Law as a Private Good


A Response to Tyler Cowen on the Economics of Anarchy


In a recent article in this journal, Tyler Cowen (1992) argues that an anarcho-capitalist system of private protection agencies of the sort proposed by myself and others is unworkable. I believe he is mistaken. In explaining why, I will start with a brief explanation of how I believe that such a system would work, then summarize Professor Cowen's arguments and attempt to show why they are wrong.

Imagine a society with no government. Individuals purchase protection of themselves and their property from private firms. Each such firm faces the problem of possible conflicts with other firms. Private policemen working for the protection agency that I employ may track down the burglar who stole my property only to discover, when they try to arrest him, that he too employs a protection agency.

There are three ways in which this problem might be dealt with. The most obvious and least likely is direct violence-a mini-war between my agency, attempting to arrest the burglar, and his agency attempting to defend him from arrest. A somewhat more plausible scenario is negotiation. Since warfare is expensive, agencies might include in the contracts they offer their customers a provision under which they are not obliged to defend customers against legitimate punishments for their actual crimes. When a conflict occurs, it would then be up to the two agencies to determine whether the accused customer of one will or will not be deemed guilty and turned over to the other.

A still more attractive and more likely solution is advance contracting between the agencies. Under this scenario, any two agencies that faced a significant probability of such clashes would agree on an arbitration agency to settle them-a private court. Implicit or explicit in their agreement would be the legal rules under which such disputes were to be settled.

Under these circumstances, both law enforcement and law are private goods produced on a private market. Law enforcement is produced by enforcement agencies and sold directly to their customers. Law is produced by arbitration agencies and sold to the protection agencies, who resell it to their customers as one characteristic of the bundle of services they provide.

One attractive feature of such a system is that the usual economic arguments for the efficiency of market outcomes apply to the legal system and the rules it generates. To see why, imagine that there is some change in the legal rules currently prevailing between two enforcement agencies which would yield net benefits to their customers. If it benefits both sets of customers, then it is in the interest of the protection agencies either to persuade their arbitration agency to make the change or to shift to one that follows the superior set of rules. If it benefits the customers of one agency but imposes costs on the customers of the other, with net costs smaller then net benefits, then it is in the interest of the two agencies to agree to make the change, with the loser compensated either directly or by some other change elsewhere in the legal rules. In practice, since it is the arbitration agencies that specialize in legal rules, we would expect them to try to develop superior legal codes in the process of competing for customers. The result should be a set of legal codes that are economically efficient in the conventional sense.[1]

While this argument implies an efficient set of legal codes, it does not tell us which efficient set. Legal rules have distributional as well as allocational consequences. Imagine, for example, that agency X and agency Y represent customers with different tastes in legal rules. Perhaps the customers of X support the death penalty and the customers of Y oppose it. The argument of the previous paragraph implies that whichever group most values its preferred legal rule will get it. But it does not tell us which agency will have to pay the other in order to get its way in disputes between their customers. Will X have to pay Y to get its agreement to a pro-death penalty court for disputes between their customers, or will Y have to pay X if it wants an anti-death penalty court? I have described the logic of the bargaining, but not the starting point-the default rules from which mutually beneficial changes will be made.[2]

The answer is that the distributional starting point is the solution to a bilateral monopoly bargaining game between the agencies. Each agency can threaten to refuse to agree to any arbitrator, subjecting both to the costs of occasional violence, or at least ad hoc negotiation to avoid violence. Each knows that the other would prefer even a rather unfavorable set of legal rules to no agreement at all. The situation is analogous to a union management negotiation or the negotiations determining borders, trade policies, and the like between neighboring countries. While there is no good theoretical account of exactly what determines the outcome of bilateral monopoly bargaining, experience suggests that some reasonably efficient equilibrium usually exists. Most unionized firms manage to settle their differences without lengthy strikes, and most nations are at peace with most of their neighbors most of the time.[3]

So we may imagine the market for law as starting out with a set of default rules between each pair of protection agencies, representing the result of bargaining backed by threats of refusal to agree on an arbitrator. From there, the agencies bargain to an efficient set of rules. Where the change benefits both, it may occur without side payments. Where the change is preferred by only one agency, it must pay the other enough to obtain its agreement. The distributional outcome is the result of an implicit threat game between the agencies; the allocational outcome is the result of a (logically subsequent) bargaining game to move from the starting point to the Pareto frontier.

