Professional Documents
Culture Documents
This Content Downloaded From 219.223.223.250 On Fri, 09 Sep 2022 05:34:59 UTC
This Content Downloaded From 219.223.223.250 On Fri, 09 Sep 2022 05:34:59 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic
Literature
This essay surveys the new monetarist approach to liquidity. Work in this litera
ture strives for empirical and policy relevance, plus rigorous foundations. Questions
include: What is liquidity? Is money essential in achieving desirable outcomes? Which
objects can or should serve in this capacity? When can asset prices differ from fun
damentals? What are the functions of commitment and collateral in credit markets?
How does money interact with credit and intermediation? What can and should
monetary policy do? The research summarized emphasizes the micro structure of
frictional transactions, and studies how institutions like monetary exchange, credit
arrangements, or intermediation facilitate the exchange process. (TEL E24, E31, E42,
E44, E52, G10, G21)
This, as I see it, is really the did not seem to them a promising
central issue in the pure theory of field for economic analysis. This is
money. Either we have to give an where I disagree. I think we have to
explanation of the fact that people look frictions in the face, and see if
do hold money when rates of interest they are really so refractory after all.
are positive, or we have to evade —Hicks (1935)
the difficulty somehow ... The great
evaders ... would have put it dotvn Progress can be made in monetary
to "frictions," and since there was theory and policy analysis only by
no adequate place for frictions in modeling monetary arrangements
the rest of their economic theory, a explicitly.
theory of money based on frictions —Kareken and Wallace (1980)
* Lagos: New York University. Rocheteaui University delivered by Wright as the 2011 Toulouse Lectures over a
of California, Irvine. Wright: University of Wisconsin week at the Toulouse School of Economics, and he thanks
Madison, FRB Minneapolis, FRB Chicago, and NBER. them for their generous hospitality. Wright also acknowl
A previous version of this essay circulated as "The Art edges the NSF and the Ray Zemon Chair in Liquid
of Monetary Theory: A New Monetarist Perspective" Assets at the Wisconsin School of Business for research
(to pay homage to Clarida, Gali, and Gertler 1999). support. Lagos acknowledges the C.V. Starr Center for
For comments we thank the editor and four referees. Applied Economics at NYU as well as the hospitality of
Particularly useful input was also provided by Davidthe Cowles Foundation for Research in Economics at
Andolfatto, Ed Green, Ricardo Cavalcanti, Mei Dong, Yale University, the Federal Reserve Bank of Minneapo
John Duffy, Pedro Gomis-Porqueras, Ian King, Yiting lis, and the Federal Reserve Bank of St. Louis. The usual
Li, Narayana Kocherlakota, Andy Postlewaite, Danielladisclaimers apply.
Puzzello, Mario Silva, Alberto Trejos, Neil Wallace, + Go to https://doi.org/10.1257/jel.20141195 to visit the
Steve Williamson, and Sylvia Xiao. Parts of the essay werearticle page and view author disclosure statement(s).
371
Research discussed below borrows from arise from having agents set prices in dollars
GE, and some models adopt the abstraction and making it difficult or costly to change.
of competitive markets, when convenient,If money were really such a hindrance, how
but others rely heavily on search theory,did it survive all these centuries? The work
which is all about agents trading with each reviewed here tries to get the relevant phe
other. They also use game theory, since the nomena, like monetary or credit arrange
issues are inherently strategic, e.g., what ments, to arise endogenously, as beneficial
one accepts as a medium of exchange or colinstitutions. This can give different answers
lateral might depend on what others accept. than reduced-form models, and allows
At the intersection of search and game theadditional questions. How can one purport
ory, the papers often use bargaining, which to understand financial crises or banking
leads to insights one might miss if exclu problems using theories with no essential
sively wed to a Walrasian auctioneer. Otherrole for payment or settlement systems in
ways to determine the terms of trade arethe first place? How can one hope to assess
studied, too, including price taking, post the effects of inflation—a tax on the use of
money—using theories that do not incor
ing, and more abstract mechanisms. Key
frictions include spatial and temporal sepaporate the frictions that money is meant to
ration, limited commitment, and imperfect remedy? We do not claim the papers dis
information. These can make arrangementscussed below have definitive answers to all
these questions, but contend that they pro
involving money socially beneficial, which is
not the case in models with cash-in-advance vide useful ways to think about them.
(CIA) or money-in-utility-function (MUF) To hint at where we are going, consider
assumptions. Such reduced-form devices these well-known examples of work that
presumably stand in for the idea that mon shares our stance on avoiding shortcuts
etary exchange helps overcome some diffi to model monetary phenomena. There is
culties—but then, one might ask, why not a large body of work using overlapping
model that? generations (OLG) economies, with major
contributions by Samuelson (1958), Lucas
In CIA models, having to use cash hurts; in
sticky-price models, nothing but problems (1972), and Wallace (1980). Since Kiyotaki
and Wright (1989), more monetary econ
omists employ search theory (Jones 1976,
most of the models of an economy which we have, and Oh 1989, and Iwai 1996 are other early
I am thinking here of many besides those of Arrow and
Debreu, have no formal account for the exchange pro attempts at search-based models of money).
cess." Clower (1970) similarly says: "conventional value However, as discussed below, spatial sep
aration per se is not crucial, as clarified
theory is essentially a device for logical analysis of virtual
trades in a world where individual economic activities are
costlessly coordinated by a central market authority. It by Kocherlakota (1998), in line with ear
has nothing whatever to say about delivery and payment lier work by Ostroy (1973) and Townsend
arrangements, about the timing or frequency of market (1989). On banking, we discuss research
transactions, about search, bargaining, information and
other trading costs, or about countless other commonplace
drawing on Diamond and Dybvig (1983),
features of real-world trading processes." On middlemen, Diamond (1984), and Williamson (1986,
Rubinstein and Wolinsky (1987) say: "Despite the import 1987). On intermediation, related contri
ant role played by intermediation in most markets, it is
largely ignored by the standard theoretical literature. This
butions include Rubinstein and Wolinsky
is because a study of intermediation requires a basic model (1987) and Duffie, Garleanu, and Pedersen
that describes explicitly the trade frictions that give rise to (2005). On secured credit, the models share
the function of intermediation. But this is missing from the
standard market models, where the actual process of trad themes with Kiyotaki and Moore (1997,
ing is left unmodeled." 2005), while on unsecured credit, they
make use of Kehoe and Levine (1993), and we show how Lagos and Rocheteau (2009)
Alvarez and Jermann (2000).4 relax it in models of over-the-counter (OTC)
By way of preview, we start with the asset markets along the lines of Duffle,
first-generation search models of money by Garleanu, and Pedersen (2005). We also dis
Kiyotaki and Wright (1989, 1993), Aiyagari cuss information-based models.
and Wallace (1991, 1992), and others, to Before getting into theory, let us clarify
illustrate trade-offs between asset returns a few terms and put them in historical per
and acceptability, and to show how econospective. It is commonly understood that
mies where liquidity plays a role are pronedouble-coincidence problems plague direct
to multiplicity and volatility. We also present barter: in a bilateral meeting between indi
elementary versions of Kocherlakota s (1998)viduals i and j, it would be a coincidence if
results on the essentiality of money, andi liked j s good, and a double coincidence
Cavalcanti and Wallaces (1999b) analysis of if j also liked i's good.5 Reference is often
inside and outside money. We then move tomade to a medium of exchange, and on this
the second-generation models by Shi (1995),it is hard to improve on Wicksell (1911): "an
Trejos and Wright (1995), and others, withobject which is taken in exchange, not for
divisible goods. This permits further exploits own account, i.e., not to be consumed by
ration of the efficiency of monetary exchangethe receiver or to be employed in technical
and the idea that liquidity considerations lead production, but to be exchanged for some
to multiplicity and volatility. We also develop thing else within a longer or shorter period
a connection with the above-mentioned of time." A medium of exchange that is also
models of intermediation. a consumption or production good is com
We then move to divisible assets, as in com
modity money. By contrast, Wallace (1980)
defines^# money as a medium of exchange
putational work by Molico (2006) and others,
that is intrinsically useless (neither a con
or the more analytically tractable approaches
in Shi (1997a) and Lagos and Wright (2005).
sumption good nor a productive input) and
These models are easier to integrate with
inconvertible (not a claim on consumption
mainstream macro and allow us to examine or production goods). This usefully delin
many standard issues in a new light. We alsoeates a pure case, although assets other than
discuss the effects of monetary policy on laborcurrency can convey moneyness—i.e., can be
markets, the interaction between money and more or less easy to use in transactions.
other assets in facilitating exchange, and the A challenge in monetary economics is to
theme that observations that seem anoma describe environments where an institution
lous from the perspective of standard finanlike money is essential. As introduced by
cial economics can emerge naturally when Hahn (1973), essentiality means welfare is
there are trading frictions. Also, just as some higher, or the set of incentive-feasible allo
papers relax the assumption of indivisible cations is larger, with money than without
assets in early search-based monetary theory, it. In some applications, it is fairly evident
that one should want to use models where
money is essential—e.g., again, to under and early example of the methods used in
stand the cost of the inflation tax, one might the literature.7
want to first understand the benefits of mon Time is discrete and continues forever.
etary exchange, and the fact that money is There is a [0,1] continuum of infinitely lived
so prevalent over time suggests that as an agents that meet bilaterally at random. To
institution, it must be welfare improving. introduce gains from trade, assume they
That this is nontrivial is evidenced by the have specialized tastes and technologies:
fact that money is not essential in standard there are N types of agents and N goods,
theories. As Debreu (1959) says about his, where type j agents consume good j but
an "important and difficult question ... not produce good j + 1 modulo N (i.e., type N
answered by the approach taken here: the agents produce good N + 1 = 1). For now,
integration of money in the theory of value." the fraction of type / is rij = 1/N, and we set
An objective here is to clarify what ingredi N — 3. Although we usually call the agents
ents are relevant for getting liquidity consid consumers, obviously similar considerations
erations into rigorous theory, and to show are relevant for producers. Instead of say
how economics gets more interesting when ing individual i produces good i + 1 but
they are there.6 wants to consume i, it is a relabeling to say
firm i uses i to produce i + 1, generating the
same motives for and difficulties with trade.
2. Commodities as Money
However, although these difficulties hinder
We begin with Kiyotaki and Wright barter, the double-coincidence problem per
(1989). The goal is to derive equilibrium se does nothing to rule out credit. For that
transactions patterns endogenously and see we need a lack of enforcement/commitment
if they resemble trading arrangements in plus imperfect information, as assumed here
actual economies, in a stylized sense, such and discussed in detail in section 3.
as certain commodities acting as media of Goods are indivisible. They are stor
exchange, or certain agents playing the role able only one unit at a time. Let pj be the
of middlemen. While abstract relative to return on good j—i.e., the flow utility agents
much of what follows, this goal still seems
get from a unit of it in inventory. One can
relevant, and a discussion seems appropriate,
interpret pj > 0 as a dividend, or fruit from
because this setting provides a rudimentary
a Lucas (1978) tree, and pj < 0 as a storage
cost. As discussed in Nosal and Rocheteau
(2011, chapter 5), it is a venerable idea that
the intrinsic properties of objects influ
ence which can or should serve as media
6 Related surveys or discussions include Wallace (2001,
2010), Wright (2005), Shi (2006), Lagos (2008), Williamson
of exchange,
and Wright (2010a, 2010b), Nosal and Rocheteau (2011), and storability is the property
and Waller (2015). By way of comparison, in this essay in
we: focus here. Typej agents also get utility
(1) discuss liquidity generally with money a special case;
u > 0 by consuming good j, and then pro
(2) connect more to finance and labor; (3) provide sim
duce a new unit of good j + 1 at cost c = 0.
plified versions of difficult material not found elsewhere;
Type j agents always accept good j in trade,
(4) highlight a few quantitative results; and (5) cover work
over the past five years. Surveys of the very different new
Keynesian approach include Clarida, Gali, and Gertler
(1999) and Woodford (2003). To be clear, while there
are disagreements, it is certainly desirable that different
practices are allowed to flourish. Also, it is hard not to 7be
Without much loss of continuity, it is possible to slap
impressed by Keynesian success with policy makers and toin
section 6 in order to reach the frontier more quickly, or
the media; the more modest goal here is to communicate to skip to section 3 for a model that came later but is easier
to scholars that there is an alternative. than this one.
and immediately consume, given |/0y| is not are: rx = 1 if Vi3 > V12; r, = 0 if V13 < V12;
too big (more on this below). and Tj = [ 0, 1 ] if V13 = V]2. Calculation
The following aspect of strategies is to be implies V13 — V12 takes the same sign as
determined: Will type j agents trade good
j + l for good j + 2 in an attempt to facilitate (4) A] = p3 - p2 + a[n3m3(l — r3)
acquisition of good y? Or will they hold onto
good j + 1 until trading directly for good j? -n2(l-m2)\u.
If Tj is the probability type j agents trade
good j + 1 for good j + 2, a symmet In (4), p3 — p--i is the return differential
ric, stationary, strategy profile is a vector from holding good 3 rather than good 2. If
t = (r1; r2, r3). If Tj > 0, typej agents use returns were all agents valued, this would be
good /' + 2 as a medium of exchange. Also to the sole factor determining T\. But the other
be determined is the distribution of inven term is the difference in the probability of
tories. Since typej agents consume good j, acquiring Is desired good when holding good
they always have either good j + 1 or j + 2. 3 rather than 2, or the liquidity differential.
Hence, m = (t7ihm2,m3) gives the dis Whether type 1 agents should opt for indi
tribution, where m, is the proportion of rect exchange, swapping good 2 for 3 when
type / agents with goodj + 1. The probability ever they can, involves comparing return and
type i agents meet type j agents with good liquidity differentials. This reduces the BR
j + 1 each period is anj-nij, where a is the condition for T\ to a check on the sign of A^
probability of meeting anyone and rij = 1 /N. Indeed, for anyj:
The appendix derives the steady state (SS)
condition for each /, 1 if A, > 0
(5) Tj = [0,1] if Aj = 0.
(1) rrij nJ+1 mj+1 Tj — (1 - rrij) nJ+2 mJ+2.
0 if Aj < 0
To describe payoffs, let r be the rate of
time preference and V = ( where Vy the A stationary, symmetric equilibrium is a list
value function of type i holding goodj. For (V, m, t) satisfying the DP, SS, and BR con
notational convenience, let the utility from ditions. There are eight candidate equilibria
dividends be realized next period. Then, for in pure strategies, and for each such t, one
type 1,
can solve for to, and use (5) to determine the
parameters for which t is a BR to itself (see
(2) rV12 = p2 + an2{l- m2)u the appendix).
To present the results, assume p1? p2 > 0
+ an3m3T3u = io3, so we can display outcomes in the
positive quadrant of (ph po) space. Figure 1
+ an2m2r1( V13 — V12), shows different regions labeled by t to
indicate which equilibria exist. There are
two cases, model A or B, distinguished by
(3) rVj3 = p3 + an3m3(u + V12 — V13), Pi > P2 or Pi > Pi-8 1° figure 1, model A
corresponds to the region below the 45° wants. In contrast, type 2 agents always
line, where there are two possibilities: if accept good 2 but only have good 1 with
p2 > p-2 the unique outcome is t = (0,1,0); probability m2 = 1/2. Hence, good 3 allows
and if p2 < P2 it is t = (1,1,0). To under type 1 agents to consume sooner. If p2 > p2
stand this, note that for type 1, good 3 is more this liquidity factor does not compensate for
liquid than good 2. It is more liquid since a lower return; if p2 < p% it does. The rea
type 3 agents accept good 3 but not good 2, son type 1 is pivotal is this: for type 2 agents,
and type 3 agents always have what type 1 trading good 3 for good 1 enhances both
liquidity and return, as does holding onto
good 1 for type 3. Hence, only type 1 agents
but easier. To see how they differ, consider model A with have a trade-off.
Pi > p2- Then at least myopically it looks like a bad idea for
In model A, t = (0,1,0) is called the fun
type 1 to set = 1, because trading good 2 for 3 lowers
his return. Similarly, it looks like a bad idea for type 3 to damental equilibrium. It features good 1
set t3 = 1, and a good idea for type 2 to set r2 = 1. Hence, as the universally accepted commodity
exactly one type is predisposed to use indirect trade based
on fundamentals. In model B, types 2 and 3 are both so money, and has type 2 agents acting as
predisposed. middlemen by acquiring good 1 from its
producers and delivering it to its consumers. types and N goods, and prove existence of an
While this is a natural outcome, if p2 < ~p2 equilibrium where the highest-return good is
we instead get t = (1,1,0), called a specu universally accepted (but there can also exist
lative equilibrium. This outcome has type 1 equilibria where this good is not universally
agents trading good 2 for the lower-return accepted). This is nontrivial because stan
good 3 to improve their liquidity position, dard fixed-point theorems do not guarantee
and both goods 1 and 3 are used for indirect equilibrium with a particular exchange pat
exchange. Theory delivers cutoffs for type 1 tern (one of Hahn's 1965 problems; see also
to sacrifice return for liquidity, but there is Zhu 2003, 2005). An extension by Kiyotaki
a gap: for p2> p%> P2 there is no station and Wright (1989) and Aiyagari and Wallace
ary, symmetric equilibrium in pure strate (1992) is to add fiat currency. We postpone
gies. Kehoe, Kiyotaki, and Wright (1993) discussion of this, but mention that it pro
show there is one in mixed strategies, where vides one way to see that equilibria are not
type 1 agents accept good 3 with probabil generally efficient: for some parameters,
ity t* € (0,1). They also show there can be equilibria with valued fiat money exist and
multiple stationary mixed-strategy equilib dominate other equilibria. In terms of com
ria, but the set of such equilibria is finite. paring commodity-money equilibria when
Whenever p2 > P2 > P2 they also construct they coexist, it may seem better to use the
nonstationary equilibria with r* cycling over highest-return object as money, but some
time—an early (perhaps the first?) exam agents may prefer to have other objects so
ple of production and exchange fluctuating used, like those who produce these objects,
as a self-fulfilling prophecy due to liquidity reminiscent of the bimetalism debates (e.g.,
considerations.9 see Friedman 1992).
In model B, above the 45° line, there is To study how we get to equilibrium, sev
always an equilibrium with r = (0,1,1). This eral papers use evolutionary dynamics.10
is the fundamental equilibrium for model B, Wright (1995) has a general population
where type 1 agents hang on to good 2, which n = (rii,n2,n^), and in one application
now has the highest return, while types 2 agents can choose their type. This can be
and 3 opt for indirect exchange, with goods interpreted as choosing preferences, or tech
1 and 2 serving as money. For some param nologies, or as an evolutionary process where
eters, there coexists an equilibrium with types with higher payoffs increase in num
t = (1,1,0), the speculative equilibrium for bers due to reproduction or imitation. In
this specification, where good 2 is not uni model A, with n( evolving according to stan
versally accepted even though it now has dard Darwinian dynamics, for any initial n0,
the best return. The coexistence of equilib and any initial equilibrium if n0 admits mul
ria with different transactions patterns and tiplicity, n, —> iigo where at the unique
liquidity properties shows that these are not
necessarily pinned down by fundamentals.
This is the baseline model. In an exten 10These include Matsuyama, Kiyotaki, and Matsui
(1993), Wright (1995), Luo (1998), and Sethi (1999).
sion, Aiyagari and Wallace (1991) allowRelatedly,
N Marimon, McGrattan, and Sargent (1990) and
Ba§£i (1999) ask if artificially intelligent agents can learn
to play equilibrium in the model. There are also studies
in the lab. In these experiments, Brown (1996) and Duffy
9Oberfield and Trachter (2012) show that the setand of Ochs (1999) find subjects have little problem finding
the fundamental equilibrium, but can be reluctant to adopt
dynamic equilibria shrinks as the length of the period gets
small in a version of the model. Still, a recurrent theme
speculative strategies. Duffy (2001) shows they can learn
below is that economies where liquidity plays a role areto do so. Duffy and Ochs (2002) also experiment with ver
generally prone to multiplicity and volatility. sions including fiat currency.
equilibrium is speculative. Intuitively, with randomize exist for many parameter val
fundamental strategies type 3 agents get the ues, and in such equilibria goods with lower
highest payoff, since they produce a good p have higher acceptability. Intuitively, to
with the best return and highest liquidity. make agents indifferent between lower- and
Ergo, n3 increases, and as type 1 agents higher-return objects, the former must be
interact with type 3 more often, they are more liquid. This outcome, which can be
more inclined to sacrifice return for liquidity. socially desirable, is related to Gresham's
Motivated in part by criticism of random law, and more generally captures rigorously
matching (e.g., Howitt 2005 or Prescott a robust idea: Abstracting from risk, for the
2005), Corbae, Temzelides, and Wright sake of illustration, whenever agents are
(2003) redo the model with directed search. indifferent between two assets (e.g., savings
Generalizing Gale and Shapley (1962), at and checking deposits), as they must be if
each t the population partitions into sub they are willing to hold both, the one with a
sets containing at most two agents such lower return must be more liquid.12
that there are no profitable deviations in
trade or trading partners for any individual
3. Assets as Money
or pair. For model A with nt — 1/3, with
directed search the fundamental outcome Adding other assets allows us to illustrate
t— (0,1,0) is the unique equilibrium in additional results: (1) assets can facilitate
a certain class. On the equilibrium path,intertemporal exchange; (2) this may be true
starting at m = (1,1,1), type 2 trades withfor flat currency, an asset with a 0 return, or
type 3 while type 1 sits out. Next period, even for those with negative returns; (3) for
at m = (1,0,1), type 2 trades with type 1money to be essential, necessary conditions
while type 3 sits out, putting us back atinclude limited commitment and imperfect
m = (1,1,1). Different from random matchinformation; (4) the value of fiat money is
ing, there is a unique outcome and it features both tenuous and robust; and (5) whether
good 1 as money. Heuristically, with randomassets circulate as media of exchange may
search, in speculative equilibrium type 1not be pinned down by primitives.13
cares about the chance of meeting type 3 Goods are nonstorable and produced on
with good 1; with directed search, chance isthe spot for immediate consumption, at cost
not a factor because the endogenous trans c > 0. Hence, they cannot be retraded (one
action pattern is deterministic. Indeed,
type 2 agents cater to type 1 agents' needs by
acting as middlemen, delivering consump 12While this should suffice to illustrate how an early
tion every second period. Hence, one mightformalization of the liquidity concept works, additional
results are available—e.g., Camera (2001) discusses inter
say that some randomness is needed to makemediation in more detail, while Cuadras-Morato (1994)
operative type Is precautionary demand forand Yiting Li (1995) incorporate recognizability consider
liquidity, as seems natural.11 ations in versions with private information. We return to
these topics in sections 5 and 10.
