Full text of DOI:10.1162/003355300554737

THE NONNEUTRALITY OF MONETARY POLICY
WITH LARGE PRICE OR WAGE SETTERS*
DAVID SOSKICE AND TORBEN IVERSEN
Monetary rules matter for the equilibrium rate of employment when the
number of price-wage setters is small, even when assuming rational expectations,
complete information, central bank precommitment, and absence of nominal
rigidities. If the central bank is nonaccommodating, sufficiently large unions,
bargaining independently, have an incentive to moderate sectoral money wages,
and thereby expected real wages. The result is an increase in the real money
supply, and hence higher demand and employment. This does not hold with
accommodating monetary policy since unions’ wage decisions cannot then affect
the real money supply. A similar argument holds for large monopolistically
competitive price setters.
I. INTRODUCTION
Under a wide range of assumptions, the choice by a central
bank of a monetary rule does not affect the equilibrium rate of
employment. We show in this article, however, that with a finite
number of wage or price setters this is no longer necessarily the
case. In particular, we show that a switch by the central bank from
an accommodating to a nonaccommodating monetary rule leads to
an increase in the equilibrium rate of output or employment, and
that this increase is greater the smaller the number of price or
wage setters. The result does not challenge the weak neutrality of
money theory; given the choice of monetary rule, a change in the
money supply has no effect on real variables.1 Rather, it shows
that the strategic interaction of price-wage setters and monetary
authorities can have important effects on the equilibrium rate of
output and employment. In other words, with a finite number of
wage or price setters, the character of the monetary rule is
nonneutral. While it may appear at first sight counterintuitive
that an increase in nonaccommodation should increase equilib-
rium employment, we show that this is consistent with data for
seventeen OECD economies over the period 1973–1993.
* We would like to thank Christopher Allsopp, Thomas Cusack, Robert
Franzese, J effry Frieden, Peter A. Hall, William Novshek, and in particular
Alberto Alesina, Olivier Blanchard, and two anonymous referees for their many
useful comments and suggestions. The first author benefited from the environ-
ment of the RSSS in the Australian National University in revising this article.
1. Bleaney [1996] shows this to be the case under a similar set of assumptions
used here. This result is simply a restatement of the money neutrality thesis—a
thesis we obviously do not challenge.
2000 by the President and Fellows of Harvard College and the Massachusetts Institute of
Technology.
௠
The Quarterly J ournal of Economics, February 2000
265

266
QUARTERLY J OURNAL OF ECONOMICS
A sufficient set of assumptions under which choice of mone-
tary rule has no effect on equilibrium employment are (i) rational
expectations, (ii) perfect competition, (iii) complete information,
(iv) capacity for the central bank to credibly precommit, and (v)
absence of nominal rigidities. A large literature on central banks
has subsequently shown that the core conclusion is robust to
changes in assumptions (iii), (iv), and (v).2 We deliberately retain
all these assumptions (in addition to (i)), and show instead what
happens when assumption (ii) is dropped. In its place, we present
two general equilibrium models, both based on Blanchard and
Kiyotaki [1987] and Blanchard and Fischer [1989], which assume
that either prices are set by monopolistically competitive produc-
ers or wages are set by monopoly unions. The first, a ‘‘yeoman
farmer ’’ model of monopolistic competition in which each of N
farmers produces unaided one product and consumes all N
products, is simple, but gives the intuition of our argument. This
is set out in Section II. The second model assumes a fully
unionized multisector economy in which monopoly unions in each
sector set wages independently and simultaneously. This is set out
in Section III. In both cases, the economic agents care only about
real variables, so there is no money illusion.3
The key to our argument is that price or wage setters with
some monopoly power can affect the real money supply depending
on the monetary rule used by the central bank. If the central bank
fixes the nominal money supply, the effect of increased sectoral
prices or wages on aggregate prices will translate into reduced
real demand and hence lower output or employment. There is
therefore an incentive for producers or unions to lower their
nominal prices or wages relative to those expected in other
2. In Barro and Gordon, for example, inability of the central bank to credibly
precommit to a monetary rule raises inflation, but has no effects on equilibrium
unemployment (see also Kydland and Prescott [1977]). Other work has focused on
the consequences of assuming incomplete information. For example, when voters
do not know how competent governments are in producing public goods for a given
fiscal revenue, governments may try to signal their competency by engaging in
expansionary preelection fiscal and monetary policies (see Rogoff and Sibert
[1988], Cukierman and Meltzer [1986], and Rogoff [1990]). Other models assume
nominal rigidities in the presence of economic business cycles that create a
trade-off between inflation conservatism (which reduces inflation) and flexibility to
respond to exogenous shocks (which reduces employment variability). This has
consequences of the optimal design of central bank contracts, but does not affect
equilibrium employment. For examples of this approach, see Rogoff [1985],
Lohmann [1992], Persson and Tabellini [1993], and Svensson [1996].
3. This contrasts with a recent paper by Cukierman and Lippi [1999] which
assumes that the rate of inflation matters to unions. Absent this, the paper shows a
nonneutrality result (in a different direction to ours) only when relative wages as
well as real wages are added to the employment demand function.

