Our legal system has two quite different sets of rules designed to do the same thing—deter people from injuring others by making it costly to do so. This apparent redundancy was demonstrated in two recent high profile cases. In one, O.J. Simpson was first acquitted of the crime of killing his wife and then convicted of the tort of killing his wife. In another, Michael Jackson was accused of child molestation. The civil case settled out of court, at which point the criminal case was dropped, presumably because the witnesses were no longer willing to testify.
The existence of two sets of legal rules for the same purpose raises at least three interesting questions:
I. Is one approach clearly superior for some sorts of offenses and the other for other sorts, providing a good functional reason for the existence of both, or could we have a functioning legal system in which all offenses were treated as torts or all offenses as crimes, perhaps with some modifications to existing tort or criminal law?
II. Is there a reason to sort offenses between the two systems as we do, a reason, for example, why a burglary is normally treated as a crime rather than a tort and an auto accident as a tort rather than a crime?
III. Tort law and criminal law are each a bundle of legal rules. Is there a reason why the rules are bundled in that particular way, why, for example, the system in which offenses are prosecuted by the victim requires a lower standard of proof for conviction and relies more heavily on monetary punishments than the one in which offenses are prosecuted by the state?
We know it is possible to have a functioning legal system in which all offenses are torts because the Icelanders had such a system, and it functioned for more than three centuries. But that does not tell us whether such a system would work for us, nor does it tell us whether it would be for us (or was for them) superior to a system in which some offenses are torts and others crimes. So it is worth examining the theoretical question of whether there are good reasons why some offenses should be publicly prosecuted, as in current criminal law. A variety of such reasons have been proposed. They include:
1. The victim of an offense may not have sufficient resources to prosecute it.
This problem can be dealt with by making tort claims transferable, as was done in saga period Iceland. A victim with inadequate resources gives or sells his claim to someone better able to prosecute it. He might, as a result, end up with no compensation for his loss—but then, under criminal law at present, victims receive no compensation either.
2. Some offenses cause diffuse injury, so nobody has an adequate incentive to prosecute them.
This is dealt with under current law by class actions. It could be better dealt with by making claims for torts that had not yet been litigated, including ones that had not yet occurred, transferable. Middlemen would buy bundles of small claims from potential plaintiffs, then rebundle them for sale to prosecution firms, with each firm buying claims associated with the particular offenses it is litigating.
3. Some offenses result in a diffuse injury difficult to observe, along with an observable injury to a single victim. The standard example is a crime that both injures the victim and imposes fear on potential victims.
This problem is less severe under tort law, where the victim/prosecutor, if successful, gets reimbursed for his injury than under a criminal system, where the victim gets nothing. But even under tort law, victims are not fully reimbursed ex ante, since there is some chance they will fail to collect damages, and, in most circumstances, they must pay their own legal costs. One possible solution, discussed back in chapter 14 in the context of punitive damages, is to increase the amount of damages to allow for such problems.
4. If an offender is judgement-proof, there is no incentive for the victim to prosecute him, so prosecution must be by the state.
There are at least three possible responses to this problem.
Whether defendants are judgement-proof is in part a function of the legal rules on collecting judgements. If convicted defendants who were unable to pay money damages could be sold into slavery or dismembered for organ transplants, fewer defendants would be judgement-proof, not only because some would have sufficient value as slaves or spare parts but because the existence of such unattractive outcomes would provide offenders an incentive to make sure they were able to pay damages in money instead. Some of the risks of such a system were discussed in chapter 15.
The state could pay the fines of judgement-proof offenders to the victim/prosecutors, thus providing them an incentive to prosecute, and impose criminal punishments, thus deterring offenders. Such a bounty system, analogous to a voucher system for schooling (or the GI Bill, which was a voucher system for higher education), combines private prosecution with public funding.
The victim of a tort may commit himself to prosecute even judgement-proof defendants in order to deter offenses against him, provided that the legal system provides non-pecuniary punishments for convicted defendants unable to pay damages. We saw an example of this pattern in the prosecution associations of 18th century England.
