If someone shoots you, you call a cop. If he runs his car into yours, you call a lawyer. Crimes are prosecuted publicly, torts privately. This chapter is devoted to torts, the next to crimes. A later chapter examines both, asking whether there are good reasons to have two different systems, whether there are good reasons why each is associated with a particular set of legal rules, and whether there is a good reason why our legal rules sort offenses into the categories of crime and tort in the way in which they do&endash;whether, for example, it might be better to treat burglary as a tort and auto accidents as crimes.
A tort is a wrong that is privately prosecuted, typically for damages, although some tort suits seek injunctions. Unlike contract law, which enforces obligations voluntarily agreed to, tort law deals with obligations imposed by law. To do so it must answer four questions:
1. What makes an act wrongful? When I go into competition with you I make you worse off, yet competition is not a tort. Implicit in tort law is some set of rights defining what acts are or are not wrongful.
2. What does it mean to say that my act caused your loss? Events in the real world have multiple causes. I would not have run you over had you not been walking down the road to have dinner with a friend&endash;did he cause the accident?
3. Under what circumstances is someone liable for a loss he causes? What if I caused an accident, but there was no reasonable precaution by which I could have avoided it? What if both parties were negligent: I was driving too fast, but you were standing in the middle of the road paying no attention to traffic?
4. If I am liable, how much am I liable for?
As we will see, all of these questions can be analysed in terms of economic efficiency. Whether doing so explains current law is less clear.
Think of tort damages as a way of forcing potential tortfeasors to take account of the costs their acts impose on other people, a legal mechanism to internalize externalities and thus produce efficient choices. As I argued back in chapter 3, your loss when I go into competition with you is only a pecuniary externality, not a net externality—you are worse off, but our customers are better off. On net, I impose no external cost so should not be charged for your loss. Competition should not be a tort and, at least since the 15th century, has not been.
To make the point clearer, imagine that I become my city's 101st physician. Before I did so, the other 100 physicians were each seeing a hundred patients a month at $10 a visit. My competition drives the price down to $9.90. At the lower price, patients are a little more willing to see a doctor—total visits increase by 100/month, just enough to keep me busy.
Each of the other physicians is now worse off by $10/month as a result of my entering the business, ten cents less a month for each of a hundred patients, which may explain why none of them invite me to join their golf game. That loss, however, is precisely matched by their patients' gain, since the patients of each doctor are now paying a total of $10/month less in medical bills. The additional hundred visits that are being made as a result of the lower price must be worth less than ten dollars apiece to the patients, since they did not make those visits at the old price, but more than $9.90 apiece, since they do make them at the new. So my services must be worth between $990/month and $1,000/month. I am getting $990/month for them. My reward almost precisely measures the value of what I produce, so I have almost precisely the correct incentive to become a physician. Competition is not a tort.
Neither is dressing badly. When I wear a green shirt with purple pants or forget to mow my lawn for a month, I impose a real externality on my neighbors. Yet the former is never tortious and the latter rarely. The externality is hard to measure and its size too small to be worth the cost of dealing with it through the legal system, so we use less formal mechanisms instead. Acquaintances ask, in a pointed way, if perhaps I am color blind. My neighbors mow my grass for me—along with my newly planted fruit tree (true story).
Ordinary tort damages are a liability rule, so another reason not to treat something as a tort is that we prefer a property rule. If I steal your car, you don't sue me for the resulting inconvenience, you call the cops. Since I do not want to be arrested, I buy your car instead.
So we now have three different reasons why acts that impose costs might not be tortious: that they result in a transfer, not a net cost, that they impose a net cost but are not worth controlling through the tort system, or that they are best controlled by a property rule rather than a liability rule. Acts that impose costs on others but do not fit any of those categories are quite likely to meet the requirements for a wrongful, hence tortious, act.
The next question is, given that a cost was imposed, what does it mean to say that I imposed it—that I caused the loss.
Falling Safes and Very Dead Hunters
I stop my friend in the street to chat. He continues on down the street. As he passes by an office building, a safe falls out the window and crushes him. Have I caused his death? Should I be liable?
One sense of "I caused his death" is "had I not acted as I did, he would not have died"—the "but for" definition of causality. In that sense, I killed my friend—if I had not delayed him, he would not have been under the safe when it fell. Yet it would seem odd to blame me and odder still to hold me liable. Why?
This puzzle is based on a real case. A tree fell on a moving trolly, injuring passengers. One of them sued. He succeeded in demonstrating that in order for the trolly to be where it was when the tree fell on it the driver had to have driven faster than the speed limit at some point during the trip. Breaking the law is per se negligence, so the driver was legally negligent whether or not his driving was actually unsafe. If he had not driven over the speed limit the bus would not have been under the tree when he fell, so, the plaintiff argued, his negligence caused the injury.
The plaintiff lost; the court held that the driver's negligence had not caused the accident in the legally relevant sense. Logically, the court was wrong; economically, it was right. Why?
The driver decided how fast to drive before he knew when and whether the tree would fall. Driving faster did not increase the ex ante probability that the tree would fall on the trolly; if it had fallen a few seconds later, the extra speed would have prevented the accident instead of causing it. We could, in principle, give the driver the right incentives by penalizing him when his decision causes an accident and rewarding him when his decision prevents one but it is a great deal easier to do neither. Since his decision imposes no ex ante cost, imposing no liability gives him the right incentives.
The same argument applies to the falling safe. Ex post, judged by what actually happened, your decision killed your friend. Ex ante, judged by the information available to you at the time, your decision had no effect on the probability your friend would die. There was a very tiny chance that delaying him would put him under a falling safe, just balanced by the very tiny chance that not delaying him would put him under a falling safe. You caused your friend's death. But for the law to take notice of that sort of causation produces less efficient outcomes than for the law to ignore it.
Coincidental causation is an extreme case of the general problem of foreseeability. The main reason to award tort damages, from the economic standpoint, is to give people an incentive not to do things that impose costs on others. But that does no good if the actor does not and cannot know that his action will impose costs and so lead to his paying damages. Hence unforseeability is a defense to tort liability. That conclusion does not apply where the actor did not know the consequence of his actions but readily could have found out. In such situations, the fact that he will be liable if damage is done gives him a desirable incentive—to find out what the consequences of his actions are likely to be and, having found out, to modify his behavior accordingly. Many real cases fall somewhere between "was, or easily could have been, foreseeable" and "entirely unforeseeable," leaving a difficult line drawing exercise for the courts. We will leave them with it, and go on to a second complication in the logic of causation.
Two hunters simultaneously, independently, and accidentally mistake a third hunter for a deer and shoot him, one through the head and one through the heart. The widow sues. Each hunter responds that he will be happy to pay for the damage he has caused, but that the difference between the effect of one bullet and of two is small—perhaps a slight increase in the undertaker's bill. Put in the language of economics, each points out that while the average cost per bullet was half a life, the cost imposed by the marginal bullet was close to zero and, if we conceive of tort law as a system of incentives, it is the marginal cost that matters. The conclusion seems paradoxical, yet the argument is logically sound.
If you find that implausible, try working the problem through with numbers. Suppose that if either Al or Bill hunts, his chance of shooting Carl is ten percent. If both of them hunt, each has a .1 probability of shooting Carl but Carl has only a .19 probability of being shot— a .09 probability that Al will shoot him, a .09 probability that Bill will, and a .01 probability that both will.
Suppose we (somehow—see chapter 9) value Carl's life at a million dollars. Further suppose that the law accepts the argument offered above. If either Al or Bill shoots Carl, he pays a million dollar fine. If both of them shoot Carl, neither pays anything. Does this give them the right incentives—make it in their interest to properly allow for the probabilistic cost they impose on Carl by deciding to go hunting?
If Al is not hunting, Bill's decision to hunt imposes a .1 probability of a million dollar loss, giving an expected cost of $100,000. It also costs Bill $100,000 in expected damages, in the form of a .1 probability of having to pay a million dollars to Carl's widow. Bill has the right incentive. He will go hunting only if doing so is worth $100,000 (perhaps he is a very enthusiastic hunter).
Next suppose that Al is hunting. Bill's decision to hunt now increases the probability of Carl dying from .10 to .19, imposing an expected cost of only $90,000. Bill's expected damages are also $90,000, since he has a .09 chance of killing Carl all by himself and owing a million dollars in damages and a .01 chance of firing one of the two bullets that kills Carl and owing nothing. Again, the rule gives Bill the right incentive.
