1In any civil society, at whatever stage of its political, social and economic development, money is one of the most crucial instruments used in the coordination process between agents and between agents and their government. As a matter of fact, even Rousseau’s mythical “noble savage” or Smith’s “rude and primitive societies” could not function without recognizing at least the unit-of-account function of money. In today’s complex societies, in its multifarious financial forms, money has acquired such an all-powerful role in the co-ordination process between agents and countries that any dysfunctional tendency in the financial system precipitates the entire economic system into crises.
- 1 Ever since falling under Frank Hahn’s intellectual spell during his Cambridge years, the present au (...)
2Even if it is clear what this main-building block of any human society does (“Money is what money does” used to say F.A. Walker, 1877), historians, anthropologists and of course monetary theorists, are far from being clear about 1) the origin of money and, 2) the reasons why, today, legal fiduciary money without any intrinsic value and dominated in terms of rate of return by any other asset is nevertheless held (and very much so after the 2008 crisis) by agents and, hence, displays a positive price. This second issue is what monetary theorists call the “Hahn problem”1. The connection between the origin of money and its positive price is an obvious consequential problem: once you are in a position to explain and not simply to witness that money ‘exists’ you are then in a situation to look for a serious explanation of why it displays a positive value.
3Starr’s brilliant little volume is the latest of a long line of attempts to integrate money and price (general equilibrium) theory. Within 160 pages and 14 dense chapters, the author synthesizes forty years of careful and painstaking research. He sets them against the background not only of the tradition inaugurated by Walras and carried on by Pareto, Hicks, Patinkin, Samuelson, Clower, Hahn, Kyotaki and Wright (to name but a few) but also of the no-less famous conflict between trust and authority at the origin of money (Frankel, 1977). Hence, and before discussing Starr’s trading post model, it seems not out of place to recall briefly the analytical history of these two central riddles in monetary theory.
4Following Starr’s chapter 1, to start with a grossly simplified reminder of the two main philosophies behind the origin of money seems to be a useful starting point. On the one hand (and this is what is called the “trust” approach), there are those who argue that the value of a currency depends mainly, or exclusively (?), on the intrinsic value of the backing of that currency. Accordingly, the price of that particular asset is determined like the price of any other good. Hence money evolves as a private sector, market-oriented response to overcome via a price mechanism the transaction costs inherent in the lack of double coincidence of wants characterizing a barter economy. With Charles Goodhart (a distinction also adopted by Starr), let us call these economists Mengerians.
5Menger’s theory is a brilliant conjectural history of how money might have evolved as the result of individual choices by individuals seeking to reduce their transactions costs in an economic environment that is changing from subsistence farming into a market economy characterized by the division of labour. Some agents, realizing that certain commodities are easier to trade than others, would begin holding inventories of those goods beyond their immediate demands, thereby enhancing their ability to find trading partners. But by holding inventories of commodities endowed with a greater saleableness (to use Menger’s jargon), these individual agents would independently do two things. First they would make it even easier to trade in those commodities; and second they would induce other people to follow their example. As others followed their example, the costs of trading in those commodities originally identified as low cost commodities to trade would be reduced still more. This is an early description of what have recently come to be known as network effects or network externalities (of which Starr’s trading post model seems to be an extension). A good characterized by a network effect is a good for which the demand increases, as more people demand it. In his famous 1892 article (relentlessly quoted by Starr), Menger describes the historical process through which a commodity would emerge as money owing to the network effects inherent in being used as a medium of exchange; and not to any interference of any ‘non agent’ such as a sovereign, a government or a Central Bank.
6In a nutshell, within a decentralised economy, the origin of money is the unattended consequences of the uncoordinated behaviours of rational agents trying to minimise their transaction costs. Menger considers that the medium of exchange arises endogenously in a trading situation as the most convenient and liquid carrier of value between trades. The choice of a unique monetary medium reflects its inherent suitability in terms of low transaction costs (portability, verifiability, divisibility and durability to use Jevons traditional terms). Money is a commodity and its monetary functions (and its price) are an outcome of a market process. However, it raises immediately a nagging question: Should the sovereign not sanction this monetary medium? Should not authority support trust?
