This paper is also available on SSRN as a PDF.
Monetary theory has been concerned since Menger (1892) with the question, just what feature of the world makes money necessary? The answer depends crucially on what we take to be the relevant analytical alternative to monetary exchange.
In what we will call the Coasean tradition, the alternative is a frictionless general equilibrium (GE) economy, in which direct barter is sufficient to achieve a Pareto-optimal allocation of resources. The key difference of the real world, of course, is the presence of frictions, or transaction costs, and a great part of the literature is occupied with identifying the set of necessary and sufficient transaction costs that result in monetary exchange. The GE economy’s ability to make do with pure barter, where one good is as good as any other for exchange, lends itself to a “veil” view of the money economy in which the pattern of economic activity is best understood as isomorphic to some underlying or ideal pattern of barter relationships, however impeded its realization may be in fact by the transaction costs that necessitate money in the real world.
It must be emphasized, however, that the central question of the Coasean approach – “What specific frictions distinguish the world from a GE model?” – is not the same as the question, “How did money in fact arise?”. This latter question we will identify with the “historical” approach. Rather than beginning with a frictionless model and introducing frictions to bump it out of a barter equilibrium, we may get a better sense of the relevant alternative(s) to monetary exchange by taking our starting point from the actual historical precursors to monetary exchange. From this perspective, money cannot be understood as a veil over some ideal pattern of relationships; instead, money is constitutive of those relationships.
Unfortunately, the two approaches have not to this point been clearly distinguished and separated from other relevant questions. Without the benefit of a detailed history, Menger’s conjectural history of the evolution of monetary exchange out of barter has been taken both by friends and foes as an actual history.1 Similarly, without the benefit of a detailed theory, proponents of the historical approach are often derailed by historical accidents and elevate them as necessary conditions (cf. Salter & Luther 2014). For this reason the historical approach is often identified with a “state money” paradigm (Goodhart 1998; Wray 2004), in which money gains currency and value from the diktat of some central authority. This paper does not directly engage with the state money literature, but rather offers an alternative approach: a historically grounded theory and a theoretically grounded history, at the same time evolutionary and historical, unlike the evolutionary-but-ahistorical Mengerian approach and the historical-but-constructivist state money approach. It does, however, take the anthropological evidence marshaled by state money proponents seriously, as well as bringing to bear additional evidence from cognitive science and sociobiology.
The first section argues for the inadequacy of the Coasean approach to monetary theory, at least in its current incarnations, in a world where the division of labor is limited by the extent of the market. The next two set out a framework for classifying exchange institutions as involving a tradeoff between fixed and marginal cognitive costs: up-front investment in increasingly higher fixed-cost exchange institutions lowers marginal costs of exchange, resulting in increasing returns to the division of labor as the size of the market expands. Those costs account for the persistence of more or less direct barter in more primitive societies (i.e. societies with a small scope of trade), despite the inevitability of monetary exchange that seems to be a feature of Menger’s story in “On the Origin of Money”. The remaining sections identify the relevant fixed costs of money and its institutional substitutes throughout the world. The paper concludes with some speculation on the future development of exchange institutions.
Monetary exchange arose out of a long chain of less satisfactory solutions to the problems it solves. There are two basic issues with interpreting this history using a Coasean model. First, such a model almost by necessity makes monetization a one-step process. Such analytically soluble models excel at isolating an equilibrating force in a well-defined problem context, such as the emergence of monetary exchange out of atomistic barter. Motion through time from one equilibrium to another, except as a smooth transition along a small and limited number of dimensions, quickly makes an analytical model unwieldy and insoluble. In contrast, the increasing-returns model presented in this paper has no such neat result as the emergence of a single medium of exchange out of the interaction of optimal trading strategies. Instead, it sets up a broad framework to encompass heterogeneous transitions among various equilibria, each of which is perhaps individually amenable to future analytical models, but which together are too multidimensional to be tractable using the simultaneous-equation techniques characteristic of the Coasean tradition.2
Second, the canonical set of Coasean models are motivated not by explanatory power for historical events, but by current monetary policy concerns. Their initial conditions do not correspond in any meaningful way to the conditions out of which money in fact emerged. Indeed, they are not meant to do so. For this reason they should be interpreted as fundamentally synchronic rather than diachronic, despite their featuring “emergence” of money as an equilibrium result.
Before presenting a truly diachronic alternative, it will be worth highlighting the necessity of such in light of several extant classes of Coasean models.
Search-theoretic models identify the relevant transaction cost that results in monetary exchange as search costs. In order to make a mutually agreeable exchange with a second party, I must not only have something they want, but they must also have something I want. In other words, there must be a double coincidence of wants (DCW). The cost of arranging a DCW was identified as a ground for money by Jevons (1875), set into a dynamic model of the emergence of money by Menger (1892), and formalized by Kiyotaki & Wright (1989, 1993).
