Law & Economics Lecture 8: Tort

I. The Main Questions of Tort Law Law & Economics Lecture 8: Tort Tort is a branch of civil law that deals with a variety of legal wrongs that involve harm to individuals; examples include product liability, malpractice, personal injury, and wrongful death. Many of our previous discussions have actually been about tort, though we haven't always called it that. Tort is a very broad area, and it is sometimes hard to distinguish it from other areas; DDF defines a tort as simply "a wrong that is privately prosecuted," and he notes that while contract law "enforced obligations voluntarily agreed to," tort law deals with obligations imposed by law. There are four questions we must ask in analyzing the law of tort: What does it mean to be harmed? That is, what do we really mean by the word "harm" in a legal context? If I dump my girlfriend, clearly she is worse off, but have I harmed her in the legally relevant sense of that term? If I go into competition with you, is that a legal harm? What does it mean to say you caused a harm? That is, what do we really mean by the word "cause"? If I dump my girlfriend, and she goes and sleeps with some sleazy guy out of grief, and as a result she catches an STD, have I caused her to be harmed? After all, she would not have caught that STD if I hadn't dumped her. But is this how the law uses the word "cause"? When should you be liable for a harm that you cause? Just because you caused a harm doesn't mean you should necessarily be held liable -- that is, punished in some fashion -- for it. For example, what if I caused a harm to you, but you shouldn't have been there to be harmed? That is, what if you and I both caused the harm to you -- should I be punished? If you are liable, what remedy should the court apply? Should the court force you to take a particular action? Or should it just force you to pay money? And if it forces you to pay money, how much should you pay? We will take each of these questions in turn. II. What Is Harm? As we've observed, there are some "harms," in the vernacular, that the courts do not punish, such as dumping your girlfriend or going into competition with someone else. What does economic efficiency have to tell us about what should be a harm? Remember that, from an efficiency perspective, transfers are irrelevant unless they lead to wasteful rent-seeking. What matters is the overall effect on wealth. Now, what is the

effect of going into competition with existing suppliers of a service? Let's suppose that this is an imperfectly competitive industry. Then if I go into the business, I'll cause the price to fall and the quantity to increase. Now, this is bad for the existing businesses, because they get a lower price and possibly lose some of their customers to me. But where is the money that they lose going? Some of it goes to me (because I get some of their customers), and some of it goes to the customers (because they get a lower price). Either way, that's just a transfer. In addition, because of the lower price, more customers get served, and they are clearly better off. So wealth is actually increased by my entry into the industry. If I could be sued for doing that, I wouldn't have an incentive to increase wealth in that way. [Show picture of a monopolist, with transfer and DWL marked. Show how a lower price from more competition would reduce the size of the transfer (transferring it back to customer) and also reduce the DWL.] There are some other "harms" that are not considered legally actionable. DDF gives the example of his wearing ugly shirts; I give the example of dumping my girlfriend. Why are these not dealt with by the court system? Several reasons. First, in the case of ugly shirts, the harm is most likely very small. The cost of dealing with "fashion police" cases in court would probably far exceed the benefits. (Besides, isn't clothing a kind of signaling mechanism?) Second, the benefits are extremely hard to measure. This is especially true in the case of dumping your girlfriend. How can the court measure the benefit to me and the cost to her? Do we really want the courts to be making that kind of decision? (Plus, we'd also have to consider the benefits to other young ladies who now stand a chance with me.) Third, some of these kind of harms are easy enough to deal with through private transactions; this is the Coase Theorem approach. At least when we're talking about an ongoing harm, it may be possible to negotiate a better agreement (e.g., your spouse rewards you for wearing a nicer shirt). There is yet another reason why we might not want to treat some harms as torts: because we'd rather treat them with a property rule instead. Recall that property rules are generally best when we want to protect subjective valuations and encourage private bargaining. III. Causation and Foreseeability In some sense, everything that happens has been caused by events that happened far, far in the past. If it weren't for the Big Bang, no torts would ever have taken place. If my parents hadn't given birth to me, I wouldn't have wrecked your car. If I hadn t stopped you for directions on the street, you would have left before the piano fell where you were standing. Maybe it seems ridiculous even to consider that kind of cause in a legal context. But it's important for us to ask why, because there are many borderline cases. For instance, in what sense are the tobacco companies responsible for the government's high Medicare and Medicaid bills? Are gun manufacturers responsible for the crimes committed with their weapons? Are parents responsible for the crimes committed by their children (e.g.,

