enforcement, let us at least explore the possibility and see whether it would be workable.The most obvious question about restitution is whether it would be feasible for criminals to pay for their jail accomodations and pay off their victims. Isn’t it unrealistic to think that impoverished criminals could have sufficiently high marginal products to do so? It is difficult to know for sure, but throughout history slaves and indentured servants have commanded high prices. (The median price of a slave in 1860 was $1500, at a time when median annual income for free workers was $200.)65 Even in primitive times, the discounted income stream of a slave’s net earnings (marginal product minus the costs of food, shelter, and monitoring) was positive. And Ransom and Sutch, in their study of American slavery, found that the “rate of exploitation” (the percentage of a slave’s income that the owner was able to expropriate) for 1859 was about 53.7%.66 Consider further that slaves were relatively unskilled workers in a semi-industrial society. As a society grows wealthier, the fraction of the average worker’s income that goes toward basic necessities declines. Consequently, the feasible rate of exploitation should increase in step with per capita income, making restitution systems more feasible in modern societies.A second fear is that private prisons would mistreat prisoners. It is difficult to imagine that any private prison could be worse than public prisons, with their reputation for abuse and violence. Still, prisoner abuse in private prisons could be a serious problem. Courts might help by grading prisons according to their strictness (e.g., parole with weekly checkups; evening and weekend detention only; minimum security; moderate security; maximum security); then they could set a maximum “security level,” along with the restitution fee. The convicted criminal might then bargain with the victim for better living conditions in exchange for more restitution. Alternately, the court could fix the total restitution without reference to prison costs, so that the victim would have to weigh the greater security of stricter prisons against the lower return. Another possible way to protect criminal rights is to give inmates the right to switch to any comparably restrictive prison that would take them.Even if there were no court-imposed checks on the treatment of prisoners, they would probably be much better off than today. While scholars of American slavery differ on some points, they agree that slave-owners had a strong incentive to make sure that their working slaves were adequately fed, housed, and otherwise cared for. (Non-working slaves, such as children, received markedly worse treatment.67) In particular, profit-making firms would have strong incentives to prevent prisoners from using violence against one another. Some obvious tactics would be to strictly separate violent and non-violent criminals, to segregate prisoners according to might and size, etc. There would also be incentives to monitor abuse by guards.Prisons might invest in prisoners’ human capital, since the firms could collect a large fraction of the benefit. One common complaint about prisons is that they release prisoners without any job skills; hence, the notoriously high recidivism rate of our public prison system. While a private prison could not reap the full benefit of training unless a criminal’s debts added up to a life sentence, it would at least have some incentive to invest in job training. Most students of slavery argue that prejudice and social sanctions prevented slave-owners from training slaves for skilled professions. Prison entrepreneurs, in contrast, might try to turn inmates into skilled workers.Another way that private prisons might reduce recidivism would be to hire criminals after their sentence runs out. Since by assumption the criminal was profitable to employ while serving time, he should still be profitable to employ after he pays off his debt. Indeed, since he would no longer owe restitution, and since his monitoring costs would fall, he would likely get a raise. While the failure of public prisons to rehabilitate criminals surely owes a great deal to the criminal personality itself, profit-making prisons might better prepare criminals to become self-supporting members of society. (Whereas prisons, as the quip goes, prepare criminals to become better criminals.) Prisoners would have an incentive to work as hard as possible because they would only get released once they paid off their debt. For this reason, restitution has been called a “self- determinative sentence.” The prison time is fixed not in terms of years, but in dollars of work; criminals who work hard and cooperate with management get out sooner. (To correct for the problem of super-wealthy criminals, courts might grade restitution according to the criminal’s wealth.) This is the central positive incentive to work hard in restitution-based systems. Prisons might hold out other positive incentives, such as television, nicer cells, free time, and so on. It is difficult to know a priori what incentives would best elicit work effort from criminals, but this is surely a possibility. Public prisons are so conservative that they ignore attractive innovations; but anecdotal evidence from a few experimental prison programs is quite heartening. During the 1970’s the Maine State Prison (conveniently located along a major tourist route) allowed prisoners