Coase's work provides one possible approach to the problem of constructing efficient legal rules. To make that approach clearer, I spend most of this chapter using Coase's ideas to analyze how legal rights ought to be defined in the context of a particular real-world problem, one that Pigou used to explain the problem of externalities and that Coase in turn used to show why Pigou was wrong.
Having seen how, in principle, we would find the efficient legal rules for that particular situation, we then generalize the argument to the question of how to decide whether rights ought to be protected by property rules ("steal a car, go to jail"), liability rules ("dent my car, pay to fix it"), or fines.
Railroad trains in the nineteenth century threw sparks; the sparks sometimes started fires in adjacent fields. Railroad companies could reduce the problem either by running fewer trains or by installing a spark arrester, an apparatus on the train's smokestack designed to keep sparks from getting out. Farmers could reduce the problem by leaving land near the railroad track bare or planting some crop unlikely to catch fire.
For purposes of my example I am going to simplify the situation a little. Farmers have only two alternatives: wheat, which burns, and clover, which does not. I assume that growing clover is less profitable than growing wheat, since otherwise everyone grows clover and the problem disappears. The railroad also has only two alternatives: to install or not install a spark arrester. All costs in the example can be thought of as annual costs: the lost revenue per year from growing clover instead of wheat, the lost revenue per year from occasional fires in wheat fields, the annual cost of spark arresters on railroad engines. Finally, I assume that the area we are considering contains one railroad company and a hundred farmers and that all farmers are equally at risk from sparks.
We consider four different legal rules:
1. Property right by the railroad. The railroad is free to throw sparks if it wants to.
2. Property right by the farmers. The railroad may only throw sparks if it has permission to do so from all the farmers; any one farmer can go to court and enjoin the railroad from throwing sparks.
3. Liability right by the farmers. The railroad is free to throw sparks but must compensate the farmers for any damage that results.
4. Liability right by the railroad. Any farmer may enjoin the railroad from throwing sparks but must then compensate the railroad for the cost of having to put on a spark arrester.
This way of stating the rules partly obscures the symmetry of the alternatives. In particular, alternative 4 (the legal term for which is "incomplete privilege") seems less natural than the other three. The symmetry is clearer if we think of each rule as answering two questions:
Who decides whether the railroad throws sparks?Who bears the costs implied by that decision?
There are two possible answers to each question: Railroad or farmers. Combining the possible answers gives us our four rules:
1. Railroad decides, farmers bear
2. Farmers decide, railroad bears
3. Railroad decides, railroad bears
4. Farmers decide, farmers bear
In addition to four rules, we also have four possible outcomes:
A. Sparks+Wheat=Fires: Railroad throws sparks, farmers grow wheat, the result is occasional fires.B. Sparks+Clover=No Fires: Railroad throws sparks, farmers grow clover, no fires because there is nothing to burn.
C. Spark Arrester+Wheat=No Fires: Railroad installs the spark arrester, farmers grow wheat, there are no fires because there are no sparks to start them.
D. Spark Arrester+Clover=No Fires: Railroad installs the spark arrester, farmers grow clover, no fires.
Each legal rule can lead to any of several different outcomes, either directly or via a transaction between railroad and farmer, as shown in figure 5.1. Each outcome has a cost, as shown in table 1. The objective of the game is to pick the right starting point, the efficient legal rule. As will become clear, which one that is depends on where you want to go or, more precisely, on how likely each outcome is to be the efficient one.
We begin with rule 1: The railroad has a right to throw sparks and is liable to nobody for the consequences. What happens?
First suppose that the damage done by fires in wheat fields is $400, switching to clover costs $800 ($8 per farmer), and a spark arrester costs $1,000 (column a of the table). The cost of the fires is less than the cost of either way of preventing them, so the efficient outcome is A: Sparks+Wheat. That outcome occurs with no transactions necessary between the parties and costs $400. The railroad throws sparks; the farmers reject the possibility of switching to clover because it would cost them more than the savings from preventing fires. The farmers could try to pay the railroad to put on a spark arrester, but the most they would offer would be $400, and the cost of the spark arrester is $1,000 so there would be no point to making the offer. We have the efficient outcome, and we have gotten there with no transactions among the parties and no transaction costs, as indicated by the solid arrow linking rule 1 to outcome A.
