Convergence and Then Downstream Divergence in Torts and Other Law – Article by Saul Levmore

 

From Volume 92, Number 4 (May 2019)
DOWNLOAD PDF


 

Convergence and then Downstream Divergence in Torts and other Law

Saul Levmore[*]

Introduction

All legal systems discourage theft, negligence, and other destructive behaviors, but they do so in disparate ways. They converge as far as many of their aims and starting points are concerned, but eventually they diverge in their details. My claim in this Article is that, at least in the case of tort law, there is a pattern to this descent. The divergence has a recognizable cause and is therefore fairly predictable. Specifically, I argue that convergence is associated with laws that are both efficient and attractivea term that conveys a shared ethical intuition in the society governed by these laws. Divergence, on the other hand, arises around rules or standards that are unnecessary on efficiency groundsbecause alternatives seem equally efficientand that do not reflect some commonly held ethical sensibility.[1] Inasmuch as these terms, as well as this ambitious claim, are open to debate, it is useful to consider a number of legal problems and treatments, both legislative and judicial, in order to clarify and appreciate the claim. The argument draws on elements of comparative law, the reality of shared ethical sentiments within a society, the economic analysis of tort theory, and the limits of empirical work. In this Article, the claim is limited to torts but, as suggestive examples will show, it can be turned into a conjecture that includes other areas of law,[2] and that pertains to both rules and standards.[3] It might also say something about areas in which law outsources decisionmaking to markets, but that subject is left for another day.

I.  Binary Constructions: Contributory and Comparative Negligence

In many jurisdictions, tort law long ago settled on an axiom that is binary in form: liability is assigned to a negligent party who caused harm; causation is also an all-or-nothing game. Law, more generally, vacillates between binary rules and smoother, multifactor ones. A either murdered B or did not; there is no punishment when there is a 30 percent chance that A did it. Similarly, in tort law’s treatment of negligently caused injuries, the long-standing tradition was, first, if C was 70 percent likely to have negligently caused D’s injury, C paid 100 percent, and second, that liability was completely forgiven when the injured party, D, was also negligent.[4] To be sure, there was, and remains, divergence at the first step because some activities were subject to strict liability; still, when this regimen was in force, liability was withdrawn in certain cases if the injured party’s negligence contributed to the injury.[5] Tort liability was thus regularly allornothing, or what I am calling binary, even when law might instead have shared liability between the injurer and injured parties. Several features of this characterization require examination. A newcomer to the legal system might ask where these categories came from and why they are necessary. A more difficult but somewhat tangential question is why and when laws come in binary form. This is especially relevant here, because the binary character is an important starting point for the claim that convergence is located where efficiency and shared ethical sentiments find common ground.

It is apparent that categories as well as binary rules facilitate simplified instructions to those who must make everyday decisions with legal threats hovering nearby, and this is itself a form of efficiency. Categories also smooth the work of government officials asked to judge or enforce laws. The more complicated an organization, and the lower the cost of false positives, the more attractive are binary rules. For example, a law requiring drivers to be at least sixteen years old might be sensible, even though age is an extremely rough proxy for what the law really wants of drivers, including skill, informed judgment, and an appreciation of risks. This is because the cost of barring skilled but underage drivers is low, especially when they are on notice not to invest in early training or in jobs that require them to drive, or where there is public transportation. It might be more efficient to count on insurance markets, concerned parents, or direct and cleverly designed fees to encourage safety training, or even to determine the age at which driving is permitted for particular individuals. However, generalization seems low cost and politically acceptable (note the hint of shared ethical intuitions here); it also makes enforcement easier as very young (illegal) drivers can be identified and removed. In fact, once automobiles came into wide use, virtually every jurisdiction established a minimum age requirement for drivers, as opposed to a standard or to no intervention by the state at all.

However attractive legal systems find binary categorization,[6] it comes at a cost. Categories can be foolish. Thus, people are unlikely to be entirely introverted or extroverted, or cruel or kind, however natural it may be to imagine these binary forms.[7] In some cases, binary legal judgments are attractive because more refined, multifactor decisionmaking is not only costlier to process but also likely to bring on the risk of corruption, at least where the new variables call for judgment rather than straightforward measurement. A standard gives a regulator more discretion, and thus invites common sense, but also abuse.[8] Binary rules make it easy to treat like cases alike, and thus lower a source of discontent by disappointed parties. Much of this is familiar from the rules-versus-standards literature, but some caution is required. An apparent rule might require a great deal of discretionary decisionmaking; thus, negligence and contributory negligence (discussed below), like “net income” in tax law and so many other legal “rules,” are better thought of as standards than as rules. This is in part why it is misleading to associate either binary or multifactor guidelines with what are conventionally described as rules.