7On the other hand (and this is the “authority” approach), there are those who argue that a currency becomes money with a positive price primarily because the coins/notes bear the signs of the authority of a sovereign/state/government and not so much because they are made of gold, silver, copper or… paper! For these Cartalists (as they are usually called after Knapp’s terminology), the State has generally played a central role in the origin, the evolution and the use of money. Menger calls them the contractualist theorists because, for them, money derives “from a contract… an agreement between men or [from] the expression of their collective will (the law), especially with respect to money in its minted form” (Menger,  1976, 315). To put things even in simpler terms, for the Cartalists, a lump of gold is difficult to evaluate and hence problematic to use as an instrument to save on transaction costs. It is only because this lump of gold is struck with the face of a Roman emperor (or that of the Queen of England) that, thanks to a monopoly called seignoriage, this coin acquires a positive value by means of the information it carries (notably its ability to be used to pay taxes). As a matter of fact, besides the guarantee given to money via the seignoriage, a government with the power to tax seems to create the value of its fiat money not by announcing bluntly that its money is valuable but rather by accepting it for something that is valuable, i.e. the forgiveness of the tax obligations. As the Mengerian would argue, the State cannot create money by simply printing fiat money; but the Cartalists would reply that the State can create money with a positive value by simply considering itself this fiat instrument as genuinely valuable.
8Authority without trust is a non-starter; but trust without authority is unstable. Hence the origin and the positive value of money seem to rely on both authority and trust. Modern fiat money (or legal tender) is backed by the power of the government (not by gold) and the government’s ability to impose taxes payable (and often exclusively payable) in that fiat currency. But the positive value of this fiat money would vanish (and has actually often vanished) whenever the trust in the government’s authority disappears.
9Economic historians and anthropologists have shown time and again that the Cartalist approach preceded Menger’s argument, which is closely linked to the division of labour and the spread of market exchanges connected with the industrial revolution. But these same social scientists have also drawn the attention on the obvious many difficulties for the sovereign to maintain a level of confidence compatible with a stable value of money. And similarly, the recent financial crisis is yet another demonstration of the difficult relationship between a Mengerian financial system fraught with instabilities and the role played by the State’s authority to stabilise the value of money.
10Moreover, once the Cartalist link between money creation and taxation is understood, the move from a metallic currency (initially necessary to the Mengerian argument) to a fiat paper money becomes much easier to understand. Even if one could accept the Mengerian approach to the origin of metallic coins as currency, it is problematic (if not downright impossible as Hahn argued) to use this theory in its pure form to explain why agents should suddenly (like in August 1971 courtesy of President Nixon) all be willing to jump from using gold-backed paper money to paper notes backed by no specific assets and dominated in terms of rate of return by all other assets.
- 2 His first contributions to the topic date back to the early 1970s.
- 3 Historians of economic thought have still to explain how and why the most articulate Mengerians are (...)
11For nearly four decades,2 Starr has been one of the foremost and most articulate representatives of the Mengerian approach. It goes without saying—horresco referens—that, ever since Walras himself, the numerous attempts to formalise Menger’s initial intuition have been ceaselessly undertaken within a… Walrasian general equilibrium model.3 In other words, the Hahn problem is nothing but the straightforward question of finding room for a positive price to the nth good in a model in which the prices of the (n-1) goods are clearly positive but defined in a barter model in which no transaction technology and hence no transaction costs are defined: money is simply an adding-on variable introduced as a mere after-thought in a model that can logically dispense with it. As old Walras already said imprudently in 1874:
[Money] is in fact a practical simplification and a theoretical complication; for that reason once we will have defined clearly its role, we shall very often ignore it in the study of economic phenomena. (Walras, 1874, 175, same passage in OEC, VIII, 544).
12Extensively developed elsewhere (and by far more competent writers than this reviewer), a potted version of the genealogy of the Hahn problem might nevertheless be useful at this stage.
13From Walras (1900) down to Patinkin (1965), and with the notable help of Hicks (1935), the first idea was to introduce ‘simply’ money within the agents’utility functions. Nothing was said about the reasons why, within the Walrasian general equilibrium model, money had to display a positive marginal utility, and hence a positive price. After nearly seventy years of MIUF (money in the utility function) variations (including of course the real-balance effect), Patinkin eventually had to admit in the preface to his 1989 edition of Money, Interest and Prices that, “[m]ost discussions of monetary theory ... simply assume (as I too do in this book), that money exists and serves as a unique medium of exchange in the economy “ (Patinkin,  1989, xxix ; italics added).