Menger’s illustration3 is of a man who has produced some wares, and now seeks to bring them to market in exchange for other items he desires – which suggests a barter society that has already achieved some degree of specialization in production. Kiyotaki & Wright (1993) postulate an exogenous parameter x ∈ [0,1] that “captures the [inverse of the] extent to which real commodities and tastes are differentiated.” This parameter also represents the probability of one agent desiring the good offered by another random agent. But because trade requires both agents to desire the good the other is offering, the probability of executing a successful trade with any given agent will be x². This baseline case of random meeting we will call “atomistic” barter.
It is this assumption of pre-existent specialization that makes the search-theoretic story “ahistorical”.4 A sufficiently specialized society that somehow found itself without a medium of exchange would surely converge upon one. But specialization is itself a function of exchange institutions – an insight as old as economics itself. As Adam Smith famously put it in a chapter title (1776, bk. 1, ch. 3), “the division of labor is limited by the extent of the market”, i.e. by the extent of the money economy. This will be the key insight that organizes our alternative account. Because x is a decreasing function of the number of regularly traded goods, to the extent that arranging trades is costly, members of such a society will rarely find it worthwhile to specialize in the first place.
Walrasian models of money differentiate themselves from search-theoretic models by the assumption of organized markets, in which DCW can be more or less assumed. Banerjee & Maskin (1996), formalizing the argument in Alchian (1977), show that in such a world, information asymmetries over goods quality and adverse selection problems can lead to the emergence of a single medium of exchange, even without search frictions.
Like the search-theoretic models, Walrasian models begin with something like atomistic barter. However, in addition to assuming pre-existent specialization, these models also have to assume the pre-existence of well-organized markets. If Smith’s dictum about the division of labor is valid, this is an even more heroic and question-begging assumption. In the presence of such frictions as characterize the real world, not least of which is travel time between markets, a commonly accepted medium of exchange is a prerequisite for organized markets, not the other way around.
In addition, to posit a well-organized market without contract enforcement also begs the question. The model stipulates no contracts, i.e. no ability to commit to time-separated credit exchange. And yet, for the operation of actual markets, the range of “defecting” behavior is much larger than contract non-fulfillment. Importantly, it includes a basic respect for property. One may wonder, then, why the same implicit institution that supposedly causes Walrasian agents to respect property does not also cause them to respect contracts.
There is, of course, no evidence of any pre-monetary society using anything like atomistic barter as a primary exchange institution.5 From the perspective of a modern market economy, a Walrasian situation without contracts or third-party enforcement, which rule out exchange on credit, might seem appropriately primeval to serve as the initial condition. However, the historical record suggests that incentive-compatible credit exchange was in fact the first problem to be solved in the chain of exchange institutions that led to money.
It must again be emphasized that such considerations are not meant as criticisms of the models themselves, but of their application to history. Both Banerjee & Maskin and Kiyotaki & Wright are motivated not by an interest in the origins of money, but by current policy concerns such as the costs of inflation. We have no quarrel with their use for this purpose. Nevertheless, a truly diachronic model of the emergence of monetary exchange must begin with the initial conditions provided to us by history.
In order to see the problems monetary exchange solves compared to its its actual historical precursors, we must first have a sense of the basic impediments to the regularization of exchange that any exchange institution must solve. We begin therefore with autarky: the size of the market being zero, and therefore without specialization. There is, of course, little doubt as to the historicity of autarky: social cooperation is relatively rare among animals, and a great number of mammals – including some primates – are essentially solitary.
It is very nearly a first principle in economics that voluntary exchange is a pure gain over autarky. Because it benefits both parties, it may seem not to require any particular explanation. Indeed, Smith (1776) again referred to the “propensity to truck, barter, and exchange” as a basic feature of human nature.
Consider, then, the prospects of two identical and autarkic individuals (call them Crusoe and Friday) considering an exchange. If exchange necessarily precedes specialization, then the prototypical exchange cannot be between separate goods, as in the Coasean models. Instead, our prototype must be the exchange of undifferentiated services. More specifically, services performed on the body of the other party – for example grooming or a back rub, which must be performed sequentially – are a more likely prototype for the first exchanges than other services that might be performed contemporaneously, because the former are easier to monitor.6
Crusoe and Friday are, of course, by hypothesis, both better off exchanging back rubs. But both would also rather receive a back rub and then go swimming without returning the favor. Similarly, conditional on not receiving a back rub, each party would rather avoid the hassle and just stay at the ocean. Despite the potential gains from trade, swimming is a dominant strategy. The basic problem with getting trade started in the first place is that the time-separated quality of service exchange poses a prisoner’s dilemma and opens the door for defection. Without this first step, there is no opportunity for the development of specialization and goods exchange.
In this more general sense, exchange is primarily impeded – not by search costs, since at this stage any individual is more or less as good as any other at performing an available service – but by a commitment problem. The pertinence of the dilemma can be seen even more readily in the animal kingdom, where the fitness costs of being the “dupe” preclude most forms of cooperation and exchange. Even in this setting completely lacking in specialization, if Friday cannot commit to return the back rub, there is no exchange.7
The distinctiveness of human social behavior is not, therefore, as is usually assumed, merely that man is able to appreciate gains from trade whereas animals are too dull to do so. Rather, humans are distinctive in their ability to commit on a non-rational basis to act contrary to their own narrow interest. It is not because humans are more rational than animals, but indeed because they are less rational in this narrow sense that they can be trusted, on the whole, to reciprocate cooperative behavior even at some cost to themselves.