The popularity of a negligence or strict liability approach, accompanied in the academic literature by a binary contributory negligence defense, is from this perspective unsurprising. Its efficiency is well known.[9] In the case of automobiles, for example, a driver has reason to be careful because he has no reason to expect that another driver will also be negligent in a way that causes the two to crash. The second driver also has reason to take care because she will know that recovery is unavailable if she does something like speeding out of a driveway without looking and is then hit (even) by another speeding vehicle unable to be stopped in time. Again, describing the law as binary, refers to the fact that liability is allornothing when it might easily have been shared between the two parties.

A more global explanation for the prevalence, or convergence, of the binary character of many laws, is that it suppresses evidence of voting paradoxes.[10] If a decision is to be made by multiple voters or other decisionmakers and there are more than two variables, it is easy for the decisionmakers to be unable to reach a stable result. They might cycle among the available options or be disgruntled when they can see, after the smoke clears, that a majority opposes the decision reached by a legislature or other body of voters that claimed to be governed by majority rule. A majority may prefer outcome X over Y, and then Y over Z, but Z might be preferred over X. Motion-and-amendment voting, found in legislatures around the world and popularized by Robert’s Rules of Order in the case of sizeable private organizations in the United States, finds a Condorcet winnerwhich is to say an option that defeats all other options in head-to-head competitionif one exists.[11] Motion-and-amendment voting allows X to win if it is preferred over both Y and Z, and all other proposals, but it hides the fact that there is no real, consistent majority winner where X beats Y, Y defeats Z, and Z is preferred over X. This voting method does this by always asking for binary decisions and then making it difficult to revisit an option once the option has lost in a head-to-head competition. With its binary decisions, law thus hides evidence of cycling preferences but does an excellent job of finding desirable, Condorcet winners where they exist. This might explain why legislatures are normally asked to vote a proposal, or any amendment or substitution to it, up or down, and why panels of judges on appellate courts, in binary style, affirm or reject rulings that they hear on appeal.[12] Indeed, this is a remarkable example of convergence across legal systems, though I am unsure whether to call it efficient (as it avoids repetitive cycling) or simply politically attractive (as it prevents the outrage expected of people when they see that a majority opposes the result found by another alleged majority). Though this is not unique to tort law, it is useful to see that where there is little risk of cycling, as when preferences are “single peaked,”[13] law is more hospitable to putting multiple variables or ranges on the table.

A good example of the emergence in tort law of a non-binary approach where there is little risk of cycling, and where there is no loss of efficiency and even a gain in terms of law’s ability to conform to political and ethical expectations, is the emergence of comparative negligence. Its predecessor, contributory negligence, is binary in form; a wrongdoer is absolved if the injured party was also negligent. In contrast, comparative negligence, under which the fact-finder spreads liability according to the relative wrongdoing of multiple parties, is more continuous as well as increasingly popular.[14] There are two useful ways to think about the emergence of comparative negligence. First, citizens find comparative negligence attractive because they seem to dislike the idea that mere negligence can reduce an injured party’s recovery all the way down to zero. The shared ethical sentiment is to favor comparative over contributory negligence. This sentiment may be motivated by a fear of mistaken fact-finding, though I hesitate to put too much weight on this conjecture because it introduces a kind of efficiency argument to what is essentially one about ethical sentiments. It seems unlikely that many citizens or legislators think about enforcement costs when they opine about fairness in tort law. A better description might therefore draw on a theory of justice or an evolved taste for retribution. An obvious question is why older law did not reflect these ethical intuitions. One answer is that perhaps we cannot know what ancient shared ethical intuitions looked like. Another, and one with some support in the historical record, is that some lawmakers inserted a version of comparative negligence by quietly reducing damages when the injured party was also negligent.[15]

The second characteristic of the emergence of comparative negligence in tort law returns us to cycling preferences. Imagine decisionmakers deciding whether to award full damages, partial damages, or zero damages to someone who was injured by a speeding driver, but who was arguably contributorily negligent, perhaps by suddenly emerging from a hidden driveway. One fact-finder might rank the desirability of results as Full-Partial-Nonewhere Full means that all of the liability is assigned to the speeding driver on the roadway, and is the fact-finder or decisionmaker’s first choice, Partial means that the first driver pays a fraction of the loss suffered by the negligent victim, so that the two divide the loss, and so forth. Another voter might have the opposite inclination, reasoning that law should strongly discourage drivers from darting into a public road, so that recovery by the negligent injured party should be completely denied. But it is hard to see why anyone would rank Partial last. A voter, or jury member, who thinks the plaintiff should recover full damages will surely be happier with a partial recovery by this plaintiff than with no recovery at all. Similarly, one who thinks that contributory negligence should bar recovery, would certainly prefer partial recovery over full recovery if she cannot have her first choice of None. If so, there is no danger of cycling; Full will beat None, or None will defeat Full, or there could be a tie between these two, but there is no chance that the winner of the Full-None competition will then lose in competition with the Partial option. Partial liability can certainly be a voter’s first choice, and it is almost surely the second choice of those whose first choice is either Full or None, but it is hard to imagine Partial (liability) as anyone’s third choice, and for this reason there will be no cycling. Similarly, the common rule barring anyone under the age of sixteen from operating a motor vehicle is a binary rule where there is no danger of cycling.[16] It is telling, and certainly interesting, that law has moved away from the binary approach of contributory negligence just where there is no apparent risk of cycling preferences.