Repeat the argument for Al to get the same result. Substitute a different rule, say one in which a double bullet killing results in each hunter owing half a million, run through the calculations, and you will find that each has an inefficiently high incentive to stay home.
The argument clashes sharply with most people's intuition of what the law should be. One reason is that, in the two bullet case, it provides no money for Carl's widow. If the purpose of the tort system were providing insurance, making neither Al nor Bill liable wouldn't do it.
Tort law also provides Carl's widow nothing if Carl trips over a root while he is hunting and shoots himself. For reasons that will be covered at greater length later in the chapter, tort law is a poor way of providing insurance. It makes more sense to use insurance companies to provide insurance and tort law to deter actions that impose cost on others.
A second reason it seems wrong is that the numbers don't add up. Al and Bill, between them, have done a million dollars worth of damage, yet I am claiming that the sum of the damage done by Al and the damage done by Bill is zero. The confusion here is between marginal damage and average damage. The marginal damage done by Bill, given what Al did, is zero; Carl is no more dead than if Bill hadn't been hunting. Ditto for Al. The average damage is half a life each.
If this seems puzzling, consider one of the most famous puzzles in the history of economics—the diamond/water paradox. Diamonds are much less useful than water—compare a world with no diamonds to a world with no water. Yet they cost much more. Why?
The answer is that although the total value of all the water I use is very large, the marginal value of the last gallon, used to water the lawn, is low. In market equilibrium (for details, see a good price theory book) price equals marginal value, because people keep increasing the amount they use of something as long as a little more costs less than it is worth. They stop when the value to them of one more gallon of water is just equal to its price. They pay the same price per gallon for all of the gallons they get, including the (very valuable) gallon that makes the difference between life and death. They are buying water at its marginal value, which is much lower than its average value. Similarly, Al and Bill each pays the marginal cost he imposes, which is much less than the average cost.
Why do I spend so much time on such a counterintuitive argument? Precisely because it is counterintuitive. As long as economics tells us things we already believe, it is easy to ignore the logic and focus only on the conclusion. An economic argument that logically derives a conclusion that feels wrong forces us to think more carefully about both the argument and the intuition it contradicts.
In this particular case, the logic of the argument is correct but the legal conclusion, at least for the case of two hunters, is wrong. To see why, we must expand the argument beyond the issue of causality.
One problem with the conclusion is that it depends on my assuming that Al and Bill are acting independently, that the double killing was a freak accident. If we accept the legal principle of no liability for double shootings, the next one may not be an accident. My proposed legal rule presents an obvious opportunity for getting away with murder.
A second problem arises from my implausible assumption that Al and Bill could actually pay fines that properly compensated Carl for the cost of being killed, at least in the sense explained in chapter 9. This brings us back to the issue of ex ante vs ex postante vs ex post punishment discussed in chapter 7. One argument for ex ante punishment is that ex post requires large punishments (imposed with low probability), which the offender may not be able to pay. We then have the choice of either a fine that is lower than the damage done and so underdeters the offense or a more severe but more costly punishment such as execution or imprisonment.
One solution to the problem posed by judgement-proof hunters is to supplement the inadequate ex-post damage payment for killing Carl with an ex ante punishment for careless hunting. One way of doing so is to fine the hunters in the double bullet case. Neither hunter did any serious damage, since Carl would have been just as dead from only one bullet. But the fact that they shot him is evidence that they were hunting carelessly. Hunting carelessly is an activity we would like to punish. On exactly the same grounds, we would also punish them for a near miss, assuming we could prove it, although that, too, does no harm.
While this argument implies that Al and Bill ought to be liable even in the two bullet case, it would be a mistake to generalize that conclusion. One can imagine other cases of redundant causation where independence was clear, the parties were not potentially judgement proof, and the argument I have offered would go through, implying that neither party should be liable.
One can also imagine cases where the same line of argument leads to a counterintuitive result in the opposite direction. Suppose two people independently commit tortious acts such that either act alone does no damage, but the two together do a great deal of damage. This time marginal damage is greater than average damage; the same argument that implied that neither Al nor Bill owed anything now implies that each party should be separately liable for the full amount of the damage. Ingenious readers are invited to invent their own examples.
There is a delightfully logical illogic to causation in law, examined through an economic lens. In the case of coincidental causation, you did in fact cause your friend's death—but for your act he would be alive—yet the law pretends you did not and, economically speaking, is correct to do so. In the case of redundant causation, neither Al nor Bill caused Carl's death, yet the law pretends they did and, economically speaking, is correct to do so.
Someone running a nuclear reactor makes a mistake and radioactive gas leaks out. Careful calculations predict that the result will be to increase the local cancer rate for the next twenty years from ten cases a year to eleven. Is the owner of the reactor liable, and if so for how much to whom?
Under traditional common law, the answer was "no." If a cancer victim sues, the defendant points out that since only one cancer case in eleven is due to his negligence the odds are ten to one that this particular case would have happened anyway. A tort plaintiff must prove his case by a preponderance of the evidence, which is usually interpreted as requiring him to show that it is more likely than not that the defendant is guilty. Since the probability that the defendant is responsible for the injury is only about nine percent, the plaintiff loses.
How might the legal system handle such cases? Not, surely, by holding the reactor liable for all the cancer cases. If we do that, the reactor is charged eleven times the damage done by its negligence, which is both excessive and inefficient. Generalizing that principle, any time a hunter is injured by a stray bullet and the guilty party is unknown every hunter in the woods would be found guilty and required to pay the full cost of the injury.
One possibility might be to permit tort actions for probabilistic wrongs. A cancer patient could demand 1/11th of the damages he would have gotten if the reactor had been solely responsible. Alternatively, everybody in the area could sue, asking damages based on the ex ante cost, the increased risk of getting cancer.
There are problems with this approach. To begin with, if we are going to apply it consistently, all civil verdicts should take a probabilistic form. If someone who is nine percent likely to have caused a loss owes nine percent damages, then someone who is sixty percent likely should owe only sixty percent damages and not, as at present, a hundred percent. An additional problem, if we permit suit for the risk of cancer cases that have not yet happened, is that before the cases have occurred we have very little evidence on what effect the radiation release actually had. And pooling claims by (say) a hundred thousand people requires a class action or something similar, raising serious incentive problems—how to make sure the attorney serves the benefit of the class instead of his own. Given the limitations of a trial as a way of solving a complicated problem, there seems much to be said for the all or nothing nature of verdicts under present law.
A less radical and more elegant solution would be suits by groups of victims. Suppose a hundred and ten cancer victims pool their cases and sue together. They have no way of showing which of them got cancer because of the release. But they can use statistics to show that the best estimate of the consequence of the release is that among them they suffered ten more cases than they otherwise would have, and that the defendant therefore owes the group a corresponding amount of damages. Alternatively, if our legal system treated tort claims as transferable property (as the legal system of Iceland, which we will encounter in chapter 17, did a thousand years ago), a law firm could buy up the claims of a large number of victims and then sue the reactor owner. That is a possibility we will return to in chapter 18.
So far as I know, the joint suit approach was never actually applied to such cases. But there is at least one famous case that was handled in an analogous way, with the roles of plaintiffs and defendants reversed. It involved Diethylstilbestrol (DES), a fertility drug that turned out to produce medical complications in the daughters of some of the women who had taken it. The complications took a minimum of ten to twelve years to appear, and the most serious, a form of cancer, appeared, according to varying estimates, in between one in 250 and one in 10,000 of the daughters.
The case raised two issues. The first was whether there was any basis for holding the producers of the drug liable. In order to avoid releasing a drug with a side effect that takes a generation to manifest itself, a drug company would have to test each new drug for a full generation before releasing it, carefully monitoring the medical condition of not only everyone who took it but all of their children as well. And they would have to do it on a very large scale in order to detect so small an effect. That seems an excessive level of precaution, likely to result in much more expensive drugs and a lot of unnecessary death and misery to those denied new drugs during such an extensive testing period. So one could plausibly argue that the effect was not foreseeable at any reasonable cost and the drug companies therefore ought not be held liable. By reaching the opposite result, the court provided no benefit in greater precautions, since even if the companies were liable the precautions necessary to prevent such a problem were not worth the cost of taking, and it generated a considerable cost in expensive litigation.
