The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
w o r k i n g
p
a
p
e
r
02 08
Search, Money, and Capital:
A Neoclassical Dichotomy
by S. Borağan Aruoba and Randall Wright
FEDERAL RESERVE BANK OF CLEVELAND
Working papers of the Federal Reserve Bank of Cleveland are preliminary materials
circulated to stimulate discussion and critical comment on research in progress. They may not
have been subject to the formal editorial review accorded official Federal Reserve Bank of
Cleveland publications. The views stated herein are those of the authors and are not necessarily
those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal
Reserve System.
Working papers are now available electronically through the Cleveland Fed’s site on the World
Wide Web: www.clev.frb.org.
September 2002
Working Paper 02-08
Search, Money and Capital:
A Neoclassical Dichotomy
by S. Borağan Aruoba and Randall Wright
Recent work has reduced the gap between search-based monetary theory and mainstream
macroeconomics by incorporating into the search model some centralized markets as well as some
decentralized markets where money is essential. This paper takes a further step towards this integration
by introducing labor, capital and neoclassical firms. The resulting framework nests the search-theoretic
monetary model and a standard neoclassical growth model as special cases. Perhaps surprisingly, it also
exhibits a dichotomy: one can determine the equilibrium path for the value of money independently of the
paths of consumption, investment and employment in the centralized market.
S. Boragan Aruoba is at the University of Pennsylvania. Randall Wright is at the University of
Pennsylvania and is an institute scholar at the Central Bank Institute of the Federal Reserve Bank of
Cleveland. The authors thank K. Burdett, G. Eudey, R. Lagos, M. Molico and
C. Waller for their input, as well as the NSF and the Federal Reserve Bank of Cleveland for research
support. Address correspondence to Randall Wright, Department of Economics, University of
Pennsylvania, 3718 Locust Walk, Philadelphia, PA 19104.
1
Introduction
There seems to be a big distance between standard macroeconomics and the branch of
monetary theory with explicit microfoundations based on search, or matching, theory. As
Azariadis (1993) put it, \Capturing the transactions motive for holding money balances in
a compact and logically appealing manner has turned out to be an enormously complicated
task. Logically coherent models such as those proposed by Diamond (1982) and Kiyotaki
and Wright (1989) tend to be so removed from neoclassical growth theory as to seriously
hinder the job of integrating rigorous monetary theory with the rest of macroeconomics." As
Kiyotaki and Moore (2001) more recently put it, \The matching models are without doubt
ingenious and beautiful. But it is quite hard to integrate them with the rest of macroeconomic
theory { not least because they jettison the basic tool of our trade, competitive markets."
Recent work has gone some distance towards closing the gap between the search-based
approach and mainstream macroeconomics. An example is the model in Lagos and Wright
(2002a), hereafter referred to as LW. The innovation in LW is to bring competitive markets
back on board in a way that maintains an essential role for money and at the same time
greatly increases the tractability of the search framework. In the LW environment there is
decentralized trade in anonymous markets with bilateral random matching, as in a typical
search model, but after each round of decentralized trade a centralized market convenes.
In the centralized market agents not only produce and exchange goods for consumption
purposes, they also trade to adjust their money balances, which may have changed from the
desired level during the previous round of decentralized trade. Under the assumption that
utility is quasi-linear in one of the goods traded in the centralized market, it turns out that
all agents adjust to the same money balances. Hence, at the start of every period there will
be a degenerate distribution of money holdings. 1
1
See Lagos and Wright (2002b), Rocheteau and Wright (2002), Berentsen, Lagos and Rocheteau (2002)
2
This resolves a complicated technical problem { solving for and keeping track of the money
distribution { which often forced people to make undesirably strong assumptions in earlier
search-based models, like severe restrictions on how much money agents can hold (typically it
was restricted to 0 or 1 unit). The LW framework allows one to address many issues for which
models with these severe restrictions are ill-suited, and yet it is very simple. The simplicity
comes at a cost since, after all, having an endogenous non-degenerate distribution of money
holdings may be interesting and relevant for some questions.2 Presumably, however, there are
some interesting questions in monetary economics for which an endogenous non-degenerate
distribution is not critical. For such questions, the LW framework provides a tractable model
with explicit microfoundations, and no restrictions on money holdings, which means that it
can be more easily used to discuss monetary policy and other issues that were di±cult in
earlier search models.
This is the sense in which we mean recent work has gone some distance towards integrating search-based models theory and mainstream macroeconomics. The point of the current
paper is to show that with a little e®ort one can take a much bigger step towards this integration. Existing versions of the LW framework still do not look much like the neoclassical
growth model. Indeed, not much happens in the centralized market in these models, and
it is there mainly to render the distribution of money holdings in the decentralized market
degenerate. Yet once this centralized market is up and running, one can do a lot more. Here
we introduce labor, capital, and a neoclassical production function. The result integrates a
and Berentsen, Rocheteau and Waller (2002) for extensions and applications of the basic framework. A
related but also quite di®erent approach, dating back to Shi (1997), uses the assumption of large families
rather than competitive markets to render the money distribution degenerate. In Shi (1999) and also Faig
(2001), these families produce specialized goo ds that they can either trade or keep within the household to
be used as capital. Here we will also introduce capital, but as a general good that is traded on a centralized
market, much more in the spirit of standard macroeconomics.
2
See Molico (1999) for an example where a non-degenerate distribution is interesting; see Wallace (2002)
for a general discussion. The standard references for models that assume m 2 f0; 1g, so as to avoid dealing
with this distribution, include Kiyotaki and Wright (1993), Shi (1995) and Trejos and Wright (1995).
3
standard growth model and the search-theoretic monetary model; indeed, these two models
emerge as special cases.
Perhaps surprisingly, when we specify the model in what we think of as a very natural
way, an interesting dichotomy emerges: it is possible to partition the equilibrium conditions
in such a way that one can solve independently for the allocation in the centralized and
decentralized markets. The nominal price level ties these markets together, since money is
traded in both markets, but it turns out that although the price level a®ects the allocation
in the decentralized market in an important way, in the centralized market it does not a®ect
aggregate activity or welfare. Many policy implications follow from this result. For example,
a change in the rate of monetary expansion can a®ect the in°ation rate and hence the price
level, and this a®ects consumption in the decentralized market, but is completely neutral in
terms of the aggregate labor market or capital accumulation. 3
These policy implications ought to be interpreted cautiously. First, the fact that in°ation
has no impact on the aggregate labor market or capital accumulation does not mean that
in°ation does not matter, since it does a®ect economic activity in the decentralized market
and hence welfare. Second, the dichotomy and its implied policy implications of course
depend crucially on certain assumptions. So, while our model does integrate neoclassical
growth theory and monetary models with explicit microfoundations in a simple and natural
way, we think of it mainly as a benchmark from which policy discussions can proceed. Thus,
it may or may not be that monetary policy has real e®ects on centralized markets in actual
3 Our
result is di®erent from the classical dichotomy. As Sargent (1979) put it, \A macroeconomic model
is said to dichotomize if a subset of equations can determine the values of all real variables with the level
of the money supply playing no role in determining the equilibrium value of any real variable. Given the
equilibrium values of the real variables, the level of the money supply helps determine the equilibrium
values of all nominal variables that are endogenous but cannot in°uence any real variable. In a system that
dichotomizes the equilibrium values of all real variables are independent of the absolute price level." This
is not quite the case here, since the amount of output that one gets for a dollar in decentralized trading (a
real variable) does depend on the absolute price level and hence on monetary policy { but the real variables
from the centralized market (employment, consumption and investment) are independent of the price level
and monetary policy.