Next suppose we reverse the costs of clover and fires; switching to clover now costs $400, and fires do $800 worth of damage (column b). The railroad throws sparks. The farmers consider their options and switch to clover, since the savings from eliminating fires make it worth doing. This time outcome B, Sparks+Clover, is the efficient one, and we have again gotten there with no transactions between the parties and no transaction cost, shown by another solid arrow.
Finally, suppose the cost of a spark arrester drops to $200 (column c). The railroad is still legally free to throw sparks if it wants to. But now it is in the interest of the farmers to offer to buy the railroad a spark arrester, since doing so will cost them less than switching to clover, which is their next best alternative.
It is in their interest, but it may not happen. The reason is the public good problem mentioned earlier. As long as enough money is contributed to pay for the spark arrester, farmers who choose not to contribute get a free ride. With a hundred farmers, one farmer's contribution is unlikely to make the difference between success and failure. Many farmers may refuse to pay, figuring that if everyone else refuses, their contribution will not be enough to make a difference, and if enough other farmers pay, they can get a free ride: receive the benefit of a spark-free world without paying for it.
There are a variety of ways in which the farmers might try to overcome such a problem. For example, they might draw up a contract by which each agrees to contribute only if all of the others do. A potential holdout will realize that if he refuses, the project will fall through, while if he agrees but someone else refuses, his agreement will not cost him anything, so it is in his interest to agree.
This solution depends on our assumption that all farmers are equally at risk from sparks. If that is not the case, then what each farmer's fair share is becomes a complicated and debatable issue. If even one farmer thinks he is being charged more than the spark arrester is worth to him, the whole deal falls through. And a sufficiently strong-minded farmer may simply announce that he is not going to contribute; if the others want to raise the money, they had better draw up a new contract with his name omitted. As this example suggests, the situation provides a lot of opportunity for bargaining, bluffing, threats, and counterthreats. The result may well be no spark arrester. That is why rule 1 is linked to outcome C by a dashed line, blocked by a box representing the public good problem faced by the farmers.
What about outcome D—Spark Arrester + Clover? As long as both the spark arrester and the switch to clover are costly, that can never be the efficient outcome, since both precautions are costly and either one is sufficient. That is why there are no arrows linking any of the four rules to outcome d. That would change if the invention of diesel locomotives eliminated sparks at no cost while the import of lots of foreign wheat made clover the more profitable crop, but under those assumptions the problem we are analyzing vanishes.
Turn now to rule 2. This time it is the farmers who have the absolute property right; the railroad is permitted to throw sparks only if it has permission from every farmer. What happens?
If the efficient outcome is C—No Sparks + Wheat—it happens immediately with no need for any transactions. The railroad could offer to buy the farmers' permission, but to do so itthat wouldhave to at least reimburse them for the costof switching to clover. Thatcosts more than a spark arrester, so it is not in the railroad's interest to make the offer. Rule 2 is linked to outcome C by a solid line.
Suppose, however, that the efficient outcome is A or B—sparks, with or without a switch to clover. To get there the railroad must buy permission to throw sparks from the farmers. Such a transaction, at any price lower than the cost of a spark arrester and higher than the cost of either fires (A) or clover (B), is in the interest of both railroad and farmers.
It is in their interest but, again, may not happen, this time because of a holdout problem. Suppose the railroad offers each farmer $6 for permission to throw sparks; the cost to the farmers (fires or clover, whichever is cheaper) is only $4. It occurs to one of the farmers that if the deal goes through, the railroad reduces its cost from $1,000 (for a spark arrester) to $600 (paid to farmers), for a net savings of $400. It also occurs to him that if he refuses to agree to the deal, the $400 savings will vanish. Being a reasonable man, he offers to split the gain; he will agree, but only if the railroad pays him an extra $200.
If only one farmer adopts this strategy, the deal still goes through, and the efficient outcome is reached, although there may be some grumbling when the other farmers discover the terms. But if more than two farmers decide to charge $200 for their assent, the railroad refuses to pay and the deal collapses. And it is quite likely that multiple farmers will do so. Their risk is the loss of the $2 profit they make by agreeing to accept $6 for something that costs them $4. Their potential gain from strategic bargaining is $200. Here again, there are ways that the farmers might try to resolve the situation, but none that is guaranteed to work. So rule 2 is linked to outcomes A and B by dotted arrows blocked by the holdout problem.