As a final application, Renero (1998,1999) 13The setup here follows Kiyotaki and Wright (1991,
shows that equilibria where all agents1993), simplified in various ways. In the original model,
as in Diamond (1982), agents go to one "island" to pro
duce and another to trade; here they produce on the spot.
Also, in early versions, agents consume all goods, but
11 However, these results have only been established forlike some more than others and have to choose which to
model A and nt = 1/3 fixed exogenously; it is not knownaccept; here that choice is made obvious. Also, in early for
what happens in model B, for 1/3 or for n deter mulations agents with assets could not produce, and so had
mined endogenously, so there is still work to be done onto use money even in double-coincidence meetings; that is
this model. relaxed in this presentation, following Siandra (1990).
might think of them as services). Agents matching process. However, if they cannot
still specialize, but now, when i and j meet, commit, we must check that said promise is
the probability of a double coincidence is credible. This entails an incentive condition,
8, the probability of a single coincidence IC, when an agent is supposed to produce
where i likes j's output is a, and symmet in a single-coincidence meeting, and of
rically the probability of a single coinci course it depends on the consequences of
dence where j likes i's output is also a. So deviating. If VD is the deviation payoff, IC is
the double-coincidence arrival rate is aS,
with a and S capturing search and matching, (6) -c + Vc>/iVD + (l-/x)Vc,
respectively, and the equations below hold
ing in discrete time or continuous time with where fi is the probability deviators are
a interpreted as a Poisson parameter. Any caught and punished. Thus, /x < 1 captures
good that i likes gives him the same utility imperfect monitoring, or record keeping,
u > c; all others give him 0. There is a stor so that deviations are only probabilistically
able asset that no one consumes but yields a detected and communicated to the popu
flow utility p. If p — 0 it is fiat currency. For lation at large. The best punishment is the
now, assume that agents neither dispose of harshest, which is banishment to autarky,
nor hoard assets (we check this below). but that may not be feasible, depending on
Let A 6 [0,1] be the fixed asset supply, details, so we take the best punishment to be
and for now continue to assume assets are a loss of future credit (these are of course the
indivisible and agents can hold at most same
1 if <5 = 0). Then VD = VB, in which case
unit, a G {0,1}. To begin, however, let A = 0,
(6) holds if and only if
so that if credit is not incentive feasible,
barter is the only option. The flow barter(7) r <fc = naa(u — c)/c.
payoff is rVB = a5(u — c). Given 6 > 0,
this beats autarky: V > VA — 0. But given As is standard, low r is necessary for coop
a > 0, it does not do all that well, because
erative behavior. Or, one can alternatively
say that /( must not be too small, meaning
in some meetings i wants to trade but / will
not oblige, which is bad for everyone ex ante.the monitoring/communication technol
Suppose we try to institute a credit system,ogy must be sufficiently good. If credit is
where agent i produces forj whenever j likesnot viable, consider monetizing exchange.
i's output. This is credit because agents pro
Let Va be the value function for agents with
duce while receiving nothing by way of quid a £ {0,1}, and call those with a = 1 (a = 0)
pro quo, with the understanding—call it abuyers (sellers). Then
promise—that someone will do the same for
them in the future. The flow payoff to this (8) rV0 = ad \u — c)
arrangement is
+ aaAroT^Vi — V0 - c),
rVc = aS(u — c) + ami — aac
(9) rVj = ad{u — c)
= a(6 + <t) (m — c). + aa( 1 - A) Tq t±(u + V0 - Vx)
Ti = T0 = 1
The above discussion concerns symmet called bankers, are monitored in all meetings;
ric meetings in a large population. Jin and the rest, called nonbankers, are never mon
Temzelides (2004) have some meetings itored. Agents can now issue notes, pieces
involving people who know each other and of paper with names on them, having no
others involving those who do not. Hence, coupon (p = 0) but potentially having value
credit works in some meetings but not others. in exchange. Notes issued by anonymous
With a finite set of agents, Araujo (2004) agents are never accepted for payment—
shows that even if deviations cannot be com why produce to get a note when you can
municated to the population at large, some print your own for free?—but notes issued
times social norms and contagion strategies by monitored agents may be accepted, which
can be used to enforce credit: if someone fails is why these agents might be interpreted as
to produce for you, stop producing for oth banks. This setup can be used to compare an
ers, who stop ... and eventually this gets back outside money regime (only fiat currency)
to the original deviant. This cannot dissuade to an inside money regime (only notes). The
deviations if the population is large, however, outside money regime is similar to our base
consistent with the stylized fact that money line model, except we can now exploit the
is used in large or complicated societies but fact that some agents are monitored.
not in small primitive ones. Even with a large Let W be average utility across monitored
population, we need imperfect monitoring, and unmonitored agents, and, for illustra
and n < 1 is just one way of modeling this.16 tion, suppose monitored agents never hold
In Cavalcanti and Wallace (1999a, 1999b), money (given A this does not affect W). Let
a fraction nm of the population, sometimes us try to implement an outcome where mon
itored agents produce for anyone that likes
their output. For simplicity, set 5 = 0 to rule
"'This random monitoring formulation follows Gu out barter. Then a monitored agents flow
et al. (2013a, 2013b) and Araujo et al. (2015). Cavalcanti payoff is aa(nmu — c), since he produces for
and Wallace (1999a, 1999b) have some agents monitored
and not others. Sanehes and Williamson (2010) have some anyone but only consumes in a meeting with
meetings monitored and not others. Koeherlakota and probability nm, since anonymous agents do
Wallace (1998) and Mills (2008) assume information about
not produce without quid pro quo. In some
deviations is detected with a lag. Amendola and Ferraris
(2013) assume information about deviations is sometimes applications, monitored agents can always
lost. See also Carmona (2015). become anonymous, but suppose here they
can be punished by autarky, making it easy These models are in some ways crude.18
to compute their IC. For anonymous agents, Yet they nicely illustrate how money can be
beneficial and clarify the requisite frictions.
(12) rV0 = aanmu + aa(l — nm) In applications, Kiyotaki and Wright (1993),
Camera, Reed, and Waller (2003), and Shi
x(l-A)(« + V1 —Vo), (1997b) endogenize specialization to show
how money enhances efficiency. Burdett et
(13) rV1 = aanmu al. (1995) analyze who searches, buyers or
sellers, and Victor E. Li (1994, 1995) con
+ acr( 1 - nm)A(—c + V0 - Vj), siders taxing money to correct externalities
from endogenous search intensity. Ritter
which modifies (8)-(9) by recognizing that (1995), Green and Weber (1996), Lotz and
they can consume but do not have to pro Rocheteau (2002), and Lotz (2004) con
duce when they meet a monitored agent, sider introducing new currencies, while
and A now denotes the asset supply per non Matsuyama, Kiyotaki, and Matsui (1993) and
monitored agent. Generalizing (11), a moni Zhou (1997) consider international curren
tored agent's IC is cies. Corbae and Ritter (2004) further ana
lyze credit. Amendola (2008) discusses ways
(14) r < aa{l — nm)(l — A)(u — c)/c. in which one might rule out the nonmone
tary outcome, making monetary exchange
Notice W is maximized at A* = 1/2 (as more robust. Araujo and Shevchenko (2006)
in the baseline model, this maximizes trade and Araujo and Camargo (2006, 2008) study
volume). In the other regime, with no out learning, which is especially interesting
side money, monitored agents can issue because with bilateral matching, information
notes. This allows them to consume when diffuses through the population slowly due
they meet nonmonitored agents. Also, one to the same search frictions that are part of
can specify that with some probability, bankwhat make money useful. There are several
ers require nonmonitored agents with notes models introducing private information (see
to turn them over to get goods, to adjustfn. 58). While simple, based on all this work,
the measure of notes in circulation. Againwe contend that these models have many
A = 1 /2 is optimal. It is easy to check that Wfeatures that undeniably ring true.
is higher with inside money, because it lets
monitored agents trade more often by issu
ing notes as needed. While this is not too sur
prising, the virtue of the method in general
1® In particular, everything is indivisible, and that does
is that it allows one to concretely discuss thedrive some results. Shevchenko and Wright (2004) argue
relative merits of different arrangements.1' that any equilibrium with partial acceptability, r0 € (0,1),
is an artifact of indivisibility, but then show how adding
heterogeneity yields a similar multiplicity and robust par
tial acceptability. Note also that indivisibilities introduce a
complication that we ignore: as in many nonconvex envi
ronments, agents may want to use lotteries, producing in
exchange for a probability of getting a (Berentsen, Molico,
l7See Mills (2008), Wallace (2010, 2013, 2014b), and Wright 2002; Berentsen and Rocheteau 2002; Lotz,
Wallace and Zhu (2007), and Deviatov and Wallace (2014) Shevchenko, and Waller 2007). Of course, lotteries are not
for more applications. Related but different studies of at all the same as mixed strategies. In figure 2, e.g., if one
banking in this kind of model include Cavalcanti, Erosa, allows lotteries, what happens is this: for large p, the sellers
and Temzelides (1999, 2005), Cavalcanti (2004), He, gives his goods to the buyer for sure, and with some prob
Huang, and Wright (2005), and Lester (2009). abily he gets the asset.
Trejos and Wright (2016) characterize theand Trejos and Wright (2016) also construct
outcomes. Figure 3 shows the case of a relsunspot equilibria where q fluctuates ran
atively high 61, with subcases depending ondomly, while Coles and Wright (1998) con
p. Starting with p — 0 (middle panel), therestruct continuous-time cycles where q and
are two steady states, q = 0 and a uniqueA revolve around steady state. This formal
qe > 0 solving e(q") = 0. There are alsoization of excess volatility in asset values is
dynamic equilibria starting from any q0 £different from results in ostensibly similar
(0,qe), where q—> 0, due to self-fulfilling models without liquidity considerations
inflationary expectations. For an asset with a (Diamond and Fudenberg 1989; Boldrin,
moderate dividend (right panel), e(q) shiftsKiyotaki, and Wright 1993; Mortensen
down, leaving a unique steady state qe £1999), which require increasing returns.
(0,q) and a unique equilibrium, since any
Here dynamics are due to the self-referential
other solution to q = e(q) exits [0,Zy], If wenature of liquidity. Also note that to get an
increase the dividend further to p >p (notasset, sellers incur a cost above the funda
shown), e(q) shifts further, the steady statemental value, pIr. In standard usage this is
with trade vanishes, and assets get hoarded.a bubble: "if the reason that the price is high
For a moderate storage cost p & (p, 0) (left today is only because investors believe that
panel), there are two steady states with trade,the selling price is high tomorrow—when
qeH G (0,q) and ql €E (0,qeH), plus equilibria 'fundamental' factors do not seem to justify
where q —► ql due to self-fulfilling expec such a price—then a bubble exists" (Stiglitz
tations. For p < p (not shown), there is no1990). Now, one could argue that liquid
equilibrium with trade and agents disposeity services are fundamental (e.g., Tirole
of a. 1985), but rather than discuss semantics, we
Multiple steady states and dynamics emphasize the economics: liquidity consid
arise because the value of liquidity is partly erations alone can generate deterministic or
self-fulfilling: if you think others give low qL stochastic fluctuations.
for an asset then you only give a little to get In terms of efficiency, clearly q = q* is
it; if you think they give high qH for it then desirable, where u'[q*) — c'(q*). Related to
you give more. Shi (1995), Ennis (2001), Mortensen (1982) and Hosios (1990), with
fiat money and a fixed A, one can construct
conceptual merit of connecting q = T(q) to t = T(t) in 61 such that q = q * in equilibrium if and only
section 3. if 91 — 9J, where 9\ < 1 if and only if r is not
First best
Second best
too big. Hence, if agents are patient 6\ <panel1 of figure 4, drawn assuming A >c(q*),
achieves the first best; otherwise 9i =q*1 can be achieved for some 6\ at the tan
achieves the second best. One can also take gency between the frontier and the 45° line,
61 as given and maximize W with respect to and q is too low (high) if Q\ is below (above)
A. With q exogenous, as in section 3, the 6\. In in the right panel, drawn assuming
solution is A* = 1/2. With q endoge A < c(q*), output is too low for all 0ly and
nous, if q <q* then A* < 1/2 due to an the second best obtains at Oi = 1.
envelope-theorem argument. This captures, Many results do not rely on a particular
in a very stylized way, the notion that mone bargaining solution, and various alternatives
tary policy should balance liquidity provision used in the literature can be nested by letting
and control of the price level. Although it v(q) be a generic mechanism describing how
depends here on the upper bound for a, it much value a buyer must transfer to a seller to
illustrates the robust idea that it is the distri get q, so the terms of trade solve A = «(</).21
bution of liquidity that really matters.
For a diagrammatic depiction of wel
fare, let Si = u(q) — A and S0 = A — c(q) 21 In terms of the literature, Shi (1995) and Trejos and
denote the buyers and sellers surplus. Any Wright (1995) use symmetric Nash bargaining; Rupert,
Schindler, and Wright (2001) use generalized Nash; they
trade must satisfy the ICs, Sj > 0 and S0 > 0. all consider threat points given by continuation values and
The relationship between S | and S0 as q by 0, both of which can be easily derived from strategic
changes, the frontier of the bargaining set, isbargaining in nonstationary settings (Binmore, Rubinstein,
and Wolinsky 1986). Coles and Wright (1998) use a strate
S0 = — c[u_1(S1 + A)] + A in figure 4.gic solution for nonstationary equilibria. Trejos and Wright
Kalai s solution selects the point on the fron (2016) use Kalai bargaining. Others use mechanism design
tier intersecting the ray S0 = (9o/61) Sj. As (Wallace and Zhu 2007; Zhu and Wallace 2007), price
posting (Curtis and Wright 2004; Julien, Kennes, and
0] increases, this ray rotates and changes Sj, King 2008; Burdett, Trejos, and Wright 2016) or auctions
S0, andq. Let S* = u(q*) — c{q*). In the left (Julien, Kennes, and King 2008).
Consider, e.g., generalized Nash bargain applications include Curtis and Waller
ing with threat points given by continua (2000), Waller and Curtis (2003), Wright and
tion values: nuix(/ \u[q) — A]01 [A -c{q)\\ Trejos (2001), Camera, Craig, and Waller
The FOC can be written A = v[q) with (2004), and Craig and Waller (2004).
Let ay be the rate at which i meets j (there when he has a & {0,1}. In fact, from these
is always a population n consistent with this). DP equations, one might observe that this
In PC matches, P gives the indivisible good model looks a lot like the one in section 4,
to C for qCP. In MC matches, if M has a = 1 where the indivisible good was money,
he gives it to C for qcM-1'1 MP matches, they except here u(q) = c(cj) = c/.24
cannot trade if a = 1, and may or may not The BR conditions are standard, e.g.,
trade if a = 0, but if they do, M transfers qMP e — I if VP > 0. So is the SS condition.
to P. Let m be the measure of type M with Equilibrium satisfies the obvious condi
a— I. The qtj that i gives j splits the surplus, tions, and implies payments in direct trade
where 6y is the share or bargaining power qCP = 6PCu, wholesale trade qMP = 9PMA,
of i, with Oji + 6j = 1, consistent with Nash and retail trade qCM — 0MCu + 9CMA. The
or Kalai bargaining, since u[q) = c[q) — q. spread, or markup, is qCM - qMP = 6MCu +
Similar to entry by firms in Pissarides (2000), (9MC ~ @mp) A. Equilibrium exists and is
we need to determine the fraction of P that generically unique, as shown in figure 5 in
produce, say e, and the fraction of M that {—Pp,—Pm) space (remember that — p is
the storage cost). When —pP is high, e = 0
actively trade, say r. In stationary, pure-strat
egy, asymmetric equilibria, a fraction e of and the market closes. When — p? is low
type P and a fraction r of type M are always and —Pm high, e = 1 and r= 0, so there is
active while the rest sit out. In terms of eco production but not intermediation. When
nomics, a key feature is that storage costs are both are low we get intermediation. For
sunk when agents meet, implying holdup some parameters, P enters with probability
problems.23 e £ (0,1), with m adjusting endogenously to
To illustrate the method suppose that make VP = 0. Also, note that when P has a
t— 0, so C can only trade directly with P. poor storage technology, a low rate of finding
Then rVc = CYCPe{u — qCP), where u — qCp C, or low bargaining power, intermediation
is Cs surplus, since the continuation value is essential: the market opens if and only if
Vc cancels with his outside option Vc. Notice middlemen are active.
e appears because we assume uniform ran Rubinstein and Wolinsky (1987) use
dom matching, in the sense that C can con 9y — 1/2 \/ij and Pp — Pu — 0. In that case,
tact P even if the latter is not participating P is always active, and M is active if and only
(imagine contacts occurring by phone, with if aMC > aPC, as is socially efficient. More
aCP the probability per unit time C and P generally, again related to Mortensen (1982)
are connected, but P only picks up if he is and Hosios (1990), Nosal, Wong, and Wright
on the market). Bargaining implies u — qCp (2015a, 2015b) show that equilibrium is effi
= dcpu, where u is total match surplus since cient if and only if the 6s take on particular
for P the continuation value also cancels with values. Related work includes Li (1998),
his outside option. Then rVc = acp£0cpu. Shevchenko (2004), and Masters (2007,
Similar expressions hold for r ^ 0, and for VP 2008); see Wright and Wong (2014) for a
and Va, where the latter is Ms value function recent paper with more citations to related
papers on middlemen in similar settings.
Pm
(e, r) = (1,0)
(e, r) = (0, 0)
Vine**
(e, t) = (1, 1)
(e. r)
= (e\ 1)
-P,,
Vpc "
j £ {0,1} and P\> po, and they switch as fol One can interpret qA — qD as the fee D
lows: in any state j, shocks implying j' = 1 charges when he gets an asset for I in
and j' — 0 arrive at Poisson rates Uj and u>q. the interdealer market, and similarly for
Again, there is a divisible good anyone qD — qB. The round-trip spread, similar
can consume and produce, for u(q) = q and to the markup in Rubinstein-Wolinsky, is
c((j) — q (interpretable as transferable util s = qA — qs — #d(Ai — A0) > 0.
ity, although some people say it loosely rep Trading strategies are summarized by
resents a bank account that agents can use r = (rA, rB), where ta is the probability that
to save and borrow). It is to be determined when D meets I with a = 0 and j = 1 they
if I trades with D, but I trades with another agree to exchange the asset for qA, while tb
I if and only if one has a — 1 and j = 0 while is the probability that when D meets I with
the other has a = 0 and j — 1, a double a = 1 and j = 0 they agree to exchange the
coincidence. Let Vaj be Is value function asset for qB. The BR conditions are again
with asset position a and valuation j, so that standard, e.g., ta = 1 if Ay > qD, etc. The
Aj — Vy — V()j is the value to I of acquiring reason I with a — 0 and j = 1 might not
the asset when he is in state j. When I trades trade when he meets D is that qD>
with I, the total surplus is Sj = — A0 and is possible. Similarly, I with a = 1 and
the one that gives up the asset gets a trans j = 0 might not trade when he meets D if
fer qt (the subscript indicates I trades with qD < A0. For market clearing, because the
I). This yields the party that entered with asset is indivisible, it will be typically the
asset position a a share 9a, with 60 + 61 = 1. case that qD 6 {A0, A,}, and the long side
The individual gains from trade are qt — A0 of the market will be indifferent to trade. If
= 01 Sj and A [—<:/; = 0() Sj, and therefore,
the measure of D trying to acquire an asset
qi = + $1^1. exceeds the measure trying to off-load one,
The rate at which I meets I is ah then
and qA
the= qD = Aj and rA 6 (0,1). In the
probability that I has asset position opposite
a and valcase, qB = qD = A0 and tb e (0,1).
uation j is maj. The rate at which I Since
meets theD measure of I trying to acquire
is aD. When I meets D, if they trade, assets isone
m10 and the measure trying to divest
can think of the latter as trading in the themselves
inter of assets is m01, the former is on
dealer market on behalf of the former for a the short side if and only if m10 < m(ll if and
fee. Obviously this is only relevant when D only if A < A = u>i/(w0 +
meets I with a — 0 and / = 1 or meets I with The SS and DP equations are stan
a — 1 and j — 0, since these are the only type dard—e.g., the flow payoff for I with a — 1
I agents that currently want to trade. If I gets and low valuation is the dividend, plus the
the asset, D gets qA (for ask); if I gives up an expected value of trading with I or D, plus
asset, D pays qB (for bid). Since the cost to D the capital gain from a preference shock:
of getting an asset on the interdealer market
is q D, when D gives an asset to I in exchange
r^io = Po + + aDTB[qB — A0)
for qA, the bilateral surplus is SA — A1 — qD.
And when D gets an asset from I the sur
plus is SB = qD — A0. If 0D is D's bargaining + wi(Vn — V10).
power when he deals with I, then
An equilibrium is a list (V, t, m) satisfying
(19) 9a — @dAi + (1 — #d)<7d and the usual conditions, and it exists uniquely.
The terms of trade are easily recovered, as is
qB — 0dA o + (1 — 0D)qD. the bid-ask spread.
This stylized structure, with a core of A = [0, oo), and studies the case of flat cur
dealers and a periphery of investors that rency, allowing the supply to evolve accord
may trade with each other or with dealers, ing to A+1 = (1 + vr) A, where the subscript
is a reasonable representation of many OTC indicates next period, and ir > 0 is the rate
markets. The proportion of intermediated of monetary expansion, generated by a lump
trade is aD /(aD + az). If aD is small, most sum transfer T.29
exchange occurs between investors, as in As above, agents can be buyers or sell
markets for specialized derivatives or fed ers depending on who they meet, but now
eral funds; small a.; better approximates they are not constrained by a G {0,1}.