5. It is impossible to construct legal rules that give victims the optimal incentive to prosecute. Hence the best possible system of private enforcement is inferior to an ideal system of public enforcement, although not necessarily to an actual system of public enforcement.
[article link]
This, the most sophisticated and interesting of the arguments against private enforcement, is due to Landes and Posner. As we saw back in chapter 15, for any offense there exists some optimal amount and probability of punishment. But once we set the amount of punishment in a system with private prosecution, we have no way of controlling the probability. That will be determined by the profit-maximizing behavior of the victims—how much they find it worth spending on identifying and convicting offenders in order to collect damages from them. There is thus no way save luck of getting to the optimal combination of probability and punishment.
To put the argument more mathematically, we are trying to separately get two variables, probability and punishment, to their optimal values. But we have only one control variable to do it with—the level of punishment. An ideal public enforcement system, on the other hand, could set probability and punishment independently, making an ideal public enforcement system superior to even an ideal private one, at least in that respect.
The solution to this problem is for the legal system to set, not the punishment if convicted, but the expected punishment: fine paid times probability of conviction. To make that possible, we must first add a set of middlemen to our privately enforced legal system, prosecution firms which buy claims from victims, prosecute them, and collect the fine, if any. As in chapter 15, we will limit ourselves for simplicity to a single crime, say mugging.
Think of any punishment as consisting of a fine paid and a fine collected. The fine paid measures how much worse off the convicted criminal is as a result of having the punishment imposed on him. The fine collected measures how much better off other people are as a direct result of having the punishment imposed on him. The difference between fine paid and fine collected is the punishment cost. Consider some examples:
1. A cash fine, with no administrative costs of collecting it. Fine paid equals fine collected, punishment cost equals zero.
2. Execution. Fine paid equals the value of one life. Fine collected equals zero—a little less if execution has positive costs.
3. Imprisonment. Fine paid equals the value to the criminal of not being in prison. Fine collected is negative—the cost to the rest of us of running the prison.
Suppose a prosecutor has bought a thousand offenses, for each of which he is permitted (and required) by the court to collect an expected punishment, a fine paid, of a thousand dollars. If he catches and convicts every offender—an expensive proposition—he charges each a thousand dollar fine, collecting a total revenue of a million dollars. If he catches and convicts only a hundred offenders, he is legally entitled to impose a ten thousand dollar fine on each but, since many cannot pay that much, he ends up having to imprison some instead—and pay for the prisons. In deciding how many to catch and convict he must trade off the lower apprehension cost of catching fewer criminals against the higher punishment cost of punishing those he catches more severely.
Generalizing the argument we can see (the formal demonstration is at the end of the chapter) that it is in the private interest of the prosecutor to choose the optimal punishment/probability combination, the combination that minimizes the sum of apprehension cost and punishment cost. He wants to minimize the sum of the two costs because he is paying both: apprehension cost directly, since it is up to the prosecutor to apprehend and convict offenders, and punishment cost indirectly, because the higher the punishment cost, the less of what offenders lose goes to him and the more he has to pay to punish them.
So all the court has to do is to set the correct expected value for fine paid. Profit maximization by the enforcer will produce that expected value in the optimal way, solving the problem raised by Landes and Posner.
There is a problem with this approach to expanding tort law to provide private prosecution of offenses that are now crimes. Consider an offense for which the average return from catching and punishing an offender is negative; it costs more to catch and convict him than the fine paid. In order for prosecutors to stay in business, victims must pay enforcers to take over their claims and prosecute them. The victims are still selling their claims but at a negative price.
A victim who sells an offense for a negative price is paying for the enforcer's agreement to impose a specific expected punishment on the offender. The reason to do so is deterrence. The victim wants potential offenders to know that if they commit an offense against him they risk punishment.
