Economic Analysis of Accident Law

by Steven Shavell,

Harvard University Press, Cambridge, Massachusets; London, England 1987


The first issue of the Journal of Law and Economics  was published in 1958. In the ensuing thirty years, the economic analysis of law has become increasingly important in both economic and legal theory. For economists, it provides not only a chance to apply our tools to a new set of problems but also a unique opportunity to improve those tools. Concepts such as property, contract, or causation look much simpler in an economics textbook than in a courtroom. The attempt to understand legal doctrines in economic terms forces economists to think through more carefully some of the fundamental concepts underlying economic theory. The most important example of this is the set of ideas originating with Ronald Coase's famous article on externalities.1

The economic analysis of law has been even more important, and more controversial, in the law schools. In the hands of writers such as Judge Richard Posner, it claims to provide the unifying theory that existing legal analysis lacks, making it possible to understand a wide variety of legal issues in terms of one central set of interrelated concepts.

Steven Shavell is one of the leading practitioners of economic analysis of law; his book is a clear and careful discussion of what we currently know about the economic analysis of accident law, much of it based on the author's own work. The result is in some ways less interesting than a broader work such as Posner and Landes's recent book on torts,2 but within its limited area it is very much more complete. One can be fairly certain that where something important is missing from Shavell's description of accident law, the fault is with the theory not the description. The book thus provides a good snapshot of the strengths and weaknesses of economic analysis of law as it now exists.

The economic analysis of accidents starts with the observation that they are not entirely accidental. I do not choose to run my automobile into a pedestrian, but I do choose what kind of a car I drive, how often and at what speed I drive it, and how often to have my brakes checked. These decisions and many more affect the probability that I will be in an accident and thus the cost my driving imposes on other people. It then seems natural to ask, as Shavell does, what set of legal rules will lead me to make such decisions in the most nearly efficient manner. The answer to that question is most of the book.

The simplest approach to generating efficient behavior is direct regulation. Let the law state how cars must be built, how many miles people may drive and at what speed, how often their brakes must be checked. This solution runs into problems familiar to any economist. In order to set efficient values for all of the variables, the legislature would require detailed information about individual tastes and abilities that it has no way of getting; if it had the information, using it to calculate optimal behavior would involve daunting mathematical problems. Even if the legislature could calculate and enforce optimal behavior, there is no obvious reason why it would be in the interest of the legislators to do so. Where the state does directly regulate driving, the regulation is limited to only a few of the more easily observed variables such as speed, and almost entirely ignores relevant differences among individual drivers.

Similar problems arise if we try to control accidents by Pigouvian taxes on the behavior that causes them. The probability that I will be involved in an accident depends on many choices, only some of which are observable. If by driving an extra mile I impose an expected cost of five cents on potential accident victims, then a tax of five cents a mile will induce me to do the efficient amount of driving. But what tax will prevent me from paying a more than optimal amount of attention to the radio and a less than optimal amount to the road?

The solution to this problem is to charge by results; if I cause an accident I must pay the cost. Externalities are then internalized; I have an incentive to engage in an efficient level of accident provision on every margin. In our legal system, such costs are imposed mainly through civil suits for damages.

This produces new problems. Driving becomes a lottery with large negative prizes. If drivers are risk averse they have an incentive to insure--and, by doing so, reduce their incentive to take precautions. Many drivers will be judgement proof--unable to pay the cost of a major accident. That can be solved by requiring drivers to be insured--but again with negative effects on incentives.

There is another and deeper problem. As Coase pointed out, the Pigouvian approach contains a fundamental error; it treats external costs as if they were the result of only one party's action. The probability that I will run you down depends on your decisions as well as mine, on how carefully you cross the street as well as on how fast I drive. Ideally both of us should take all cost justified precautions. But if the driver is responsible for all costs of the accident, the pedestrian has no incentive to take precautions.