The other issue was, if the producers were liable, which producer was liable to which victim. DES was an unpatented drug produced by a number of different companies, no one of which had a majority of the market. By the time the side effect finally manifested itself, there were few surviving records to show which company had made the dose used by any particular woman. If a woman sued a drug company for damages, the company could (truthfully) reply that the odds were much less than even that it had produced the dose she took. The court settled the problem by allocating damages to the drug companies in proportion to market share, thus involuntarily pooling the defendants in a way roughly analogous to the voluntary pooling I have suggested for plaintiffs.
"It is just and reasonable that if a person uses a dangerous machine, he should pay for the damage which it occasions; if the reward which he gains for the use of the machine will not pay for the damage, it is mischievous to the public and ought to be suppressed... ."
(Ramwell, L.J. in Powell v. Fall, 5 Q.B. 597 (1880))
After a long detour through the jungles of causation, it is time to return to a very simple question. I take actions that may impose costs on others—drive a car, shoot a rifle, blow up rocks with dynamite. The size and likelihood of those costs depend on what precautions I take. How can we use tort law to give me an incentive to take those precautions that are worth taking, and only those?
Our objective is not to eliminate the risks entirely—we could do that by banning cars, rifles, and dynamite. Our objective is to get the efficient level of precautions, and thus the efficient level of risk. We want a world where I get my brakes checked one more time if, and only if, doing so reduces expected accident costs by at least as much as it costs. We want a world where I break up rock with dynamite instead of a sledge hammer if and only if the savings in cost to me at least makes up for the increased risk to my neighbors. What we want is not a world of no accidents—that costs more than it is worth—but a world with only efficient accidents, only those accidents that cost more to prevent than preventing them is worth. We want the world we would have if everyone took all and only cost-justified precautions.
To simplify things, I start with the simplest case—unicausal accidents. I am engaged in an activity, flying a small airplane, which has some chance of injuring other people's persons and property. The probability of such injury depends on what precautions I take but not on what precautions they take. There is nothing other people can do, short of armoring their roofs with several feet of reinforced concrete, a precaution we are confident is not worth the cost, to protect themselves against the risk that I might crash my plane into their houses.
We also start with a simple legal rule—strict liability. If I cause damage, I am liable. In deciding whether and how often to fly, how safe (and expensive) a plane to use and how often to get it checked, I will now take account of both costs to me and costs to others, since I bear the former and am liable for the latter (we will assume that if my plane goes down I will have time to get out with a parachute, and will thus survive to pay damages). So it is in my private interest to take all precautions that are worth taking. Strict liability leads, in this case at least, to the efficient level of precautions.
Next consider a different legal rule—negligence. Under a negligence rule, at least as economists define it, I am liable for an accident if and only if I did not take all cost-justified precautions to prevent it. This view of negligence appears in the law as the Hand formula: A party is negligent if there is a precaution he could have taken to reduce the probability of the accident such that the cost of the precaution was less than the reduction in probability times the cost of the accident. A more intuitive way of putting it is that a party who imposes costs on others does so negligently if he has failed to take precautions that a reasonable person would have taken if he himself were the one bearing the cost of the accident.
Suppose airplane crashes are governed by a negligence rule. I have two alternatives: I can take all cost-justified precautions (taking account of costs to others as well as to myself) or not. If I do not take all cost-justified precautions and there is an accident, I will be liable, just as if the rule were strict liability. But as we have just seen, if I am going to be liable it is in my interest to take all cost-justified precautions. So under a negligence rule, just as under a rule of strict liability, it is in my interest to take all and only cost-justified precautions.
How does which rule we use affect how much litigation we can expect? Under strict liability, I am liable even if I took all cost-justified precautions, so we would expect strict liability to lead to more lawsuits than negligence. On the other hand, under strict liability lawsuits ought to be simpler, since the plaintiff does not have to prove that the defendant was negligent. So the overall effect is unclear.
So far I have assumed that courts are fully informed, that they know not only what the damage was and who caused it but also what precautions the tortfeasor took and what precautions he should have taken. Let us now weaken that very unrealistic assumption. Precautions are of two sorts: observable and unobservable. In the case of observable precautions, the court knows both what you did and what you should have done. In the case of unobservable precautions, the court either does not know what you did (whether, for example, I was paying attention to traffic while driving down the road or thinking about how to revise this chapter) or it does not know what you should have done. Since, in the case of unobservable precautions, the court cannot tell whether you were negligent, negligence now means failure to take all cost-justified observable precautions.
Under the new assumptions, strict liability and negligence produce different results. Under strict liability the court does not have to know what precautions you took or should have taken, so the result is unchanged; it is in your interest to take all and only cost-justified precautions. But under negligence, you have a more attractive option available. You take all precautions that are both observable and cost-justified. Having done so you are non-negligent, so if an accident happens you will not be liable. You now choose the optimal level of unobservable precautions based on the cost of an accident to you but ignoring the cost to the victim.
Suppose I am involved in a minor accident. At trial, I succeed in showing that I was driving at a reasonable speed, swerved when a deer ran in front of my car, and as a result dented a parked car. No cost-justified improvement in how I drove would have prevented that accident.
I could, however, have prevented the accident by not driving. A perfectly informed judge might calculate that the net value to me of that particular trip was only ten cents, the expected damage to people and property I might run into in the course of the trip was twenty cents, so not taking the trip was thus a cost-justified precaution. If so, then I was negligent from the moment I started the car and should be held liable for any damage I did thereafter.
In the real world, judges do not have that sort of information; they know whether I took the trip but not whether I should have. How much I drive, my activity level, is thus an unobservable precaution. In the law and economics literature, the distinction between observable and unobservable precautions is often given as the distinction between precautions and activity level. That is a convenient shorthand, since activity level is an important unobservable precaution, but a misleading one, since it is not the only such.
Strict liability does, and negligence does not, give me an incentive to take account of external costs in deciding on unobservable precautions such as activity level. So the more important unobservable precautions are, the stronger the case for strict liability over negligence. Judge Posner has proposed this as an explanation for the legal rule that holds ultrahazardous activities, such as removing tree stumps with dynamite or keeping a pet lion in a cage in your back yard, to be subject to strict liability. His (plausible) argument is that "ultrahazardous" is simply the legal term for the sort of activities for which the efficient precaution is often not to do them.
To see the same argument applied in the opposite direction, consider the Texas Supreme Court's decision in Turner v. Big Lake Oil Company, mentioned several chapters back in my description of how courts make law. The question was whether or not to follow the precedent of Fletcher v. Rylands, the English case that established strict liability for bursting reservoirs. The court declined to do so, in part on the grounds that storing large quantities of water might have been an unusual use of land in England but was essential if one wished to grow anything in the parched lands of West Texas. Putting their argument in our terms, there was no point in using strict liability to give farmers an incentive to prevent accidents by not having reservoirs, since they were going to have reservoirs on their property whatever the legal rule was.
One common feature of the two cases, one in England and one in Texas, is that in each the court was clearly conscious of the wider implications of its verdict. The English judges discussed the potential loss of life from bursting reservoirs, despite the fact that the actual accident being litigated killed nobody. Presumably they had in mind catastrophes earlier in the century in which bursting dams had killed hundreds—in one American case thousands—of people, and wanted to establish legal rules that would make such accidents less likely in the future. The Texas court based its result in part on the importance of reservoirs for agriculture, although the particular reservoir whose failure was being litigated contained, not water for irrigation, but salt water produced as a byproduct of pumping oil. Both sets of judges were looking at the case from what I earlier described as a forward-looking perspective. They were determining not merely how to deal with a particular accident that had already happened but how future accidents would be dealt with, and thus affecting the incentives of the people whose future actions would determine how likely future accidents were to occur.
So far, by assuming that only one party can take precautions, I have avoided the complications which Coase introduced to the theory of externalities. We now shift the example from flying a plane, which might crash into a house, to driving a car, which might run into someone else's car. How likely I am to run into you depends both on how I drive and on how you drive. Auto accidents are subject to dual causation; how likely they are to happen depends on decisions by both parties.
Auto accidents are also dual events in another sense; when two cars run into each other, both are damaged. That feature of the problem makes the analysis somewhat more complicated without adding anything interesting to its essential logic. I will therefore simplify the discussion by assuming that all collisions occur between automobiles and tanks. Both drivers affect the chance of the collision but only the car is injured. If we wanted to drop that assumption, we could do it by treating each collision as two torts, one each way, and applying our analysis to each in turn.