With autarky as the backdrop, the problem facing any exchange institution will be: how to overcome the prisoner’s dilemma of reciprocal altruism to make exchanges self-enforcing? There are factors that both help and hinder as we move toward a consideration of more familiar large-scale human interaction. Repeated play allows Crusoe to punish Friday if he defects. However, this equilibrium also becomes more and more difficult to monitor and sustain as Friday’s brothers, Wednesday and Thursday, arrive to the island. As Bowles & Gintis (2008) show, repeated play on its own is generally insufficient to sustain cooperation in groups larger than about four.
Cooperation in a group larger than four individuals therefore requires an exchange institution, which consists of norms and strategies that facilitate:
Much of the literature on early human development has focused on the punishment aspect, especially as used to suppress overt violence, and its historical trajectory toward more centralized administration. Comparatively little has focused on either the use of punishment in the context of exchange institutions (which may or may not overlap with the mechanisms and strategies used to punish overt violence), or the accounting and publication necessary to orient such punishment. Widespread “mooching” is at least as deleterious to social cooperation as overt violence, but is generally punished in a different manner. We therefore focus the following taxonomy on these aspects.
The problem with cooperation in large groups is that the accounting necessary to prevent defection from becoming a dominant strategy requires increasing cognitive capacity as group size increases (Dunbar 1992, 1995). Humans do have an advantage here among animals: from their very emergence as a separate species, anatomically modern humans have been characterized by efficient facial recognition and a relatively large memory for keeping track personal obligations. Still, impressive as it is compared to other primates, humans’ natural capacity for accounting obligations allows them to maintain group sizes averaging a few dozen, and maxing out around a few hundred.
However, humans do have another advantage: the ability to extend their cognition “outside the skull”, so to speak, by imbuing elements in their environment with symbolic significance (Clark & Chalmers 1998). Writing is the prototypical example: we can extend our effective memory by replacing huge quantities of information with a reference; i.e. with the knowledge of where to look. Similarly, investment in the capacity to offload the accounting of exchanges into the environment – to publish trading histories in an objective form – has the potential to vastly increase the potential scale and complexity of social organization.
We can use this idea of extended cognition to categorize exchange institutions in terms of a fixed cost and a marginal cost. The environment does not present itself to us in such a fashion that we may costlessly use it to extend our cognition. To do so requires convergence on certain norms, symbols, and meanings. The necessary learning and habituation are what we will call the fixed costs of the institution, taking biological and evolutionary costs as given.8 To put it another way, any pattern of behavior can in principle be identified as having a joint basis in biology and culture. The cultural pattern may build upon, interact with, or even override biological patterns (cf. Hayek 1960, p. 93). In this sense, the greater the cultural basis of an institution, the more distant from any biological basis, and the greater the extent to which it conflicts with or overrides biological predilections, the greater the institution’s fixed cost.
The marginal cost of an institution, on the other hand, is the cognitive cost of accomplishing a particular exchange. Importantly, the marginal costs are a diminishing function of the fixed costs: the more extensive the symbolic convergence, the less cognitively costly the accounting on the margin, and the larger the potential scale of human organization. To anticipate the following sections somewhat, organizing exchange using only the natural cognitive faculties (what we will call “natural credit”) involves a relatively high marginal cost for each relationship – say, 1/150 of the brain’s social memory. Thus, despite its low fixed cost, high marginal costs of maintaining ongoing relationships mean that natural credit as a method of exchange and social organization exhausts cognitive limits at a very low scale. Monetary exchange, by contrast, involves a high fixed cost but extremely minimal marginal costs. Any given transaction requires relatively little thought to accomplish, and once completed, can be safely forgotten in a way that would destroy the self-enforcing quality of natural credit exchanges.
The formulation of exchange institutions in terms of fixed and marginal costs suggests an analogy to capital theory, and a sense in which institutions can legitimately be called “social capital”.9 Lachmann (1956) asks the same question of physical capital that we have just asked of institutions:
It will not pay to install an indivisible [i.e. high-fixed-cost] capital good unless there are enough complementary capital goods to justify it. Until the quantity of goods in transit has reached a certain size it does not pay to build a railway. A poor society therefore often uses costlier (at the margin) means of transport than a wealthy one. . . . [N]ew indivisibilities account for the increasing returns [to capital].
Similarly, even for a society on the brink of subsistence, investing in high-fixed-cost institutions is not “worth it” – i.e. is not an equilibrium outcome – until a sufficient volume of exchange has built up to amortize the cost. Tribal foraging societies, limited as they are in scale, simply cannot sustain enough exchange to make the fixed cost of monetary exchange – namely, a system of writing and mathematics – worth the cost to develop and sustain.