So far we have been dealing with property rules; one side or the other gets to decide whether or not sparks are thrown. Next we consider liability rules: One party gets to decide but owes compensation to the other for the resulting costs.
Rule 3 is a liability rule in which the railroad controls the decision of whether to throw sparks but the farmers have a right to be compensated for any resulting damage. What happens?
Suppose first that the costs are those shown by column a on the table, making A the efficient outcome. The railroad can either throw sparks and pay the farmers $400 for the resulting fires or else pay $1,000 for a spark arrester; it chooses the former option.
The result is a little more complicated than it was under rule 1; this time when the railroad's sparks start fires, the farmers sue for damages. We have gotten to the efficient outcome, but with an additional cost: the litigation cost of the resulting tort suits.
Implicit in that conclusion is an important assumption: that the court can accurately measure damages. Suppose that is not the case. The fires actually do $400 worth of damage, but the court, impressed by pictures of smoldering fields, overestimates the damages and awards $1,200 in compensation to the farmers. If the overestimate is predictable, it is in the interest of the railroad to put on the spark arrester instead even though that is an inefficient outcome.
The solution to this problem requires another transaction. The railroad goes to the farmers and offers to buy their liability right, to pay each of them $8 in exchange for his agreement not to sue. Just as with rule 2, the parties may be able to bargain to an efficient outcome, but to do so they must overcome a holdout problem. Fortunately it is an easier one than before. A single farmer can no longer force the railroad to put on a spark arrester, since the railroad has the option of throwing sparks and paying him $12 in damages instead. So the most the holdout can hope for is a $12 payment for his assent.
So far we have been assuming that switching to clover costs more than the damage done by the fires, making A the efficient outcome. Suppose we now reverse the assumption: Clover costs $400, fires do $800 worth of damage. B is now the efficient outcome. How do we get it?
The answer depends on how much the court knows when it awards damages. If the court simply measures the damage from fires, the obvious tactic for the farmers is to grow wheat, suffer fires, and send the bill to the railroad. The cost to them is zero. If instead they switch to clover, they pay the $400 cost of the switch and receive no reimbursement—since there are no longer any fires, hence no visible damage to sue for. So they end up at outcome A, even though B is efficient.
We can solve this problem by having the railroad pay the farmers to switch to clover. Here again there is a potential holdout problem, but it is reduced by the fact that the railroad can pay some farmers to switch to clover, eliminating fires in their fields and the resulting liability, and pay damages to the ones who refuse to switch.
There is an easier route to the efficient outcome if the court is smart enough. Under conventional principles of common law, the victim of a tort is obliged to take action after the fact to minimize the damage. A sufficiently well-informed court, applying the same principle to precautions against a continuing problem, could refuse to award the farmers the full $800 cost of their burning wheat on the grounds that half the cost is their fault; they should have switched to clover. If the farmers do switch to clover, the sufficiently smart court will recognize that they are suffering a real cost from the sparks even though there are no visible fires and award them $400 in damages. Under those rules, farmers are better off switching to clover ($400 cost, fully reimbursed by what the court makes the railroad pay them in damage payments) than staying with wheat ($800 cost, only $400 of it reimbursed). We go directly to B, with no further transactions necessary. Do not stop at Go. Do collect $400.
As we will see in more detail in chapter 14, the competence of courts is a crucial factor in determining efficient legal rules. If courts were fully informed about everything, if they knew, in detail, what every party ought to do, we would not need a liability system. The court could simply announce what everyone ought to do and hang anyone who didn't do it. That is the centralized solution to the problem of coordinating human activity. We have massive historical evidence that it does not work, save for very small societies, and lots of economic theory to tell us why.
At the other extreme, if courts knew nothing at all and produced their verdicts by rolling dice, legal liability, law in general, would make little sense as a solution to the coordination problem. To make sense out of our legal system we require a picture of the courts somewhere between total ignorance and omniscience. Exactly where is one of the critical assumptions of our theory, and one rarely made explicit.