This works for offenses for which it is possible to make deterrence a private good, such as burglary. A potential victim pays an enforcer to agree to prosecute his claim for any offenses committed against him. The victim passes on the relevant information to potential burglars by posting a notice on his door announcing that that particular enforcer has purchased his claims and will enforce them. The victim has sold his claims, in advance, at a negative price, and gotten deterrence in exchange.
This is not a purely imaginary arrangement. Even in our present system, where the enforcement of criminal law is nominally entirely public, there are firms that sell similar services. In 18th century England, potential crime victims joined prosecution associations, paying a price to buy deterrence. A number of the Icelandic sagas describe situations where a weak victim suffers injury from a much more powerful offender. The victim, unable to successfully pursue legal remedies, transfers his claim to a more powerful party. In some cases, the victim also offers additional inducements to the enforcer, presumably because the amount of damages that could be collected would not, by itself, be a sufficient incentive to undertake the costs and risks of collection.
So private enforcement, generalized tort law, could work for all offenses that sell at a positive price, offenses for which the amount collected from convicted offenders is, on average, more than the cost of catching and convicting them. It also could work for offenses that sell at a negative price but for which deterrence can be made a private good. It does not work for offenses that sell at a negative price for which deterrence cannot be made a private good because the offender does not know enough about the victim to be deterred—what I earlier described as anonymous-victim offenses. These were the sort of offenses—highway robbery, for example—that provided a special problem for the system of private enforcement of criminal law in 18th Century England.
How serious the problem is depends in part on the technology of committing offenses and apprehending and punishing offenders and in part on features of the legal system that affect the cost of apprehension and punishment. A legal system that permitted penal slavery, or one that allowed the sale of organs from executed felons, or one set in a society where individual reputation was important and stigma thus provided a powerful and efficient punishment, would have lower punishment costs than one without those features.
In modern legal systems some offenses are privately prosecuted as torts, others publicly prosecuted as crimes. This raises an obvious question: does the tort/crime distinction in modern law correspond reasonably closely to the distinction between those offenses that are and are not dealt with adequately by private enforcement?
Landes and Posner argue that it does. They argue that torts are, generally speaking, offenses detected with probability near one, hence for which the problem of getting the proper probability/punishment combination disappears.
But even if detecting many torts is cheap, prosecuting them is not. The more the victim spends, the more likely he is to win his case. The higher the damage payment he will receive if he wins, the greater the incentive to spend money in order to win it. The damage payment set by the law is both the punishment for committing a tort and the reward for successfully prosecuting one. There is no reason to suppose that the same damage payment produces the optimal level of both. The problem Landes and Posner raised in order to show why it is inefficient to apply tort law to crimes also implies that it is inefficient to apply tort law to torts!
An alternative defense of the current crime/tort division is the argument that criminals are more likely to be judgement-proof against the optimal punishment, unable to pay the corresponding fine, than are tortfeasors. Translated into the language I am using here, that suggests that tort offenses are more likely to sell at a positive price, more likely to produce more in fines than it costs to prosecute them, than are criminal offenses.
There are two reasons why we might expect criminals to be more often judgement-proof than tortfeasors. One is that (some) crimes, such as burglary, are hard to detect, so the optimal probability of conviction is low, so the fine must be high in order to give an adequate expected punishment. The other is that, if offenses are hard to detect, producing even a small probability of conviction may be costly. Both of these are related to the fact that crimes are intentional, so criminals are likely to spend resources concealing them.
One problem with this comparison is that it is based on a biased sample of torts: the ones that get litigated. Since tort law is privately enforced, anonymous-victim torts with negative price don't get litigated. The offenses we observe being prosecuted in the tort system tend to be the ones with positive value.
A second problem is that while the offenses we now classify as crimes are more likely than torts to be negative price offenses, they are less likely than torts to be anonymous-victim offenses. A burglar deciding which house to burgle can choose to avoid houses with notices saying that their owners have paid in advance to have burglars prosecuted. A driver cannot readily adjust his level of care to take account of which other cars have notices on them saying that their owners have paid in advance to have drivers who run into them prosecuted. So intentional offenses, typically crimes rather than torts, are less likely to be anonymous-victim offenses, hence more likely to be offenses for which deterrence can be made a private good.