The analogy to capital theory is also appropriate in taking the form of a mutual feedback – in Lachmann’s example between investment in high-fixed-cost transportation capital and the demand for transport; in our case between investment in high-fixed-cost exchange institutions and the volume of exchange. This feedback is characteristic of increasing returns models with multiple equilibria, and makes it impossible to give either element strict causal priority. While the exchange institution limits the division of labor and the volume of exchange, so that investment in higher-fixed-cost exchange institutions makes possible increases in the volume of exchange, increases in the volume of exchange also make investment in higher-fixed-cost exchange institutions “worthwhile” (cf. Kaldor 1972). In this case the uniderectionality of both Say’s Law and it’s Keynesian opposite with respect to the volume of exchange will be analytically inadequate.
The zero-fixed-cost case, relying entirely on the brain’s innate capacities for accounting and publication, will correspond to the earliest known form of social organization in cognitively modern humans, as well as the form characterizing the most primitive hunter-gatherer societies today. We will call it the “natural credit” economy.
Natural credit is the terminus of a more or less straightforward progression from autarky through dyadic and small-group cooperation as cognitive capacity increases. Most importantly, the time-separated quality of exchange is preserved through the whole chain, even as the exchange of fishing for gathering becomes reified into an exchange of fish for berries – hence the “credit” aspect (Wray 2004), ruled out by hypothesis in Coasean models.10
The key institution that facilitates both accounting and punishment is the tribe, within which both functions are performed in a relatively decentralized manner. Regular interaction with a more or less definite group of people facilitates “gossip”, through which reputational information is regularly disseminated. Punishment, likewise, consists of diffuse social pressure that escalates into ostracism. Each member has familiar knowledge of all or most other members,11 and the “trading history” of each member – unlike under atomistic barter – is more or less common knowledge.
With the regular publication of reputations and the collective exercise of social pressure on defectors, the tribe is actually able to harness the time-separated quality of early exchange to create mutual rents among tribe members sufficient to enable them to commit to cooperate with one another for the foreseeable future. Conversely, self-amortizing exchange – by giving up the leverage of repeated play within a local structure – would make it relatively infeasible to punish defection.
The marginal cost of exchange under natural credit, of course, is extremely high, which places a hard limit on both group size and the division of labor. Because cooperation in a reputational game depends on knowledge of the identity and history of each potential trading partner, it breaks down when a sufficient number of players are anonymous. The scale of natural credit organization is therefore limited by humans’ cognitive capacity for keeping track of relationships and obligations, thought to average around 150 relationships (Dunbar 1992, 1995). Indeed, natural credit societies do not get far past 150 members before free riding and conflict become endemic, and such communities have regular institutions for cleaving once they get too large. It is an anthropological truism that tribes and villages will fission after reaching a few hundred people – about twice Dunbar’s number – in the absence of investment in higher-fixed-cost “integrative institutions” (cf. Bandy 2004; Chagnon  2009), more on which below. As a community approaches the size where any two individuals are as likely to know (i.e. to have ongoing mutual obligations with) each other as not, free-riding becomes the dominant strategy, conflict (or the anthropological literature’s somewhat more antiseptic term, “scalar stress”) rends it in two, and one half is forced to establish a camp elsewhere.
Similarly, the division of labor under natural credit remains mostly ad hoc (cf. Hooper et al. 2015), with little capacity for permanent specialization beyond gender roles. There are simply not enough potential trading partners for it to be worth it for any of them to relinquish food production and to specialize in, say, pottery manufacture.
Such societies must form the starting point for a theory of money. Unlike models of atomistic barter, where commitment is impossible and unlimited cognitive power allows for costless optimization, early human society is in fact characterized by robust institutions for long-term commitment, and extremely limited cognitive accounting capacity. Though there are no markets per se in a natural credit society, even such a basic organization of in-kind multilateral exchange offers sufficient incentive for defection that institutions for accounting and punishment are necessary to make it work. Evidence suggests that tribal institutions, by making exchange ongoing and personal, do indeed enable sufficient commitment to sustain cooperation at a scale of a few dozen to a few hundred individuals. As compared with Menger’s story, these credit relationships render in-kind and time-separated barter sufficiently tolerable to forestall the spontaneous development of money.
The problem in breaking out of the natural credit equilibrium will be to expand the size of the market to include anonymous others. The first step will be the formation of trade networks, which allow trade to take place on the basis of group affiliation rather than personal knowledge.
There are two basic preconditions for the formation of trade networks. First, agents must be organized into groups within which some combination of internal (e.g. Greif 1993) and external (e.g. North, Wallis, & Weingast 2009, ch. 2) forces suppresses defection. These mechanisms are analogous to prosocial preferences on the individual level. Tribal institutions accomplish just this in the natural credit economy: they allow the group to act as a “superorganism” (Wilson & Gowdy 2015) and ensure its trustworthiness vis-a-vis outside groups.
Second, agents trading with the group must posses the cognitive faculty of hierarchical syntax, whereby groups of objects can be “chunked” and represented internally as a single object (Potts 2000, pp. 68, 116). This allows agents to respond to trustworthiness with trust, and in turn makes trustworthiness a worthwhile investment. To the extent that the group’s defection-suppressing mechanisms are reliable, outside agents can use information about an agent’s class or tribe as a substitute for personal knowledge. Indeed, Marwick (2003) provides archaeological evidence that the development of inter-tribal trading networks arose with the development of the language faculty. He argues,
The ability to express symbolic categorizations of social systems allows individuals to identify and interact with unrelated individuals in terms of symbolic categories rather than as unique individuals. This allows for relationships based on mutual rights and obligations rather than the histories of interpersonal relations that require renegotiation at each encounter.
