-->THE EMERGENCE OF FIAT MONEY:
A RECONSIDERATION
Kevin Dowd
One of the fundamental questions in monetary economics is why
fiat money has value: Why do rational agents trade real resources for
intrinsically worthless pieces of paper? Monetary economists have
long understood that part of the explanation relates to the superiority
of a monetary equilibrium over a barter one. However, it is one thing
to explain why fiat money is better than barter, and quite another to
explain how fiat money actually emerges. Recognizing this point, a
number of recent studies (e.g., Kiyotaki and Wright 1991, 1992, 1993;
Ritter 1995; and Williamson and Wright 1995)1 have sought to explain
the emergence of fiat money by means of a hypothetical direct jump
from barter to a fiat money equilibrium.2
This paper suggests that these attempts are fundamentally misconceived.
They suffer from three main problems. The first is the “start
problem”3—that is, the difficulty of ensuring an initial demand for
fiat money balances. If fiat money is ever to emerge from barter,
someone must be the first to exchange real goods for pieces of paper
money that, by definition, do not provide any direct consumption
Cato Journal, Vol. 20, No. 3 (Winter 2001). © Cato Institute. All rights reserved.
Kevin Dowd is Professor of Financial Risk Management at Nottingham University Business
School.He thanks Charles Goodhart and George Selgin for very helpful correspondence on the
issues discussed in this paper. The usual caveat applies.
1Some of the issues raised in this literature are also discussed by Goodhart (1997) and
Selgin (1993, 1994, 1997).
2This research program implicitly assumes that agents act noncooperatively and decide
independently of each other whether to accept intrinsically worthless assets from an initial
barter equilibrium. This implies that there is no institutional mechanism to coordinate
agents’ move from barter to a fiat money equilibrium. If such a mechanism existed, we
could easily explain the emergence of fiat money simply by invoking it to coordinate the
move to a fiat money equilibrium. However, any such explanation is obviously spurious. We
must therefore rule out such mechanisms if we are to provide a satisfactory explanation for
the emergence of fiat money.
3I thank George Selgin for suggesting this terminology.
467
services to the holder.4 The problem is that no agent will ever give up
real goods for intrinsically worthless pieces of paper, unless he believes
that others will accept them too, and an agent living in a barter
economy has no reason to expect them to. Each agent would therefore
refuse to accept fiat money, because he expects others to refuse
it as well. People will not move away from barter equilibrium, precisely
because it is an equilibrium.
Second, even if we could solve this first problem, there is the
difficulty that agents generally face a choice of potential monetary
objects, in which case we get a multiplicity of potential equilibria
involving the use of any one or more of these objects as money.5 The
equilibrium that actually results will then depend on agents’ expectations
about the acceptability of different objects, and it is not easy,
to say the least, to manipulate these expectations to produce any one
particular equilibrium. It will be fortuitous if all agents happened to
choose only one object, and even more fortuitous if they all chose
government fiat money.
Finally, the predictions of these models are at odds with the historical
evidence. Fiat money did not in fact evolve in the way these
models postulate: nowhere did fiat money ever emerge by means of
a great leap forward from barter. Nor did fiat monies ever emerge out
of thin air. Instead, fiat monies have always developed out of some
previously existing money. If we are to explain the emergence of fiat
money, we must therefore explain how it emerges from previously
existing money, not from barter.
The “Start Problem” in Recent Models of
Fiat Money
The “start problem” is best illustrated using a simple version of the
Kiyotaki-Wright model. Following Kiyotaki and Wright (1992: 19-
4This problem is quite different from and logically prior to the “terminal problem” that has
traditionally preoccupied monetary economists (Gale 1992: 226). The latter problem relates
to the point that people must have some confidence in the future value of money if money
is to have a positive value now (Cass and Shell 1980: 252). Solving this terminal problem
requires that we give a plausible reason to expect money to have a positive future value, but
even if we solve this problem—and doing so is not easy (Faust (1989: 872, 879)—any
explanation that says that fiat money has value now because it is expected to have value in
the future begs the main issue (i.e., how fiat money gets accepted at all). The “start
problem” is discussed further in Selgin (1993).