NASDAQ. The case aD — 0 is interesting Maintaining the commitment and infor
for making connections between money and mation assumptions precluding credit, and
finance: this model has gains from trading setting S = 0 to eliminate barter, we let
a due to heterogeneous valuations, with q q(aB,as) be the quantity and d(aB,as) the
a payment instrument; the one in section 4 dollars traded when the buyer has aB and the
has gains from trading q, with a the pay seller as, assuming for simplicity (aB,as) is
ment instrument. For intermediation theory, observed by both. Then
otj = 0 is nice since the Vs and qs are inde
pendent of m. This makes it easy to see that (20) V(a) = (1 - 2cur) 0V+1(a + T)
spreads are decreasing in aD and increasing
6d, and as r —+ 0, qA, qB, and qp go to the
same limit, which is p0 / r ifA > A and p j /r if
A < A. There are also implications for trade + 0V+1[a-d(a,as)
volume, often associated with liquidity, but
these results are sensitive to a £ {0,1}. The + T]}dF(as)
time has come to relax this restriction, first in
monetary then financial economics.27
the expected value of not trading; the second liquidity around. At the same time, inflation
is the value of buying from a random seller; detrimentally reduces real balances, and pol
and the third is the value of selling to a ran icy must balance these effects.30 Chiu and
dom buyer. Molico (2010, 2011) extend the analysis to let
To determine terms of trade, consider agents sometimes access a centralized mar
generalized Nash bargaining, ket where they can adjust money balances
(more on this below); in one version, they
(21) maxSB (q, d, aB, as)eSs (q, d,aB, ag)1"6, must pay a fixed cost for this access, resem
q,d
bling Baumols (1952) inventory approach,
where SB(-) = u(q) + /3V+l(aB + T - d) - but using general equilibrium and not just
PV+i(aB + T) and Ss(-) = -c(q) + decision theory.
/3V+i(as + d + T) — /3V+1(as + T) are theThose papers focus on stationary equilib
surpluses. The maximization is subject to
ria. Chiu and Molico (2014) and Jin and Zhu
d <aB. One sometimes hears that this looks (2014) extend the methods to study dynamic
like a CIA restriction, but in this context transitions
it after various types of monetary
is simply a feasibility condition saying that
injections, and show how the redistributional
impact of policy on F(a) can have rather
agents cannot hand over more than they
have in quid pro quo exchange. That does notinteresting effects on output and prices. In
make this a CIA model, because the analysis Jin and Zhus formulation, for some param
eters, output in a match q(aB, as) is decreas
explicitly describes agents trading with each
other, and while it happens that barter and ing and convex in as. This means a policy
credit are ruled out by the environment in that increases dispersion in real balances
increases average q. Now, there are other
this specification, it is not hard to let some
barter or credit back in. That is different effects, and this is a numerical result about
from tacking on a CIA restriction over and the net effect (explaining why the findings
above the usual budget constraint in classicaldiffer from Molico or Chiu-Molico). The
competitive models (more on this below). important point is that there are cases where
There is a law of motion for F(a) with a a monetary injection increases dispersion in
standard stationarity condition. A stationaryreal balances and hence average q, and that
equilibrium is a list (V,q,d,F) satisfying leads to slow increases in prices after the
these conditions. Molico (2006) analyzes injection, where again p = d/q. The reason
the model numerically. He studies the rela is not that prices are sticky—indeed, q and d
tionship between inflation and dispersion in are determined endogenously by bargaining
prices, defined by p — d/q, and asks what in every single trade—but the increase in q
happens as frictions vanish. He also studies keeps p from rising quickly during the tran
the welfare effects of inflation. Increasingsition. This implies that nominal rigidities
A by giving agents transfers proportional toare not needed to capture time-series obser
their current a has no real effect—it is like vations that suggest money shocks affect
a change in units. But a lump-sum transfer mainly output in the short run and prices in
T > 0 compresses the distribution of real the longer run.
balances because it raises the price level, and
when the value of a dollar falls it hurts those
with high a more than Iowa. Since those with 30WaIlace (2014b) conjectures that in any economy
low a don't buy very much, and those with with this trade-off there genetically exists an incentive
compatible scheme, with transfers nondecreasing in wealth
high a don't sell very much, this compression and not necessarily lump sum, that is inflationary and raises
stimulates economic activity by spreading welfare relative to laissez-faire.
A different modeling approach when realism: much activity in our economic lives
A = [0, oo) tries to harness the distribu is fairly centralized (it is easy to trade, prices
tion F(a) somehow. A method due to Shi are taken parametrically, etc.), as might be
(1997a) assumes a continuum of households, approximated by classical GE theory, but
each with a continuum of members, to get there is also much that is decentralized (it is
a degenerate distribution across households. not easy to find counterparties, to use credit,
The decision-making units are families, etc.), as in search theory. While one might
whose members search randomly, as above, imagine different ways to combine markets
but at the end of each trading round they with and without frictions, the Lagos-Wright
return home and share the proceeds. By setup divides each period into two subperi
a law of large numbers, each family starts ods: in the first, agents interact in a decen
the next trading round with the same a. tralized market, or DM; in the second, they
The approach is discussed at length by Shi interact in a frictionless centralized market,
(2006).31 Another method due to Menzio, or CM. Sometimes the subperiods are called
Shi, and Sun (2013) uses directed search and "day" and "night" as a mnemonic device, but
free entry, so that while there is a distribu we avoid this, since other people seem to dis
tion F(a), the market segments into submar like it. More substantively, we maintain the
kets in such a way that agents do not need frictions that make money essential.33
to know F(a). The useful feature is called We now go through the details of this
block recursivity. Unfortunately, analytical model, which has become a workhorse in the
tractability is lost if there is money growth literature. DM consumption is still q, while
implemented with lump sum transfers, since CM consumption is a different good x.34 For
the real value of the transfer is proportional
to aggregate real balances, in which case
the model needs to be solved numerically. the rest of macroeconomics." Or, as Kiyotaki and Moore
However, the version in Sun (2011) is tracta
(2001) put it, "The matching models are without doubt
ingenious and beautiful. But it is quite hard to integrate
ble even with money growth. them with the rest of macroeconomic theory—not least
Another method due to Lagos and Wright because they jettison the basic tool of our trade, competi
(2005) delivers tractability by combining tive markets." The Lagos-Wright setup brings some com
petitive markets back on board, to continue the nautical
search-based models like those presented metaphor, but also maintains frictions from the search
above with frictionless models. One advan based approach—one doesn't want to go too far, since, as
tage is that this reduces the gap between Hellwig (1993) put it, the "main obstacle" in developing a
framework for studying monetary systems is "our habit of
monetary theory with microfoundations
thinking in terms of frictionless, organized, i.e. Walrasian
and mainstream macro.32 Another virtue is
markets."
33 Lagos and Wright (2007) argue that the presence of
the CM does not imply agents can communicate DM devi
31 In addition to work cited below, see Head and Kumar
ations to the population at large, even if one may get that
(2005), Head, Kumar, and Lapham (2010), Head and Shiimpression from Aliprantis, Camera, and Puzzello (2006,
(2003), Kumar (2008), Peterson and Shi (2004) and Shi
2007b). To clarify, in their model, agents can communicate
(2001, 2005, 2008, 2014). Rauch (2000), Berentsen and
DM behavior when they reconvene in the CM, which is
Rocheteau (2003b), and Zhu (2008) discuss technicallegitimate, but so is the assumption that agents cannot do
issues in these models. so—i.e., spatial separation and limited communication are
different frictions, as should be clear from the first- and
32That the gap was big is clear from Azariadis (1993):
"Capturing the transactions motive for holding money second-generation models. Moreover, even in their setup,
balances in a compact and logically appealing manner hasAliprantis, Camera, and Puzzello (2007a) show how to
ensure money is essential. For more on this see Araujo et
turned out to be an enormously complicated task. Logically
coherent models such as those proposed by Diamond al. (2012) and Wiseman (2015).
(1984) and Kiyotaki and Wright (1989) tend to be so34 It is easy to make x a vector without changing the
removed from neoclassical growth theory as to seriously results. Also, Julien et al. (2015) study the model where q
hinder the job of integrating rigorous monetary theory with
is indivisible, which delivers some interesting results.
now, x is produced one-for-one using labor t, linear with slope (p + p; and (3) a is inde
so the CM real wage is 1. In the DM, agents pendent of wealth. By (3) we get history inde
can be buyers or sellers depending on who pendence (a is orthogonal to a), and hence
they meet. In the former case period utility F(a) is degenerate when there is a unique
is U(x, 1 — £) +u(cj\ and in the latter case maximizer a. In most applications there is a
it is U{x, 1 — £) — c{cj), where U(-) is mono unique such a. In some, like Galenianos and
tone and concave while u(-) and c(-) are as Kircher (2008) or Dutu, Julien, and King
in section 4. To ease the presentation, we (2012), F is nondegenerate for endogenous
begin with U(x,l — £) — U (x) — I and dis reasons, but the analysis is still tractable
cuss alternatives later. due to history independence. Similarly, with
The DM value function V( •) is like (20) exogenous heterogeneity (see below), F is
with one change: wherever /3V+1( ■) appears only degenerate after conditioning on type,
on the RHS, replace it with W( ■), since but that is enough for tractability.
before going to the next DM agents now visit It is important to know these results actu
the CM, where W( •) is the value function. ally hold for a larger class of specifications.
It satisfies Instead of quasi-linear utility, we can assume
U(x, 1 — i) is homogeneous of degree 1, as a
W(a) = max{ U(x) — £ + f3V+1 (a)} special case of Wong (2016), and hence we
x,£, a
can use common preferences like x7(l — £)7
subject to or [x7 + (1 — £)7]1/7. Alternatively, as shown
in Rocheteau et al. (2008), the main results
x = (j>(a — a) + pa + £ + T,
all go through with any monotone and con
cave U(x, 1 — i) if we assume indivisible
where a and a are asset holdings when trad
labor, I £ {0,1}, and use employment lotter
ing opens and closes, <p is the price of a in ies as in Hansen (1985) or Rogerson (1988).
terms of x, p is a dividend, and T is a trans
Any of these assumptions, U quasi-linear
fer of new money. Note that we allow T < 0,
or homogeneous of degree 1, or £ £ {0,1},
so lump-sum taxes can be used to contract
imply a is independent of a. There is one
the money supply. This was not possible in caveat: we need interior solutions at least in
Molico (2006), e.g., because an individual's a
some periods.
may be too low to pay the tax, but now they We now move to the DM, characterized
can use CM labor £. Also, nothing changes
by search and bargaining. Here it makes a
except £ if, instead of taxes or transfers, gov
difference whether a is real or fiat money.
ernment adjusts the money supply by trading
One reason is that in a stationary equilib
CM goods. rium the constraint d < aB, which again
There are constraints x > 0, a > 0, and
says agents cannot turn over more than they
£ 6 [0,1] that we ignore for now. Then, elim
have, binds with fiat money, but it need not
inating £, we have bind with real assets. For now, let p — 0, so
that d = aB. Then consider generalized Nash
(22) W(a) = {(/) + p)a + T + max{[/(x) — x} bargaining,
+ max{-(j)d + (3V+i(a)}.
a
(23) max[u(q) - (f>aB]e \4>aB -c{q)\l~e.
Several results are immediate: (1) x=x* The simplicity of (23) is due to W\a) = 4>
is pinned down by U'(x*) = 1; (2) W(a) isbeing independent of aB or as (which, by the
way, means agents do not need to observe Here we keep the mechanism v(q)
each others a to know their marginal val general, imposing only monotonicity plus
uations). Indeed, (23) is the same as Nash this condition: cpaB>v(q*) implies a
bargaining in section 4, except the value buyer pays d = v(q*)/(p and gets q*; and
of the monetary payment is <fiaB instead <paB < v(q*) implies he pays d = aB and gets
of A. Hence, the outcome is (f>aB = v(q) q = u_1((/>aB). One can show this holds
instead of A = v(q), but v(q) is still given automatically for standard bargaining solu
by (Is). tions, competitive pricing, and many other
We can now easily accomodate a variety of mechanisms, and indeed it can be derived
alternative solution concepts. For Kalai bar as an outcome, rather than an assumption,
gaining, simply use v(q) from (17) instead of given some simple axioms (Gu and Wright
(18). Aruoba, Rocheteau, and Waller (2007) 2016).
advocate the use of Kalai in this model, For any such v(q), the DM value function
while recognizing that it has some peculiari satisfies
ties (e.g., interpersonal utility comparisons),
for the following reasons: (1) it makes buy (24) V(a) = W(a) + aa{u[q(a) ]— (pa}
ers' surplus increasing in a; (2) it does not
give an incentive to hide assets; (3) it makes + aa J {(pa - c[q(a)]}dF(a),
V(fl) concave; and (4) it is easy. These results
are not always true with Nash bargaining, where a is money held by an individual while
although Lagos and Rocheteau (2008b) pro a is held by others, which we allow to be ran
vide a fix. In any case, it seems best to be dom at this stage, although in equilibrium
agnostic and allow different bargaining solu a = A. Then one can derive
tions for different applications. We can also
use Walrasian pricing by letting v(q) = Pq, V'(a) = 4> + eta\u'[q(a)]q'(a) - 4>}
where P is the price of DM goods in terms
of x that buyers take parametrically, even f u'\q(a)1 1
though P = c'(cj) in equilibrium. To motivate = 4>< 1 + aa [ \ - aa V
I v'ma)\ J
this, Rocheteau and Wright (2005) describe
DM meetings in terms of large goups, as using q'(a) = (p/v'{q), which fol
opposed to bilateral trade—think about the (pa = v(q). We now insert this in
Lucas-Prescott (1974) search model, as from the previous CM, </;_] = f3V
opposed to Mortensen and Pissarides (1994). the —1 subscript indicates last p
Or, following Silva (2015) (see also Laing, Li, result is the Euler equation
and Wang 2007, 2013), we can use monopo
listic competition.35 (25) 4>^i = + aa\(q)],
where X(q) = u'((j)/v'(cj) — 1 is the liquid
35 One might say that using perfectly or monopolistically
ity premium,
competitive markets means giving up to some extent on or equivalently, the Lagrange
multiplier
microfoundations, compared to search-based on the constraint in the problem
models that
go into more detail with respect to the meeting/trading
max
{u(q) — £>(<7)} subject to v(q) < a(p.
process, but obviously it can still be prudent to use these
standard solution concepts in applications. Indeed, one can
make the DM look even more like a standard competitive
market by setting aa = 1, to avoid search andfor matching
unforeseen opportunities of advantage." Also
frictions. However, aa < 1 nicely captures the precaution
avoids a problem in many models where velocity
ary demand for liquidity, described by Keynes 1 (Lagos and
(1936) in Rocheteau 2005; Telyukova and Vi
2013).
terms of "contingencies requiring sudden expenditure and
allocations obtained from any mecha That's the basic model. Many assump
nism that is individually rational and pair tions can be relaxed, but quasi-linearity, or
wise efficient. A mechanism is a mapping one of the other options mentioned above,
from the asset holdings in a buyer-seller is needed for history independence. The
DM meeting, (aB,as), into a trade, (q,cl). framework is of course well-posed without
Due to lack of monitoring or record keep that, but then it requires numerical methods.
ing, the mechanism cannot use individ By analogy, while heterogeneity and incom
ual trading histories, and cannot induce plete markets are worth studying computa
agents to truthfully reveal past defections tionally in mainstream macro, it is useful to
from proposed play except through money have standard growth theory as a benchmark
balances.37 With a constant money sup to analyze existence, uniqueness versus
ply, one can show that (q, 0) is incentive multiplicity, dynamics, and efficiency The
feasible if q<q*, c(q) < (pA < u(q) and framework just presented is our benchmark
tw\u{q) — c(q)] >r<j>A, where the last for monetary economics.
condition says the expected DM surplus is
worth the cost of accumulating real balances.
7. Extending the Benchmark
It follows immediately that the efficient q *
is implementable when Some versions of our baseline environment
have the CM and DM open simultaneously
(28) r<aa[u(q*)-c(q*)\/c(q*). with agents transiting randomly between
them (Williamson 2006); others have multi
Notice (28) is identical to (11) from the ple rounds of DM trade between CM meet
second-generation model, except for the ings or vice-versa (Berentsen, Camera, and
term 1 — A, since now the asset distribution Waller 2005; Telyukova and Wright 2008;
is degenerate. Hence, if agents are patient, Ennis 2009, Jiang and Shao 2014); and others
q* can be implemented without deflation. use continuous time (Craig and Rocheteau
Also, thanks to quasi-linearity, mechanism 2008b; Rocheteau and Rodriguez-Lopez
design is especially tractable. Thus, Gu and 2014). An extension in Lagos and Rocheteau
Wright (2016) show that we can restrict atten (2005) and Rocheteau and Wright (2005), on
tion to a mechanism v[q) that is very simple which we spend more time, has two distinct
(linear except when IC conditions in the DM types, called buyers and sellers because in
would otherwise bind). See Wong (2015) for every DM the former want to consume but
an analysis without quasi-linearity. In other cannot produce while the latter produce but
applications, Rocheteau (2012) studies the do not consume. This provides a natural set
cost of inflation, Hu and Rocheteau (2013, ting to consider price posting and directed
2015) and Araujo and Hu (2015) analyze the search, sometimes called competitive search
coexistence of money and credit or other equilibrium.
assets, while Chiu and Wong (2015) investi One version of this has market makers set
gate alternative payment systems. up submarkets in the DM to attract buyers
and sellers, and charge them entrance fees,
although due to free entry into market mak
37The CM is kept the same, as can be justified by ing the equilibrium fee is 0. In general, buy
the equivalence between competitive and core alloca ers and sellers meet in the DM according to
tions. More generally, for a discussion of subtleties in a standard matching technology, where a(n)
this approach, see Rocheteau (2012), and for additional
motivation of the mechanism design approach, see Wallace
is the probability a buyer meets a seller and
(2010). a(n)/n is the probability a seller meets a
buyer, with n now denoting the seller/buyer surplus between buyers and sellers. In other
ratio. Notice a = 1 here, given that directed applications it is assumed that sellers have
search avoids matching problems (but see a cost k to enter the DM. Then Ss — k,
Dong 2011). A submarket involves posting and (30)-(31) determine q and n jointly.
([q,z,n) in the CM, describing the next DM In either case, l — 0 implies q — q*, and
by the terms of trade—agents commit to given this, with endogenous entry, it implies
swapping q units of output lor z = (pa real a'(n*)[u(q*)-c(q*)] = /c.39
balances if they meet—as well as n, which Lagos and Roeheteau (2005) fix n but
they use to compute the probability of endogenize a through buyers' search inten
meeting. Market makers design (q,z,n) to sity. With random search and bargaining, the
maximize buyers' surplus subject to sellers time it takes buyers to spend their money
getting a minimal surplus of Ss.38 increases with 7r, counter to the well-known
Algebra reduces the market maker prob hot-potato effect of inflation (see, e.g.,
lem to Keynes 1924, p. 51). This is because n, as a
tax on money, reduces the DM surplus and
(29) max ja(n) [u(q) — z] — iz} hence search effort. But with price posting,
although 7r lowers the total surplus, they
subject to show it raises buyers' share at low tt. Hence,
when 7r increases buyers spend money faster
a(n) [z -c(q)] =n~Ss. by increasing search effort. Alternatively, in
Liu, Wang, and Wright (2011) buyers spend
Eliminating z and taking the FOCs, we get their money faster at higher n, even with
random search and bargaining, because they
(30) a(n)u'(q) = [a(n) + L]c'(q), (rather than sellers) have a DM entry deci
sion, and higher 7T increases n. Relatedly, in
Nosal (2011) buyers spend money faster at
(31) a'(n) [u(q) - c(q)\ = Ss higher n with random search because higher
7r reduces their reservation trade, endogeniz
where e = na'(n)/a(n) is the elasticity of ing <x, as Kiyotaki and Wright (1991) did in
matching with respect to participation by a first-generation model. See also Victor E.
sellers. In some applications, it is assumed Li (1995), Shi (1997b), Jafarey and Masters
that all agents participate in the DM for free, (2003), Wang and Shi (2006), Faig and Jerez
so that n is given. Then (30) determines q, (2007), Ennis (2009), and Hu, Zhang, and
while (31) determines the split of the total Komai (2014). These exercises are germane
because they concern duration analysis, for
38The market maker story comes from Moen (1997).
which search theory is well suited, and con
We can instead let sellers post the terms of trade to attract stitute models of velocity based on explicit
buyers, or vice-versa. Often these are equivalent, but not search, entry or trading decisions.
always—Faig and Huangfu (2007) provide an example
where they are not equivalent precisely because trade is
monetary. Other models with money and competitive
search include Faig and Jerez (2006), Huangfu (2009), 39 One can show l — 0 achieves efficiency on both the
Dong (2011), Dutu, Huangfu, and Julien (2011), Bethune, intensive and extensive margins, (q *, n *). In contrast, Kalai
Choi, and Wright (2014), and Choi (2015). Also, note that bargaining with DM entry by sellers yields n* if and only
while some directed search theory takes the matching if 1 — 0 = a'(n*)n*/a(n*), the Hosios condition saying
technology as a primitive, this is not always the case—e.g., that sellers' shares should equal the elasticity of match
see Lagos (2000) or Burdett, Shi, and Wright (2001) for ing with respect to their participation. Hence, one might
models where it is endogenous. Whether this is an issue say that competitive search delivers the Hosios condition
depends on the application. endogenously.
Rocheteau, Weill, and Wong (2015) ana nonbuyers can earn interest on balances that
lyze a version of the model with a nondegen would otherwise lay idle, with interest pay
erate distribution of a that is still tractable. ments covered by borrowers. Hence, banks
Suppose the constraint on labor, i < 1 binds. reallocate liquidity towards those that could
It then takes several periods for a buyer to use more, similar to Diamond and Dybvig
accumulate desired real balances. There (1983), although here bankers realistically
are some equilibria where, in a DM match take deposits and make loans in cash rather
with a seller, the buyer spends the money than he goods.
accumulated so far (there can also be equi In some applications, bankers can abscond
libria with partial depletion that need towith be deposits or borrowers can renege on
analyzed numerically). In these equilibria, debt (Berentsen, Camera, and Waller 2007).