Start with a legal rule of no liability; if a car gets crushed, that is the driver's problem. Drivers have the proper incentive to take precautions, since they will have to pay the cost of any accident. Tanks bear no costs and so take no precautions. The rule leads to efficient behavior by drivers of cars but not by drivers of tanks.
To solve that problem, we switch to a rule of strict liability: tanks are strictly liable when they damage cars. Now the tank driver takes full account of accident costs in deciding how to drive (and how often to have his treads checked and whether to engage in random target practice while rolling down the highway). But the driver of the car, knowing that if a tank crushes his car the driver will have to pay to replace it (all cars in my hypothetical are equipped with ejection seats, so only the car is injured), takes no precautions at all. We have solved the incentive problem for the tank but created one for the car.
Next consider a negligence rule. When an accident occurs, the tank driver is liable if and only if he was negligent. If we assume a fully informed court, the result is efficient. Tank drivers take all cost-justified precautions, for the same reason as in our earlier discussion: If they don't they will be liable, and if they are going to be liable it is in their interest to take all cost-justified precautions. Since tank drivers take all cost-justified precautions, tanks are never negligent, hence never liable. Since tanks are never liable, car drivers know they will have to pay the cost of accidents. So it is in their interest to take all cost-justified precautions as well. A negligence rule solves the problem of getting efficient incentives in a world of double-sided causation.
So does a rule of strict liability with a defense of contributory negligence. Under that rule, the tank is strictly liable unless the car was negligent. Car drivers know that if they fail to take all cost-justified precautions they will have to bear the full costs of accidents, and if they are going to bear the full cost it is in their interest to take all cost-justified precautions. Since car drivers are never contributorily negligent, tank drivers know they will be held strictly liable so it is in their interest to take all cost-justified precautions as well. We now have not one solution to the problem but two.
To make these solutions work we had to assume that the court was fully informed, that it knew what everybody did and what everybody ought to do. But under that assumption, there is an even simpler solution to the problem of dual causation. The court announces that, any time there is an accident, any driver who has not taken the cost-justified level of precautions will be shot. Nobody wants to be shot, so everybody takes the cost-justified level of precautions.
We have just worked our way back to the simplest solution to the problem of externalities—a central authority who figures out what everyone should do and tells him to do it. Both evidence and theory suggest that, save in very simple situations, that solution does not work very well. It requires the central planner to have information he is unlikely to have and it depends on the interest of the planner being the same as the interest of the planned, which is rarely the case.
In order to include some of these problems in our analysis, we again drop the assumption of fully informed courts and divide precautions into observable and unobservable. The result, applying the earlier arguments to the more complicated case of dual causation, is summarized in table 14-1.
The table has an interesting symmetry: columns 1 and 2 are identical, save that the roles of tank and car are reversed, and similarly for 3 and 4. That is not an accident. No liability and strict liability are the same rule, with the parties reversed: Under no liability, the car always pays for the damage; under strict liability, the tank does. Negligence and strict liability with a defense of contributory negligence are the same rule as well—one party pays the cost unless the other party is negligent, in which case the negligent party pays. The rules seem different because it is the car and not the tank that gets smashed, so if the law does nothing it is the car owner who must pay the cost. But if we think of the cost as jointly produced by the decisions of both and the liability rule as a way of deciding who will bear it, the symmetry becomes clear.
Optimal Level of Liability +Contributory Care by Tank Activity by Tank Care by Car Activity by Car
Optimal Level of
Care by Tank
Activity by Tank
Care by Car
Activity by Car
Table 1 might be useful if we were actually designing a legal system. Suppose there is some category of accident that produces a cost initially born by one party (Car) as a result of decisions by both (tank and car). If the party in the best position to take precautions is car, if there are lots of things worth doing by the driver to prevent the accident and few or no cost-effective precautions the tank can take, then no liability is a sensible rule. If we reverse the assumption, strict liability is a sensible rule.
Suppose neither works. Some of the precautions worth taking should be taken by the car, some by the tank. The next step is to separate precautions into observable and unobservable ones. Suppose the important precautions cars can take are observable—not driving drunk, not veering in front of tanks, and the like—but the main thing tanks can do to reduce the number of accidents is to stay off the road. Staying off the road counts as an unobservable precaution, since although the court knows whether the tank was on the road—if it had not been on the road it would not have run over my car—the court does not know whether it should have been. Under these assumptions, the efficient rule is strict liability with a defense of contributory negligence. Reverse the pattern and the efficient rule would be negligence.
There is one more rule we ought to consider. Suppose we make tank drivers strictly liable, but instead of paying damages to car drivers they pay a fine to the state. Now both tanks and cars have an incentive to take the efficient level of precautions, observable and unobservable alike. Tanks bear the full cost of accidents in the form of a fine; cars bear the full cost because the driver receives no damage payment if a tank crushes his car. And under this rule, the court no longer needs to either observe or evaluate precautions, so it requires less information than under a rule of either negligence or contributory negligence.
This elegant solution to the problem of dual causation has one unfortunate consequence: Accidents never get reported. After a tank crushes a car, the tank driver pays the car driver not to report the accident. That becomes even clearer if we shift to a more realistic picture: Everyone is driving a car and a collision injures both vehicles. The rule now takes the form of double liability. Each driver is responsible for his own costs and pays a fine to the state equal to the other driver's costs. That gives both drivers an efficient incentive to avoid accidents, provided that accidents are reported and fines collected, but it provides no reason for either driver to report, since by not reporting he avoids the fine.
What I have been discussing is a normal part of our legal system—but not of tort law. When we convert the damage payment into a fine, we also convert the offense from a tort to a crime and must shift the enforcement mechanism from private law suits to enforcement by the state. For more on that subject, stay tuned.
We have made our assumptions significantly more realistic and shown why some legal rules are better suited to some sorts of accidents. But we have still not explained why, under a negligence rule, anyone is ever found liable. The argument so far implies that actors will always choose the efficient level of observable precautions—and it is observable precautions that determine the court's judgement of negligence.
The next step is to recognize that both the court and the parties may make mistakes. The court might be wrong about what the efficient level of precautions is and so find someone negligent when he was not or not negligent when he was. Tortfeasors might underestimate the efficient level of precautions or gamble that the court would underestimate it. Once we introduce court error some actors will find it in their interest to take more than the efficient level of precaution, in order to make it less likely that a court will (mistakenly) find them negligent when they are not. Others will find it in their interest to take less than the efficient level, in the hope that if there is an accident the court may mistakenly find them non-negligent. All parties will find it in their interest, if an accident occurs and someone is sued, to spend resources trying to generate court errors in their favor. Including such possibilities substantially complicates the analysis. It also explains why people are sometimes found negligent and why there are so many jobs for lawyers—and law professors.
One reason courts may make mistakes is that they do not know enough about the characteristics of the actors whose precautions they are judging. The optimal speed for you to drive depends, among other things, on how fast your reflexes are and how well you deal with emergency situations in real time. Since courts do not have that sort of information, they generally define negligence not in terms of what precautions are cost effective for you but in terms of what precautions would be cost effective for an imaginary "reasonable man."
Suppose you are a professional race driver and believe, correctly, that the speed at which it is safe for you to drive is substantially faster than the speed which is safe for the average driver. Further assume there is no speed limit, but that if you have an accident how fast you were going at the time will be one of the things determining whether a court will hold you negligent. Finally assume that the court, following a reasonable man standard, considers any speed above sixty miles an hour to be negligent. How fast should you drive?
You have two options. You can stay down to sixty, thus guaranteeing that if you are in an accident you will not be found negligent. Or you can drive faster than that, at the cost of being found liable if there is an accident. The second option puts you, in effect, in a strict liability world, where you will, for the usual reasons, choose the efficient speed.
In deciding between those alternatives, you must trade off the benefit of being able to drive at what is, for you, the optimal speed against the cost of being liable for whatever accidents occur at that speed. If the cost happens to be larger than the benefit, it will be in your interest to drive inefficiently slowly. The same argument, applied to someone who is an unusually bad driver, implies that it may sometimes be in his interest to drive at sixty even though that is inefficiently fast for him, since at sixty he will not be liable for any accidents he causes.
Strict liability eliminates the problem. The court does not have to decide what the optimal level of precaution is; the driver, knowing he will bear all of the costs, makes that decision for himself, incorporating all of his private information about his own abilities.
