., more "conservative")
1 than the policymaker himself. The key is
Var(y)= 1 b)2 that the agent is independent (i.e., he can-
not be removed ex post).3
where E(*) indicates expected values.
Thus, the policy rule reduces output vari-
ance but induces positive average inflation, 3In fact, after expectations are set, the policymaker
has an incentive to remove the agent, choose monetary
without increasing average output. The in- policy directly, and be time-inconsistent, causing unex-
flation bias can be eliminated if the policy- pected inflation. Thus, central-bank independence alle-
maker could commit ex ante to follow the viates a problem of time-inconsistency.

Comparing the solutions with and without

the independent central bank, it is straight- (8) J7R b=(+bk
forward to show that, with the independent (1 + bD)- P(bD - bR) k
agent, inflation is lower and more stable,
but output is more variable, while expected
output is the same. bR
1 + bR
II. Political Uncertainty

Following Alesina (1987), consider the Thus, it follows that, if D is elected,

case of two competing parties, D and R,
with the following loss functions
(1 D -b R)
( g) y D= P)p(b k
D 1
=72 2bD 2
(1 + bD) -P(b D - bR)k
LD +-y(y-k)

1 bR 2 1
LR = _72 + -(y-k 2
2 2 1 + bD

where 0 < bR < bD. Thus, party D cares

more about output stabilization relative to If, instead, R is elected, we have
inflation than party R. The timing of events
is as follows: first, expectations (wages) are
set; then elections take place. Party D wins (10) yR
P(bD bR) k
with probability P, which is exogenous; party
R wins with probability 1-P. After the (1 + bD)- P(b D-b R)
election, the shock E occurs; finally, the
party in office chooses policy.4 The same 1
sequence of events is repeated in every pe-
riod. Thus, a "period" coincides with the 1 + bR
length of a wage contract and with a term of
office. Nothing hinges on the latter assump- From (9) and (10) we then obtain our key
tion: we could have multiperiod terms of equation:
office. The crucial point instead is that the
inflation expectation embodies electoral un-
certainty: 7e = PEO,TD)+(1 - P)E(7rR).
Taking this into account, the policies cho- (11) Var(y)=
sen by the two parties if in office are [(1 + bD) -p(bD - bR)]2

b D(1 +R)
(7) 7D (1 + bD) - P(bD - bR) k
+ - p
(1 +bD)2 + ~~~(1-R22 0-2.
R1 jb

- +bD
The variance of output can be decom-
posed into two parts. The first term in (11)
is the politically induced variance. This term
reflects the fluctuations of output induced
4The probability of electoral results could be made
exogenous as a function of individual preferences, fol- by the electoral uncertainty. It can be shown
lowing work by Alesina and Howard Rosenthal (1995). to be increasing in (bD - bR), namely, with
Since this is not our focus, we keep things simple here. the difference in parties' preferences; it also

vanishes if P = 0 or P = 1, namely, with no Without an independent central bank we

electoral uncertainty. The second term in have instead
(11) derives from the exogenous shock v,
the effects of which are dampened by the
bR(1 + bD) + P(bD - bR)
stabilization terms in (7) and (8).
Ev=(1 + bD)- P(bD -bR)
Suppose now that the two parties con-
sider the appointment of an independent
central banker. Specifically, we ask the fol- E(y) = 0
lowing question: can the two parties im-
prove upon the outcome described above by
agreeing before the election to appoint an Pk2[(bD)2(1 + bR)2_ (bR)2(l + bD)2]
independent central banker who then, after Var(7)= [(1+bD) p(bD bR)]2
the election, chooses policy and cannot be
removed from office? More precisely, con-
sider the following timing: first the two par- Pk2[ - 2bR(bD - bR)(1 + bD) - p(bD - bR)2]
ties appoint a central banker, namely, they
[(1 + bD) -P(bD
agree on an agent characterized by a certain -bR)]2

b in his loss function; then expectations are

formed, followed by elections; then s is
bD bR
realized; and finally, the independent agent 0-2 p
+ 1 P)
chooses policy. +
+ bD 1+bR
+b 2j