THE NONNEUTRALITY OF MONETARY POLICY
267
sectors; and hence to increase production or employment. The
smaller the number of independent monopoly price-wage setters,
the greater the individual effect on the aggregate price level, and
the greater the incentive for real price-wage restraint. In the case
of unions, the standard trade-off between the real wage and the
employment level along the sectoral employment demand curve is
thus altered: greater real wage restraint lowers the relative price
of the sector, and this raises the real money supply (given
wage-price setting elsewhere). By contrast, if the central bank
fixes the real money supply, the incentive to exercise restraint is
absent since unions in this situation cannot affect real demand
and hence employment. Because all unions reason similarly, real
wage restraint and equilibrium employment are higher, the fewer
the number of unions, and the more nonaccommodating the
monetary regime.
Since wage-setting tends to be more monopolized than price-
setting, the argument applies to all economies with a small
number of independent unions. It does not apply to economies
where there are a very large number of unions (or where
individuals bargain wages), because here the aggregate price
effects of individual unions are too small. Nor does it apply to
economies where a single union sets wages, or where a small
number of unions can coordinate their wage policies effectively
enough to act as a single union, since the union(s) can then choose
full employment independently of the monetary regime.4 Rather,
the argument is intended to solve, and is motivated by, a
persistent empirical puzzle in comparative political economy: the
capacity for good unemployment performance by some, but not
other, economies with intermediately centralized systems of wage-
setting. Such systems are often portrayed as institutional dilem-
mas where large unions are capable of inflicting great harm on the
economy, yet incapable of solving their collective action problems
[Olson 1982; Calmfors and Driffill 1988]. But while it is true that
some of these systems have performed poorly, it is frequently
noted that others are among the most successful, and institution-
ally stable, economies in the world (see Soskice [1990], Hall
[1994], and Iversen [1999]).
The purpose of this article is analytic rather than empirical,
but the data in Table I help to illustrate the puzzle (the table is
4. The case of multilevel bargaining is more complicated; see Iversen
[1998,1999].

268
QUARTERLY J OURNAL OF ECONOMICS
TABLE I
LONG-RUN AVERAGE UNEMPLOYMENT RATES FOR 17 OECD COUNTRIES DEPENDING
ON THE MONETARY RULE AND THE CENTRALIZATION OF WAGE BARGAINING,
1973–1993 (N, IN PARENTHESES)
Centralization
Very high
Intermediate
Very low
Mean:
Monetary rule
Accommodating
Nonaccommodating
Difference
3.9 (3)
5.6 (2)
Ϫ1.7
7.6 (3)
3.6 (4)
4.0*
7.1 (3)
7.4 (2)
Ϫ0.3
6.2 (9)
5.0 (8)
5.6 (17)
* Difference is significant at a .001 level (one-sided test).
Monetary rule refers to the independence of the central bank except in the case of J apan where
a
dependent bank has followed a nonaccommodating rule. Centralization refers to the prevalent level and
concentration of wage setting. Very high centralization means that a small number of national unions or
confederations bargain wages and routinely coordinate their wage demands. Very low centralization means
that wages are predominantly bargained at the plant or firm levels by local unions or individuals.
Intermediate cases are characterized by a limited number of bargaining areas, but without any effective
coordination of union wage demands. Beginning with the top row, the countries in the table are from left to
right (with brackets demarcating cells): [Finland, Norway, Sweden]; [Australia, Belgium, Italy]; United
Kingdom, France, New Zealand]; [Austria, Denmark]; [Germany, J apan, Netherlands, Switzerland]; [Canada,
United States].
Sources: OECD [1996, 1997]; Cukierman, Webb, and Neyapti [1992]; and Hutchison, Ito, and Cargil
[1997].
literally only illustrative). The table shows the equilibrium rate of
unemployment in seventeen OECD countries characterized by
different types of unions and monetary rules.5 With one exception,
we use Cukierman, Webb, and Neyapti’s [1992] index of legal
central bank independence as a proxy for the monetary rule,
dividing the sample into an accommodating and a nonaccommodat-
ing category.6 With regard to the number of unions, we use
centralization of wage bargaining as a proxy. Although there is no
consensus on the classification of every bargaining system, most
industrial relations specialists agree on the ones that are either
highly decentralized (Canada, France, New Zealand, the United
Kingdom, and the United States), or highly centralized (Austria,
Denmark, Finland, Norway, and Sweden). We treat the remainder
5. Equilibrium unemployment rates are proxied by OECD’s [1996] estimates
for the nonaccelerating wage rates of unemployment in the period 1973–1993 (i.e.,
post-Bretton Woods), except in the cases of Switzerland and New Zealand where
these are not available (we use national definitions instead).
6. The exception is J apan where the monetary authorities began to adhere to
an unambiguously nonaccommodating rule after 1973, even though the central
bank remained legally dependent (see Cargil [1993], Hutchison and J udd [1989],
and Hutchison, Ito, and Cargil [1997]). This is recognized by Cukierman, Webb,
and Neyapti [1992] who attribute nonaccommodating monetary policies in J apan
to the influence of an exceptionally ‘‘conservative’’ Ministry of Finance [1993, p.
372].

THE NONNEUTRALITY OF MONETARY POLICY
269
(Australia, Belgium, Germany, Italy, J apan, Netherlands, and
Switzerland) as intermediate cases.7
Note that monetary rule appears to be largely unrelated to
unemployment when the number of unions is either very large or
very small, but that nonaccommodating rules are associated with
significantly lower unemployment than accommodating rules
when the number of unions is intermediate.8 Although data of this
nature can only be suggestive, a number of more detailed empiri-
cal studies support the notion that monetary rules matter for
unemployment when their interactive effects with the centraliza-
tion of wage-setting are taken into account (see Iversen [1998,
1999], Hall [1994], Hall and Franzese [1998], and Garrett and
Way [1995]). Specifically, if our argument is correct, it would help
us to understand why countries with intermediately centralized
wage-setting systems are characterized by such widely divergent
long-term unemployment performance.
II. THE YEOMAN FARMER ECONOMY
To convey the basic logic of the argument, we begin by
considering a simple ‘‘yeoman farmer ’’ economy with monopolisti-
cally competitive producers who provide their own labor input. As
in Blanchard and Fischer [1989, pp. 376–380] we assume, with
one modification, that there are N identical monopolistic competi-
tors each producing a single good indexed by i with price Pi and
output Qi (i ϭ 1, . . . , N ). (The modification is that in each sector,
while price and total output are decided by a single ‘‘marketing
agent,’’ we assume that there are many identical producers whose
production quota is set by the agent but who take individual
consumption decisions; the agent maximizes their utility indi-
rectly. This enables us to focus on strategic price-setting behavior
when there are a small number of sectors, while avoiding the
complications of strategic consumption behavior. It requires mi-
nor changes—set out in the Appendix—to the Blanchard-Fischer
argument). Blanchard and Fischer conclude that the macroeconom-
7. The centralization of wage bargaining in different countries is reviewed in
a recent OECD study [1997]. Our classification is consistent with the most
frequently used classifications of centralization. See also Flanagan, Soskice, and
Ulman [1983], and Iversen [1998].
8. Any attempt to capture these relationships in a simply additive model
would fail, and it is not surprising that past studies employing a simple additive
methodology have not found any strong employment effects of either central bank
independence [Alesina and Summers 1993; Bleaney 1996], or bargaining structure
[OECD 1997].