4
economies, but if so, a reasonable model of this will have to do something di®erent from
what we do here. 4
The rest of the paper is organized as follows. Section 2 presents the basic structure by
reviewing the LW model. Section 3 shows how to introduce capital accumulation. Section 4
adds labor as well as capital. Section 5 endogenizes search intensity, or shopping time. All
of these models display the strong dichotomy referred to above. Section 6 concludes.
2
The Basic Model
Time is discrete and continues forever. There is a [0; 1] continuum of in¯nitely-lived agents.
There are two types of commodities: a general good, and a set of special goods. All goods
are nonstorable and perfectly divisible. All agents consume the general good, but each agent
derives utility from only some subset of special goods. All agents can produce the general
good, but each has a technology that allows him to produce only one special good. No
agent consumes the special good he produces. For a random pair of agents, we assume the
following: with probability ¢ both like the special good the other can produce (called a
double coincidence ); with probability ¾ one likes the other's good but not vice-versa (called
a single coincidence); and with probability 1 ¡ ¢ ¡ 2¾ neither likes the other's good, where
¢ ¸ 0, ¾ > 0 and ¢ + 2¾ · 1.
In addition to consumption goods, there is another object called money that cannot be
consumed or produced by any private agent. Money is perfectly divisible and storable, and
agents can carry any non-negative quantity of money. Let F t(m) denote the CDF of money
R
holdings across agents, where mt dFt (mt ) = Mt is the total amount of money, at date t.
The money supply changes over time according to Mt+1 = (1 + ¿ t)Mt , where the growth
rate ¿ t need not be constant. New money is injected in the form of lump-sum transfers, or
4
By analogy, in the basic growth model, the welfare theorems hold, which obviously also entails some
stark conclusions about policy, yet this model serves us well as a benchmark for policy discussions.
5
taxes if ¿ t < 0. To be precise, we assume each period is divided in two subperiods { say, day
and night { and money transfers occur at the end of the second subperiod. Agents discount
between periods at rate ¯, but not between day and night within a period (this is without
loss in generality).
During the day (i.e. in the ¯rst subperiod), agents participate in a decentralized market
with bilateral random matching. The probability of meeting anyone is ® and each meeting
is a random draw from the population. These meetings are anonymous, which prevents
agents from trading any promises to be ful¯lled in the future or even later that same period
(Kocherlakota [1998]; Wallace [2001, 2002]). Also, during the day agents can produce special
goods but not general goods. By contrast, during the night agents can produce general but
not special goods, and they participate in a centralized market. Given this environment,
the feasible trades are as follows: special goods can be traded for other special goods or for
money during the day; and general goods can be traded for money at night.
In any single coincidence meeting in the decentralized market, we call the agent that
likes that other's good the buyer, and the other agent the seller. In such a meeting let
qt (m; m)
~ be the amount of goods and dt (m; m)
~ the amount of money they exchange, where
m is the money holdings of the buyer and m
~ is the money holdings of the seller. Also, let
Bt (m; m)
~ be the payo® from a trade in a double coincidence meeting when the agents hold
m and m.
~ These variables will be determined by bargaining. By contrast, in the centralized
market that convenes at night agents behave competitively { i.e., they trade general goods
and money taking prices parametrically. We normalize the price of a general good in the
night market to 1 and let Át be the amount of general goods that a dollar will buy; thus
pgt = 1=Át is the nominal price of general goods at t.
The utility of consuming q units of a special good that one likes is u(q), and the cost of
producing q units of a special good is c(q). Assume u and c are Cn (n times continuously
6
di®erentiable) with n ¸ 3, where u0 > 0, c0 > 0, u00 < 0 and c00 ¸ 0. Also, u(0) = c(0) = 0
and u(¹
q ) = c(¹
q ) for some q¹ > 0. For certain results we need an assumption on u000 which
is conveniently stated by saying that marginal utility is log-concave (i.e., the log of u0 is
concave). Let q ¤ denote the e±cient quantity, which solves u0 (q ¤) = c0 (q¤); q¤ is the amount
agents would agree ex ante that they should trade in each decentralized meeting if they could
commit to such an arrangement { but of course, they cannot so commit, since if they could
money would be inessential. For general goods, U and C are the utility of consumption and
cost of production. Assume U and C are Cn with n ¸ 2, where U 0 > 0, C 0 > 0, U 00 · 0,
C 00 ¸ 0, and U 0 (x¤ ) = C 0 (x¤) for some x¤ > 0 with U(x¤ ) > C(x¤). We need either U or C
to be linear; for now we take C(y) = y.5
Let W (s) be the value function of an agent entering the night market and V (s) the value
function of an agent entering the day market with individual state variable s. For now, one's
state is simply one's money holdings, s = m, but we introduce this notation since s will
include other objects in the models analyzed below. The aggregate state is the distribution
F , which will remain implicit in the functional notation. Bellman's equation is
Z
Vt(m) = ®¾ fu [qt (m; m)]
~ + Wt [m ¡ dt (m; m)]g
~
dFt( m)
~
Z
+ ®¾ f¡c [qt ( m;
~ m)] + Wt [m + dt (m;
~ m)]g dFt (m)
~
Z
+ ®¢ Bt(m; m)dF
~
~ + (1 ¡ 2®¾ ¡ ®¢)Wt(m):
t (m)
(1)
The ¯rst term is the expected gain from buying in a single-coincidence meeting; the second is
the expected gain from selling in a single-coincidence meeting; the third is the expected gain
from a double-coincidence meeting; and the last term is the expected value of not trading in
the day market and going to the centralized market with m. We are not restricting anything
5 Below
we show the case where U is linear and C strictly convex is basically identical. Once we introduce
a neoclassical production function we assume utiltiy is linear in lesiure. The reason we need linearity
somewhere in preferences over general goods is to eliminate wealth e®ects, because then all agents will take
the same amount of money out of the centralized market, regardless of their histories. If this were not the
case, the model would still be well-speci¯ed, but it would be much less tractable.
