II. What Is Law? A Coordination Account of the Characteristics of Legal Order

By eschewing the assumption that a legal system is defined as a body of rules that are written and enforced by a centralized public authority with effective coercive power, we face the problem of choosing an alternative basis on which to say that law exists in a given environment. If, for example, medieval Iceland lacked any government enforcement authority, then what does it mean to assert that Icelandic society in the 11th century had a legal system?

Although economists and positive political theorists have generally not explored this question systematically (Benson [Reference Benson1989a] and Kornhauser [Reference Kornhauser2004] are exceptions), legal anthropologists and sociologists have. Anthropologists have faced the question of determining whether preindustrial communities such as foragers or hunter-gatherers possess a legal system and, if so, how to identify its content. Hoebel (Reference Hoebel2006, 28), for example, uses the following definition to analyze the legal practices and rules among those he called “primitive man”: “A social norm is legal if its neglect or infraction is regularly met, in threat or in fact, by the application of physical force by an individual or group possessing the socially recognized privilege of so acting.”

Sociologists have considered the question of the various forms of social order and the distinctive features of legal order. Weber defined legal order as order achieved at least in part by the probability that a “coercive apparatus” would be mobilized to impose a penalty, physical or psychological, in reaction to a norm violation simply because the norm is violated—and not to achieve any other material benefit. The “coercive apparatus” is understood by Weber to be “one or more persons whose special task it is to hold themselves ready to apply specially provided means of coercion (legal coercion) for the purposes of norm enforcement” (Reference Weber1978, 313). Recognizing that “a consociation specifically dedicated to the purpose of ‘legal coercion’ … has not always been the monopoly of the political community” (317), Weber explicitly takes account of the possibility that law may exist in environments that lack a state and, indeed, may possess instances of “legal order” that coexist with the state but are not enforced by the state.

We approach the problem of identifying law without presuming state enforcement by first placing law within this broader framework of distinguishing among different types of social order.Footnote ³ All human societies display various forms and degrees of social order, meaning behavior that is patterned in identifiable ways. People shake hands when they meet; they pay their bills; they stand an appropriate distance apart when speaking to a stranger. Order arises from many sources including biology, technology, morality, and social sanction. If we are to call the resulting order “legal,” we suggest, it should at least possess the following necessary characteristics. Note that we are not making the reverse claim: that any social order that possesses the following characteristics is legal; nor are we claiming that these characteristics are sufficient to identify an order as legal.

1. 1. Behavior is patterned on a normative classification; that is, the orderliness of behavior is with reference to a designation of some behaviors as preferred to other behaviors.

2. 2. The content of the normative classification is capable of being deliberately chosen, articulated, and identified by an identifiable actor or entity.

Hadfield and Weingast (2012) model an environment that potentially meets these initial criteria so that we can say that we may be modeling a legal order. That is, we consider an environment that includes institutions that are capable of supplying a deliberately chosen normative classification of conduct and in which the incentive for actors to choose behavior that the classification designates as preferred (we will say “not wrongful”) is to avoid incurring a penalty. We do not, however, assume that an institution that supplies the normative classification is an institution also capable of imposing penalties for wrongful behavior. We model instead an environment in which the only way in which behavior classified as wrongful by an institution can be punished is through decentralized enforcement efforts. Specifically, we look at a form of collective punishment whereby delivery of an effective penalty depends on independent and simultaneous decisions made by individual (nonofficial) actors to punish a wrongdoer.Footnote ⁴ We then ask the question, Is it possible to secure an equilibrium in which wrongdoing, according to the classification of some institution, is effectively deterred by this form of collective, decentralized, punishment without resort to a centralized, state-based source of coercion? If so, we argue, we will have shown the potential to achieve a social order that meets the above criteria for a legal order and so may be properly characterized as a legal order.

A key assumption in our model is that individual judgments about what constitutes wrongful behavior are the product of what we call an idiosyncratic logic employed by each potential punisher. By idiosyncratic we mean that the potential punisher’s reasoning is (in the extreme) inaccessible to others. This assumption captures an important and economically valuable form of heterogeneity, arising from the division of labor and specialization: the potential punishers (and beneficiaries of a stable system of rules) may be in different countries or industries; they may employ different production methods; they may organize their relationships and contracts with others in different ways. Social welfare in this economy is higher if rule violations can be deterred. But in the absence of a centralized third-party institution capable of punishing rule violations, deterrence must be achieved through the delivery of decentralized punishments. Our model assumes in particular that the only available form of punishment to secure deterrence is decentralized collective punishment.

Achieving deterrence therefore requires coordinating collective punishment in response to particular actions. This presents two essential problems. First, because each potential punisher has an idiosyncratic logic for assessing wrongfulness, none are able to determine unilaterally when to punish in response to possible rule violations. The punishers need a coordination device that tells them when to punish. Second, because punishment is individually costly, punishers need an incentive to punish. Note that unlike the law and social norms literature, as well as some recent legal philosophy literature, this is not a pure coordination game: coordination alone provides an insufficient incentive to participate.

Hadfield and Weingast (Reference Hadfield and Weingast2012) show that there exists an equilibrium in this repeated game in which a third-party institution supplying a publicly accessible normative classification system—what we call a common logic—can resolve the coordination and incentive problems and secure deterrence of actions that are deemed wrongful by the common logic. Thus the equilibrium satisfies the minimal criteria we suggest are necessary (but not sufficient) to reasonably identify social order as legal order: behavior is patterned on the basis of a normative classification, the content of which can be deliberately supplied and changed by an identifiable entity.

Moreover, as we investigate the characteristics of that equilibrium, we see that it possesses characteristics that are generally thought by legal scholars and philosophers to be distinctively legal. In particular, we argue that the equilibrium we have identified is characterized by attributes that are frequently identified as marking the existence of law by legal theorists such as Fuller (Reference Fuller1964), Raz (Reference Raz1977), and Waldron (Reference Waldron2008).Footnote ⁵ These attributes are as follows:

• generality, stability, prospectivity, and congruence

• qualified universality

• clarity, noncontradiction, uniqueness: authoritative stewardship

• impersonal, neutral, and independent reasoning

Each of these attributes plays a role in resolving either the coordination problem or the incentive compatibility problem facing the reliance on decentralized collective punishment.Footnote ⁶

Consider the coordination problem first. The common logic serves as a coordinating device by providing a publicly accessible, clear, and unique classification of conduct based on impersonal reasoning and either general rules or generalizable categories. Potential punishers can determine the classification of conduct by the common logic, even in light of new or novel circumstances (which we expect to arise in the context of diversity and economic growth). The common logic reaches a unique classification for any performance, and thus punishers receive the same signal about whether to punish or not. Moreover, the classification is common knowledge. Hence, contingent on each potential punisher having concluded that the other punishers who are needed to make punishment effective also have an incentive to punish, all punishers can predict that if they punish, the others will too. Furthermore, it is common knowledge that the potential wrongdoer can also access the common logic, implement the same impersonal reasoning, and make this prediction. The punishers can therefore predict that the potential wrongdoer will avoid conduct that triggers collective punishment.