270
QUARTERLY J OURNAL OF ECONOMICS
ics is unaltered as we move from perfect competition to monopolis-
tic competition, and that ‘‘. . . money is neutral under monopolistic
competition just as it is under perfect competition’’ [p. 381].
However, if the monopolistic competitor ’s price Pi has a finite
effect on the general price level, changes in the monetary rule may
be nonneutral. We assume that the central bank can precommit to
a monetary rule, which we define by a parameter ␣:9
(1)
M ϭ P ␣
0 Յ ␣ Յ 1.
Specifically, two limiting cases can be distinguished. When ␣ ϭ 0,
M ϭ 1, the monetary rule consists in fixing the nominal money
supply equal to unity, and thus does not accommodate any price
increases. When ␣ ϭ 1, M/P ϭ 1, and the monetary rule consists in
fixing the real money supply equal to unity, and thus accommo-
dates all price increases.10 It is convenient to assume that central
banks are endowed with a fixed value of ␣ which might be thought
of as the result of past reputation-building. In what follows, we
assume a sequence of three moves: (1) the central bank precom-
mits to setting M contingent on P. (2) The farmers’ agents each set
Pi simultaneously, in the knowledge of the monetary rule. (3) The
central bank then sets M given P (P1, . . . , PN). Since stage 3 is
automatic, the game is in effect concentrated on stage 2. If Ui is
the objective function of the ith agent, we show that the ith agent
*
*
*
*
should choose P i such that Ui(P i, P Ϫi) Ն Ui(Pi, P Ϫi) for all i,
*
*
where P i is the equilibrium price of the ith agent and P Ϫi is the
vector of equilibrium prices of the N Ϫ 1 agents.
Using the Blanchard and Fischer model [1989, p. 433] with
minor notational changes, the demand for good i is given by
(2)
Qi ϭ m/N · pϪi ␩
,
where pi ϵ Pi/P is the relative price, and m ϵ M/P is the real
money supply. (It is shown in the Appendix how (2) is derived from
individual maximization behavior by farmers. This relies on the
assumption that the agent can fix the supply of the farmer ’s
output to equilibrium demand.) The aggregate price level P is
defined as a constant elasticity of substitution (CES) index:
1/(1Ϫ␩)
1
(3)
P ϭ
·
P1i Ϫ␩
.
͚
11 2
2
N
9. We are indebted to Chris Allsopp for suggesting this form.
10. When ␣ ϭ 1, M/P could be set equal to any constant other than 1 without
affecting the results (as we discuss below). Using 1 is for convenience.

THE NONNEUTRALITY OF MONETARY POLICY
271
Assuming that there are a fixed number of n farmers in each
sector, with constant returns to scale production, it is shown (also
in the Appendix) that maximization of the direct utility function of
the farmer with CES preferences over the N goods, disutility of
labor, and utility of real money balances, and who makes optimal
consumption and real cash balance choices, is secured by maximi-
zation by the farmer ’s agent of the following indirect utility
function:
(4)
max Ui ϭ pi · Qi Ϫ (d/␤) · Q␤i ϩ m/N,
p
i
where dЈ/␤ · (Qi/n)␤ is the individual farmer ’s utility cost in terms
of own-labor of production Q (␤ Ͼ 1),11 with d ϵ dЈ/n␤Ϫ1, and
nϪ1 · ( pi · Qi ϩ m/N ) is the farmer ’s wealth.12 Substituting (2)
into (4) and differentiating with respect to pi gives the first-
order condition for utility maximization (see footnote 13). We
focus on a symmetric equilibrium, implying that pi ϭ 1 and hence
Qi ϭ m/N, for all i. Together with the first-order condition this
yields13
1/(␤Ϫ1)
␩ Ϫ 1 Ϫ 2Ѩ ln m/Ѩ ln pi
(5)
Qi ϭ Q* ϭ
.
1
2
d␩ Ϫ d · Ѩ ln m/Ѩ ln pi
where Q* is the equilibrium level of total output per sector (so the
output per farmer is Q*/n). The crucial term in (5) is Ѩ ln m/Ѩ ln pi,
11. This is identical to equation (3) in Blanchard and Fischer [1989, p. 378].
12. We assume that a change in M (when the central bank’s monetary rule is
applied after prices have been set) is distributed across the farmers in proportion
to their original holdings of M.
13. To see this, first substitute (2) into (4) to derive the first-order condition:
␤
m
1N2
m
(1 Ϫ ␩) и pϪi ␩
и
ϩ d␩pϪi ␩␤Ϫ1
и
1N2
␤Ϫ1
1
1N2
Ѩm
Ѩpi
d
1N2
m
Ѩm
Ѩpi
ϩ (1 ϩ p1i Ϫ␩) и
и
Ϫ
pϪi ␩␤
и
и
ϭ 0.
1N2
Then divide through by (m/N ) and impose the equilibrium conditions ( pi ϵ 1 and
Qi ϭ m/N ) to obtain
Ѩ ln m Ѩ ln m
(1 Ϫ ␩) ϩ d␩Q␤i Ϫ1 ϩ 2
ϭ
· dQ␤i Ϫ1
,
Ѩ ln pi
Ѩ ln pi
which can be reformulated as in (5). The second-order condition that d2U/dp2 Ͻ 0 is
satisfied; the proof is available from the authors upon request.