Another important issue, touched on in chapter 15, is the problem of deliberately fraudulent claims. Under any system where offenses sell at a positive price, there is an incentive to manufacture them, to frame potential defendants. Hence any legal changes that shift the system toward more efficient punishment by, for example, making it easier to collect fines, also risk encouraging prosecutorial fraud.
The English criminal system encountered this problem in the mid-eighteenth century. Because of concern that incentives for private prosecution were too low, the Crown established substantial rewards for successful prosecution of certain offenses such as highway robbery. The result was a series of scandals in which it appeared that the convicted offender had either been entrapped into committing the offense in order that he could then be betrayed for the reward or had been framed for an offense that had not occurred. Analogous modern cases arise under both civil forfeiture and punitive damages.
This problem provides a possible explanation for one of the more puzzling features of tort law, the absence of probability multipliers. A successful litigant is entitled to have his injury made good, to receive a damage payment equal to the damage done. But if victims of torts are successful litigants with probability less than one, as is surely the case, the result is an expected punishment predictably less than the damage done. The obvious solution is to add in a probability multiplier, to scale up the punishment of tortfeasors who are successfully sued to compensate for the failure to punish those who are not.
One explanation for the lack of such probability multipliers is that they would be an invitation to fraud. Under current law, someone is never better off as a result of being a victim of a tort; even if he successfully litigates it, all he gets is an amount sufficient to make up for the damage he has suffered. With a substantial probability multiplier, someone who could set up bogus torts, with witnesses suitably placed and primed to guarantee successful litigation, might do very well for himself. He faces much the same incentives as the homeowner whose insurance company has carelessly permitted him to insure his house for twice its value.
This suggests a second category of offenses for which private enforcement may work poorly: positive value offenses for which it is relatively easy to manufacture false positive verdicts. It is not clear, however, that public enforcement avoids the problem. Public enforcers must have some incentive to get convictions if they are to do their job, hence some incentive to get false convictions. Furthermore, public enforcers frequently use the threat of conviction to extract services in the form of inf ormation or testimony. The threat need not involve a real offense, as a number of examples, including a recent police scandal in Philadelphia, have demonstrated. While a legal system that limits enforcers to inefficient punishments makes prosecutorial rent seeking more difficult, it does not eliminate it, since an inefficient punishment can be converted into an efficient one by an out of court settlement.
This suggests one important way in which the incentives of an enforcer whose objective is private deterrence are superior to those of an enforcer whose incentive is the profit from collecting an efficient punishment or the political or bureaucratic gains from maintaining a high conviction rate. A policy of arresting the first indigent who passes a crime scene, convicting him in front of a hanging judge, executing him, and auctioning off his organs, may be an effective way of collecting income through prosecution, private or public. But since your chance of being punished under such a policy does not depend on whether or not you are guilty, it is a very ineffective way of deterring crime. Private deterrence is the one incentive that makes it in the direct private interest of the prosecutor to convict the right person.
[text link]
I have written, so far, as if the defining difference between criminal law and tort law was the distinction between public and private prosecution. While that approach seems natural to an economist, other differences might seem more central to scholars approaching the question in other ways. Thus, for example, it is sometimes said that the essential difference is that a crime is seen as a moral fault, a source of stigma, and a tort is not; accusing someone of being a criminal is more of an insult than accusing him of being a tortfeasor. Crimes are sometimes thought of as distinguished by certain sorts of punishment, notably imprisonment and execution. And criminal law is distinguished from tort law, at least in modern Anglo-American legal systems, in a variety of other respects. Crimes have a high standard of proof, require intent, are guaranteed a jury trial, have punishments often much higher than the damage done, pay fines to the state rather than the victim, … .
This suggests an obvious question: Are there good reasons why these features go together? Are there good reasons why we have one category called "tort" with one set of rules, another called "crime" with another set?