It follows that if Posner's conjecture that the common law is efficient is correct, courts should tend to impose strict liability for torts where potential tortfeasors vary a great deal in characteristics relevant to the optimal level of precaution, since in such cases a negligence rule based on the characteristics of a reasonable man will over deter some and under deter others. More generally, we should see strict liability rules not merely where activity level is important (Posner's point) but where unobservable precautions are important—including precautions that are "unobservable" not because the court cannot see them but because the court cannot judge if they are worth taking. I do not know whether such a pattern actually exists or not.
You negligently injure me; I sue and win. How much do you owe me? The traditional answer is "enough to make whole the injury," to make me as well off as if the injury had not occurred. chapter 9 dealt with one problem raised by that rule, the problem of adequately compensating someone for being killed or injured. Here we will consider others.
Enough to make whole the injury appears, at first glance, to be exactly what is implied by the Pigouvian approach to externalities, putting aside complications of double causation and the like. I impose a cost on you, so I pay a sum equal to the damage done, giving me an incentive to properly include the external cost in my decisions.
That works if every tortfeasor is detected, sued, and loses. But in the real world, information is imperfect, litigation is costly, and courts less than omniscient. We can expect real-world tortfeasors to be found liable with a probability less, perhaps substantially less, than one. If so, tortfeasors will, on average, pay out as damages less than the full damage they do.
One solution might be to scale up the amount awarded to compensate for the less than unit probability of a successful suit. Criminal law sometimes works that way. When you pay a hundred dollar fine for littering, the reason is not that your one Coke can did a hundred dollars worth of damage but that a high fine is required to make up for the low chance that any single act of littering will be detected. Tort law, however, only requires that the victim who prevails in his suit be made whole. It calculates damages as if they were imposed with certainty. One reason why that may be a sensible rule will be discussed in chapter 18.
There is an exception to this rule that has become increasingly important in recent decades: Punitive damages. Ordinary damages are supposed to be based on damage done. Punitive damages may be more, sometimes much more.
Until very recently, almost all damage awards in civil suits were based on the rule "enough to make the victim whole." Punitive damages, also called exemplary damages, were rare, so rare that in the nineteenth century a respectable minority opinion (supported by, among others, the Supreme Court of New Hampshire and one of the chief legal commentators) held that no such thing existed.
Before trying to make sense of punitive damages, it is worth looking at some of the early cases. One of the earliest, Huckle v. Money, involved a conflict between the British Crown and its critics. An anonymous article appeared in an anti-government publication. Some King's Messengers—roughly speaking, the eighteenth century equivalent of Secret Service agents—illegally forced their way into the house of the suspected author, holding him prisoner while they searched his papers. He sued. They argued that they only owed him for the actual cost inflicted by their efforts, a few hours of his time plus the inconvenience of cleaning up the mess. They lost; the court awarded him punitive damages.
A slightly later example involved a case of deliberate insult. One landowner swore that he would shoot birds on another's property without the owner's permission—and did. The owner sued and was awarded damages for substantially more than the damage actually done.
In the 18th and 19th centuries, such cases were rare; more recently, awards of punitive damages have become much more common. The legal requirement is that the tort be deliberate or reckless. The law puts no limit on the damages that may be awarded.
I have so far collected six different explanations for this doctrine, four that are explicitly economic, two that are not. I start with the latter:
1. Punitive damages do not exist. What are misinterpreted as punitive damages are simply damages for hard to measure injuries. Having someone hold you prisoner in your house or shoot on your land without your permission humiliates you, lowers your reputation, and the court includes those costs in calculating damages. This explanation is consistent with the early cases I mentioned and was a respectable minority opinion in the nineteenth century. If it is correct then modern punitive damage law is a mistake.
2. Ordinary damages imply no moral judgement of the tortfeasor—he is merely required to make up for the injury he has done. Punitive damages serve to express public condemnation. But why express outrage by giving a windfall gain to the victim? Indeed, why use tort law at all, why not use criminal law instead to deal with outrageous behavior?
The next two are economic explanations, offered by Judge Posner and his frequent co-author, William Landes:
3. Punitive damages are a probability multiplier to compensate for the chance that a tortfeasor may never be sued or the victim may be unable to win his case If so, why is the criterion "deliberate or reckless?" One might argue that someone who commits a deliberate tort will also try to hide it, making detection less likely, but that hardly applies to a reckless tort such as riding down the street in the back of a pickup firing your rifle into the air. If punitive damages are a probability multiplier, they ought to be based on how likely the offense is to be lead to a successful tort suit, not on whether it was reckless or deliberate.
4. Punitive damages are a way of playing safe if damage is hard to measure but efficient offenses are unlikely. To understand this argument, we start by asking what is wrong with imposing too large a penalty. The answer is that it gives people an incentive to be too careful, deterring efficient offenses. So if we are confident that all offenses of a particular sort, all deliberate torts, for example, are inefficient, we don't have to worry about the risk that we are imposing too high a punishment. If we are not sure exactly how much damage the tort did, we might as well guess high.
This explanation works a little better than the previous one. Part of what makes us describe a tort as reckless is the failure of the tortfeasor to take even the simplest and most obvious precautions. That suggests that his behavior was clearly inefficient, so we need not worry about over deterring it. In the case of a deliberate tort, the tortfeasor is paying some cost to commit it, unlike an accident, where you must pay a cost to avoid it. So the optimal level might well be zero.
Then again, it might not; perhaps the benefit to the tortfeasor is sufficiently large to more than cover the cost to both tortfeasor and victim. Even then, permitting the tort is inefficient if the same benefit could be obtained at lower cost via a market transaction. In the case of an accidental tort, I do not know whom I am going to injure, which makes it hard to buy his permission in advance. In the case of a deliberate tort, that is less of a problem. I could have paid you for permission to shoot birds on your property if my doing so was worth more to me than my not doing it was to you. So one way of viewing punitive damages is as a tool for enforcing a property rule.
But what about a tort that is both deliberate and efficient: polluting when you know it is tortious, are willing to pay for the damage, but cannot arrange the transaction because there are too many victims? What about throwing sparks in the cases where doing so, and paying damages, is efficient? In any case where damages can be measured accurately, setting the damage payment equal to damage done will be sufficient to deter all inefficient torts, so additional offenses deterred by a higher damage payment must be efficient ones—which we don't want to deter.
The final two explanations for punitive damages are my own. I am not entirely satisfied with either, but I think I prefer them to the alternatives.
5.When we correctly take account of the cost of litigation in calculating efficient damages, it turns out that more deterrable torts should be punished more severely than less deterrable torts relative to the damage they do. Punitive damages are for a class of particularly deterrable torts.
In order to explain this, I must introduce a factor so far ignored: litigation cost. We would like to deter all inefficient torts, but not at any cost. If a particular offense does net damage, damage to victim minus gain to tortfeasor, of a hundred dollars, but deterring it costs a thousand dollars in litigation costs, we are better off not bothering.
The next chapter works through the logic of optimal punishment in some detail in the context of criminal law. Its conclusion is that the Pigouvian rule of imposing a punishment equal to the damage done in order to deter all and only inefficient acts is efficient only if doing so is costless. If detecting, litigating, and punishing offenses is costly, the rule for calculating the efficient level of damages becomes more complicated; the result depends both on the damage done and on how the number of offenses and the associated costs vary with the punishment.
If the supply of offenses is very inelastic, if it requires a large (and costly) increase in punishment to produce a modest decrease in the number of offenses, it is then only worth deterring offenses that are not merely inefficient but very inefficient. We do so by making the punishment less than the damage done. We thus deter only very inefficient torts, torts for which the benefit to the tortfeasor of committing them (or, equivalently, what it would cost the tortfeasor to take the precautions necessary to avoid committing them) is much less than the damage done.
Higher damage awards lead to more litigation cost per case. But if the supply of offenses is very elastic, higher damage awards also result in many fewer cases. In such a situation increasing the punishment, for example by imposing punitive damages, lowers litigation costs. It becomes efficient to set the damage payment higher than the actual damage done, deterring some efficient offenses in order to avoid the cost of litigating them.
The details of the analysis can be found in chapter 15—and, in a more mathematical form, on the web page. What is important here is the relation between this argument and punitive damages. Arguably, deliberate torts are particularly deterrable. In the case of an accident, liability simply adds one more cost to whatever costs the actor already bears from the accident. The actor cannot simply choose to avoid liability by not having the accident, he can only, at some cost in precautions, decrease its likelihood. Deliberate torts are deliberate, so the actor can deliberately not commit them. Reckless torts are a harder case, but presumably part of what makes them reckless is that the tortfeasor easily could have, and should have, avoided them.
