A Unified Framework For Monetary Theory and Policy Analysis: Ricardo Lagos

A Unified Framework for Monetary Theory and
Policy Analysis
Ricardo Lagos
New York University and Federal Reserve Bank of Minneapolis
Randall Wright
University of Pennsylvania and Federal Reserve Bank of Cleveland
Search-theoretic models of monetary exchange are based on explicit

descriptions of the frictions that make money essential. However, trac-
table versions of these models typically make strong assumptions that
render them ill suited for monetary policy analysis. We propose a new
framework, based on explicit micro foundations, within which macro
policy can be studied. The framework is analytically tractable and easily
quantifiable. We calibrate the model to standard observations and use
it to measure the cost of inflation. We find that going from 10 percent
to 0 percent inflation is worth between 3 and 5 percent of consump-
tionmuch higher than previous estimates.
I. Introduction
Most monetary models in macroeconomics are reduced-form models. By
this we mean that they make assumptions, such as putting money in the
Many people provided extremely helpful input to this project, including S. B. Aruoba,
A. Berentsen, V. V. Chari, L. Christiano, N. Kocherlakota, D. Krueger, G. Rocheteau, S.
Shi, N. Stokey, N. Wallace, and C. Waller. Research support from the C. V. Starr Center
for Applied Economics at New York University, the Suntory-Toyota International Centre
for Economics and Related Disciplines at the London School of Economics, Equipe de
Recherche sur les Marches, lEmploi et la Simulation at University of Paris II, and the
National Science Foundation is gratefully acknowledged. In addition to the usual dis-
claimer, we note that the views expressed here are those of the authors and not necessarily
those of the Federal Reserve Banks of Minneapolis or Cleveland or the Federal Reserve
System.
[ Journal of Political Economy, 2005, vol. 113, no. 3]

2005 by The University of Chicago. All rights reserved. 0022-3808/2005/11303-0001$10.00
463
This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
464 journal of political economy
utility function or imposing cash-in-advance constraints, that are pre-
sumably meant to stand in for a role of money that is not made explicit;
for example, it helps overcome spatial, temporal, or informational fric-
tions. There are models that provide micro foundations for monetary
economics based on search theory, with explicit descriptions of meet-
ings, specialization, information, and so on, but they are ill suited for
the analysis of monetary policy as it is usually formulated. The reason
is that to make the models tractable, people typically adopt extreme
restrictions on how much cash agents can hold.
In this paper we propose a new framework, with explicit micro foun-
dations and without these extreme restrictions, that is easy to use for
policy analysis. To illustrate, we use the model to measure the welfare
cost of inflation. Calibrating parameters to standard observations for
the U.S. economy, we estimate that the gain to reducing inflation from
10 percent to 0 percent is worth between 3 percent and 5 percent of
consumption. This is much higher than earlier estimates based on
reduced-form models. Our interpretation of the results is that building
monetary models with explicit micro foundations not only is theoreti-
cally appealing but also can make a significant difference for quantitative
results.
There are previous attempts to study search models without extreme
restrictions on money holdings. Trejos and Wright (1995) present a
model in which agents can hold any m , but like Shi (1995), they
have analytic results only for the case m {0, 1}. Papers that make some
progress on the general case include Green and Zhou (1998), Camera
and Corbae (1999), Zhou (1999), and Zhu (2003), but they are suffi-
ciently complicated that it is difficult to get substantive analytic results.
One can also study the model numerically, as in Molico (1997), but we
think that it is useful to have a framework that delivers analytic results.
For one thing, simple models with sharp results are often a better vehicle
than computer output for developing economic understanding. Also,
numerical methods make it difficult to say much about issues such as
existence, uniqueness or multiplicity, and dynamicsissues that are es-
pecially relevant in monetary economics. Finally, it is useful to have a
benchmark that is analytically tractable even if (especially if) one ulti-
mately wants to complicate things for the sake of realism, policy rele-
vance, and so forth.
What complicates the analysis in previous search models is the en-
dogenous distribution of money holdings, F(m). Our assumptions make
F(m) degenerate. We accomplish this by assuming quasi-linear prefer-
ences and giving agents periodic access to centralized markets in ad-
dition to the decentralized markets that make money essential as in the
typical search model. Quasi linearity means that there are no wealth
effects in the demand for money, so all agents in the centralized markets

monetary theory and policy analysis 465
choose the same m. Hence F(m) is degenerate across agents in the
decentralized market. This allows us to get sharp analytic results. It also
makes the framework about as easy to use as standard reduced-form
models for addressing policy questions such as the cost of inflation,
although, as we shall see, the answers are different.1
The rest of the paper is organized as follows. Section II describes the
environment. Section III defines equilibrium and derives some basic
results. Section IV introduces monetary policy considerations. Section
V uses a calibrated version to quantify the welfare cost of inflation.
Section VI presents conclusions and discusses some extensions to the
model.
II. The Environment

Time is discrete. There is a [0, 1] continuum of agents who live forever
with discount factor b (0, 1). Each period is divided into two sub-
periods, say day and night. Agents consume and supply labor in both
subperiods. In general, preferences are U(x, h, X, H ), where x and h (X
and H) are consumption and labor during the day (night). Although
in principle we do not need any special restrictions to define or discuss
equilibrium in this model, to get the distribution of money degenerate
we need U to be linear in either X or H. Here we use
U(x, h, X, H ) p u(x) c(h) U(X ) H. (1)
Assume that u, c, and U are twice continuously differentiable with
u 1 0, c 1 0, U 1 0, u ! 0, c 0, and U 0. Also, u(0) p c(0) p 0,
and suppose that there exists q* (0, ) such that u (q*) p c (q*) and
X * (0, ) such that U (X *) p 1 with U(X *) 1 X *.
As in a typical search model, during the day agents interact in a
decentralized market with anonymous bilateral matching, where a is
the probability of a meeting. The day good x comes in many varieties,
of which each agent consumes only a subset. Each agent can transform
labor one for one into one of these special goods that he himself does
not consume. For two agents i and j drawn at random, there are four
possible events. The probability that both consume what the other can
produce (a double coincidence) is d. The probability that i consumes
1
A related model is developed by Shi (1997), who also gets F degenerate but by differ-
ent means: he assumes that the fundamental decision-making unit is not an individual,
but a family with a continuum of agents, and appeals to the law of large numbers. In our
working paper (Lagos and Wright 2004), we provide a detailed discussion of the two
approaches; suffice it to say here that our model avoids some technical problems in the
infinite-family model and seems, at least to us, more natural for many applications. Also,
independent of making the money distribution degenerate, there are many other reasons
why it may be interesting to integrate some competitive markets into search models; see
Sec. VI for some examples.