the parents of the Columbine killers)? These are live issues, and they relate to the issue of legal causation. Coincidental Causation. Consider DDF's example of a trolley that was hit by a falling tree, injuring some passengers (Berry v. Bourough of Sugar Notch, 1899). It turns out that the trolley must have been speeding, and if it had gone the speed limit, it would not have been hit by the tree. So, did the driver of the trolley "cause" the accident? In a physical sense, the clear answer is "yes." The driver's actions were part of a whole series of events that led up to the outcome. Yet the court held the driver did not cause the accident. In DDF's words, "Logically, the court was wrong; economically, it was right." From an economic perspective, the question is one of incentives. Does punishing the driver in this case provide any kind of important incentive? Yes, it might cause him not to speed. But what if the tree had fallen just a few seconds later? Then in that case, going the speed limit would have led to an accident, and speeding would have prevented the accident. Clearly, the speed limit is supposed to limit the risks created by speeding -- but avoiding falling trees is not one of those risks. More broadly, the point is that punishing the driver for damage caused by falling trees is not likely to change his behavior in an economically efficient way. However fast or slow he drives, the probability of getting hit by a tree is the same from his perspective. Punishing him for tree damage will not make tree damage any less common. All it will do is transfer wealth from the driver to the passengers, and incur administration and litigation costs in the process. Foreseeability. All of this relates to the issue of foreseeability. For a punishment to have any effect on incentives, it must be the case that the agent can foresee the incident occurring, and also that his actions can have a predictable and systematic effect on the likelihood of the incident occurring. Neither is satisfied in the trolley example. But there are many less clear cases. Consider the case of Palsgraf v. Long Island Railroad (1928). In this case, a man was running to catch up with a train as it pulled out of the station. He was carrying a brown cardboard box. Two railroad employees assisted the man in getting onto the moving train, and this led to the box falling out of his hands. Turns out the box contained fireworks that exploded on hitting the tracks, causing vibrations throughout the station. As a result, some scales at the other end of the platform tipped over and injured a woman named Palsgraf, who sued the railroad for her injuries. The question is, was this outcome foreseeable? In one sense, it was not. The railroad employees could not have foreseen that the man they helped would be carrying fireworks, that helping him would cause the fireworks to explode, and that the explosion would cause someone to be injured by falling scales. As Judge Cardozo argued in the majority decision, the railroad owes a duty of care only to specific persons in the foreseeable zone of danger. But it another sense, the harm was foreseeable. The railroad employees could not foresee that their act would lead to this

particular harm, but they could foresee that their act could lead to a variety of sorts of harm. Helping someone board a moving train is an inherently dangerous activity. As Judge Andrews argued in the dissent, the railroad has a general duty of care to society at large. If you define an outcome narrowly enough, you can always say it was not foreseeable. For instance, if I drive recklessly, can I predict that my actions will (or might) cause an accident that causes Mr. Gary Smith to suffer a severe fracture of his left pelvis at 3:15 p.m. on Saturday, October 9, 2002, while it happens to be raining and a dog is crossing the street? The likelihood of that specific event is approximately zero. But the probability of an event in that general class (recklessness leading to medical damage) is greater than zero. From an economic standpoint, the issue of foreseeability is important because your behavior will be affected by a risk only if you can foresee the risk. Suppose that driving without a side rearview mirror will increase the likelihood of a particular type of accident (side-swiping) by 2%. If you're aware of that fact, and you're aware you'll be held responsible for the results, then you may be induced to fix your rearview mirror. But if you're not aware that your activity increases the risk of this type of accident, then holding you responsible for the results will not change your behavior. It will simply transfer wealth from you to the victim, without increasing wealth in any way. So from an efficiency point of view, questions of foreseeability should be addressed in the following manner: Will holding people liable for damage of this type (i.e., damage resulting from accidents like the one in the present case) alter the behavior of other potential litigants? And if so, will the change in behavior reduce expected damages by more than the cost of the change? If the answer to both questions is yes, then the outcome is foreseeable in the economically relevant sense of the term. IV. Probabilistic Causation Almost all causation is probabilistic in some sense. My reckless driving does not lead to an accident with certainty, but it increases the probability of an accident. But that is not the sense in which the word probabilistic is used here. Here, I'm using the word probabilistic to refer to an accident or harm that has already occurred, but whose cause we can only determine probabilistically. To use DDF's example, consider a nuclear reactor's error that leads to the leaking of radioactive gas. This leak leads to a 10% increase in the rate of cancer in the area (say, from 10 cases a year to 11). But you cannot tell whether any particular cancer victim s condition was caused by the reactor leak. For any given cancer victim, there is a 10 in 11 (or about 91%) chance of their condition having been caused by something else. The problem here is that courts typically use a "preponderance of evidence" standard of evidence, meaning the jury must think it's more likely than not that the defendant's actions caused the harm in question. You can think of the preponderance standard as