A key feature of the common logic helps to ensure that the logic can perform this coordination role. We refer to this as authoritative stewardship. By this we mean that there is a single identifiable institution whose responsibility is to resolve any ambiguity or uncertainty about the classification reached by the common logic. The steward also helps adapt the common logic to changing circumstances. This institution serves to ensure both uniqueness and common knowledge of uniqueness. Suppose that the conduct in question is particularly novel or complex, requiring the elaboration of a general principle in the context of previously unseen facts. For the common logic to continue to perform its role in coordinating punishment, it must be that all agents are known to treat the steward as the authoritative and final means of resolving ambiguity in the logic.

Other features of the common logic help to resolve the incentive compatibility problem of participation in collective punishment. In our model, the incentive to participate is that it secures the benefits of the coordinated equilibrium in which rule violations are deterred. This benefit is enjoyed by an individual agent only if the common logic classifies as wrongful conduct that the agent judges, according to his or her idiosyncratic logic, to be undesirable. The concordance between the common and the idiosyncratic logic need not be perfect, but the two logics must be sufficiently convergent that the agent is better off in the coordinated equilibrium with some deterrence than the uncoordinated equilibrium with none. As we show in Hadfield and Weingast (Reference Hadfield and Weingast2012), by participating in punishment activity, an individual agent signals to the other agents that the common logic that coordinates deterrence is or continues to be sufficiently convergent with its own idiosyncratic logic to make continued coordination valuable.

The resolution of the incentive problem thus requires that each agent be able to assess the value of future coordination under the common logic, that is, to determine how often the common logic will classify as wrongful conduct that the agent judges to be wrongful. As with the coordination problem, this requires a clear, accessible reasoning scheme that is impersonal in the sense that its classifications do not depend on the identity of the individual who is implementing the reasoning. The common logic must be universal in the qualified sense that it includes rules that benefit all those who are necessary to make collective punishment effective. To allow necessary punishers to assess how the common logic might be of benefit to them, the common logic will have to consist of general principles that each agent can elaborate on the basis of his or her private information about the nature of the facts and circumstances that would be relevant if the agent suffered a wrong. It is also likely in a complex world that the common logic will possess open processes for elaborating general principles, allowing agents who object to conduct as wrongful to introduce privately known facts and idiosyncratic reasoning about why those facts should result in a finding of wrongfulness under the common logic. With such openness, a system can achieve higher levels of convergence between idiosyncratic and common logic and hence is more likely to be adopted and remain stable over time. Finally, the common logic must be stable for a period of time sufficient to allow punishers to recoup their up-front investments in costly punishment to signal participation.

The virtue of our model, then, is that it provides a positive basis for predicting that an equilibrium in which the deterrence of wrongful behavior is supported exclusively by decentralized collective punishment is likely to possess attractive normative characteristics frequently associated with the rule of law.Footnote ⁷ Because these legal attributes play an instrumental role in establishing an equilibrium legal order in the absence of centralized coercive force, we can therefore look to see the extent to which a particular historical setting lacking centralized coercion displays these attributes. We expect that the efficacy and robustness of an effort to secure legal order on the basis of decentralized enforcement alone will be correlated with the presence of the legal attributes we identify. We thus have more theoretical tools to bring to bear on what Greif (2006, 350–76) calls the project of “interactive context-specific analysis.” With this in mind, we turn to examine our three specific case studies.

III. Medieval Iceland

The age of blood feuds in Iceland was also an age of extensive and complex laws and litigation (Bryce Reference Bryce1901; Miller Reference Miller1990). Shortly following settlement by the Vikings, the first general assembly, known as the Althing, was held in 930 at Law Rock. Here, freemen assembled every year to vote on rules and, more importantly, hear lawsuits “argued with an elaborate formality and a minute adherence to technical rules far more strict than is now practised anywhere in Europe” (Bryce Reference Bryce1901, 274). Lawsuits were argued on the basis of both customary rules and those generated at Law Rock. They were heard by ad hoc panels of judges selected by the chieftains of the individual Things into which the Republic of Iceland was divided. Individual Things might hear disputes throughout the year; the annual Althing decided disputes that transcended the capacity or authority of the individual Things.

Despite elaborate procedures for litigation, the Icelandic system had no centralized authority to enforce its rules. Again according to Bryce (Reference Bryce1901, 281), “There was no police, no militia, no fleet, no army. … Such State organizations as existed came into being for the sake of deciding lawsuits. There it ended. When the decision had been given, the action of the Republic stopped. To carry it out was left to a successful plaintiff.” In fact, the only government official in the entire Icelandic republic was an individual known as the Law Speaker. He was elected at the Althing to a 3-year term and paid out of public coffers. His job was, however, not to adjudicate disputes. His sole duties were to memorize and recite at the Althing the entirety of the law, composed of both customs and legislated rules; to answer queries from potential litigants about the law; and to serve as the only person recognized as able to declare with finality the content of the rules in the event of uncertainty in a court proceeding.

The Icelandic system satisfies our axioms about what constitutes a legal system: behavior was organized on the basis of a normative classification, as articulated by the Law Speaker and chosen by the Althing. The Icelandic rules governed almost every detail of mundane life (Miller Reference Miller1990), not just killing and plunder: moving bones to relocate a cemetery, how to separate and account for hay blown into a neighbor’s field, baptismal practices, marriage contracts, loan agreements, sales of goods, and more.

Friedman (Reference Friedman1979) uses Iceland as an example of a legal system independent of governmental authority for enforcement that instead relied on a purely private system to enforce rights (especially rights of property and against personal injury). He analyzes the Icelandic system as a relatively “pure” case in which victims obtain the exclusive private right to prosecute, and receive monetary compensation from, a wrongdoer. This “property” right in prosecution, he suggests, could have produced a relatively efficient system of deterrence: the holder of the right could have contracted with private “enforcers” to extract payments and such enforcers, advertising their customer list to potential wrongdoers, could have internalized the externalities enjoyed by the community from the prosecution of individual injuries. Friedman interprets the voluntary association of freemen into Things and other family and friend-based coalitions as an example of the formation of groups that took on the obligation to assist in the prosecution of injuries and collection of compensation on behalf of their members.

This analysis leaves a series of questions unresolved. If harms were felt at the level of the coalition, why did the coalitions look to public rules and litigation controlled by those outside the coalition to determine when and how they sought compensation? The analysis also provides no explanation of why coalitions limited their compensation to the legislated amounts rather than set their own rules for when and how they retaliated against wrongs they perceived. The analysis accounts for only intercoalitional harms: What governs relations within the coalition, and how did the coalitions solve the free-rider problem? Finally, analyzing the Icelandic system as one based on intercoalitional injury and retaliation reduces the system to an instance of the clan-based warfare that characterized many early (and some modern) human societies. Were the legal aspects of the system—the Althing, the courts, the convoluted rules of procedure, the specified levels of compensation, the detailed regulation of everyday life, the pronouncements of the Law Speaker—merely epiphenomenal, playing no functional role?