to go into business for themselves. The prison issued canteen coupons for currency, allowed workers to hire each other and form “firms”, and eventually set up a patent system for novelties. Over two-thirds of the prisoners participated, and out of this labor pool arose some exceptional entrepreneurs. One criminal-entrepreneur bought out the prison canteen and ran it at a profit; another employed 30-50 other workers and branched out into television sales and rentals. Nevertheless, this highly successful program got shut down in 1980 due to bureaucratic resistance.68 While we should be cautious about generalizing from one case, the Maine State Prison experiment does hint at the innovative potential of private prisons. After all, it takes entrepreneurial talent to be a drug-dealer. Maybe the problem with past rehabilitation programs is that they tried to completely remold inmates’ characters rather than appeal to their entrepreneurial spirit.A final complaint is that restitution systems could only work for crimes with a small number of identifiable victims. This is largely correct, although class action suits against pollution and other public bads exist. But there remain other crimes, like drug possession or prostitution, that seem difficult to fit within the restitutive model. Rather than digress, I will briefly point out some possibilities. First, one might get rid of victimless crime laws, precisely because they don’t have an identifiable victim. Second, (this is apparently what Stigler and Becker suggest) whoever “narcs” on the law violator would collect fine , even though the user did the informant no harm. Third, one might use restitution for crimes-with-victims and government enforcement for victimless crimes. Although I find the first option ethically superior, disagreement should not lead us to dismiss restitution altogether.Now that I have discussed the popular objections to restitution-based criminal enforcement, we shall turn to Judge Posner’s technical economic criticisms. Posner argues that private enforcement for torts and contracts and public enforcement of criminal law is economically rational because p (=the probability of apprehension and conviction) is likely to be near unity in the former and low in the latter case. As Posner suggests, “The victim of a breach of contract knows who breached it; the victim of an automobile accident usually knows the identity of the other driver; but the victim of a burglary does not know the burglar’s identity.”69If you recall section 4.4, Posner’s complaint about the Becker-Stigler bounty system was that it did not give incentives for an optimal severity-probability mix. But Posner embraces private enforcement for tort and contract because, he says, the probability is near unity in civil cases. Hence higher severity won’t induce greater expenditures because p is already at a maximum. In contrast, crimes have low p’s; therefore, higher severity induces greater spending on enforcement. And as Posner pointed out earlier, such expenditures are suboptimal because we could get the same level of deterrence by merely raising the severity and slashing the enforcement budget (and thereby the probability of conviction).Posner is not in his best form here; he forgets that p=the probability of apprehension and conviction, not just the former. All that his examples prove is that victims of torts and breach of contract are more likely to know who did it than victims of crimes. But that hardly shows that the probability of conviction is 1 without any expenditures! All of our statistics on the amount of money spent on civil suits indicates that the sum spent on the enforcement is large indeed: Cooter and Ulen estimate that the social cost of trials is at least $400/hour, and the litigants pay at least half of that out of their own pockets.70 And this doesn’t even consider pre-trial expenses and the cost of legal battles settled out of court. Evidently, the parties to these disputes think that they can adjust p by spending more money. Posner is too quick to assume that civil suits don’t encourage enforcement expenditures; plainly they do.Anyway, even if Posner were correct that p=1 for civil suits, we could still get cheaper enforcement by arbitrarily increasing the level of damages and decreasing the number of suits tried. We may not be able to further increase the probability of conviction, but we could certainly decrease it by, for example, letting the government take over the prosecution of civil suits and permitting selective enforcement. (Alternately, the government could let would-be plaintiffs draw straws for the right to bring suit.) And if Posner is correct that optimal enforcement requires high severity and low probability, that is the the correct strategy. Overall, Posner’s effort to rationalize private enforcement of civil law and government enforcement of criminal law is poorly thought out.The preceding passages appear to justify government enforcement of both civil and criminal wrongs. Both of them have the same inefficiency that Posner brought up earlier — they spend excessive resources on enforcement, when it would always be possible to get the same level of deterrence by arbitrarily increasing the severity and slashing enforcement. Most people, while admitting the economic logic of Posner’s suggestion (i.e., near-infinite severity and near-zero enforcement expenditures), find it morally appalling. I share this reaction, but I also