5Most authors respond to this problem by assuming that there is only one particular object
that could satisfy a public demand for money, and then go on to identify this object with
government-issued fiat money. However, such an assumption merely dodges the issue.
CATO JOURNAL
468
20), suppose there is a large number of agents who consume a fraction
x of a large number of perfectly storable consumption goods.
Each good is produced by a fraction x of agents, and agents do not
consume the goods they themselves produce. In period 1, a fraction
(1-M) of these agents is randomly endowed with consumption goods,
and the rest with intrinsically worthless paper money. Each agent has
one unit of a good or one unit of money. Agents meet bilaterally and
at random each period, and trade only if it is in their mutual interest.
An agent who consumes then immediately produces another good, so
all agents always have either one unit of a good or one unit of money.
An agent who accepts a good is assumed to incur a positive transaction
cost, but one who accepts money incurs a transaction cost of zero.
Each agent now chooses a trading strategy to maximize his expected
utility net of production and transaction costs, and we seek Nash
equilibria in which agents take each others’ trading strategies as given.
We also make the simplifying assumptions that all agents are identical
and that all goods are equally acceptable in trade. It is then immediately
apparent that an agent will only accept a good in exchange if
he wishes to consume that particular good. There is no point incurring
transactions costs to obtain a good that one does not wish to
consume, and is no better for trading purposes than the good one
already has. It also follows that trade will only take place if there is a
double coincidence of wants, and a double coincidence of wants
occurs with probability x2. Barter thus involves fairly high search
costs.
The next step is to see if we can induce any agent to accept money
in exchange for real goods. If _ is the probability that a typical agent
attaches to other agents accepting money from him, then Kiyotaki and
Wright (1992: 20; 1993: 66-68) show that our typical agent will always
accept money if _ > x, but always refuse it if _ < x. (The intuition is
simple: given that our agent does not wish to consume fiat money, it
only makes sense to accept it if it is more acceptable to others than
the good he already has, i.e., if _ > x.) Hence, the decision whether
to accept money depends on _, the probability of other agents accepting
money from him. The equilibrium consequently depends on
expectations. If a typical agent expects a sufficiently large number of
others to accept money, he will accept it himself, and we will get a
monetary equilibrium in which everyone accepts money. But if our
agent believes that an insufficient number of other people will accept
money, he will refuse it himself, in which case no one accepts it and
we get a nonmonetary equilibrium (i.e., barter). Expectations about
the acceptance of money have a self-fulfilling character.
As an aside, the monetary equilibrium also has the property that it
THE EMERENGE OF FIAT MONEY
469
relieves agents of dependence upon a double coincidence of wants
and thereby produces a considerable increase in the efficiency of
exchange. An agent who accepts money in a monetary equilibrium
can expect to obtain the good(s) he wants next period with probability
x, rather than the probability x2 he faces under barter. The agents
involved would clearly prefer the monetary equilibrium to the nonmonetary
one.
Nonetheless, there is a very awkward corollary: If agents start from
barter, they have no way to get to the monetary equilibrium. The
reasoning is simple: In a barter equilibrium agents do not expect each
other to accept fiat money, so _ = 0; yet one or more agents must be
willing to accept fiat money in exchange for goods if the economy is
ever to move from barter, and we already know that an agent will only
accept fiat money if _ > x. However, the conditions _ = 0 and _ >
x cannot be simultaneously satisfied for any positive value of x. Hence,
the economy can never move from the barter equilibrium to the
monetary one. The economy remains stuck in barter, and it would
remain stuck even if the agents involved all recognized that they
would be better off in the monetary equilibrium. They remain stuck
because they are effectively prisoners of barter expectations: They
would accept money if their expectations of each others’ willingness
to accept money could be shifted, but it is not individually rational to
accept money because there is no way to change those expectations.