Our model analyzes explicitly what Friedman treats in summary fashion: what it takes, institutionally, for the coordination of collective punishment of a wrongdoer by ordinary individuals to be effective and to be sustained as an equilibrium. By focusing explicitly on the problem of coordinating collective punishment, our approach accounts for the distinctively legal features of the medieval Icelandic system and suggests that they were essential to its achievement of some degree of legal order.

Consider in more detail the collective punishment system. If a person was declared in violation of the republic’s rules by a court, the court announced the prescribed penalty. If monetary compensation was not paid, then the violator was declared either a lesser or greater outlaw (Bryce Reference Bryce1901, 162). A lesser outlaw had all his property confiscated and was banished from the republic for 3 years; a greater outlaw lost all his property, was banished permanently, and could be killed by anyone with impunity. Once declared an outlaw, anyone might punish the offender unilaterally—taking the offender’s property or killing him, for example.

Outlawry is a collective punishment. To be effective, it requires all individuals to participate by refusing to give shelter and aid to an outlaw. Why did people participate in this punishment? It must surely have been costly to some, such as those who depended on the outlaw for support or income. Even where the Icelandic punishment regime appears to rest on unilateral action—such as by authorizing anyone to take an outlaw’s property or to kill him—efficacy still rests on voluntary participation by multiple individuals and is an example of a regime based on decentralized collective punishment.

The coordination problem associated with decentralized collective punishment becomes visible in this system when we consider how the rest of the community—whether a member of the same Thing or another—responds when someone retaliates against the convicted offender. Suppose that A perceives an injury done by B, files a lawsuit, and obtains a court judgment that B owes A compensation but that B does not pay. Person A brings another lawsuit and has B declared an outlaw. Person C then helps himself to B’s sheep; C is engaging in unilateral action to enforce the outlawry judgment. How does the rest of the community respond to C’s action?

For retaliation to succeed against those declared in the wrong, a retaliator needs to be assured that the community at large interprets his or her action as an appropriate response to the declaration rather than as an independent harm. Abstractly, this requires that the community coordinate on what constitutes authorized retaliation; that in turn requires coordination on what constitutes a rule violation and when the obligation of compensation is considered unsatisfied and thus cause for a decree of outlawry in the first place. Those who refrain from punishing retaliators acting in accord with the public rules and decisions participate, if passively, in this form of collective punishment. Without coordination on the classification of authorized retaliation, individuals cannot be confident about the deterrence effect of decentralized punishment.

The institutional features that characterized the Icelandic system helped to resolve the problems of coordination and incentive compatibility in the service of supporting decentralized collective punishment. Consider the function of the Law Speaker. Our framework explains the use of a single designated law speaker by the need for an authoritative steward to coordinate and incentivize participation in decentralized collective punishment. Our model suggests that individuals have an interest in determining how the common logic classifies particular actions. This public classification allows them to avoid committing a wrong and to determine whether their own participation in punishment is called for. The Law Speaker’s duties specifically included an obligation to recite the rules publicly and to respond to any queries posed to him about the rules. The Law Speaker thus serves as a quintessential example of an institution that exercises authoritative stewardship over the common logic.

Other features of the Icelandic regime can be understood in terms of their role in resolving the incentive compatibility problem facing the decentralized collective punishment on which the regime relied. Although adjudication was performed by ad hoc groups of judges appointed to resolve particular disputes, decisions were not ad hoc. Legal rules and reasoning were expressed in general terms. Miller (1990, 62) recounts, for example, a law declaring, “It is prescribed that there shall be no such things as accidents [but that] if a man does worse than he intends to do and damage [to livestock] results from his clumsiness, that is not punishable at law and he shall make amends for the damage within two weeks time as it is evaluated by five neighbors. Otherwise it shall not be judged an accident.” Stated in this manner, a general rule is also universal. A universal rule tells a person who invests in punishing the person who fails to make amends for damage done to a neighbor’s livestock that he can do so in the confidence that the rules about accidents will apply to him in the same way if he suffers uncompensated damage in the future.

The Icelandic laws appear to have drawn fewer distinctions between free men and women than did other Anglo-Saxon regimes. All were (at least officially) entitled to have their death compensated at the same rate, for example (Miller Reference Miller1990, 29). Our model suggests a possible account for this: in a regime that depended significantly on women’s participation in punishments—refusing food or shelter to an outlaw, for example—we predict a common logic that attends sufficiently (relative to their alternative payoff if they decide not to participate) to their interests.

The incentive to help the wronged neighbor to punish was also supported by the stability and prospectivity of the Icelandic rules. In this system, leaders could not emerge at random times and announce a new set of rules—whether for sincere policy reasons or to achieve corrupt objectives. Instead, the rules either were derived from custom, which is inherently slow to change, or were legislated at an annual public event—the midsummer Althing at Law Rock. This system gave a basis for A’s reasonable expectation that if A helped this year to punish the trespasser who damaged B’s livestock, the same rules would be in place 3 years hence to generate an incentive for B to help out A if A suffered similar harms. The stability and prospectivity of Iceland’s general rules also gave A the basis to feel reasonably confident that his participation in punishment on B’s behalf today would not later be deemed a wrongful act itself by a change in the rule that authorized retaliation at the time he decided to participate.

The Icelandic rules also displayed the use of impersonal reasoning. Ad hoc courts were composed of groups of individuals; this implies that the rules were expected to be capable of interpretation and application by individuals in consistent ways. The designation of a single individual with special authority to resolve uncertainty about the content of rules—the Law Speaker—also demonstrates the reliance on impersonal reasoning. For the Law Speaker was elected only to a 3-year term, indicating that the content of rules was capable of being articulated by anyone who might be chosen for that office and not dependent on the reasoning of a particular person.

The violence of Viking society puts the importance of these institutional attributes into sharp focus. Violence is far more likely to spiral out of control in the absence of centralized coercion given ambiguity or instability about what counts as justified violence. Ambiguity about judgments is likely to plague a system based solely on emergent social norms. An unwillingness to channel violence to track a system of rules is likely to be characterized by ad hoc judgments according to poorly articulated, unstable, or highly personalized criteria. Icelandic society was not close-knit; freemen lived fiercely independent lives only loosely affiliated with a Thing headed by a chieftain who exercised little coercive power over his Thingmen (Miller Reference Miller1990, 21–28). This environment is unlikely to have spontaneously developed a unique, common knowledge classification of behavior across the myriad circumstances that might breed conflict.