272
QUARTERLY J OURNAL OF ECONOMICS
and we need to evaluate this. Since M ϭ P␣, m ϭ P(␣Ϫ1), and
ln m ϭ Ϫ(1 Ϫ ␣) · ln P. Hence,
Ѩ ln m
Ѩ ln pi
Ѩ ln PE(i)
Ѩ ln pi
Ѩ ln PE(i) Ѩ ln Pi
ϭ Ϫ(1 Ϫ ␣) ·
ϭ Ϫ(1 Ϫ ␣) ·
·
,
(6)
Ѩ ln Pi Ѩ ln pi
where PE(i) is the ith agent’s expectation of P. Note that, if PjE,(i) is the
value of Pj expected by the ith agent, PEj,(i) is independent of Pi since
agents set prices simultaneously and independently. So from (3)
Ѩ ln PE(i)
Ѩ ln Pi
1
P
1
␩Ϫ1
(7)
ϭ
·
ϭ
1 2
N
Pi
N
given that P ϭ Pi in equilibrium. Consequently,
Ϫ1
Ѩ ln Pi
Ѩ ln pi
Ѩ(ln Pi Ϫ ln PE(i)) Ϫ1
Ѩ ln PE(i)
Ѩ ln Pi
N
ϭ
ϭ 1 Ϫ
ϭ
.
1
2
1
2
(8)
Ѩ ln Pi
NϪ1
Substituting (7) and (8) into (6) yields
Ѩ ln m
Ѩ ln pi
1
N
1Ϫ␣
(9)
ϭ Ϫ(1 Ϫ ␣) ·
·
ϭ Ϫ
,
N
NϪ1
NϪ1
and, substituting (9) into (5), the farmers’ agent thus sets Qi in
equilibrium to satisfy
␩Ϫ1 ϩ 2 (1 Ϫ ␣)/(N Ϫ 1) 1/(␤Ϫ1)
(10)
Q* ϭ
.
1
2
d␩ ϩ d · (1 Ϫ ␣)/(N Ϫ 1)
Note that if the monetary rule is completely accommodating, ␣ ϭ
1, or if N ϭ ϱ, we get the Blanchard-Fischer result [1989 p. 380,
equation 7]:
1/(␤Ϫ1)
␩Ϫ1
(11)
Q* ϭ
.
1
2
d␩
Comparing (10) and (11), we can see that the right-hand side of
(10) is greater than that of (11) as long as ((1 Ϫ ␣)/(N Ϫ 1)) Ͼ 0.14
Thus, we reach the following conclusion: in the yeoman farmer
monopolistic competition model, if ␣ Ͻ 1, so that monetary policy is
14. Writing ((1 Ϫ ␣)/(NϪ1)) as K, the right-hand-side bracket of (10) becomes
(␩ Ϫ 1 ϩ 2K)/(␩ ϩ K) ϭ ((␩ Ϫ 1)/␩) · (((1 ϩ 2K)/(␩ Ϫ 1))/((1 ϩ K)/␩)) and this is
greater than the right-hand-side bracket of (11) when 2K␩ Ͼ K · (␩ Ϫ 1). Since by
assumption ␩ Ͼ 1, this requires K Ͼ 0.

THE NONNEUTRALITY OF MONETARY POLICY
273
not completely accommodating, and N is finite (but greater than 1),
equilibrium output will be higher than when monetary policy is
completely accommodating; the difference will be larger the smaller
is N, and the smaller (i.e., the more nonaccommodating) is ␣.
The reason for these effects is that monopolistic competitors
will take into account their own effect on the aggregate price level,
and hence the real money supply, when they set prices; and this
effect is contingent on the monetary rule. Given ␣ Ͻ 1, if the ith
agent is to reduce the aggregate price level P, that implies Pi must
fall relative to prices in other sectors. Since all sectors are
identical, the output, and thus marginal cost in terms of own-
labor, of each farmer must be sufficiently large to eliminate the
incentive to engage in further relative price-cutting. This will
then be the equilibrium output and price level.
Another way of looking at this is in comparative static terms.
Suppose that the central bank reduces ␣, thus tightening mone-
tary policy. With small enough N, each agent will now have an
increased incentive to reduce his price (hence to undercut the
others) to take advantage of the increased impact that this will
have on the real money supply. In effect, the real marginal
revenue curve has shifted out because the elasticity of the real
money supply and hence demand with respect to the sector ’s
relative price, Ѩ ln m/Ѩ ln pi, has increased; this elasticity plays an
exactly parallel role to the direct elasticity of demand. Competi-
tive undercutting will thus continue until the equilibrium be-
tween marginal revenue and marginal costs has been restored at
a higher level of output, with a higher real money supply and a
lower aggregate price level.
This does not mean, however, that output is also nonneutral
with respect to the nominal money supply. To see this, we can
break the monetary rule into two parts: (a) the degree of accommo-
dation of the central bank, and (b) the scale of the nominal money
supply. The accommodation part of the rule is
Ѩ ln m
(1a)
ϭ Ϫ(1 Ϫ ␣) .
Ѩ ln P
Integrating (1a) gives us
(1b)
m ϭ K · PϪ(1Ϫ␣)
.
In the models of this article we have taken K ϭ 1 to simplify the