requiring 51% likelihood in order to find someone liable for damage. In this case, for any particular victim, there's only a 1 in 11 (about 9%) chance that any given victim s cancer was caused by the reactor leak, so the preponderance standard is not met. And this would be true for all 11 victims, so the nuclear reactor would never be held legally responsible for its actions. How could this problem be dealt with? One obvious answer is to have the firm pay 9% (1/11) of the damages. But as DDF points out, to be consistent, we'd have to say the same thing about cases where the jury thinks it's 60% likely that the defendant caused the harm -- he should only have to pay 60% of the damages. Under the status quo, it's all or nothing. So we'd have to change the system to ask juries to assign probabilities to their verdicts, rather than just saying "yes" or "no." Moreover, this would drastically increase the number of cases in the courts, because all the potential cases currently dropped or settled (because the litigants realize there's only a small likelihood of the victory) would now be viable. Indeed, we'd no longer have just two types of verdicts -- all verdicts would take the form of probability assessments. A better solution is to use class-action suits. In a class action suit, a group of people get together and claim that they, as a class, have been harmed by the action of a defendant. In this case, all the victims of this kind of cancer in the area would band together and sue the reactor. For this approach to work, though, it's important that the amount of damage awarded to the group be the marginal amount of damage caused by the action of the defendant. The reactor should not be charged for the entire cost of treating cancer for the whole class, only for 1/11 of that cost. But class-action suits can involve large transaction costs, especially when there are especially large groups involved. For that reason, DDF proposes a small-scale alternative: suits by groups of individuals. In this example, you wouldn't need to get all the victims of cancer together; getting 110 together would be sufficient. This group could prove to the jury that the defendant's action, more likely than not, caused approximately 1/11 of the group's aggregate cancer treatment costs. V. Negligence versus Strict Liability: Unilateral Care Now we finally get to the issue of when we will and when we won't hold a defendant liable for damage that we conclude that they caused. Broadly speaking, there are two types of liability rule that the courts have applied in tort cases: negligence and strict liability. A strict liability rule holds a defendant liable for damages as long as they meet the requirements of causation. For example, a strict liability rule applied to products would say that if you were hurt by your lawnmower (you got your toe cut off while mowing), the producer of the lawnmower would have to pay. A negligence rule holds a defendant liable for damages only if, in addition to having met requirements of causation, he also failed to take "due care." Due care is the level of care