- 4 Despite having the advantage of motivating the presence of money in the utility function by tying t (...)
14In 1990, Starr (in a joint paper with Ostroy) finally hammered the ultimate nail into the coffin of the Walras-Hicks-Patinkin tradition by arguing rightly that “putting money, even real money balances, into the utility function is an unreliable choice-theoretic short cut for modelling the transactions role of money” (1990, 6)4.
- 5 Even in the more sophisticated stochastic cash-in-advance models à la Svensson (1985) or Lucas and (...)
15Even if still used as a standard argument in many contemporary macro-models, the cash-in-advance or Clower constraint (1967) suffers from the same logical weakness. To add a cash constraint to the consumer’s budget constraint by assuming that “money buys goods and goods buy money; but goods do not buy goods” (Clower, 1967, 208-09) is hardly a profound theoretical argument to turn a barter model into a fully monetized general equilibrium model. Money has a positive value because… the theorist says so by postulating an exchange technology in which money displays a positive price!5
16While concentrating more on money as a store of value than as a means of exchange, and by introducing a formal period analysis, Samuelson’s 1958 intergenerational model “with the contrivance of money” appears at first to be ‘more realistic’ by introducing money as a social contract between generations allowing the reallocation of purchasing power over time. However, and among many reservations successively introduced, this model seems to suffer from a severe flaw linked to the infinite horizon necessary to the logical validity of a positive price for money: the positive price of money today is sustained by the expectation of a positive price in the future.
- 6 Any good graduate textbook in monetary theory would provide the reader with all the relevant techni (...)
17Last but one in this petit tour for schoolchildren,6 with their search model, Kyotaki and Wright initiated in 1989 the last great wave of interest in solving the Hahn problem. The main issue this time is to give, within a general equilibrium framework, a rational explanation of the emergence of a monetary technology.
18As a matter of fact these models try to endogenise the process through which a unique and socially accepted means of exchange is adopted. A substantial part of their central logic is based on the probabilities agents attribute to the money they (wish to) hold to be later on exchanged against goods. If this probability is high, the agent will accept money in exchange for the goods/services he produces; if the probability is low, the agent will not accept to part with his goods/services. In an equilibrium position, no one has an interest not to accept money if all other agents are ready to use it. For the first time, the positive value of money results not exclusively from individual agent’s constraint maximisation, but also depends upon a social agreement on the (monetary) exchange technology. But do these models fulfil their promises?
19To put things in a brutally simplistic way, in these search models, the selection of a monetary equilibrium seems to depend exclusively on the exogenously given value of the probability of acceptance of money. Walras’s mythical deus ex machina auctioneer (who, through the tâtonnement mechanism alters prices on which individual agents have by definition no influence) is simply replaced by the exogenously and parametrically determined probability of acceptance of money. In both cases, either the price adjustment mechanism or the definition of the probability of acceptance of money are determined by an ‘institution’ which is, by definition, a non-agent. In both cases, the hoped-for decentralisation postulated by a general equilibrium model is brutally replaced by a deus ex machina which, in the case of search models, could well be a government endowed with a monopoly on the supply of legal fiduciary money.
20To suggest models integrating an exchange technology allowing a positive price for fiduciary money is clearly a theoretical advance; but, once again, what is left of such a step forward if the choice of technology (money or barter) is left unexplained? Or, at least left to an exogenously determined probability? Following an argument similar to that applied to the general equilibrium highly problematic stability theorem, one is sent back to a question much more complex than ‘simply’ demonstrating why, when in equilibrium, money may have a positive value. Since the choice of a monetary or non-monetary technology of exchange is a collective (and not a individual) choice, should one not refer to a disequilibrium social process rather than to a simple definition of equilibrium conditions? Hence, and despite Kyotaki and Wright’s clever theoretical prowess, money appears once again as a social and conventional institution: the social convention at the origin of the positive value of fiduciary money must be defined before the social game starts.