F[a) is a truncated geometric distribution,
Others study the role of banks in investment
and the value and policy functions can and
be growth (Chiu and Meh 2011; Chiu, Meh,
solved in closed form. The model is trac and Wright 2017). Related work includes
table thanks to ex ante distinct buyer and Li (2006, 2011), He, Huang, and Wright
seller types because, since sellers hold a —(2008),
0 Bencivenga and Camera (2011), and
, there is only one-sided heterogeneity acrossLi and Li (2013). In some of these, money
matches. Still, equilibrium features a nonde and banking are complements, since a bank
generate distribution of prices, as in Molicois where one goes to get cash; in others, they
(2006). Under some conditions, inflation can are substitutes, since currency and bank lia
raise output and welfare. Moreover, one can bilities are alternative payment instruments,
use this model to analytically characterize allowing one to discuss not only currency but
transitional dynamics following a monetary checks or debit cards. Ferraris and Watanabe
(2008, 2011, 2012) have versions with invest
injection, and show that price adjustments are
sluggish in the sense that aggregate real balment in capital or housing used to secure cash
ances increase following the money injection.loans from banks; in these models investment
Rather than permanently distinct types, can be too high and endogenous dynamics
it is equivalent to have type realized each
can emerge. Williamson (2012, 2014, 2016)
period if the realization occurs before the
has models where banks design contracts like
CM closes—(27) still holds, with agents Diamond-Dybvig, incorporating multiple
conditioning on all the information they assets and private information. His banks hold
have when they choose a. Related to this,diversified portfolios that allow depositors to
Berentsen, Camera, and Waller (2007) share the benefits of long-term investments
introduce banking. After the CM closes, but
with less exposure to liquidity shocks. Ennis
(2014) extends that model and uses it to dis
before the DM opens, agents realize if they
will be buyers or not, at which point banks
cuss some other contemporary policy issues.40
are open where they can deposit or with
draw money. Or, to make this look more
40 Gu et al. (2013a) propose a related theory of banking
like standard deposit banking, let everyonebased on commitment. Agents have various combinations
put money in the bank in the CM and then, of attributes affecting the tightness of their ICs, and hence
when types are revealed, buyers will withdetermining who should be bankers. Bankers accept and
invest deposits, and can be essential (see also Araujo and
draw while others leave their deposits alone.
Minetti 2011). Monitoring can also be endogenized as in
A bank contract involves an option to conDiamond (1984) (see Huang 2015). Moreover, as in some
vert on demand interest-bearing deposits
of the settings discussed above, here bank liabilities, claims
on deposits, serve to facilitate transactions with third
into cash, or transfer claims on deposits toparties—i.e.,
a serve as inside money (see also Donaldson,
third party. This enhances efficiency becausePiacentino, and Thakor 2015 and Sanches 2015). This is
As regards labor markets and monetary convening between the CM and DM.43 Since
policy, theory is flexible. If c^(q, k) = 0, labor is allocated in the LM, they make CM
the model just presented dichotomizes, with utility linear in x, as in typical Mortensen
q solving (33), and (k,x,£) solving (34)-(36). Pissarides models. Then
This means dq/dit < 0, but dk / dir = dx / dir
= dl/dir = 0. The idea of putting k in c(q,k) We (a) = max{x + /3Le(a)}
x,a
was precisely to break the dichotomy. It
leads to d£ /dir < 0. An alternative is to subject to
interact q with x in utility, say U[q, x) — £.
Rocheteau, Rupert, and Wright (2007) x = (f>[a — a) + ew\ + (1 — e)wo + T,
and Dong (2011) do this in models with
indivisible labor, where 1 — £ is genuine where Le (a) is the LM value function
unemployment. They still get dq / dir < 0, indexed by employment status e G {0,1}, wx
but now, U12 > 0 =» dx/d-n < 0, and given i is employment income, and w0 is unemploy
is used to make x, this implies 81/ dir < 0; ment income. They assume W\ is determined
symmetrically, t/12 < 0 01 / dn > 0. So by bargaining in the LM (but paid in the
the Phillips curve can slope up or down, CM, with no loss in generality). Generalizing
depending on [/12. Notice this is a fully the benchmark results, a is now independent
exploitable long-run trade-off. Even without of a and e. Unemployed agents in the LM
nominal rigidities, this can rationalize tradi find jobs at a rate determined by a matching
tional Keynesian prescriptions: if Ul2 < 0, function taking unemployment and vacan
permanently reducing 1 — I by increasing ir cies as inputs, as is standard.
is feasible. But it is not desirable, as absent More novel is the assumption that
other distortions t = 0 is still optimal. This worker-firm pairs generate output, but do
illustrates a key point. One reason it is worth not consume what they produce. Instead,
starting with fundamentals like preferences, all households and a measure n of firms par
rather than taking as a structural relationship ticipate in the DM, where n is the employ
the Phillips curve and adopting some loss ment rate and hence also the measure of
function over (7x,i), is that it allows us to use firms with workers. The DM arrival rate is
properties of U to evaluate the mechanics determined by a matching function, taking
and the merits of policy. While an exploitable as inputs the measures of buyers and sellers.
Phillips curve relationship may exist, exploit This establishes one link between the LM
ing it might be a bad idea. and DM: higher n allows buyers to trade
A different approach to inflation and more frequently, endogenizing the need for
employment is developed by Berentsen, liquidity and thus leading to higher q. A sec
Menzio, and Wright (2011), who add a ond link is that higher q leads to higher real
Pissarides (2000) frictional labor market to revenue for firms, as long as they have some
our benchmark model, lets call it the LM,
43 See also Liu (2010), Lehmann (2012), Gomis
Porqueras, Julien, and Wang (2013), Petrosky-Nadeau
and Rocheteau (2013), Rocheteau and Rodriguez-Lopez
lower-frequency observations, where it does fairly well.
Aruoba, Davis, and Wright (2016) and He, Wright, and (2013), Dong and Xiao (2013), Bethune, Rocheteau, and
Zhu (2015) are versions built to study housing, whileRupert (2015), and Gu, Mattesini, and Wright (2016).
Independent of monetary applications (one can always
Aruoba and Schorfheide (2011) is one with sticky prices.
assume perfect credit), this application of our benchmark
Quantitative work incorporating capital is also performed
in the model of Shi (1997a) by Shi (1999a, 1999b), Wangmodel provides a neat way to put Mortensen-Pissarides
and Shi (2006), and Menner (2006), and in the modelinto of general equilibrium. See Shi (1998) for an alternative
Molico (2006) by Molico and Zhang (2005). approach.
DM bargaining power, which leads to more following Lucas (1980), but Haug and King
vacancies and hiring. By taxing real balances, (2014) use more advanced time-series meth
inflation lowers q, and this lowers n. In other ods and deliver similar results. Because this
words, unemployment and inflation are posi may not be well known, we extract figure 7
tively related in the model (although that can from their findings, showing filtered infla
be altered by changes in assumptions). tion and unemployment with a phase shift
Unemployment is shown in figure 6 from of 13 quarters, as dictated by fit. We agree
US data, 1955 to 2005, and from the cali with their conclusion that, in terms of the
bration in Berentsen, Menzio, and Wright old monetarist notion of "long and variable
(2011) generated by inputting actual values lags" between policies and outcomes, the lag
of l, and counterfactually assuming produc between inflation and unemployment may
tivity, taxes, and demographics are constant. be long but it is in fact not that variable.44
While it cannot match the 60s, the theory
accounts for much of the broad movement
in unemployment over a half century with 44 These findings are related to old monetarism in sev
eral ways. Friedman (1968) says there can be a trade-off
monetary policy as the only driving force.between inflation and unemployment in the short run, but
It is important to emphasize that in filteredin the longer run the latter gravitates to its natural rate.
data, unemployment comoves with both nFriedman (1977) says "This natural rate will tend to be
attained when expectations are on average realized ... con
and l, with the latter probably more relevant,sistent with any absolute level of prices or of price change,
since what matters is expected inflation andprovided allowance is made for the effect of price change
for that i may be a better proxy than conon the real cost of holding money balances." This is the
effect operative in the model. The data in Friedman (1977)
temporaneous 7T. Berentsen, Menzio, and also suggested to him an upward-sloping Phillips curve, if
Wright (2011) argue this using scatter plots,not as strongly as figure 7.
Inflation component
Unemployment component shifted 13 quarters forwards
The framework can also explain sticky for all p G V, because low-p sellers earn less
price observations, defined as some firms per unit but make it up on the volume. This
leaving their nominal prices the same when is because buyers that see multiple prices
the aggregate price level increases. In Head choose the lowest p, obviously. Given infla
et al. (2012), the DM uses price posting a tion, sellers that stick to p are letting their
la Burdett and Judd (1983).45 Each seller real prices fall, but their profits do not fall
sets p taking as given the distribution of because volume increases. Of course, in the
other sellers' prices, say G(p), and the longer run they must change, because with
behavior of buyers. The law of one price fails persistent inflation p ^ V eventually.
because some buyers see one but others see Although G(p) is unique, theory does not
more than one p at a time. As is standard pin down which seller sets which p. Headet al.
in Burdett—Judd models, this delivers an (2012) introduce a tie-breaking rule: when
equilibrium distribution G(p) on a nonde indifferent to change, sellers stick to p with
generate interval V = [p,p]. Heuristically, some probability. One can choose this prob
G(p) is constructed so that profit is the same ability to match average duration between p
changes in, e.g., Klenow and Kiyvtsov (2008)
data (sellers end up sticking with probability
45 See Wang (2016), Liu, Wang, and Wright (2014), 0.9). Then there is a unique distribution of p
Burdett, Trejos, and Wright (2016), and Burdett and
Menzio (2014) for related work. Earlier, Head and Kumar changes. With other parameters calibrated
(2005) and Head, Kumar, and Lapham (2010) put Burdett to other moments, the results are shown for
Judd pricing in the Shi (1997b) model, but did not discuss the model and data in figure 8. From this it
nominal rigidities. The idea here is more closely related
to Caplin and Spulber (1987) and Eden (1994), but the is apparent there is no puzzle in the price
microfoundations are different. change data, since it is just what simple
0.3
■ Model
0.25
0.1
0.05
A similar
monetary search models predict. There point is
concerns studies of opti
also no definitive information about Calvo
mal monetary and fiscal policy by Aruoba
arrival rates or Mankiw menu costs in these and Chugh (2010) and Gomis-Porqueras
data, since the model makes no use of such and Peralta-Alva (2010). Their results over
devices. This finding should influence the turn conventional wisdom from the reduced
way economists interpret price change data: form literature. One manifestation is that
even if prices look sticky, one cannot con MUF models, e.g., imply l = 0 is optimal
clude that central banks should follow policy even when other taxes are distortionary
recommendations based on an alleged cost or under the assumption that utility over goods
difficulty of changing prices. Again, explicitly and money is homothetic (Chari, Christiano,
modeling the exchange process, and in this and Kehoe 1991). But Aruoba and Chughs
case asking why prices are sticky, can matter (2010) analysis shows that making utility
a lot.46 homothetic over goods does not imply the
value function is homothetic. Again, the
46 The model is simultaneously consistent with sev
eral facts that are challenging for other approaches: real
istic duration, large average price changes, many small
changes, many negative changes, a decreasing hazard, and observations on money demand and the micro observa
behavior that varies with inflation. However, the version tions payment methods. Liu, Wang, and Wright (2014) fix
in Head et al. (2012) cannot match these plus the macro this by introducing costly credit.
"'[</(*)] ~ c'[</(z)] |
»' [<?(z)l J
references). By contrast, in models along the To quantify the effects, a typical str
lines of Lagos and Wright (2005), with Nash egy is this: Take U(x) = log(x), u(
bargaining and 0 calibrated to match retail — rt/1_7/(l — 7), and c(q)=q. Set /3
markups, eliminating 10 percent inflation is match an average real interest rate. Since
worth around 5.0 percent of consumption— is not so easy to identify, set aa =1/2 an
an order of magnitude higher. later check robustness (it does not matt
To explain this, as in Craig and much over a reasonable range). Then set 6
Rocheteau (2008a), figure 9 generalizes the match a markup of 30 percent.49 Finally, se
welfare-triangle analysis of Bailey (1956). the parameters F and 7 to match the emp
With Kalai bargaining, write (27) as ical money demand curve, i.e., the relati
ship between 1 and real balances scaled b
income. A typical fit is shown in figure
j'u'[q(z)\ 1 )
i(z) = aa - 1 f, drawn using US data on Ml (it might be b
{v'[q(z)\ J ter to use MIS from Cynamon, Dutkowsk
and Jones 2006 or M1J from Lucas a
which can be interpreted as money demandNicolini 2013, but the goal here is mainl
with z = 4>a denoting real balances. As to discuss
i —> 0, the method). This delivers resu
z —> z0 = Oc(q*) + (1 — 9)u(q*), andclose to Lagos and Wrights 5.0 percen
there
is an upper bound T = aa6/( 1 — 9) above
Note that the calibration does not imply t
which z = 0. If 0=1, then Zo —DM c(q*)
accounts for a large fraction of to
and T = oo, and the welfare cost of output—it
going is only about 10 percent. N
from l = 0 to 11 > 0 is the area under
alsothe
that as we change the frequency fr
curve, ABC, because annual to quarterly or monthly, the relevan
values of aa and 8 change, but as long as
are below the bound aa = 1, the results a
£° Uz) dz = aa{u [q( z0)] - c [q( z0)]}
approximately the same—a big advanta
of search-based models over those where
- Oi<r{u[q(zi)\ -c[9(z1)]}.
agents spend all their cash each period.
In an extension with private informa
But if 9 < 1, the cost of inflation no longer
tion, Ennis (2008) finds even larger effects,
between 6 percent and 7 percent. Dong
coincides with this area, as buyers receive
(2010) allows inflation to affect the vari
only a fraction of the increase in surplus
coming from z. The true cost of inflation
ety of goods, and hence exchange on the
is the area ADC, not ABC, because ABC and intensive margins. She gets
extensive
ignores the surplus of the seller, related
numbersto between 5 percent and 8 percent
the holdup problems discussed above.48
with bargaining. Wang (2016) uses Burdett
Judd pricing to get price dispersion, so that
inflation affects both the average and the
variability of prices. He gets 3 percent in a
48 Some mechanisms, including Kalai bargaining or
Walrasian pricing, deliver the first best at i — 0. Hence,
by the envelope theorem, a small i > 0 has only a second 49 Lagos and Wright (2005) targeted a markup of
order effect. Nash bargaining with 6 < 1, however, does 10 percent, but the results are similar, since 10 percent
not deliver the first best at b = 0, so the envelope theo already makes 0 small enough to matter. Our preferred
rem does not apply, and even a small t > 0 has a first-order 30 percent is based on the Annual Retail Trade Survey dis
impact. This is an additional effect from Nash, but even cussed in Faig and Jerez (2005) (although Bethune, Choi,
with Kalai bargaining, the cost of inflation is greater than it and Wright 2014 say the data imply 39 percent, the differ
is with Walrasian pricing because of the holdup problem. ence does not matter that much for our purposes).
Mi/py
calibration trying to match dispersion, and Dutu (2006) and Boel and Camera (2011)
7 percent when he matches a 30 percent consider other countries.
markup. In a model with capital, Aruoba, Faig and Jerez (2006), Rocheteau and
Waller, and Wright (2011) get around 3 per Wright (2009), and Dong (2010) use compet
cent across steady states, although some of itive search, and find costs closer to 1 percent
that is lost during transitions. Faig and Li or 1.5 percent, since this avoids holdup prob
(2009) add a signal extraction problem and lems. Relatedly, with capital, in Aruoba,
decompose the cost into components due Waller, and Wright (2011) switching from bar
to anticipated and unanticipated inflation; gaining to Walrasian pricing brings the cost
they find the former is far more important. down, even though the effect of n on invest
Aruoba, Davis, and Wright (2016) add home ment is much bigger with Walrasian pricing.
production, which increases the cost of infla Although the choice of mechanisms matters
tion the same way it magnifies the effects of, for welfare results, models with bargain
e.g., taxation in nonmonetary models. Boel ing, price taking, and posting can all match
and Camera (2009) consider the distribu the money demand data. Rocheteau (2012)
tional impact in a model with heterogeneity. also shows a socially efficient mechanism,
as in Hu, Kennan, and Wallace (2009), can effect, optimal 7r is around 3.5 percent (obvi
match money demand, but it implies the ously a second-best result). Bethune, Choi,
welfare cost of a low inflation is nil. Bethune, and Wright (2014) can get i > 0 optimal
Choi, and Wright (2014) have informed and because equilibrium tends to have too many
uninformed agents, as in Lester (2011), and sellers catering to the uninformed, and since
combine directed search with posting and such sellers use more cash, inflation reduces
random search with bargaining. They also their number. Molico (2006), with a non
introduce the option to use costly credit and degenerate F(a), has examples of a positive
show how all of these ingredients matter redistributive effect, although in Chiu and
quantitatively. All of these results underscore Molico's (2010, 2011) calibrated models, this
the importance of understanding the micro effect reduces the cost of inflation but 7r < 0
foundations of information and price forma is still optimal. The same is true of Dressier
tion in decentralized markets. (2011a, 2011b), although his majority-voting
A more radical extension is Aruoba and equilibrium has t > 0. Based on all this, while
Schorfheide (2011), who estimate a model understanding the effects of inflation is an
integrating new monetarist and Keynesianongoing project, progress to date provides lit
features. They compare the importance oftle support for the dogmatic position in some
sticky price distortions, which imply 7r = 0Keynesian research that monetary consider
is optimal, and the effect emphasized here,ations of the type considered here are irrele
which implies l = 0 is optimal. They estimate
vant and can be ignored in policy discussions.
the model under four scenarios, depending
on the DM mechanism and whether they
9. Liquidity in Finance
fit short- or long-run money demand. With
bargaining and short-run demand, despite Assets other than currency also con
large sticky-price distortions, i — 0 is optivey liquidity.50 To begin this discussion, we
mal. The other scenarios even with parameemphasize that assets can facilitate trans
ter uncertainty never imply 7r = 0. Craig and actions in different ways. First, with per
Rocheteau (2008b) reach similar conclu fect credit, there is no such role for assets.
sions in a menu-cost version of our bench Perfect credit means default is not an option.
mark model, as in Benabou (1988, 1992)
But if sellers worry buyers will renege, they
and Diamond (1993), except in Diamondmay insist on getting something, like an
Benabou money is merely a unit of account.
It matters: Diamond (1993) argues that infla 5° Early papers where real assets facilitate trade in
tion usefully erodes the market power of sellour benchmark model include Geromichalos, Licari, and
Suarez-Lledo (2007) and Lagos (2010a, 2010b, 2011), who
ers; but Craig and Rocheteau show that that
have equity a in fixed supply, and Lagos and Rocheteau
is dwarfed by the inefficiency emphasized (2008b), who have capital k. See also Shi (2005),
here for reasonable parameters, making
Rocheteau (2011), Lester, Postlewaite, and Wright (2012),
deflation, not inflation, optimal. Li, Rocheteau, and Weill (2012), Rranch, Rocheteau, and
Petrosky-Nadeau (2016), Nosal and Rocheteau (2013), Hu
However, as mentioned above, sometimes
and Rocheteau (2013), and Venkateswaran and Wright
a little inflation can be good. In Craig and(2013). There are applications using such models to study
Rocheteau (2008b) or Rocheteau and Wright financial issues like the credit-card-debt puzzle (Telyukova
and Wright 2008), on-the-run phenomena (Vayanos and
(2009), with endogenous entry or search
Weill 2008), the equity-premium and risk-free-rate puz
intensity, optimal 7r is around 2 percent. In
zles (Lagos 2008, 2010a), home bias (Geromichalos and
Venkateswaran and Wright (2013), since cap Simonovska 2014), repos (Monnet and Narajabad 2012),
the term structure (Geromichalos, Herrenbrueck, and
ital taxation makes k too low, and this is par
Salyer 2016; Williamson 2016), and housing bubbles (He,
tially offset by a version of the Mundell-Tobin
Wright, and Zhu 2015).
asset, by way of quid pro quo. Or, instead, (2011), rather than using the assets to secure
as in Kiyotaki and Moore (1997), they may his promise, a buyer can hand them over
require assets to serve as collateral that can and finalize the transaction in the DM. It is
be seized to punish default. equivalent for assets to serve as a medium of
To pursue this, first consider our bench exchange or as collateral. One can imagine
mark model with perfect credit. Then exceptions—e.g., if it is "inconvenient" to use
part of your house as a payment instrument,
W(a,D) = max{[/(x) — i + f3V{a)} you may prefer to get a home-equity loan—
but unless that is modeled explicitly, secured
subject to credit a la Kiyotala-Moore is not distinct
from assets serving as media of exchange a la
x = 4>{a — a) + pa + t + T — D, Kiyotaki-Wright.52
An ostensible distinction is that with
where D is debt from the last DM, denom secured credit you can only pledge a frac
inated in numeraire (one-period debt is tion x °f assets, but we can just as well say
imposed without loss of generality). Any you can only hand over a fraction % of your
q can be purchased in the DM if a buyer assets, and again we show how this can
promises to make a payment in the CM of arise endogenously in section 10. While one
D = v(q). For any Pareto efficient mech might tell different stories, the equations are
anism, q = q* and V(a) =W(a, 0)+ the same. Another such distinction is that
a<j[u(q*) —c(q*)]. Since a does not affect the Kiyotaki-Moore literature usually talks
DM trade, the Euler equation is <j)_i about producer credit, not consumer credit,
= /3(p + 4>), the only bounded solution to but in terms of theory, as we said earlier, that
which is the constant solution <f> = <j>* = p/r, is merely a relabeling. A less delusory dis
where <p* is the fundamental price. tinction may be this: In the models presented
Now let debtors default. Also, suppose we above, assets are useful for the acquisition of
cannot take away defaulters' future credit.51 (j. Suppose what you want is not q, but more
Then the only punishment is seizing assets of the same asset, like a producer increasing
pledged as collateral—assuming, say, that business capital or a homeowner increasing
they have been assigned to third parties with housing capital. It will not do to exchange old
commitment. As in standard Kiyotaki-Moore k for new k. But it might be useful to pledge
models, borrowers can pledge only a fraction the old k to get new k on credit if we assume
X < 1 of their assets, with x exogenous (it is the former is pledgeable but the latter is not,
endogenized in section 10). The IC for hon which is perhaps arbitrary but not crazy.
oring obligations is D <(</> + p) xa, as for
In any event, let us include the parame
ter x, and consider assets with p > 0. Now
higher D it is better to forfeit the collateral.