The emergence of a formal institution (the Law Speaker) for resolving ambiguity, formal institutions (ad hoc courts) for applying rules based on impersonal reasoning, and a formal institution (the Althing) for publicly designating general rules in a stable and slow-changing way thus indicates a critical shift in the mechanism by which Iceland achieved some level of social order. We argue that the adoption of these institutions is usefully identified as a critical step in creating legal order, despite the absence of a centralized coercive authority.

IV. The California Gold Rush

“When gold was discovered on January 24, 1848, the territory [of California] had none of the usual legal institutions such as a legislature, courts, police or jails” (McDowell Reference McDowell2004, 772).Footnote ⁸ Yet disputes over claims were generally rare, and surprisingly little violence arose, at least over the right to engage in the hard work of extracting gold from a digging (Zerbe and Anderson Reference Zerbe2001). Instead, “a common or customary law of the diggings emerged that allowed a miner to hold a small claim for as long as he was working it or left his tools in his hole. When diggings looked promising, however, and likely to attract many miners, those who were on the spot held a meeting to pass a more detailed mining code for that particular area[,] … [choosing] a chairman, appoint[ing] a committee to draft a code, and a short time later, approv[ing] it by majority vote” (McDowell Reference McDowell2004, 778).

The substantive content of the codes “varied in detail from camp to camp, and they could be modified at a subsequent miners’ meeting, in which case the rights of claim holders might change from one day to the next” (McDowell Reference McDowell2004, 773). Notably, most codes settled on limiting each miner to possession of a single claim—although the size of the claim did not settle to a uniform dimension—and required that a claim be actively worked; a miner who staked a claim and then disappeared no longer held a claim. In most cases, neither the procedural nor the substantive rules distinguished between those who held claims and those who did not at the time of the meeting, nor between those who were already at the camp and those who arrived later. The rules established at the camps made those working claims vulnerable to legitimate claim jumping by newcomers, for example, if the miner who held the claim failed to meet the requirement of sustained working of the claim.

When disputes arose between American miners, they were often referred to third parties including ad hoc arbitrators and juries of miners. The miners at Jackass Gulch, for example, specified in their code that “as soon as there is sufficiency of water for working a claim, five days absence from said claim, except in case of sickness, accident or reasonable excuse, shall forfeit the property.” They declared that any disputes—about what counted as a “sufficiency” of water or as a “reasonable” excuse, for example—would be decided by a jury of five persons (Umbeck Reference Umbeck1977, 436).

“Almost all litigants complied with [jury or arbitrator] decisions without further ado” (McDowell Reference McDowell2004, 788). In the apparently rare cases in which they did not comply immediately, the community would announce clearly to violators that they were to abandon the diggings or risk community punishment (800). These collective efforts were decentralized in the sense that every participant had to decide whether or not to incur the cost of rising up to protect someone else’s claim.

Umbeck (Reference Umbeck1977), drawing on the early property rights literature, highlights the private incentive to create and protect property rights when the value of private property exceeds the cost of the institution of property rights. He suggests that property rights in California were stable despite the absence of an external enforcer because it was in the interests of all miners to agree to a set of rules and participate in enforcing them. While an important contribution, Umbeck’s work is undertheorized: it fails to explain the particular features of the rules and procedures of the mining camps, and it provides no compelling account of the incentive for individuals to either abide by the rules or help enforce them.

Later contributions attempt to remedy these defects using the framework of coordination games. Zerbe and Anderson (Reference Zerbe2001) use Kreps’s (Reference Kreps and Shepsle1990) theory of corporate culture to argue that the particular rules and procedures in the camps can be explained as focal points based on shared cultural norms, which selected an equilibrium on which miners coordinated. Participation in enforcement, they suggest, is also explained by shared norms of fairness and culturally induced preferences for seeing wrongdoers punished.Footnote ⁹ McDowell (Reference McDowell2002) challenges Zerbe and Anderson’s views about shared culture, emphasizing the ambiguity surrounding concepts of fairness, even among Americans, and several ways in which the mining rules departed from prevailing American norms, such as by making owners vulnerable to losing their property if they did not actively work it. She emphasizes “the great mystery” of why the claim holders who devised the first codes in a mining camp did not allocate to themselves full property rights, assuming (as Umbeck does) that the residents of a camp were sufficiently numerous to fight off newcomers. Why limit miners to a single claim and authorize jumping of an unworked claim? McDowell appeals to prevailing Jacksonian ideology, but she also argues that miners recognized that even if they held a claim today they were likely to find themselves in search of new claims tomorrow; behind a “Rawlsian veil of ignorance,” she suggests they settled on rules that treated existing claim holders and newcomers on equal terms.

McDowell (Reference McDowell2004) supplements this analysis of the choice of particular rules with an account of the enforcement mechanism, relying on Sugden’s (Reference Sugden1986) theory of the spontaneous emergence of property rules as a solution to the strategic interaction captured by a hawk/dove game.Footnote ¹⁰ The rules established in the mining camps, McDowell argues, were closer to law than norms because they were deliberately chosen, but they were nonetheless self-enforcing in light of the incentive to coordinate to avoid wasteful contests. Clay and Wright (Reference Clay2005, 163) endorse McDowell’s view that miners chose rules that balanced the interests of claim holders and claim jumpers because of the inevitable “‘search’ and ‘race’ aspects of gold mining.” They argue that ambiguity surrounding the implementation of these balancing rules—which required an assessment of the validity of excuses for not actively working a claim—ultimately undermined the efficacy of decentralized third-party enforcement.

Our framework advances the understanding of the features of the system developed in gold rush California. Like McDowell, we characterize the achieved order as a legal order because it created a system for supplying deliberate content to rules. Moreover, as McDowell and Clay and Wright recognize, both ambiguity and arbitrariness are at work in this system of deliberate rule creation. And like all of those in the more recent literature, we agree that coordination games are a useful framework for thinking about how the rules the miners developed were (at least to some extent) enforced.

We disagree, however, that the solutions the miners developed are adequately explained as equilibria in pure coordination games played with respect to primary (claim holding and claim jumping) behavior. Modeling gold rush California as a pure coordination game fails to explain several problems. In a pure coordination game, no conflict arises over which equilibrium to play. Further, no problems of enforcement arise. Coordinating players’ expectations in such a game is sufficient to achieve an equilibrium with honorable behavior. But why would a miner follow the rule of respecting claim markers or limiting himself to a single plot of land 15 feet square just because others did so? Why would other miners participate in costly punishment of those who violated the rules? Modeling the problems faced by miners as a pure coordination game, as with most of the existing literature, sets aside this diversity of interests and views about what should count as allowable behavior and rights.Footnote ¹¹

We also agree with two of Umbeck’s (Reference Umbeck1977) claims: in the early days of gold discovery, spontaneous norms governing claiming (allowing a miner to claim a digging hole indefinitely by leaving his tools there) worked reasonably well given the abundance of potential diggings and the small population; and population pressure created the risk of greater conflict over diggings so that the costs of establishing a more complex system were warranted.