what j produces but not vice versa (a single coincidence) is j. Sym-
metrically, the probability that j consumes what i produces but not vice
versa is j. And the probability that neither wants what the other produces
is 1 2j d. In a single-coincidence meeting, if i wants the special good
j produces, we call i the buyer and j the seller.
At night agents trade in a centralized (Walrasian) market. With cen-
tralized trade, specialization does not lead to a double-coincidence prob-
lem, and so it is irrelevant whether the night good X comes in many
varieties or one; hence we assume that at night all agents produce and
consume a general good. Agents at night can transform one unit of labor
into one unit of the general good. The general goods produced at night
and the special goods produced during the day are perfectly divisible
and nonstorable. There is another object, called money, that is perfectly
divisible and storable in any quantity m 0. For now the total money
stock is fixed at M, but later we allow it to change over time.
Given our assumptions, during the day the only feasible trades are
barter in special goods and the exchange of special goods for money,
and at night the only feasible trades involve general goods and money.
Special goods cannot be traded at night nor general goods during the
day because they are produced in only one subperiod and are not stor-
able.2 Money is essential in this model for the same reason it is essential
in the typical search model: since meetings in the day market are anon-
ymous, there is no scope for trading future promises in this market, so
exchange must be quid pro quo (see Kocherlakota [1998] and Wallace
[2001] for detailed discussions).
III. Equilibrium
Let F(m)
t
be the measure of agents starting the decentralized day market
at t holding m m . Similarly let G(m)
t
be the distribution at the start of
the centralized night market. The initial distributioneither F0 or G 0,
depending on whether we start the model at t p 0 during the day or
nightis given exogenously. Since for now the total money stock is fixed,
mdF(m)
t p mdG(m)
t p M for all t. Let ft be the price of money in the
centralized market (i.e., 1/ft is the nominal price of general goods).
There is no uncertainty in the basic model except for random matching.
Hence, an individuals decisions in a given period depend on only his
money holdings, m. That is, at each t, since aggregate variables such as
Ft, Gt, and prices are taken as given by individuals, we can characterize
2
Some extensions that allow general goods to be storable are discussed in Sec. VI. Of
course, whether or not goods are storable, we can allow the exchange of intertemporal
claims across centralized markets. Since agents are homogeneous in this market, such
claims will not trade, but we can still price assets like real or nominal bonds.

their decisions in terms of common value functions with m as the only
argument.3
Let V(m)
t be the value function for an agent with m dollars when he
enters the decentralized market and W(m)t the value function when he
enters the centralized market. Since trade is bilateral in the day market,
a sellers production h must equal a buyers consumption x. Hence, we
denote their common value by q(m,t
and use d t(m, m)
m) to denote the
dollars the buyer pays; these may depend on the money holdings of the
buyer m and seller m. Also, in double-coincidence meetings, let B(m, t
be the payoff for an agent holding m who meets someone with m
m) .
Then
V(m)
t p aj {u[q(m,
t
Wt[m d t(m, m)]}dF(m)
m)] t

aj m)] Wt[m d t(m,

{c[q(m,
t
m)]}dF(m)
t

ad B(m,
t
(1 2aj ad)W(m),
m)dF(m)
t t (2)
where the four terms represent the expected payoffs to buying, selling,
bartering, and not trading.
A version of (2) appears in Trejos and Wright (1995) and Molico
(1997), except bVt1(m) replaces W(m)
t , because in those papers there
is no centralized market and all agents can do with money is carry it
forward to the next round of decentralized trade. Here, they get to go
to the centralized market, where they solve
W(m)
t p max {U(X ) H bVt1(m )} (3)
X,H,m
subject to X p H fm t fm
t

, X 0, 0 H H, and m 0, where H
is an upper bound on hours, m is money taken out of the market, and
again ft is the price of money. We assume an interior solution for X
and H. This is guaranteed for X under standard assumptions, but things
are more problematic for H because of quasi linearity. Our approach
is to assume interiority, characterize equilibria, and then check that
3
As editor Nancy Stokey emphasized to us, our equilibrium concept is a blend of
traditional Arrow-Debreu components describing aggregates as functions of time t and
recursive components describing individuals problems as functions of t and individual
state variables. Equilibrium also specifies the distributions of the individual state for all t.
As is usual in monetary models, there may be multiple equilibria, some of which are
nonstationary (there may also be sunspot equilibria, but they are ignored here; see Lagos
and Wright [2003]). The t index indicates the prices and distributions that are relevant,
but there are no aggregate state variables. In Lagos and Wright (2004), we stay closer to
conventional recursive methods, but the results are less general; see nn. 6 and 8 below.

0 ! H ! H is satisfied. Thus, when we say equilibrium here, we always
mean equilibrium with 0 ! H ! H.
Given the value functions, we now consider the terms of trade in the
decentralized market. In double-coincidence meetings, we use the sym-
metric Nash bargaining solution with threat point given by the contin-
uation value W(m) t . It is easy to show that this implies that, regardless
of the money holdings of the agents, each gives the other q* as defined
by u (q*) p c (q*), and no money changes hands (see Lagos and Wright
2004). Hence, B(m, t
p u(q*) c(q*) W(m)
m) t . In single-coincidence
meetings, we use the generalized Nash solution in which the buyer has
bargaining power v 1 0 and threat points are again given by continuation
values. Hence, (q, d) maximizes
[u(q) W(m
t d) W(m)]
t
v
[c(q) W(m
t
d) W(m)]
t
1v (4)
subject to d m and q 0, where m and m are the buyers and sellers
money holdings.
This leads us to the following definition.
Definition. An equilibrium is a list {Vt, Wt, X t, Ht, m t, qt, d t, ft, Ft,
Gt}, where, for all t, V(m) t and W(m)
t are the value functions; X t(m),