In the particular example we have been working with, a property rule, 1 or 2, only requires a court smart enough to tell whether or not the railroad has been throwing sparks and starting fires. Rule 3 gets us directly to outcome A (sparks, wheat, and fires) when that is efficient, provided the court is smart enough to measure the cost actually imposed by fires that actually happen. It gets us directly to outcome B (sparks and clover) only if the court is smart enough to measure not merely the actual cost of the fires, but the cost of precautions taken to avoid that cost (farmers who switch to clover) and the potential cost of precautions not taken that would have avoided that cost (farmers who don't switch to clover), and include that information in its calculation of damages. Whichever outcome we are headed for, if the court is not smart enough to get us there directly, the Coasian solution cuts in: The parties bargain to the efficient rule, provided the transaction costs associated with such bargaining are not too high.
Before I so rudely interrupted myself, we were applying rule 3 to situations where the efficient solution was for the railroad to continue to throw sparks. We now shift to the opposite case. The efficient outcome is C: the railroad should buy a spark arrester, leaving the farmers free to plant wheat without worrying about fires. As long as the court can accurately measure damages, there is no problem. The railroad calculates the cost in damage payments of continuing to throw sparks and puts in an order for a spark arrester.
Rule 4 is rule 3 run backwards. The farmers get to decide whether or not a spark arrester goes on but are liable for the cost of their decision; if they insist on a spark arrester, they must pay for it. Getting to outcomes with sparks is now easy: Each farmer correctly calculates that putting up with fires or growing clover, whichever is cheaper, costs less than buying the railroad a spark arrester. Rule 4 is linked to outcomes A and B by solid arrows.
Getting to C is harder because of the same sort of public good problem that we observed earlier. If any one farmer enjoins the railroad from throwing sparks, he will have to pay the full cost of the spark arrester, giving the other farmers a free ride. In order to get to C, the farmers must somehow agree to share the cost of the spark arrester, just as under rule 1.
Under rule 1, the price paid by the farmers to the railroad to get it to stop throwing sparks was determined by bargaining among the parties; under rule 4, it is determined by the court. If one or more farmers is willing to pay for the spark arrester, the railroad no longer has the option of refusing the deal in hopes of being offered a higher price. Which transaction works better depends on the costs of bargaining and litigation and the relative accuracy of the two procedures for reaching the correct price.
We have now completed our spaghetti diagram. What do we do with it?
What we do with it is choose a legal rule. In order to do so, we need additional information: estimates of how likely it is that each of the alternative outcomes will be the efficient one and of how serious the various transactional problems are.
Suppose we knew that, in all of the cases the legal rule would apply to, outcome A was the efficient one. We choose legal rule 1. It takes us directly to A with no transactions and no need for the court to measure damages. A real-world example is the externality that I impose every time I exhale, increasing the amount of carbon dioxide in the world and thus contributing my little bit to global warming. The cost may be real, but we are quite confident that continuing to breath and putting up with the consequences is the least costly solution to the problem, so we give each individual a right to breathe, free of tort liability for costs that his breathing may produce.
Suppose, on the other hand, we were sure it was outcome C that would always be efficient. Now the same argument implies that we want rule 2. An example is my absolute property right not to be deliberately shot at by my neighbors. While losing the opportunity to practice their marksmanship and take out their frustrations may cost them something, we are pretty sure that it costs them less than being shot costs me.
Things get more complicated if we do not know which outcome will be optimal. Now we have to estimate the probabilities, how likely it is that A, B, or C will be the right answer. Given the probabilities, we can look at each rule and see how easily we can get from it to each of the places we might want to go. Whichever rule performs best on average, minimizes the summed cost of transactions that happen and inefficiency due to transactions that don't happen, is the one we want.
The calculation requires estimates not only of probabilities but also of transaction costs. If we know that courts can measure damages accurately and cheaply, taking into account all possible precautions, then rules 3 and 4 look very attractive; either takes us directly to A, B, or C without requiring any transaction between farmers and railroad. If we believe that courts are cheap and accurate but can measure only direct damages, then how attractive rule 3 looks depends on how likely we think it is that the efficient outcome is B, since that is the only one that requires bargaining to get to.
Similarly, if we believe that it is easy for the farmers to solve their public good problem, perhaps because there are not very many of them and they are all friends and neighbors, rule 1 looks attractive. It takes us directly to outcomes A and B and indirectly but inexpensively to C. It is especially attractive, relative to the alternatives, if we believe courts are costly to use and incompetent at estimating damages.
The basic logic of what we are doing should now be clear. Given beliefs about the world we face (probability of each outcome being efficient) and our technology for dealing with it (costs of transactions, accuracy of court judgments), we can deduce the average cost resulting from each legal rule and pick the least costly. We have, in principle, an intellectual apparatus for designing law.