Experience suggests that there is enormous inertia in mutual threat games of this sort. National boundaries do not move half a mile one way or the other each time one nation becomes a little richer or a little more powerful. In practice, an anarcho-capitalist society will probably be built not so much on an ongoing mutual threat game as on a mutual threat game played out in the distant past. That suggest that, once the initial equilibrium has been established, the success of a protection agency will be based mainly on its ability to produce protection for its customers, not its ability to defeat rivals in open warfare.

While it is always possible for one firm to threaten to withdraw from its arbitration agreement with another unless the terms are renegotiated de novo, such threats are unlikely to be either common or successful. Other agencies have a strong incentive to insist on basing their bargaining on the existing rules, in order to prevent the costs of continual renegotiation and the costs of violence when negotiations break down.

I have now described anarcho-capitalism as I believe it would function in a modern society.[4] What are Cowen's reasons for believing that such institutions would be unstable?

Cowen (1992) writes:


"... the argument is that Friedman's scenario is not an independent alternative. Competing law codes are stable only if they evolve into a dominant agency or arbitration network ... .

Agencies might eschew warfare in favor of arbitration and interagency cooperation. Agencies would agree in advance how interagency conflicts will be settled. Common standards would be applied for criminality, punishment, and criminal procedures when disputes occur ... .

A systematic arbitration network would arise to encourage the orderly application of law. Although intra-agency conflicts might be settled differently from interagency conflicts, society would possess effectively a single legal code. ... At the very least, agencies abide by higher-order arbitration. ... "


So far I agree with Cowen,[5] provided that the accent is put on the final sentence and that it is recognized that what is described is an equilibrium, not a constraint. Firms almost always abide by arbitration because it is almost always in their interest to do so. Describing this as a single legal code is, however, somewhat misleading, since there may be as many legal codes as there are pairs of agencies.[6]

The distinction between a market equilibrium and a constraint is not merely a verbal one. Consider the analogous case of an ordinary competitive market. Economic theory tells us that firms selling identical goods will all charge the same price. That does not mean that firms are not free to change their price if they wish, nor that a change by one firm will somehow force every other firm to make an identical change. On the contrary, the analysis of what the price will be depends on the assumption that each firm is free to set whatever price it wishes, and deduces both the existence and level of the common price from that assumption. Similarly, protection firms under anarcho-capitalism will agree on arbitrators to settle disputes between them, but that is a consequence of their profit maximizing behavior not a constraint upon it. The fact that they are free to refuse to agree to arbitration is one of the elements that determines what the actual terms of arbitration will be.

Cowen then writes:


"Unlike Nozick's ultraminimal state, the network consists of more than one firm. ... The presence of a network gives rise to contractual relations that induce firms to behave cooperatively, as if they were one large firm. Whether the common arbitration network is "one big firm," or "many cooperating smaller firms" is primarily a matter of semantics. The network can just as well be considered a single firm with separate divisions that compete to some degree. Each division has its own set of residual claimants, but the behavior of divisions is constrained to favor the interests of the entire network."


So far as I can tell, this final assertion is nowhere justified, and I believe it to be false. What Cowen describes as a "network" is simply a set of private firms-protection and arbitration agencies-linked by a large number of contracts. Each pair of protection agencies has a contract specifying an arbitrator for disputes between their customers, and each protection agency has contracts with one or more arbitration agencies specifying the terms on which they will arbitrate its disputes with specified other protection agencies.