If this argument is right, punitive damages are awarded for torts in relatively elastic supply, ordinary damages for torts in relatively inelastic supply, and doing so is at least roughly efficient. Ordinary damages undercompensate, because they contain no probability multiplier—and they should undercompensate, since the optimal punishment, allowing for the cost of imposing it, is less than damage done if the supply of offenses is sufficiently inelastic. Punitive damages overcompensate, and should, since the optimal punishment is more than the damage done if the supply of offenses is sufficiently elastic.
6. Punitive damages are designed to deter strategic torts.
Consider a simple example of a deliberate tort: beating up the guy who is trying to date the same girl you are. The reason you do it is not that you enjoy beating people up but that you want to deter people from going out with "your" girl.
If ordinary damage payments fully compensate the victim of a tort, this won't work. You beat me up, I sue you, the court awards me enough money to fully compensate me for being beaten up. I have suffered no net injury, so neither I nor anyone else has a good reason to avoid offending you in the future. In the words of a popular bumper sticker, "Hit me, I need the money."
But ordinary damage payments do not fully compensate the victim of a tort, for at least three reasons. One is that damages are not imposed with certainty, so ex ante the victim is only partly compensated. Another is that (under U.S., but not British, law) the victim normally must pay his own attorney's fees. Finally, ordinary damages often ignore costs such as pain and suffering that are hard to quantify. The result is that if I expect to get beaten up for trying to date your girl I may date someone else instead, even though I know I can sue you and may collect. You have successfully employed the same strategy as the railroad company, back in chapter 5, when it successfully deterred the farmers into switching to clover and saving it the cost of a spark arrester.
When you beat me up, was that an efficient tort? Probably not. Your gain may well have been greater than my loss, as suggested by the fact that you chose to do it even though you knew you might end up having to pay for both my damages and your lawyer. But I am not the only one who lost. By demonstrating your willingness to beat people up, you also imposed a cost on everyone else who might have wanted to do something you disapproved of, including all of the other men who might have wanted to date "your" girl—and on her. Hence in the case of a strategic tort, the damages observed by the court and used to calculate what it takes to make the victim whole may greatly underestimate the real damage done.
Beating someone up, in our legal system, is usually treated as a crime rather than a tort, making my example an artificial one designed to link with the analysis of the economics of bar room brawls back in chapter 8. For more realistic examples consider the early British cases, with a few added details.
I am a wealthy landholder in 18th century England. One cause and consequence of my wealth is that I control several seats in Parliament. The reason I control them is that my tenants vote the way I tell them to, my neighbors' tenants vote the way they tell them to, and, in matters political, my neighbors follow my lead.
Recently one of my neighbors has been acting uppity, threatening to run his own candidate for a seat where his tenants make up a sizable fraction of the voters. Something must be done; it is time to teach him a lesson.
I publicly announce, toward the end of a local ball, that my neighbor is a rascal and a coward, that nobody with any sense would have anything to do with him, and that I think so little of him that I plan to shoot birds on his property next weekend, whether he likes it or not. I do it. He sues for trespass. I hire a good lawyer. Five years, many hours of time, much frustration and many thousands of pounds later, he wins his suit. I pay him the cost of a day's rent of a grouse-shooting field—and nobody else challenges my control over local politics. I have successfully committed a strategic tort. In order to stop me, the court must award my victim punitive damages. Doing so not only raises the cost of my tactic, it also reduces its effectiveness, since what I am losing my neighbor is getting.
A similar analysis can be used to make sense of the earlier case. One objective of the King's messengers was to find out who wrote the anonymous article. Another was to make it unpleasant to be known as a critic of the Crown. If the court had awarded only ordinary damages, they would have succeeded.
Punitive damages are not, as it happens, available against our equivalent of the King's Messengers. Our government, under the doctrine of sovereign immunity, can only be sued with its permission. The Torts Claims Act, which waives sovereign immunity for the federal government in a variety of contexts, does not permit punitive damages. When a Texas court found that the United States Secret Service, in the course of its search and seizure of Steve Jackson Games, not only acted in direct violation of federal law but continued to do so for several months after the relevant facts had been pointed out by the victim's attorney, it awarded the victim only a very conservative estimate of ordinary damages.
One difference between tort law and criminal law is that under tort law, the offender pays a damage payment to the victim instead of a fine to the state. Why do it that way?
One answer was mentioned earlier—to give the victim an incentive to prosecute. This is a good reason, but not quite as good as it might seem. If the punishment is a fine going to the state, the victim still has an incentive to prosecute—and then offer to drop the charges in exchange for suitable compensation. Arguably this is what happened in English criminal law in the 18th century, under a system where criminal prosecution was not by police—there were no police—but by a private prosecutor, usually the victim.
If the fine goes to the state, any settlement greater than zero makes the victim better off than if he continues to press charges and any settlement less than the fine makes the tortfeasor better off. Having the fine go to the victim instead of the state reduces the bargaining range, since he will only accept a settlement that gives him at least what the court would have awarded, net of any legal costs avoided by settlement. Reducing the bargaining range is likely to reduce bargaining costs.
That argument would apply to any private prosecutor, not just the victim—and under 18th century English law, any Englishman could prosecute any crime. There are, however, three good reasons for giving the right to prosecute to the victim, as is done in modern tort law. One is that the victim is the person most likely to know that the tort occurred, hence the one in the best position to report it if given a suitable incentive to do so. A second is that the victim is likely to be an important witness, so giving him the right to the fine eliminates transactions that would otherwise be required between him and the prosecutor. A third reason is that the victim has an additional incentive to prosecute; prosecuting this offense may deter future offenses against him. A rule that gives the victim the fine combines the incentives in one person, eliminating costly transactions. These argument will appear again in chapter 18.
A more common, but less persuasive, justification for awarding tort damages to the victim is that doing so compensates him for his loss. As I have already suggested, and will discuss at greater length shortly, tort law makes a poor form of insurance. And, as we saw in the case of the car and the tank, compensating the "victim" can have perverse incentive effects, since the compensation reduces his incentive to take precautions to prevent the accident.
One exception is the case of the strategic tort where, as we saw back in chapter 5, compensation reduces the effectiveness of the tortfeasor's threat. When the fine for starting fires was converted into a damage payment, in a world without complications such as litigation costs, the railroad's strategic position collapsed and it installed the (efficient) spark arrester.
This suggests the importance of distinguishing between an anonymous tort, such as an auto accident, and a tort such as deliberate trespass where the tortfeasor knows the identity of the victim. With an anonymous tort, strategic behavior is not an option. I could drive dangerously in the hope that others would respond by being very careful but, since I am only one driver out of many, the effect on their behavior would be negligible. The closest I can think of to a real-world example of that tactic is a friend who never took dents out of her car, on the theory that other drivers would conclude that since she wasn't careful they should be.
Non-anonymous torts create not only the risk of strategic behavior but also an opportunity for bargaining. In interactions between neighbors, we may not need tort law to move us to the right outcome. As we move toward less anonymous torts, we also move toward a world where property rules may make more sense than liability rules.
Who Pays for Coke Grenades?
Product liability is a legal issue located on the border between tort and contract that has been responsible for a great deal of litigation in recent decades. The fundamental question is who is to be held responsible when something unexpected goes wrong in the use of a product, when a Coke bottle explodes or a lawn mower injures the person pushing it. Do we follow a rule of caveat emptor (Latin for "let the buyer beware"), under which the buyer takes the good as he finds it, complete with any defects, or a rule of caveat venditor, under which if anything goes wrong the seller is liable? Or are there, perhaps, better alternatives?
This way of putting it makes the distinction too sharp. If your wife takes out your gun and shoots you, something has gone wrong; that isn't what you bought the gun for. But even under caveat venditor the manufacturer of the gun is unlikely to be held liable. Product liability law deals with the subset of "things going wrong" associated with defective products. Even then, the category is large and vague. Lawn mowers are supposed to cut things; when they cut the wrong things it is not always clear whether the defect is in the machine or the man pushing it.
The obvious basis for choosing a liability rule is its effect on the incentives of people to do things that prevent accidents, to make Coke bottles that don't blow up or to design safer mowers. If Coke is liable, that gives them an incentive to maintain high levels of quality control in order to make sure bottles rarely blow up, which is an argument for caveat venditor.