Hence, the Kiyotaki-Moore debt limit we cannot be sure d < x(<P + p)a binds. If it
is D = (0 + p) \a. But as noted in Lagos binds, the analog of (25) is
where X(q) is as before. Using v(q) (j)> <f>* for intermediate values of x> the
— x(0 + p)a to eliminate (f> and c/)_h we claim follows. Hence, as assets become more
get a difference equation in q analogous to pledgeable, there can emerge an endoge
(26). While p > 0 rules out equilibria where nous price boom and bust (He, Wright, and
4> = 0 or cj) —> 0, there can still exist cyclic, Zhu 2015).
chaotic and stochastic equilibria (Lagos and To check if the liquidity constraint
Wright 2003; Zhou 2003; Rocheteau and binds, assume that it does, and use v(q)
Wright 2013). Thus, it is not the fiat nature = x{(/> + p)a to eliminate q from (37). In
of money that generates dynamics, but an stationary equilibrium, the result can be
inherent feature of liquidity, which applies to interpreted as the long-run demand for a
assets serving as a means of payment or col as a function of <j). One can show demand is
lateral, whether they are nominal or real. We decreasing for q < q0, defined by A(c/0) = 0,
also claim that <f» is nonmonotone in %; first, even if X(q) is not monotone (Wright 2010).
c/) = (j)* when x = 0 and when x is so big that Then define A* by v(q0) = x{4>* + p)A*,
the liquidity constraint is slack; then, since where <j>* = p/r, so that a = A* satiates
a buyer in liquidity. The resulting asset budget equation, but as in the benchmark
demand curve is shown in figure 11, which model, T adjusts to satisfy this after a pol
truncates (37) at 4>*. It is now immediate icy change. The nominal returns on illiquid
that A > A* implies liquidity is plentiful and real and nominal bonds are still 1 + r— 1/(3
(f> = <f>*, while A < A* implies liquidity and 1 + t = (1 + tt)/d. The nominal return
is scarce and bears a premium (f> > <p*. on the liquid real bond—i.e., the amount of
We also mention that some applications cash accruing in the next CM from a dollar
use one-period assets instead of long-lived put into these bonds in the current CM—
assets. Then (37) changes to 0_j_ is denoted if,, and generally differs from
= /3p[ 1 + aaLX[q)] where now v(q) = XPa> Suppose in the DM, with probability am a
since the asset pays a dividend p but has no buyer meets a seller that accepts only am,
resale value in the next CM; otherwise the and they trade qm; with probability en/, he
results are similar. meets one that accepts only they trade qy,
The above analysis applies to assets in and with probability a2 he meets one that
fixed supply. Neoclassical capital is simi accepts both, they trade q2. One can also say
lar except the supply curve is horizontal that conditional on being accepted, Xm and
instead of vertical, so liquidity considerations Xb are the fractions of the buyers am and a/,
are manifest not by (j> > <j>* but by k > k*. that can be used in the transaction. While
Housing is similar when it conveys liquidity X,n = Xh — 1 is a fill'- special case, again,
via home-equity lending, except supply need these are endogenized in section 10, and
not be horizontal or vertical. Recent research
it is shown how Xj < 1 can emerge in
(see fn. 50) studies models with different private-information settings where sellers
combinations of these kinds of assets plus are worried about counterfeits.
currency. One application concerns open In any case, buyers always spend all the
market operations, or OMOs, in economies cash they can, but may or may not use all
with money plus a real, one-period, pure-dis their bonds. Suppose they do use all their
count government bond issued in the CM bonds in type-Z? and type-2 meetings, which
and paying a unit of numeraire in the next happens when liquidity is scarce. The Euler
CM. The following discussion encompasses equations are
features in Williamson (2012, 2014, 2016),
Rocheteau and Rodriguez-Lopez (2014), (38) L = amXmMqm) + <*2XmMq2),
and Rocheteau, Wright, and Xiao (2014).53
Suppose these bonds are partially liq
uid—i.e., can be traded in some DM (39) s = ahXbMqb) + <X2XbMq2),
meetings.54 Bonds affect the government
where s = (i — th)/{ 1 + ^b) ^le spread
between illiquid and liquid bond returns.55 As
53 See also Dong and Xiao (2013), Han (2014), and
Ennis (2014). Also, while these papers consider both real
and nominal bonds, we concentrate here solely on the for Bech and Monnet (2016), and Chiu and Monnet (2015).
mer, but most of the results are similar. See also section 11.
54 While it may be unusual for households to use bonds 55 This is related to what Krishnamurthy and Vissing
as payment instruments, it is common for financial firms Jorgensen (2012) call the "convenience yield" and measure
to use them as collateral, and just like we can reinterpret
by the difference between government and corporate bond
agents as producers instead of consumers, many applica yields (see also Nagel 2014). One might say the model here
tions interpret them as banks. See Berentsen and Monnet rationalizes their reduced-form specification with T-bills
(2008), Koeppl, Monnet, and Temzelides (2008, 2012), in the utility function, but its not clear if that's desirable,
Martin and Monnet (2011), Chapman, Chiu, and Molico any more than saying the models presented above ratio
(2011, 2013), Berentsen, Marchesiani, and Waller (2014), nalize money in the utility function. The sine qua non of
Rocheteau, Wright, andXiao (2014) point out, generate a liquidity trap, with low q's, and
Ly < 0 is possible in this model, although we ineffectual OMOs, because changing Ay
still have/, > Oforilliquidbonds. When < 0, crowds out real balances, with total liquidity
agents are willing to hold aj, because its staying put, similar to Wallace (1981, 1983).
expected liquidity benefit exceeds that of At this juncture we offer some general
cash, or because it is more pledgeable, and comments on the coexistence of money and
note that this does not violate no-arbitrage riskless, perfectly recognizable bonds with
as long as bonds issued by private agents in positive returns. Papers trying to address
the CM do circulate in the DM. Whether this coexistence (or rate of return domi
nance) puzzle generally impose some asym
or not it is empirically relevant, this shows
how Lb < 0 is a logical possibility (see also
metry between money and bonds. Aiyagari,
He, Huang, and Wright 2008, Sanches and Wallace, and Wright (1996), e.g., study
Williamson 2010, or Lagos 2013b). a second-generation model with money
and two-period government bonds. As
The effects of ir or i are still equivalent
in Victor E. Li (1994, 1995), Aiyagari and
since, again, l + t = (l + 7t)//3. As long as
Wallace (1997), or Li and Wright (1998),
a2 > 0, higher t raises (!)/, and s as agents
try to shift out of cash and into bonds. Thethere are government agents that act like
effect of i on i./, is nonmonotone: a Fisheranyone else except: in meetings with pri
effect raises the nominal return for a givenvate sellers, they may either pay with cash
real return, but a Mundell effect lowers theor issue a bond; they may refuse to accept
real return as demand for bonds rises, and not-yet-matured bonds from private buyers;
either can dominate. For OMOs, note that and in any meeting they can turn matured
bonds into cash. Equilibria with valued
buying bonds with cash is formally equiva
money and interest-bearing bonds exist
lent to buying them with general tax revenue:
higher Am means higher prices but the samebecause of asymmetry in the way govern
<pmAm. Hence, any real impact comes from ment treats the assets.56
changing Aj. One can show dz / dA}, < 0Such asymmetries are adopted because of a
and dqm / d.Aj, <0 if a2> 0- Intuitively,
belief that, under laissez-faire, absent exoge
higher Aj, makes liquidity less scarce nous
in assumptions that favor money, there are
type-2 meetings, so agents try to economize
no equilibria where it coexists with default
on cash, but this comes back to haunt them free,
in interest-paying, nominal bearer bonds.
type-m meetings. Since an OMO that raisesYet arguably there are episodes where such
Am results in higher z, prices go up by less
securities and money both circulated, a strong
instance of the rate-of-return-dominance
than Artx. One could misinterpret this as slug
gish prices. In fact, z increases because Ay,
puzzle. Lagos (2013b) addresses this in a ver
falls. Similarly, this OMO leads to a reduc
sion of the benchmark model where currency
tion in ih. One could misinterpret this byconsists of notes that are heterogeneous in
extraneous attributes—e.g., serial numbers—
thinking the increase in Am reduces nominal
returns, but again lower ij, actually happens
called moneyspots, to make a connection with
because Ay falls. Hence, it is not so easysunspots.
to These payoff-irrelevant characteris
make inferences about monetary neutrality.tics are enough to get money coexisting with
For some parameterizations, as Williamson
(2012) emphasizes, the model can also
56 A different asymmetry is studied in Zhu and Wallace
(2007) and Nosal and Rocheteau (2013), where the trading
our approach is modeling exchange explicitly, not deriving
protocol gives a larger share of the surplus to agents with
indirect utility functions from primitives. more money.
In particular, (47) says L-type buyers have no applies: agents with L-type assets make the
incentive to offer [qH, dH). complete-information offer while those with
The least-cost separating offer satisfies II-type assets make an offer that others do
(46)-(47) at equality, so dH = c(qH)/pH not want to imitate. When t —» 0, agents hold
(z(0,dL), while qH£[0,qL) solves enough currency to buy q *, and no one uses
~ c^\h)PlIph = u{qL) - c(qL). Thus, the asset in DM trade. For t > 0, they do not
H-type buyers retain a fraction of their hold enough currency to buy q * and spend it
holdings as a way to signal quality, and all plus a fraction of their risky assets. Asset
hence qH < qL (interpretable as over liquidity as measured by this fraction clearly
collate ralization, as in DeMarzo and Duffie depends on monetary policy.60
1999). When the L-type asset is a pure lemon, Asset quality can also depend on hidden
pL = 0, both qL and qH tend to 0.59 In con actions. One rendition allows agents to pro
trast to models with exogenous constraints, duce assets of low quality, or that are outright
agents here turn over all their assets in trade worthless, as when through history coins
with probability 1 — C ar|d a fraction dH /a were clipped or notes counterfeited (Sargent
with probability £, where dH la depends on and Velde 2003; Mihm 2007; Fung and
pL and pH. With a one-period-lived asset, Shao 2011). Li, Rocheteau, and Weill (2012)
while (f>* = /3[CPn + (1 — C ) Pi] is the funda have a fixed supply of one-period-lived
mental value, here we get assets yielding p. At some fixed cost k > 0,
agents can produce counterfeits that yield 0.
Assume counterfeits are confiscated by the
0 = 0* + PaaeiCp"X{qH)pLMqL) government after each round of DM trade,
{PHMqH) + Ph~ Pl
so they do not circulate across periods, sim
ilar to Nosal and Wallace (2007). Then with
the payoff to accumulating counterfeits, cost a cost to become informed. This endoge
k plus DM utility. nizes the fraction of matches where an asset
Given a = d, in equilibrium, (48) can is accepted (Kim 1996 and Berentsen and
be rewritten as an upper bound on the Rocheteau 2004 similarly let agents pay a
amount of asset that can be transferred, as in cost to better recognize the quality of goods).
Kiyotaki-Moore models, To simplify the analysis, Lester, Postlewaite,
and Wright (2012) assume fraudulent assets
(49) d < are worthless and can be produced at no
<p — <f>* + flaaOp'
cost: pL — k = 0. This implies sellers only
accept assets if they recognize them, and so
This endogenous bound increases with
it is possible the
to use standard bargaining the
cost of counterfeiting k, whileory—when sellers recognize assets there is
it decreases
with the cost of holding assetsno<f>
private
— information;
<f>* and when they don't the
the frequency of DM trading opportuni
assets are simply not on the bargaining table.
ties aa. If a a — 1, so there are no search
Interpret a as the or
probability of a single
matching frictions in the DM, coincidence times the
(49) says the probability a seller
assets value cj)d must be less than
can the
discern, andcost
hence accept, a buyer's
assets.
of fraud k. As k —> 0 the asset stopsAt the beginning of the DM, sell
circulat
ing. Notice an increase in theers (endogenous)
choose a at cost C(cr), satisfying the
usual conditions.
price (p tightens the constraint, with policyThe decision of a seller
implications described in Li, to
Rocheteau,
become informed,and so that he can accept
Weill (2012). If (49) binds and
assetsdin<the
A,DM, then
is similar to a decision to
the asset is illiquid at the margin and
enter <f>in
market = the
ft*;
first place or the search
if (49) does not bind the asset is perfectly
intensity decision discussed above. In order
liquid. There is an intermediate
to givecase
themwhere
some incentive to invest, sell
the liquidity premium, k/A ers must—have
(3aa9p,
some bargaining power, so
increases with k and decreases with
consider /3acr9.
Kalai's solution with 6 6 (0,1). The
Notice the threat of fraud can affect asset FOC is C'(cr) = a6[u(q) — c(q)\, where as
prices even if there is no fraud in equilib always v(q) = min{Ap,u(g*)}, with p the
rium, and assets with identical yields can bedividend on genuine assets. This equates
priced differently.61 the marginal cost of becoming informed
Given that sellers may be uninformedto the expected benefit from being able
about the quality of buyers' assets, Lester,to accept assets. As the marginal cost of
Postlewaite, and Wright (2012) let them payinformation decreases, a and <f> increase.
With long-lived asset, Lester, Postlewaite,
and Wright (2012) show there are strategic
complementarities
61 If the asset subject to fraud is fiat money, as in Li and between buyers' asset
Rocheteau (2009), then (49) becomes c(q) < k//3(t demand
+ aad) and sellers' information invest
—another case where acceptability is not exogenous,
ment,
but depends on the policy variable t. Other applications and this can naturally generate mul
include Li and Rocheteau (2011), Li, Rocheteau,tipleand equilibria.
Weill (2012), Shao (2014), Williamson (2014), and Gomis
Suppose there are two short-lived assets,
Porqueras, Kam, and Waller (forthcoming). The last paper
includes two currencies to study exchange rates. In both
gen yielding p. Asset 1 is perfectly recog
nizable
eral, there can be many assets, each with cost of fraud Kj
at no cost in a fraction a of all
and supply A-, with aggregate liquidity a weighted average
meetings while asset 2 requires an ex ante
of Aj with weights depending asset characteristics, consis
investment
tent with the notion of money suggested by, e.g., Friedman C(cj2 ) to be acceptable in a frac
and Schwartz (1963). tion cr2 of meetings. Thus, in an endogenous
fraction a2 of matches buyers can use both models to a more general yet still tractable
assets, while in fraction a1 = a — a2 they framework. The new model captures aspects
can use only asset 1. The investment deci of illiquid markets, like participants adjust
sion satisfies ing positions to reduce trading needs. For
simplicity, suppose all trade goes through
dealers, as in section 5 with aD > 0 = at.
(50) C'( a2) = ad0{ [u(q2) - c{q2)]
Let a € K+, and assume uJa) is the flow
utility of an agent with preference type
- [u(qi)-c(q{)]}.
j € {0,..., I}. Each I draws a preference
The RHS of (50) is the sellers benefit of type j'6{0,...,/} at Poisson rate Up with
being informed, the extra surplus from hav }2i Wi = u>. When I with preference j con
ing a payment capacity of (Aj + A2) p instead tacts D, they bargain over the dj that I takes
of A\p. If there is an increase in the supply out of the meeting, and a payment that
of recognizable assets A1; agents invest less includes D's cost of the transaction qlaj — a),
in information, asset 2 becomes less liq where q is the real asset price in the inter
uid, and 02 falls. If the recognizable asset dealer market, plus a fee <pj(a).
is fiat money, the acceptability of the other The choice of (ij solves maxfl,{uj (a') —
asset is affected by policy. At l = 0, we get rqa'}, where
(ji = <72 — q* and agents stop investing—
reflecting the old idea that the use of money (r + ri)Uj(a) + Y,kUkUk{a)
saves information costs. uAa) = : ,
J r+rj + uj
There is more to be done on information,
but existing results help in understanding and r) = aD(1 — 6D) is
phenomena related to acceptability and adjusted for bargaining po
pledgeability, as well as the coexistence of
assets with different returns, as discussed in 9D{ui(af) —u,(a) — rq
section 9 in the context of money and bonds. (51) <Pj(a) = [jUj [
Information theory is a natural and venerable
notion to bring to bear on Hicks' suggestion is the intermediation fee. Equilibrium
with which we began this essay.62 given by desired asset positions (ait.. . ,aj)
the fee ipj (a), the q that clears the interdeale
market and the distribution (n^) satisfyin
11. Generalized OTC Markets
the usual conditions. To focus on the impli
tions for asset prices and measures of liqu
Papers spurred by Duffie, Garleanu, and
Pedersen (2005) maintain traetability by ity, consider ut(a) =ipiloga, where Vy'i
restricting a £ {0,1}; Lagos and Rocheteau < • ■ • < ipi, and let xp =
(2009) relax this. In spirit if not detail, the
Then the post-trade position of type j wh
idea is to do what section 6 does for mon just met a dealer is
etary theory: extend second-generation
holdings between a?° and a, with the weight in illiquid markets, I trades very little, so aj30
assigned to a*J increasing in i). is close to A, and <f is small. The model also
Hence, frictions concentrate the asset predicts a distribution of transactions, with
distribution. As frictions decrease, dj -* a.j° spreads increasing in the size of a trade, as
and the distribution becomes more disperse, well as varying with I's valuation.64 The model
but aggregate demand is unchanged. A mes can also be extended to allow heterogeneity
sage is that one should not expect to identify across I in terms of arrival rates or bargaining
frictions by looking at prices alone. Trade power, and those with higher 77 trade larger
volume is riy \cij — ai\/2, lower than quantities at a lower cost per unit.
a frictionless economy. It increases with 77, Trading delays are an integral feature of
capturing the idea that large volume charac the microstructure in OTC markets. The
terizes liquid markets, where I can switch in time it takes to execute a trade not only
and out of positions easily. With unrestricted influences volume and spreads, but is often
asset holdings, frictions affect volume used directly as a measure of liquidity.
through the extensive and intensive margins. Lagos and Rocheteau (2007, 2008a) endog
If aD increases, the number of investors who enize aD with entry by D, and derive some
are able to trade rises, but the number who new predictions, including a change in the
are mismatched with their current portfolio equilibrium set due to the nonmonotonic
falls. Also, higher aD shifts the distribution ity mentioned above. The model can gen
across desired and actual holdings in a way erate multiple equilibria: it may be illiquid
that increases volume. This effect is shut because participation by D is low given a
off if one restricts a E {0,1}, which has difbelief that I will only trade small quanti
ferent predictions for trade volume after ties; and it is rational for I to take conser
changes in the microstructure of the market. vative positions given long trading delays.
In Duffle, Garleanu, and Pedersen (2005),Tight spreads are correlated with large vol
with a £ {0,1}, trade volume is indepen ume and short delays across equilibria, and
dent of dealers' market power; here volumescarce liquidity can arise as a self-fulfilling
decreases with 0D.m prophecy. Subsidizing entry can eliminate
Another conventional measure of financial this multiplicity. Even when equilibrium is
liquidity is the bid—ask spread or intermedia unique, the model has novel predictions,
tion fee. As long as a € {0,1}, an increase in like lower market power for dealers promot
aD raises 7s value of search for alternative ing entry and reducing delays by increasing
traders, so spreads narrow. With unrestricted trade size. Similarly, a regulatory reform
asset holdings, spreads still depend on aD, or a technological innovation that gives I
but also on the extent of mismatch between more direct access to the market (e.g., elec
asset positions and valuations. Hence, the tronic communication networks) implies an
relationship between the spread and 77 can increase in market liquidity and intermedi
be nonmonotone: under reasonable condi ated trade.
tions, one can show the spread vanishes asThe model also provides insights on wel
r) —> 0 or as 77 —* 00. In liquid markets I has
fare in illiquid markets. At least when contact
good outside options, and hence ip is small;
rates are exogenous, welfare is unaffected by but learn their valuations infrequently. They
9d, which only affects transfers from I to D. also provide a class of utility functions nest
When a is not restricted to {0,1}, equilib ing Duffie, Garleanu, and Pedersen (2005)
rium is inefficient unless 0D = 0. Indeed, and Lagos and Rocheteau (2009). Lagos and
if D captures any of the gains that I gets Rocheteau (2008a) allow investors to have
from adjusting his portfolio, I economizes both infrequent access to the market, where
on intermediation fees by choosing cij closer they are price takers, and to dealers, where
to a, thus increasing mismatch. When aD is they bargain. Pagnotta and Philippon (2011)
endogenous, the equilibrium is generically study marketplaces competing on speed,
inefficient, again related to Hosios (1990). endogenizing the efficiency of the trading
Efficient entry requires that aD equal the technology, and entry/investment decisions
contribution of dealers to the matching pro from an industrial organization perspective.
cess, but efficiency along the intensive mar Melin (2012) has two types of assets traded
gin requires that aD = 0. As in monetary in different markets, one with search and
models, those two requirements are incom one frictionless. Mattesini and Nosal (2016),
patible and the market is inefficient, although Geromichalos and Herrenbrueck (2016),
as in many other models, a competitive Lagos and Zhang (2013), and Han (2014)
search version can deliver efficiency (Lester, integrate generalized models of OTC markets
Rocheteau, and Weill 2015). Branzoli (2013) with the monetary models in section 6, which
finds that the most effective way to promote would seem an important step in the program.
trading activity is the introduction of weekly One reason it is important is the following:
auctions where investors trade bilaterally. Duffie, Garleanu, and Pedersen (2005) and
When oti = 0 < aD, investors trade only many subsequent papers have something
with dealers who continuously manage posi called a liquidity shock, which is presumably
tions in a frictionless market. As mentioned meant to capture a need to offload financial
earlier, some markets are well approximated assets in favor of more liquid payment instru
by this, while others are better represented ments, but is more rigorously interpreted
by Oil > 0 = aD, such as the federal funds as a change in the utility of consuming the
market, where overnight loans are traded, fruit of a Lucas tree. The papers that inte
typically without intermediation. Afonso and grate models of money and generalized OTC
Lagos (2015b) model the fed funds mar markets take this a level deeper by having a
ket explicitly, providing another case where shock to the utility of a good that is acquired
the traders in the model are interpreted as in the DM, where liquid assets are required
banks. They have a £ M+ and trade bilater for exchange, as in many of the specifications
ally, so the state variable is a time-varying presented above. Then the reason for trying
distribution of asset holdings Ft(a). Afonso to sell off assets is not that you have gone
and Lagos show existence and uniqueness, off fruit, but that you want something that is
characterize the terms of trade, and address easier to get using a more liquid asset. This
various positive and normative questions, seems more realistic, and more elegant, and
including quantitative questions facing cen moreover it allows one to study many other
tral banks. Afonso and Lagos (2015a) con issues in a rigorous way—e.g., the effects of
sider the special case a £ {0,1,2}, similar to monetary policy on OTC markets.
some monetary models. In particular, in Lagos and Zhang (2013)
In other applications, Biais, Hombert, and an asset a called equity with dividend p is
Weill (2014) give a reinterpretation where held by I with time-varying idiosyncratic
agents have continuous access to the market, valuations. There are gains from trade in a
from heterogeneous valuations, and I partic This is again driven by the self-referential
ipates in an OTC market like the DM in our nature of liquidity.
benchmark monetary model, but intermedi Another area where new monetarist mod
ated by D, who again has access to a friction els can provide useful and practical advice,
less interdealer market. In this market, fiat somewhat related to the above-mentioned
money is essential as a medium of exchange, work on fed funds, concerns payment and
and as usual some mechanism like bargain settlement systems in general. Central banks
ing determines the terms of trade between around the world oversee such systems, e.g.,
I and D. The DM alternates with a friction the Division of Reserve Bank Operations
less CM where agents rebalance portfolios.and Payment Systems oversees the Federal
Equilibrium entails a cutoff preference type Reserve Banks as providers of financial ser
such that I below this who meets D trades vices to depository institutions and fiscal
all his equity for cash and I above this trades agency services to the government. A cash
all his cash for equity. When I has all the barless new Keynesian framework, standard GE
gaining power, he trades equity for cash at theory, or reduced-form monetary models
the price in the interdealer market. If D has cannot provide much guidance to the regu
all the bargaining power, I trades at a price lators of these systems, where trade is bet
higher (lower) than the interdealer market ifter characterized as bilateral. See Freeman
I wants to buy (sell) assets. (1996) for an early effort at modeling this
This implies a bid-ask spread determinedactivity carefully. See Williamson and Wright
by monetary policy, and details of the mar(2010b, Sec. 5), Nosal and Rocheteau (2011,
ket structure, such as the speed at which Ch.