or precaution the court says you should have taken. Usually, due care is defined as the amount of precaution that a "reasonable man" in the same position would have taken. In the case of the lawnmower, a negligence rule might say that the lawnmower company will avoid liability if it included safety instructions, put a strong guard over the blades of the mower, etc. If it failed to do these things, it would have to pay. Which of these rules is efficient? First, we need to be clear what is meant by efficiency here. It does not mean preventing all accidents. Rather, it means taking all cost-justified precautions, a concept we've covered extensively before. A precaution should be taken if its marginal cost is less than the marginal gain in terms of avoided accident damage. Which rule assures cost-justified precautions are taken? Assume for now that only one party, the potential defendant, can take actions that will prevent (or reduce the probability) of accidents. This is known as "unilateral" or "alternative" care. A case where this seems to be true is DDF's example of flying small airplanes. There is very little that homeowners can do to protect their homes against the damage done by falling airplanes, but there is much that the pilots and owners of such planes can do to prevent crashes. It turns out that, in cases like this, strict liability and negligence both achieve efficiency. First, take strict liability. Under this rule, you'll be held responsible for all damage caused by your acts. Anything you do that reduces the likelihood of damage also reduces your expected liability, and by just the right amount. Say that doing $5000 of maintenance on your plane each year will reduce the probability of a crash by 1%, and a typical crash will cause $600,000 in damage (to you and other parties). Then by spending $5000 a year you can save an average of $6000 a year. Obviously, it's in your interest to do so. On the other hand, if you did $5000 more maintenance, it would only reduce the probability of an accident by an additional 0.5%, for additional expected savings of only $3000, so you won't do it. So you end up taking the optimal amount of care. Second, take negligence. This rule will also induce optimal care, so long as "due care" is defined as "economically efficient" care. Take the maintenance above. The court, observing that the cost of the first $5000 of maintenance is less than the expected benefit, will say that due care requires making that expenditure. But it does not require the second $5000 as part of due care. The negligence rule in this case says that you'll be held liable if you spend less than $5000 on maintenance, but you'll have to pay nothing if you spend $5000 or more. So, in deciding whether to spend the first $5000, you realize that you're legally exposed to all the expected savings for doing so, and you do it. In deciding whether to spend the second $5000, you're not legally exposed to the expected savings of doing so, so you don't. Again, we get the efficient outcome. [Draw a diagram showing a horizontal marginal cost of care, with a downward-sloping marginal benefit of care. The marginal benefit is reductions in expected accident losses. Where they cross is the optimal amount of care. This is all independent of the legal rule. The strict liability rule is represented by a private marginal benefit identical to the actual

marginal benefit. The negligence rule is represented by a private marginal benefit identical to the actual when care is less than the optimal, and horizontal at zero when care is greater than the optimal.] VI. Negligence versus Strict Liability: Bilateral Care So far, negligence and strict liability seem equally efficient. But now we add a wrinkle: both the defendant and the plaintiff can take actions that reduce the likelihood of an accident. This is known as "bilateral" or "joint" care. The strict liability rule turns out not to be efficient here. If the plaintiff (victim) knows that all damages will be paid by the defendant, then he will not have an incentive to take cost-justified precautions of his own. For instance, in the car-pedestrian accidents, holding car drivers strictly liable could induce pedestrians not to be careful enough (failing to look both ways before crossing the street, for example). Notice that this is a moral hazard problem. The rule may turn out to be inefficient in another way, too: knowing that the plaintiff won't take enough care, the defendant may be led to take too much care, relative to the amount that would have been cost-justified if the plaintiff acted responsibly. For instance, drivers may drive excessively slow, or plastic bad makers will feel the need to put warnings on bags saying that you shouldn t put them over your head and inhale. Note that a rule of no liability would produce the same result in reverse: the defendant will not take enough care, and the plaintiff may take too much care. The negligence rule, however, does turn out to be efficient. The defendant, as per our reasoning earlier, will take the right amount of care. As a result, he will not be held liable, which means the defendant can expect to have to bear damages from whatever accidents occur. As a result, the defendant has an incentive to take any cost-justified precautions, because he will experience all the full marginal benefits (in damage avoidance) of doing so. There is another alternative: strict liability with contributory negligence. This is just like strict liability, except that the court will specify a level of due care that should be taken by the plaintiff. If he takes that level of care or more, the defendant is held strictly liable, but if he takes less, then the defendant pays nothing. Under this rule, the plaintiff has an incentive to take all cost-justified precautions, because if he doesn't, he's exposed to all expected benefits of doing so. Knowing that plaintiffs will be taking due care, the defendant expects to be held liable, and so he also takes all cost-justified precautions. We could also have negligence with contributory negligence. It works like this: If the defendant takes due care, the plaintiff bears the damages (regardless of his care level). If the defendant fails to take due care, but the plaintiff also fails to take due care, the plaintiff still bears the damages. If the defendant fails to take due care and the plaintiff does take due care, the plaintiff is held liable. This rule is also efficient, for the same reason that negligence without contributory negligence is.