think that there are two solid economic objections to Posner’s proposal.First, criminals tend to be strong risk-preferrers. As the probability of conviction drops, they are sufficiently myopic to ignore the severity. Overall, they are exceptionally short-sighted and present-minded, and would probably not respond well to a Posnerian criminal justice system.71 Indeed, brief reflection shows that high severity, low-probability harm frequently fails to deter law-abiding citizens. When you drive, you expose yourself to a small risk of death. The probability is low and the severity is high (indeed, infinite). If incentives like this don’t even induce regular people to give up driving, why should we think that they would induce criminals to give up murder, rape, or theft? Posner builds his whole model on the assumption of risk-neutrality; but especially in the case of criminal law, the assumption is ill-founded. (In all fairness, Posner alludes to this problem on p.170 of his Economic Analysis of Law, but does not otherwise discuss it.)Second, the higher the severity of punishment, the greater the procedural safeguards normally become. Neither judges nor the public would tolerate gruesomely severe punishments unless they were extremely certain of the guilt of the accused. This is a common explanation for the increasing difficulty of conviction: judges are reluctant to sentence people to prison because the punishment is so horrible.72 Posner briefly notes that since the error costs are higher in criminal suits than civil suits, there is an economic rationale for their respective required levels of proof.73 What I am arguing is that at least in democratic countries the level of procedural safeguards is often a function of the severity of the punishment. We can easily imagine that as a Posnerian fine-setter kept increasing the severity of punishment asymptotically to infinity, the judicial system would respond by asymptotically decreasing the probability of punishment to zero. This phenomenon is especially evident in death penalty cases: because judges are afraid of executing the wrong person, the trials drag on for years and cost the taxpayers millions of dollars.There are, then, at least two reasons to doubt that the optimal severity-probability mix is actually infinite severity with near-zero probability. First, since criminals tend to be risk- preferrers, effective deterrence would probably require a considerably higher probability. Second, as punishments become even more severe, the judicial system responds by making the probability of conviction ever more minute. In the real world the severity-probability mix requires a markedly higher probability and lower severity than Posner suggests. What is the optimal mix, then? That question is hard to answer. But since Posner’s mix is inefficient in the real world, it doesn’t matter if restitution-based enforcement systems (or private enforcement of contract and tort law, for that matter) deviate from his prescription.Posner, if you recall, argued that apprehension is easy in civil cases and difficult in criminal cases. “The victim of a breach of contract knows who breached it; the victim of an automobile accident usually knows the identity of the other driver; but the victim of a burglary does not know the burglar’s identity.”74 While there is some truth in this point, he perhaps neglects the possibility that the likelihood of apprehension would be greater if the victims had some incentive to assist the apprehension and conviction process. There are, after all, high costs of participation in the criminal justice system. Prof. Benson describes the victim’s incentives aptly: “The victims’ cost of cooperating with prosecutors can be staggering. In addition to the initial loss to the criminal, victims face the costs of transportation, babysitting, and parking. More importantly, they can lose wages and they endure seemingly endless delays and continuances. There are also considerable emotional and psychological costs of having to confront an assailant, for example, or enduring a defense attorney’s questions.”75 Just as the possibility of damages induces the victims of civil wrongs to help apprehend and “convict” tortfeasors and contract-breakers, restitution gives an incentive for crime victims to help apprehend and convict criminals. Cictims of most crimes have no incentive to report them. Unless revenge is a sufficient impetus, they are better off if they just forget what happened. Nonreporting of rape is well-known in our society; but this is only part of a general trend. Summarizing the well-known Figgie Report on Fear of Crime, Benson states that, “an estimated 60 percent of all personal larceny cases where there is no contact between the thief and his victim go unreported; and less than 50 percent of all assaults, less than 60 percent of all household burglaries, less than 30 percent of household larcenies, and only a little more than half of all roberries and rapes are reported.”76 If victims knew that they could get restitution if their assailant were caught, they would be more likely to report crimes. Similarly, they would probably be more willing to use their time to assist police if they knew that restitution awaited them once the crime was solved.Restitution incentives would also give victims an asset with which they could hire lawyers on contingency. Under the current system, the district attorney chooses which cases to prosecute and which