For the parties to achieve an agreement, D(1-+(1D-PR)( I

it must be that for each of them the ex-
pected loss when acting "noncooperatively" P(lP)(bD-bR)2k2
(i.e., choosing policy directly after the elec- (13) Var(y) =
[(1 + bD) - p(bD - bR)]2
tion) is larger than the expected loss when
the appointed agent chooses policy inde-
+ [+D)2 + (+R)2]
In this setting we would expect the inde-
[(I + b) (I + bR
pendent agent to have two benefits: first, to
reduce the inflation bias problem, as in
Rogoff (1985); second, to eliminate politi- It can be shown that an appropriate
cally induced output variance. choice of b, in the range of the parameters
We are now ready to illustrate the punch- that make both parties better off, allows
line of this paper, which is a comparison of them to achieve a lower expected inflation.
the outcomes with an independent central The variance of inflation can also be lower.
bank and without. With the independent More importantly, the variance of output
central bank we have in (13), without an independent central
bank, can easily be larger than the vari-
ance of output with an independent and
E(,i-)=bk E(y)=0
inflation-averse central bank, (12). To see
this, consider the extreme case in which
1b 2 bR = bD = b, and b <b. A comparison of
Var(r) = ( bA (2 (12) and (13) in this case reproduces the
standard Rogoff (1985) model. If bD bR,
the parties have similar preferences for in-
1 flation, and only a little political uncertainty
(12) Var(y)= - A2 is introduced in the system. As (bD-bR)
(1+ b) increases, the role of the "political" vari-
ance also increases. For (bD- bR) suffi-
where b is the parameter characterizing the ciently large (relative to q2), the first term
appointed agent. in (13) becomes predominant, and the vari-

ance of output is significantly lower with an real economic variability. In other words, an
independent central bank.5 independent central bank can achieve both
lower average inflation and lower output
III. Discussion and Conclusion variability. The empirical results in Alesina
and Summers (1993) which show a strong
The institution of an independent and positive correlation between independence
inflation-averse central bank has two bene- and inflation, but no correlation between
fits: first it reduces average inflation; sec- independence and output (or unemploy-
ond, it eliminates "politically induced" out- ment) variability, are consistent with this
put variability, since monetary policy is not model.
under the direct control of governments with
changing preferences. The elimination of REFERENCES
this type of real variability may or may not
more than compensate for the increased Alesina,Alberto."Macroeconomic Policy in a
variability caused by the fact that the infla- Two-Party System as a Repeated Game."
tion-averse central bank does not make a Quarterly Journal of Economics, August
big effort at stabilizing output shocks. 1987, 102(3), pp. 651-78.
The basic thrust of these results should Alesina, Alberto and Gatti, Roberta. "Central
be quite general. We have modeled a rather Bank Independence and Political Cycles."
specific form of "political" uncertainty: un- Unpublished manuscript, Harvard Uni-
certain election results in a two-party sys- versity, 1995.
tem. However, other forms of "political" or Alesina,Albertoand Rosenthal,Howard.Parti-
policy-induced uncertainty may bring about san politics, divided government and the
similar considerations. For instance, the case economy. Cambridge: Cambridge Univer-
of uncertainty about political outcomes in sity Press, 1995.
coalition governments with shifting weights Alesina, Alberto and Summers, Lawrence.
within the coalition; unpredictable out- "Central Bank Independence and Macro-
comes of interparty struggles between fac- economic Performance: Some Compara-
tions; and so on. From an empirical stand- tive Evidence." Journal of Money, Credit,
point, these results suggest that one should and Banking, May 1993, 25(2), pp. 151-62.
not expect that an increase in the measured Barro,Robertand Gordon,David."Rules, Dis-
degree of independence of central banks is cretion, and Reputation in a Model of
necessarily associated with an increase in Monetary Policy." Journal of Monetary
Economics, July 1983, 12(1), pp. 101-22.
Cukierman,Alex. Central bank strategy, credi-
sAn interesting question is how the two parties bility, and independence. Cambridge, MA:
would choose a particular b within the range that MIT Press, 1992.
makes both of them better off relative to the noncoop- Rogoff, Kenneth. "The Optimal Degree of
erative solution. It is easily verifiable that for bD = 3, Commitment to an Intermediate Mone-
bR = 0, P = 1/2, k2 = 1, and o-,2= 1 a choice of b = 0.5
would benefit both parties by lowering average infla-
tary Target." Quarterly Journal of Eco-
tion and output variability. For a detailed discussion nomics, November 1985, 100(4), pp.
see Alesina and Gatti (1995). 1169-90.