There are at least three solutions to this problem, two of which are discussed at length by Shavell. One is a rule of negligence. The driver is responsible for the cost of the accident if and only if he is negligent, with negligence defined as failure to take all cost-justified precautions. Under that rule, drivers find it in their interest to take the efficient level of precautions; pedestrians, knowing that drivers will not be negligent and thus not be liable, find it in their interest to take an efficient level of precautions as well.

A second solution is a rule of strict liability with a defense of contributory negligence. A pedestrian who has taken less than the efficient level of precautions cannot collect damages from the driver. This rule again leads to an efficient level of precaution by both parties.

Both of these suffer from the same problem as the first and simplest solution to accidents--direct regulation. In order to apply either negligence or contributory negligence, the court must be able to calculate the efficient level of precaution and observe whether it is taken. They are an improvement only in restricting the problem to cases where an accident actually occurs.

Shavell discusses this difficulty at some length, mainly in the context of "level of activity." The court can judge whether the driver was negligent in how he drove but not whether he was negligent in how much he drove--whether his marginal trip was worth taking, given the expected accident costs it produced. Under a negligence rule, drivers drive too much since, having taken the efficient level of precaution, they are no longer liable for damages. Under a rule of strict liability, with or without contributory negligence, there is an efficient amount of driving but an inefficiently large amount of walking, since pedestrians are reimbursed by drivers for the cost of accidents.

The solution to this problem carries us outside of the context of civil damages. If the driver pays damages but the pedestrian does not receive them--more generally, if each party to the accident must separately pay its full cost--then each has the efficient incentive to avoid the accident. The damage award has been converted into a fine.

This solution generates new problems. If the injured party receives no damages he has no incentive to sue, so the accident never gets reported. In converting damages into fines we have gone from a private to a public system of law, and must provide some public mechanism to report damages and institute cases.

This is one example of a problem that Shavell considers at some length. In a civil system, the damage award provides both the incentive to the victim to sue and the incentive to the potential offender not to commit the offense. There is no good reason why the optimal incentives for the two purposes should be equal. This point has been discussed at some length in the literature on private vs public enforcement of law.3

A similar problem arises when Shavell drops his initial assumption of risk aversion and considers damage payments as providing both deterrence and risk spreading; here again, the two objectives imply different levels of damages. If the optimal damages for deterrence are lower than the optimal damages for risk spreading, the potential victim can solve the problem by buying insurance. If they are higher, Shavell concludes that there is no way to get the efficient outcome with a system of damages--one must introduce fines as well.

This is, I think, one of the few mistakes in the book. If damage payments overcompensate victims from the standpoint of insurance--if the marginal utility of a dollar is lower to the victim after collecting damages than it would be if the accident had not occurred--potential victims will find that it pays them to sell  insurance on themselves.4 There is no obvious reason, in either the real world or Shavell's models, why insurance companies should be less willing to buy than to sell. The fact that we do not observe such contracts may indicate either that damage awards are not inefficiently high or that our legal system (inefficiently) fails to treat damage claims as fully marketable property.

By now, the reader should have some feeling for the complexity of the problem Shavell is dealing with, and why it takes him some three hundred pages to work part way through what seems, at first, a simple exercise in the control of externalities.

Within the limitations of our present knowledge of the economics of accident law, this is a very satisfactory book. It covers everything and it covers it carefully and in great detail. For those in the field, it provides a summary that should be a useful reference. For students, economists, or legal theorists wishing to learn the economic theory of accident law, it is a complete, if somewhat dry, introduction.

By presenting the material as verbal chapters with mathematical appendices, the author has produced a book that should be comprehensible to the intelligent non-economist, with no serious loss of rigour. Only in a few places does the verbal description misrepresent the mathematical proof, as when a proof that something is true if a variable is sufficiently small is converted into the claim that it will "tend to" be true--but might be false if the variable is large enough.