H(m)
t , and m t(m) are the decision rules in the centralized market;
q(m,
t

m) and d t(m,
m) are the terms of trade in the decentralized market;
ft is the price in the centralized market; and Ft and Gt are the distri-
butions of money holdings before and after decentralized trade. The
equilibrium conditions are as follows. For all t, (i) given prices and
distributions, the value functions and decision rules satisfy (2) and (3);
(ii) the terms of trade in the decentralized market maximize (4), given
the value functions; (iii) ft 1 0 (i.e., we focus on monetary equilibria);
(iv) centralized money markets clear, m (m)dG(m) t p M (goods markets
clear by Walras law); and (v) {F,t Gt} is consistent with initial conditions
and the evolution of money holdings implied by trades in the centralized
and decentralized markets.4
We now characterize equilibria. Here is an outline of what will follow.
We begin by deriving some properties of the solution to the centralized
market problem. We use these properties to solve the bargaining prob-
lem in the decentralized market and then use these results to simplify
Vt and to solve an individuals problem of choosing m t(m). In particular,
we show that under certain conditions, m t p M for all agents regardless
of the value of m with which they entered the centralized market (i.e.,
4
To see what this last condition entails, consider an agent with m entering the decen-
tralized market at t. With probability aj, he buys and leaves with m dt(m, m) , where m
is a random draw from Ft. Similarly, with probability aj, he sells and leaves with m
m), and with probability 1 2aj , he neither buys nor sells (although he might barter)
dt(m,
and leaves with m. This maps Ft into Gt. Later that period, in the centralized market, an
agent with m chooses mt(m) p mt1, and this maps Gt into Ft1.

Ft1 must be degenerate) in any equilibrium. Finally, we combine the
solutions to the centralized and decentralized market problems to re-
duce the model to a single difference equation.
To begin, substitute for H from the budget equation to write (3) as
W(m)
t p fm
t max {U(X ) X fm
t

bVt1(m )}. (5)
X,m
This immediately implies several things. First, X t(m) p X *, where

U (X *) p 1. Also, m t(m) does not depend on m.5 Third, Wt is linear in
m with slope ft. Given this linearity, the bargaining problem (4) sim-
plifies to
max [u(q) fd]
t
v
[c(q) fd]
t
1v
(6)
q,d
subject to d m and q 0.
We claim that the solution to (6) is
q(m,
t
p
m) {q(m)
q*
t if m ! m*t
if m m*,
t
m if m ! m*t
p{
d (m, m) (7)
t
m* if m m*,
t
where q(m)
t is the qt that solves fm
t p z(qt), with
vc(q)u (q) (1 v)u(q)c (q)
z(q) { , (8)
vu (q) (1 v)c (q)
and m*t p z(q*)/ft. To verify this, notice that if we ignore the constraint
d m, then necessary and sufficient conditions for a solution are
v[c(qt) fd
t t]u (qt) p (1 v)[u(qt) fd

t t]c (qt) (9)
and
v[c(qt) fd
t t] p (1 v)[u(qt) fd
t t]. (10)
Thus u (qt) p c (qt), or qt p q*, and d t p m*t p [vc(q*) (1

v)u(q*)]/ft. If m m*t , the constraint is not binding. If m ! m*t , the so-

lution is given by (9) with d t p m, which easily yields fm t p z(qt). This
verifies the claim.
In words, if the buyers cash is at least m*t , the constraint d m is not
binding and he gets q* for m*t dollars; otherwise the constraint is bind-
5
The result that m does not depend on m suggests that it is reasonable to look for
equilibrium where Ft is degenerate; but we are after bigger game. We want to show that
Ft is degenerate in all equilibria, and even if m does not depend on m, it is possible that
there are multiple solutions to (5) and different agents select different m . Below we give
conditions implying that Vt is strictly concave in the relevant range, and hence there is a
unique solution to (5).

ing, and he spends all of his money to get q(m)t . Notice that the solution
does not depend on the sellers money holdings m at all, and so we
write q(m,
t
p q(m)
m) t
p d t(m) in what follows. For all
and d t(m, m)
m ! m*t , qt(m) p f/z
t

(qt), where
u c [vu (1 v)c ] v(1 v)(u c)(u c c u )
z p 1 0. (11)
[vu (1 v)c ]2
It is a simple matter to check that q(m) t r q* as m r m*t . Hence,
q(m)
t
t
p q(m) is strictly increasing for m ! m*t , is continuous at m*t , and
is constant at q(m)
t p q* for all m m*t .
We can now use what we know about the bargaining solution and
W(m)
t to simplify (2) to
V(m)
t p v(m)
t fm
t max {fm
t

bVt1(m )}, (12)
m
where
v(m)
t { aj{u[q(m)]
t fd
t t(m)} aj c[q(m)]}dF(m)
{ftd t(m) t
t

ad[u(q*) c(q*)] U(X *) X *. (13)

By repeated substitution we have
V(m
t t) p v(m
t t) fm
t t
jpt
j j1 b[vj1(m j1) fj1m j1]}.
b jt max {fm
m j1
(14)
This reduces the choice of the sequence {m t1 } to a sequence of simple

problems defined in terms of primitives, since vt1 is a known function.6
Notice that

vt1(m t1) p aj{u [qt1(m t1)]qt1

(m t1) ft1d t1

(m t1)} (15)
is zero for all m t1 m*t1 by (7). Hence, ft ! bft1 implies that the prob-
lem of choosing m t1 in (14) has no solution, since the objective function
is strictly increasing for all m t1 m*t1. This means that any equilibrium
6
This is not quite true since vt1 depends on Ft1, which we do not know yet; but this
merely influences the intercept and not the choice mt1 . The point is that we eliminated
Vt1 from the problem. In Lagos and Wright (2004), we used standard dynamic program-
ming methods by writing V as a stationary function of (m, f) and showing that, even
though V is unbounded, one can apply the contraction mapping theorem to prove ex-
istence and uniqueness. That is more work and is less general because there are equilibria
that cannot be represented in terms of time-independent functions of (m, f), or any other
aggregate state variable; see n. 8 for an example. However, if one is willing to sacrifice
generality (e.g., to focus on stationary equilibria), then the results in Lagos and Wright
(2004) are useful because one can study the problem maxm {fm bV(m , f)} in (12)
directly.

must satisfy ft bft1. Therefore, the minimum inflation rate consistent
with equilibrium is f/f t t1 p b, which is the Friedman rule. Given
ft bft1, for all t, the objective function in (14) is nonincreasing in
m t1 for m t1 m*t1.
The slope of the date t objective function in (14) as m t1 r m*t1 from
below is proportional to ft bft1 bajft1 S, where
u (q*)2
S{ 1 (16)
u (q*) v(1 v)[u(q*) c(q*)][c (q*) u (q*)]
2
is the buyers marginal gain from bringing an additional dollar into a

single-coincidence meeting evaluated at q p q*. Note that S 0, and
the inequality is strict except when v p 1. So, unless ft p bft1 and
v p 1, the slope of the objective function in (14) as m t1 r m*t1 is strictly
negative, and therefore any solution must satisfy m t1 ! m*t1. In the ex-
treme case in which ft p bft1 and v p 1, the slope of the objective
function at m*t1 is zero; in this case, however, we consider only solutions
that are limits when either ft bft1 r 0 from above or v r 1 from
below. Hence, m t1 ! m*t1 as long as we are not in the extreme case
ft p bft1 and v p 1, and in this case we take the limit.7
For m t1 ! m*t1, we have vt1
p aj[u (qt1)(qt1