21In the book under review, Why is there Money?, Starr puts eventually together nearly four decades of work on his trading post model—the last attempt so far to solve Hahn’s problem.
22Within the dense but short compass of 150 pages, Starr’s difficult, but fascinating book, tries precisely to solve two questions. As the subtitle makes clear, the author intends to answer first the question Why is there money? by offering “Walrasian General Equilibrium Foundations” to monetary theory. Moreover, by so doing, and second, he also ambitions to reconcile the Mengerians and the Cartalists who “of course are both correct” (7). Or put in slightly different words, with his trading post model seen as an extension of a standard “parsimonious” (14) Arrow-Debreu model, Starr is determined not only to solve Hahn’s problem but also to elucidate the century-old trust vs authority riddle. The remainder of this review article is devoted to a discussion of Starr’s demanding model and to a tentative assessment of his twin-theoretical ambitions.
- 7 Segmenting the trading opportunities is in fact an alternative way to reopening trade over time (Ha (...)
- 8 Again, and in his own words, and contrary to the Arrow-Debreu scheme, in Starr’s model there is, by (...)
- 9 Within a pure exchange economy, with linear transaction costs and when double coincidence of wants (...)
23In contrast with the Arrow-Debreu model where all goods trade simultaneously in a single market at one point in theoretical time (with a single budget constraint for each household and a single expression of profit for firms), Starr’s “trading post model7 segments transactions at each trading post to take place in commodity pairwise trading markets, with a budget constraint applying to each trader’s transaction at each trading post” (15). This will appear later as a first and most crucial ex ante hypothesis.8 Clearly, such market segmentation creates a demand for a ‘carrier’ of value to bridge the gaps between the different trades/markets. In short, Starr adds to the standard Arrow-Debreu model (in which ex definitio transaction costs are nil) a resource-using transaction technology made up of two elements: exchanges have to take place at commodity-pairwise trading posts and—in the Hicksian tradition—trade implies transaction costs. Prices at each trading post are defined as bid and ask rates of exchange between the two goods traded at the post; and, logically, the bid-ask spread measures the transaction costs at that trading post. The most liquid good displaying the lowest bid-ask spread (i.e the lowest transaction costs) is destined to be used as the universal medium of exchange and is called… money! In a nutshell, as Starr puts it vividly, “liquidity creates money” (43). Hence, the price system seems to provide not only information on relative scarcity and desirability of goods but it also prices liquidity. Clearly, network externalities encourage all transactions to use this single common medium of exchange. Hence, Starr’s trading post model describes an economy in which trade is not only mutually advantageous but where direct (barter) trade between agents at most trading posts may even be more costly than indirect trade trough the lower transaction cost instrument. In a fascinating chapter 6, Starr demonstrates how the uniqueness of the medium of exchange and the exclusive use of monetary transactions (even in the presence of double coincidence of wants) are explained by economies of scale in transaction costs.9
- 10 Liquidity resulting from high trading volumes, accordingly a high-volume good becomes money leading (...)
24To put things in a slightly different way, for any two distinct commodities there is a separate trading post for their pairwise exchange. Starting with n commodities there are n (n − 1)/ 2 separate trading posts with bid ask prices reflecting transaction costs. When most trading posts are active in market equilibrium, the economy operates like a barter economy. But when, in equilibrium, most posts are inactive, and activity is concentrated on (n-1) posts trading a single good against the remaining (n-1) other goods, then the economy is purely monetary with the single actively traded good as commodity money. In Starr’s own words: “Money is a natural monopoly; in equilibrium it will be unique” (58).10
- 11 Starr’s next sentence hammers down even further the fact that the holy grail of using price theory (...)
25The budget constraint applies to each pairwise transaction separately. Exchange is a resource using activity. Transaction costs are evident to buyers and sellers from a spread between bids and ask prices. At each of many separate transactions, the requirement that payment be made for acquisitions implies a role for a carrier of value between trades (there are consecutive transactions after all). Households and firms choose a trading plan among trading posts that will optimise their utility subject to the trading post balance constraints at prevailing prices. Scale economy in transaction costs leads to a unique common medium of exchange: “one commodity has, by assumption, distinctively low transaction cost” (43; italics added). The critical assumption behind the Clower constraint is no longer an assumption: “even in the presence of double coincidence of wants, trade is monetary” (137), because the uniqueness of a positively valued money in equilibrium “is logically derived from price theory”(p. 69; italics added).11 QED.