The answer, as I will attempt to show below, is that there are reasons, although perhaps not always compelling ones. I begin with a list of features in tabular form:
Characteristic Who Controls Prosecution Who Collects Punishments Form of Punishment Standard of Proof Probability Multiplier? Right to Jury Trial? Desired level of offenses Requires intent Stigma to Conviction
The tort and crime columns in the table represent stylized versions of the two systems. Fines are used in criminal law, but they tend to be for offenses, such as speeding or illegal parking, which have some of the other characteristics of torts. A parking ticket does not convey stigma and, in practice, does not require proof to the criminal standard. Punitive damages can be interpreted as a form of probability multiplier, and have been by some scholars, but they are also connected with stigma. For purposes of simplicity, the table ignores such intermediate cases.
The table also ignores features that have dropped out of modern law. Eighteenth Century English criminal law combined private prosecution with criminal punishments. Early medieval English law went even further; the appeal of felony was, like a modern tort action, entirely private—the private prosecutor could drop charges and the crown could not pardon a convicted offender—but a successful prosecution resulted in criminal penalties. These systems provide important evidence of alternative ways in which legal rules might be bundled, but I am concerned here with a narrower range of options.
Looking at the table, we observe that under both systems the same actor controls prosecution and collects punishments. There are two obvious reasons for this. One is that collecting a damage payment provides an incentive to prosecute. The other is that the party who controls the prosecution can also drop the prosecution or, if that is forbidden, prosecute badly. Any system in which one party controls prosecution and another collects the punishment raises the possibility that the former will divert the fine to himself via an out of court settlement.
There is also a disadvantage to the combination: the opportunity it provides for fraudulent prosecution. If the claim and the right to prosecute are held separately, fraudulent prosecution requires a transaction between the owner of the claim and the prosecutor.
The victim of an offense has an incentive to prosecute in order to deter future offenses against him. He is also, in many cases, the chief witness. By making him the owner of the claim, we unite in one person two incentives to prosecute plus an important input to successful prosecution, eliminating the transaction costs of getting separate people playing those roles to work together. Here again, there is also a possible disadvantage: the greater the victim's incentive to get a conviction, the less trust can be put in his testimony. One of the reasons for the abandonment of the system of rewards in 18th century England was that juries, knowing that witnesses might share in the rewards, rationally distrusted their testimony. The same problem arises today when juries know that witnesses for the prosecution in criminal trials are paid informants or have agreed to testify in return for a reduced sentence.
Having the claim belong to the victim has both advantages and disadvantages so far as its effect on the victim's incentive to prevent the offense, for reasons that will be discussed later in the chapter.
We next notice that the tort system is associated with relatively efficient punishments, the criminal with relatively inefficient. That makes sense in terms of the previous analysis, since offenses that require inefficient punishments are likely to be negative price offenses and so more difficult to prosecute privately. Of course, the fact that an offense is currently subject to inefficient punishment does not guarantee that it must be.
The inefficiency of criminal punishments provides, as first pointed out back in chapter 1, an explanation for the higher standard of proof required to impose them. The lower the cost of punishment, the lower the (net) cost of imposing it incorrectly. So it makes sense to combine less efficient punishments with a higher standard of proof.
Or perhaps not. One problem with efficient punishments is that, by making successful prosecution profitable, they create an incentive for fraudulent prosecutions. One way of controlling that is a high standard of proof. An alternative way is by making fraudulent prosecution—in the limiting case, any unsuccessful prosecution—itself tortious. A weaker version of that is to require the losing party in a tort suit to pay the other side's legal expenses as is currently the rule in England, although not in the United States
There are three plausible explanations for the absence of probability multipliers in tort law. One I have already mentioned: Probability multipliers make it profitable to be the victim of an offense, provided your probability of successful prosecution is sufficiently high relative to the average probability from which the multiplier is calculated. In a privately prosecuted system, that provides an incentive for fraudulent prosecution.