"In principle" conceals a multitude of problems, since most of what we need in order to pick the right rule is information we don't have. The point of the exercise is, by seeing how we would do the exact calculation, if we had all the necessary information, to understand how to attempt the much more approximate calculation which is the best we can do with our very imperfect data. While we may not be able to prove for certain what rule is best for determining the liability of railroads for fires started by their sparks, we can show what factors make one rule or another more attractive.
Before restating the argument in this form, however, it is worth considering one more possible rule—controlling sparks by Pigouvian taxes. We therefore add to the argument (but not to the diagram, which is quite complicated enough already)
5: Fine for sparks. Railroad pays a fine equal to the damage done by its sparks.
At first glance, this rule seems to solve all problems. If the damage done is less than the cost of the spark arrester (outcome A or B is efficient), the railroad pays the fine and throws the sparks. The farmers, who are now receiving no compensation, grow wheat or clover according to which minimizes their cost. If the damage done is more than the cost of the spark arrester, the railroad buys the spark arrester and we are safely at outcome C. No transactions between farmers and railroad are required.
There are some problems, however. Suppose the regulatory agency assessing the fine, like the only moderately smart court of an earlier example, can measure and charge only for direct damages. Further suppose that switching to clover costs less than putting up with fires but more than a spark arrester. The efficient outcome is C; the costs are shown in column c of the table.
The railroad continues to throw sparks and pay fines. When questioned by irate stockholders about why it is paying fines when it would be cheaper to put on a spark arrester, its president tells them to be patient. The farmers, observing that the railroad continues to throw sparks, switch to clover. The fires cease. So do the fines.
Railroad and farmers are engaged in a game of mutual bluff. The farmers could try to force the railroad to put on a spark arrester by refusing to switch to clover, just as the railroad can try to force the farmers to switch to clover by refusing to put on a spark arrester. But it is a game where the railroad is in much the stronger position—because there is one railroad and a hundred farmers. Each individual farmer has an incentive to switch to clover in order to avoid the cost of fires, while hoping that enough others will grow wheat to maintain the railroad's incentive to put on the spark arrester. The farmers, in trying to keep each other in the business of burning wheat until the railroad backs down, face a public good problem.
A second problem was mentioned in the previous chapter: Coase plus PigouCoase plus Pigou is too much of a good thing. Suppose the spark arrester costs $1,000, switching to clover costs $800, fires cost $600 (column d). The efficient solution is to keep throwing sparks. Under this solution, each year the railroad pays a $600 fine, and the farmers suffer $600 worth of damage. If the railroad installs a spark arrester, both fine and damage disappear, for a combined gain of $1,200 at a cost of $1,000. The farmers offer to pay the railroad $500 toward the cost of a spark arrester, the railroad agrees, and we have gotten to the inefficient outcome C.
We are now finished with the details of this particular example. The next step is to see what it has taught us, not about railroads and farmers but about the advantages and disadvantages of different sorts of legal rules.
If someone breaks a window or a contract, he gets sued for damages roughly equal to the cost he has imposed on the victim. If instead he steals a car and is caught and convicted, he ends up in jail. The right to performance of a contract, or the right not to have your property accidentally damaged, is a liability right, protected by civil (contract and tort) law. The right to possession of your car is a property right, protected by criminal law.
A liability right to something means that if it is taken, the taker owes you compensation. A property right means that if it is taken, the taker is punished in a fashion intended to make it in his interest not to have taken it. Very roughly, we may think of the liability rule as "damages equal to damage done" and the property rule as "damages high enough always to deter." Real-world property rules, of course, do not always deter, an issue that will be discussed in chapter 15. But it will simplify the present discussion if we ignore that complication for the moment.
Imagine a world with only liability rules. If someone wants your car, he doesn't have to come to you and offer you a price. He simply takes the car and lets you sue him for its value. What is wrong with that approach? Why would it not lead to an efficient outcome?