To the extent that group or class affiliation can substitute for personal knowledge in the evaluation of trustworthiness, higher fixed-cost institutions can regularize ad hoc trading into permanent inter-group specialization, with the trading relationships between groups analogous to the relationships between individuals in the previous section.
Inter-group trade is the prototype of self-amortizing exchange. Indeed, to demand repayment at the time of exchange is seen as a peculiar form of trade reserved for “outsiders”, and protects to some degree against defection in a situation where the two groups are not in regular enough contact to make something like natural credit relationships feasible on an inter-group scale.
Consider a society consisting of inland and coastal villages. Because the ecologies of the two village types are so distinct, the gains from trade are readily apparent: it pays to organize an ongoing division of labor between the coastal villages, which provide fish, and the inland villages, which provide vegetables. But even with self-amortizing exchange, the two villages still face a problem similar to Crusoe and Friday: permanent specialization implies reliance on the other party, which allows it to “exploit” the other for better terms of trade (cf. Williamson 1983). In order to make specialization worth the cost, the village must be assured of a steady – or at least predictable – demand for its product. Even if the gains from trade are substantial, a society just embarking on the division of labor will have extremely limited means of smoothing consumption. For a risk-averse agent, an increase in the variance of his consumption is only worth it for a substantially higher mean (see the discussion in Wiessner 1977).
Our question, then, will be: how can a tribe or village organize itself internally in order to credibly commit to forswear opportunism vis-a-vis other tribes or villages, and what forces lead it to do so?
As a matter of history, the transition out of the paleolithic natural credit era and into the neolithic customary exchange era was marked by a number of significant cultural shifts: first, the emergence of belief in powerful and moralistic gods rather than immanent and self-interested spirits, with whom interaction is mediated by a specialized priest or priestly class. A credible signal of belief in such gods – the more vengeful the better (Shariff & Norenzayan 2011) – serves to commit the community of believers to forswear opportunistic behavior with trading partners (Leeson & Suarez 2015). Indeed, the more permanent a particular trading relationship, the more likely it is to involve ceremony of some sort or another, and the more elaborate the ceremony is likely to be (e.g. Malinowski 1926, ch. 3).
Ceremony and religion are examples par excellence of investment in symbolic convergence. It is not difficult to imagine forces driving investment in such institutions without anyone deliberately aiming at the result – as indeed we must, for an instrumental approach to such institutions would more than likely destroy any commitment power. If we assume some degree of random variation in exchange institutions, population pressure will select for higher fixed cost institutions, simply because a mismatch where the size of the society significantly outpaces the exchange institution will imperil that society’s continued survival. A society with numerous groups in close proximity must find a way to live peaceably with most of them most of the time, or face perpetual conflict and possibly extinction.12 In this sense, whatever other aspects it may have, any integrative institution that results in expanding the feasible social scale must necessarily be an exchange institution – i.e. it must involve symbolic convergence for the purposes of accounting and/or punishment.
As evidence for this account, a striking feature of contemporary natural credit societies without permanent trading networks is that they persist in areas with little to no space constraint.13 Environmental pressure, then – whether high birth rates or more intense competition for land – is a likely candidate for the kind of exogenous synoetic pressure that would push a society out of the natural credit equilibrium, and into either extinction or a higher fixed-cost exchange institution.
In a fairly small or sparsely populated society, the probability of failing to find trading partners willing to buy one’s wares in sufficient quantities to safeguard against starvation is high enough that investment in most specialized human capital is not worth the time. As the size of the “market” increases with the development of inter-group trade and groups are able to settle into permanent patterns of specialization, individuals are able to find productive ways to specialize in intra-group roles as well. Increased demand for particular products arises with specialization and production. It is at this point that the process of monetization begins to look properly Mengerian: each trader’s desire to hold some stock of the most saleable good leads the group to converge spontaneously on a single medium (or a common set of media) of exchange. But note how far along the process is already by this point: specialization is very far from an analytical primitive!
This process of convergence on a medium of exchange is an important fixed cost and a precondition for a great deal of further investment. Specialization does not yet enable purely anonymous exchange, but less personal knowledge is necessary between individuals. For this reason agents can trade with each other as “outsiders” – i.e. with self-amortizing exchanges, although natural credit remains important within an agent’s close circle. The investment of individuals in permanent specialization creates sufficient rents to bind them to their community and prevent defection as long-term tribal obligations decline in importance.
With the increased importance of impersonal and self-amortizing exchange, the medium of exchange takes on additional importance: calculation becomes important in order to ensure against exploitation.14 Natural credit societies are characterized by languages with no precise numbers beyond one, two, three, many; and in at least one case no exact numbers at all (Everett 2005). As specialization requires more and more transactions to be made on an outsider basis, there is pressure for the emerging administrative and merchant classes to develop or adopt systems of writing and arithmetic. Evidence suggests that both were originally developed for accounting use; indeed, the earliest known example of human writing is the record of a debt (Wray 2004).