7
to be stationary here, although we sometimes drop the subscript t when there is no risk of
confusion.
The problem of an agent in the centralized market is
W (m) = max0 U(x) ¡ y + ¯V (m0 + ¿ M)
x;y;m
s:t: x = Á(m ¡ m0 ) + y:
(2)
(3)
Thus, he chooses general good consumption and production, xt and yt , and takes mt+1 =
m0t + ¿ tMt dollars into the next day, where m0t is money left over after trading and ¿ t Mt is
the lump sum transfer. We impose x ¸ 0 and m0 ¸ 0, but we do not impose y ¸ 0. For
technical reasons it is easier to allow y < 0 when solving this problem, and then after ¯nding
an equilibrium, one can impose conditions to guarantee y > 0; this is what we do here (see
LW for a discussion).
The following result describes several useful features of the solution, including the linearity of W (m).
Lemma 1 In the centralized market, for all agents and for all t, xt = x¤, m0t is independent
of mt , and Wm = Át .
Proof: Substituting y from (3) into (2), we have
W (m) = Ám + max0 fU (x) ¡ x ¡ Ám0 + ¯V (m0 + ¿ M)g ;
x;m
which implies that W is linear in m with slope Á, and that the choices of x and m0 are
independent of m. Di®erentiating, we get the ¯rst order conditions
1 = U 0 (x)
(4)
Á = ¯Vm(m0 + ¿ M);
(5)
the ¯rst of which implies x = x¤ . ¥
8
The next step is to discuss the terms of trade in the decentralized market. In doublecoincidence meetings we adopt the symmetric Nash solution where the threat point of an
agent with m dollars is given by W (m). It is straightforward to show that, for an agent
with m dollars, this implies B(m; m)
~ = u(q¤ ) ¡ c(q¤ ) + W (m); i.e., regardless of (m; m),
~
in double-coincidence meetings agents produce the e±cient quantity q = q ¤ for each other
and no money changes hands. In single-coincidence meetings we use the generalized Nash
solution, with µ denoting the bargaining power of the buyer, and again the threat point of
an agent with m dollars given by W (m). The solution is characterized in the next lemma,
where we write q = q (s; s~) and d = d (s; s~) since then we can use the same notation in
models where s contains more than just m.
Lemma 2 In single coincidence meetings in the decentralized market, for all t, the bargaining solution is
qt (s; s~) =
½
q¤
if m ¸ m¤t
q~t (m) if m < m¤t
and d (s; ~s) =
½
m¤t if m ¸ m¤t
m if m < m¤t
(6)
where q~t (m) solves g(q) = Á tM , with
g(q) ´
µc(q)u0 (q) + (1 ¡ µ)u(q)c0 (q)
;
µu0 (q) + (1 ¡ µ)c0 (q)
(7)
and m¤t = g(q ¤)=Át .
Proof: This is a special case of the bargaining solution in Lemma 5 below. ¥
Figure 1 shows the solution. An important observation is that, since the function g(q)
depends only on exogenous objects, qt(m) is a ¯xed function of the buyer's real balances,
zt = Á tm. As long as z t ¸ z ¤, in real terms the buyer spends z ¤ and gets q¤, where
z ¤ = Á tm¤t = g(q¤) = µc(q ¤) + (1 ¡ µ)u(q ¤) is constant. If zt < z ¤ the buyer spends all his
cash and gets q < q ¤. Since q and d depend on s and s~ only through the buyer's money
9
Figure 1: The Single-Coincidence Bargaining Solution
holdings, in what follows we write q (s; ~s) = q(m) and d (s; ~s) = d(m). Note that for m < m¤ ,
q~ is di®erentiable and q~0 (m) = Á=g 0 (q), where
g0 =
u0 c0 [µu0 + (1 ¡ µ)c0 ] + µ(1 ¡ µ)(u ¡ c)(u0 c00 ¡ c0u00 )
> 0:
[µu0 + (1 ¡ µ)c0 ] 2
(8)
Using Lemmas 1 and 2 we can simplify (1) to
V (s) = ®¾fu [q (m)] ¡ Ád (m)g + ®V0 + W (s);
(9)
where again we write s for m so that we can use the same notation below, and
V0 = ¾
Z
f¡c [q (m)]
~ + Ád (m)g
~ dF (m)
~ + ¢[u(q ¤) ¡ c(q ¤)]
(10)
does not depend on m. Given u and c are Cn, V is Cn¡1 for all m 6= m¤. For m < m¤,
Vm = ®¾u0 q~0 ¡ ®¾Á + Á > 0;
(11)
and for m > m¤ , Vm = Á. A simple calculation shows that the limit of Vm as m ! m¤ from
below is strictly less than Á, and so V has a kink at m¤. For m < m¤ we have
Vmm = ®¾ q~02u00 + ®¾u0q~00 ;
10
(12)
which cannot be signed in general since it depends on q 00 which depends on u000. However, it
can be shown that Vmm < 0 for all m < m¤ under the assumption that either µ is close to 1
or u0 is log-concave (see LW).
Figure 2: The Centralized Market Problem
Given all this, V must be as shown in Figure 2, which illustrates the problem of deciding
how much cash to take out of the centralized market at t, maxf¡Át mt+1 + ¯Vt+1(mt+1)g. As
should be clear from the picture, if Át < ¯Át+1 this problem has no solution. 6 Hence, we can
assume Át ¸ ¯Á t+1 without loss of generality. Then it is clear that any solution is to the left
of m¤t+1, simply because of the kink. This result does not depend at all on concavity, but
under the assumptions stated above that guarantee Vmm < 0 we have the additional result
6 Formally,
suppose Át < ¯Át+1 for some t, and consider the following arbitrage argument. At t, you can
raise your general good production by dy > 0 and sell it for dy=Át dollars. Then at t + 1, you can use the
money to reduce general good production by dyÁt+1 =Át without changing anything else. The net utility gain
from this is dy(¡1 + ¯Át+1 =Át ) > 0; hence we cannot have Át < ¯ Át+1 . In this argument we did not worry
about the possibility of y t+1 = 0; however, it is not hard to show that xt > 0 for all t, and therefore y t+1 > 0
for at least one agent in any equilibrium, and this is all that is necessary for the desired result.
11
that there is a unique mt+1 < m¤t+1 solving the problem. We summarize as follows.
Lemma 3 In the centralized market, for all t, all agents choose the same mt+1 = m0t + ¿ tMt ,
and mt+1 < m¤t+1.