)2 u (qt1)qt1

]. We

would like to be able to conclude that vt1 ! 0, since then there will be
a unique solution m t1. In numerical work one can check it directly, but
it seems useful to have some conditions to guarantee it more generally.
For example, suppose that we set c(q) p q, mainly to reduce notation.
Then if we insert q and q , one can check that v takes the sign of
G (1 v)[u u (u )2 ], where G ! 0 but is otherwise of no concern.
The problem in general is the presence of u , but this vanishes if v
1. Alternatively, for an arbitrary v, we can make assumptions on pref-
erences; a sufficient condition is u u (u )2, which holds if u is log
concave.
In summary, we now have simple conditions either on v or on pref-

erences to guarantee vt1 ! 0, and this implies that there is a unique
choice of m t1 in any equilibrium. That is, Ft1 must be degenerate at
m t1 p M. In any monetary equilibrium, the first-order condition eval-
uated at m t1 p M is ft p b[vt1
(M ) ft1], or
ft p b{aju [qt1(M )]qt1

(M ) (1 aj)ft1 }. (17)
7
It is standard in monetary theory to consider only equilibria at the Friedman rule that
are the limit of equilibria as inflation approaches the Friedman rule. Here we can actually
do more, by considering any equilibria at the Friedman rule as long as v ! 1 , or as long
as v p 1, but we consider only equilibria that correspond to a limit as v r 1.

Inserting ft p z(qt)/M and q (M ) p f/z (qt), from the bargaining solu-
t

tion, we arrive at
u (qt1)
z(qt) p bz(qt1) aj [ z (qt1)
1 aj , ] (18)
a simple difference equation in qt. A monetary equilibrium is now char-

acterized by any path for {qt} satisfying (18) that stays in (0, q*), since
qt ! q* follows from the result that m t ! m*t .
For the rest of this paper we focus on stationary equilibria, or steady
states.8 Notice, however, that it was important not to restrict attention
to such equilibria earlier, since the argument that Ft1 is degenerate did
not use stationarity. In any case, in steady state (18) yields
u (q) 1b
p1 . (19)
z (q) ajb
Consider first the limiting case v p 1, which means z(q) p c(q). Then
it is easy to see that a unique solution q 1 0 to (19) exists under standard
conditions, such as u (0) p . For v ! 1, a steady state also exists under
this condition, but we cannot be sure of uniqueness because
u (q)/z (q) may not be monotone. One can show that it is monotone
under certain additional conditions, such as v 1, or c linear and u
log concave (Lagos and Wright 2004). One can also show that
u (q)/z (q) is increasing in v; hence when the solution is unique, we know
that q/v 1 0 (Lagos and Wright 2004). Similarly, we know that
q/a 1 0, q/j 1 0, and q/b 1 0. Also note that at v p 1, (19) implies
q r q* as b r 1; for v ! 1, however, q ! q* even in the limit as b r 1.
To summarize the results, we have shown that in any equilibrium the
distribution of money is degenerate across agents exiting the centralized
and entering the decentralized market. Also, from the choice of m t1
at date t, we know that m t1 m*t1 with strict inequality except in the
extreme case of f/f t t1 p b and v p 1. This means that in single-
coincidence meetings at t 1, the buyer exchanges all his money,
d t1 p M, for qt1 p q t1(M ). A steady state exists under the usual con-
ditions, and although it may not be unique, in general, it will be under
more stringent conditions on preferences or v. Steady states have natural
8
Dynamics are studied in detail in Lagos and Wright (2003), including cyclic, chaotic,
and sunspot equilibria, but we do want to mention one thing here. Equilibrium condition
(18) defines a function qt p Q(qt1) that may not be invertiblei.e., qt1 p Q1(qt) may
be a correspondenceas is standard in monetary models. In this case there is nothing
that precludes equilibria in which we select different qt1 at different dates given the same
qt. For example, under certain conditions, we can construct equilibria in which we select
qt1 from Q1(qt) so that qt cycles between qL and qH for all t ! T and then select qT1
{qL, qH} from Q1(qT). There is no way to represent such an equilibrium in terms of a time-
invariant function of any aggregate state.

comparative static properties, and in particular q is increasing in both
b and v. One thing we want to emphasize is that the steady state is
efficient if and only if q p q*, which requires both b p 1 and v p 1.
To close this section, recall that so far we have simply assumed an
interior solution for H. Suppose that we want to guarantee H 1 0. Be-
ginning at t p 0 with the centralized market, let m 0 be the upper bound
of the initial G 0 distribution. In the candidate equilibrium, X p X * and
m t1 p M, so agents endowed with m work H(m) p X * f 0(M m)
hours. Since in any equilibrium ft f* p z(q*)/M, we have H 0(m) 1 0
for all m as long as
m0 ! M
X*
f*
p M1 [
X*
z(q*)
. ]
One can similarly show that if the lower bound satisfies
X* H
[
m0 1 M 1
z(q*) ],
then H 0(m) ! H for all m. Finally, for t 1, one can show that X * 1
z(q*) and X * ! H z(q*) guarantee 0 ! Ht ! H. Hence, simple condi-
tions imply that the constraint 0 Ht H will be slack for all t.
IV. Changes in the Money Supply

