Feud as Law Enforcement, Ancient and Modern
Why is There a Patent Troll Problem and How Can it be Solved?
Feud  is one of the mechanisms by which legal rules are enforced. Its essential logic is simple: If you wrong me, I threaten to harm you unless you compensate me for the wrong. The critical requirement for it to work is some mechanism that makes my threat more believable when you actually have wronged me than when you have not, some way of converting right into might, in order to prevent the enforcement mechanism from being used instead for extortion.
For a simple example, consider the feud system of the Rominchal gypsies, the largest gypsy population in England. If you wrong me, I threaten to beat you up. Both of us know that if you have wronged me, as judged by the norms of our community, my friends will back me and your friends won't back you, making it in your interest to either compensate me or leave town.
Feud systems have existed in many human societies. In addition to the Rominchal, well recorded examples include saga period Iceland and traditional Somali. In the Icelandic case, the mechanism for converting right into might was an explicit law code and a court system. You sued the person who wronged you. If you won, the verdict was a damage payment he owed you. If he failed to pay, he was outlawed and had two weeks to leave Iceland, after which it was legal to kill him and tortious for anyone to defend him. That system functioned for about a third of a millennium, producing substantial amounts of violence only in the final fifty year period of breakdown. The Somali version was somewhere between the Icelandic and the Rominchal, with customary law and customary mechanisms for setting up courts to arbitrate disputes, along with a fascinating system of prefabricated coalitions to deal with both paying damages and enforcing their members' claims.
There is evidence that many, perhaps most, legal systems were built on top of feud systems. That includes Jewish law and Muslim law, both of which contain fossilized evidence of preexisting feud systems, along with Anglo-American common law and Roman law.
Feud systems are not a matter of only historical interest—de facto, if not de jure, they still exist. One current example is patent litigation among high-tech companies such as Apple and Samsung.
To see the logic of the situation, imagine that Apple is considering suing Samsung for a patent violation of which Samsung is not actually guilty—the equivalent, in the modern context, of a Rominchal Gypsy or medieval Icelander wronging someone by attacking him for no good reason. There are at least two reasons why Apple might sue even if it did not believe in its own case. One is the chance that the court will mistakenly decide in Apple’s favor. The other is that, even if Apple loses the case, the litigation imposes significant costs on a rival. Not only will Samsung, like Apple, have to spend money on lawyers, the public perception that Samsung might end up having to withdraw products from the market or modify them will cost Samsung sales, some of which will go to Apple.
If courts were perfectly able to judge cases and measure costs, there would be no mistaken decisions and the court could award Samsung, if it prevailed, damages against Apple for Samsung’s lost sales. In practice, however, courts sometimes reach mistaken verdicts and are unlikely, perhaps unable, to calculate such indirect effects and award damages on the basis of them. Hence, in the real world, Apple might find it profitable to sue even if it knows it has a weak case.
In the real world, however, there is a reason for Apple not to sue that does not depend on the court always getting the result right—the risk that Samsung will retaliate by suing Apple. If we assume that neither company has actually violated the patents of the other, the countersuit may still be profitable for the same reason the initial suit was. And even if the countersuit is not profitable as a gamble on court error or a way of reducing Apple’s sales in favor of Samsung’s, being committed to such a countersuit is one way of deterring the initial suit, just as being committed to vengeance against anyone who kills your kin is one way of keeping your kin from getting killed. The implicit feud system in modern patent litigation provides a mechanism for deterring meritless suits that might otherwise be profitable, just as traditional feud systems deter other forms of otherwise profitable wrongs, such as robbery.
What about suits that are not meritless—what if Samsung actually has infringed Apple’s patents? The threat of countersuit is still a cost to Apple of suing. If courts reached their verdicts at random, the situation would be the same as in the meritless case and the feud system would equally deter suits in both cases.
But courts do not reach their decisions at random. If Samsung is guilty, that raises, one hopes substantially, the chance that Apple will win. That increases the benefit to Apple of suing. If it increases it by enough to outweigh the cost of Samsung’s (hypothetically meritless) countersuit, Apple sues.