Proof: Obvious from the discussion in the text. ¥
This result is what makes the model simple. First, the result that all agents choose
the same mt+1 implies the distribution of money at the start of each day is degenerate at
m = M . Second, the result that mt+1 < m¤t+1 implies buyers spend all their money, so we
know d = M and q = q~(M) < q¤. Hence, at the close of each decentralized market, the
fraction ®¾ of agents who were buyers have 0 dollars, the fraction ®¾ who were sellers have
2M dollars, and the rest have M dollars. Since they all exit the centralized market holding
M and they all consume x = x¤ , the budget constraint implies individual supply of y is
8 ¤
< x + ÁM for buyers
x¤ ¡ ÁM for sellers
y=
(13)
: ¤
x
for others,
where when we write \for buyers" we mean \for agents who were buyers in the previous
subperiod" and so on. Aggregate supply is simply y = x¤.
At this stage we can consider the issue of nonnegativity. Recall that we have not imposed
y ¸ 0 so far. Given this we have shown that in equilibrium xt = x¤, mt+1 = M and y is
given by (??). We can guarantee y ¸ 0 for all agents if we can be sure x¤ ¸ ÁM = g(q),
where g is given in (7). Since q < q ¤ and g is monotonically increasing, we can guarantee
what we want if we impose
x¤ ¸ g(q¤ ) = µc(q ¤) + (1 ¡ µ)u(q ¤):
(14)
Hence, we have a simple condition to rule out y < 0 in equilibrium.7
7
This is true if we start by giving all agents the same endowment of money, m 0 , or at least as long as the
initial distribution F0 is not too disperse. If some agent starts with a very large initial m 0 and we impose
12
We now simplify things by reducing the model to one equation in one unknown. First,
insert Vm from (11) into (5), being careful to index all objects by the appropriate date, to
get the following expression:
£
¤
Át = ¯ ®¾u0 (qt+1)~
q0 (mt+1) + (1 ¡ ®¾)Á t+1 :
(15)
Now substitute q~0 (q) = Á=g 0 (q) and Á = g(q)=m from the bargaining solution, as well as
m = M, to arrive at
·
¸
g(qt )
g(qt+1)
u0(qt+1 )
=¯
®¾ 0
+ 1 ¡ ®¾ :
Mt
Mt+1
g (qt+1)
(16)
Given any exogenous path for M t, this is a di®erence equation in qt . An equilibrium can now
be de¯ned as a solution to (16) that stays in [0; q ¤] for all t; it is a monetary equilibrium if
qt > 0 for all t.
Given the q path, one can recover Á = g(q)=M and all of the other variables. The
aggregate values of the centralized market variables are easy, since y = x = x¤, but we can
also disaggregate into the amount y supplied by buyers, sellers, and others as described in
(13). We can compute the nominal price of a special good, ps = M=q, and the general good,
pg = 1=Á. It is immediate from (16) that the model displays classical neutrality: for any
¸ > 0, if we change the money supply sequence so that Mt becomes ¸Mt for all t, then all
nominal variables (Át , pst , pgt,:::) also change by a factor ¸ while all real variables (qt , yt , zt ,:::)
stay the same.8 The model does not, however, display superneutrality: generally, changing
y ¸ 0, he will set y0 = 0, x 0 > x¤ and m1 > M . Thus, his money holdings can stay above M for several
periods, but eventually any such agent will spent down his initial riches and then (??) guarantees y ¸ 0
for all agents. To avoid this nuisance we can simply asume F0 is not too disperse; to be precise, y0 ¸ 0 if
x ¤ ¸ g(q ¤ )m 0 =M , or
m0 ·
M X¤
µc(q ¤ ) + (1 ¡ µ)u(q¤ )
for all agents. This condition on F 0 togther with (??) is su±cient to make nonnegativity a nonissue.
8 More precisely, given a path for M 1 , suppose there is a set of equilibrium paths for q , and for each q
t
t
t
there are associated values for each of the other real and nominal variables. Then for any ¸ > 0, given the
path Mt2 = ¸Mt1 the set of equilibrium paths for qt is the same, and for each qt the associated values of each
the other real variables is the same while the associated values of each of the nominal variables changes by
the factor ¸.
13
the growth rate of M will a®ect at least some of the real variables, as we will discuss in detail
below.
A case that makes sense when ¿ t = ¿ is constant is a steady state monetary equilibrium,
which is a constant solution q > 0 to (16) with M t+1 = (1 + ¿)M t:
·
¸
¯
u0 (q)
1=
®¾ 0
+ 1 ¡ ®¾ :
1+¿
g (q)
(17)
In a such an equilibrium Á t = g(q)=Mt falls as Mt grows, but real balances zt = ÁtM t = g(q)
remain constant; i.e., the in°ation rate equals the rate of monetary expansion, ¿ . It is
straightforward to establish the existence of a monetary steady state, and either uniqueness
or multiplicity, depending on assumptions (see LW). In some special cases the analysis is
especially easy; e.g., if µ = 1 (take-it-or-leave-it o®ers by the buyer) then (7) implies g(q) =
c(q), and (17) is really quite simple.
In steady state, the general result ¯Át+1 · Át implies ¯Át+1Mt+1 · Át Mt (1 + ¿ ), and
hence ¯ · 1 + ¿ . This puts a constraint on policy: we cannot contract the money supply
faster than the so-called Friedman Rule, ¿ = ¯ ¡ 1; if we try, the monetary equilibrium
will break down. It is not hard to check that q is increasing in
¯
1+¿
¯
,
1+¿
and that q ! q¤ as
! 1 i® µ = 1. This implies that the Friedman Rule is the optimal policy, since it gets q
as close to q ¤ as possible before the equilibrium breaks down, but it cannot achieve the fully
e±cient outcome q = q ¤ unless µ = 1. This can have interesting implications for some issues,
including the welfare cost of in°ation (see LW for further discussion).
Although the equilibrium is ine±cient if we have either ¿ > ¯¡1 or µ < 1, this ine±ciency
manifests itself only in the decentralized market { in the centralized market agents always
consume the e±cient quantity x¤. Indeed, the model displays a very strong dichotomy: one
can solve independently for the allocations in the decentralized and centralized markets. That
is, at least the aggregate allocation in the centralized market, x = y = x¤, is independent of
the solution q to (16), and vice-versa. The value of money Á = g(q)=M does depend on q,
14
and this does a®ect how much y di®erent individuals supply as seen in (13), but this does
not a®ect aggregate supply. Hence, for example, an increase in ¿ will lower q but does has
no e®ect on x or y.