The fiat money is stuck at its launching pad (Selgin 1994: 809-11;
Dowd 1996: chap. 10).
This example illustrates very clearly that we must do more than
merely demonstrate that the monetary equilibrium exists and involves
higher utility than barter, if we wish to show that the economy will
reach a monetary equilibrium in which fiat money has a positive
value.6 The monetary equilibrium can exist and have higher utility
than barter, and yet the economy can still remain stuck in barter. If
the economy is ever to get out of barter, we must also demonstrate
that one or more agents living in barter will at some point have a
private incentive to accept money, that is, we must overcome the start
problem. Unfortunately, this problem cannot be overcome because
the attempt to do so involves a contradiction: agents will only accept
6Iwai (1988: 1-3) also recognizes the implications of the bootstrap nature of the monetary
equilibrium. As he puts it, “any student of the speculative philosophy knows [that] there is
a wide gap between the potentiality and the actuality, and, however tempting it is, we
cannot immediately jump from [a] proposition about the potential ubiquity of money to the
assertion that ’therefore, money evolves naturally in any economy.’ The logic of money
should not be confused with the genesis of money” (emphasis added).
CATO JOURNAL
470
fiat money if they think that a sufficient number of other agents will
accept it, and yet the very fact that they are starting in barter equilibrium
implies that they do not expect others to accept it.7
Which Money Will Agents Choose?
There is another problem. Even if we can explain why agents might
start accepting intrinsically worthless monetary objects in a state of
barter, we must also explain why their demand for such objects translates
into a demand to hold government fiat money. A priori, there is
no reason to suppose that they have only one possible object to choose
from, or to suppose that all such agents would choose the same asset.
Moreover, even if they did choose the same object, there is no reason
to suppose that it would be government fiat money.8
To see what is involved, suppose there are only two possible fiat
monies, a green one and a blue one. We can assume that the government
prefers the greenback, but the two monies are otherwise the
same. Each agent now has to decide which money to accept, or
whether to accept both, and we get a variety of possible equilibrium
outcomes (Matsuyama, Kiyotaki, and Matsui 1993).9 One outcome
arises where each agent happens to think that the greenback will be
acceptable, but the blueback will not be. The resulting equilibrium is
then one in which the greenback is accepted as money and the blueback
is not. This is the outcome that is normally assumed to occur
when the demand for money is presumed to translate into a demand
for government fiat money. However, there is also an equilibrium
outcome in which the blueback is accepted as money and the green-
7This criticism also applies to Ritter (1995), despite his claim to model the emergence of fiat
money. While Ritter talks of the emergence of fiat money, his analysis only addresses the
question of whether a monetary equilibrium exists, and he provides no analysis at all of how
the economy actually reaches this equilibrium starting from a state of barter. Ritter thus
falls foul of the same start problem that affects the rest of this literature.
8One standard response is to assume that agents select as their fiat money an asset in limited
natural supply, such as cowry shells. However, this response begs the point and fails to
explain why agents use cowry shells instead of available alternatives such as snail shells. It
also ignores the point that cowry shells were never adopted as some equivalent to fiat
money. They were valued first as ornaments, and then adopted for monetary purposes. The
fact that most modern economists have little use for cowry shell ornaments should not blind
us to the fact that cowry shells were an example of commodity money.
9Matsuyama, Kiyotaki, and Matsui (1993) use a version of the Kiyotaki-Wright model to
examine which fiat money is used, in a framework in which each region has its own
currency, and find that there is a variety of possible equilibria depending on agents’ (selffulfilling)
expectations about the acceptability of the two currencies. Kiyotaki and Wright
(1993: 74-75) also have a model in which either or both monies could circulate, but
explicitly examine only the dual-currency equilibrium.