To put it differently, the mechanism through which right makes might in the modern feud context is the court system. As long as the plaintiff is more likely to prevail when he is in the right than when he is in the wrong, suing someone for infringing your patents produces a larger benefit to the plaintiff and a larger cost to the defendant when the defendant actually has infringed the plaintiff’s patents than when he has not. Provided that the cost imposed by the threat of countersuit is greater than the benefit of a meritless suit but less than the benefit of a legitimate suit, the result is to deter the former but not the latter. That result depends on courts working to some degree, but it does not require them to work as well as they would have to work in order to themselves deter plaintiffs from filing meritless suits. The high tech feud system working through the court system is a better mechanism for law enforcement than the court system alone in at least one important respect: It requires less ability on the part of the courts in order to work.
The clearest anecdotal evidence that what I have described is how the system actually works is the practice of high-tech companies accumulating large inventories of patents, many of which they are unlikely to use. It is the modern equivalent of the medieval Icelander accumulating weapons and allies just in case he ever needs them to prosecute his side of a feud. As in that case and the somewhat higher stakes version played by nations under the name of Mutual Assured Destructions, if the strategy works the weapons need never be used.
There remains, however, one hole in the system.
The Invulnerable Plaintiff: Non-Practicing Entities and the Patent Troll Problem
Samsung and Apple both produce cell phones and so both can be at least plausibly accused of violating relevant patents; they are thus both vulnerable to threats of retaliation. A firm that produces nothing is not. A non-practicing-entity, referred to by critics as a patent troll, owns a collection of patents, sues practicing entities for infringement, but faces no risk of an infringement countersuit. It is invulnerable.
I described two reasons why Apple might sue Samsung even if the suit was meritless. Only one applies directly to the case of the non-practicing entity. It is not producing phones, so gets no direct benefit from persuading people not to buy phones from Samsung. Like Apple, however, it has the possibility of profiting by court error, of winning a case it should have lost and collecting damages. And, although imposing costs on Samsung provides no direct benefit to the non-practicing-entity, it does give Samsung an additional incentive to settle instead of letting the case go to trial.
In the case of Samsung, there is an obvious incentive not to settle—paying off one plaintiff with a weak case will encourage others. That incentive is weaker in the case of a much smaller firm, unlikely to be the target of multiple extortion attempts and at risk of being destroyed by a single law suit. Hence we get what critics of non-practicing-entities allege to be their usual tactic—suing small firms in order to be paid off to drop the suit.
It follows that, even if the feud system is adequate as a way of controlling patent suits among producing companies, it is impotent to control patent suits by non-practicing entities. Which suggests that we may need something else.
The Athenian Rule: A Modest Proposal for Revising Tort Law
Under the so-called English rule, the losing party to a tort suit owes the prevailing party compensation for its legal costs. To see why that is not an adequate deterrent to meritless suits, consider the situation that exists if litigation costs are zero—lawyers and judges work for free—but there is some probability of court error. As long as the plaintiff has some chance of prevailing it pays him to sue, since suing is a “heads I win, tails I break even” bet. If litigation costs are not zero, some such suits may be deterred by the fact that the plaintiff will pay more in litigation than he can expect to receive through erroneous verdicts in his favor, but whether that happens depends on the relative size of the litigation cost and the potential damage award and on the probability of an erroneous favorable verdict.
Add in the possibility of an out of court settlement and the situation gets worse. If we assume that the cost of negotiating such a settlement is negligible and that the parties settle for the middle of the bargaining range—the defendant pays the plaintiff his expected return, the probability that he will prevail times the damages he will be awarded if he does—we are back in a world where litigation costs are zero and it pays to sue as long as there is any chance of winning. Defendants who are repeat players, such as a firm the size of Apple or Samsung, may be able to commit to never settling for enough to cover the plaintiff’s litigation costs, provided that those are high enough to make actually going to trial unprofitable for the plaintiff. The problem is more serious for defendants who do not expect to be repeat players and so are not in a position to commit themselves.