A second explanation is that, because torts are not intentional, they are usually not concealed, and the probability of apprehension and conviction is thus arguably higher than for crimes. If so, the need for a probability multiplier is less. It might be argued in the opposite direction, however, that where the probability is low, it is the privately enforced system that particularly requires the multiplier. In both systems the multiplier is desirable in order to provide adequate deterrence. With private prosecution, it may also be desirable in order to give an adequate incentive to the prosecutor. That relation is made explicit in the efficient system for private prosecution that I described in Part I, which differs in important ways from tort law as it actually exists.
A third explanation links the lack of a multiplier to the importance of efficient punishments. The higher the punishment, the less likely it is that the offender can pay it as a fine. Hence one problem with probability multipliers is that they may push the punishment too high, increasing the fine paid but decreasing the fine received and thus reducing the incentive to prosecute.
The fact that Anglo-American law guarantees the defendant the opportunity for a jury trial for criminal cases but not always (except in the United States) for civil cases is puzzling. One possible explanation is that it serves the function of making fraudulent prosecution more difficult. Under a criminal system, the prosecutor, judge, and claimant are all, in some sense, the same party; despite attempts to guarantee judicial independence, judges are ultimately employees of the state. Hence in a corrupt criminal system, a single party is sufficient to indict and convict. Introducing a jury changes that situation. Of course, there is still the risk that, under a corrupt civil system without juries, judges could conspire with fraudulent claimants. That risk is particularly serious when agents of the state play the role of civil plaintiff, as they do in forfeiture cases.
The category "desired level of offenses" requires some explanation. There is a sense in which the desired number of murders is positive—given the cost of preventing them. And the desired number of auto collisions would be zero—if we could avoid collisions at no cost.
What I mean here by the desired level is the level that would be efficient if the cost to the legal system of achieving that level were zero. If we could costlessly deter all murders we would, despite the inconvenience to those of us with rich uncles. If we could costlessly deter all traffic accidents we would not, because the ways in which people avoid having traffic accidents when the expected penalty is high, for example by not driving, sometimes cost more than the reduction in accidents is worth.
As this example suggests, part of the reason the optimal level of crime is typically zero is that crimes are intentional; perpetrators benefit by the occurrence of successful crimes and so spend resources to make them occur. This makes it likely that crimes will be inefficient for two reasons. First, many crimes are transfers, and if we add to a transfer the cost of bringing it about, the result is a net loss. Second, since the auto thief typically chooses whom he will steal a car from, he could choose to buy it instead. Where market transactions are practical they are typically less expensive than coerced transactions, since in the latter case the other party is spending resources trying to prevent the transaction.
This brings us to the issue of intent. Given that intentional offenses are more likely to be inefficient, and thus worth deterring if doing so were costless, why are they also worth prosecuting publicly rather than privately?
Two answers have already been suggested in earlier discussions. One is that intentional offenses are more likely to be concealed, hence harder to detect, hence more likely to be negative price offenses. The other is that intentional acts are more likely to have market substitutes than accidental acts, making it less likely that they are efficient, and may be more deterrable; if so it makes sense to use a property rule, enforced by criminal law, rather than a liability rule enforced by tort law, to deal with them. An argument in the opposite direction is that intentional offenses are less likely to be anonymous victim offenses, hence it is easier to deter them by making deterrence a private good.
We come finally to the question of stigma, the one punishment with negative cost. It is an unusual form of punishment in another respect as well: the direct benefit from imposing it is a public, not a private, good. A successful prosecution produces information about the defendant that is valuable to those who will deal with him in the future, but that value is not available to help pay the cost of prosecution. Stigma does, however, help solve the problem of deterring offenses that are costly to punish. Hence it is not surprising to see its use combined with the use of inefficient punishments—i.e. with the criminal law.