In our view, the rise in population created not only greater resource pressure but also increased diversity; people from all over the country and indeed the world, motivated to rationalize their access to a limited resource, were bound to bring or develop their own views about appropriate norms. Similarly, inventions of new mining techniques and the mining of different types of deposits also generated diversity in private valuation systems over alternative rules (what our model refers to as idiosyncratic logics). Finally, as both McDowell and Clay and Wright emphasize, substantial diversions of interest existed between those who had begun work on a prospective deposit and those still searching and between those who were already at a mining camp and those who arrived afterward.

We begin with the idea that the rules the miners adopted constituted an effort to create a common logic—a common knowledge system for classifying behavior as wrongful or honorable. Our approach isolates the central role played by the problems of both coordination and incentive compatibility in the enforcement game.

We interpret the mining rules as enforced by a threat of collective punishment. This account focuses on the game of collective punishment, which as our model makes clear is not a game of pure coordination. Individuals will participate in collective punishment to enforce a particular rule if they expect others to do so and if they evaluate themselves as better off in the equilibrium coordinated under that rule than the alternative. The alternative in gold rush California was whatever each miner could secure by his own wits and might in a world in which no one follows rules, an environment akin to the Hobbesian “state of nature” characterized by the “war of all against all” (Hobbes Reference Hobbes1651/1991).

The incentive constraint on the equilibrium provides an explanation of several features that are poorly explained by the existing accounts. Consider McDowell’s “great mystery” of the mining rules: miners who already were working the area agreed to rules that restricted the size and security of their preexisting claims, allowing newcomers to take over the area in specified circumstances. This feature is better explained by the constraints on effective collective punishment than it is by a specification of the particular shape of the preferences of miners.Footnote ¹² Specifically, suppose that effective collective punishment required the participation of most of the inhabitants of a mining camp at the time of a violation. Miners had various enforcement techniques, including social disapproval and ostracism as well as collective physical efforts to scare off or even injure someone attempting to jump a legitimately held claim.Footnote ¹³ Stability of the regime required that third parties participate in collective punishment, in putting out an illegitimate claim jumper or defending the interests of a legitimate one.

Our view also provides an explanation for why the mining rules treated everyone at the camp alike—old-timer and newcomer, lucky claim holder and unlucky searcher. By doing so, the rules increased the likelihood that they satisfied the incentive compatibility constraint on effective collective punishment: any miner in the camp had to perceive the common logic of the rules to be sufficiently convergent with his own idiosyncratic logic—to rationalize his personal expenditure of effort and risk to help secure stability under the rules. This is an example of qualified universality: the rules address the interests of all those necessary to an effective punishment scheme.Footnote ¹⁴

Universality as a characteristic that supports equilibrium enforcement may also help to explain the rules that limited miners to a single claim and required them to consistently work the claim in order to maintain their rights. Allowing existing miners to exert rights over all the territory in a mining district sets up clear “insider” and “outsider” groups, organizes the interests of each as a group, and limits the size and strength of the insider group. In contrast, holding open the prospect of accommodating as many miners as possible in a district and rewarding a newcomer who participated in camp life with the possibility of taking over an (apparently) abandoned claim has the effect of extending the pool of those with a potential stake in upholding the rules. This may have contributed to stability of the regime.

The rules limiting miners to a single worked claim may also have served to generate other legal attributes supporting legal order. Such rules promote the capacity for any individual miner to engage in the impersonal reasoning necessary to make assessments of when intrusion on a claim counted as punishable claim jumping—not only for the purpose of assessing his own claim making and claim jumping behavior but also for the purpose of identifying when he would be expected to participate in enforcement efforts. A one-man-one-claim rule implies that an individual can rely on having seen a miner working a particular claim as clearly implying that this miner has no right to challenge claimants elsewhere; any other rule gives rise to claims by absentee owners, which are difficult for a casual observer to evaluate. Similarly, the requirement of working a claim to hold it supports easy—and common—inferences about who has a valid claim and who does not. This rule relies on easily observable behavior (someone is there working most of the time or is not) and thus enables miners working nearby claims—who are the most likely to be in a position to raise an alarm—to decide when they should stand up for their neighbor and when for a newcomer.

The absence of other criteria in the rules supports the impersonality of the reasoning necessary to predict how the common logic will classify an event. Suppose, for example, that the rules discriminated in favor of old-timers against newcomers. This more complex approach to classification of rightful and wrongful claiming would make it more difficult for any ordinary miner to apply the reasoning of the rules to classify events for himself. This complexity means that some would fail to correctly classify particular miners; this possibility, in turn, presents a great excuse for not helping to throw off a claim jumper: “I had no idea if he was at the first meeting or grandfathered in.” In contrast, impersonal rules—the one-man-one-claim rule, the requirement of actively working a claim, and the specification of uniform dimensions and uniform means of marking a claim—promoted clarity in the rules and allowed any miner to apply them.

A final important feature of our approach missing from the pure coordination account concerns the interaction of ambiguity (as Clay and Wright [2005] emphasize) and hence the need for an authoritative mechanism for resolving this problem. Consider the Jackass Gulch rule noted earlier about 5 days’ absence from a claim (subject to qualifications about sufficient water and sickness or accidents) resulting in forfeiture of the claim. This rule treats many circumstances definitively (e.g., when there is plenty of water), but not all. Good-faith disputes inevitably arise, for example, over how much water is sufficient; was an accident sufficiently bad? Hence the provision for what we call authoritative stewardship: in the event of dispute, the matter is decided with finality by a jury or designated arbitrator. A robust use of this mechanism, far from demonstrating the fragility of rules (as Clay and Wright [2005] suggest, pointing to significant levels of litigation), may well serve the important function of reassuring everyone at the camp that residual uncertainties are resolved according to the common logic (including its procedures) and not by other means. These procedures serve at least in part the function of making the common logic a system of public and open reasoning.

Clay and Wright (2005) make the case that the miners’ rules produced order but did not produce efficiency because they promoted a wasteful race for gold; a better regime might well have established exclusive rights that made claim jumping everywhere illegitimate. And had California been organized as a legal territory or as a state antecedent to the discovery of gold, a more efficient system of property rights might have emerged. But as Clay and Wright ask, would such a system have been enforceable? Our model focuses expressly on this question, and as our analysis above suggests, it is possible that the constraints on securing reasonable legal order relying exclusively on decentralized collective punishment limited the efficiency of the underlying rules.