This completes our review of the basic LW model. We close the section by sketching
an alternative version where, instead of assuming that C(y) = y is linear, we assume that
C 00 (y) > 0 and that U(x) = x is linear. With this speci¯cation the centralized market
problem becomes
W (m) = max0 x ¡ C (y) + ¯V (m0 + ¿ M)
x;y;m
(18)
subject to (3). We do not impose x ¸ 0 here, for the same reason we did not impose y ¸ 0
earlier. Substituting for x from (3) and di®erentiating with respect to y and m0 , we get
1 = C 0 (y)
(19)
Á = ¯Vm(m0 + ¿ M):
(20)
Hence, y = y¤ where C 0 (y¤) = 1, and m0 satis¯es the same condition as before.
In the original model x = x¤ is constant across individuals and y varies according to
whether an agent was a buyer or seller in the previous subperiod, while here y = y¤ is
constant and x varies according to:
8 ¤
< y ¡ ÁM for buyers
x=
y¤ + ÁM for sellers
: ¤
y
for others.
(21)
We can guarantee x ¸ 0 with a condition like (??), except y¤ replaces x¤. Otherwise, things
are exactly the same. In any case, we summarize the key result for our purposes as follows:
Proposition 1 The basic model dichotomizes: one can solve for the equilibrium path of q
and the equilibrium path of aggregate (x; y) independently, and monetary policy a®ects the
former but not the latter.
15
3
Capital
Here we introduce capital and neoclassical production. As in the standard one-sector growth
model, capital is the same as the general consumption good. Later we introduce ¯rms
explicitly, but for now we let each agent have access to a technology for producing general
goods f(k), with the usual properties f (0) = 0, f 0 > 0 and f 00 < 0. A very important
assumption is that one's capital is not mobile: it can be traded in the centralized market,
but cannot be carried into the decentralized market. Nor can claims to capital be traded in
the decentralized market, since agents are anonymous and hence could renege on any such
claim without fear of retribution. These assumptions are made simply to guarantee that
capital or claims to capital do not replace money as a medium of exchange { that is, to
guarantee that money is still essential. 9
The individual state variable now includes one's money holdings and capital stock, s =
(m; k), with joint distribution F (s). In this environment, Bellman's equation is the natural
generalization of (1):
Z
fu [q (s; s~)] + W [m ¡ d (s; s~) ; k]g dF (~
s)
Z
+®¾ f¡c [q (~
s; s)] + W [m + d (~
s; s) ; k]g dF (~
s)
Z
+®¢ B(s; s~)dF (~
s ) + (1 ¡ 2®¾ ¡ ®¢)W (s):
V (s) = ®¾
(22)
Again, the aggregate state F and the date t are implicit in the notation, but we emphasize
again that we are not imposing stationarity. The centralized market problem is
W (s) =
max x + ¯V (m0 + ¿ M; k 0 )
x;m0;k0
s:t: x = Á(m ¡ m0 ) + f(k) + (1 ¡ ±)k ¡ k 0;
9 Obviously
(23)
(24)
modeling at a deeper level the restriction that capital cannot be traded in the decenteralized
market may be worthwhile, and presumably it would be interesting to have some capital or claims to
capital circulate along side of currency. One possible route is to assume some agents are anonymous in the
decentralized market while others are not, along the lines of Cavalcante and Wallace (1999), perhaps.
16
where ± is the depreciation rate. Notice we are using the version of the model in the previous
section with linear U(x) = x, and as was the case there we do not impose x ¸ 0, but we can
check that this is true once we ¯nd an equilibrium.
We have the following versions of Lemmas 1 and 2.
Lemma 4 In the centralized market with capital, for all agents and for all t, m0 and k 0 are
independent of s = (m; k), Wm = Á and Wk = f 0 + 1 ¡ ±.
Proof: Substituting from (24) into (23), we have
W (s) = Ám + f(k) + (1 ¡ ±)k + max
f¡Ám0 ¡ k 0 + ¯V (m0 + ¿ M; k 0)g ;
0 0
m ;k
and everything follows. In particular,
1 = ¯V k(m0 + ¿ M; k0 )
(25)
Á = ¯V m(m0 + ¿M; k 0 ):
(26)
are the ¯rst order conditions. ¥
Lemma 5 In the model with capital, for all t, the single-coincidence bargaining solution is
exactly the same as in Lemma 2.
Proof: The generalized Nash problem when the buyer has s = (m; k) and the seller
s~ = (m;
~ ~k) is
h
i1¡µ
max [u(q) + W(m ¡ d; k) ¡ W (m; k)] µ ¡c(q) + W (m
~ + d; ~k) ¡ W (m;
~ ~k)
q;d
subject to d · m. Lemma 4 implies W (m ¡ d; k) ¡ W (m; k) = ¡Ád and W (m
~ + d; ~k) ¡
W (m;
~ ~k) = Ád, and so this problem reduces to
max [u(q) ¡ Ád] µ [¡c(q) + Ád] 1¡µ
q;d
17
subject to d · m. Notice m,
~ k and ~k have vanished.
The Kuhn-Tucker conditions, which are necessary and su±cient here, are
µu0(q)
(1 ¡ µ)c0 (q)
¡
= 0
u(q) ¡ Ád ¡c(q) + Ád
¡µÁ
(1 ¡ µ)Á
+
¡¸ = 0
u(q) ¡ Ád ¡c(q) + Ád
¸(m ¡ d) = 0
where ¸ ¸ 0 is the multiplier. If ¸ = 0, the ¯rst two conditions yield u0 (q) = c0 (q), or q = q¤ ,
and then d = m¤. If ¸ > 0, the solution is d = m and q = q~(m), where q~(m) solves the ¯rst
Kuhn-Tucker condition, which can be rearranged into mÁ = g(q) with g de¯ned in (7). By
the implicit function theorem, q~0 > 0. Hence, for m < m¤ we have d = m and q = q(m)
~
and
for m ¸ m¤ we have d = m¤ and q = q¤ . ¥
~ we again write
Since the bargaining solution here depends on m but not m,
~ k or k,
q = q(m) and d = d(m). Hence, Bellman's equation again reduces to (9) in the previous
section and Vm and Vmm are again given by (11) and (12). The conditions that guarantee V
is strictly concave in m from the previous section (µ ¼ 1 or u0 log-concave) still apply, and
this yields the generalized version of Lemma 3.
Lemma 6 In the centralized market with capital, for all t, all agents choose the same m0t
and kt0 , and mt+1 = m0t + ¿ t Mt < m¤t+1.
Proof: Given Vmm < 0, V mk = 0 and Vkk = f 00 < 0 imply that V is strictly concave, and
hence there exists a unique solution to (25) and (26). The result mt+1 < m¤t+1 follows from
the same argument used in Lemma 3. ¥
As in the previous section, F is degenerate at (m; k) = (M; K), although here of course
the aggregate capital stock K is endogenous, and buyers spend all their money in every
18
single-coincidence meeting, d = M . In equilibrium we have the following version of (21):
8
< X ¡ ÁM for buyers
x=
X + ÁM for sellers
(27)
:X
for others
where X = f (K) + (1 ¡ ±)K ¡ K 0 . Thus, individual consumption in the centralized market
depends on whether one spent or acquired money in the previous subperiod, but aggregate
consumption is simply X.