to dismiss. Prof. Benson suggests that district attorneys’ value in the private sector depends mainly on the number of successful prosecutions and their conviction rate. In consequence, they pick cases where it is easy to get a conviction, even if the crime is relatively unimportant compared to alternative cases with a lower probability of conviction. Moreover, since DAs seek to maximize their conviction rate, they do so at the expense of sentence severity.77 (And police often get rated by total number of arrests, so they arrest people for crimes that they know the district attorney won’t prosecute because the probability of conviction is too low.) The moral here is that bureaucracies cannot directly measure their success according to consumer satisfaction. Instead, they must use imperfect proxies, such as conviction rate and number of arrests. Once you evaluate job performance on this basis, you create incentives to elevate the proxy variable above the end, to sacrifice actual crime deterrence to conviction rates and arrest statistics. These features of our criminal justice system echo the inefficiencies of the Soviet economy. Since performance was not measured by profitability, the central planners had to come up with proxies for good performance: typically, numerical quotas. For example, shoe firms were evaluated according to the number of shoes they produced. Since any sort of shoes would count, they predictably overproduced the cheapest quota-fulfilling footwear: baby shoes.78 Just as measuring performance according to the total number of shoes leads to too many baby shoes, measuring the job performance of police according to arrests alone leads to too many unprosecutable arrests. In a parallel fashion, measuring the job performance of district attorneys according to conviction rates leads to too many prosecutions of unimportant crimes and excessively low severity of punishment.In contrast, a restitution-based system would lead lawyers to pursue victim satisfaction directly. So long as the court properly sets the restitution sums (i.e., correctly ranks them according to severity), the market will automatically spend more resources on the more important crimes. Since the lawyers would probably be paid on contingency (and, in any case, would be answerable to the victim who hired them), they would not have incentives to improperly trade off severity of punishment for certainty of conviction. Most economists who look at the plea bargaining system see nothing but good in it; but they should think about the principal-agent problem and realize that DAs might pursue their own interest at the expense of crime deterrence.Another problem with public prosecution is that it does not consider the possibility that a victim may value justice at an unusually high level. Just as home-owners often value their homes at above the opportunity cost, crime victims may value justice at above the social deterrence level alone. In a restitution-based system, such a victim could offer his attorney a higher share of the award (e.g., 50% instead of 33%); in exchange the attorney would prosecute a more difficult case. One of the chief problems with the public prosecution system is that it gives much discretionary authority to DAs, then gives them no incentive to respond to unusual cases. The market is more flexible: since there is free bargaining, it is possible to take special circumstances into account.If you recall, I made two criticisms of Posner’s severity-probability mix. First, since criminals are risk-preferrers, probability of conviction would need to be higher to deter criminals. Second, since procedural safeguards often increase in proportion to severity of sentence, increasing the severity may be self-defeating. How does a restitution-based system cope with these problems?First, it handles the problem of risk-aversion by giving incentives to victims to increase the probability of conviction. Since victims get restitution if the criminal who wronged them gets convicted, they have incentives to report crimes, assist the police, and prosecute the case. This might dramatically raise the probability of punishment at a reasonable cost.The second problem is that procedural safeguards often increase in direct proportion to the severity of punishment. Restitution handles this problem in several related ways. First, while the monetary sanctions might be quite high, the prison conditions would be more humane. Non-violent criminals might not even be in prison at all; or perhaps they would merely be confined during their “free time.” And imprisoned criminals would enjoy protection from violence from other prisoners or guards, and receive adequate food, shelter, and medical attention. And if an error were ever uncovered, all of the restitution could be returned; while the costs and injustice of error would remain, they would be easier to rectify. A related innovation would be to modify the rules of evidence. Courts could require proof beyond a reasonable doubt to imprison someone, but only a preponderence of evidence to extract restitution. Under such a system, the harm of error would be no greater than the harm of a mistaken ruling in a civil suit. For all these reasons, a restitution-based system could humanely reduce the level of procedural safeguards, because the injustice and costs of judicial error would be far less. The likely effect would be to markedly raise the probability of conviction. The system