274
QUARTERLY J OURNAL OF ECONOMICS
presentation. But in general K is the nominal scale part of the
rule. Defining
(1c)
K ϭ µ1Ϫ␣
,
we have that P ϭ m*1/(1Ϫ␣) · µ so that equilibrium output is neutral
with respect to µ, the nominal scale parameter.
But when N is finite, equilibrium output is nonneutral with
respect to the degree of accommodation in the monetary rule (␣),
and the main objective of this section has been to show that the
analysis of macroeconomics changes fundamentally as we move
from a large number of monopolistic competitors to a small
number; hence also from a perfectly competitive economy to a
world of imperfect competition with a small number of price
setters. It can be objected, however, that the practical significance
of this insight is limited since price setting rarely occurs at a level
that would make general price considerations relevant for indi-
vidual pricing agents. But this objection need not apply to wage
setting, which is frequently conducted at the industry, sector, or
even national level. Indeed, relatively centralized wage setting is
the rule rather than the exception in a majority of OECD
countries [OECD 1997]. Because centralized wage setters exert
an effect on the general price level, our argument has considerable
practical importance for understanding real wage behavior and
hence employment performance. In the following section we show
this by introducing sectoral labor markets where each sectoral
wage is set by the sectoral monopoly union. To ensure that our
results do not depend on large price setters, we assume that there
are two or more Bertrand competitors in each sectoral product
market.
III. A MODEL WITH MONOPOLY UNIONS
Nominal wages Wi are set in each of N identical sectors by a
sectoral monopoly union; there are n workers in each sector, all of
whom are union members, and there is no mobility between
sectors. As before, the demand for the good in each sector is a
constant elasticity function of the relative price of the sector ’s
good multiplied by the real supply of money deflated by the
number of sectors; this is equation (2). Hourly labor productivity
is constant and equal to unity, and there are two or more Bertrand
competitors in each sector; thus, the price in sector i, Pi, is equal to
the constant marginal cost Wi. Producers have to hire at the

THE NONNEUTRALITY OF MONETARY POLICY
275
union-determined wage, which determines total hours employed
Ei, and each worker is hired for the same number of hours, Ei/n.
Wages and prices in each sector are set independently of wage and
price setting in other sectors. Finally, the aggregate price level is
derived from a constant elasticity of substitution index of the
individual sectoral prices, as in equation (3) in the last section.15
When analyzing this case, it is helpful to keep in mind the
implied game of complete and quasi-perfect information. The
game consists of four stages.
Stage 1: The CB precommits to the monetary rule implied by ␣;
Stage 2: unions simultaneously and independently choose Wi
(i ϭ 1, . . . ,N );
Stage 3: producers simultaneously and independently set Pi
and Ei (i ϭ 1, . . . , N );
Stage 4: the CB sets M contingent on P as predetermined by ␣
(for all ␣ Ͼ 0 since ␣ ϭ 0 is a noncontingent strategy).
The solution is found by backward induction, but in Stage 4 the
central bank acts as an automaton in setting M equal to P ␣ (as a
result of the precommitment assumption); in Stage 3 Bertrand
competitors in each sector i set Pi ϭ Wi; and in Stage 1 the CB
precommits to a monetary rule (␣) that it will subsequently
adhere to. So we simply focus on union wage-setting behavior in
Stage 2. All we need to retain from Stage 3 are the two constraints
on union wage-setting. The first is
(12)
pi ϭ wi,
which is implied by Bertrand pricing, using wi ϵ Wi/P. The second
provides the union with a trade-off between total hours employed
and the real wage, and is found by substituting (12) into (2), and
using the assumption of constant unit labor productivity, Qi ϭ Ei:
(13)
Ei ϭ m/N · wϪi ␩
.
Equations (12) and (13) hold as equilibrium results for any
possible value of 5W1, . . . ,WN 6, i.e., for each possible subgame
starting at Stage 3. J ust as the marketing agent in the yeoman
farmer model sets production quotas for individual farmers, we
assume here that the union sets the number of hours worked by
its members. This assumption ensures that the union’s supply of
15. Hence, if effective collusion occurs between several unions, this would
alter the number of sectors as defined here.

276
QUARTERLY J OURNAL OF ECONOMICS
labor always satisfies firms’ demand for labor, and hence that
there is an equilibrium for any value of 5W1, . . . ,WN 6.16
As in the yeoman farmer model we assume that each union
member has a utility function in which consumption and real
money balances affect utility positively and directly and which is
negative in hours worked. We show in the Appendix how the
demand for labor schedules in each sector can be derived from
utility maximization by individual workers. In addition, if the ith
union maximizes the utility of the representative union member,
this implies that the union should choose wi to maximize the
indirect utility function:
(14)
Ui ϭ wi · EiϪ(d/␤) · E␤i ϩ m/N,
where d ϵ (dЈ/n␤Ϫ1) and (dЈ/␤) · (E/n)␤ is the disutility of work and
nϪ1 · (wi · Ei ϩ m/N ) the wealth of the individual union member
(see the Appendix). In Stage 2 unions choose their sectoral money
*
*
*
wage simultaneously such that Ui(W i,W Ϫi) Ն Ui(Wi,W Ϫi) for all i.
This is equivalent to choosing wi to maximize (14) subject to (12)
and (13).
We focus on the symmetric equilibrium pi ϭ wi ϭ 1, for all i.
Maximizing (14) subject to (13) then implies that the equilibrium
number of hours of employment per sector is given by17
1/(␤Ϫ1)
␩ Ϫ 1 Ϫ 2Ѩ ln m/Ѩ ln wi
(15)
E* ϭ
.
1
2
d␩ Ϫ d · Ѩ ln m/Ѩ ln wi
Since Pi ϭ Wi, since the aggregate price index is defined in the
same way as in Section II above, and since pi ϭ 1 in equilibrium,
we can use the same reasoning as in that section to establish
Ѩ ln m
Ѩ ln wi
1
N
1Ϫ␣
NϪ1
(16)
ϭ Ϫ(1 Ϫ ␣) ·
·
ϭ Ϫ
N
NϪ1
so that the equilibrium employment rate is given by
␩ Ϫ 1 ϩ 2 · (1 Ϫ ␣)/(N Ϫ 1) 1/(␤Ϫ1)
(17)
E* ϭ
1
2
d␩ ϩ d · (1 Ϫ ␣)/(N Ϫ 1)
with equilibrium hours per worker of E*/n. Thus, exactly as output
16. One referee generously wrote out a formal proof of the existence of
equilibria for all possible vectors of Wi, and pointed out that this is not the case in a
standard model with market-determined wages. We will pass the proof on to
anyone interested.
17. Recall that the number of workers per sector, n, is fixed, so the equilib-
rium number of hours worked per worker is simply E*/n.