It is noteworthy that the non-binary form of modern comparative negligence was first introduced in European admiralty law, where judgment was rendered by a single lawmaker who was, therefore, in no danger of cycling.[17] Indeed, it is tempting to say that legal systems that rely on a single judge rather than a jury will be quicker to adopt comparative negligence, though this hardly explains Hammurabi’s Code and other long-standing uses of contributory negligence. But the argument here suggests that juries are unlikely to cycle within a comparative negligence framework, so it is unsurprising that comparative negligence emerged in a wide variety of jurisdictions.[18] Similarly, it is not surprising that comparative negligence comes in many forms; none of these forms would come in last place, and any could emerge in a divided group of voters.

This first example illustrates the idea that there is convergence up high where ethical sentiments and efficiency considerations are in sync. Virtually all legal systems find a means of discouraging negligent behavior toward strangers. This example also supports the claim that divergence arises downstream, when it comes to detail, where ethical intuitions and efficiency considerations are at odds, or where efficiency can be promoted with a variety of legal directives.[19] To be sure, the argument would be better if there were no excluded class; once there is convergence upstream, we do not get to observe whether downstream divergence proves much. My claim might seem obvious or even circular. A better example might be one with upstream divergence and then downstream convergence. It is thus convenient to point to the fact that virtually all systems converge on awarding single damages in tort law, presumably to avoid moral hazard. At first, this example of upstream divergence and downstream convergence will seem contrary to the theory advanced here, and certainly to the title of this Article, but it shows that the theory is hardly obvious or tautological. Where different rules are all efficient, legal systems can afford to be divergent, though they might eventually evolve to a single rule if there develops a shared ethical intuition in its favor. But where efficiency favors a single ruleas it does for single indemnity to combat moral hazardand there is no shared ethical intuition that contradicts this rule, we find convergence. Here we have rare cases where this convergence-divergence pattern is of the downstream-upstream form, respectively, though usually it is the reverse.[20]

II.  Other Examples in Tort Law

A.  Negligence: Strict Liability and Rescue

The choice between negligence and strict liability is perhaps the most obvious and far-reaching example of the theory developed here. Another, yet narrower, example concerns the familiar question of whether there ought to be a duty to rescue.[21] But inasmuch as both of these examples are now obvious and inadvertently addressed elsewhere, they merit just a few words. Again, legal systems must discourage avoidable and intentionally caused injuries. They can do so with a variety of approaches, and of course we find divergence among legal systems.[22] The convergence is with respect to a law that requires one actorincluding negligent and wrongful onesto pay for the loss inflicted on an innocent party. The divergence is in the choice of negligence or strict liability, and then in the treatment of exceptional cases, where a negligencebased system opts for strict liability, or where a system or subsystem that chooses strict liability as its default switches to a negligence formulation.[23] Both approaches are consistent with the idea that tort law promotes efficiency, although reasonable people can disagree as to which is superior once activity levels and enforcement costs are taken into account. This uncertainty might be said to “explain” the divergence between the two approaches. The same is true once ethical sentiments come into play, as the approaches diverge just as these sentiments do.

In the case of a failure to rescue a person in danger, or even to save another person’s property that is at risk, there is variety among legal systems[24] and also disagreement about the efficiency of the various approaches.[25] Where human life is at stake, the shared ethical sensibility is quite clear; it is widely thought that one ought (or is even bound) to rescue another where there is little danger to the rescuer.[26] Indeed, this is an excellent example of the proposition that shared ethical sentiments alone do not predict law, and certainly not convergence of law. There is less agreement about rescuing another’s property when the rescue might be unsuccessful, where rescue requires trespass, and where destruction is unlikely to be immediate. But as an efficiency matter, there is the problem, perhaps just theoretical, that a duty to rescue persons might encourage people to avoid “rescue spots,” and also the problem that so many potential rescues involve multiple parties who might truly believe that someone else has called for help; over-rescue can be inefficient or even dangerous.[27] It might also be difficult to judge when the risk is immediate; a rescuer might claim there was imminent harm when in fact professional help could have been summoned, or there was no threat of harm in the first place. Again, we see variety among laws where there is not agreement between ethical intuitions and efficiency considerations.