So, the analysis here suggests that negligence is superior to strict liability. But we're not done yet. VII. Negligence versus Strict Liability: Unobservable Actions Suppose that there are many things that a party can do to reduce risk, but only some of those things are readily observable. For instance, a driver can reduce his speed and follow the rules of the road; he can also just drive less often. The former is generally observable, but the latter is not. A negligence rule involves having the court set a due care level. But due care can only be defined in terms of observables. It's easy enough to introduce evidence about your obedience to rules of the road, but it's much harder to figure out whether you were taking too many joyrides. Moreover, it's also very hard to deduce whether your level of driving was optimal, because doing so requires knowing the value of taking car trips to you. The same goes for other unobservables, such as whether you were paying too much attention to the radio. As a result, a negligence rule does not really require the defendant to take all costjustified precautions; it only requires him to take all cost-justified and observable precautions. Under such a rule, the defendant will tend to take the observable precautions, but he will not take the unobservable precautions, because the benefits of taking them (reductions in expected damages) will be borne by plaintiff. A strict liability rule, on the other hand, will induce the defendant to take all cost-justified precautions, even unobservable ones. This is because the strict liability rule concentrates all the benefits of risk reduction on the defendant, thus giving him an incentive to take any action that reduces the expected damage by more than the cost of the precaution. Consider a situation of unilateral care, where the defendant can take both observable and unobservable precautions. Before, we concluded that both strict liability and negligence were efficient. But now, with unobservable precautions, we conclude that only strict liability is efficient. Now consider a situation of bilateral care, where the defendant can take both observable and unobservable precautions, but the plaintiff can only take observable precautions. Before, we concluded that both negligence and strict liability with contributory negligence were efficient. But now, with unobservable precautions, we conclude that only the latter is efficient. Only strict liability can induce the defendant to take unobservable precautions. Thus, now we seem to have an argument for why strict liability (possibly with contributory negligence) is superior to negligence. But there is another wrinkle.

What if both the defendant and the plaintiff can take unobservable precautions? A due care level can only take into account observable precautions. The only way to induce an agent to take all cost-justified precautions is to expose him to the full marginal benefit of those precautions. So we can make the following predictions: A no liability rule will induce the plaintiff to take all cost-justified precautions, but it will not induce the defendant to take any precautions. A strict liability rule will induce the defendant to take all cost-justified precautions, but it will not induce the plaintiff to take any precautions. A negligence rule will induce the plaintiff to take all cost-justified precautions, and it will induce the defendant to take observable, but not unobservable, precautions. A strict liability with contributory negligence rule will induce the defendant to take all cost-justified precautions, and it will induce the plaintiff to take observable, but not unobservable, precautions. To summarize, here is a table, much like DDF's table on p. 203: Observable care by defendant Unobservable care by defendant Observable care by plaintiff Unobservable care by defendant No Liability Strict Liability Negligence Strict Liability + Contributory No Yes Yes Yes No Yes No Yes Yes No Yes Yes Yes No Yes No Residual liability. Another way to think about this is in terms of the residual liability. The residual liability is the legal liability that remains after everyone has satisfied the explicit requirements of due care, if any. Anyone who is exposed to residual liability will take all cost-justified precautions. Under a rule of no liability, there is no requirement of due care, so residual liability rests on the plaintiff. So the plaintiff takes all cost-justified care, but the defendant does not take any kind of care. A rule of strict liability is the mirror image of no liability: since there is no requirement of due care, residual liability rests on the defendant. So the defendant takes all costjustified care, but the plaintiff does not take any kind of care. A negligence rule places a due care requirement on the defendant, but once it's satisfied, residual liability rests on the plaintiff. So the plaintiff takes all cost-justified care, but the defendant takes only observable care.