Nothing in this situation requires or implies a single firm controlling the whole, nor anything analogous to one. The network as I have described it has no decision making body. Its "decisions," the set of legal codes it enforces, are the outcome of independent profit making decisions by the individual firms and bargaining between pairs of firms. Nothing in the logic of the market for protection and arbitration implies that the outcome will maximize the summed profits of the firms, as Cowen seems to assert. Indeed, ordinary economic theory suggests that in equilibrium this market, like any competitive market, will yield zero profit to the firms that make it up.

Consider Cowen's argument applied to a less exotic industry-groceries. As a practical matter, any grocery that wishes to stay in business must have contracts with a number of large suppliers, such as Kraft and General Mills, either directly or through distributors that function as intermediaries. Thus all grocery stores are linked together by contracts with common intermediaries. The whole collection of firms-grocery stores, producers, wholesalers-could be described as a network in the same sense in which Cowen describes the protection agency as a network.[7] Does it follow that, in the grocery industry, "contractual relations ... induce firms to behave cooperatively, as if they were one large firm?" Is there any reason to believe that the behavior of the separate firms "is constrained to favor the interests of the entire network?"

Grocery stores and protection agencies are indeed constrained, but it is not their own interest that they are constrained to follow. Grocery stores are constrained to follow policies that maximize the welfare of their customers, and protection agencies are constrained to enforce legal codes that maximize the welfare of their customers, for essentially analogous reasons. In both cases the constraint is only approximate, due to the familiar problems of imperfect competition, imperfect knowledge, externalities, and the like. But nothing in the logic of either market leads to maximization of the interests of the industry.

Having asserted that the protection industry is in effect a single firm, the next step in Cowen's argument is straightforward.


"The existence of a common arbitration network creates a vehicle for protection agency collusion. Members of the network find it profitable to write a contract agreeing not to compete with each other. The agencies restrict output and raise prices, thus reaping monopoly profits. Network membership requires contractual acceptance of jointly determined prices and outputs, as well as legal procedures. ..."


Let us see how this works-remembering that the "network" is not a firm but a set of contracts among a large number of firms. Firm A announces that it will only agree to arbitration agreements with other firms that agree to restrict output and raise price. Firm B treats this offer like any other move in its negotiations with firm A-it accepts it if the agreement, along with any compensation offered by firm A for agreeing, makes it better off, otherwise it insists on sticking to the old terms.

But firm B could have raised prices and restricted output without any demand from firm A. The reason it did not was that doing so would have lowered its profits. It will accept A's demand only if A is willing to pay enough to make up for the resulting losses. The situation is no different than in any industry (without anti-trust laws) where one firm attempts to create a cartel. As in any industry, it is possible to have a profitable cartel if all of the firms can somehow agree to and abide by a cartel agreement, while keeping out new entrants. One reason that is difficult, here as elsewhere, is that if some subgroup of the industry forms a cartel it is in the interest of all the non-members to undercut the members.

Cowen seems to imagine the collusion occurring not at the level of the firm but at the level of the network. He writes:


"The ability to collude successfully is inherent in the nature of the network. The network can internalize the externalities problem behind peaceful adjudication only by suspending quality competition - that is, by offering a uniform set of laws or higher-order adjudication procedures. The ability to engage successfully in quality collusion, however, implies that other kinds of collusion are possible also."


But, as we have already seen, nothing in the logic of the system requires either uniform laws or any single body determining such laws. It is, of course, possible that there will be one or more bodies offering model legal codes, and that many firms may adopt such codes in order to reduce the costs of legal diversity. To the extent that reasonably uniform standards prevail, collusion will be somewhat easier since there will be fewer dimensions on which the collusive agreement must be defined and adherence to it monitored. But a standard setting body does not, as Cowen seems to assume, provide an enforcement mechanism for a cartel. Non-member firms can abide by non-price standards while chiselling on the cartel's price.