Such cases are the exception, not the rule. The only serious faults in the book are those that reflect limitations in the existing theory. One way of seeing them is to extend that theory a little, in ways that are both consistent and implausible.

Shavell demonstrates that either negligence or strict liability with contributory negligence leads to an efficient level of care. Under his assumptions, these are not the only rules that do so. Consider, as a simple alternative, replacing all accident law with one simple rule: anyone involved in an accident who has not chosen the efficient level of care will be shot. Everyone chooses the efficient level, noone is shot, the objective is achieved costlessly, and we avoid the nuisance of actually having to calculate and pay out damage payments.

If you find that unconvincing, I have another rule to suggest. Whenever there is an accident, let its full cost be imposed separately on everyone who could conceivably have anything to do with it--say everyone within ten miles of the accident. The resulting revenue is distributed among everyone farther than ten miles from the accident. The result is that everyone has an incentive to take all cost-justified precautions to prevent the accident--which is the efficient result. Of course, most of the people paying the fines will have nothing to do with the accident--but since they have nothing to do with the accident, a fine conditional on the accident happening has no effect on their behavior and is thus costless.

What is wrong with both of these suggestions is that they ignore the costs and errors of the legal system. They would work in the model that underlies most of Shavell's book, where courts make and enforce decisions costlessly and without error. When we drop those assumptions, the efficiency of my proposals goes with them. So does most of the existing theory of accident law.

Shavell is aware of this problem and devotes considerable space to trying to analyze problems associated with court error. Indeed, he is arguably more aware of it than his most immediate competitors. Landes and Posner, in their book, spend a chapter defending the present state of product liability law, including its unwillingness to accept explicit contractual limitations on liability. Their essential argument is that since, in a world of costless and error-free courts, no rational customer would agree to such limitations, we may safely dismiss those who in the real world do agree to them as irrational.

The problem is not that Shavell and others are unaware of the problems of an imperfect system of enforcement but that they have not yet integrated those problems into the underlying structure of their model. Now that the implications of the existing models have been fairly well worked out, it is time to do so.

One way might be to start from the observation that we have two mechanisms for measuring costs and values in order to determine what behavior is efficient--revealed preference on the market and estimates by a court. The first mechanism is clearly superior to the second. Unfortunately, there seems to be no way of solving the problems of accident law using it alone. Alternative solutions involve different ways of dividing the problem between the two mechanisms, and must somehow be evaluated on that basis.

The analysis should also include the economics of the court system itself. Whether to litigate and how much to spend litigating are economic decisions that affect the costs of running the system and the quality of its decisions. That is true both under a private system, where how much to spend is decided by plaintiff and defendant, and under a public system, where the plaintiff is replaced by a public prosecutor. The analysis should also, if possible, include the incentives acting on judges and legislators; in the ideal model, everything is endogenous. We cannot get that far, but we should be able to get a good deal farther than we have.

The initial steps in this direction have been taken--by Shavell, in his attempts to analyze the effects of cost and error, by Posner in his arguments for why common law judges find in their interest to generate efficient law, and by others. So far, they are ad hoc additions to a structure whose underlying assumption is still a perfect court system functioning costlessly. In this regard, law and economics resembles the economics of public policy just before it was revolutionized by public choice and rational expectations. We have gotten this far in thirty years. The next step may take another thirty.



Ronald Coase, The Problem of Social Cost, 3 Journal of Law and Economics  1 (1960).

David Friedman, "Efficient Institutions for the Private Enforcement of Law." Journal of Legal Studies , June 1984.

David Friedman, "What is Fair Compensation for Death or Injury?" International Review of Law and Economics, 2, 1982.

William M. Landes and Richard A. Posner, The Economic Structure of Tort Law, Harvard University Press, Cambridge 1987.

William M. Landes and Richard A. Posner, The Private Enforcement of Law, 4 J. Legal Studies  1 (1975).

Back to the list of articles.

Back to my home page.