The fundamental problem is that a plaintiff who sues an innocent defendant in a system with legal error imposes a cost on him—the risk of losing the case and being found liable for damages. We cannot compensate the defendant for that risk when it actually eventuates, because if he loses we do not know he is innocent. But tort law treats a defendant who is guilty by the preponderance of the evidence as if he were guilty with certainty, so it seems natural to treat a defendant who is innocent by the preponderance of the evidence as if he were innocent with certainty, making the plaintiff guilty of suing an innocent defendant. We can then use damages owed by the losing plaintiff to the prevailing defendant as a proxy for damages for the cost imposed by a plaintiff who sues an innocent defendant and wins. The logic is analogous to the case for punishing unsuccessful criminal attempts. Shooting at someone and missing does no harm. Punishing someone who shoots and misses serves as a proxy punishment for imposing a risk of death, imposed when the risk does not eventuate in order to provide an additional disincentive beyond the largest punishment we are able and willing to impose on the successful criminal.
These arguments suggest that the losing tort plaintiff should be liable to the defendant for damages, possibly based on the amount the plaintiff claimed and thus the size of the risk imposed, possibly also on some measure of how badly the plaintiff lost and thus how strong the evidence is that he was innocent. Making damages depend on how badly the plaintiff lost would correspond to the rule for private prosecution of most categories of criminal offenses in Periclean Athens; a prosecutor who failed to get at least 20% of a large jury to vote for conviction was himself fined. Making the damages depend on the amount claimed would correspond to the rule applying in Athens to at least some of their equivalent of our tort cases; the losing plaintiff owed the defendant one sixth of the amount claimed. It is not known whether that penalty depended on getting less than 20% of the jury to vote for conviction, like the equivalent in criminal cases, or applied to any acquittal.
By suitably adjusting the damages owed by the losing plaintiff and those owed by the losing defendant, the legal system could deter would-be plaintiffs from suing defendants they believe to be innocent while providing potential tortfeasors with the desired level of deterrence.
I have sketched the idea of the Athenian rule, damages owed to a prevailing tort defendant to compensate him for the risk of being wrongfully found liable, in the context of the patent troll problem, but the argument is a general one. Suing an innocent defendant in a system with some rate of legal error imposes costs beyond the cost of the litigation, for which the plaintiff should be liable. The rule is particularly important in the patent troll case only because that is a situation where deliberately suing innocent defendants and then proposing settlement appears to be a serious problem. There may well be others.
The Code of Maimonides, Book Eleven, The Book of Torts, Kyman Klein tr., Yale University Press, New Haven, 1954.
The Twelve Tables, from Ancient
Roman statutes : translation, with introduction, commentary, glossary, and
index, by Allan Chester Johnson, Paul Robinson Coleman-Norton, Frank Card
Bourne ; general editor, Clyde Pharr Austin : University of Texas Press, 1961
Forte, David F., Studies in Islamic Law, Austin and Winfield, Lanham, N.Y., Oxford, 1999.
Friedman, David, “Private Creation and Enforcement of Law — A Historical Case.” Journal of Legal Studies, (March 1979), pp. 399-415.
Friedman, David, JLS XX, (January 1991), webbed at: http://www.daviddfriedman.com/Academic
Lewis, I.M., A Pastoral Democracy: A Study Of Pastoralism And Politics Among The Northern Somali Of The Horn Of Africa, Published for the International African Institute by the Oxford University Press London New York Toronto, 1961.
MacDowell, Douglas M., The Law in Classical Athens, Cornell university Press, Ithaca, N.Y., 1978
Weyrauch, Walter O., Ed., Gypsy Law, University of California Press
 Often, but misleadingly, referred to as “blood feud.” The description is misleading because, as with other systems of law enforcement, people get killed only when the system fails—if it works, it either deters wrongs or settles the resulting conflict by damage payments to the victim or his kin. And while kinship often plays a role in such system—another possible justification for “blood feud”— it need not. In both the Icelandic and Somali systems, the coalitions that enforced claims were based in part on kinship, in part on alliance. The modern feud system described in this essay makes no use of kinship.