B.  Incapacity Exceptions

It is tempting to deflect criticism that this Article is engaging in a form of ex post data fitting by claiming that the ideas offered here fit every nook and cranny of tort law.[28] It is more useful, I think, to limit our inspection to one additional, small application in the law of torts, and then to postpone for future work an expansion to other areas of law. There is great divergence among jurisdictions when negligent behavior by children, insane persons, and other incapacitated defendants have caused injury.[29] Again, there is no shared ethical intuition, as we sometimes excuse the incapacitated person and at other times empathize with the injured victim. On the other hand, there is some de facto convergence, as when an insanity defense tends to be followed by a period of confinement about the same as the length of incarceration for more typical, convicted criminals. Note that ethical intuitions support this similarity in detention.[30] There is also no clear category or principle to separate cases where it is clear that efficiency calls for liability or an exception to the default category of negligence (or strict liability). In the case of young children, for example, liability might lead to optimal supervision by their guardians, but it might also detract from desirable learning, something courts have noted.[31]

III.  Expanding the Theory Outside of Tort Law: From Trust Games to Voting

The ideas offered here are interesting but in some areas they are hard to pin down. It becomes apparent that both efficiency and convergence are imprecise terms, and yet the theory is enticing even outside of conventional legal topics. Consider the first-year curriculum of most law schools. The schools teach torts, civil procedure, and contracts, and thus converge. They then diverge, as predicted, one or two steps downstream when it comes to the length of these courses, the cases that are assigned, the mix of theory and doctrine, and the time spent on particular subjects.[32] The argument would be that it is sensible or “efficient” to begin with these subjects because they prepare students for more advanced courses, though these courses might themselves reflect arbitrary but useful categorization; the schools could just as easily have converged on a course titled unjust enrichment or free-riding. It would be difficult to insist on shared ethical intuitions and, as for efficiency, we do not really know that the convergence in curricula, if it is that, is efficient. There is certainly no controlled experiment in which the emergence of good lawyers or even of impressive bar-passage rates is compared with that of comparable students who experience a different first-year curriculum. To the contrary, no law school appears to discriminate among applications to be admitted as transfer students based on the details of the first-year courses these applicants experienced in other schools. Here, as in so many other institutional practices, there is initial convergence because of intuitions or imitation and, more interestingly, subsequent divergence where there is no way to assess or theorize about the most efficient or ethically attractive details.

Returning to legal systems, rather than the way they are taught, it is noteworthy that they have also converged over time in the way they punish or disable convicted criminals. Nevertheless, they diverge as to the length of imprisonment. It is easy to regard this as another example of basic convergence and subsequent, downstream divergence, but much harder to associate that which converges to a combination of efficiency and shared ethical sentiments. The best we can say is that there is less agreement about these things in questions of length, rather than manner, of punishment. In some cases, efficiency is a matter of theory rather than empirical evidence, as it is for negligence and strict liability, as well as contributory and comparative negligence. The pairs are interchangeable as a theoretical matter, and that is enough for the theory advanced here. If so, we can say that, in theory, imprisonment should deter crime, but when it comes to the length of sentences there are decent arguments for shorter and longer sentences, not to mention more or less substantial gaps as one moves from less serious to more serious crimes.[33] No doubt, divergence also comes from disparate circumstances. A society with little water will probably impose greater criminal penalties for the destruction of a well; cattle rustling (or “duffing”) was treated with severe penalties in the Old West as it was in many countries, and not because the thieves were harder to apprehend. These examples of downstream divergence are best described as responding to different shared ethical sentiments or political pressures, although strained efficiency arguments can be made. Either way, law and politicians obviously respond to outcries, though we might think of these as examples of local ethical sentiments that are fairly predictable.

This optimism suggests that we ought to find divergence even on basic principles where there is neither a common ethical intuition nor empirical evidence or theoretical support for one legal tactic.[34] Many examples will come to mind, but two should suffice. Consider the choice of voting systems. There might be theoretical and even ethical reasons to expect democracy, but divergence begins almost immediately when it comes to its form. We cannot identify the best voting system and of course we find great diversity in this area.[35] The example suffers from the fact that there is a great deal of copying among jurisdictions, but even if we subscribe to the idea of looking for convergence through spontaneous ordering,[36] the divergence is impressive. Convergence is thus limited to questions that are essential for democratic systems, and this fits nicely with the idea that we find convergence where efficiency considerations (broadly defined in this case) and shared ethical sentiments converge.