26So far, and as Starr makes amply clear, within the Mengerian logic, money is one of the n goods, i.e. the common medium of exchange resulting from the working of trading post models is commodity money. In chapter 8, Starr develops the idea that his trading post monetary model also provides for the use of a unique monopolistic, government-issued, fiduciary money used in most if not all transactions. Accordingly, Starr introduces ex hypothesis a government endowed with a monopoly to print fiat money and to declare it acceptable in payment of taxes (probably the second most crucial hypothesis). The economies of scale in transaction costs inherent to the use of money in trading post models and the government’s large size compared to all other agents lead the economy “to a corner solution where government money is the natural monopoly medium of exchange” (83). Menger is reconciled with Knapp; or, alternatively, the unique use of government-issued fiat money seems to be at last given proper micro-foundations.
27For Starr, these ‘elementary’ economic conditions permit the model to conclude, in a market general equilibrium, that (to paraphrase pages 13 and 142):
Trade is monetary. One side of all transactions is the economy’s common medium of exchange.
Money is unique: the existence of a unique common medium of exchange in equilibrium is logically derived from price theory.
Even transactions suitable for barter solutions, where the two parties have reciprocal demands and supplies are conducted with money.
As a natural monopoly, money is government-issued fiat money, trading at a positive value though it conveys no direct utility.
28And these central conclusions are clearly for Starr consequences of his model and not simple assumptions as posited by earlier theorists. This expanded Arrow-Debreu model emphasizes transaction costs and the multiplicity of separate pairwise transactions each fulfilling a budget constraint. For Starr, it allows the theory to conclude “the monetary structure of trade is shown to be a consequence of price theory. The medium … of exchange is … a consequence of general equilibrium, not a separate assumption” (110). Or alternatively, payment takes monetary form, and “as a common medium of exchange, money is not an assumption but a conclusion of the theory based on elementary assumptions” (123; italics added). Hence, Why is there Money? seems to fulfil the long stated need for a microeconomic foundation of monetary theory. But does Starr’s book really foot this very tall bill? How elementary are his assumptions? Don’t these elementary assumptions on the exchange technology imply, by definition, a positive price for money?
29First of all, and less optimistically than in his first thirteen chapters, Starr cautiously and rightly remarks in his concluding chapter that the conclusions reached in chapters 3-8 “are examples, not general results … which only provide clues to our comprehension” (138-39) of the questions at hand. In other words, and as for most previous attempts at solving Hahn’s problem, any expanded version of the Arrow-Debreu model, even if parsimoniously and properly set up, is bound to lead to the conclusion that payment takes a monetary form. To paraphrase Edgeworth’s quip about Walras’s tâtonnement mechanism, Starr’s explanation (that emphasizes transaction costs and the multiplicity of separate pairwise transactions each fulfilling a budget constraint) is a solution and not the solution to the Hahn problem. Money might no longer be an assumption but a conclusion of a theory based on elementary assumptions. But those elementary assumptions cannot be but a set of conditions on the transaction technology that allows the theorist to conclude that there is a common medium of exchange, that payments take monetary form and that money has a positive value. As for any general equilibrium, “the clues provided to our comprehension” are purely heuristic but do not in any way offer the only “true” solution to Hahn’s problem. Moreover, just like for any classical Arrow-Debreu model, to use old Walras’s dictum, “pure theory does not expect any confirmation from the real world”. Hence, Starr’s attempt at solving Hahn’s problem and at reconciling the Mengerians and the Cartalists is no more, but no less as well, than a clue to our comprehension of the workings of a monetary theory of exchange.
- 12 Epitomised by the Mantel-Sonnenschein-Debreu 1970s negative results.
30In such models, there is of course the all-important convergence/stability issue. At least since Walras and Edgeworth, any student of general equilibrium knows the mind-boggling problem linked to tâtonnement (a sheer algorithm without any connection with ‘real’ markets) and the related issue of convergence towards a vector of equilibrium prices.12 Starr concedes explicitly that his chapter 7 on “Monetization of General Equilibrium” is “just an example” of very limited generality. He even goes on to admit candidly that, in terms of convergence to monetary equilibrium, “it is difficult to see how pure theory can give us fully general results” (141).