In the previous section we reduced the model to one equation by substituting Vm into the
¯rst order condition for m0 and then inserting the bargaining solution. The same procedure
here yields exactly the same result, which we repeat for convenience:
·
¸
g(qt )
g(qt+1)
u0(qt+1 )
=¯
®¾ 0
+ 1 ¡ ®¾ :
Mt
Mt+1
g (qt+1)
(28)
Similarly, substituting Vk into (25) we get
1 = ¯[f 0(K t+1) + 1 ¡ ±]:
(29)
This is the familiar condition from the standard (nonmonetary) neoclassical growth model.10
Equilibrium can now be de¯ned as a path (q; K ) that solves (28) and (29) subject to the
¹ where as is standard K
¹ is the
usual side conditions, K0 is given, q 2 [0; q ¤], and k 2 [0; K]
maximum of K0 and the upper bound on the sustainable capital stock.
The main point is that when we introduce capital the model still dichotomizes: (28) and
(29) can be solved independently. The set of equilibrium q paths is the same as in the basic
LW model while the K path is the same as in a nonmonetary growth model. As in the model
from the previous section, q a®ects Á and hence individual consumption in the centralized
10 That is, the familiar condition when utility is linear, U (x) = x. Normally, U 0 (x ) appears on the left
t
and U 0(xt+1 ) on the right side of the Euler equation; these cancel in this case not only in steady state but
for all t. This implies we jump to steady state in one period, ignoring nonnegativity constraints on x, which
is valid if f (K0 ) + (1 ¡ ±)K0 ¸ K s where K s is the steady state. We emphasize that this has nothing to
do with money and also holds in the standard growth model with U(x) = x. In the next section U will be
strictly concave and hence we do not jump to the steady state immediately.
19
market as seen in (27), but aggregate consumption X = f(K )+ (1¡±)K ¡K 0 is independent
of Á and q. In terms of policy implications, for example, in this model in°ation will a®ect
the value of money in decentralized trade, but not aggregate consumption or investment in
the general goods market.
We summarize as follows.
Proposition 2 The model with capital dichotomizes: one can solve for the equilibrium path
of q and the equilibrium path of aggregate (X; K ) independently; monetary policy a®ects the
former but not the latter.
4
Capital and Labor
The previous section may help move search-based monetary theory somewhat towards the
mainstream, but does not go all the way. In this section, instead of having agents produce
general goods themselves, we assume there is a representative ¯rm with a constant returns
to scale production function f(K; H; Z) that hires capital K at rate r and labor H at wage
w. The state of technology Z evolves exogenously according to Zt+1 = ³(Zt; " t), where " is
an i.i.d. random technology shock observed at the start of period t. Every night, individuals
supply labor and capital and buy general goods in the centralized market. We assume utility
is separable and linear in leisure, given by 1 ¡ h (total time is 1 and h is hours worked). If
there were no decentralized trade or money, this would be a standard macroeconomic model
{ indeed, except for some minor di®erences in notation it would be identical to the model in
Hansen (1985). 11
As Section 3, during the day agents meet in a decentralized market where they cannot
bring their capital, nor can they trade claims to capital because of anonymity. Bellman's
11 Hansen
(1985) does not directly assume linearity, but uses Rogerson's (1988) indivisible labor model
with lotteries to derive a reduced-form utility function that is linear in hours worked. In terms of its
macro implications this is of no consequence, and one can view the linearity in that model as a primitive.
Alternatively one could also reinterpret the linearity in our model in terms indivisible labor and lotteries.
20
equation is again given by (22), except now the aggregate state is (Z; F ) but in any case this
is subsumed in the notation. The centralized market problem is
W (s) =
max
U(x) + A(1 ¡ h) + ¯EV (m0 + ¿ M; k0 )
0 0
x;h;m ;k
s:t: x = Á(m ¡ m0 ) + wh + rk + (1 ¡ ±)k ¡ k 0 ;
(30)
(31)
where the expectation is with respect to future prices. As is standard, pro¯t maximization
implies these prices will satisfy w = fh (K; H; Z) and r = fk (K; H; Z) in equilibrium. Note
that we do not impose h ¸ 0 here, for the same reason we did not impose y ¸ 0 or x ¸ 0 in
the earlier models, but we can check that this is true later.
Lemma 7 In the centralized market with capital and labor, for all agents and for all t, x,
k 0 and m0 are independent of s = (m; k), Wm =
A
Á
w
and Wk = A
(r + 1 ¡ ±).
w
Proof: Substituting for h from (31) into (30), we have
W (s) = A +
A
w
[Ám + (r + 1 ¡ ±)k]
©
ª
A
0
0
0
0
+ max
U(x)
¡
(x
+
Ám
+
k
)
+
¯EV
(m
+
¿M;
k
)
;
w
0 0
x;m ;k
and everything follows. In particular,
U 0 (x) = A=w
(32)
A=w = ¯EVk (m0 + ¿M; k 0 )
(33)
ÁA=w = ¯EVm (m0 + ¿M; k 0 ) :
(34)
are the ¯rst order conditions. ¥
Lemma 8 In the model with capital and labor, for all t, the single-coincidence bargaining
t m) where g is still given by
solution is the same as in Lemma 2 except now q~t (m) = g ¡1 ( AÁ
wt
(7) and
A
Ám¤t
w
= g(q¤).
21
Proof: Lemma 7 reduces the generalized Nash problem to
·
¸ ·
¸
AÁ µ
AÁ 1¡µ
max u (q) ¡
d
¡c (q) +
d
q;d
w
w
subject to d · m. A argument involving Kuhn-Tucker conditions similar to the one in
Lemma 5 completes the proof. ¥
Earlier, m¤ and q~ depended only on z = Ám, or real balances denominated in the general
good; now they depend on real balances denominated in leisure units,
q~0 =
A
Á=g 0.
w
A
w Ám,
and we have
Otherwise, the bargaining solution is the same as before. Bellman's equation
becomes
©
ª
V (s) = ®¾ u [q (m)] ¡ A
Ád(m) + ®V0 + W (s);
w
where V0 looks like (10) in the previous section except that
A
wÁ
(35)
replaces Á. Hence, V m and
Vmm are given by expressions that look like (11) and (12) except that
A
wÁ
replaces Á. The
conditions that guarantee V is strictly concave still apply, and so we have:
Lemma 9 In the centralized market with capital and labor, for all t, all agents choose the
same m0t and kt0 , and mt+1 = m0t + ¿ Mt < m¤t+1.