In a world where courts could accurately measure value and enforce liability judgements at negligible cost, it would. The damage payment equals the cost to you of having your car stolen, so it pays to steal a car only if it is worth more to the thief than to the owner, in which case stealing it is an efficient transaction. And, unlike the (uncompensated) theft discussed in chapter 3, such a legal rule would lead to no rent seeking. I would have no incentive to spend time and money guarding my car against thieves, since if it were stolen I would be fully compensated—and besides, theft would be rare in a world where the thief had to fully compensate the victim. The cost to thieves of stealing cars would be negligible, since owners would take no precautions. All you have to do is walk up to a car, open the (unlocked) door, and start it with the key that the owner has left in the ignition.
In the real world courts cannot measure the value of my car to me very accurately, not nearly as accurately as I can. Furthermore, the cost of finding out who stole the car, proving that he did it, and establishing an estimate of its value, is substantial. Protecting my right to my car with a liability rule would be an expensive approach and one that could frequently lead to inefficient transfers.
That would not be a sufficient reason to reject that approach if we had no better alternative, but we do. The better approach is a property rule: You can take my car only with my permission. If you take it without my permission very bad things will happen to you.
The reason that is a better approach was sketched out by Anne, Mary and John back in chapter 2. If my car is worth more to you than to me you will be willing to make me an offer that I will be willing to accept. You don't have to steal it; you buy it instead. Value is demonstrated not by a court's best guess but by what offers you are willing to make and I am willing to accept. The transaction cost of selling you the car is a great deal less than the transaction cost of your stealing the car and my then suing you and collecting.
If property rules work so well, why do we ever use liability rules? The answer is that there are some cases in which the transaction cost of selling is higher than of suing. Every time I pull out of my driveway I impose a tiny risk of injury on every driver and pedestrian for miles around. The transaction cost of getting permission from all of them before I turn on the ignition is prohibitively high. So instead of enforcing their right not to be run into or over by a property rule, we enforce it by a liability rule: If I injure someone under circumstances in which the court finds me at fault, I must compensate him.
These examples suggest a very simple conclusion. Property rules are attractive when the cost of allocating rights by market transactions is low. Liability rules are attractive when the cost of allocating rights by litigation is low. We use property rules in contexts, such as allocating cars, where allocation through the market works well, allocation by courts works badly. Reverse the situation and we use liability rules. In a world where both costs were zero, either approach would lead to an efficient outcome. In a world where both sets of costs are high, we may have to look for other alternatives.
Consider the distinction between trespass by people and trespass by cattle. If I deliberately trespass on your land, I am liable without regard for how much damage I did or how reasonable my action was. If my cattle stray onto your land, I am liable only for the damage they actually do. Thus your right not to have people walk on your land without your permission is enforced by a property rule; your right not to have cattle stray onto your land is enforced by a liability rule.
The obvious explanation is that if crossing your land is worth more to me than it costs you, I can buy permission—permanently, in the form of an easement, if that is what I require. But if I want to keep cattle on my land, I cannot readily buy the permission of every landowner onto whose property they might stray.
For a second example, consider the difference between the legal treatment of an accident and the legal treatment of a continuing nuisance. If my haystack spontaneously ignites and the fire spreads to your nearby buildings, you sue me for damages. If the smell from your tannery makes my apartment building unfit for human habitation, I go to court and ask to have the tannery enjoined as a nuisance. In the former case I am asking the court to apply rule 3, to give me a liability right not to have my buildings burned down. In the latter I am asking it to apply rule 2, to give me a property right not to have my building fumigated without my consent.
Part of the explanation for the difference may relate to the ability of the court to determine who is the least-cost avoider. With a continuing nuisance, the cost to either party of eliminating or tolerating the nuisance is to some degree observable. So the court has a reasonable chance of figuring out who is the least-cost avoider and assigning the property right accordingly, thus getting directly to the efficient outcome either on a case-by-case basis or by general rules such as the doctrine of coming to the nuisance. But accidents are uncertain, making their expected cost uncertain, making it hard, perhaps impossible, for the court to know who is the least-cost avoider. By using a liability rule, it puts the burden of deciding whether to take precautions, by keeping the hay wet or locating my haystacks farther from my neighbor's building, on me. All the court has to do is to estimate damages when and if the accident happens.
A second reason for the difference may be that reallocating rights by a private transaction is often more practical in the case of the continuing nuisance. If my tannery is polluting your factory, both of us know it. So if the wrong person is assigned the right by a court deciding whether or not to grant an injunction, the other person can buy it from him, at least as long as only one or two people are affected. This doesn't work as well for accidents. If the legal system gives you an absolute right not to have your buildings burnt down by accident, enforced after the fact with statutory damages or criminal punishments, I may not know which of my neighbors I need to buy what rights from in order to avoid being found liable if an accidentoccurs.
occurs.