31In parallel, the issue of exchange decentralisation raised in search models comes back with a vengeance. The trading post model is a set of institutions; the non-agent that partially and exogenously replaces the traditional fictitious auctioneer. Moreover, the exogenously given large size of the government (once again a non-agent in a general equilibrium setting) and its natural monopoly on the medium of exchange is bound to lead to the hoped for corner solution.
32In more general terms, in order to provide microeconomic foundations for the use of a unique monopolistic government-issued fiduciary money used for virtually all transactions (and to reconcile en passant the Mengerians and the Cartalists), like all previous attempts at solving Hahn’s problem, Starr’s trading post model seems to keep adding up assumptions on a standard Arrow-Debreu model until one solution (and not the solution) is found. But is such an extended Arrow-Debreu model still connected in a substantial way with the highly parsimonious initial structure on which rely most of its optimising results?
33For more than a century, these various attempts to give microeconomic foundations to money (including of course Starr’s trading post model) bring the reader to a crucial meta-theoretical distinction, which is the analytical parallel to the opposition between the Mengerian and Cartalist conception of money. Either the validity of these various models relies on ad hoc hypotheses and, hence, money is simply a theoretical afterthought added to a general equilibrium model fundamentally erected on barter assumptions; or, money is no longer a variable to be simply added to real variables but an intrinsic component of monetary exchanges the technology of which it is necessary to explain and not simply to postulate. In other words, either the theoretical logic of an exchange economy can be apprehended without money (the neutrality tradition with an inessential money); or, money is an essential characteristic of the fundamentals of an exchange economy and, thus, should be, from the very beginning, part of its analytical representation (the essentialist tradition).
34The growing dis-interest of economic theorists for finding a solution to Hahn’s problem, or alternatively the growing acceptance of any ad hoc solution to the positive price of money in most contemporary macro-models, is the sign of the progressive degeneration of this research programme and, more fundamentally, the progressive realisation that no general solution can be found within the narrow limits of pure economic theory. As a matter of fact, the suspicion that historians, economic historians, anthropologists and political philosophers have more interesting things to say than economists on the reasons why “there is money” goes probably a long way to explain the progressive waning of this research agenda.
35Eventually, and once again, money appears as a social and conventional institution; and the rational choice between a barter, mixed or monetary economy seems to result from ex ante rules of the game; the choice of which cannot be the outcome of an ‘equilibrium’ game. As in any social game, as in any kind of social contract, the social convention at the origin of the positive value of fiduciary money must be defined before the game starts, even if this ex ante rule will certainly not be always adhered to during the social game.
36Even if a century of economic theory has shown unsurprisingly that authority without trust is a non-starter and that trust without authority is unstable, a genuine reconciliation between the Mengerians and the Cartalists cannot be left to economic theorists alone. Long ago, none other than Pareto himself, “a nominalist among the nominalists” in economic theory, had already fully understood the ambivalent status of money within economic theory:
Theories have some connections with the other facts of the environment from which they grow out. If one analyses them impartially, without taking sides, if in other words one makes a theory of theories, one will note first that they can be split into two broad classes. In the one class, theories converge indefinitely towards some limit; one can say, using the mathematical language, that they have an asymptote… In the other class, theories instead of converging toward their asymptote do oscillate perpetually in between two boundary points… As regards the theory of money, one can say roughly that the two boundary points of oscillation are on the one hand the theory of commodity-money and on the other hand the theory of fiat money. The former corresponds to an exclusive economic point of view, while the latter corresponds to a political and social point of view. Any concrete phenomena being likely to be considered from both points of view, it is pretty clear how and why the theory of money continually oscillates between both extremes. (Pareto, , as in OC, IX, 175 and OC, XXI, 152–3; italics added
37Hence, even if the various attempts to solve the Hahn problem are of central heuristic importance to understand the logic of a monetary economy, none of them can reasonably be expected to explain alone the reasons why there is money and why money displays a positive value.