THE NONNEUTRALITY OF MONETARY POLICY
277
in the yeoman farmer model, if N is finite and greater than unity,
the equilibrium rate of employment is higher when ␣ Ͻ 1 (less than
fully accommodating monetary policy) than when ␣ ϭ 1 (fully
accommodating monetary policy); given N, the difference increases
as ␣ falls to zero (fully nonaccommodating policy); and given ␣, the
difference is greater the smaller is N. Analogous to the yeoman
farmer model, the reason is that unions have an incentive to take
into account the effect of their own wages on aggregate prices and
hence on the real money supply. The smaller is ␣ and N (for
N Ͼ 1), the greater is this incentive.
If N is very large—so that (1 Ϫ ␣)/(N Ϫ 1) approximates
0—the equilibrium employment rate is independent of the central
bank’s monetary rule. If N ϭ 1—when there is either a single
encompassing union, or when a small number of unions are
capable of acting as one—(17) is undefined, and our model does
not cover this situation. However, we can sketch an Olson-type
argument for expecting an encompassing union to act so as
produce full employment for 0 Յ ␣ Ͻ 1, (although a different
argument is needed in the limiting case of a fully accommodating
monetary policy). The encompassing union controls W hence P,
and therefore can choose m (which is feasible if ␣ Ͻ 1). The union
will choose m to maximize utility function (14) subject to the
equilibrium values w ϭ 1 and E ϭ m; this implies an equilibrium
employment rate of E* ϭ (2/d )1/(␤Ϫ1). It is can be easily checked
that this is greater than E* for any other value of N. This
argument does not apply in the special case of ␣ ϭ 1, since the CB
can set real money supply at any level, as long as it is less than or
equal to (2/d )1/(␤Ϫ1). But apart from this limiting case, when N ϭ 1,
it is reasonable to expect full employment in the sense that no
worker could be made better off by working more. For all
intermediate cases, however, when there is a small number of
independently acting unions, the monetary rule of the central
bank matters for equilibrium employment.
The results are illustrated in Figure I which plots the
equilibrium rate of employment as a function of N for different
values of ␣.18 Note that as the number of sectors decreases,
employment rises (except in the limiting case of ␣ ϭ 1). For a
finite number of sectors greater than one, the more nonaccommo-
dating the central bank, the higher the equilibrium rate of
employment; and this effect increases as N decreases. When N ϭ
18. N is treated as a continuous variable by assuming a continuum of sectors.

278
QUARTERLY J OURNAL OF ECONOMICS
FIGURE I
The Effects of N and ␣ on Equilibrium Employment
1, a single encompassing union can, and will, choose full employ-
ment. The special case of N ϭ ␣ ϭ 1 is not covered by the model.
Although the central bank can choose full employment in this
case, we have not defined a CB objective function that ensures it
will.
We can also interpret our main result in Layard-Nickell
terms, in which equilibrium employment is determined by the
intersection of the real wage imposed by business pricing behavior
(the ‘‘Feasible Real Wage’’) and the union-determined ‘‘Target
Real Wage’’ (see Layard, Nickell, and J ackman [1991]; and
Blanchard and Fischer [1989]). This is done in Figure II, where
the union-determined real wage schedule is obtained from equa-
tions (13) and (17). Because unions have an effect on aggregate
prices, they can influence the real money supply and hence
employment. The less accommodating the central bank (the
smaller ␣), and the smaller the number of unions (the smaller N ),
the greater the effect of each union’s wage on the real money
supply, and the greater the incentive to exercise restraint. This
shifts the union-determined real wage schedule downward and
raises the equilibrium level of employment.

THE NONNEUTRALITY OF MONETARY POLICY
279
FIGURE II
Layard-Nickell Interpretation of the Main Result
IV. CONCLUSIONS
The main result of this article can be summarized as follows:
under the assumptions of rational expectations, complete informa-
tion, credible precommitment, and a finite number of price or wage
setters, the accommodating or nonaccommodating nature of mone-
tary rules affects the equilibrium level of employment. The conven-
tional result that the monetary rule is unrelated to the equilib-
rium rate of employment emerges as a special case in our model,
when N = ϱ. In that case private agents are not engaged in a
strategic interaction with the CB: very small unions or marketing
agents cannot affect the general price level. We also sketched out a
complementary model of the case of an encompassing union
(N ϭ 1) and showed how monetary policy was also neutral in this
case—apart from a set of measure zero. Here the lack of strategic
interaction reflected the ability of the encompassing union to
determine the price level. But whenever a limited number of
unions or agents set wages or prices independently, the monetary
rule of the CB affects the equilibrium employment (or output) rate
since it determines the extent to which unions (or marketing
agents) can affect the real money supply, and hence the level of
demand and employment.
Although our model is deliberately based on a stringent set of