Writing and arithmetic, like the division of labor, are important drivers of the feedback loop between population density and monetization, and the large institutional fixed costs of a proper money economy that has mostly displaced long-term credit relationships. The development of accounting and mathematics makes it easier to identify profitable opportunities for specialization, opportunities which could be seen only impressionistically in a natural credit economy. This specialization in turn increases the demand for and usefulness of the medium, and so on. The development of indirect exchange provides a society not only with the original impetus to abstract away from the countable numbers into a system of accounting relationships in order to reckon terms of trade more precisely, but also with a division of labor fine enough to support occupations in which an abstract system of mathematics will pay.
Indirect exchange’s major cognitive advance over natural credit is that the accounting is embodied in a definite quantity of some medium of exchange. The accounting apparatus by which humans organize the division of labor is now offloaded into a reified good which serves not only as a unit of account, but also as a store of value – i.e. an indication of a positive account in the balance of reciprocal altruism (cf. Kocherlakota 1996). As Simmel ( 1978, p. 128) argued, “the ability to construct symbolic objects attains its greatest triumph in money.”
The crucial transition from natural credit with trading networks to customary exchange comes with the geographic coexistence of permanently specialized trading classes, which entails more or less regular contact among the various members of those classes, as opposed to the mediation of trade through tribal representatives.15 Customary exchange is characterized by three important aspects of symbolic convergence:
The institutions that support anonymous exchange within a given group are a straightforward development from those that support trading networks. The priestly class that formerly suppressed defection within one group now grows into a bureaucracy and suppresses defection in many groups. The groups themselves commingle, but rely on this external bureaucracy to punish the defection of other groups in order to maintain trustful relations with them.
Because these class divisions are the repository of information on the accounts necessary to sustain a division of labor for the great majority of people, they cannot be modified except at great cost; certainly not at the whim of any single individual in an occupation. The convergence upon their meaning is the institution’s fixed cost. As Leijonhufvud (1977) noted,
In largely non-monetary economies, important economic rights and obligations will be inseparable from particularized relationships of social status and political allegiance and will be in the same measure permanent, inalienable, and irrevocable.
Customary exchange, with its more centralized administration of punishment, entails dramatically lower marginal cognitive costs than natural credit, and for this reason it scales up far better as a method of social organization. Rather than exchanging with a particular person in the context of an ongoing relationship, a trader gains the ability to trade with a member of a permanent class of people who specialize in some production, without worrying about the identity of his particular trading partner. In this way, because the delineation of its particular elements lies outside any single mind, the division of labor can reach far greater complexity than what was possible under natural credit organization. Indeed, the first proper markets begin to appear with the development of customary exchange economies: with a larger pool of people to whom one can sell, permanent specialization becomes a more profitable investment – hence the well-known correlation between the emergence of agriculture, money, and cities (Clower 1995).
The emergence of indirect exchange and political centralization, by themselves, should not be oversold: monetary exchange is certainly not a sufficient condition for economic development. In the “hockey stick” graph of economic growth in world history, the kink does not correspond to the monetization of a premonetary society – indeed, it does not come until after indirect exchange had already been in use (with an interruption at the opening of the Middle Ages) for millennia.
In order for ad hoc monetary exchange to displace customary exchange, the marginal cost of monetary exchange must be extremely low. This entails a great deal of investment in high-fixed-cost exchange institutions, both in terms of the money itself and the money users. For the money itself, there are a number of developments in minting technology that require the division of labor to be sufficiently advanced to develop. The screw press, for example, dramatically lowered the cost of coinage and increased the quality and standardization of coins. The serrated edge gave assurance that the coin had not been clipped or shaved, allowing exchanges to be conducted without high-marginal-cost haggling or verification. A coin whose quality must be haggled over is not a solid foundation for the explosion of ad hoc exchange.17
There were, in addition, important institutional developments, particularly in England, that contributed to low-marginal-cost exchange. The displacement of bimetallic standards by the gold standard required institutions sufficiently credible to issue gold-backed token coins, as opposed to the full-bodied silver coins that had been used for small change to that point (Redish 1990), but which had been beset by frequent shortages due to Gresham’s law. The development of institutions for credibly restraining the administrative class from predation enabled the development of liquid banking and finance institutions (North & Weingast 1989), which contributed to the stability of the value of the monetary unit (Harwick 2016) and (therefore) to the widespread use of reliable paper money.
More broadly, transportation and infrastructure were an important fixed cost in increasing the size of the market and thus the equilibrium level of investment in fixed-cost institutions. The emergence of market prices doesn’t mean much in an isolated village; indeed, controlled prices may be observationally equivalent to market prices. It is only when markets begin to become integrated across long distances – requiring prior investment in navigable roads and rivers between them – that market prices begin to play a crucial role in driving anonymous trade and the expansion of merchant classes (Nye 1991).