The same might be said for how democracies function. Consider the decision of a city’s mayor to divert some funds from the anticipated purchase of police cars to the idea of putting some officers on bicycles. The intuition is that these officers could then better patrol a recreational area or interact with citizens. Bicycles might also help keep some officers physically fit, and citizens might simply like them or be inspired to be more fit themselves. The decision is almost impossible to evaluate. Crime or fitness may have increased or decreased after the introduction of bicycles, but so many other factors could cause these changes, and it would be absurd to attribute a dramatic change in crime rates to a small change in the mixture of police vehicles in a large city. It may be ethically and politically irresponsible to run an experiment in which the number of bicycles is randomly altered, and in any event even a sophisticated use of artificial intelligence will find it difficult to overcome the problem of omitted variables.[37] In practice, there is divergence among jurisdictions where efficiency is difficult to determine, and where there is unlikely to be a shared ethical intuition. I suspect that when a serious crime occurs in a location accessible by bike but not by car, there is the intuition that some officers on bicycles would be a good idea. For this or other reasons, there may be convergence by imitation so that we expect at least a few bicycles in every large police force.

Finally, there are subjects where other forces are likely to overwhelm the boundaries suggested here between convergence and divergence. Convergence may, for example, come about because it is of lower cost than innovation and divergence. One can hire officials or draft laws at lower cost if they have been trained or written elsewhere in a particular mode. On the other hand, divergence may be attractive because there is a market for it. Following Professor Charles Tiebout’s well-known hypothesis,[38] a jurisdiction may opt for more or less spending on a good, or on a different legal approach, in order to attract businesses or citizens that prefer that law in the mix they will encounter in the adopting jurisdiction. This form of strategic, market-like diversification might be more apparent the less fundamental the laws, but in some cases it is seen upstream with respect to basic principles, as in the promise by some jurisdictions not to have an income tax. This source of divergence seems unlikely for tort law. A business might choose not to sell a product in an area known to award very high damages in cases of products liability, but it is unlikely that a jurisdiction will gain much by choosing to apply the negligence principle, or an exemption from any liability claim at all, to products, if only because mobile firms normally expect that many of their products will be used in other jurisdictions. One jurisdiction might care more about income redistribution than another, but I prefer to think of this as a subset of what this Article has called ethical intuitionsand thus expect divergence. Though, I note that while academics associate strict liability in tort law with redistribution to poorer consumers, strict liability in fact is often capped at absurdly low amounts that promise to leave wealth in the hands of the owners of many corporations. In any event, Tieboutian logic in favor of diversity is probably better associated with a divergence in jurisdictions’ spending decisions than with threats of liability.[39]

Conclusion

Complete convergence among legal systems is implausible for several reasons. Decisionmakers have reason to copy the laws and practices of other jurisdictionsif only to avoid criticism when things go wrongbut also to distinguish themselves, and then hope for the best. Somewhat similarly, judges have good reasons to follow precedent, but also to diverge on occasion, whether openly or by asserting that a case is one of first impression. This Article has suggested a different, but coexisting, reason for convergence and divergence in tort law, as well as in other far-flung areas of the law. Convergence is associated with efficiency alongside shared ethical sentiments. Where a variety of legal techniques seem equally likely to be efficient, or where ethical intuitions do not match identifiable efficient laws, divergence is expected. A second part of the theory advanced here is that the details of laws are less likely to be uniquely efficient and also less likely to match shared ethical sensibilities. As such, there is almost always divergence as legal systems take fundamental principles and then experiment with, or simply use their experiences or political and social characteristics to fashion, details on the ground. An empirically minded optimist might look at the examples provided here and suggest that divergence offers lawmakers natural experiments, and that these will soon enough lead to more convergence further downstream. I have suggested, and will extend the argument in future work, that this intuition about the impact of empirical work is too optimistic, because as we move downstream it becomes harder to assemble useful empirical evidence. Rules may come to dominate the terrain,[40] but we should continue to expect divergence in the detailsor more technically, in areas where we cannot identify the most efficient legal rule and where there is no widely shared ethical sentiment. These claims about convergence and divergence were demonstrated in this Article with cases drawn from tort law, and in future work I hope to show not only that these claims hold true for many other areas of law, but also that they are necessarily true for the reasons hinted at here.

 


[*] *. William B. Graham Distinguished Service Professor, University of Chicago Law School. I am grateful for comments received at a University of Chicago Law School workshop, from participants in the conference memorialized in this issue of the Southern California Law Review, and over the years from Richard Epstein and Ariel Porat. Better colleagues would be impossible to find.

 [1]. Note that these intuitions can come from very different reasoning processes. Twenty citizens are likely to agree that murderers should be disabled, but they might have twenty different reasons for this view or even for the question of what incarceration will accomplish.

 [2]. As the claim is expanded beyond tort law, it is easier to find exceptions to the idea advanced here. Exceptions may well prove the rule, but it is useful or simply safer to think of the thesis developed here as limited to torts, and to expand the argument to other areas of law in the future.