Suppose we used stigma to punish offenses that we wished to ration but not to eliminate, such as traffic accidents. There would be a risk of overdeterrence, especially since the amount of stigma, unlike the size of a fine or the length of a prison sentence, is not something that the criminal justice system has direct control over. That is not a problem for murder. That suggests a second reason why it makes sense for stigma to be associated with criminal rather than civil offenses. A third reason is that, since stigma imposes costs on the defendant and provides its benefits to people other than the prosecutor, a private prosecutor may be willing, for a suitable consideration, to agree to a secret out of court settlement, thus keeping the information that would have been generated by a guilty verdict from those who could have used it to guide their future dealings with the defendant.
Legal rules affect behavior on many margins. They affect incentives to commit offenses, incentives to prosecute them, and incentives to prevent them. For some offenses, tort law may provide better incentives than criminal law on one of those margins and worse on another. If so, the question of what ought to be a tort and what ought to be a crime will be ambiguous, at least until we develop a theory good enough to predict not only the sign but also the size of such effects.
In this section of the chapter, I take one of the margins and argue that, judged by the incentives provided by the alternative legal regimes on that margin, our rules are backwards. Things we treat as crimes, such as burglary, ought to be torts; things we treat as torts, such as auto accidents, ought to be crimes. In doing so, I hope to demonstrate one reason why figuring out whether our present allocation of offenses between the two systems is efficient is a hard problem.
The incentive I have chosen to examine is the incentive for potential victims to prevent offenses. From that perspective, there is a simple and striking difference between tort law and criminal law. The victim of a tort suffers an injury but also obtains an asset, a claim for damages against the offender. The victim of a crime suffers an injury but obtains no corresponding asset since the fine for the offense, if any, goes to the state. So one effect of treating an offense as a tort instead of a crime is to reduce the net damage suffered by the victim and thus his incentive to prevent the offense.
Reducing the incentive to prevent the offense is not necessarily a bad thing; here as elsewhere, what we want is the right incentive, neither too much nor too little. The ideal legal system, seen from this standpoint, is one in which the potential victim receives the net social benefit produced by his defensive precautions and thus has an incentive to take them if and only if they are worth the cost. The question for any particular class of offense is whether the reduction in incentive provided by tort law moves the potential victim toward or away from the efficient level of precaution.
To make the argument easier, we start with a world in which catching and punishing offenders is costless. All offenders are detected and none are judgement proof. Within this simplified framework, what is the right incentive for the victim to prevent the offense and what system gives it?
Start with automobile accidents. For simplicity, consider accidents in which a car runs into a pedestrian and only the pedestrian is injured. We may attempt to deter such accidents either with a tort rule, in which the driver of the automobile is liable to the victim for the damage done, or with a criminal rule, in which the driver pays a fine but the victim does not receive it.
From the standpoint of the driver, the two rules (with the same penalty) generate the same incentive. Any reduction in the probability of an accident due to the driver's production reduces his liability by an amount equal to the savings in net social cost. What about the pedestrian?
Under the tort rule, the pedestrian suffers no cost from being run into, since his damages are fully compensated by the driver. It follows that the pedestrian has no incentive to take precautions. Under the criminal rule, on the other hand, the pedestrian pays the full cost of the accident. It follows that he has the efficient incentive to take precautions to prevent it. It follows that, from the standpoint of the incentive for victim precautions, such offenses should be crimes, not torts.
Generalizing beyond the simplified example in which only one party suffers a loss, the efficient rule is that each party bears his own loss and pays a fine equal to the loss that the accident imposes on the other party. Thus each bears the full cost of the accident and has an incentive to take any cost-justified precautions to reduce its probability.
Consider next burglary, still in a world of costless enforcement. The legal system sets the penalty for burglary equal to the damage done, the value of what is stolen. Any burglar willing to pay that price commits an offense—and should. The system generates only efficient burglaries, those for which the gain to the offender is greater than the loss to the victim. The textbook example is Posner's hunter, lost and starving, who comes across a locked cabin in the woods, breaks in, feeds himself, and telephones for help.