We now allow M to change over time, with new money injected as lump-
sum transfers in the centralized market. The generalization of (18) is
u (qt1)
z(qt)
Mt
pb
z(qt1)
M t1 [
aj
z (qt1)
1 aj . ] (20)
If M t1 p (1 t)M t with t constant, it makes sense to consider steady

states in which q and real balances fM p z(q) are constant, that is, in
which f/ft t1 p 1 t. As in the previous section, f/f
t t1 b is necessary
for equilibrium to exist; hence, we have t t F p b 1, and a lower
bound on feasible t is given by the Friedman rule.
The steady-state condition is now
u (q) 1tb

p1 . (21)
z (q) ajb
As is standard, we can also state this in terms of the nominal interest

rate i, defined by 1 i p (1 r)(1 p), where p p t is the equilibrium
inflation rate and r p (1 b)/b the equilibrium real interest rate:
u (q) i

p1 . (22)
z (q) aj
Assuming a unique monetary steady state, we have q/t ! 0 or, equiv-
alently, q/i ! 0.
If v p 1, then z(q) p c(q), and the efficient outcome q* obtains if and
only if t p t F or, equivalently, i p 0. If v ! 1, however, then q ! q* at
t F. Since a necessary condition for monetary equilibrium is t t F or,
equivalently, i 0, the Friedman rule is always optimal here; but when
v ! 1, it does not achieve q*. The reason is that in this model there are
two types of inefficiencies: one due to b and one to v. The b effect is
standard: when you acquire cash, you can turn it into future consump-
tion; but because b ! 1, you are willing to produce less than the q* you
would produce if you could turn the proceeds into immediate con-
sumption. The Friedman rule corrects this by generating a real return
on money that compensates for discounting. Notice that although this
effect is standard, the model does generate some novel insights about
it, since the frictions show up explicitly: (21) or (22) makes it clear that
the effect depends on search and specialization through the term aj.
The more novel effect is the wedge due to v ! 1. One intuition for
this effect is the notion of a holdup problem. An agent who carries a dollar
into next period is making an investment with cost f, since he could
have spent the cash on general goods. When he uses the money in the
future, he reaps the full return on his investment if and only if v p 1;
otherwise the seller bargains away part of the surplus. Thus v ! 1 reduces
the incentive to invest, which lowers the demand for real balances and
hence q. Therefore, v ! 1 implies q ! q* even at the Friedman rule. The
Hosios (1990) condition for efficiency says that the bargaining solution
should split the surplus so that each party is compensated for his con-
tribution to the surplus in a match. Intuitively, the surplus in a single-
coincidence meeting is all due to the buyer, since the outcome depends
on m but not on m. Hence, efficiency requires v p 1 here.9
The wedge due to v ! 1 is important for issues such as the welfare
cost of inflation. Measuring welfare by V, when v p 1 we know that V
is maximized at t F and achieves the efficient outcome V *, where
(1 b)V * p a(d j)[u(q*) c(q*)] U(X *) X *, (23)
as shown in figure 1. With v p 1, small deviations from t F have very
9
This holdup problem does not arise in the paper by Shi (1997) because of the way
he solves the bargaining problem, although it would if he used a more standard bargaining
solution, as Rauch (2000) points out. See Berentsen and Rocheteau (2003) for a discussion.

Fig. 1.Welfare cost of moderate inflation
small effects on V because of the envelope theorem, just as in the typi-

cal reduced-form (e.g., cash-in-advance) model. When v ! 1, t F is a con-
strained optimum: t ! t F would achieve a higher q and V if it were
feasible, but it is not. Hence the slope of V with respect to t is steep at
t F and the envelope theorem does not apply. A moderate inflation
therefore has a larger welfare cost when v ! 1. We quantify this statement
in the next section.
V. Quantitative Analysis
We parameterize the model as follows. Assume that u(q) p [(q
b)1h b 1h]/(1 h), where h 1 0 and b (0, 1). This generalizes stan-
dard constant relative risk aversion preferences by including b, which
forces u(0) p 0 (a maintained assumption); this does not matter much
quantitatively, however, because we set b 0 for this exercise. Since
q* p 1 b, this means that q* 1. Next, assume that U(X ) p
B log (X ). Notice that this implies X * p B. Finally, assume that c(h) p
h, which makes the disutility of labor the same in the two markets.
We begin with a yearly model, mainly to facilitate comparison with

the existing literature (below we show that a monthly model yields sim-
ilar results). The annual rate of time preference is r p 0.04. We can
normalize a p 1 since results depend only on the products ad and aj.
We set d p 0, but this actually matters little for the results. We shall take
two approaches to setting j: first we shall estimate it along with the
preference parameters, using a procedure discussed below; then, since
it does not matter much for the results, we shall simply fix it at j p
0.5, which means that every agent always has an opportunity to either
buy or sell in each meeting of the decentralized market.10
We now describe the method we use to fit the parameters (h, B, j).
The idea, exactly as in Lucas (2000), is to look at the relationship be-
tween the nominal rate i and L { M/PY. This relationship represents
money demand in the sense that desired real balances M/P are
proportional to Y, with a factor of proportionality L that depends on
the cost of holding cash, i. To construct L in the model, note that
nominal output in the centralized market is X */f p B/f, and nominal
output in the decentralized market is jM. Hence, PY p (B/f) jM and
Y p B jfM. In equilibrium, M/P p fM p z(q), and so
M/P z(q)
Lp p . (24)
Y B jz(q)
Condition (22) gives q and hence L as a function of i; (24) is the money

demand curve implied by theory.
We want to fit this relationship to the data by choosing (h, B, j). We
again follow Lucas (2000) and let i be the commercial paper rate and
let M be M1 (there are issues concerning how to measure all these
variables, perhaps especially M; again our choices are made mainly to
facilitate comparison with previous studies). The sample period is 1900
2000. The fitted values of (h, B, j) are described below. Before we can
proceed, however, we need to discuss the bargaining power parameter,
v. Our method is to try three alternatives: v p 1, which eliminates the
holdup problem and makes our setup closer to previous studies; v p
0.5, which means symmetric bargaining; and v p vm, where vm is the value
that generates a markup m (price over marginal cost) consistent with
the evidence, which we take to be m p 1.1.11
Given v p 1 or 0.5, we fit the parameters to the money demand
data; given vm, we fit them subject to the constraint m p 1.1 at a bench-
10
Setting j p 0.5 maximizes the importance of the decentralized market, but it is still
fairly small in the calibrations reported below: at a benchmark inflation rate of 4 percent,
it contributes less than 10 percent to aggregate output.
11
See Basu and Fernald (1997) for the evidence. To compute m in the model, note that
price over marginal cost in the decentralized market is fM/q , whereas in the centralized
market it is one. Aggregate m averages these markups using the shares of output from
each sector.