that I outlined in this section would be more sophisticated in the real world. Insurance companies might spring up; their job would be to compensate victims, who would in turn transfer their right to restitution to the insurance company. The insurer would then require, as condition of payment, that the victim help the police and participate in the trial. The alienability of the right to restitution existed in restitution-based systems throughout history. As David Friedman explains, “One obvious objection to a system of private enforcement is that the poor (or weak) would be defenseless. The Icelandic system dealt with this problem by giving the victim a property right – the right to be reimbursed by the criminal – and making that right transferable. The victim could turn over his case to someone else, either gratis or in return for a consideration.”79 Other possible innovations include punitive damages for wealthy criminals, or an option for pure punishment without restitution in especially horrific cases.The point of this section has not been to describe in detail every feature of a system of restitution, but merely to explore its chief advantages over the current system. Overall, it looks like the same advantages that lead our society to accept private prosecution of civil offenses might very well lead us to accept private prosecution for crimes. While loopholes exist in restitution-based systems, they don’t look more severe than those of public prosecution. None of the most popular objections to restitution make much sense. Where does this analysis lead us? Well, at the very least victims should have an option for restitution and self-enforcement. If it is feasible for restitution to give adequate incentives to prosecute and punish criminals, why not at least permit it as an alternative to our current system? Perhaps private enforcement just wouldn’t work. In that case, the present system would continue. But there is a good reason to think that restitution could work in some cases. So why not at least give it a chance?4.6 Security guards and private policeUp to this point, I have only discussed non-violent forms of private enforcement. These seem relatively unthreatening and noncontroversial in our society, though they face legal hindrances. The right to use violence seems to be a necessary monopoly of government, so much so that anyone who suggests relaxing it appears to be less than sane. Despite all this intellectual baggage, within our current society private security firms employ almost twice as many guards and private policemen than state and local governments combined! In 1982, there were approximately 1.1 million security employees, compared with 653,579 public police in 1979. Payroll expenditures for private security forces in 1982 were about $21.7 billion, compared with $13.8 billion for public police in 1979.61 Evidently our culture’s aversion to private use of force is far more theoretical than practical.Of course, security guards and private police cannot legally imprison anyone; they mainly deter criminals and stop them in the act. Yet the latter activity is an implicit delegation to private parties of the government’s exclusive right to use coercion. Despite the competitive disadvantage of private enforcers (since their authorized powers are narrower than the government’s), the security industry’s growth rate is startling: from 1964 to 1981, the number of employees in the security industry grew 432.9%.62 Clearly, security guards and police are not merely a theoretical candidate for private supply, but one of the most significant forms of private law enforcement (indeed, all law enforcement) in our society.But aren’t police a pure public good? That is what the textbooks tell us, but it isn’t always true. Private parties find that private benefits and substantial; moreover, these benefits can usually be internalized more readily than economists are wont to suppose. Economics textbooks repeatedly ask us to imagine private supply of patrol cars: since each member of a neighborhood has an incentive to free ride, the textbooks argue, police have to be supplied by the government. This argument ignores the fact that many of the benefits of police are exclusive. For instance, if you want police to respond if your alarm activates, or if you want security guards in your store or business, the benefits are private. Indeed, the free-rider argument outlined above is especially weak since many neighborhoods don’t get patrolled regularly by public police either. There are perhaps moderate positive externalities in the security market, but that hardly makes it a public good. (In fact, section 4.5 implies that externalities disappear if crime victims may exact restitution from convicted criminals.)Imperfect information cause the most significant security externalities. When it is possible to “advertise” one’s security resources, the benefit is private; but when security measures are difficult to publicize, the benefit is largely public. This point may be illustrated with an example. Imagine two possible rules for gun-owners. The first rule permits them to wear their guns openly. The second permits them to have guns, but not to expose them by wearing them in a belt holster. Under the first rule, would-be criminals spot the dangerous targets; and since people want to appear dangerous, gun-owners would wear them openly. Thus, people who do not display a gun would probably not have