I 9), and the papers cited in fn. 54 for
contacts D or bargaining power. A nonmone models designed to address the issues, and
tary equilibrium always exists, and when ir isMcAndrews, Nosal, and Rocheteau (2011)
large it is the only equilibrium. In nonmon for a survey of related work.
etary equilibrium only I holds equity, since
there is no trade in the OTC market, and
12. Conclusion
the equity price is the expected discounted
value of the dividend stream for the average The literature summarized above covers
I. Monetary equilibrium exists for lower n.much territory, including rudimentary the
For 7r not much larger than /3 — 1, there isories of commodity money, variations on
a unique stationary monetary equilibriumand contributions to the general search-and
where D holds all the equity overnight, bargaining literature, fully articulated quan
while I holds it intraday. The asset price is
titative macro systems, and micro depictions
higher than in nonmonetary equilibrium,of financial institutions trading in OTC mar
because OTC exchange props up the resale kets, like the market for fed funds. The mod
value of assets. As usual, the Friedman ruleels differ in various ways, but share some
implements the efficient allocation, and real
basic features, including an attempt to take
asset prices decrease with it because money the microfoundations of exchange seriously.
is complementary with (used to purchase) a. Is it worth the effort? Lucas (2000) says
With entry by D, Lagos and Zhang (2013) "Successful applied science is done at many
generate sunspot equilibria with recurrent levels, sometimes close to its foundations,
episodes that resemble financial crises— sometimes far away from them or without
when a sunspot shock hits, spreads spike,them altogether." This sounds reasonable,
while volume, trading frequency, marketespecially when there are important issues
making activity, and asset prices collapse.that we think we need to address for which
the foundations are poorly developed. But not justify impositions on behavior rather
that is not the case in monetary economics, than on the environment. Is having agents
where work on the microfoundations is rich unable to adjust portfolios (e.g., increase
money balances) after they meet, or unable
and fertile, as we hope to have demonstrated.
When embarking on research we all have to find the right counterparty in the first
to decide what distance from foundations place, on par with assuming sticky prices
makes us comfortable, and here we offer or CIA constraints? We think not, because
some opinions on this. We do not endorse in models with sticky prices or CIA con
models that impose on agents behavior thatstraints agents trade inefficiently after theij
they do not like. An example is the impositionmeet, leaving gains from trade sitting right
of a CIA constraint in the model of bilateral there on the table. It seems different to have
trade in Diamond (1984). If two agents meetinefficient decisions made before meetings,
and want to barter, who are we to precludelike underinvestment in liquidity or search,
that? The situation is worse in CIA mod than to have agents ignore gains from trade
els that otherwise adhere to GE methods, conditional on a meeting. Still, one can think
where agents trade only along their budget about ways to reallocate liquidity after meet
ings, say through banks, or to encourage effi
lines, because when they are not trading with
each other one cannot even ask how they cient search and investments, say through
competitive search or creatively designed
might like to trade. At least in Diamonds
mechanisms, as discussed above.
example, this can be addressed by specifica
tions for specialization precluding barter, asWe do not endorse the use of models where
fiat currency—or bonds or bank deposits—
in Kiyotaki and Wrights (1991) version of his
model, but one has to also preclude credit,enter utility or production functions directly.
From a theoretical perspective, deriving
which has something to do with commitment
and memory issues, as Kocherlakota (1998) the value of assets endogenously may mean
makes clear. A related example of imposi having to work harder to tackle some issues,
tions on agents is assuming that they trade but from a policy perspective, anything that
taking prices as given in situations where the deviates from this discipline is obviously sub
prices are all wrong due to nominal or other ject to the Lucas critique. The appropriate
ness of assumptions depends on the issue at
rigidities. It is natural to take prices as given
in Arrow—Debreu, where they efficiently hand, of course, but it is hard to imagine why
allocate resources. If prices are all wrong, anyone would prefer reduced-form models,
due to rigidities, however, wouldn't the unless the alternatives were too hard. But
economy evolve to other ways of allocating there is nothing especially hard about the
material presented above. To be clear, real
resources? If not, in our opinion, this needs
to be explained, not assumed. assets may generally appear in production or
One sometimes hears that devices like utility functions—e.g., capital, housing, and
CIA constraints or nominal contracts are "nowine all belong there. But if they somehow
different" than limiting what agents can serve
do to facilitate transactions, that is worth
by specifying the environment in particular modeling explicitly.65
ways. We disagree. Frictions like spatial or
temporal separation, limited commitment,65 The distinction can be subtle. The OTC model of
and imperfect information are features
Duffie, Garleanu, and Pedersen (2005) and Lagos and
Rocheteau (2009) has agents getting flow utility from
best made explicit, so one can work out allassets. This can be taken at face value for some assets, like
of their implications. With such frictions,housing
it or fruit-bearing trees. Or, it can be interpreted as a
is atypical to get the first best, but this does
reduced-form for various liquidity and hedging services in
Our position is not, and cannot be, based around participation, search intensity and
on models that strive for deeper microfoun reservation trade decisions inteact with the
dations empirically outperforming relatively effect of inflation, and for some parameter
reduced-form models. Whatever primitives izations they imply that some inflation can
one adopts, including frictions and mech be desirable; and (5) distributional consid
anisms, the same results (for a given set of erations can also imply that some inflation
observables) obtain if one starts with the may be desirable. While (1) is clear from
value function having assets as arguments. standard reduced-form models, (2)-(5) are
In our benchmark model, V(a) = W(a) + not.66 Also, while a nondegenerate F(a) may
aa [«(</) — v(q)], where q depends on a, and be important for some issues, because it can
one can take this V(a) as a primitive to get be a natural outcome of decentralized trade,
an observationally equivalent reduced-form and captures a fundamental tension between
model. It is not so easy, however, to come two roles of monetary policy—providing
up with a good guess for V(a), or even for favorable returns on liquidity and addressing
its properties, out of thin air. Recall that liquidity-sharing considerations (see Wallace
Aruoba and Chugh (2010), e.g., show homo 2014b)—for some purposes it is appropriate
thetic utility over goods does not imply V(a) to work with models with degenerate distri
is homothetic, and this has implications for butions (as in Shi 1997b or Lagos and Wright
optimal policy. Also, key ingredients in V(a) 2005). Again, this depends on the question
are a and er, capturing search and matching, at hand.
and v(q), nesting various mechanisms. In Other issues discussed include the wis
some models the distribution F(a) is also an dom of trying to reduce unemployment
important element, as are pledgeability and using inflation, and the interpretation of
acceptability, as functions of private informa sticky prices. Search and bargaining are not
tion. How does one know how these features critical for making the first point qualita
figure into the reduced-form without deriv tively, and bargaining is replaced by posting
ing it? for the second. Still, Berentsen, Menzio,
Moreover, these features provide new and Wright (2011) argue that both search
avenues of exploration for policy issues. On and bargaining are quantitatively important
inflation, in particular, the models presented in accounting for unemployment and infla
incorporate several effects: (1) inflation is a tion, and search is obviously the key to sticky
tax on real balances; (2) bargaining can com prices in Head et al. (2012). Both search
pound this wedge; (3) so can search-and and bargaining are relevant for the quan
matching frictions; (4) distortions revolving titative effects of inflation. Lagos (2010a)
shows how frictions help us understand
issues in financial economics quantitatively,
the case of financial securities, like municipal or corporate
bonds (Duffle, Garleanu, and Pedersen 2007). We view
the latter interpretation as acceptable when the focus ®>We is reiterate that while the constraints in some of our
not on why people trade assets, but on the consequences models "look like" CIA restrictions, they are in fact feasibil
of trading in frictional markets. However, in our baseline
ity conditions. It is clear that in Kiyotaki and Wright (1989),
alternating-market model this is less of a problem: assets
e.g., traders cannot turn over something they do not have,
can give off coupons or dividends in CM numeraire but or, this cannot be called a CIA constraint because the
baseline model does not even have cash. Moreover, the
realistically, dollars. To further emphasize the point, assets
in the baseline Kiyotala-Wright model also directly transactions
gen pattern is endogenous and may not be unique.
We reject the idea that the approach is "the same as" CIA
erate utility. But rather than saying money enters utility
or MUF models, despite the obvious result that one can
functions, it seems more accurate to say that some goods
always reverse engineer a reduced form to "look like" a
in the utility function end up endogenously playing a role
commonly ascribed to money. microfounded model (e.g., as in Camera and Chien 2013).
while Rocheteau and Wright (2013) argue work is needed on the theories that combine
that these kinds of models are consistent money and credit.6'
with outcomes that appear anomalous from We close by highlighting a few issues
the perspective of standard asset-pricing where the methods covered above are espe
theory. The approach also sheds new lightcially useful or provide novel insights. First,
on inside versus outside money (Cavaleantithey deliver endogenous exchange pat
and Wallace 1999a, 1999b), banking terns that illustrate the interplay between
(Berentsen et al. 2007; Gu et al. 2013a), intrinsic characteristics (e.g., storability or
investment (Shi 1999a, 1999b; Aruoba, recognizability) and beliefs in determin
Waller, and Wright 2011), OTC financial
ing which objects will or should be used to
markets (Duffie, Garleanu, and Pedersen
facilitate transactions. They determine the
2005; Lagos and Rocheteau 2009), and botheffective supply of liquidity, depending on
conventional and unconventional mone the environment and policy. They allow us
tary policy (Williamson 2012; Rocheteau, to study monetary, credit, and intermediary
Wright, and Xiao 2014), to mention a few arrangements, and allow us to clarify the
additional examples. essential frictions. Importantly, the approach
At the frontier, research is trying toisfur amenable to implementation theory and
ther develop models with multiple assets mechanism design, mapping frictions like
having different returns, without ad hoc commitment or information problems into
restrictions on their use in transactions. This
incentive-feasible allocations, and identify
"modified Hahn problem" (Hellwig 1993) ingisinstitutions with good welfare properties.
challenging, but there are currently expla This is problematic in reduced-form mod
nations under discussion; (1) certain els, pair where the frictions are not well speci
wise trading mechanisms can deliver this fied.
as Also, the framework is flexible enough
an outcome (Zhu and Wallace 2007); (2) to so accommodate a variety of market struc
can private information about asset quality This is relevant for understanding
tures.
(Li, Rocheteau, and Weill 2012; Lester,
how, e.g., the impact of inflation depends on
Postlewaite, and Wright 2012); (3) sowhether
can the terms of trade are determined
assumptions about safety, e.g., from theftby bargaining, price posting or price taking,
(He, Huang, and Wright 2008; Sanches whether
and search is random or directed, etc.
Williamson 2010); (4) it can also be a self-ful
Future work should continue to explore dif
filling prophecy (Lagos 2013b); and (5) ferent
it is micro market structures.
sometimes socially efficient (Kocherlakota The framework also illustrates how econ
2003; Hu and Rocheteau 2013). Are these omies where liquidity considerations matter
explanations satisfactory? Which are most can be prone to multiplicity and interesting
dynamics, where endogenous transaction
relevant? While progress has been made,
these are still important open questions.
Another direction is to pursue qualitative
®^They prove that in many natural environments,
and quantitative models combining improvements
New in credit conditions are irrelevant in mon
Monetarist and Keynesian features, as inequilibrium, because such improvements simply
etary
crowd out
examples by Aruoba and Schorfheide (2011) real balances one for one. This is very much
like other irrelevance results (e.g., Modigliani-Miller or
and Williamson (2015). More quantita
Ricardian equivalence), in that there may well be excep
tive work on all this would be welcome. So tions to the baseline results, as Gu, Mattesini, and Wright
would further research on banks and other (2016) discuss, but in many standard models the results
hold, and more generally, the results of changes in credit
intermediaries as providers of liquidity. Gu, conditions can be very different in monetary and nonmon
Mattesini, and Wright (2016) suggest more etary economies.
patterns are not unique nor stationary. This curve and sticky nominal prices, both of
is true of simple first- and second-generation which are commonly assumed in macro,
models, as well as the more sophisticated ver but usually with different implications. The
sions designed to study the macro economy theory can be used not only to make con
and financial markets. We mentioned how ceptual points about these issues, but also to
elements like a, a, v, F, and x open up new organize and interpret micro data, e.g., as in
avenues of exploration, e.g., in discussion of
the studies of sticky prices discussed above.
the hot-potato effect, where velocity dependsIt can also deliver time series observations
on explicit search, entry, and trading deciwhere after a monetary injection, quanti
sions, or in the discussion of pledgeability andties first rise, then later prices rise, without
acceptability based on information theory. sticky-price assumptions, but with prices and
The models are also set up to analyze creditquantities determined bilaterally efficiently.
in the context of bilateral relationships, orThe models generate a demand for liquidity
pairwise meetings, where private informationthat helps understand correlations between
is naturally accommodated. Sometimes themoney holdings and nominal rates or bond
models generate novel perspectives on topiholdings and spreads, without sticking assets
cal issues, as with the zero lower bound prob in utility. Finally, we mention how the mod
lem in models with Nash bargaining, whereels allow one to analyze many dimensions
l < 0 would be desirable but is not feasible. of liquidity in a unified framework, includ
The models can also contribute to discus ing acceptability, pledgeability, moneyness,
sions of informational sensitivity, liquidityvelocity, trade volume, bid-ask spreads, and
traps, and price bubbles. The theory canliquidity premia. We hope our presentation
be used to better understand the impact of of this approach will be useful, especially
OMOs and less conventional policies. It cangiven current interest in liquidity in econom
be used to deliver a fully exploitable Phillipsics and finance.
Appendix on Notation
a,r, (5 — arrival rate, discount rate, discount factor
p = (utility of) dividend if p > 0 or storage cost i F p < 0
a, A = individual, aggregate asset holdings
0, 5 — single- and double-coincidence probability
K,e = cost and probability of entry a la Pissarides
nj = measure of type /, mt = measure inventory i
//, = monitoring probability
n = fraction of monitored agents in CW
r — trading strategy
r=T(f) = best response correspondence
u,c = utility, cost of DM good
q,Q = quantity in monetary, barter trades
VA, Vs, Vc, VD = value functions for autarky, barter, credit, and deviation
Va = value function for a 6 {0,1}
6a = bargaining power of agent with a £ {0,1}
v(q) = cost of q—i.e., a general mechanism
rj — aO — arrive rate times bargaining power
P,M,C — producer, middleman, and consumer in RW
1,D = investor and dealer in DGP
Sj = surplus in DGP
LOj — probability of preference shock j in DGP
F(a) = asset distribution
V(a), W(a) = value function for a € R+ in DM, CM
(f>, z = (<f> + p) a — price and value of a
U(x) — £ — CM utility
S(q) =u[q) — c[cf) = DM surplus
d,p = d/q = DM dollars and price
it, i = money growth (or inflation) and nominal interest rate
G, T = gov't consumption and transfers
A — liquidity premium or Lagrange multiplier
Wf , wk = factor prices for labor and capital
tj, tk = tax rates on labor and capital income
5^ = depreciation rate on k
7, r = DM utility function, I q1-7/(I — 7)
D, x = KM debt and haircut parameter
£b> 0; — wedges on shares and bonds in Lagos
R — 1 + l = gross returns
tp = DM preference shock, i/)u(q)
1 — £ = probability that p = pH, p = pL
2
(l + r)Vi2= p2 + a«iVi2 (1,1,0) (V2/2.V2 - 1,1) maybe
3 (1,0,1) never
(V2 — 1,1, V2/2)
+ Qrl2w2[Tl V]3 + (1 — Ti) V12] 4
(1,0,0) (1/2,1,1) never
5 (0,1,1) maybe
(l, V2/2, V2 — l)
+ cm2(l - m2) (« + Vj2) 6 (0,1,0) (1,1/2,1) maybe
7 (0,0,1) (1,1,1/2) never
+ (1 — r3) V12]
+ an3(l -m3)V12.
The RHS is type Is payoff next period x (1 — r2), but we can ignore that
from the dividend, plus the expected value either r2 = 1 or m2 = 1. Equating the
of: meeting his own type with probability sure of type 1 that switch from good
nh which implies no trade; meeting type 2 good 3 and the measure that switch bac
with their production good with probability get (1).
n2m2, which implies trade with probabil Table 1 lists candidate equilibria, with exis
ity meeting type 2 with good 1, which tence results for the case n,- = 1/3. Consider
implies trade for sure; meeting type 3 with case 1. After inserting m, the BR conditions
good 1, which implies trade with probability reduce to
73; and meeting type 3 with good 2, which
implies no trade. Algebra leads to (2). Aj > 0, or p3 — p2 > w/6;
Now to explain the SS condition, consider
type 1 and pure strategies. If 1 has good 2, A2 > 0, or pi — p3 > w/6;
he can switch to good 3 only by trading with a
2 that has good 3 (since 3 never has good 3). ^3 > 0, or p2 — p\ > u/6.
For this, 1 has to meet 2 with good 3,which
occurs with probability n2m2, then trade, Since these cannot all hold, this is never an
which occurs with probability Tj (since 2 equilibrium. In contrast, for case 2, the BR
always wants good 2). And if 1 has good 3, he conditions reduce to
switches to good 2 only by acquiring good 1,
consuming and producing a new good 2 (he Ai > 0, or p3 — p2 > (1 — V2 )u/3;
never switches from good 3 to 2 directly, since
if he preferred good 2 he would not trade it A2 > 0, or pi - p3 > 0;
for good 3 in the first place). He trades good
3 for good 1 either by trading with 3 that has A3 < 0, or p2 — Pi < (1 — a/2/2)m/3.
good 1, which occurs with probability n3m3,
or with 2 that has good 1 but prefers good For some parameters, these all hold and this
3, which occurs with probability n2( 1 — m2) is an equilibrium. The rest are similar. I
Christopher Waller. 2014. "Optimal Disclosure Azariadis, Costas. 1993. Intertemporal Macroeconom
Policy and Undue Diligence." Journal of Economic ics. Oxford and Cambridge: Blackwell.
Theory 149: 128-52. Babus, Ana, and Tai-Wei Hu. 2015. "Endogenous
Brown, Paul M. 1996. "Experimental Evidence on Mechanism for Sharing Capital: Diamond and Dyb
Money as a Medium of Exchange." Journal of Eco vig Meet Kiyotaki and Wright." Economic Theory 24
nomic Dynamics and Control 20 (4): 583-600. (4): 769-88.
Burdett, Kenneth, Melvyn Coles, Nobuhiro Kiyotaki, Cavalcanti, Ricardo de O., Andres Erosa, and Ted Tem
and Randall Wright. 1995. "Buyers and Sellers: zelides. 1999. "Private Money and Reserve Manage
Should I Stay or Should I Go?" American Economic ment in a Random-Matching Model." Journal of
Review 85 (2): 281-86. Political Economy 107 (5): 929-45.
Burdett, Kenneth, Mei Dong, Ling Sun, and Randall Cavalcanti, Ricardo de O., Andres Erosa, and Ted
Wright. 2016. "Marriage, Markets and Money: A Temzelides. 2005. "Liquidity, Money Creation and
Coasian Theory of Household Formation." Interna Destruction, and the Returns to Banking." Interna
tional Economic Review 57 (2): 337-68. tional Economic Review 46 (2): 67.5-706.
Burdett, Kenneth, and Kenneth L. Judd. 1983. "Equi Cavalcanti, Ricardo de O., and Neil Wallace. 1999a.
librium Price Dispersion." Econometrica 51 (4): "Inside and Outside Money as Alternative Media of
955-69. Exchange." Journal of Money, Credit and Banking 31
Burdett, Kenneth, and Guido Menzio. 2014. "The(3 Part 2): 443-^57.
(Q,S,s) Pricing Rule." Unpublished. Cavalcanti, Ricardo de O., and Neil Wallace. 1999b. "A
Burdett, Kenneth, Shouyong Shi, and Randall Wright. Model of Private Bank-Note Issue." Review of Eco
2001. "Pricing and Matching with Frictions ."Journalnomic Dynamics 2(1): 104^36.
of Political Economy 109 (5): 1060-85. Chapman, James, Jonathan Chiu, and Miguel Mol
Burdett, Kenneth, Alberto Trejos, and Randall Wright.ico. 2011. "Central Bank Haircut Policy." Annals of
2001. "Cigarette Money." Journal of Economic TheFinance 7 (3): 319^18.
ory 99 (1-2): 117^2. Chapman, James, Jonathan Chiu, and Miguel Molico.
Burdett, Kenneth, Alberto Trejos, and Randall Wright.2013. "A Model of Tiered Settlement Networks."
2016. "A New Suggestion for Simplifying the TheoryJournal of Money, Credit and Banking 45 (2-3):
of Money." Unpublished. 327-47.
Camargo, Braz, and Benjamin Lester. 2014. "Trading Chang, Briana. Forthcoming. "Adverse Selection and
Dynamics in Decentralized Markets with AdverseLiquidity Distortion." Review of Econoynic Studies.
Chari, V. V., Lawrence J. Christiano, and Patrick J.
Selection." Journal of Economic Theory 153: 534-68.
Camera, Gabriele. 2001. "Search, Dealers, and the Kehoe. 1991. "Optimal Fiscal and Monetary Policy:
Terms of Trade." Review of Economic Dynamics Some 4 Recent ResultsJournal of Money, Credit and
(3): 680-94. Banking 23 (3 Part 2): 519-39.
Camera, Gabriele, and YiLi Chien. 2013. "Model Chiu, Jonathan, and Thorsten V. Koeppl. 2016. "Trad
ing Monetary Economies: An Equivalence Result." ing Dynamics with Adverse Selection and Search:
Federal Reserve Bank of St. Louis Working Paper Market Freeze, Intervention and Recovery." Review
2013-009.. of Economic Studies 83 (3): 969-1000.