Fig. 2.Model and data
mark inflation rate of 4 percent. For example, given v p 1, the best fit
is (h, B, j) p (0.266, 2.133, 0.311). However, things are not precisely
identified; if we fix j p 0.5, we estimate (h, B) p (0.163, 1.968) with
virtually no sacrifice in fit and no change in the welfare implications,
as we shall see below. Hence we often simply set j p 0.5. Figure 2 shows
the fitted relationship for the case v p 1 as the solid line and for the
case v p 0.5 as the dashed line, where in each case we set j p 0.5 and
fit the preference parameters; clearly there is little in these data to
recommend one v over another.
Our measure of the cost of inflation asks how much agents would be
willing to give up in terms of total consumption to have inflation zero
instead of t. For any t, steady-state utility is
(1 b)V(t) p U(X *) X * aj{u[q(t)] q(t)}. (25)

TABLE 1
Annual Model (19002000)
vp1 vm p .343
v p .5 vp1
j p .31 j p .50 j p .50 j p .50 j p .50
h p .27 h p .16 h p .30 h p .39 h p .39
B p 2.13 B p 1.97 B p 1.91 B p 1.78 B p 1.78
(1) (2) (3) (4) (5)
q(t) .243 .206 .143 .094 .522
q(0) .638 .618 .442 .296 .821
q(tF) 1.000 1.000 .779 .568 1.000
1 D0 .014 .014 .032 .046 .012
1 DF .016 .016 .041 .068 .013
If we reduce t to zero but also reduce consumption of both general

and special goods by a factor D, utility becomes
(1 b)VD(0) p U(X *D) X * aj{u[q(0)D] q(0)}. (26)
We measure the cost of t as the value D 0 that solves VD 0(0) p V(t); agents
would give up 1 D 0 percent of consumption to have zero rather than
t. We also consider DF, which is how much they would give up to have
the Friedman rule t F rather than t. Our experiments use t p 0.1 (i.e.,
10 percent inflation), but we also report the costs of a wide range for
inflation at the end of the section.
In table 1, column 1 presents results for the case v p 1 and the fitted
(h, B, j). To focus on one number, we find that going from 10 percent
to 0 percent inflation is worth 1.4 percent of consumption. The column
2 results pertain to the case in which we fix j p 0.5 and refit (h, B). As
mentioned, the results are very similar, especially for welfare. The main
point we want to make is that these numbers are similar to, if slightly
larger than, typical estimates in the literature, including those in Lucas
(2000), which reports a range for 1 D 0, depending on the exact spec-
ification, but typically slightly under 1 percent. We interpret the results
with v p 1 as being in line with, if slightly higher than, the consensus
view in the literature.12
Since j does not matter much for the results, the remaining columns
in table 1 vary v, while fixing j p 0.5 and reestimating (h, B) for each
v. The goodness of fit is basically the same for each v, but the welfare
costs increase and q decreases sharply with v. This indicates that the
12
Lucas actually uses r p 0.03 rather than r p 0.04 , but this has a small effect on our
estimates. Cooley and Hansen (1989, 1991) report even smaller numbers. Molico (1997)
gets small numbers, and sometimes negative numbers, because inflation in his model
beneficially redistributes liquidity to those who need it most; this effect is absent in our
model, of course, because F is degenerate. Wu and Zhang (2000) get bigger numbers
because they assume monopolistic competition, and they also provide references to some
other related studies.

TABLE 2
Annual Model (19592000)
vp1 v p .5 vm p .404 vp1

j p .50 j p .50 j p .50 j p .50
h p .27 h p .45 h p .48 h p .48
B p 3.19 B p 2.92 B p 2.71 B p 2.71
(1) (2) (3) (4)
q(t) .392 .192 .135 .590
q(0) .752 .409 .307 .852
q(tF) 1.000 .602 .478 1.000
1 D0 .008 .025 .031 .007
1 DF .009 .035 .046 .008
holdup problem is serious and is in fact what generates a relatively large

cost of inflation. In the case of v p 0.5, the aggregate markup is m p
1.04. Column 4 generates m p 1.10 at 4 percent inflation and implies
that the welfare costs are 1 D 0 p 0.046 and 1 DF p 0.068substan-
tially larger than the consensus view. To verify that it is indeed the holdup
problem that lies at the heart of these effects, as opposed to the dif-
ferences in other parameters across the columns, column 5 uses the
same parameters as column 4 but sets v p 1. This yields fairly low costs.
Hence, it is v ! 1 and not the other parameters that generates the big
effects.13
Table 2 reports similar experiments fitting the model to a shorter
sample, 19592000. Although the welfare costs are slightly lower, the
main conclusion is the same: decreasing v from 1 to 0.5 or to vm increases
the welfare costs considerably.14 Table 3 reports a final robustness check
by recalibrating so that the period is a month; that is, we transform the
data for Y and i to make them monthly. For these results we estimate
j along with the preference parameters in every case. While the estimates
change when we go from an annual to a monthly model, the overall fit
is about the same.15 What we want to emphasize is that the welfare costs
13
There is a sense in which we may be overestimating the cost of inflation by using
vm, since vm is calibrated to generate an average markup of 10 percent under the assumption
that the centralized market is perfectly competitive; if there were a noncompetitive markup
in the centralized market, we would not need such a low v to match the average markup.
14
Intuitively, the cost of inflation is lower in the shorter sample because the estimated
money demand curve has a much flatter slope in the shorter sample. This is not meant
to be a rigorous explanation, however, especially since we shall argue below that the area
under the money demand curve is not necessarily the right way to measure the cost of
inflation.
15
In particular, estimates of B and j are smaller in the monthly data, for the following
simple reason. First, in equilibrium, centralized market consumption is X* p B , and ob-
viously monthly consumption is less than annual. Second, j is the probability of a single
coincidence, and obviously this probability is lower per month than per year. Intuitively,
this last point is important because it means that we can match velocity equally well when
we vary the period length.

TABLE 3
Monthly Model (19002000)
vp1 v p .5 vm p .315 vp1

j p .033 j p .052 j p .052 j p .052
h p .20 h p .23 h p .33 h p .33
B p .17 B p .15 B p .14 B p .14
(1) (2) (3) (4)
q(t) .230 .151 .101 .552
q(0) .623 .476 .329 .831
q(tF) 1.000 .845 .644 1.000
1 D0 .013 .030 .049 .010
1 DF .015 .038 .069 .011
here are very similar to those in table 1, and so the main conclusion is
robust to changing the period length as well as the sample. That con-
clusion is that bargaining power seems to be a quantitatively important
consideration in estimating the welfare cost of inflation, and one that
previous analyses have missed entirely.
To illustrate the effects of less moderate inflations, figure 3 shows the
welfare cost 1 D 0 for inflation rates ranging from 0 to 150 percent.
The upper curve pertains to vm and the parameters from column 4 of
table 1, whereas the lower curve pertains to v p 1 and the same param-
eters. The difference in the curves is due to the holdup problem. Notice
that the difference gets smaller at big inflation rates, because q gets very
small for big t regardless of v. As the figure makes clear, the costs
basically converge when t reaches 150 percent since decentralized trade
has all but shut down by this point. Hence, the difference between
models with v p 1 and v ! 1 is especially relevant for small to moderate
inflation rates.
Finally, we want to contrast our method with the traditional way of
measuring the cost of inflation, which is to compute the area under the
money demand curve (see the discussion and references in Lucas
[2000]). Our results show that this procedure does not work in general.
If we start with a value for v and fit parameters to money demand and
then change v and refit the parameters, we match the data equally well
but get very different values for the welfare cost. Knowing the empirical
money demand curve is not enough: one really needs to understand
the micro foundations, and especially how the terms of trade are de-
termined, in order to correctly estimate the welfare cost of inflation.
VI. Conclusion
We have presented a new framework for monetary economics, explicitly
based on the frictions used in search theory, but without the restrictions