One way of treating pollution and similar externalities is as torts; if your neighbor's noisy candy factory makes it impossible to use your consulting room, you sue him for damages. Under modern environmental law they are more commonly treated as violations of federalregulations and punished with fines. If our objective is efficiency, how do we choose between the two approaches?
One answer is that the two approaches have different effects on the incentives of the victim to take precautions. If, when a railroad throws sparks, farmers whose wheat burns are compensated for the loss, they have no incentive to switch to clover even if that would be a cheaper solution than either fires or a spark arrester. If the railroad is fined but the money does not go to the farmers, they have an incentive to minimize the damage by switching to clover. That is one respect in which fines are a better approach. How important the advantage is depends in part on how easy it is for the railroad to negotiate with the farmers. If transaction costs are low, a liability rule leads to the efficient solution too, because the railroad will pay the farmers to switch to clover.
Low transaction costs not only make a liability rule work better, they also make a fine work worse, for a reason already discussed. If transaction costs are low and the fine does not go to the farmers, the railroad has a double incentive to put on the spark arrester; not only can it eliminate its fine, it can also get the farmers to help pay the cost. So it may be in the railroad's interest to install a spark arrester even if doing so costs more than the damage done by the sparks.
So far I have ignored a second basis for choosing between these two rules: the incentive each provides to report and prosecute the offense. In order for someone to be punished, his violation must be reported and prosecuted. A liability rule gives the victim an incentive to do so, since he is the one collecting the damage payment. A fine does not.
One way of thinking of tort liability is as a Pigouvian tax, designed to force tortfeasors to internalize their externality. Another is as a way of compensating victims, a sort of court ordered insurance policy. A third is as a bounty system, in which private law enforcers, also known as tort victims, bring malefactors to justice and are rewarded with the fines paid by the guilty parties. Only the first approach has been dealt with here; the others will make their appearance in later chapters.
One possible response to the problems I have been discussing in this chapter is that I have not considered enough legal rules. One might, for example, argue that the solution to both the public good and holdout problems faced by the farmers is a legal rule that allows a majority of farmers to set the terms on which all farmers will deal with the railroad. If the majority decides that each farmer will chip in $4 to buy a spark arrester, every farmer must do so; if the majority agrees to permit sparks or to grow clover in exchange for a payment from the railroad, every farmer is bound by that agreement. Versions of that rule are used not only in democratic governments and many private organizations but in some legal contexts as well, such as the rules for unitizing oil fields.
That particular set of institutions solves the problems I have discussed here at the cost of producing a different set of problems. Consider, for example, the costs shown in column b of the table, with a slight modification of our assumptions: Forty of the farmers are already growing clover; the other sixty have soil suited only for wheat and suffer the whole $800 cost of the fires.
The legal rule is 1; the railroad is free to throw sparks. One of the farmers who is at risk of fire proposes that each farmer contribute ten dollars toward the cost of buying the railroad a spark arrester. For the farmers who are at risk, it is an attractive proposal, since they are losing more than ten dollars to fires. The proposal passes by sixty to forty. The result is outcome C, even though A would be more efficient. The railroad ends up with a spark arrester that costs more than the fires it prevents.
As this example suggests, it is not enough to come up with a legal rule that works better under one set of assumptions; it may work worse under another. I could have added this additional rule and several others to the figure, along with additional columns on the table to describe a greater range of possible cost distributions. Doing so would make the analysis, which already occupies most of a chapter, still more complicated but would not change the essential point—a demonstration not of what the efficient rule is but of how one goes about finding it.
One point that may have occurred to you is that choosing the right legal rules requires a good deal of specific information about the problems they are going to be applied to—information that a court may not have. That is one argument for the principle of freedom of contract, which allows the parties to a contract to set their own rules for the little two-person society that the contract creates; they, after all, are the experts on the problems that particular society is likely to encounter. More broadly, it suggests that there may be much to be said for a legal system sufficiently decentralized so that legal rules are created by people reasonably close to the problems those rules will apply to.
Coase, Ronald, "The Problem of Social Cost," Journal of Law and Economics 3, 1-44 (1960).