Finally, we should not take for granted the costs of learning, not to mention teaching an entire society, to count indefinitely high and to compare two arbitrarily large numbers (Deheane 1997): the rise of mass education following the Protestant Reformation (Dittmar & Meisenzahl 2016), by raising literacy and basic numeracy, enabled large swaths of previously excluded people to participate in the monetary economy. This had the effect of dramatically expanding the size of the market, and thus the division of labor.
These cumulative investments in higher fixed-cost institutions, each of which lowered the cost of exchange on various margins, took place over the course of hundreds of years in Western Europe. Once marginal costs are sufficiently low, and once the market reaches a sufficient size, to money’s initial embodiment of accounts is added the reification of global data on conditions of supply and demand as market prices in a common monetary unit, around which separate and anonymous (to each other) human actors can coordinate their plans impersonally at low marginal cost. The emergence of market prices reflects perhaps mankind’s most significant cognitive advance since the dawn of language (cf. Horwitz 1992), a distinction borne out by the phenomenal explosion in economic growth following the widespread integration of disparate markets first within England, and later much of Europe.
Market prices are crucial in the emergence of what North, Wallis, and Weingast’s (2009) (NWW hereafter) call the “open access order”.18
Limited access [i.e. customary exchange] prevents market prices from allocating resources between competing uses. . . . Rather than capturing rents by charging a high price, the possessor of a privilege may exploit it by charging a low price and allocating the resource to political allies. . . . When elites charge less than market clearing prices to secure political ends, the result is that prices cannot be used for impersonal coordination of the behavior of individuals. [Customary exchange] thus cripple[s] the price mechanism as a means to convey information about marginal benefits, marginal costs, and scarcity. . . . [I]t is not surprising that modern economic analysis of the price mechanism did not develop until open access orders with competitive markets began to develop. Competitive use of resources have existed since the dawn of human existence, but, with a few notable exceptions (such as ancient Greece), competitive markets with prices that convey information capable of coordinating human action are a recent development.
Having long ago substituted permanent specialization for ad hoc specialization with the adoption of customary exchange, the progressive substitution of ad hoc exchange relationships for customary relationships is equivalent with increasing social mobility. Thus, in the sense of expanding choice in both occupation and consumption, we may agree with Mitchell (1944) that money – once it becomes a sufficiently low-cost exchange institution – is perhaps the most liberating institutional innovation in human history:
When money is introduced into the dealings of men, it enlarges their freedom. . . . As a society learns to use money confidently, it gradually abandons restrictions upon the places people shall live, the occupations they shall follow, the circles they shall serve, the prices they shall charge, and the goods they can buy.
Interpreting the past, of course, is a simpler task than predicting the future. Nevertheless, in order to 1) demonstrate the usefulness of the framework “out of sample”, and 2) avoid the impression that monetary exchange in its current form represents “the end of history”, it will be worth imagining the shape of future investments in exchange institutions and advances in the division of labor.
Recall the two necessary constituents of an exchange institution: 1) accounting, and 2) punishment. It must first of all be remembered that relatively frictionless monetary exchange is still a rarity in much of the world. The lack of investment in institutions that can both credibly punish defectors and refrain from defecting themselves (Acemoglu 2003) inhibits the development of low-marginal-cost exchange institutions and technologies. Much of the future advancement must be in “catchup”, which will be – one must hope – an easier task than the first investments with the path already blazed by the Western world.
That said, there is still much potential in the developed world as well. One important recent innovation with implications for both punishment and accounting is decentralized blockchain technology, especially with “smart contracts” and decentralized autonomous organizations (DAOs). These are all tools that direct and manage the ownership of records19 according to pre-specified rules, written in code and distributed across many different computers across the internet.
Obviously the widespread use of blockchain technology for payments and record-keeping requires a huge investment in the fixed costs of communications infrastructure and a staggeringly intricate division of labor to support the mass production of the computers necessary to use them. But having made this investment, the technology has the potential to address a major drawback that has beset the use of external tokens for accounting since their very emergence: the fact that external accounts can be manipulated.
Theft under natural credit is necessarily theft of specific items. With accounting kept track of by personal knowledge, there is no way to “forge” the accounts and receive more than one gives. The introduction of indirect exchange vastly increased the returns to theft: by stealing a cache of the medium of exchange, the thief could induce his none-the-wiser peers to render services of more value than he had rendered himself.
The political control of money beginning with customary exchange organization amplified the problem. While petty theft came to be more or less effectively punished by the new administrative class, the administrative class’ control of the accounting tokens gave it a large (though not infinite) latitude to accumulate resources to itself without contributing through exchange, or directly coercing. Devaluations, inflations, and extortion of financiers have been the norm throughout history. It is only with the relatively recent development of credibly restrained political institutions that fiat currencies – which enable the administrative class to manipulate the accounting system much more easily and extensively compared to full-bodied currencies – have become viable.