one; they would be easy targets for criminals. If potentiual victims realize this, they would have an extra incentive to buy one, since armed individuals would get left alone and weaponless individuals would be likely victims. In consequence, the percentage of the public that carried guns would rise. Under the second rule, in contrast, criminals need only consider the average probability that someone carries a weapon. There is no extra incentive to carry a weapon, since gun-bearers and non-gun-bearers would get attacked with equal frequency. Hence, the percentage of the public bearing weapons would fall. What we have here is a typical tragedy of the commons, in which individuals pay the marginal cost but receive the average benefit.What is the point of this example? The point is this: Private enforcement is most effective when one can credibly show that one protects oneself. Stores can hire security guards and put them in public view. Security guard companies may give their customers “armed response” warning signs. Or consider this common business practice: Stores post signs announcing “All shoplifters will be prosecuted.” Since they can credibly commit to prosecution, they reap the deterrent benefit. If I get mugged, prosecution would be a public good, since my decision to press charges would not improve my future safety. But a firm is so permanent and stable that an investment in a reputation for vigorous prosecution can pay off.Consider a final point on externalities and the security industry: different property rights could internalize the externalities of police and security. Imagine, for example, if housing developers, rather than deeding streets and sidewalks over to the local government, kept them in private hands. They might set up a homeowners’ association, and require home-buyers, as part of the contract, to pay for local public goods. This could easily internalize the seemingly public benefits like police patrols. Many benefits of police protection are private; furthermore, most of the externalities are local and might be solved with different property rights. This is approximately what malls do in order to provide their “local public goods.” The mall-developer rents store space to different firms, and then maintains common areas, supplies security guards, organizes patrols, etc. If the firm supplies local public goods for its tenants, it can charge higher rent. That is the mall-owner’s incentive. And the tenants know that they will have higher sales (since customers will be happier in a safe and clean environment) and lower costs (due to lower crime). So long as there is competition for the market, arrangements of this kind should supply an efficient level of local public goods — most notably, police. [footnote Demsetz, "Why Regulate Utilities?"]Many people who ponder the role of profit-making firms in our criminal justice system find it frightening. Isn’t this a dangerous power to leave in private hands? I doubt it. First, while the aggregate figures for private protection are large, this power is extremely decentralized. In 1981, there were 7126 firms in the security industry, with a growth rate of 285.5% for the period 1964-1981. The average number of employees per firm was 46.5.63 There is no plausible danger from such small firms, however impressive their aggregate might. One need not be a true believer in the structure-conduct-performance model to worry about the political risks of a highly concentrated security industry; but economies of scale so limited that we needn’t worry.On top of this, private firms are usually more accountable for their misdeeds than public police. One need not mention Rodney King to realize that police can get away with a lot more than private citizens such as farmers, teachers, engineers, and — security guards. Our legal system gives special preference to accused public police. Their testimony gets greater weight, and the doctrine of sovereign immunity makes it difficult to sue the police. Since the police is a public body, it has little incentive to avoid lawsuits by monitoring its employees. In contrast, private security firms have no special legal status, and it is not especially difficult to sue them. Managers have an incentive to monitor their employees for abuse of power, because the firm may pick up the tab. While most people assume that the profit motive encourages abuse of power and democratic control prevents it, the respective incentives of public and private organizations suggest the opposite.This section has two surprising conclusions. First, private supply of security guards and police is quite feasible; indeed, it may be our society’s most significant deterrent to crime. The common assumption that such services are a public good considers only a fraction of the types of police services. Conventional wisdom also underestimates the market’s ability to supply local public goods with creative use of property and contract. Second, there is no particular reason to be fearful of private supply of police services. Their power is unconcentrated, and they face more incentives to restrain themselves than government police do. People concerned about police brutality should look into the potential of the private sector to supply us with protectors we can trust.4.7 Monopoly on coercion: the source of the limits of arbitrationLegal privatization eventually runs into a seemingly insurmountable obstacle: It cannot resolve conflicts between comp