Fig. 3.Welfare cost of higher inflation
on money holdings usually made in those models. A key innovation is

to allow agents to interact periodically in centralized as well as decen-
tralized markets. Given that agents have quasi-linear preferences, the
distribution of money is degenerate in equilibrium, and this keeps the
model tractable. We characterized equilibria and discussed some policy
issues. The Friedman rule is optimal but does not achieve the first-best
for v ! 1. We found that this has sizable implications for the cost of
inflation: going from 10 percent to 0 percent inflation here is worth
between 3 percent and 5 percent of consumptionmuch larger than
most previous estimates. This indicates that building monetary theories
with explicit foundations matters for quantitative analysis. We think that
all of this constitutes progress in terms of bringing micro and macro
models of money closer together.
We also think that we have only scratched the surface, and much
more can be done. In Lagos and Wright (2004), we report several ex-
tensions. For example, we add real shocks, either match-specific or ag-
gregate, and either independently and identically distributed or persis-
tent. Although in that model the constraint d m may not bind with
probability one, we show that F is still degenerate. We also discuss the
effects of uncertainty in M. One experiment is to keep total M constant
and randomly transfer m across agents. Again d m may not bind with
probability one, and we show that a mean-preserving spread in m always

reduces welfare, even though it may increase f and q if u 0 because
of a precautionary demand effect. We also consider transfers t that
are the same for all agents but are random over time. This version
delivers natural results and remains fairly tractable: we show that if t is
independently and identically distributed, then f and q are constant;
and if t is persistent, f and q are smaller in periods of high t because
this implies forecasts of higher future inflation.
Other extensions include the paper by Aruoba and Wright (2003),
who add neoclassical firms and capitalthat is, they make general goods
storableand integrate the framework with a standard real business
cycle model. However, they assume that capital is not used in the de-
centralized market, which implies some very special results; Waller
(2003) and Aruoba, Waller, and Wright (2004) generalize this. Lagos
and Rocheteau (2004) make the general good storable and allow it to
compete with money as a medium of exchange. Rocheteau and Wright
(2005) add heterogeneity and free entry, which allows one to analyze
effects on the extensive margin (number of trades), in addition to the
intensive margin (quantity per trade). Lagos and Rocheteau (2005)
endogenize search intensity and study the effects of inflation on velocity,
output, and welfare. Some of these papers also consider alternative
pricing mechanisms, such as price taking or posting, instead of bar-
gaining. See also Ennis (2004), Faig and Huangfu (2004), and Roche-
teau and Waller (2004). Ennis (2004) and Rocheteau and Wright (forth-
coming) redo the quantitative experiments in this paper under some
of these alternative pricing mechanisms.
Faig (2004) asks when credit and insurance markets can replace quasi-
linear utility. Williamson (forthcoming) studies policy in a version with
seasonal and other fluctuations in the demand for liquidity. Lagos
(2005) extends the basic environment to study liquidity and asset prices.
Bhattacharya, Haslag, and Martin (2005) study policy with heteroge-
neous agents. Reed and Waller (2004) discuss risk sharing. Berentsen,
Camera, and Waller (2004) and He, Huang, and Wright (2005) intro-
duce roles for banks. Rocheteau and Craig (2004) consider sticky
prices. Berentsen et al. (2005) assume that agents may be in the de-
centralized market for more that one round of trade, which makes the
distribution no longer degenerate but still tractable. Kahn, Thomas, and
Wright (2004) assume that utility is not quasi-linear; this model can be
solved numerically to show that when wealth effects are not too big, the
results are close to those derived here. While not an exhaustive list, this
gives a sense of a few of the applications and extensions that are possible.
References
Aruoba, S. Boragan, Christopher Waller, and Randall Wright. 2004. Money and
Capital. Manuscript, Univ. Maryland.

Aruoba, S. Boragan, and Randall Wright. 2003. Search, Money, and Capital: A
Neoclassical Dichotomy. J. Money, Credit and Banking 35, no. 6, pt. 2 (Decem-
ber): 10851105.
Basu, Susanto, and John G. Fernald. 1997. Returns to Scale in U.S. Production:
Estimates and Implications. J.P.E. 105 (April): 24983.
Berentsen, Aleksander, Gabriele Camera, and Christopher J. Waller. 2004.
Money, Credit and Banking. Manuscript, Notre Dame Univ.
. 2005. The Distribution of Money Balances and the Non-neutrality of
Money. Internat. Econ. Rev. 46 (May).
Berentsen, Aleksander, and Guillaume Rocheteau. 2003. On the Friedman Rule
in Search Models with Divisible Money. Contributions to Macroeconomics 3 (1).
Bhattacharya, Joydeep, Joseph H. Haslag, and Antoine Martin. 2005. Hetero-
geneity, Redistribution, and the Friedman Rule. Internat. Econ. Rev. 46 (May).
Camera, Gabriele, and Dean Corbae. 1999. Money and Price Dispersion. In-
ternat. Econ. Rev. 40 (November): 9851008.
Cooley, Thomas F., and Gary D. Hansen. 1989. The Inflation Tax in a Real
Business Cycle Model. A.E.R. 79 (September): 73348.
. 1991. The Welfare Costs of Moderate Inflations. J. Money, Credit and
Banking 23, no. 3, pt. 2 (August): 483503.
Ennis, Huberto. 2004. Search, Money, and Inflation under Private Information.
Discussion Paper no. 142 (August), Inst. Empirical Macroeconomics, Fed.
Reserve Bank Minneapolis.
Faig, Miguel. 2004. Divisible Money in an Economy with Villages. Manuscript,
Univ. Toronto.
Faig, Miguel, and Xiuhua Huangfu. 2004. Competitive Search in Monetary
Economies. Manuscript, Univ. Toronto.
Green, Edward J., and Ruilin Zhou. 1998. A Rudimentary Random-Matching
Model with Divisible Money and Prices. J. Econ. Theory 81 (August): 25271.
He, Ping, Lixin Huang, and Randall Wright. 2005. Money and Banking in
Search Equilibrium. Internat. Econ. Rev. 46 (May).
Hosios, Arthur J. 1990. On the Efficiency of Matching and Related Models of
Search and Unemployment. Rev. Econ. Studies 57 (April): 27998.
Kahn, Aubhik, Julia Thomas, and Randall Wright. 2004. The Distribution of
Money in Search Equilibrium: Quantitative Theory and Policy Analysis. Man-
uscript, Univ. Minnesota.
Kocherlakota, Narayana R. 1998. Money Is Memory. J. Econ. Theory 81 (August):
23251.
Lagos, Ricardo. 2005. Asset Prices and Liquidity in an Exchange Economy.
Manuscript, New York Univ.
Lagos, Ricardo, and Guillaume Rocheteau. 2004. Money and Capital as Com-
peting Media of Exchange. Staff Report no. 341 (August), Res. Dept., Fed.
Reserve Bank Minneapolis.
. 2005. Inflation, Output, and Welfare. Internat. Econ. Rev. 46 (May).
Lagos, Ricardo, and Randall Wright. 2003. Dynamics, Cycles, and Sunspot Equi-
libria in Genuinely Dynamic, Fundamentally Disaggregative Models of
Money. J. Econ. Theory 109 (April): 15671.
. 2004. A Unified Framework for Monetary Theory and Policy Analysis.
Staff Report no. 346 (September), Res. Dept., Fed. Reserve Bank Minneapolis.
Lucas, Robert E., Jr. 2000. Inflation and Welfare. Econometrica 68 (March): 247
74.
Molico, Miguel. 1997. The Distribution of Money and Prices in Search Equi-
librium. PhD diss., Univ. Pennsylvania.