V. Medieval Merchants

With the expansion of trade in Europe and the Muslim Mediterranean beginning in the 11th century came the challenge of securing the terms of commercial transactions in an increasingly diversified and unfamiliar world, especially for exchanges involved in long-distance trade (Greif Reference Greif1989, Reference Greif1993, Reference Greif2006; Milgrom et al. Reference Milgrom, North and Weingast1990; Greif et al. Reference Greif and Weingast1994). Although individual towns or regions might benefit from the protections offered by a powerful local ruler or religious organization capable of enforcing rules, a great deal of trade took place between local residents and merchants (or their agents) who traveled from afar. Enforcement by or against foreign traders in disputes with local merchants was a major challenge: the foreign merchant who cheated a local might be long gone before a dispute could be resolved by local courts. Conversely, foreign traders who suffered at the hands of local merchants might find local judicial systems deaf to their concerns as rulers sought to benefit their own citizens at the expense of foreigners. In many cases, a centralized enforcement body was practically unavailable to support transactions between local and foreign merchants. Even local merchants who lived in a territory governed by a powerful ruler or a religious organization were likely to face gaps in centralized enforcement of their local transactions, as the availability of enforcement resources was frequently strained by the fairly constant challenges to authority that rulers experienced in this period before the consolidation of stable states.

The achievement of a degree of legal order in commercial transactions in the world of the medieval merchant therefore depended on a fairly robust decentralized enforcement mechanism to fill sometimes-large gaps in the efficacy or availability of centralized enforcement. Several authors have studied this problem and demonstrated that enforcement of commercial rules in the Middle Ages rested significantly on organizations such as merchant coalitions, merchant guilds, and communes (Greif Reference Greif1989, Reference Greif1993, Reference Greif2006; Milgrom et al. Reference Milgrom, North and Weingast1990; Putnam Reference Putnam1993; Greif et al. Reference Greif and Weingast1994).Footnote ¹⁵ These organizations provided sets of rules and enforced them, primarily through ostracism and boycott: excluding violators from profitable transactions or from group benefits such as protection.

The multiplicity of enforcement institutions stands in contrast to the modern environment in which legal order is largely secured through the elaborate legal systems implemented by nation-states.Footnote ¹⁶ This raises fundamental questions about how and why the modern system of territorially based and exclusive legal jurisdictions evolved and the extent to which centralized states are required for the provision of law. As we discuss in the next section, our framework for analyzing legal order sheds light on these questions.

Trakman (Reference Trakman1983) and Benson (1989b) argue that the legal order in the medieval world evolved spontaneously, guided by a decentralized process not unlike the invisible hand of markets. From their perspective, the emergence of centrally organized legal order through state governments was unnecessary; in fact, it degraded the efficiency and quality of medieval merchant law.

Our framework challenges this institution-free view of the evolution of commercial legal order. Although we agree that private commercial organizations are likely to arise to meet the demand for legal rules to coordinate commercial activity, Benson’s and Trakman’s accounts gloss over the difficult questions of how decentralized commercial enforcement is incentivized and coordinated. Both accounts also idealize custom, suggesting that it evolves to preserve efficient rules because merchants simply avoid rules that undermine efficiency. But this claim ignores the powerful coordination nature of equilibria: even if an individual merchant comes to recognize, accurately, that a particular rule is suboptimal for his trade, if other merchants effectively punish violations of existing rules, then inefficiency can survive as an equilibrium. Our express focus on the problems associated with coordinating and incentivizing equilibrium demonstrates the role for centralized institutions even if they do not engage in coercive enforcement. Benson and Trakman, however, conflate government with governmental coercion and thus overlook key questions about the comparative capacity for different types of institutions—public and private—to coordinate legal order.

Building on Trakman’s and Benson’s work, a robust literature provides theoretically grounded accounts of how decentralized merchant institutions can secure legal order without government coercive authority. Greif (Reference Greif1989, Reference Greif2006) shows how a closed group of traders in the 11th-century Muslim Mediterranean—the Maghribi traders—secured their agency contracts through a decentralized enforcement system in which all members of the group refused to deal with (boycotted) a merchant who had a reputation for cheating. Greif argues that these traders, with a shared Jewish heritage, were able to draw on collectivist cultural beliefs according to which individuals are expected to respond to injuries done to others to coordinate collective punishment of wrongdoing. Milgrom et al. (Reference Milgrom, North and Weingast1990) demonstrate that a similar multilateral reputation equilibrium can be coordinated in the absence of shared cultural beliefs by a group of merchants using an institution they call the Law Merchant. This set of institutions incentivized merchants to undertake costly behavior to facilitate punishment for cheating, including to obtain and share information about cheating by other merchants with a merchant court and to punish a merchant with a public record of cheating because they themselves would be punished (cheated) if they let the cheater get away with cheating.

Other contributions investigate how private organizations with coercive power over their members act to enforce contracts between members and nonmembers. Greif et al. (Reference Greif and Weingast1994) model a game in which a merchant guild, by threatening a boycott of a foreign ruler’s territory, induces the ruler to live up to promises of security and fair legal treatment in the ruler’s courts. Guilds were able to make this threat effective through their power to exclude from trade any of their members who violated the boycott. Greif (Reference Greif2006) argues that prior to the 13th century, communes were able to punish members who violated agreements with nonmembers (members of other communities); the communes had an incentive to do so because of a community responsibility system in which the foreign community would punish all members of the transgressing merchant’s community. If a merchant from Cologne, for example, cheated a merchant in London, all merchants from Cologne who traveled to London were liable to have their goods seized if the authorities in Cologne failed to punish the cheater.Footnote ¹⁷ The threat of seizure in London, in turn, provided merchants in Cologne with the incentives to police the behavior of all merchants in their community.

This theoretical literature has advanced significantly our understanding of the complexities of sustaining legal order through decentralized or private mechanisms, exposing the limits to Panglossian views of the naturally orderly world of customary commercial trade. But the pictures of “law” in this literature are relatively thin. Nothing in these models characterizes the rules as lawlike.

Consider Greif’s account of the Maghribi traders. He focuses on a solution to the problem of enforcing agency contracts that is naturally limited to a group of individuals who share (as a matter of common knowledge) the particular collectivist culture that sustains equilibrium beliefs and strategies. The challenge is to explain why this method of generating beliefs about retaliation did not transcend its cultural bounds. Greif argues that the individualistic beliefs shared by, for example, the Genoese merchants were not consistent with such a system and that the dominance of individualistic culture in medieval Europe forced Europeans to turn to a formal legal system—by which Greif means centralized coercive enforcement by an organized state with a local monopoly over the legitimate use of force.

Our model questions this line of analysis. We show that even purely self-interested traders, lacking the kind of prosocial preferences that Greif attributes to the Maghribis, can sustain an equilibrium based exclusively on decentralized collective punishment of the type that sustains the Maghribi equilibrium provided that there exists a centralized institution with legal attributes. Rather than solving the problem of beliefs by a switch to coercive enforcement, our approach suggests solving the problem of beliefs by a switch from shared culture to formal institutional structure. Of course, coercion may have aided this system by lowering the costs of punishment in some cases. But coercion is not the central feature of this system. We emphasize instead the coordinating mechanisms of a body of rules or laws with legal attributes in combination with an authoritative steward.