A rule of strict liability with contributory negligence is the mirror image of the negligence rule: it places a due care requirement on the plaintiff, but once it's satisfied, residual liability rests on the defendant. So the defendant takes all cost-justified care, but the plaintiff takes only observable care. What should we conclude from this analysis? We've found that each type of rule has some kind of disadvantage. Which rule is the best in any given type of case depends on the facts of the situation. Keep in mind that, when two rules will both induce equally efficient behavior by the agents, the rule that requires fewer transfers of wealth through the legal system will generally be preferable. (Why? Because litigation is a costly rentseeking process.) Suppose there's unilateral care by the defendant, and there are lots of unobservable things he could do to reduce risk. Then go with strict liability. Example: transportation of hazardous materials; keeping wild animals as pets. Suppose there's unilateral care by the plaintiff, and there are lots of unobservable things he cold do to reduce risk. Then go with no liability. Example: the assumption of risk doctrine, which eliminates (most cases of) liability for damage that occurs during dangerous sports like boxing. Suppose there's bilateral care, and only the defendant can take unobservable precautions. Then go with strict liability with contributory negligence. Example: an amusement park might be held liable for most damage to guests, but it might successfully raise a contributory negligence defense if a plaintiff deliberately went into a restricted area. Suppose there's bilateral care, and only the plaintiff can take unobservable precautions. Then go with negligence. Example: products liability. Even though products liability is governed by strict liability in name, negligence is brought in through the back door by inserting notions of reasonability into the definition of a product defect. Etc. VIII. Information Requirements of Negligence Rules There is one more disadvantage to negligence rules, and that's that they require more information to apply. In order to set due care levels optimally, the courts must possess a great deal of specific information. This can be very costly, and there is good reason to believe the courts may get it wrong. If so, then the efficiency results above may not hold. Suppose the court sets a defendant's due care level too low. That means the defendant can avoid all liability by taking a low level of care, so naturally he'll do so. The result is inefficiently low care. In addition, the plaintiff may respond by taking too much care (example: over-the-counter medicines with foot-long instructions wrapped around the bottle).

Now suppose the court sets a defendant's due care level too high. You might think this would lead the defendant to take too much care -- and that's true, as long as the due care level is not that much higher than the optimal. But if the due care level is high enough, some defendants will decide that's too expensive; they might as well be under a strict liability rule. These defendants will actually take optimal care, just as under a strict liability rule, because they figure they'll be exposed to the full liability regardless, and thus they ll be exposed to all benefits of whatever care they take. (This is assuming, of course, that this is a unilateral care situation.) Complicating matters more is the fact that the particular cases in a class of cases are not usually identical. What would be optimal care in one case is not necessarily optimal in another. Here, the court faces a choice. On the one hand, it can set a different care level for every single case. The problem here is that agents will have a difficult time predicting the behavior of the courts. On the other hand, it can set a single care level for all cases. The problem here is the rule will err by both over- and under-inclusion: the care level will be too high for some cases, too low for others. When choosing a rule that will apply to an entire class of cases, the courts have to realize that inefficiencies from both over- and under-inclusion will occur. They ll have to consider all types of inefficiency, finding the rule that minimizes the total wealth loss. To do this, they need to know the distribution over all potential cases in the class. The problem is that the courts are unlikely to possess this information, because the set of cases that come before the courts are a biased sample. They do not represent a random sample of all possible cases, because the cases brought to court depend on the legal rule currently in effect. IX. Punitive Damages In everything that has gone so far, we have implicitly assumed that someone who commits a tort (and meets the legal requirements for being held liable) will definitely have to pay damages. But this is certainly false. Not every tort is detected. Not every victim sues, even if he detects the damage and knows who did it, because litigation can be very costly. And not every valid case brought to court is decided correctly. For these reasons, the probability of actually having to pay damages may be substantially less than 100%. This is important economically, because a low probability of payment reduces the incentive to take efficient precautions. Take the earlier example of an airplane owner. The first $5000 worth of maintenance saves an expected $6000 in damage (.01*$600,000). But say there is only a 75% chance that, if an accident occurs, the case will be brought to court and decided in the plaintiff's favor. Then the expected damage payment is actually only.75*$6000 = $4500. Since this is less than $5000, the owner doesn't take the proper precautions. The obvious answer to this problem is to scale up the amount of damages to create the right incentive. If D is the amount of damages awarded by the court, and it gets awarded