THE EMERENGE OF FIAT MONEY
471
back is rejected, as well as a dual-currency equilibrium in which both
green and blue monies circulate as currency. It would therefore be
fortuitous if agents happened to converge on one single money, and
even more fortuitous if they happened to converge on the government’s
preferred money, assuming, of course, that they were free to
choose. Instead, the natural outcome is a dual-currency equilibrium
in which both green and blue assets circulate as money. The odds
against any one single-currency equilibrium also rise with the number
of choices that agents have. If agents have a choice of n assets, we get
possible equilibria in which any of the n assets, or any combination of
the n assets, circulate as money. The odds in favor of multiplecurrency
equilibria therefore rise with the number of potential monetary
assets.10
The outcome depends on agents’ expectations, and yet there is no
obvious way in which expectations can be coordinated to ensure that
any particular outcome actually results. In practice, governments can,
and do, attempt to promote their preferred money in various ways.
First, the government might attempt to promote its preferred
money by announcing that it will only deal in that particular money
(e.g., when trading on its own account). However, unless the government
is very large relative to the rest of the economy (in which case
the probability of a private trader meeting a government trader is very
high), this measure will not generally suffice to ensure that private
agents accept only the government’s preferred money. The government
might encourage the adoption of the green currency, but it
cannot guarantee that only the green money will be adopted.
Second, the government could promote its preferred money by
announcing that it will deal in that currency alone when collecting
taxes. However, making a particular money uniquely acceptable for
tax payments does not in itself enhance its acceptability for nontax
transactions: the link between acceptability for tax purposes and commercial
acceptability still remains to be established. In any case, there
is also the problem that no centralized agency with tax-collection
powers actually exists in the Kiyotaki-Wright type of model environment,
and it is difficult to see how one could be introduced without
10This poses a serious problem, because multiple-currency equilibria are the exception
rather than the rule historically. Most of the time most historical agents made use of one
particular currency, and it was generally efficient that they did so (e.g., because of lower
accounting costs). Where agents did use multiple currencies, there also tended to be
observable reasons for doing so (e.g., traders might operate both locally and internationally,
and use different currencies for each type of trade).
CATO JOURNAL
472
undermining the noncooperative framework within which agents are
presumed to operate.
Finally, the government might seek to promote its favored money
by imposing legal restrictions against the use of alternatives. Perhaps
the most obvious means of enforcing legal restrictions in the assumed
model environment is for the government to send out enforcement
agents disguised as legitimate traders and give these agents powers to
punish those agents who offer them trades involving blue money (e.g.,
by confiscating their holdings). More fundamentally, the argument
that the government could use its powers to promote its favored
money also fails to address the question of what powers the government
might have in the Kiyotaki-Wright world. If all agents are decentralized
traders, then who or what is the government and where
do its powers come from? This problem is usually ignored in this
literature. The only exception seems to be Ritter (1995) and his attempts
to deal with it merely highlight the problems involved. In his
model, the issuer of money is assumed to have monopoly access to a
technology that enables member-agents to costlessly coordinate with
each other, and the monopoly itself is simply taken for granted. However,
his optimum is then one in which all agents become members
of the issuing coalition (Ritter 1995: 145). All agents therefore have
implicit access to a costless coordination technology, and Ritter fails
to explain how this technology fits alongside the assumed lack of
coordination that gives rise to the use of money in the first place.11
Historical Evidence
Recent literature on the emergence of fiat money is also historically
falsified. The historical experience indicates that fiat money actually
emerged from previously convertible currency, and did not emerge
directly from barter. The convertible currency itself arose from a
commodity currency, which in turn arose from barter.12 The process
therefore began with the emergence of commodity money from barter.
In time, the commodity exchange medium gave way to a convertible
paper currency that displaced commodity money as the dominant
medium of exchange. The paper currency was anchored to the
11In any case, unless the government resorts to draconian punishments, we still cannot rule
out a blue-money or mixed-money equilibrium in which agents regard the threat of punishment
as a form of taxation they must live with (Kiyotaki and Wright 1993: 74-75). Again,
the government lacks the power to ensure that only its preferred asset is chosen as money.