Greif reports that the Maghribi system was limited to merchants who were of recognizable membership in the group. Its success depended on social networks that both facilitated and made obligatory the exchange of commercial information. And it depended on the existence of informal customs to coordinate beliefs about what constitutes cheating by an agent. Greif argues that the feasibility of this solution limits the size of the group: if the group grows too large, this solution falls apart. Moreover, this solution depends on an exogenously supplied collectivist belief system.

Our approach suggests that other features of the Maghribis’ solution limited their ability to expand. Reliance on an informal and organic means of resolving ambiguity about what constitutes cheating is likely to have posed a substantial limit. With the growth of trade came an expanded division of labor and increasing specialization, as well as greater diversity of trading partners and settings. The innovation of commercial instruments and organizational forms that took place over the course of the medieval commercial revolution added to the complexity of trade. All of these features increase the problems of ambiguity.

Greif (Reference Greif2006, 69–71) recognizes the problem of resolving ambiguity among the Maghribis, framing the ambiguity problem as one of incomplete contracting—what an agent should do in an uncontracted-for contingency. He suggests that cultural rules can fill these contractual gaps and calls these cultural rules merchants’ law. He also tells us that Maimonides, “an important Jewish spiritual leader living in Fustat” (where the Maghribis were centered), provided some guidance in his legal code by prescribing that merchants should fill gaps “with the custom current in the land in regard to the merchandise they deal with” (71).

The express recognition of custom as a gap filler and, possibly, Maimonides’s code as a key to classification frames a series of questions, largely unexplored in the current literature, about the extent to which the body of rules employed by the Maghribis displayed the legal attributes we highlight. A body of cultural rules is likely to reflect both generality and impersonal reasoning; indeed, an important feature of custom is the accessibility of the reasoning to a wide group and the expression of obligations in ways that can be applied across a range of (customary) settings. Cultural rules also achieve common knowledge, as Chwe (Reference Chwe2001) has persuasively argued. For custom to succeed at the gap-filling role, the system of rules had to be clear; we expect there were cases in which it was indeed clear what custom required of merchants.

What we do not know, however, is how the system handled ambiguity as it inevitably arose. Greif provides no evidence that an authoritative steward existed, an institution that would have been needed when new contingencies arose not covered by the existing understanding of customs. Was Maimonides’s code authoritative in this sense? Was there a body of experts to whom the merchants turned for an authoritative declaration when it was unclear what “the custom current in the land” required? Greif (Reference Greif2006, 71) reports, “unfortunately, neither the legal literature nor the geniza documents reflect exactly how the merchants’ law was formulated and changed.” The evidence suggests that the merchants’ “law” may have been a regime of organic social norms still in the process of acquiring the attributes of law.

Our framework emphasizes the importance of the distinction between spontaneous or organic systems for classifying conduct and stewarded systems. Greif’s analysis collapses the distinction between law and culture in this regard; indeed he notes that a term he translates as “merchants’ law” can also be translated as “the way of the trade” (Reference Greif2006, 70). We believe in preserving the important distinction between custom and law as coordinating devices precisely because an institution that offers a unique authoritative steward—think of the Icelandic Law Speaker—can coordinate beliefs more extensively in settings of greater novelty and idiosyncrasy: the conditions we propose were increasingly important as the commercial revolution progressed. A system that depends on the slow adaptation of culture and customs about what constitutes “cheating” will be less effective at adapting and coordinating an equilibrium based on decentralized collective punishment.

The apparent lack of a formal steward as a coordinating device in the Maghribis’ reliance on culture adds an important insight to Greif’s account about why the formal legal institutions developed in Europe came to replace their culturally based system. Greif attributes the European turn to formal legal institutions to a difference in culture. As trade expanded in Europe, this difference in culture required, he argues, a complete change in the mechanism of enforcement: from a system of decentralized collective punishment to a regime of centralized coercive authority. But we think this view makes too big a leap. We agree that the different beliefs held by Europeans are likely to have facilitated the move to more formal institutions. But we disagree that the most important feature of the central institutions involved coercion. We emphasize instead the formal development of a coordinating device with legal attributes, including an authoritative steward. Coercion may have aided this system, but it is not the system’s central feature.

The multiple European solutions that merchants had available to them—the rule systems provided by sovereigns, religious bodies, merchant guilds, and cities—possessed to a significant extent the legal attributes we claim are needed to support widespread legal order based on decentralized collective punishment. To varying degrees, these rule providers struggled to devise decentralized collective enforcement systems that were characterized by a public common logic, capable of implementation through impersonal reasoning, and under the authoritative stewardship of an identifiable and neutral entity.

As an example, consider a typical set of rules, those established by Flemish cloth merchants in 1240: “The Ordinance of Those Men of Ypres and Douai Who Go to England” (Moore Reference Moore1985, 297). The first provision stated that “if a merchant returns a cloth after he has bought it, giving no reason for the return … then henceforth no man of Ypres or Douai shall let him take any cloth away from any of our shops until he shall have paid the full price.” This is an express example of the setting we model in Hadfield and Weingast (2012): a buyer who violates the rules established by a group of sellers is punished by a set of simultaneous decisions among all of them to boycott the offender. Punishing an offender is individually costly: refusing to deal with a merchant will sometimes mean forgoing the profit on a transaction. This collective punishment scheme therefore faces problems of coordination and incentive compatibility.

The coordination problem was addressed by the creation of the document articulating the rules—which stated that it was to be “read out before the community in every fair”—and the establishment of procedures for determining ambiguous cases: Flemish merchants traveled in convoys to English fairs with wardens with the authority to adjudicate disputes (Moore Reference Moore1985, 96–97). As stated in a subsequent guild ordinance of 1261 dealing with the purchase of wool by Flemish merchants from English sellers, “there will be in each of these cities one man to view and judge the grievances” (301). In the language of our model, these individuals served as the authoritative stewards of the common logic encapsulated in the Flemish cloth merchants’ ordinances. Their role in elaborating rules to accommodate new and unforeseen circumstances was reflected in provisions empowering wardens to add “any good rule which is not included here,” with the agreement of the community; their authority as managers of a reasoning process without wielding legislative power, however, was also reflected in the limitation that while they could expand on the rules, they could not abolish provisions unilaterally.

This attention to the distinctive attributes of institutions that coordinate legal order also helps to identify a weakness in the analysis of the Law Merchant provided by Milgrom et al. (Reference Milgrom, North and Weingast1990). The authors demonstrate, in contrast to a system relying exclusively on culture, that a multilateral reputation system—a form of decentralized collective punishment—can be built in a context that does not presuppose shared cultural beliefs. Milgrom et al., as with Greif, pay little attention to the problems of ambiguity in the concept of cheating and why the merchant judges produced a body of rules that were recognized as law. Their account of the Law Merchant, as a consequence, lacks attention to attributes of this institution beyond its capacity to serve as a common repository of judgments about cheating. Our framework, by focusing explicitly on the institutional attributes that support coordination among heterogeneous agents when ambiguity is a critical consideration, supplies these attributes.