 [3]. It is worth insisting at the outset that characterizing laws as binary, convergent, divergent, or multifactor is not, as we will see, the same as marking them as rules or standards. There is of course the danger that the upstream-downstream distinction is too malleable. But in many cases that support the theory offered here, the distinction seems unassailable. A country is upstream from a city; deterring theft and murder is upstream from the number of years of imprisonment a thief or murderer is ordered to serve.

 [4]. See William M. Landes & Richard A. Posner, The Economic Structure of Tort Law 3, 88 (1987) (noting that by the end of the nineteenth century, tort law “had assumed essentially its modern shape” in which the default standard of liability was negligence, and further elaborating on the role of contributory negligence within a negligence standard).

 [5]. See William K. Jones, Strict Liability for Hazardous Enterprise, 92 Colum. L. Rev. 1705, 1712 (1992).

 [6]. The prevalence of categories is a deep but manageable question, and it suggests the harder problem of justifying the placement of the negligence convention (and the convergence around it) far upstream in legal systems. The very development of a legal system, not to mention the establishment of a judiciary, can be placed yet further upstream. The characterization of rules and standards as either upstream or downstream is arguably a matter of taste and semantics, and it is not explored here; my claim about the descent from upstream to downstream, and from convergence to divergence, is more interesting than is identifying a starting point for, and the precise boundary of, convergence.

 [7]. See generally Merve Emre, The Personality Brokers: The Strange History of Myers-Briggs and the Birth of Personality Testing (2018) (describing the selling of the widely used Myers-Briggs personality test, replete with binary self-characterizations).

 [8]. Note that while I use public choice theory in explaining the binary character of many legal rules, this Article minimizes the role of public choice in developing rules and standards, and in bringing about convergence and divergence more generally. Powerful interest groups surely have great impact on law, and probably more on downstream than upstream decisions, but they hardly explain convergence and divergence quite generally. For example, something close to strict liability has become common for defective products, and that is hardly at the behest of well-organized interest groups. In other areas of law, like patents, powerful groups simply do not know in advance what rules will benefit them.

 [9]. See Saul Levmore, Rethinking Comparative Law: Variety and Uniformity in Ancient and Modern Tort Law, 61 Tul. L. Rev. 235, 235–87 (1986).

 [10]. This theory should be traced to Leo Katz, Why the Law Is So Perverse 6–7 (2011). On the implications for the binary quality of some law then developed, see generally Saul Levmore, Public Choice and Law’s Either/Or Inclination, 79 U. Chi. L. Rev. 1663 (2012) (reviewing Katz’s extraordinary book).

 [11]. See Henry M. Robert et al., Robert’s Rules of Order 42528 (11th ed. 2011).

 [12]. I ignore remands, but these too will eventually lead to a binary decision if they are returned to an appellate court. Note that when a lower court determines damages or other results, these can be binary, but the decision is made by a single judge who is not subject to cycling, as discussed presently. See Judiciary Act of 1891, ch. 517, 26 Stat. 826, 829.

 [13]. See Duncan Black, On the Rationale of Group Decision-Making, 56 J. Pol. Econ. 23, 24 (1948).

 [14]. See William M. Landes & Richard A. Posner, The Positive Economic Theory of Tort Law, 15 Ga. L. Rev. 851, 919 (1981). As a matter of efficiency, the old approach was just fine, and indeed strict liability with a defense of contributory negligence was favored by some legal academics and economists. See, e.g., A. Mitchell Polinsky, Strict Liability vs. Negligence in a Market Setting, 70 Am. Econ. Rev. 363, 367 (1980); Richard A. Posner, Strict Liability: A Comment, 2 J. Legal Stud. 205, 207 (1973).

 [15]. Note the similarity to the divergence found in the law regarding the good-faith purchase of stolen property. Saul Levmore, Variety and Uniformity in the Treatment of the Good-Faith Purchaser, 16 J. Legal Stud. 43, 44–45 (1987).

 [16]. Inasmuch as a driver needs to meet the age requirement and also pass a driving test, the rules might be described as multifactor and not binary. But I will opt for binary because there is no chance that someone would prefer an underage driver who had failed the exam. There are two requirements, but little danger of cycling. Note that this argument assumes that it took years of experience before cycling became apparent and law therefore accommodated the shared ethical intuition in favor of comparative negligence. These evolutionary stories are plausible but sufficiently doubtful. As such, I do not want to make them a centerpiece of the theory offered here.

 [17]. See A. Chalmers Mole & Lyman P. Wilson, A Study of Comparative Negligence, 17 Cornell L. Rev. 333, 341–46 (1932). To be fair, there are situations in which a single decisionmaker can face something like a cycling problem, but this twist does not alter the present argument. See Katz, supra note 10, at 25–68.