Cowen also argues that "anarchy is orderly only under the condition that the network can act collectively to prevent outlaw firms from gaining sizable market share. If the network can implement successful sanctions against outlaws, however, the network can also implement successful sanctions against potential competitors."

Here again, he is misinterpreting the anarcho-capitalist system, or at least the version of it that I proposed and he earlier cites. In a system of a hundred agencies of equal size, one of which is an outlaw, each of the ninety-nine others settles ninety-nine percent of its conflicts by arbitration and one percent by violence. The outlaw agency settles a hundred percent of its conflicts by violence. Since violence is much more expensive than arbitration, the outlaw's costs are much higher than the costs of its competitors. Agencies that are unwilling to sign arbitration agreements acceptable to most other agencies with which they are likely to come in conflict are prevented from gaining market share not by some collective action by "the network" but by the difficulty of selling a product when your production cost is much higher than your competitors'.

There is, it is true, one special feature of this market which might make cartelization easier. If all the existing firms do agree on a common anti-competitive policy, they may well have the physical force necessary to enforce it by keeping new firms from forming. I discussed this possibility at some length in Friedman (1989), where I wrote:


"The protection agencies will have a large fraction of the armed might of the society. What can prevent them from getting together and using that might to set themselves up as a government?

... our present police departments, national guard, and armed forces already possess most of the armed might. Why have they not combined to run the country for their own benefit? Neither soldiers nor policemen are especially well paid; surely they could impose a better settlement at gunpoint.

... A brief answer is that people act according to what they perceive as right, proper, and practical. The restraints which prevent a military coup are essentially restrains interior to the men with guns.

We must ask, not whether an anarcho-capitalist society would be safe from a power grab by the men with the guns (safety is not an available option), but whether it would be safer than our society is from a comparable seizure of power by the men with the guns. I think the answer is yes. In our society, the men who must engineer such a coup are politicians, military officers, and policemen, men selected precisely for the characteristic of desiring power and being good at using it. They are men who already believe that they have a right to push other men around--that is their job. They are particularly well qualified for the job of seizing power. Under anarcho-capitalism the men in control of protection agencies are selected for their ability to run an efficient business and please their customers. It is always possible that some will turn out to be secret power freaks as well, but it is surely less likely than under our system where the corresponding jobs are labeled 'non-power freaks need not apply.'

In addition to the temperament of potential conspirators, there is another relevant factor: the number of protection agencies. If there are only two or three agencies in the entire area now covered by the United States, a conspiracy among them may be practical. If there are 10,000, then when any group of them start acting like a government, their customers will hire someone else to protect them against their protectors.

How many agencies there are depends on what size agency does the most efficient job of protecting its clients. My own guess is that the umber will be nearer 10,000 than 3. If the performance of present-day police forces is any indication, a protection agency protecting as many as one million people is far above optimum size.

My conclusion is one of guarded optimism."

(Friedman (1989), pp. 123-124).




Buchanan, James M. 1974, review of The Machinery of Freedom: Guide to a Radical Capitalism by David D. Friedman, Journal of Economic Literature XII, 3 (September 1974), pp. 914-915.

Cowen, Tyler 1992, "Law as a Public Good: The Economics of Anarchy," Economics and Philosophy, 8 (1992), pp. 249-267.

Friedman, David 1993, "A Positive Account of Property Rights," forthcoming in Social Philosophy and Policy.

Friedman, David 1979, "Private Creation and Enforcement of Law -- A Historical Case." Journal of Legal Studies 8, 399-415 (March 1979).

Friedman, David, 1989 The Machinery of Freedom, Guide to a Radical Capitalism, 2nd ed. La Salle, IL: Open Court.

Posner, Richard 1992, Economic Analysis of Law, 4th edn., (Boston: Little Brown, 1992)


Back to the
list of articles.

Back to my home page.