12The main points summarized here can all be verified by consulting any of the standard
histories of early money and banking (e.g., Carlisle 1901, Burns 1927, and Quiggin 1963).
THE EMERENGE OF FIAT MONEY
473
value of the earlier commodity money by virtue of the issuers’ commitment
to redeem it on demand and, while banks economized on
the reserves held to redeem their currency issues, there is no evidence
of any natural market tendency toward a displacement of the
convertibility guarantee as such.13 However, governments then
started to intervene to suspend the convertibility guarantee—for example,
the British government did so in 1797 and the U.S. government
in 1861, both times prompted by the fiscal exigencies of war.
The suspensions were meant to be temporary and convertibility was
subsequently restored, but after alternating periods of convertibility
and temporary government-imposed suspension, convertibility was
finally suspended in the early 1970s with no pretense of any intention
to restore it. Today the major currencies of the world have became
true fiat currencies with no link at all to any commodity anchor.
Nor have fiat monies ever emerged de novo, by fiat command or by
other means. Selgin (1994) reviews the experience of a variety of new
fiat currencies starting from John Law’s fiat currency in the early 18th
century, and could not find a single case where a fiat currency had
been successfully launched de novo without first tying it in to some
existing currency or currency unit:
The [new currency] units (or fractional counterparts) had in every
case been in use to some extent in the marketplace prior to the
reforms embodying them in new currencies. The units’ definitions
had, moreover, to be rendered operational through some kind of
convertibility, typically involving one or more fixed or nearly fixed
exchange rates to foreign monies. Legal tender and public receivability
provisions were, on the other hand, either nonexistent or of
obviously secondary importance. Experience therefore supports the
conclusion... that a new fiat money must be operationally linked to
some established money if it is to achieve a positive value [Selgin
1994: 823, emphasis added].
13Inconvertible currency did not therefore emerge spontaneously from previously convertible
currency, as suggested by McCallum’s comment that “since it is costly for banks to
maintain the requisite commodity reserves and since the convertibility option is very rarely
exercised, banks might eventually do away with this guarantee” (McCallum 1985: 29) as if
it served no particular purpose. He went on to suggest that the abolition of convertibility by
the Bank of England in 1797 might be an example of such a process. However this example
does not bear out this interpretation. Far from being the outcome of a spontaneous market
process, the suspension of 1797 is a classic example of the government intervention mentioned
in the text. The British government had already borrowed more from the Bank than
the Bank directors wanted, and a point came when a relatively minor shock set off a run that
the Bank did not have the resources to meet. The government then intervened to prohibit
the Bank from making further specie payments. The role of the government in this suspension
could hardly be more clear-cut.
CATO JOURNAL
474
Conclusion
The essential problem with the recent literature is that it attempts
to tell an impossible story: it attempts to explain how fiat money can
emerge directly from barter. The response is that fiat money cannot
emerge directly from barter—and, as a matter of historical fact, did
not do so. Of course, no one is denying the fact that we all now use
fiat money, but any explanation for the use of fiat money must be
consistent with the historical evidence. I would therefore suggest that
the correct story is that barter gives rise to commodity money, commodity
money then gives way to convertible paper currency, and the
currency subsequently remains convertible for as long as competitive
forces are allowed to maintain convertibility. The reason is that convertibility
offers the public a guarantee of the value of the money they
use. The public values this guarantee, and so competitive pressures
force the banks of issue to maintain it (Dowd 1996: chap.1). Convertibility
then ends when the government intervenes to suppress
convertibility. Only at that point does the currency become a genuine
fiat currency—a currency of a type made possible only by the fiat
power of the state. Any sensible story about the emergence of fiat
money must therefore come to terms with two fundamental facts,
both of which are ignored by recent literature that attempts to explain
the emergence of fiat money by means of a great leap forward from
a barter to a fiat monetary equilibrium: the fact that fiat currencies
always arose out of previously convertible currencies, and the fact that
fiat currencies always arose from subsequent state intervention to
make those currencies inconvertible.