with 75% likelihood when the accident has in fact occurred, then we need.75d = 600,000, or D = 6000*(1/.75) = $800,000. Notice now that that airplane owner sees his expected gain from maintenance is (.01)(.75)($800,000) = $6000, exactly what it should be. More broadly, the actual damage should be multiplied by the inverse of the probability of being successfully sued. (Note: We're assuming here that this is an actual tortfeasor, someone who did in fact cause the damage in question. So the probability to use should not simply be the probability of any defendant, including the falsely accused, being found liable. It should be the probability of a defendant who actually caused the damage being found liable.) This discussion provides a potential explanation for the existence of punitive damages. Punitive damages are damage payments that go above and beyond the amount necessary to compensate the victim. They have become more and more common since the 19 th century. However, there are difficulties with the probability-scaling explanation of punitive damages: the legal requirement for punitive damages in the status quo is that the defendant's action had to be "deliberate or reckless." Why limit punitive damages in this way? The logic of probability scaling does not require deliberateness or recklessness. DDF therefore offers three other (economic) explanations of punitive damages. 1. Punitive damages are a way of "playing it safe" with offenses that are assuredly inefficient, with hard to measure damages. The problem with a damage payment that's too large is that it may over-deter; that is, it may induce inefficiently high care. But if an action is deliberate or reckless, the implication is that the agent failed to take even the most obvious and inexpensive of precautions -- in other words, precautions that are clearly efficient. So if we make damages extra large in such cases, there is no real danger of inducing too much care. A problem with this explanation is that deliberateness is a very imperfect standard. There may be torts that are both deliberate and efficient (such as throwing sparks on crops, being fully willing to pay the cost, but being unable to obtain advance permission because there are too many farmers). A high punitive damage award on grounds of deliberateness in a case like this would, in fact, induce too much care. 2. Punitive damages are a means of spending greater resources on the most deterrable torts. Obviously, we would like to prevent inefficient actions. But we only want to do so if the gain in efficiency is greater than the cost of deterrence. For instance, suppose some inefficient action creates a $500 loss in wealth; then it would not make sense to spend $1000 to prevent that inefficient action. This means that in deciding how to deploy our legal resources, we want to do so in a many that deters the greatest amount of inefficient behavior. How can we do that? We want to spend more legal resources on those actions that have the most elastic response to our efforts, and fewer on those actions that have a more inelastic response.

In the context of tort law, a higher amount of potential damages to be awarded will increase the amount of legal expenses incurred in dealing with it. Higher stakes induce greater investments in the case (a form of rent seeking). So we only want to raise the damage award if it's worth it -- if the inefficiencies deterred justify the greater legal expenditures. Thus, it makes sense to raise the damage award in just those cases where the tortious action is most responsive to such damages -- and reckless and deliberate torts are likely to be very responsive, because they involve behavior that is easy to control and alter. 3. Punitive damages are designed to deter strategic torts. A compensatory damage payment typically does not fully compensate the victim, for various reasons: there is some chance the damages won't be paid, the victim often has to pay attorney fees, there are pain and suffering costs that are hard to calculate, etc. For that reason, people would usually rather not be the victim of a tort even if compensation is paid. As a result, it may be possible for someone to commit a deliberate tort, the intention of which is to send a message to people other than the victim. (DDF gives the example of a bully beating up someone who flirts with his girlfriend; he is sending a message to anyone else who might flirt with his girlfriend.) The message is an implicit threat to impose all of the uncompensated costs of being a tort victim. If the message is successful, the tortfeasor gets a benefit from committing the tort that exceeds the damage payment he has to make. Clearly, an additional payment is necessary to deter this sort of tort. To put it another way, the total benefit to the tortfeasor of committing the tort may exceed the immediate benefit. Beating up someone who flirts with my girlfriend doesn t just keep one competitor away -- it potentially keeps many other competitors away as well. So I might be perfectly willing to pay for the damage suffered by this one guy, because I m willing to pay that much to send a message to other guys. To prevent me from committing this tort, then, the court needs to inflate the damages. (But, you might ask, isn t my tort an efficient one if the benefits to me exceed the damages to the person harmed? Yes, except that in this case, the added benefits to me result from having sent a message to other guys that I might commit torts against them as well.)