Proof: Similar to Lemma 3. ¥
We again have a degenerate distribution F , and agents spend all their money in singlecoincidence meetings. From (32) all agents consume xt = Xt = U 0¡1( wAt ), which depends on
t through the wage but does not depend on the individual state. What di®ers across agents
in this model is individual labor supply, given by
8
< H + wÁ M for buyers
h=
H ¡ wÁ M for sellers
:
H
for others
(36)
where H is aggregate labor supply,
H=
1 ¤
[x + k 0 ¡ (r + 1 ¡ ±)k] :
w
22
(37)
Hence, Ht may depend on t, but given t all individuals supply Ht plus an adjustment to
bring their money holdings to m = M .
To simplify this version of the model, ¯rst insert Vm into (34)
·
¸
A
0
0
ÁA=w = ¯E ®¾u (q)~
q (m) + (1 ¡ ®¾) Á :
w
Then insert ÁA=w = g(q)=m and q~0(m) = g(q)=mg 0 (q) to derive
·
¸
g(qt )
u0(qt+1 ) g(qt+1)
g(qt+1 )
= ¯E ®¾ 0
+ (1 ¡ ®¾)
:
Mt
g (qt+1) Mt+1
Mt+1
We can drop the expectation operator, since nothing on the right hand side is random. Hence
we are right back to (16), which we again repeat for convenience:
·
¸
g(qt )
g(qt+1)
u0(qt+1 )
=¯
®¾ 0
+ 1 ¡ ®¾ :
Mt
Mt+1
g (qt+1)
(38)
Once again, we can solve for the path of q, independently of the other endogenous variables
in the model.
To ¯nd the conditions the other variables must satisfy, insert Vk = Wk =
A
w (r
+ 1 ¡ ±)
and the equilibrium conditions w = fh and r = fk into the remaining ¯rst order conditions
(32) and (33) to yield
A = U 0(Xt )fh(Kt ; Ht ; Zt )
U 0 (Xt ) = ¯EU 0 (Xt+1 ) [fk (K t+1; Ht+1; Zt+1 ) + 1 ¡ ±] ;
(39)
(40)
where Xt = Ht fh + K tfk + (1 ¡ ±)Kt ¡ Kt+1 is aggregate consumption. As is standard, by
Euler's Theorem
Xt = f(K t; Ht ; Zt) + (1 ¡ ±)Kt ¡ Kt+1 :
(41)
Of course, (39), (40) and (41) are nothing more nor less than the standard equations characterizing equilibrium paths for (Xt; Ht ; Kt+1 ) in the stochastic growth model without money.
23
An equilibrium can be de¯ned here in the obvious way (paths for the endogenous variables
satisfying the conditions derived above, plus the usual side conditions, such as a given value
for the initial capital stock K 0). The main point is that the model still dichotomizes: the
set of equilibrium q paths is the same as in the basic LW model and the other real variables
are the same as in the nonmonetary growth model. As before, q a®ects Á and hence, in this
version, individual labor supply, but not aggregate labor supply, and not consumption or
investment, in the centralized market.
We summarize as follows:
Proposition 3 The model with capital and labor dichotomizes: one can solve for the equilibrium path of q and the equilibrium path of aggregate (X; H; K ) independently; monetary
policy a®ects the former but not the latter.
5
Shopping Time
In this section, we allow agents to choose their search intensity, or equivalently, their shopping
time in the decentralized market. One reason is that one can ¯nd in the literature models
where it is simply assumed that individuals have to spend time shopping, where the required
amount of time to purchase a given consumption bundle is some arbitrary decreasing function
of real balances (see Walsh [1998]). Another reason is to see what it does to the dichotomy
results. Thus, we assume an increase in time spent shopping, or in search e®ort, l, increases
one's arrival rate ® = ®(l) in the decentralized market but reduces the time left available for
leisure or labor.12 We assume ®0 > 0 and ®00 < 0. Here leisure is 1 ¡ l ¡ h. Otherwise things
are the same as the previous section.
The state variable when an agent enters the decentralized market is again s = (m; k),
12
It is not hard to derive the relationship ®(l) from an underlying matching technology that takes the
search intensity of agents as inputs, and every agent chooses l taking as given the aggregate search intensity
L; see Berensten, Rocheteau and Shi (2001), e.g., for details.
24
and Bellman's equation is now given by
Z
fu [q (s; s~)] + W [m ¡ d (s; s~) ; k; l]g dF (~
s)
Z
+®(l)¾ f¡c [q (~
s; s)] + W [m + d (~s; s) ; k; l]g dF (~
s)
Z
+®(l)¢ B(~
s; s)dF (~
s) + [1 ¡ 2®(l)¾ ¡ ®(l)¢]W (m; k; l)
V (s) = ®(l)¾
(42)
which is identical to (22) except for the fact that the arrival rate ® is a function of l, and
l is an argument of W since total time left to allocate between leisure and labor in the
centralized market is 1 ¡ l. The problem in the centralized market is
W (m; k; l) = max
U(x) + A(1 ¡ l ¡ h) + ¯EV (m0 + ¿M; k 0 )
0 0
(43)
x;m ;k ;h
subject to (31). Again we do not impose h ¸ 0.
The generalized versions of Lemmas 7 and 8 are:
Lemma 10 In the centralized market with capital, labor and shopping, for all agents and
for all t, x, k 0 and m0 are independent of s = (m; k), Wm =
A
w Á,
Wk =
A
w (r
+ 1 ¡ ±), and
Wl = ¡A.
Proof: Substituting for h from (31) into (43), we have
W (m; k; l) = A(1 ¡ l) + A
w [Ám + (r + 1 ¡ ±)k]
½
¾
A
0
0
0
+ max
U (x) ¡ (x + Ám + k) + ¯EV (m + ¿ M; k ) ;
x;m0 ;k0
w
and everything follows. The ¯rst order conditions are the same as in the previous section,
(32), (33) and (34). ¥
Lemma 11 In the model with capital, labor and shopping, for all t, the single-coincidence
bargaining solution is the same as in Lemma 8.
25
Proof: Obvious.
The usual procedure implies we can write Bellman's equation as
©
ª
V (m; k) = ®(l)¾ u [q (m)] ¡ A
Ád (m) + ®(l)V0 + W (m; k; l);
w
where V0 is the same as above. The only di®erence from the previous section is that l is
endogenous, and as such it must satisfy the ¯rst order condition
Then we have:
©
ª
®0 (l)¾ u [q (m)] ¡ A
Ád
(m)
+ ®0 (l)V0 + Wl = 0
w
(44)
Lemma 12 In the model with capital, labor and shopping, for all t, in the decentralized
market all agents choose the same lt , and in the centralized market all agents choose the
same m0t and kt0 , and mt+1 = m0t + ¿ Mt < m¤t+1.