 [18]. See Landes & Posner, supra note 14, at 91920              . A related question is why change comes so slowly in some areas than in others. Why did American jurisdictions take so long to move to comparative negligence? One possibility is that judges felt free to sneak in comparative negligence by lowering damages within binary regimes. And, of course, in other areas of law where we now see convergence, U.S. law moved first.

 [19]. More detailed definitions of upstream and downstream—terms that are used rather loosely here—will emerge below. They are, inevitably, a matter of taste. “Upstream” laws are sometimes, but not always, those that pre-exist their downstream offspring; a legal system might decide to abandon contributory negligence, and then over time it might gravitate from one form of comparative negligence to another. More fundamentally, a system might install a legislature or judiciary and then be equipped to choose negligence or strict liability for wrongs done to others.

 [20]. This important example shows that the title of this Article can be misleading. The theory might better be titled “Convergence Where Efficiency and Shared Ethical Intuitions Are in Sync,” but this is less catchy and fails to suggest that we can explain divergence as well as convergence. In any event, in most areas of law, and certainly in most of tort law, the efficiency/ethical intuition match is found upstream.

 [21]. I could also include the choice between all-or-nothing and probabilistic-recovery rules. Ethical intuitions favor probabilistic recovery; a defendant who is found rather certainly to have caused a harm is commonly regarded as someone who should pay more for a given harm than one who is only 51 percent likely to have caused the injury or been negligent in doing so. There is, however, good reason to celebrate all-or-nothing directives as error-minimizing in most circumstances. I set this consideration aside both because I have addressed it in earlier work and because error minimization is not a synonym for efficiency. See Saul Levmore, Probabilistic Recoveries, Restitution, and Recurring Wrongs, 19 J. Legal Stud. 691, 692 (1990).

 [22]. See generally P.S. Atiyah, Tort Law and the Alternatives: Some Anglo-American Comparisons, 1987 Duke L.J. 1002 (noting variety in such areas as workers compensation, products liability, mass torts, and pain and suffering claims).

 [23]. Consider, for example, divergence in cases of products liability. While the United States and most European jurisdictions have adopted a strict liability approach, Canada has augmented the traditional negligence-based construction with additional remedies through contract. See Council Directive 85/374, 1985 O.J. (L 210) 29–33 (EC), amended by Directive 1999/34, 1999 O.J. (L 141) 20–21 (EC); David S. Morritt & Sonia L. Bjorkquist, Product Liability in Canada: Principles and Practice North of the Border, 27 Wm. Mitchell L. Rev. 177, 180–81 (2000).

 [24]. In general, civil law countries impose a “duty” to rescue. See, e.g., Edward A. Tomlinson, The French Experience with Duty to Rescue: A Dubious Case for Criminal Enforcement, 20 N.Y.L. Sch. J. Int’l & Comp. L. 451, 452 (2000). For example, both the French civil and criminal codes provide for a duty to rescue, with the criminal code prescribing five years’ imprisonment and a fine of 75,000 euros when individuals withhold aid from a person in danger which could be performed without risk to the rescuer or third parties. Code civil [C. civ.] [Civil Code] art. 1382 (Fr.); Code pénal [C. pén.] [Penal Code] art. 223-6 (Fr.). The Polish criminal code imposes a similar duty but limits it expressly to situations threatening an immediate danger of loss of life or serious bodily injury and imposes a penalty of up to three years’ imprisonment without an associated fine. Kodeks karny [K. K.] [Criminal Code] 1997, art. 162, sec. 1 (Pol.). The Argentine criminal code also limits liability, but bases its limitations on the identity of the would-be rescuer: specifically, it singles out those who “abandonado a su suerte” (“abandon to their fate”) an incapacitated individual in their care, or whom they have themselves incapacitated. Código Penal [Cód. Pen.] [Criminal Code] art. 106 (Arg.). The Argentine code imposes a sliding penalty of between two and fifteen years’ imprisonment based on the harm to the victim but joins Poland in foregoing additional civil penalties. Id. In sharp contrast to its civil law peers, the United States largely maintains that bystanders, though not caregivers or other singled-out parties, will not be liable for failures to rescue—joining in this respect other common law countries, including England. See Damien Schiff, Samaritans: Good, Bad and Ugly: A Comparative Law Analysis, 11 Roger Williams U. L. Rev. 77, 87 (2005); Marin Roger Scordato, Understanding the Absence of a Duty to Reasonably Rescue in American Tort Law, 82 Tul. L. Rev. 1447, 1452 (2008). On the other hand, it is common in U.S. jurisdictions to find liability for a failure to rescue when the would-be rescuer is in a “special relationship” with the one in need of rescue, as explained in the works cited in infra note 25. Thus Argentinian law, for instance, is not so different from judge-made laws in U.S. jurisdictions.