280
QUARTERLY J OURNAL OF ECONOMICS
assumptions—rational expectations, complete information, cred-
ible precommitment, and absence of nominal rigidities—these
ordinarily militate against finding employment effects of mone-
tary policies. Ipso facto, we expect that if these assumptions are
relaxed our basic result will remain. From a policy perspective
only the union-determined wage assumption cannot be dropped,
but this is a permissive one that enables us to theorize about a
broad range of economies characterized by relatively centralized
wage-setting systems. Thus, we expect our model to have some-
thing to say about the interaction between monetary policies and
wage setting in a broad range of countries including Australia,
Austria, Belgium, Denmark, Finland, Ireland, Italy, J apan, Ger-
many, the Netherlands, Norway, Sweden, Switzerland, and sev-
eral newly industrializing countries. It can also be used to
understand the effects on the equilibrium rate of employment rate
of changing the size of a currency area, as in the case of the
European Monetary Union (see Soskice and Iversen [1998]).
Considering the simplicity of the model, one obviously has to
be cautious in drawing strong policy implications. However, the
results, which are at first blush paradoxical, strongly suggest that
in countries with a small number of independently acting unions,
a nonaccommodating central bank can have substantial benefits
for long-term employment. In effect, such a bank alleviates the
coordination problems between unions that are large enough to
affect the welfare of all, yet too numerous to reach effective
collusion. This is an important lesson for both countries where
union fragmentation has undermined peak-level coordination
(such as Sweden), and for transition economies where strong
unions have reemerged in a setting of ‘‘weak’’ central banks. While
there may be offsetting political or economic benefits from having
accommodating central banks (especially in the short term), for
the category of countries where our argument is most salient, the
benefits we have described are likely to outweigh the costs.
APPENDIX: THE MICROECONOMIC BASES OF THE MODELS
In this appendix we show, first, how in the yeoman farmer
model the demand function in sector i against which the ith
‘‘marketing agent’’ maximizes, and the indirect utility function
which the agent maximizes, are derived from the individual
maximization behavior of producers. Second, we show how the
analogous propositions hold in the monopoly union model.

THE NONNEUTRALITY OF MONETARY POLICY
281
The Yeoman Farmer Model
In each sector i there are n farmers producing good i under
constant returns to scale. The marketing agent sets the price Pi
and hence total supply of good i, Qi, and each farmer is then
required to produce Qi/n.
The utility of the individual farmer, s, producing good i is
g
1Ϫg
␤
Cis
Mis/P
1 Ϫ g
dЈ Qi
(A1)
Uis
ϭ
·
Ϫ
,
1 2 1
2
1 2
g
␤
n
where Cis is aggregate consumption by s, given by
N
␩/(␩Ϫ1)
(␩Ϫ1)/␩
(A2)
Cis ϭ N1/(1Ϫ␩)
·
C
jis
͚
1
2
j
so the individual farmer ’s utility depends positively on consump-
tion and real money balances and negatively on output.
s chooses the N consumption levels Cjis and demand for
nominal money balances Mis to maximize (A1) subject to the
budget constraint:
N
Qi
(A3)
PjCjis ϩ Mis ϭ Pi
ϩ Mis ϵ Iis,
͚
1 2
n
j
where Mis is initial cash balances. This yields
Ϫ␩
Pj
g
Iis
P
(A4i)
and
Cjis
ϭ
·
·
1 2
P
N
Cis Mis/P
Iis
P
(A4ii)
ϭ
ϭ
.
g
1 Ϫ g
We can now derive the demand function for the jth good:
g
Iis
P
(A5)
Qj ϭ pjϪ␩
·
·
,
͚ ͚
N
i
s
where Qj is demand for the jth good and pj is its relative price.
We need to show finally that the double sum term in (A5) is
proportional to initial real cash balances. Since Cis/g ϭ Iis/P,
Iis
(A6)
g ·
ϭ
Cis ϭ
pjCjis ϵ Y,
͚ ͚ ͚ ͚
͚ ͚ ͚
P
i
s
i
s
j
i
s

282
QUARTERLY J OURNAL OF ECONOMICS
where Y is aggregate demand. In equilibrium, aggregate demand
is equal to output which is equal to ⌺i pi · Qi. Hence, from (A3), Y ϩ
M/P ϭ ⌺i ⌺s Iis, where M ϭ ⌺i Mi · n ϭ ⌺i ⌺s Mis so from (A6)
g
M
P
(A7)
Qj ϭ pϪj ␩
·
·
.
N · (1 Ϫ g)
The marketing agent for the jth good is constrained by (A7); this is
the same as equation (2) in the main text with M/P ϵ m and
g/(1 Ϫ g) normalized to unity.
Finally, substituting the optimal values of Cis and Mis from
(A4ii) and (A3) into (A1), multiplying (A1) by n and defining d ϭ
dЈ/n␤Ϫ1 yields the indirect utility function, equation (4) in the text,
which the marketing agent for the ith good maximizes.
The Monopoly Union Model
The argument in the monopoly union model is similar to the
yeoman farmer model, with the following differences. The utility
of the individual member, s, of union i is
g
1Ϫg
␤
Cis
Mis/P
1 Ϫ g
dЈ Ei
(A8)
Uis
ϭ
·
Ϫ
,
1 2 1
2
1 2
g
␤ n
so the individual member ’s utility depends positively on consump-
tion and real money balances and negatively on number of hours
worked Ei/n (Ei is the total number of hours worked in sector i and
n is the number of union members; all workers are union
members, and there is no mobility between sectors). Utility is
maximized subject to the budget constraint:
N
Ei
(A9)
PjCjis ϩ Mis ϭ Wi
ϩ Mis ϵ Iis.
͚
1 2
n
j
The demand function in the jth sector is derived identically, noting
that profits in the monopoly union model are zero (as a result of
Bertrand pricing and constant marginal costs) so that Y ϭ ⌺iN
Wi /P · Ei.
Finally, substituting the optimal values of Cis and Mis from
(A4ii) and (A9) into (A8), multiplying (A8) by n and defining d ϭ
dЈ/n␤Ϫ1, yields the indirect utility function in the text, equation
(14), which the ith union maximizes.
WISSENSCHAFTSZENTRUM BERLIN FU¨ R SOZIALFORSCHUNG, GERMANY
DEPARTMENT OF GOVERNMENT, HARVARD UNIVERSITY