Camera, Gabriele, and Dean Corbae. 1999. "Money Chiu, Jonathan, and Cesaire A. Meh. 2011. "Financial
and Price Dispersion." International Economic Intermediation, Liquidity, and Inflation." Macroeco
Review 40 (4): 985-1008. nomic Dynamics 15 (Supplement SI): 83-118.
Camera, Gabriele, Ben Craig, and ChristopherChiu, J. Jonathan, Cesaire A. Meh, and Randall Wright.
Waller. 2004. "Currency Competition in a Funda2017. "Innovation and Growth with Financial, and
mental Model of Money." Journal of InternationalOther, Frictions." International Economic Review 58
Economics 64 (2): 521-44. (1): 95-125.
Camera, Gabriele, and Yiting Li. 2008. "Another Exam
Chiu, Jonathan, and Miguel Molico. 2010. "Liquidity,
ple of a Credit System that Coexists with Money."Redistribution, and the Welfare Cost of Inflation."
Journal of Money, Credit, and Banking 40 (6): Journal of Monetary Economics 57 (4): 428-38.
1295-1308. Chiu, Jonathan, and Miguel Molico. 2011. "Uncer
Camera, Gabriele, Robert R. Reed, and Christopher J.tainty, Inflation, and Welfare." Journal of Money,
Credit and Banking 43 (Supplement S2): 487-512.
Waller. 2003. "Jack of All Trades or Master of One?
Chiu, Jonathan, and Miguel Molico. 2014. "Short-Run
Specialization, Trade and Money." International Eco
nomic Review 44 (4): 1275-94. Dynamics in a Search-Theoretic Model of Monetary
Exchange." Unpublished.
Caplin, Andrew S., and Daniel F. Spulber. 1987. "Menu
Chiu, Jonathan, and Cyril Monnet. 2015. "Relation
Costs and the Neutrality of Money." Quarterly Jour
nal of Economics 102 (4): 703-25. ships in the Interbank Market." Unpublished.
Carmona, Guilherme. 2015. "Monetary Trading:
Chiu, Jonathan, and Tsz-Nga Wong. 2015. "Pay
ment Systems: A Mechanism Design Approach."
An Optimal Trading Mechanism under Limited
Record-Keeping." Unpublished. Unpublished.
Cho, In-Koo, and David M. Kreps. 1987. "Signaling
Carapella, Francesca, and Stephen Williamson. 2015.
"Credit Markets, Limited Commitment, and Gov Games and Stable Equilibria." Quarterly Journal of
Economics 102 (2): 179-221.
ernment Debt." Review of Economic Studies 82 (3):
963-90. Choi, Michael. 2013. "Imperfect Information Trans
mission and Adverse Selection in Asset Markets."
Cavalcanti, Ricardo de O. 2004. "A Monetary
Duffle, Darrell, Nicolae Garleanu, and Lasse Heje and Information in Competitive Search." BE. Jour
Pedersen. 2007. "Valuation in Over-the-Counter nal of Macroeconomics: Advances in Macroeconom
Markets." Review of Financial Studies 20 (6): ics 6(1): 1-34.
1865-1900. Faig, Miquel, and Belen Jerez. 2007. "Precautionary
Duffy, John. 2001. "Learning to Speculate: Experi Balances and the Velocity of Circulation of Money."
ments with Artificial and Real Agents." Journal Journal
of of Money, Credit and Banking 39 (4): 843-73.
Economic Dynamics and Control 25 (3-4): 295-319. Faig, Miquel, and Zhe Li. 2009. "The Welfare Costs
Duffy, John, and Jack Ochs. 1999. "Emergence of Expected and Unexpected Inflation." Journal of
Money as a Medium of Exchange: An Experimental Monetary Economics 56 (7): 1004-13.
Study." American Economic Review 89 (4): 847-77. Ferraris, Leo, and Makoto Watanabe. 2008. "Collateral
Duffy, John, and Jack Ochs. 2002. "Intrinsically WorthSecured Loans in a Monetary Economy." Journal of
less Objects as Media of Exchange: Experimental Economic Theory 143 (1): 405-24.
Evidence." International Economic Review 43 Ferraris,
(3): Leo, and Makoto Watanabe. 2011. "Collateral
637-73. Fluctuations in a Monetary Economy." Journal of
Duffy, John, and Daniela Puzzello. 2014. "Gift Economic Theory 146 (5): 1915^0.
Exchange versus Monetary Exchange: Theory and Ferraris, Leo, and Makoto Watanabe. 2012. "Liquidity
Evidence." American Economic Review 104 (6):Constraints in a Monetary Economy." International
1735-76. Economic Review 53 (1): 255-77.
Freeman, Scott. 1996. "The Payments System, Liquid
Dutta, Rohan. 2012. "Bargaining with Revoking Costs."
Games and Economic Behavior 74 (1): 144-53. ity, and Rediscounting." American Economic Review
Dutu, Richard. 2006. "The Costs of Inflation in Aus86 (5): 1126-38.
tralia and New Zealand." New Zealand Economic Friedman, Milton. 1960. A Program for Monetary Sta
Papers 40 (2): 127-46. bility. New York: Fordham University Press.
Dutu, Richard, Stella Huangfu, and Benoit Julien. Friedman, Milton. 1968. "The Role of Monetary Pol
2011. "Contingent Prices and Money." International icy." American Economic Review 58 (1): 1-17.
Economic Review 52 (4): 1291—1308. Friedman, Milton. 1969. The Optimum Quantity of
Dutu, Richard, Benoit Julien, and Ian King. 2012. "OnMoney: and Other Essays. London: Macmillan.
the Welfare Gains of Price Dispersion." Journal ofFriedman, Milton. 1977. "Inflation and Unemploy
Money, Credit and Banking 44 (5): 757-86. ment." JoumflZ of Political Economy 85 (3): 451-72.
Eden, Benjamin. 1994. "The Adjustment of Prices Friedman, Milton. 1992. Money Mischief: Episodes in
to Monetary Shocks When Trade Is Uncertain and Monetary History. San Diego and New York: Har
Sequential." Journal of Political Economy 102 (3): court Brace and Company.
493-509. Friedman, Milton, and Anna Jacobson Schwartz. 1963.
Engineer, Merwan, and Shouying Shi. 1998. "Asymme A Monetary History of the United States, 1867-1960.
Princeton
try, Imperfectly Transferable Utility, and the Role of and Oxford: Princeton University Press.
Fung,ofBen, and Enchuan Shao. 2011. "Modelling the
Fiat Money in Improving Terms of Trade." Journal
Monetary Economics 41 (1): 153-83. Counterfeiting of Bank Notes: A Literature Review."
Bank of Canada Review Autumn: 29-35.
Engineer, Merwan, and Shouyong Shi. 2001. "Bargains,
Gale, David, and Lloyd S. Shapley. 1962. "College
Barter, and Money." Review of Economic Dynamics
4(1): 188-209. Admissions and the Stability of Marriage." American
Ennis, Huberto M. 2001. "On Random Matching, Mathematical Monthly 69 (1): 9-15.
Calenianos, Manolis, and Philipp Kircher. 2008. "A
Monetary Equilibria, and Sunspots." Macroeconomic
Dynamics 5 (1): 132-42. Model of Money with Multilateral Matching." Jour
Ennis, Huberto M. 2008. "Search, Money, and nalInfla
of Monetary Economics 55 (6): 1054-66.
tion under Private Information." Journal of Gavazza,
Eco Alessandro. 2011. "The Role of Trading Fric
nomic Theory 138 (1): 101-31. tions in Real Asset Markets." American Economic
Ennis, Huberto M. 2009. "Avoiding the InflationReview
Tax." 101 (4): 1106-43.
International Economic Review 50 (2): 607-25.Geromichalos, Athanasios, and Lucas Herrenbrueck.
Ennis, Huberto M. 2014. "A Simple General 2016. Equi "Monetary Policy, Asset Prices, and Liquidity
in Over-the-Counter Markets." Journal of Money,
librium Model of Large Excess Reserves." Federal
Reserve Bank of Richmond Working Paper 14-14. Credit, and Banking 48 (1): 35-79.
Faig, Miquel. 2008. "Endogenous Buyer-Seller Geromichalos,
Choice Athanasios, Lucas Herrenbrueck, and
Kevin Salyer. 2016. "A Search-Theoretic Model of
and Divisible Money in Search Equilibrium." Jour
nal of Economic Theory 141 (1): 184—99. the Term Premium." Theoretical Economics 11 (3):
897-935.
Faig, Miquel, andXiuhua Huangfu. 2007. "Competitive
Geromichalos, Athanasios, Juan Manuel Licari, and
Search Equilibrium in Monetary Economies ."Journal
of Economic Theory 136 (1): 709-18. Jose Suarez-Lledo. 2007. "Monetary Policy and
Faig, Miquel, and Belen Jerez. 2005. "A Theory Asset
ofPrices." Review of Economic Diinamics 10 (4):
Commerce." Journal of Economic Theory 122761-79.
(1):
60-99. Geromichalos, Athanasios, and Ina Simonovska. 2014.
Faig, Miquel, and Belen Jerez. 2006. "Inflation, Prices,"Asset Liquidity and International Portfolio Choice."
Illiquid Bonds." Journal of Economic Theory 108 (2): Lagos, Ricardo, and Guillaume Rocheteau. 2009.
179-93. "Liquidity in Asset Markets with Search Frictions."
Kocherlakota, Narayana R., and Neil Wallace. 1998.Econometrica 77 (2): 403-26.
"Incomplete Record-Keeping and Optimal PaymentLagos, Ricardo, Guillaume Rocheteau, and
Arrangements." Journal of Economic Theory 81 (2):Pierre-Olivier Weill. 2011. "Crises and Liquidity in
272-89. Over-the-Counter Markets." Journal of Economic
Theory 146 (6): 2169-2205.
Koeppl, Thorsten, Cyril Monnet, and Ted Temzelides.
2008. "A Dynamic Model of Settlement." JournalLagos,
of Ricardo, and Randall Wright. 2003. "Dynam
Economic Theory 142 (1): 233-46. ics, Cycles, and Sunspot Equilibria in 'Genuinely
Koeppl, Thorsten, Cyril Monnet, and Ted Temzelides.
Dynamic, Fundamentally Disaggregative' Models
2012. "Optimal Clearing Arrangements for Finan
of Money." Journal of Economic Tlieory 109 (2):
156-71.
cial Trades ."Journal of Financial Economics 103 (1):
189-203. Lagos, Ricardo, and Randall Wright. 2005. "A Unified
Krishnamurthy, Arvind, and Annette Vissing-Jorgensen.Framework for Monetary Theory and Policy Analy
2012. "The Aggregate Demand for Treasury Debt." sis." Journal of Political Economy 113 (3): 463-84.
Journal of Political Economy 120 (2): 233-67. Lagos, Ricardo, and Randall Wright. 2007. "When Is
Kumar, Alok. 2008. "Inflation and the Dispersion of Money Essential? A Comment on Aliprantis, Camera
Real Wages." International Economic Review 49 (2):and Puzzello." Unpublished.
377-99. Lagos, Ricardo, and Shengxing Zhang. 2013. "A Model
Kurmann, Andre. 2014. "Holdups and Overinvestment of Monetary Exchange in Over-tne-Counter Mar
in Capital Markets." Journal of Economic Theorti
151: 88-113. Laing, DereK, Victor E. Li, and Ping Wang. 2007. "Infla
Lagos, Ricardo. 2000. "An Alternative Approach totion and Productive Activity in a Multiple-Matching
Search Frictions." Journal of Political Economy 108Model of Money." Journal of Monetary Economics 54
(5): 851-73. (7): 1949-61.
Laing, Derek, Victor E. Li, and Ping Wang. 2013.
Lagos, Ricardo. 2008. "Liquidity and the Search The
ory of Money." Society for Economic Dynamics "Decentralized Exchange and Factor Payments:
Newsletter 10 (1). A Multiple-Matching Approach." Macroeconomic
Dynamics 17 (1): 54-88.
Lagos, Ricardo. 2010a. "Asset Prices and Liquidity in
an Exchange Economy." Journal of Monetary Eco Law, John. 1705. Money and Trade Considered: With
nomics 57 (8): 913-30. a Proposal for Supplying the Nation with Money.
Lagos, Ricardo. 2010b. "Some Results on the Optimal Glasgow: R. & A. Foulis.
ity and Implementation of the Friedman Rule in Lee,
the Manjong, Neil Wallace, and Tao Zhu. 2005. "Mod
Search Theory of Money." Journal of Economic The eling Denomination Structures." Econometrica 73
ory 145 (4): 1508-24. (3): 949-60.
Lehmann, Etienne. 2012. "A Search Model of Unem
Lagos, Ricardo. 2011. "Asset Prices, Liquidity, and
Monetary Policy in an Exchange Economy." Journal ployment and Inflation." Scandinavian Journal of
of Money, Credit and Banking 43 (Supplement S2):Economics 114 (1): 245-66.
521-52. Lester, Benjamin. 2009. "Settlement Systems." BE
Lagos, Ricardo. 2013a. "Moneyspots and Coexistence Journal of Macroeconomics 9 (1).
in the Pure Theory of Money: A Reply to Neil Wal Lester, Benjamin. 2011. "Information and Prices with
Capacity Constraints." American Economic Review
Iace." Journal of Political Economy 121 (4): 796-801.
101 (4): 1591-1600.
Lagos, Ricardo. 2013b. "Moneyspots: Extraneous
Attributes and the Coexistence of Money and
Lester, Benjamin, Andrew Postlewaite, and Randall
Interest-Bearing Nominal Bonds." Journal of PolitWright. 2011. "Information and Liquidity." Journal
ical Economy 121 (1): 127-85. of Money, Credit and Banking 43 (Supplement S2):
355-77.
Lagos, Ricardo, and Guillaume Rocheteau. 2005.
Lester, Benjamin, Andrew Postlewaite, and Randall
"Inflation, Output, and Welfare." International Eco
nomic Review 46 (2): 495-522. Wright. 2012. "Information, Liquidity, Asset Prices,
Lagos, Ricardo, and Guillaume Rocheteau. 2007. and Monetary Policy." Review of Economic Studies
"Search in Asset Markets: Market Structure, Liquid79 (3): 1209-38.
ity, and Welfare." American Economic Review Lester,
97 Benjamin, Guillaume Rocheteau, and
(2): 198-202. Pierre-Olivier Weill. 2015. "Competing for Order
Lagos, Ricardo, and Guillaume Rocheteau. 2008a.Flow in OTC Markets." Journal of Money, Credit
and Banking 47 (S2): 77-126.
"Liquidity in Asset Markets with Search Frictions."
Federal Reserve Bank of Cleveland Working Paper Li, Victor E. 1994. "Inventory Accumulation in a
08-04. Search-Based Monetary Economy." Journal of Mon
Lagos, Ricardo, and Guillaume Rocheteau. 2008b. etary Economics 34 (3): 511-36.
"Money and Capital as Competing Media of
Li, Victor E. 1995. "The Optimal Taxation of Fiat
Exchange." Journal of Economic Theory 142 (1): Money in Search Equilibrium." International Eco
247-58. nomic Review 36 (4): 927^42.
Li, Ying-Syuan, and Yiting Li. 2013. "Liquidity and Lucas, Robert E. 1972. "Expectations and the Neu
Asset Prices: A New Monetarist Approach." Journal trality of Money." Journal of Economic Theory 4 (2):
of Monetary Economics 60 (4): 426^-38. 103-24.
Li, Yiting. 1995. "Commodity Money under Private Lucas, Robert E. 1976. "Econometric Policy Evalua
Information." Journal ofMonetarii Economics 36 (3): tion: A Critique." Carnegie-Rochester Conference
573-92. Series on Public Policy 1: 19^6.
Li, Yiting. 1998. "Middlemen and Private Information."Lucas, Robert E. 1978. "Asset Prices in an Exchange
Journal of Monetary Economics 42 (1): 131-59. Economy." Econometrica 46 (6): 1429^15.
Li, Yiting. 2001. "A Search Model of Money and Cir Lucas, Robert E. 1980. "Two Illustrations of the Quan
culating Private Debt with Applications to Monetity Theory of Money." American Economic Review
tary Policy." International Economic Review 42 (4):70 (5): 1005-14.
925-46. Lucas, Robert E. 2000. "Inflation and Welfare." Econo
Li, Yiting. 2006. "Banks, Private Money, and Governmetrica 68 (2): 247-74.
ment Regulation." Journal of Monetary Economics Lucas, Robert E., and Juan Pablo Nicolini. 2015. "On
53 (8): 2067-83. the Stability of Money Demand." Journal of Mone
Li, Yiting. 2007. "Inside Money, Organized Markets,tary Economics 73: 48-65.
Lucas, Robert E., and Edward Prescott. 1974. "Equi
and Specialization." Macroeconomic Dynamics 11
(3): 388-404. librium Search and Unemployment." Journal of Eco
Li, Yiting. 2011. "Currency and Checking Deposits as nomic Theory 7 (2): 188-209.
Means of Payment." Review of Economic Dynamics Luo, Guo Ying. 1998, "The Evolution of Money as a
14 (2): 403-17. Medium of Exchange." Journal of Economic Dynam
Li, Yiting, and Guillaume Rocheteau. 2009. "Liquidity ics and Control 23 (3): 415-58.
Constraints." Unpublished. Mailath, George J., Masahiro Okuno-Fujiwara, and
Li, Yiting, and Guillaume Rocheteau. 2011. "On the Andrew Postlewaite. 1993. "Belief-Rased Refine
Threat of Counterfeiting." Macroeconomic Dynam ments in Signaling Games." Journal of Economic
ics 15 (Supplement SI): 10-41. Theory 60 (2): 241-76.
Li, Yiting, Guillaume Rocheteau, and Pierre-Olivier Marimon, Ramon, Ellen McGrattan, and Thomas J.
Weill. 2012. "Liquidity and the Threat of Fraudu Sargent. 1990. "Money as a Medium of Exchange
lent Assets." Journal of Political Economy 120 (5): in an Economy with Artificially Intelligent Agents."
815-46. Journal of Economic Dynamics and Control 14 (2):
Li, Yiting, and Randall Wright. 1998. "Government 329-73.
Transaction Policy, Media of Exchange, and Prices." Martin, Antoine, and Cyril Monnet. 2011. "Monetary
Journal of Economic Theory 81 (2): 290-313. Policy Implementation Frameworks: A Comparative
Liu, Lucy Qian. 2010. "Inflation and Unemployment: Analysis." Macroeconomic Dynamics 15 (Supple
The Roles of Goods and Labor Markets Institutions." ment SI): 145-89.
Unpublished. Masters, Adrian. 2007. "Middlemen in Search Equi
Liu, Lucy Qian, Liang Wang, and Randall Wright. 2011. librium." International Economic Review 48 (1):
"On the 'Hot Potato' Effect of Inflation: Intensive 343-62.
versus Extensive Margins." Macroeconomic Dynam Masters, Adrian. 2008. "Unpleasant Middlemen."/our
ics 15 (Supplement S2): 191-216. nal of Economic Behavior and Organization 68 (1):
Liu, Lucy Qian, Liang Wang, and Randall Wright. 2014. 73-86.
"Sticky Prices and Costly Credit." Unpublished. Matsui, Akihiko, and Takashi Shimizu. 2005. "A The
Liu, Qing, and Shouyong Shi. 2010. "Currency Areas ory of Money and Marketplaces." International Eco
and Monetary Coordination." International Eco nomic Review 46 (1): 35^59.
nomic Review 51 (3): 813-36. Matsuyama, Kiminori, Nobuhiro Kiyotaki, and Akihiko
Lomeli, Hector, and Ted Temzelides. 2002. "Discrete Matsui. 1993. "Toward a Theory of International Cur
Time Dynamics in a Random Matching Monetary rency." Review of Economic Studies 60 (2): 283-307.
Model." Economic Theory 20 (2): 259-69. Mattesini, Fabrizio, and Ed Nosal. 2016. "Cash-in
Lotz, Sebastien. 2004. "Introducing a New Currency: the-Market Pricing in a Model with Money and
Government Policy and Prices." European Economic Over-the-Counter Financial Markets." Journal of
Review 48 (5): 959-82. Economic Theory 164: 187-215.
Lotz, Sebastien, and Guillaume Rocheteau. 2002. McAndrews, James, Ed Nosal, and Guillaume Roche
"On the Launching of a New Currency." Journal of teau. 2011. "Payment Systems." Annual Review of
Money, Credit, ana Banking 34 (3 Part 1): 563-88. Financial Economics 3: 259-87.
Lotz, Sebastien, Andrei Shevchenko, and Christopher Mehra, Rajnish, and Edward C. Prescott. 1985. "The
Waller. 2007. "Heterogeneity and Lotteries in Mon Equity Premium: A Puzzle." Journal of Monetary
etary Search Models ."Journal of Money, Credit and Economics 15 (2): 145-61.
Banking 39 (2-3): 703-12. Melin, Lionel. 2012. "Multi-speed Markets."
Lotz, Sebastien, and Cathy Zhang. 2016. "Money and Unpublished.
Credit as Means of Payment: A New MonetaristMenger, Carl. 1892. "On the Origin of Money." Eco
ApproachJournal of Economic Theory 164: 68-100. nomic Journal 2: 239-55.
Menner, Martin. 2006. "A Search-Theoretic Monetary Oberfield, Ezra, and Nicholas Trachter. 2012. "Com
Business Cycle Model with Capital Formation." BE modity Money with Frequent Search." Journal of
Journal of Macroeconomics: Contributions in Macro Economic Theory 147 (6): 2332-56.
economics 6 (1). Oh, Seonghwan. 1989. "A Theory of a Generally
Menzio, Guido, Shouyong Shi, and Hongfei Sun. Acceptable Medium of Exchange and Barter." Jour
2013. "A Monetary Theory with Non-degenerate nal of Monetary Economics 23 (1): 101-19.
Okumura, Tsunao. 2006. "Wealth as a Signal in the
Distributions." Journal of Economic Theory 148 (6):
2266-2312. Search Model of Money." International Economic
Review 47 (1): 87-106.
Mihm, Stephen. 2007. A Nation of Counterfeiters:
Capitalists, Con Men, and the Making of the United
Ostroy, Joseph M. 1973. "The Informational Efficiency
States. Cambridge, MA: Harvard University Press.
of Monetary Exchange." American Economic Review
Mills, David C. 2008. "Imperfect Monitoring and 63
the(4): 597-610.