References
Burns, A.R. (1927) Money and Monetary Policy in Early Times. New York:
Alfred E. Knopf.
Carlisle, W.W. (1901) The Evolution of Modern Money. London: Macmillan.
Cass, D., and Shell, K. (1980) “In Defense of a Basic Approach.” In J.H.
Kareken and N. Wallace (eds.) Models of Monetary Economies, 251-63.
Minneapolis: Federal Reserve Bank of Minneapolis.
Dowd, K. (1996) Competition and Finance: A Reinterpretation of Financial
and Monetary Economics. London: Macmillan.
Faust, J. (1989) “Supernovas in Monetary Theory: Does the Ultimate Sunspot
Rule Out Money?” American Economic Review 79: 872-81.
Gale, D. (1982) Money: In Equilibrium. Cambridge: Nisbet/Cambridge University
Press.
Goodhart, C.A.E. (1997) “Two Concepts of Money, and the Future of Europe.”
Financial Markets Group Special Paper 96. Financial Markets
Group, London School of Economics.
THE EMERENGE OF FIAT MONEY
475
Grossman, H.I. (1991) “Monetary Economics: A Review Essay.” Journal of
Monetary Economics 28: 325-45.
Hayashi, F., and Matsui, A. (1996) “A Model of Fiat Money and Barter.”
Journal of Economic Theory 68: 111-32.
Iwai, K. (1988) “The Evolution of Money: A Search-Theoretic Foundation of
Monetary Economics.” Center for Analytic Research in Economics and
Social Sciences Working Paper #88-03, University of Pennsylvania.
Kiyotaki, N., and Wright, R. (1989) “On Money as a Medium of Exchange.”
Journal of Political Economy 97: 927-54.
Kiyotaki, N., and Wright, R. (1991) “A Contribution to the Pure Theory of
Money.” Journal of Economic Theory 53: 215-35.
Kiyotaki, N., and Wright, R. (1992) “Acceptability, Means of Payment, and
Media of Exchange.” Federal Reserve Bank of Minneapolis Quarterly
Review 16 (Summer): 18-21.
Kiyotaki, N., and Wright, R. (1993) “A Search-Theoretic Approach to Monetary
Economics.” American Economic Review 83: 63-77.
Matsuyama, K.; Kiyotaki, N.; and Matsui, A. (1993) “Towards a Theory of
International Currency.” Review of Economic Studies 60: 283-307.
McCallum, B.T. (1984) “Bank Deregulation, Accounting Systems of Exchange
and the Unit of Account: A Critical Review.” Carnegie-Rochester
Conference Series on Public Policy 23: 13-46.
Menger, C. (1892) “On the Origin of Money.” Economic Journal 2: 239-55.
Quiggin, A.H. (1963) A Survey of Primitive Money: The Beginning of Currency.
London: Methuen.
Ritter, J.A. (1995) “The Transition from Barter to Fiat Money.” American
Economic Review 85: 134-49.
Selgin, G.A. (1993) “Still Searching for a Theory of Fiat Money: A Comment
on Kiyotaki and Wright.” Mimeo. Department of Economics, University of
Georgia.
Selgin, G.A. (1994) “On Ensuring the Acceptability of a New Fiat Money.”
Journal of Money, Credit, and Banking 26: 808-26.
Selgin, G.A. (1997) “Network Effects, Adaptive Learning, and the Transition
to Fiat Money.” Mimeo. Department of Economics, University of Georgia.
Williamson, S.D., and Wright, R. (1994) “Barter and Monetary Exchange
under Private Information.” American Economic Review 84: 104-23.
CATO JOURNAL
476