THE NONNEUTRALITY OF MONETARY POLICY
283
REFERENCES
Alesina, Alberto, and Lawrence Summers, ‘‘Bank Independence and Macroeco-
nomic Performance: Some Comparative Evidence,’’ J ournal of Money, Credit
and Banking, XXV(1993), 151–162.
Barro, Robert, and David Gordon, ‘‘A Positive Theory of Monetary Policy in a
Natural Rate Model,’’ J ournal of Political Economy, XCI (1983), 589–610.
Blanchard, Olivier J ., and Stanley Fischer, Lectures in Macroeconomics (Cam-
bridge, MA: MIT Press, 1989).
Blanchard, Olivier J ., and Nobuhiro Kiyotaki, ‘‘Monopolistic Competition and the
Effects of Aggregate Demand,’’ American Economic Review, LXXVII (1987),
647–666.
Bleaney, Michael, ‘‘Central Bank Independence, Wage-Bargaining Structure, and
Macroeconomic Performance in OECD Countries,’’ Oxford Economic Papers,
XLVIII (1996), 20–38.
Calmfors, Lars, and J ohn Driffill, ‘‘Centralization of Wage Bargaining,’’ Economic
Policy, VI (1988), 14–61.
Cargil, Thomas F., ‘‘The Bank of J apan: A Dependent but Price Stabilizing Central
Bank,’’ Public Budgeting and Financial Management, V (1993), 131–139.
Cukierman, Alex, and Allan H. Meltzer, ‘‘A Theory of Ambiguity, Credibility and
Inflation under Discretion and Asymmetric Information,’’ Econometrica, LIV
(1986), 1099–1128.
Cukierman, Alex, and Francesco Lippi, ‘‘Central Bank Independence, Centraliza-
tion of Wage Bargaining, Inflation and Unemployment,’’ European Economic
Review, XLIII (1999), 1395–1434.
Cukierman Alex, Steven B. Webb, and Bilin Neyapti, ‘‘Measuring the Indepen-
dence of Central Banks and Its Effect on Policy Outcomes,’’ World Bank
Economic Review, VI (1992), 353–398.
Flanagan, Robert J ., David Soskice, and Lloyd Ulman, Unionism, Economic
Stabilization, and Incomes Policies: European Experience (Washington, DC:
The Brookings Institution, 1983).
Garrett, Geoffrey, and Christopher Way, ‘‘Unions, Governments and Central
Banks in Strategic Interaction,’’ Paper presented at the annual meeting of the
American Political Science Association, Chicago, 1995.
Hall, Peter A., ‘‘Central Bank Independence and Coordinated Wage Bargaining:
Their Interaction in Germany and Europe,’’ German Politics and Society, XII
(Autumn 1994), 1–23.
Hall, Peter A., and Robert Franzese, ‘‘Mixed Signals: Central Bank Independence,
Coordinated Wage-Bargaining, and European Monetary Union,’’ Interna-
tional Organization, LII (1998), 505–535.
Hutchison, Michael M., and J ohn P. J udd, ‘‘What Makes a Central Bank Credible?’’
Weekly Letter, Federal Reserve Bank of San Francisco (1989).
Hutchison, Michael M., Takatoshi Ito, and Tom Cargil, Political Economy of
J apanese Monetary Policy (Cambridge: MIT Press, 1997).
Iversen, Torben, ‘‘Wage Bargaining, Central Bank Independence and the Real
Effects of Money,’’ International Organization, LII (1998), 469–504.
, Contested Economic Institutions: The Politics of Macroeconomics and Wage
Bargaining in Advanced Democracies (Cambridge: Cambridge University
Press, 1999).
Kydland, Finn E., and Edward C. Prescott, ‘‘Rules Rather than Discretion: The
Inconsistency of Optimal Plans,’’ J ournal of Political Economy, LXXXV (1977),
473–486.
Layard, Richard, Stephen Nickell and Richard J ackman, Unemployment. Macro-
economic Performance and the Labour Market (Oxford: Oxford University
Press, 1991).
Lohmann, Susanne, ‘‘Optimal of Commitment in Monetary Policy: Credibility
versus Flexibility,’’ American Economic Review, LXXXII (1992), 273–286.
OECD, The Fiscal Position and Business Cycles Data Base (Paris: OECD, 1996).
, ‘‘Economic Performance and the Structure of Collective Bargaining,’’ in
Employment Outlook (Paris: OECD, 1997).
Olson, Mancur, The Rise and Decline of Nations (New Haven, CT: Yale University
Press, 1982).

284
QUARTERLY J OURNAL OF ECONOMICS
Persson, Torsten, and Guido Tabellini, ‘‘Designing Institutions for Monetary
Stability,’’ Carnegie-Rochester Conference Series on Public Policy, XXXIX
(1993), 53–84.
Rogoff, Kenneth, ‘‘The Optimal Degree of Commitment to an Intermediate
Monetary Target,’’ Quarterly J ournal of Economics, C (1985), 1169–1189.
, ‘‘Equilibrium Political Budget Cycles,’’ American Economic Review, LXXX
(1990), 21–36.
Rogoff, Kenneth, and Anne Sibert, ‘‘Equilibrium Political Business Cycles,’’ Review
of Economic Studies, LV (1988), 1–16.
Soskice, David, ‘‘Wage Determination: The Changing Role of Institutions in
Advanced Industrialized Countries,’’ Oxford Review of Economic Policy, IV
(1990), 36–61.
Soskice, David, and Torben Iversen, ‘‘Multiple Wage Bargaining Systems in the
Single European Currency Area,’’ Oxford Review of Economic Policy, XIV
(1998), 110–124.
Svensson, Lars E. O., ‘‘Optimum Inflation Targets, ‘Conservative’ Central Banks,
and Linear Inflation Contracts,’’ NBER Working Paper No. 5251, 1996.