Discounting of Inside Money." International Ostroy,
Eco Joseph M., and Ross M. Starr. 1990. "The
nomic Review 49 (3): 737-54. Transactions Role of Money." In Handbook of Mone
Moen, Espen R. 1997. "Competitive Search Equilib tary Economics, Volume 1, edited by B. M. Friedman
and F. H. Hahn, 3-62. Amsterdam and New York:
rium." Journal of Political Economy 105 (2): 385^411.
Elsevier, North-Holland.
Molico, Miguel. 2006. "The Distribution of Money and
Prices in Search Equilibrium." InternationalPagnotta,
Eco Emiliano S., and Thomas Philippon. 2011.
nomic Review 47 (3): 701-22. "Competing on Speed." National Bureau of Eco
Molico, Miguel, and Yahong Zhang. 2005. "Monetarynomic Research Working Paper 17652.
Policy and the Distribution of Money and Capital."
Peterson, Brian, and Shouyong Shi. 2004. "Money,
Unpublished. Price Dispersion and Welfare." Economic Theory 24
(4): 907-32.
Monnet, Cyril. 2006. "Private versus Public Money."
International Economic Review 47 (3): 951-60. Petrosky-Nadeau, Nicholas and Guillaume Rocheteau.
Monnet, Cyril, and Borghan N. Narajabad. 2012. 2013. "Unemployment, Financial Frictions, and the
"Whey Rent When You Can Buy? A Theory Housing of Market." Unpublished.
Repurchase Agreements." Unpublished. Pissarides, Christopher A. 2000. Equilibrium Unem
Mortensen, Dale T. 1982. "Property Rights and Effi
ployment Theory, Second edition. Cambridge, MA:
ciency in Mating, Racing, and Related Games." Massachusetts Institute of Technology.
American Economic Review 72 (5): 968-79. Plantin, Guillaume. 2009. "Learning By Holding and
Liquidity." Review of Economic Studies 76 (1):
Mortensen, Dale T. 1999. "Equilibrium Unemploy
ment Dynamics." International Economic Review395^12.
40
(4): 889-914. Prescott, Edward C. 2005. "Comments on 'Inflat
Mortensen, Dale T., and Christopher A. Pissarides.
Output, and Welfare' by Ricardo Lagos and
laume Rocheteau." International Economic Review
1994. "Job Creation and Job Destruction in the The
ory of Unemployment." Review of Economic Studies46 (2): 523-31.
61 (3): 397-415. Rauch, Bernhard. 2000. "A Divisible Search Model
Mortensen, Dale T., and Randall Wright. 2002. "Com of Fiat Money: A Comment." Econometrica 68 (1):
petitive Pricing and Efficiency in Search Equilib 149-56.
rium." International Economic Review 43 (1): 1-20.
Redish, Angela, and Warren E. Weber. 2011. "Coin
Nagel, Stefan. 2014. "The Liquidity PremiumSizes of and Payments in Commodity Money Systems."
Near-Money Assets." National Bureau of Economic Macroeconomic Dynamics 15 (Supplement SI):
Research Working Paper 20265. 62-82.
Nosal, Ed. 2011. "Search, Welfare, and the 'Hot Potato'
Renero, Juan-Manuel. 1998. "Unstable and Stable
Effect of Inflation." Macroeconomic DynamicsSteady-States
15 in the Kiyotaki-Wright Model." Eco
(Supplement S2): 313-26. nomic Theory 11 (2): 275-94.
Nosal, Ed, and Guillaume Rocheteau. 2011. Money, Renero, Juan-Manuel. 1999. "Does and Should a Com
Payments, and Liquidity. Chicago and London: MIT modity Medium of Exchange Have Relatively Low
Press. Storage Costs?" International Economic Review 40
Nosal, Ed, and Guillaume Rocheteau. 2013. "Pairwise (2): 251-64.
Ritter,
Trade, Asset Prices, and Monetary Policy." Journal of Joseph A. 1995. "The Transition from Barter
Economic Dynamics and Control 37 (1): 1-17. to Fiat Money." American Economic Review 85 (1):
Nosal, Ed, and Neil Wallace. 2007. "A Model of (the 134-49.
Threat of) Counterfeiting ."Journal of MonetaryRocheteau,
Eco Guillaume. 2000. "La quantite optimale de
nomics 54 (4): 994-1001. monnaie dans un modele avec appariements alea
Nosal, Ed, Yuet-Yee Wong, and Randall Wright. 2015a.toires." Annales d'Economie et de Statistique 58:
101-42.
"More on Middlemen: Equilibrium Entry and Effi
Rocheteau, Guillaume. 2008. "Money and Competing
ciency in Intermediated Markets. "Journal of Money,
Credit and Banking 47 (S2): 7-37. Assets under Private Information." Federal Reserve
Nosal, Ed, Yuet-Yee Wong, and Randall Wright. 2015b.Bank of Cleveland Working Paper 08-02.
"Who Wants to Be a Middlemen?" Unpublished.Rocheteau, Guillaume. 2011. "Payments and Liquidity
under Adverse Selection." Journal of Monetary Eco Big Problem of Small Change. Princeton and Oxford:
nomics 58 (3): 191-205. Princeton University Press.
Rocheteau, Guillaume. 2012. "The Cost of Inflation: A Schumpeter, Joseph A. 1954. History of Economic
Mechanism Design Approach." Journal of Economic Analysis. Oxford and New York: Oxford University
Theory 147 (3): 1261-79. Press.
Rocheteau, Guillaume, and Antonio Rodriguez-Lopez. Selgin, George. 2007. "Banking." Encyclopedia
2014. "Liquidity Provision, Interest Rates, and Britannica.
Unemployment." Journal of Monetary Economics Sethi, Rajiv. 1999. "Evolutionary Stability and Media of
65: 80-101. Exchange." Journal of Economic Behavior and Orga
Rocheteau, Guillaume, Peter Rupert, Karl Shell, andnization 40 (3): 233-54.
Randall Wright. 2008. "General Equilibrium with Shao, Enchuan. 2009. "Money and Barter under Pri
Nonconvexities and Money." Journal of Economic vate Information." BE Journal of Macroeconomics 9
Theory 142 (1): 294-317. (1).
Rocheteau, Guillaume, Peter Rupert, and Randall Shao, Enchuan. 2014. "The Threat of Counterfeiting
Wright. 2007. "Inflation and Unemployment in Gen in Competitive Search Equilibrium." Journal of Eco
eral Equilibrium." Scandinavian Journal of Economnomic Dynamics and Control 47: 168-85.
ics 109 (4): 837-55. Shapley, Lloyd S., and Martin Shubik. 1969. "On Mar
Rocheteau, Guillaume, Pierre-Olivier Weill, and Tsz
ket Games." Journal of Economic Theory 1 (1): 9-25.
Shen, Ji, Bin Wei, and Hongjun Yan. 2014. "Finan
Nga Wong. 2015. "Working through the Distribution:
cial Intermediation Chains in a Search Market."
Money in the Short and Long Run." National Bureau
of Economic Research Working Paper 21779. Unpublished.
Shevchenko, Andrei. 2004. "Middlemen." Interna
Rocheteau, Guillaume, and Randall Wright. 2005.
"Money in Search Equilibrium, in Competitivetional Economic Review 45 (1): 1-24.
Shevchenko, Andrei, and Randall Wright. 2004. "A
Equilibrium, and in Competitive Search Equilib
rium." Econometrica 73 (1): 175-202. Simple Search Model of Money with Heterogeneous
Rocheteau, Guillaume, and Randall Wright. 2009. Agents and Partial Acceptability." Economic Theory
24 (4): 877-85.
"Inflation and Welfare in Models with Trading Fric
Shi, Shouyong. 1995. "Money and Prices: A Model of
tions. In Monetary Policy in Low-Inflation Econo
Search and Bargaining."/ourna/ of Economic Theory
mies, edited by David E. Altig and Ed Nosal, 89-116.
Cambridge and New York: Cambridge University 67 (2): 467-96.
Press. Shi, Shouyong. 1996. "Credit and Money in a Search
Rocheteau, Guillaume, and Randall Wright. 2013. Model with Divisible Commodities." Review of Eco
"Liquidity and Asset-Market Dynamics." Journal of nomic Studies 63 (4): 627-52.
Monetary Economics 60 (2): 275-94. Shi, Shouyong. 1997a. "A Divisible Search Model of
Rocheteau, Guillaume, Randall Wright, and Sylvia Fiat Money." Econometrica 65 (1): 75-102.
Xiaolin Xiao. 2014. "Open Market Operations."Shi, Shouyong. 1997b. "Money and Specialization."
Unpublished. Economic Theory 10 (1): 99-133.
Rogerson, Richard. 1988. "Indivisible Labor, Lotteries Shi, Shouyong. 1998. "Search for a Monetary Propa
and Equilibrium." Journal of Monetary Econornics gation Mechanism." Journal of Economic Theory 81
21 (1): 3-16. (2): 314-52.
Rubinstein, Ariel, and Asher Wolinsky. 1987. "MidShi, Shouyong. 1999a. "Money, Capital, and Redistrib
dlemen." Quarterly Journal of Economics 102 (3): utive Effects of Monetary Policies." Journal of Eco
581-93. nomic Dynamics and Control 23 (4): 565-90.
Shi, Shouyong. 1999b. "Search, Inflation and Capital
Rupert, Peter, Martin Schindler, and Randall Wright.
2001. "Generalized Search-Theoretic Models of Accumulation." Journal of Monetary Economics 44
Monetary Exchange." Journal of Monetary Econom (1): 81-103.
ics 48 (3): 605-22. Shi, Shouyong. 2001. "Liquidity, Bargaining, and Mul
Samuelson, Paul A. 1958. "An Exact Consumption
tiple Equilibria in a Search Monetary Model." Annals
Loan Model of Interest with or without the Social of Economics and Finance 2: 325—51.
Contrivance of Money."Journal of Political Economy Shi, Shouyong. 2005. "Nominal Bonds and Inter
66 (6): 467-82. est Rates." International Economic Review 46 (2):
Sanches, Daniel R. 2015. "Banking Panics and Pro 579-612.
tracted Recessions." Unpublished. Shi, Shouyong. 2006. "Viewpoint: A Microfoundation
Sanches, Daniel R., and Stephen Williamson. 2010. of Monetary Economics." Canadian Journal of Eco
"Money and Credit with Limited Commitment and nomics 39 (3): 643-88.
Theft ."Journal of Economic Theory 145 (4): 1525-^9. Shi, Shouyong. 2008. "Efficiency Improvement from
Sanches, Daniel R., and Stephen Williamson. 2011. Restricting the Liquidity of Nominal Bonds." Journal
"Adverse Selection, Segmented Markets, and the of Monetary Economics 55 (6): 1025-37.
Role of Monetary Policy." Macroeconomic Dynamics Shi, Shouyong. 2014. "Liquidity, Interest Rates and
15 (Supplement S2): 269-92. Output." Annals of Economics and Finance 15 (2):
Sargent, Thomas J., and Frangois R. Velde. 2003. The 707-^9.
Siandra, Eduardo. 1990. "Money and Specialization in Jonathan A. Parker and Michael Woodford, 227-70.
Production." Unpublished. Chicago and London: University of Chicago Press.
Silva, Mario. 2015. "Search, Product Variety and von Mises, Ludwig. 1912. The Theory of
Monopolistic Competition in a Pure Monetary Econ Money and Credit. https://mises.org/library/
omy." Unpublished. theory-money-and-credit.
Silveira, Rafael, and Randall Wright. 2010. "Search and Wallace, Neil. 1980. "The Overlapping Generations
the Market for Ideas." Journal of Economic Theory Model of Fiat Money." In Models of Monetary
145 (4): 1550-73. Economies, edited by John H. Kareken and Neil
Smith, Adam. 1776. An Inquiry into the Nature and Wallace, 49-82. Minneapolis: Federal Reserve Bank
Causes of the Wealth of Nations. New York: Modern of Minneapolis.
Library. Wallace, Neil. 1981. "A Modigliani-Miller Theorem
Stiglitz, Joseph E. 1990. "Symposium on Rubbles." for Open-Market Operations." American Economic
Journal of Economic Perspectives 4 (2): 13-18. Review 71 (3): 267-74.
Sun, Hongfei. 2011. "Monetary and Fiscal Policies in Wallace, Neil. 1983. "A Legal Restrictions Theory of
a Heterogeneous-Agent Economy." Queens Univer the Demand for 'Money' and the Role of Monetary
sity Department of Economics Working Paper 1262. Policy." Federal Reserve Bank of Minneapolis Quar
Telyukova, Irina A., and Ludo Visschers. 2013. "Precau terly Review 7 (1): 1-7.
tionary Money Demand in a Business-Cycle Model." Wallace, Neil. 1997. "Short-Run and Long-Run Effects
Journal of Monetary Economics 60 (8): 900-916. of Changes in Money in a Random-Matching Model."
Telyukova, Irina A., and Randall Wright. 2008. "A Journal of Political Economy 105 (6): 1293-1307.
Model of Money and Credit, with Application to Wallace, Neil. 1998. "A Dictum for Monetary Theory."
the Credit Card Debt Puzzle." Review of Economic Federal Reserve Bank of Minneapolis Quarterly
Studies 75 (2): 629-47. Review 22 (1): 20-26.
Tirole, Jean. 1985. "Asset Bubbles and Overlapping Wallace, Neil. 2000. "A Model of the Liquidity Struc
Generations." Econometrica 53 (6): 1499-1528. ture Based on Asset Indivisibility." Journal of Mone
Townsend, Robert M. 1987. "Arrow-Debreu Pro tary Economics 45 (1): 55-68.
grams as Microfoundations of Macroeconomics." In Wallace, Neil. 2001. "Whither Monetary Economics?"
Advances in Economic Theory: The Fifth World Con International Economic Review 42 (4): 847-69.
gress, edited by Truman F. Bewley, 379^128. Cam Wallace, Neil. 2010. "The Mechanism-Design
bridge and New York: Cambridge University Press. Approach to Monetary Theory." In Handbook of
Townsend, Robert M. 1988. "Models as Economies." Monetary Economics, Volume 3, edited by Benjamin
Economic Journal 98 (Supplement): 1-24. M. Frieclman and Michael Woodford, 3-23. Amster
Townsend, Robert M. 1989. "Currency and Credit in dam: Elsevier, North-Holland.
a Private Information Economy." Journal of Political Wallace, Neil. 2013. "Comment on 'Moneyspots: Extra
Economy 97 (6): 1323-44. neous Attributes and the Coexistence of Money and
Trejos, Alberto. 1997. "Incentives to Produce Quality Interest-Bearing Nominal Bonds' by Ricardo Lagos."
and the Liquidity of Money." Economic Theory 9 (2): Journal of Political Economy 121 (4): 793—95.
355-65. Wallace, Neil. 2014a. "An Attractive Monetary Model
Trejos, Alberto. 1999. "Search, Bargaining, Money,with Surprising Implications for Optima: Two Exam
and Prices under Private Information." International ples." Federal Reserve Bank of Minneapolis Quar
Economic Review 40 (3): 679-96. terly Review 36 (1): 2-14.
Trejos, Alberto, and Randall Wright. 1995. "Search, Wallace, Neil. 2014b. "Optimal Money Creation in
Bargaining, Money, and Prices." Journal of Political'Pure Currency' Economies: A Conjecture." Quar
Economy 103 (1): 118-41. terly Journal of Economics 129 (1): 259-74.
Trejos, Alberto, and Randall Wright. 2016. "Search Wallace, Neil, and Ruilin Zhou. 1997. "A Model of a
Based Models of Money and Finance: An IntegratedCurrency Shortag e." Journal of Monetary Economics
ApproachJournal of Economic Theory 164: 10-31.40 (3): 555-72.
Uslii, Semih. 2015. "Pricing and Liquidity in Decen Wallace, Neil, and Tao Zhu. 2004. "A Commodity
tralized Asset Markets." Unpublished. Money Refinement in Matching Models." Journal of
Vayanos, Dimitri, and Pierre-Olivier Weill. 2008. "A Economic Theory 117 (2): 246-58.
Search-Based Theory of the On-the-Run Phenome Wallace, Neil, and Tao Zhu. 2007. "Float on a Note."
non ."Journal of Finance 63 (3): 1361-98. Journal of Monetary Economics 54 (2): 229-46.
Velde, Francois R., Warren E. Weber, and Randall Waller, Christopher j. 2011. "Random Matching and
Wright. 1999. "A Model of Commodity Money, withMoney in the Neoclassical Growth Model: Some
Applications to Gresham's Law and the Debase Analytical Results." Macroeconomic Dynamics 15
ment Puzzle." Review of Economic Dynamics 2 (1):(Supplement S2): 293-312.
291-323. Waller, Christopher J. 2015. "Microfoundations of
Venkateswaran, Venky, and Randall Wright. 2013.Money: Why They Matter." Federal Reserve Bank of
"Pledgability and Liquidity: A New Monetarist St. Louis Review 97 (4): 289-301.
Model of Financial and Macroeconomic Activity." Waller, Christopher J., and Elisabeth Soller Cur
In NBER Macroeconomics Annual 2013, edited by tis. 2003. "Currency Restrictions, Government
Transaction Policies and Currency Exchange." Eco 79-104. New York: Academic Press.
nomic Theory 21 (1): 19-42. Wiseman, Thomas. 2015. "A Note on the Essentiality
Wang, Liang. 2016. "Endogenous Search, Price of Money under Limited Memory." Review of Eco
Dispersion, and Welfare." Journal of Economic nomic Dynamics 18 (4): 881-93.
Dynamics and Control 73: 94-117. Wong, Tsz-Nga. 2015. "Monetary Exchanges in the
Wang, Weimin, and Shouyong Shi. 2006. "The Variabil Second-Best and Irreducible Cost of Inflation."
ity of Velocity of Money in a Search Model." Journal Unpublished.
of Monetary Economics 53 (3): 537-71. Wong, Tsz-Nga. 2016. "A Tractable Monetary Model
Weill, Pierre-Olivier. 2007. "Leaning against the Wind." under General Preferences." Review of Economic
Review of Economic Studies 74 (4): 1329-54. Studies 83 (1): 402-20.
Weill, Pierre-Olivier. 2008. "Liquidity Premia in Woodford, Michael. 2003. Interest and Prices: Founda
Dynamic Bargaining Markets." Journal of Economic tions of a Theory of Monetary Policy. Princeton and
Theory 140 (1): 66-96. Oxford: Princeton University Press.
Wicksell, Knut. 1911. Lectures on Political Economy, Wright, Randall. 1994. "A Note on Sunspot Equilibria
translated by E. Classen, Second edition. New York: in Search Models of Fiat Money." Journal of Eco
Kelley, 1967. nomic Theory 64 (1): 234-^41.
Williamson, Stephen D. 1986. "Costly Monitoring, Wright, Randall. 1995. "Search, Evolution, and Money."
Financial Intermediation, and Equilibrium Credit Journal of Economic Dynamics and Control 19 (1-2):
Rationing." Journal of Monetary Economics 18 (2): 181-206.
159-79. Wright, Randall. 2005. "Introduction to 'Models of
Williamson, Stephen D. 1987. "Financial Intermedia Monetary Economies II: The Next Generation'."
tion, Business Failures, and Real Business Cycles." International Economic Review 46 (2): 305-16.
Journal of Political Economy 95 (6): 1196-1216. Wright, Randall. 2010. "A Uniqueness Proof for Mon
Williamson, Stephen D. 1999. "Private Money." Jouretary Steady State." Journal of Economic Theory 145
nal of Money, Credit and Banking 31 (3): 469-91. (1): 382-91.
Williamson, Stephen D. 2006. "Search, Limited Par Wright, Randall, and Alberto Trejos. 2001. "Interna
ticipation, and Monetary Policy." International Eco tional Currency." BE Journal of Macroeconomics 1
nomic Review 47 (1): 107-28. (1).
Williamson, Stephen D. 2012. "Liquidity, Monetary Wright, Randall, and Yuet-Yee Wong. 2014. "Buy
Policy, and the Financial Crisis: A New Monetar ers, Sellers, and Middlemen: Variations on Search
ist Approach." American Economic Review 102 (6):Theoretic Themes." International Economic Review
2570-2605. 55 (2): 375-97.
Zhang, Cathy. 2014. "An Information-Based Theory
Williamson, Stephen D. 2014. "Central Bank Purchases
of Private Assets." Unpublished. of International Currency." Journal of International
Economics 93 (2): 286-301.
Williamson, Stephen D. 2015. "Keynesian Inefficiency
Zhou, Ruilin. 1997. "Currency Exchange in a Random
and Optimal Policy: A New Monetarist Approach."
Jotirnal of Money, Credit and Banking 47 (S2): Search Model." Review of Economic Studies 64 (2):
197-222. 289-310.
Zhou,
Williamson, Stephen D. 2016. "Scarce Collateral, the Ruilin. 1999. "Individual and Aggregate Real
Term Premium, and Quantitative Easing." Journal Balances
of in a Random-Matching Model." Interna
Economic Theory 164: 136-65. tional Economic Review 40 (4): 1009-38.
Zhou, Ruilin. 2003. "Does Commodity Money Elimi
Williamson, Stephen D., and Randall Wright. 1994.
"Barter and Monetary Exchange under Private Infor nate the Indeterminacy of Equilibrium?" Journal of
mation." American Economic Review 84 (1): 104—23. Economic Theory 110 (1): 176-90.
Williamson, Stephen D., and Randall Wright. 2010a.
Zhu, Tao. 2003. "Existence of a Monetary Steady State
"New Monetarist Economics: Methods." Federal in a Matching Model: Indivisible Money."Journal of
Reserve Bank of St. Louis Review 92 (4): 265-302. Economic Theory 112 (2): 307-24.
Williamson, Stephen D., and Randall Wright. 2010b.
Zhu, Tao. 2005. "Existence of a Monetary Steady State
"New Monetarist Economics: Models." In Handbook in a Matching Model: Divisible Money." Journal of
of Monetary Ecoiwmics, Volume 3, edited by Ben Economic Theory 123 (2): 135-60.
jamin M. Friedman and Michael Woodford, 25-96. Zhu, Tao. 2008. "Equilibrium Concepts in the
San Diego and Amsterdam: Elsevier, North-Holland. Large-Household Model." Theoretical Economics 3
Wilson, Charles. 1979. "An Infinite Horizon Model (2): 257-81.
with Money." In General Equilibrium, Growth, and Zhu, Tao, and Neil Wallace. 2007. "Pairwise Trade and
Trade: Essays in Honor of Lionel McKenzie, edited Coexistence of Money and Higher-Return Assets."
by Jerry R. Green and Jose Alexandre Scheinkman, Journal of Economic Theory 133 (1): 524-35.