 Gypsy Law, Chapter 3.
 Not the same thing as feudal systems—the words sound similar but are unconnected in both meaning and origin.
 See D. Friedman (1979). An interesting contrast is between the conflict between pagans and Christians that ended with an arbitrated settlement in favor of Christianity in the year 1000 and the conflict between Catholics and Lutherans that led to a forced settlement in favor of protestantism more than four hundred years later. The former conflict, which occurred within a feud system where law enforcement was entirely private, was accompanied with considerably less violence and many fewer deaths (about ten) than the later conflict, after Iceland had passed under the rule of the Norwegian crown.
 For Jewish law, the clearest evidence is the role in homicide cases of the avenger of blood, the heir of the victim. If a killer is sentenced to capital punishment, it is the avenger of blood who is supposed to impose it. If he is sentenced to exile in a city of refuge, the avenger of blood has the right to kill the killer before he reaches it. Maimonides, Mishnah Torah, book XI Treatise V Chapter 1.
The evidence for Muslim law is Jinayat, the part of the legal system dealing with killing and injury, both of which are treated as torts against the victim or his kin. Forte (1999).
 The Twelve Tables, the earliest text of Roman law of which we have a reasonably good picture—the original has not survived, but parts of it are quoted in many other texts that have—includes:
Table VIII, 2. If anyone has broken another's limb there shall be retaliation in kind unless he compounds for compensation with him.
Table VIII, 12. If a thief commits a theft by night, if the owner kills the thief, the thief shall be killed lawfully.
Table VIII, 24a. If a weapon has sped accidentally from one's hand, rather than if one has aimed and hurled it, to atone for the deed a ram is substituted as a peace offering to prevent blood revenge.
All of which look like survivals of a pre-existing feud system.
 Insert a cite to a news story on this.
 The same point could be made in the context of suits by practicing entities against small firms—a small firm is less of a repeat player than a large, so less able to commit itself to a policy of retaliation. But there is a balancing advantage the other way. The small firm has much lower assets than the large firm, so its potential loss from suit is much less. And the small firm is judgement proof at a much lower level than the large, making the potential gains from suing it and winning much lower.
 In the rule as it actually exists in English law, the compensation is for what the court believes the legal costs should have been, not for whatever the costs actually were.
 For simplicity, I assume that litigation costs are the same for both parties.
 This explanation of punishment for attempts is worked out in more detail in Friedman, 1991)
 The fine was 1000 Drachma, about three years wages for an ordinary working man. (MacDowell, Douglas M. The Law in Classical Athens). If the defendant was convicted, the prosecutor, in most but not all sorts of cases, received a sizable share of the fine he paid, creating an obvious risk of prosecutors targeting men who were wealthy, unpopular, and innocent. Unlike my proposal here, the fine went to the state, not to the defendant. Along somewhat similar lines, the earliest and most important legal action in Roman law, the legis actio sacramento, required each party to deposit a sum of money as a wager on the outcome. The prevailing party recovered its deposit; the deposit of the loser forfeited.
 "In some private cases in which the prosecutor had claimed a sum of money from the defendant, on losing the case he had to pay to the defendant one-sixth of the amount which he had claimed (epobelia, one obol per drachma). Demosthenes faced this risk when he prosecuted his guardian Aphobos in 364/3 for making away with his inheritance, but Kallimakhos did not when he claimed 10,000 drachmas in 400/399, even though epobelia was payable in a diamartyria or a paragraphe at that period. Either the law was changed between those dates or epobelia was payable only on some financial claims and not all. It is also uncertain whether it was payable on every acquittal or only when the prosecutor failed to get one-fifth of the votes."
MacDowell ( 1978), pp. 252-253.
 An account of the relevant mathematics is contained in a paper currently in progress, which also considers the situation where the potential plaintiff does not know guilt or innocence with certainty but does have an estimate of probability. Suing a defendant who is probably innocent provides some deterrence, but much less than suing one who is probably guilty—and it imposes similar costs. We would like parties to properly take account of that when deciding whether or not to sue.