Proof: Given ®00 < 0 and Wl = ¡A, there is a unique l solving (44). The rest is similar
to the proof of Lemma 3. ¥
We again have a degenerate distribution F , and agents spend all their money in singlecoincidence meetings. Combining these results with (44), in equilibrium we have
®0(l)¾ fu [q (M )] ¡ c [q (M )]g + ®0(l)¢[u(q¤ ) ¡ c(q ¤)] = A:
(45)
Other than determining l the model works the same as in the previous section. Individuals
still supply more or less labor at night depending on what happened during the day, but all
leave the centralized night market with the same m = M . Again, we did not impose h ¸ 0,
but conditions can be assumed to guarantee this is true in equilibrium.
While there the may be good reasons for endogenizing search e®ort, in general, it does
nothing to change our neoclassical dichotomy. Here we can solve the system
·
¸
g(qt )
g(qt+1 )
u0 (qt+1)
= ¯
®(lt )¾ 0
+ 1 ¡ ®(lt )¾
Mt
Mt+1
g (qt+1)
A = ®0 (lt) f¾[u(qt ) ¡ c (qt )] + ¢[u(q ¤) ¡ c(q ¤)]g
26
for the equilibrium paths of (q; l) independently of the paths of (X; K; H), which are still
determined by the usual conditions for the nonmonetary growth model, (39), (40) and (41)
in the previous section. One thing this does illustrate is that our neoclassical dichotomy in
general does not say we can solve for q with a single equation independent of the rest of the
system; it says we can solve for variables determined in the decentralized market independent
of variables in the centralized market.
Proposition 4 The model with capital, labor and shopping dichotomizes: one can solve for
the equilibrium path of (q; l) and the equilibrium path of (X; H; K 0 ) independently; monetary
policy a®ects the former but not the latter.
6
Conclusion
This paper pursues the integration of search-based monetary theory and standard macroeconomics. The setup in Lagos and Wright (2002a) was extended by introducing neoclassical
production and having capital and labor trade in the centralized markets. The result is a
tractable framework that nests the search model and a simple neoclassical growth model. A
strong dichotomy emerges in the versions we considered: one can solve for the outcome in the
decentralized markets and the outcome in the centralized markets independently. The value
of money and hence monetary policy determines output and consumption in decentralized
exchange, but this does not a®ect aggregate employment, investment, or consumption of
general goods. We do not claim the dichotomy will hold in all possible versions of the model,
but we emphasize that we did not \rig" things to get the result { indeed, it was a surprise.
In future work, it may be interesting to investigate what features of similar models do or do
not lead to this kind of dichotomy.
27
References
[1] Azariadis, C. Intertemporal Macroeconomics. Oxford: Blackwell, 1993.
[2] Berentsen, Aleksander, Guillaume Rocheteau and Shouyong Shi. \Friedman Meets Hosios: E±ciency in Search Models of Money." Unpublished manuscript, 2001.
[3] Berentsen, Aleksander, Guillaume Rocheteau and Christopher Waller. \Money and Specialization." Unpublished manuscript, 2002.
[4] Berentsen, Aleksander, Ricardo Lagos and Guillaume Rocheteau. \Mechanism Design
in Monetary Models." Unpublished manuscript, 2002.
[5] Cavalcanti, Ricardo and Neil Wallace. \Inside and Outside Money as Alternative Media
of Exchange." Journal of Money, Credit, and Banking, August 1999, 31, 443-57.
[6] Diamond, Peter A. \Aggregate Demand Management in Search Equilibrium," Journal
of Political Economy, 1982.
[7] Faig, Miguel (2001) \A Search Theory of Money and Comerce with Neoclassical Production." Unpublished manuscript, 2001.
[8] Hansen, Gary D. \Indivisible Labor and the Business Cycle." Journal of Monetary
Economics, 1985.
[9] Kiyotaki, Nobuhiro and John Moore. \Liquidity, Business Cycles and Monetary Policy."
Lecture 2, Clarendon Lectures, November 2001.
[10] Kiyotaki, Nobuhiro and Randall Wright. \On Money as a Medium of Exchange." Journal of Political Economy 97, 1989, 927-954.
[11] Kiyotaki, Nobuhiro and Randall Wright. \A Search-Theoretic Approach to Monetary
Economics." American Economic Review 83, 1993, 63-77.
[12] Kocherlakota, Narayana. \Money is Memory." Journal of Economic Theory 81, 1998,
232-251.
[13] Lagos, Ricardo and Randall Wright \A Unifed Framework for Monetary Theory and
Policy Evaluation." Unpublished manuscript, 2002a.
[14] Lagos, Ricardo and Randall Wright \Dynamics, Cycles and Sunspot Equilibria in
`Genuinely Dynamic, Fundamentally Disaggregative' Models of Money." Unpublished
manuscript, 2002b.
28
[15] Molico, Miguel. \The Distribution of Money and Prices in Search Equilibrium." Unpublished manuscript, 1999.
[16] Rocheteau, Guillaume and Randall Wright. \Money in Search Equilibrium, in Competitive Equilibrium, and in Competitive Search Equilibrium." Unpublished manuscript,
2002.
[17] Rogerson, Richard. \Indivisible Labor, Lotteries and Equilibrium," Journal of Monetary
Economics 66, 1988.
[18] Sargent, Thomas J. Macroeconomic Theory. New York: Academic Press, 1979.
[19] Shi, Shouyong. \Money and Prices: A Model of Search and Bargaining." Journal of
Economic Theory 67, 1995, 467-496.
[20] Shi, Shouyong. \A Divisible Search Model of Fiat Money." Econometrica 64, 1997,
75-102.
[21] Shi, Shouyong. \Search, Money, and Capital." Journal of Monetary Economics 66, 1999,
81-103.
[22] Trejos, Alberto and Randall Wright. \Search, Bargaining, Money and Prices." Journal
of Political Economy 103, 1995, 118-141.
[23] Walsh, Carl E. Monetary Theory and Policy. Cambridge: MIT Press, 1998.
[24] Wallace, Neil. \Whither monetary economics?" International Economic Review 42,
2001, 847-870.
[25] Wallace, Neil. \General Features of Monetary Models and Their Signi¯cance." Unpublished manuscript, 2002.
29
Federal Reserve Bank
PRST STD
of Cleveland
U.S. Postage Paid
Research Department
P.O. Box 6387
Cleveland, OH 44101
Address Correction Requested:
Please send corrected mailing label to the
Federal Reserve Bank of Cleveland,
Research Department,
P.O. Box 6387,
Cleveland, OH 44101
Cleveland, OH
Permit No. 385
@FedFRASER