 [25]. See Saul Levmore, Waiting for Rescue: An Essay on the Evolution and Incentive Structure of the Law of Affirmative Obligations, 72 Va. L. Rev. 879, 881 (1986); see also Richard A. Epstein, A Theory of Strict Liability, 2 J. Legal Stud. 151, 190 (1973); Eric H. Grush, Note, The Inefficiency of the No-Duty-to-Rescue Rule and a Proposed “Similar Risk” Alternative, 146 U. Pa. L. Rev. 881, 883900 (1998); Richard L. Hasen, The Efficient Duty to Rescue, 15 Int’l Rev. L. Econ. 141, 141–42 (1995); Ernest J. Weinrib, The Case for a Duty to Rescue, 90 Yale L.J. 247, 269, 282 (1980).

 [26]. See Weinrib, supra note 25, at 279–93.

 [27]. See William M. Landes & Richard A. Posner, Salvors, Finders, Good Samaritans, and Other Rescuers: An Economic Study of Law and Altruism, 7 J. Legal Stud. 83, 119–28 (1978); Levmore, supra note 25, at 883–84.

 [28]. An objection to ex post selection of examples can of course be made to virtually all theorizing, whether in law or in medicine or any use of artificial intelligence. It is a danger that can only be overcome with more examples and with the promise that many of the examples offered here were chosen after the theory was stated.

 [29]. For example, substantial divergence has historically been present between United States and English jurisdictions, with the former choosing to hold children and the insane liable for tortious acts, and the latter denying their ability to be culpable in cases of tort. See Francis H. Bohlen, Liability in Tort of Infants and Insane Persons, 23 Mich. L. Rev. 9, 9–10 (1924). Still other jurisdictions, including Germany, Argentina, France, and Brazil, have adopted variations on the English view. See James W. Ellis, Tort Responsibility of Mentally Disabled Persons, 1981 Am. B. Found. Res. J. 1079, 1083.

 [30]. Eric Silver, Punishment or Treatment?: Comparing the Lengths of Confinement of Successful and Unsuccessful Insanity Defendants, 19 L. & Hum. Behav. 375, 377 (1995).

 [31]. See, e.g., A.R.H. v. W.H.S., 876 S.W.2d 687, 689 (Mo. Ct. App. 1994) (distinguishing verdicts following various claims of negligent supervision, in particular by weighing the ability of greater supervision to prevent harm); Kuhns v. Brugger, 135 A.2d 395, 401–02 (Pa. 1957) (describing the various considerations to be accounted for in determining juvenile negligence).

 [32]. Note that the definitions of upstream and downstream seem unobjectionable here. Upstream laws (or practices) are not necessarily earlier in time, but rather fundamental building blocks.

 [33]. See Saul Levmore & Ariel Porat, Threats and Criminal Deterrence in Several Dimensions, 2017 U. Ill. L. Rev. 1333, 1347–50. Nor is there reason to think that Artificial Intelligence will yield optimal prison sentences. See generally Saul Levmore & Frank Fagan, The Impact of Artificial Intelligence on Rules, Standards, and Judicial Discretion, 93 S. Cal. L. Rev. (forthcoming 2019).

 [34]. Some downstream divergence can be traced to a complete inability to know what a rule accomplishes. Thus, we are not surprised that there is divergence in religious beliefs or even in burial practices—and it is obvious that practitioners are unable to discern the efficiency (or after-life effects) of different rules.

 [35]. For an overview of electoral systems, and an analysis of the axioms driving their selection, see generally Pippa Norris, Choosing Electoral Systems: Proportional, Majoritarian and Mixed Systems, 18 Int’l Pol. Sci. Rev. 297 (1997).

 [36]. See generally Richard A. Epstein, How Spontaneous? How Regulated?: The Evolution of Property Rights Systems, 100 Iowa L. Rev. 2341 (2015) (discussing the relationship between spontaneous order and natural law).

 [37]. Levmore & Fagan, supra note 33.

 [38]. For Professor Tiebout’s seminal article on this hypothesis, see generally Charles M. Tiebout, A Pure Theory of Local Expenditures, 64 J. Pol. Econ. 416 (1956) (theorizing that individuals will move to communities with public goods provisions that suit their personal preferences and vice versa).

 [39]. See generally Saul Levmore, Interest Groups and the Durability of Law, in The Timing of Lawmaking 171 (Saul Levmore & Frank Fagan eds., 2017) (emphasizing lawmakers’ interest in appealing to a particular population’s preferences and doing so in a way that will last for some time).

 [40]. See Levmore & Fagan, supra note 33.