The problem is more complicated than that for two reasons. The first is that users can also prevent accidents—by not shaking Coke bottles in hot weather, by wearing shoes while mowing the lawn, and by having a couple of beers after mowing the lawn, not before. When I looked through some real cases involving exploding Coke bottles—all of them, of course, dating from before Coke switched from glass to plastic—I came across one where the bottle had been out of refrigeration for thirty-six hours, in Fresno, California, before it exploded. To the extent that liability law insures the user against accidents, it reduces his incentive to prevent them, which is an argument for caveat emptor. We are back with the problem of moral hazard discussed back in chapter 6, and the same answer: Put the incentive wherever it will do the most good.
A second complication is that even if Coke is not liable it may still have a good reason to make sure its bottles don't explode: reputation. If consumers have a reasonably accurate idea of how risky their bottles are, poor quality control will be punished by lost sales. In a world of fully informed consumers, caveat emptor results in efficient precautions by both sides, by Coke to maintain its reputation and by users because they will bear the cost of any accidents. The same would be true of Caveat Venditor if Coke were fully informed about what customers were going to do with their bottles and could adjust their prices accordingly, but it is hard to be fully informed about things that have not yet happened.
In order for reputation to provide Coke with the proper incentives, consumers must be well informed not merely about the risk of bottles blowing up but about the risk of Coke bottles blowing up, so that it is Coke, not the whole industry, that gets reputational benefits from Coke's quality control. How plausible an assumption that is will vary from one product to another. If consumers are entirely uninformed about product quality, if, for instance, they are buying medical drugs whose side effects may not show up for many years, reputation might give little or no incentive to the seller. If so, that is an argument for making the seller liable.
One way of not being blinded by an exploding Coke bottle is to take precautions, to keep it cold and, if for some reason you must handle warm bottles, wear leather gloves and glasses. Another way is not to buy it. If consumers are fully informed about risks, one result is that producers have the right incentive. Another is that consumers can make the correct decisions about what to buy. If they are ignorant of an important dimension of quality, such as the risk of defects, that no longer works. A consumer who seriously underestimates the risk of defects may buy something worth less to him than the price, while a consumer who overestimates the risk may make the opposite mistake. We are again back in chapter 6, this time with the problem of adverse selection.
If the seller knows the risk and the buyer does not, the solution is for the seller to insure the product, which is what caveat venditor makes him do. Coke is selling a bottle of Coke with an insurance policy attached and, knowing the risk, knows what both cost. The buyer is buying a bottle and an insurance policy and, because the policy will cover his losses if the bottle proves defective, he does not have to know how great that risk is in order to decide if the package is worth its price. The same argument implies that if it is the buyer who best knows the risk—knows, for example, that he plans to use his new car for drag racing, not commuting—the seller should not be liable.
These arguments give us a fairly straightforward result, although one not always easy to apply in practice. If buyers have accurate information about how risky each firm's products are, the right rule is caveat emptor. If buyers do not have such information but sellers do and sellers are in the best position to prevent accidents, the best rule is caveat venditor. In more complicated situations, such as ones where sellers know the risk but buyers can best control it, we must either figure out which consideration is more important or go with some intermediate solution, such as making the seller liable for only part of the risk.
That compromise, the liability law equivalent of coinsurance, may be the one implicit in much of tort law. Traditionally damage awards were based on a narrow definition of costs; although that has changed somewhat, they are still low enough on average so that (bumper stickers to the contrary) very few people really want someone to run into them so that they can sue him. Hence even with tort rules that make someone else liable, most of us still have at least some incentive to avoid being victims.
So far I have neglected one argument often used to justify making producers liable—the fact that they are usually in a better position than consumers to spread the risk. I may, if I am very unlucky, have one Coke bottle blow up on me in the course of my life. Coca-Cola sells many millions of bottles a year. Their loss, if they are liable for such accidents, like the loss born by an insurance company insuring many houses, is reasonably predictable.
The reason I have neglected the issue of risk spreading is that although consumers want protection against risk and tort law can provide it, there is a better alternative. Insurance policies have two important advantages over liability law as a mechanism for protecting people from risk. The first is that liability law is too selective. If I hurt my hand because a Coke bottle explodes, liability law may reimburse me for the loss, but if I slip on my own icy front stairs and hurt my hand, there is nobody to sue but myself. What I want protection against is not risk due to defective products, or risk due to other people's negligence, but risk in general. That is an argument for using insurance policies to provide protection against accidents and tort law to give people the proper incentive to prevent them.
A second advantage of insurance over tort law is that an insurance company wants a reputation for being generous in paying out benefits—not so generous as to encourage fraudulent claims, but generous enough so that people want to buy insurance from it. A litigator wants the opposite reputation. The harder he makes things for the plaintiff in this suit, the less likely it is that the next plaintiff will try his luck. The adversarial relation between plaintiff and defendant is likely to result in much higher costs than the (relatively) cooperative relation between an insurance company and a customer filing a claim. On average, of every dollar paid out by tort defendants in damages and attorney fees, tort plaintiffs, after paying their attorneys, receive about fifty cents.
This suggests a further consideration relevant to the choice of a liability rule: litigation costs. Under caveat emptor nobody is liable, nobody is sued, and nobody has to pay any lawyers. Under caveat venditor, the injured consumer must sue, or at least threaten to sue, in order to collect damages. That is a general argument, here and elsewhere in the law, for letting losses remain where they fall, to be weighed against the gains we sometimes get by shifting them somewhere else.
We have been talking in terms of two simple alternatives: caveat emptor and caveat venditor. Courts might distinguish more finely, imposing caveat emptor for some sorts of risks and caveat venditor for others, along the lines we have just been sketching out. Or they might apply negligence principles, holding Coke liable only if it could be shown that it had failed to take appropriate precautions (negligence liability) or holding it liable unless it could show the user had failed to take appropriate precautions (strict liability with contributory negligence).
So far I have assumed that the liability rule, whatever it is, is to be set by courts or perhaps legislatures. There is another alternative: contract. Every time an auto company provides a 30,000 mile guarantee, it is imposing a rule of caveat venditor on itself for the particular sorts of defects covered by the guarantee.
Under a legal regime of freedom of contract, courts set default rules but leave parties free to change them. If the court imposes caveat emptor and sellers believe that the shift to caveat venditor would be worth more to buyers than it costs them, then sellers offer guarantees and raise their price accordingly. If the court imposes caveat venditor and buyers value the resulting legal protection at less than it costs sellers to provide it, then buyers agree to sign waivers of liability, converting the rule to caveat emptor in exchange for an appropriate reduction in price. The economic analysis of what liability rule is most appropriate to any given situation is still useful under freedom of contract, both to courts choosing default rules and to parties deciding whether or not it is worth contracting around them. But the final decision is made by the parties, not the court.
The argument for permitting freedom of contract depends on consumers and sellers being rational and having some information, but it does not require the sort of full information that is needed to make caveat emptor automatically the right rule. Consumers do not have to know what the risk of Coke bottles blowing up is. They only have to know enough to decide whether, on average, they are better off with or without a guarantee, whether, for one part or another of their economic dealings, they prefer caveat emptor or caveat venditor, given the price of each.
The discussion of freedom of contract points out an important distinction that can easily be missed: between deciding what the right legal rules are and deciding what mechanism will be most likely to produce them. The analysis of the choice between caveat emptor and caveat venditor focussed on the first question: When is one rule or the other superior? The analysis of the choice between freedom of contract and mandatory liability rules set by courts or legislatures focussed on the second question—which mechanism is most likely to generate the superior rule.
"Now, even if you do get information that you did not expect, it means nothing if this new, surprising information does not change your strategy."
Fighting Fuzzy Thinking in Poker, Gaming & Life by David Sklansky
Sklansky is discussing poker—specifically the question of when it is worth paying for information about another player's strategy by making bets that reveal information but, on average, lose money—but his analysis applies more widely, as the title of his book suggests. Providing and evaluating information is costly for both producer and consumer; whether it is worth doing depends on how much the information is worth. If a bottle of Coke is worth a dollar to me and sells for fifty cents, learning that it carries with it a risk of explosion that I would evaluate as a two cent cost has no effect on my behavior. The bottle is still worth more than it costs, so I still buy it. If the bottle was worth 51 cents to me the information does affect my behavior (value 51-2=49 cents, cost 50 cents, so I don't buy it), but since the bottle is actually worth only a penny less than it costs, buying it would be only a very minor mistake.