Our approach sends us back to the historical record to look at how these attributes evolved. We wonder in particular how merchant guilds reconciled the need to preserve the attributes of uniqueness and common knowledge of classification among all those whose participation was essential for credible enforcement with the goal of expanding the power of the threat. Ideally, once a stable set of rules is in place, a guild would want to expand the set of merchants who looked to those rules to classify commercial behavior. Expanding the group of those who participate in collective punishment, however, requires universality in the rules.

We know that the merchant guilds struggled to ensure that all of the members used their rules, and only their rules, to guide their behavior. Given the multiplicity of available rule systems and courts, exclusive use of the guild’s rules was essential to secure the credibility of the punishment threat from other members of the guild. The guilds of northern Italy, for example, obliged their members to bring disputes to the guild court rather than the ordinary civil court: “the gild did not hesitate to expel members who ignored its claims to jurisdiction, and to forbid trade or commerce with them” (Mitchell Reference Mitchell1904, 42–43). The reliance on guild rules and adjudicators traveled with the merchants (51). As we have seen with the Flemish merchants, guilds selected individuals to travel with them and adjudicate disputes at English fairs (Moore Reference Moore1985, 99). Even within England, the merchants of individual English cities looked to their own guilds for rules and adjudication when they traveled to English fairs.

Guild efforts to ensure that their members used the guild’s rules exclusively for transactions were no doubt in part driven by the belief that their rules were more desirable for them than other rules. But the emphasis on guildwide adherence to a common set of rules also suggests the importance of common knowledge of the rules to support collective punishment. Our framework suggests another reason for such keen concern at the guild level. A merchant who took his case to a nonguild court not only imposed a cost on the merchant with whom he had a dispute but also created an externality for the guild as a whole, by degrading the clarity and uniqueness of the rules to which all guild members looked in deciding when to participate in collective punishment.

Achieving clarity and uniqueness in the system provided by an individual guild, however, came at a cost: the enforcement threat was limited to the collective punishment that could be supplied by the members of the guild. There would therefore, we suggest, have been pressure to attract merchants beyond a guild to the guild’s rules, to expand the power of the punishment. Greif et al. (Reference Greif and Weingast1994) describe, for example, the slow evolution of the German Hansa: increasing the number of cities willing to boycott on the basis of a centralized set of rules and procedures to discipline foreign rulers who failed to respect the charters reached with the coalition.Footnote ¹⁸

Our framework shows that expanding the set of potential punishers, however, required the achievement of universality—the creation and development of rules that sufficiently took into account the interests of those the guild sought to attract—as well as credible promises that the rules would be implemented in neutral ways, even when adjudicating between member and nonmember interests.Footnote ¹⁹ In a regime that depends significantly on decentralized collective punishment, significant expansion can occur in a stable way only if it succeeds in attracting participants by offering them something better than their alternative enforcement option.

The countervailing pressure to secure common knowledge among a group of potential punishers and yet to expand the set of punishers also sheds some light on the claim—most strongly made by Benson and Trakman—that the Law Merchant was a unified set of rules, common across merchants throughout the commercial world. The history is more complex. From the perspective of an individual merchant, commercial law must have seemed a riot of rules. Critical contract questions were resolved in very different ways by different rule systems.

The pressure to build a set of rules that all merchants recognized as the relevant reference point for evaluating commercial conduct was undoubtedly great. Again, the broader the set of merchants who recognize the rules, the more powerful the collective enforcement threat. Does a merchant acquire a reputation for reneging on contracts after they are formed? The reputational threat is much greater if merchants—whether trading at the fairs of St. Botulph or Champagne, in Bruges or in Milan—all recognize the same rules for determining what counts as a valid contract.

VI. The Role of Centralization

The essence of law is a set of rules characterized by legal attributes, such as generality and universality, and an authoritative steward for removing ambiguities and adapting the rules to changing circumstances. The definition is independent of the means of enforcement, and we have emphasized the existence of law in settings that depend solely on decentralized enforcement.

What role, then, does centralization play in a system of law and its enforcement?Footnote ²⁰

Although this history has yet to be written, our model suggests that the move to formal, centralized state-coordinated legal systems involves characteristics that go well beyond coercion. We suggest that this move has more to do with the capacity for a centralized system based on exclusive territorial jurisdiction to accomplish various ends complementary to decentralized enforcement, namely,

• to generate and sustain the legal attributes covering a wide range of activities well beyond those involving merchants,

• to sustain common knowledge about the rules in a large diverse population,

• to coordinate decentralized enforcement and to enhance the power of that enforcement by placing all members of a large society under the same rules,

• and to obtain the advantages of an authoritative steward for adapting the rules to changing circumstances.

One underappreciated feature of “state” courts that operate extensively throughout a large region, and the development of territorial as opposed to personal jurisdiction—quite apart from any coercive power of the state—is that these courts can achieve a level of common knowledge, universality, authoritative stewardship, and impersonal reasoning not available to smaller, membership-based systems. Several independent systems of laws and legal stewards necessarily involve ambiguities about the boundaries between the different systems. In contrast, a centralized system removes this ambiguity and provides a systematic means of resolving conflicts of laws arising in different areas.

Royal authority, quite apart from coercive power, may have offered these advantages and laid the groundwork for the emergence of royal courts as a single coordinating system for the realm. Consider King Edward’s Carta Mercatoria from 1303. Expressing a concern to increase the “tranquility and full security” of foreign merchants who bring their goods to England, the document first promises foreign merchants the right to enter the realm “safely and securely under our defence and protection” and free of certain taxes. It also grants foreign merchants the right to lodge in England’s cities, boroughs, and towns. Coming second only to these two basic provisions is the announcement of a uniform rule—across England and across different commodities—for determining when a contract has been formed and which rules apply.Footnote ²¹

The Carta Mercatoria thus seems a major step in creating an authoritative stewardship over a unique and (at least aspirationally) neutral and universal system of rules. Note that the document does not specify what these rules are—beyond the critical (and often overlooked) ambiguity of when a contract has come into existence. It merely establishes which local customs and rules will apply and provides promises that this system will be under the stewardship of designated individuals who are subject to royal punishment if they fail in their stewardship. Finally, we emphasize that the document does not mention royal resources dedicated to enforcing the judgments reached by these officials; the only promise of royal protection pertains to their persons and goods.

This evidence highlights the importance of the coordination of decentralized enforcement in the emergence of legal order and the evolving role of the state in medieval trade. In this we follow in the footsteps of Hart, who observed, “The history of law … strongly suggests that the lack of official agencies to determine authoritatively the fact of violation of the rules is a much more serious defect [than the absence of an official monopoly of ‘sanctions’]; for many societies have remedies for this defect long before the other” (1987, 93–94). Law is principally about legal attributes, an authoritative steward, and the coordination of decentralized enforcement. Centralization and coercion may enhance this system; but centralized coercion is not essential to law and its enforcement.