Each step along this path, toward lower-marginal-cost exchange institutions with accounting taking place more and more outside individual minds, has been accompanied by an increase in the losses due to account manipulation – whether through theft or political influence – and a concomitant increase in the resources spent to mitigate these losses. The very emergence of specialized administrators of punishment that marked the beginning of customary exchange was likely a response to the increased importance of theft following the increased use of indirect exchange. More than likely the majority of these tradeoffs have been “worth it” in the sense of resulting in a more advanced division of labor. But, assuming a diminishing marginal rate of substitution between convenience and trustworthiness, the most beneficial advances from our already high-convenience vantage point will likely be in the trustworthiness of the accounting system.
The emergence of blockchain technology, for the first time, holds forth the prospect of an increase in the reliability of the accounting system without a diminution in convenience, at least once handheld computers and smartphones are sufficiently ubiquitous. Because the protocols are administered by no individual or organization, no individual or organization has the capability to reverse transactions, forge accounts, or inflate the money supply unexpectedly. Such an advance in the trustworthiness of the accounting system, along with its convenience, has the potential to sustain advances in the division of labor far into the future, and far beyond the advances that would be exhausted under an accounting system subject to political manipulation.
On the other hand, lest this account be thought of as a Whig theory of exchange, the possibility of disinvestment in fixed-cost institutions must not be emphasized as well. Numerous examples in history show such disinvestment following a collapse in the size of the market (e.g. Humphrey 1985). Likewise, the increasing unreliability of the monetary unit at the collapse of the Roman Empire caused its money economy to fall into disuse and triggered a dramatic ruralization and spontaneous dioecism throughout Europe. Bloch’s description of the rise of feudalism (1966, p. 250f) can be understood in these terms:
Estate management requires careful account keeping, which became more and more difficult for average administrators, in the ignorance and disorder which the great distress of the opening Middle Ages brought with it. The repeated, and almost puerile, instructions which abound in the estate ordinances of the ninth century . . . show us how hard it was for the great men to make their subordinates apply the most elementary rules of book-keeping. To adopt tenancy as a solution was the line of least resistance. . . . [T]he new tone of social life and the new habits of mind were all against any effort to maintain the old, and far too complicated [i.e. high-fixed-cost], methods.
In other words, at some threshold market size between that of Roman cities and rural estates, and without a reliable monetary unit, specialization became a less attractive prospect. Enough people returned to subsistence agriculture that the whole process began to unravel, and Europe fell back for a time to the customary exchange equilibrium. Without specialization, there was little need for a medium of indirect exchange, without which there was little need for numeracy and accounting skills. As these began to atrophy, the monetary economy became more and more defunct until feudal organization was settled into as “the line of least resistance” – simple enough in its basic form to be conducted largely without the aid of arithmetic or money.20
One must hope that nothing so dramatic awaits the developed world today. Nevertheless, the dramatic slowdown in global trade over the past decade and the ballooning of unfunded obligations throughout the developed world over the past half century have the potential to trigger the “liquidation” of prior investment in exchange institutions, and especially in those more recent developments in institutions that credibly restrain political actors from manipulating their currencies too extensively in their own favor through debt monetization. Such a breakdown may necessitate, if not a return to feudal organization, at least a return to full-bodied currencies as a prophylactic against overly extensive manipulation.
Anthropologists studying natural credit societies are often impressed by the extent of the division of labor. Indeed, from a zoological perspective, the division of labor that can be sustained at essentially zero fixed institutional cost is one of humanity’s great distinctives. Its impressiveness stands out, however, because the whole complex can (and indeed must) be grasped in its entirety. The division of labor characterizing modern society, though orders of magnitude more intricate than that characterizing hunter-gatherer society, fails to impress us precisely because such intricacy does not have to be (and indeed cannot be) grasped in its entirety by a single mind. The accounting is dispersed throughout the environment through convergence on the meaning of money. When Hayek (1945) argues that “the ‘data’ from which the economic calculus starts are never for the whole society ‘given’ to a single mind which could work out the implications”, we might add that this is true only of those societies which have advanced beyond customary barter. Of course, for any society which today enjoys an appreciable level of development, customary barter lies quite a ways back.
Situated at the dawn of the era of widespread impersonal exchange, Adam Smith (1776, bk. 1, ch. 2) observed mankind’s “propensity to truck and barter”, and tentatively suggested it as “one of those original principles in human nature”. Sociobiology supports this observation. Institutions for regular and predictable exchange are a hallmark of human behavior, endowed by a long process of biological evolution. This is no less true of the language faculty, which enables humans to offload cognition into the environment through coordination upon meaningful symbols.
It is no surprise, then, that the evolution of institutions for exchange, moving progressively more of the burdens of accounting and coordination out of the head and into the environment, is the primary driver of the advancement of human behavior and cognition in a span much faster than could be provided in evolutionary time. This paper has offered a unifying, though informal, account of this development through time in the context of an increasing returns model, and shown its fruitfulness in the analysis of important developments in the history of exchange. In this sense it represents a bridge between monetary theory and the state capacity literature, approaching the former in the context of developments in the latter. To return again to Smith, it is my hope – echoing Kaldor (1977) – that the notions of increasing returns and the division of labor being limited by the extent of the market should again become central in the explanation both of economic development and monetary exchange.