Generalizing the argument, information that has only a small effect on how much a good is worth to a consumer is unlikely to affect whether he buys it; if it does push him over the edge into not buying, the result is only a small savings. So adverse selection is only important if the missing information has a large effect on the value of the good—for example, whether a car is a lemon or a creampuff. It follows that failure to provide information should not be actionable unless the information has a substantial effect on the value of the good to the consumer. Failure to recognize that principle gives us a world where the instruction manual for every product starts with fifteen pages of warnings aimed at the jury in some future damage suit, all of which the rational consumer ignores.
[link to mathematics]
To see how this requirement plays out in real-world law, consider the polio vaccination cases. About one person in a million who was inoculated with the Sabin live virus vaccine got polio as a result. Since there was no known way of immunizing against polio without that risk, courts held that the vaccine was "inherently dangerous," hence not a defective product.
While there is no way of making an inherently dangerous product safe, one can still avoid the danger by not using the product. Hence the question arose of whether the producers of the vaccine were negligent, not for producing the vaccine but for failing adequately to warn those who used it. The producers provided warning information to physicians and others who dispensed the vaccine but did not take precautions to make sure that everyone who got the vaccine also got an individualized warning.
Whether that was negligent, at least in the economist's sense, depends on whether the warning would have made a difference, whether there was a significant chance that someone who was informed of the risk would conclude as a result that it was not in his interest to be vaccinated. In Davis v Wyeth Laboratories, Inc., the court accepted that argument and attempted to calculate the relevant costs and benefits.
"The Surgeon General's report ... predicted that for the 1962 season only .9 persons over 20 years of age out of a million would contract polio from natural sources. ... Thus appellant's risk of contracting the disease without immunization was about as great (or small) as his risk of contracting it from the vaccine. Under these circumstances we cannot agree ... that the choice to take the vaccine was clear."
If the mistake is not obvious, read it again. The court is comparing the risk to Davis from being immunized—about one chance in a million of contracting polio—with the risk of not being immunized. To estimate the latter, it uses the fraction of the adult population that could be expected to contract polio from natural sources in one year. But immunization lasts for life, so the relevant comparison is to his lifetime chance of ever contracting polio, perhaps with a weighting factor to take account of the fact that the later he contracts it the smaller the fraction of his life affected.
If we assume that the risk was constant over time and measure the cost of getting polio by number of years of normal life lost, the court's calculation was off by about a factor of twenty-five, making the benefit of the vaccine more than twenty times the cost. The mistake is one that a bright high school student should have caught. If judges, like pharmaceutical companies, were legally liable for the consequences of their negligence, the judge who wrote that opinion and those who joined in it would have owed Wyeth a very large amount of money.
"The plaintiff was about 14 years of age, and the defendant about 11 years of age. On the 20th day of February, 1889, they were sitting opposite to each other across an aisle in the high school of the village of Waukesha. The defendant reached across the aisle with his foot, and hit with his toe the shin of the right leg of the plaintiff. The touch was slight. ... In a few moments he felt a violent pain in that place, which caused him to cry out loudly. ... He will never recover the use of his limb."
(Vosburg v Putney, 80 Wis. 523, 50 N.W. 403 (1891))
So begins one of the odder cases of the common law. For reasons nobody seems to have understood, something serious went wrong with Vosburg's leg after it was lightly kicked by Putney. The court concluded that although there was no way Putney could have foreseen the result, nonetheless his kick was tortious and he, which is to say his parents, legally responsible for the consequences. The result is a legal rule that still survives: A tortfeasor takes his victim as he finds him. Even if the victim happens to be unusually vulnerable, even if he turns out to have an "eggshell skull" that can be broken by a light blow, the tortfeasor is still liable for the actual costs due to his tort.
On the face of it, making me pay more when I do more damage makes sense only as long as the loss is foreseeable. If I have no way of telling who does or does not have an eggshell skull, I cannot adjust my precautions to respond to the incentive of greater liability. Yet there is an efficiency argument for the rule, based not on the (unforeseeable) damage in a single case but on the average damage across many cases.
Potential tort victims vary in how vulnerable they are. If the legal rule limited damages to what could reasonably be foreseen, then tortfeasors with the bad luck to injure particularly fragile victims would pay average damages, tortfeasors with the good luck to injure particularly sturdy victims would pay below average damages, and the average tortfeasor would pay less than the average damage done by his tort. By instead requiring tortfeasors to take their victims as they find them we give potential tortfeasors the right average incentive to take precautions.
If you find this argument for the efficiency of the rule persuasive, think back to the Himalayan photographer's lost film in chapter 12. The logic of that problem and the logic of this are strikingly similar. In each case, the party at fault has less information about the amount at risk than the victim. While Vosburg could not have known just how fragile his leg was, he knew more about the subject than Putney since he knew, and Putney did not, that he had a previous injury to that leg which was still healing. In the more general case, the fragile victim is quite likely to know he is fragile, certainly more likely to know than is the tortfeasor.
Precisely the same argument I have just made to defend the rule implied by Vosburg v Putney could have been used to attack the rule of Hadley v Baxendale under which Walgreen's escapes most of the liability for losing the photographer's film. That rule implies that negligent parties, film developers in the example, will pay below average damages for losing especially worthless films, average damages for losing especially valuable films, hence on average will pay less than the damage they do. Precisely the same argument I used to defend the rule of Hadley v Baxendale, that it gave the party with more knowledge an incentive to take the appropriate precautions, could be used to attack the rule of Vosburg v Putney. If you have an eggshell skull, it is easier to deal with the problem by having you wear a protective helmet than by having everyone in the world treat everyone else like an egg.
This matched pair of legal rules, along with the accompanying justifications (in both cases borrowed from Judge Posner), nicely illustrate the risks of evaluating the efficiency of the legal system after the fact. Once we know what the rule is, we can usually find an argument, even a plausible argument, for why it is efficient. Often enough, there is a matching argument that we could have used with equal effect if the rule had come out the other way. For a second example of the same point, consider ...
You tortiously injure someone and the court awards him damages. What form should the compensation take? Should you owe your victim each year for that year's medical costs and lost earnings, or should the court award a lump sum based on its estimate of the present value of the total stream of costs resulting from the injury?
One argument against annual payments is that they require continued involvement by the court. Another is that they distort the victim's incentives. If he succeeds in getting a job, thus demonstrating that his injury is not, after all, totally incapacitating, his damage payment will be reduced accordingly. If the court awards a lump sum instead, the victim is left free to do whatever he can to reduce the cost due to his injury without being penalized for it. This is the policy usually followed by modern courts.
There are two problems with it. The first is that it requires the court to estimate costs in advance. Victims who recover faster than the court expected are overcompensated, victims who recover more slowly are undercompensated, and both kinds have an incentive to spend time and effort persuading the court to make a high estimate of the unknown future costs.
The second problem may be illustrated by a simple story.
A tort plaintiff succeeded in collecting a large damage judgement. The defendant's attorney, confident that the claimed injury was bogus, went over to the defendant after the trial and warned him that if he was ever seen out of his wheelchair he would be back in court on a charge of fraud.
The defendant replied that to save the lawyer the cost of having him followed, he would be happy to describe his travel plans. He reached into his pocket and drew out an airline ticket—to Lourdes, the site of a Catholic shrine famous for miracles.
Lump sum payments do indeed reduce the cost of being injured. They also reduce the cost of pretending to be injured, since you can end the pretence as soon as you cash the check.
"Traffic on the highways ... cannot be conducted without exposing those whose persons or property are near it to some inevitable risk; and that being so, those who go on the highway, or have their property adjacent to it, may well be held to do so subject to their taking upon themselves the risk of injury ... and persons who ... pass near the warehouses where goods are being raised or lowered, certainly do so subject to the inevitable risk of accident. In neither case, therefore, can they recover without proof of want or care or skill occasioning the accident; ... ."
Blackburn, J. (Fletcher v. Rylands L.R. 1 Ex. 265 (1866).
The quote is from the most famous of the early cases on the choice between negligence and strict liability. Is Judge Blackburn correct in arguing that negligence is more appropriate than strict liability for the particular cases he mentions?
Hint: Whose activity level will be (efficiently) reduced if we shift from strict liability, in which victims are always compensated, to negligence, in which they are compensated only if the tortfeasor was negligent?
My Academic Pages
My Home Page
Email to me