Rauch, Bernhard. 2000. A Divisible Search Model of Fiat Money: A Comment.
Econometrica 68 (January): 14956.
Reed, Robert, and Christopher Waller. 2004. Money and Risk Sharing. Man-
uscript, Notre Dame Univ.
Rocheteau, Guillaume, and Ben Craig. 2004. State-Dependent Pricing, Inflation
and Welfare. Manuscript, Fed. Reserve Bank Cleveland.
Rocheteau, Guillaume, and Christopher Waller. 2004. Bargaining in Monetary
Economies. Manuscript, Fed. Reserve Bank Cleveland.
Rocheteau, Guillaume, and Randall Wright. 2005. Money in Search Equilib-
rium, in Competitive Equilibrium, and in Competitive Search Equilibrium.
Econometrica 73 (January): 175202.
. Forthcoming. Inflation and Welfare in Models with Trading Frictions.
In Monetary Policy in Low Inflation Economies, edited by D. Altig and E. Nosal.
Cambridge: Cambridge Univ. Press.
Shi, Shouyong. 1995. Money and Prices: A Model of Search and Bargaining.
J. Econ. Theory 67 (December): 46796.
. 1997. A Divisible Search Model of Fiat Money. Econometrica 65 (Jan-
uary): 75102.
Trejos, Alberto, and Randall Wright. 1995. Search, Bargaining, Money, and
Prices. J.P.E. 103 (February): 11841.
Wallace, Neil. 2001. Whither Monetary Economics? Internat. Econ. Rev. 42 (No-
vember): 84769.
Waller, Christopher. 2003. Comment on Search, Money, and Capital: A Neo-
classical Dichotomy by S. Boragan Aruoba and Randall Wright. J. Money,
Credit and Banking 35, no. 6, pt. 2 (December): 111117.
Williamson, Stephen. Forthcoming. Search, Seasonality and Monetary Policy.
Internat. Econ. Rev.
Wu, Yangru, and Junxi Zhang. 2000. Monopolistic Competition, Increasing
Returns to Scale, and the Welfare Costs of Inflation. J. Monetary Econ. 46
(October): 41740.
Zhou, Ruilin. 1999. Individual and Aggregate Real Balances in a Random-
Matching Model. Internat. Econ. Rev. 40 (November): 100938.
Zhu, Tao. 2003. Existence of a Monetary Steady State in a Matching Model:
Indivisible Money. J. Econ. Theory 112 (December): 30724.


A Unified Framework For Monetary Theory and Policy Analysis: Ricardo Lagos

Uploaded by

Copyright:

Available Formats

A Unified Framework For Monetary Theory and Policy Analysis: Ricardo Lagos

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Unified Framework For Monetary Theory and Policy Analysis: Ricardo Lagos

Uploaded by

Copyright:

Available Formats

A Unified Framework for Monetary Theory and

Search-theoretic models of monetary exchange are based on explicit

[ Journal of Political Economy, 2005, vol. 113, no. 3]

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

II. The Environment

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

aj m)] Wt[m d t(m,

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This immediately implies several things. First, X t(m) p X *, where

v)u(q*)]/ft. If m m*t , the constraint is not binding. If m ! m*t , the so-

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

ad[u(q*) c(q*)] U(X *) X *. (13)

This reduces the choice of the sequence {m t1 } to a sequence of simple

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

is the buyers marginal gain from bringing an additional dollar into a

ft p b{aju [qt1(M )]qt1

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

a simple difference equation in qt. A monetary equilibrium is now char-

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

IV. Changes in the Money Supply

If M t1 p (1 t)M t with t constant, it makes sense to consider steady

As is standard, we can also state this in terms of the nominal interest

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

Fig. 1.Welfare cost of moderate inflation

small effects on V because of the envelope theorem, just as in the typi-

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

Condition (22) gives q and hence L as a function of i; (24) is the money

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

Fig. 2.Model and data

(1 b)V(t) p U(X *) X * aj{u[q(t)] q(t)}. (25)

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

If we reduce t to zero but also reduce consumption of both general

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

vp1 v p .5 vm p .404 vp1

holdup problem is serious and is in fact what generates a relatively large

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

vp1 v p .5 vm p .315 vp1

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

Fig. 3.Welfare cost of higher inflation

on money holdings usually made in those models. A key innovation is

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

This content downloaded from 164.076.007.055 on August 19, 2017 17:15:55 PM

You might also like

v)u(q)]/ft. If m mt , the constraint is not binding. If m ! m*t , the so-

ad[u(q) c(q)] U(X ) X . (13)

(1 b)V(t) p U(X ) X aj{u[q(t)] q(t)}. (25)