In this world, a homeowner who puts a lock on his door is wasting his own money and the burglar's time. He is better off leaving the door open and (costlessly) collecting the value of whatever goes out. The optimal level of precaution to prevent burglary is the same as the optimal level of precaution for a supermarket to take in preventing its customers from buying its vegetables: zero.
Under a tort rule, that is the level of precaution the victim will take. He knows he will be fully reimbursed for whatever he loses, so has no incentive to spend resources preventing burglary. Under a criminal rule, on the other hand, the criminal takes from the victim but repays the state, so the victim has an incentive to prevent the burglary. Hence, at least in the simplified world of costless enforcement and considering only the incentive of the victim to defend himself, burglary ought to be a tort. Readers interested in a formal demonstration of the same result in a slightly more realistic world will find it on the web page.
[link to math]
Of course, there is no guarantee that we would reach the same conclusion if we analyzed the question in terms of incentives on other margins—for example, the incentive of the victim (under tort law) or the public prosecutor (under criminal law) to prosecute the offense. If it turns out that tort law gives the right incentives for some decisions by some parties and criminal law gives the right incentives for other decisions by other parties, we are left with no clear answer in theory to the question of which offenses should be under which legal system—and the difficult practical problem of trying to figure out which of the two imperfect solutions we should prefer.
There is a set of offenses which are especially difficult for private enforcement to deal with: negative price offenses with anonymous victims. Such offenses provide a plausible argument against a pure tort system. There is a further set of offenses which are difficult to deal with either privately or publicly: offenses where the probability of apprehension is low, requiring a probability multiplier in order to get an adequate expected punishment, where efficient punishments are possible, making successful prosecution profitable, and where it is difficult to detect fraudulent prosecutions.
The current sorting of offenses between the categories of crime and tort has at most a modest relation to what that analysis suggests would be an efficient division. In examining the bundling of legal rules into the forms we call "tort" and "crime," on the other hand, we can find plausible, but not always compelling, efficiency explanations for most of its features.
A Postcript For the Mathematically Inclined
I am a prosecutor who has purchased from the victims the right to prosecute 1000 offenses of a particular sort. The court has set an expected punishment of $1,000. That means that I may catch all offenders and impose on each a fine paid of $1,000, catch 100 and impose on each a fine paid of $10,000, or catch 10 and impose on each a fine paid of $100,000.
In choosing among these alternatives, I wish to maximize the total of fines collected minus enforcement costs. Let Fp be the fine paid by each convicted offender, Fc the fine collected, making Fp-Fc the punishment cost. N is the number of offenders convicted and punished. N/1000=p, the probability of conviction and punishment. E(p) is the per offender cost of catching an offender with probability p. Po is the price the enforcer has paid each victim for the right to enforce his claim and collect the resulting fine.
The enforcer receives the fine collected, pays for enforcement, and pays to buy offenses, so he faces a profit function:
¹=NFc-NE(p)-1000Po.
The court has set pFp=$1,000, hence NFp=$1,000,000
Hence we have:
¹=$1,000,000-1000Po] -[N(Fp-Fc)+NE(p)].
Profit = [total fine paid-price paid to buy 1000 offenses]-[the sum of total punishment cost (difference between fine paid and fine collected) and total enforcement cost]
The enforcer must decide how hard to try to catch the offenders, as measured by p, the probability with which he catches them. That decision will determine the fine paid, which he will collect in the most efficient form possible. The first bracket in Equation 2 is independent of p, so he maximizes his profit by minimizing:
N(Fp-Fc)+NE(p)
Fp-Fc is the punishment cost for one offense punished, N(Fp-Fc) is therefore the total punishment cost. NE(p) is the total enforcement cost. So the enforcer is choosing the level of p (and the punishment which, combined with that p, yields the set value of expected punishment) that minimizes the sum of punishment and enforcement cost—which is to say, the least-cost combination of probability and punishment. So all the court has to do is to set the efficient value of expected punishment, fine paid times probability of punishment, and the enforcement agency, in the process of maximizing its profit, will produce that value with the optimal combination of probability and punishment.