Edited by Gerrit De Geest
Chapter 8: Bias in the common law
[In: Volume 7, Francesco Parisi (ed) Production of Legal Rules]
According to the efficiency of the common law hypothesis, common law rules (i.e. judge-made law) attempt to allocate resources efficiently, typically in a Pareto or Kaldor-Hicks efficient way. The common law process enjoys a comparative advantage over legislation in the creation of efficient law. Initially, the arguments in favor of the efficiency hypothesis were based on examination of particular legal doctrines (see Posner, 1973; Ehrlich and Posner, 1974). Because conclusions regarding the efficiency of a particular rule depend on intuition about the relative magnitudes of costs, many economists were uncomfortable with the efficiency argument (see Rubin, 2005a). Some of them have – more in line with the standard method in economics – sought to identify a process that would explain Posner’s obervation that the common law was efficient. Several models of efficient legal evolution were created (e.g. Rubin, 1977; Priest, 1977). The essence of these models is that inefficient laws will be litigated more frequently than efficient laws because the former impose larger costs on victims. Consequently, inefficient laws will come before the court for re-examination more often, resulting in a trend toward efficiency. Other scholars soon began to critically examine these models. For example, Landes and Posner (1979) argued that litigation might strengthen or weaken a precedent without overturning it completely and showed that this consideration weakens the evolutionary pressures for efficiency. Hirshleifer (1982) showed that the law could come to favor whichever party can most easily organize and mobilize resources for litigation of unfavorable precedents and that this movement would be independent of efficiency. Zywicki (2003) added a “supply side” (i.e. competition between several court systems) to the existing “demand side” (i.e. litigants demanding efficient rules) evolutionary models. He argued that when competition is lacking, only weak or even non-existing tendencies for efficiency will exist.1 Some of the criticizing models reveal the possibility of p. 133an evolutionary bias in the production of legal rules by courts (e.g. Rubin and Bailey, 1994; Bailey and Rubin, 1994; Whitman, 2000; Fon and Parisi, 2003; Fon, Parisi and Depoorter, 2005; Hylton, 2006; Gennaioli and Shleifer, 2007a and 2007b; Miceli, 2009; Depoorter, 2010; Miceli, 2010; Luppi and Parisi, forthcoming). They mainly focus on mechanisms that may bias legal evolution toward expansion or contraction. These models are the center of attention in this chapter. In the following sections, we will look at the different sources of evolutionary bias: the long-term interests of tort lawyers (Section 2), judicial bias (Section 3), judicial path dependence (Section 4), Asymmetric information (Section 5) and the feedback effect of settlements (Section 6).
2. The Long-term Interests of Tort Lawyers
Rubin and Bailey (1994) argue that the shape of tort law, and more specifically modern product liability law, can be explained by the long-term interests of tort lawyers. They provide several reasons why the interests of attorneys in precedent will generally be larger than the interest of manufacturers and insurance companies. The costs of a firm’s liability raise the prices of the goods and services the firm sells. Firms only have a transitional interest in changes in liability (unless the transition is so extreme as to bankrupt the firm). Customers also lack an interest. The cost borne by any one customer is usually small, and problems of organizing consumer groups would prevail. Next, while plaintiff’s attorneys are potentially interested in all products, a defendant’s attorney is interested in only one product and its history. Consequently, a plaintiff’s attorney is much more interested in the stated grounds of the decision than is the defendant. Also, there are limits to the ability of firms in different industries to join together to organize for the purpose of litigation, since different types of manufacturers have different interests (see Epstein, 1988). Even for firms within the same industry, successfully organizing may be difficult. When a manufacturer is sued, other manufacturers may be ambivalent about the outcome: a loss will change the legal rule in an unfavorable direction, but it will also harm the reputation of a competitor. Of course there are some issues of interest to all businesses, such as the locus of lawsuits (businesses favor federal rather than state jurisdiction), procedural rules (e.g. limits on class actions), and the amount and form of damages (e.g. all firms may be interested in severely limiting punitive damage payments) (see also Rubin, 2005b). Next, there may be agency problems between businesses and their attorneys. Defense attorneys gain when the law becomes more complex and when liability suits become more likely. Such problems need not exist between plaintiffs’ attorneys and their clients. If the sum of expected values of litigation is positive, then the fee may be adjusted so that the client will also prefer litigation. With respect to insurance companies, their interests are mixed at best. In the short run, they may lose from product liability litigation, but, in the long run, they would expect to gain p. 134business from a more costly products liability system (see also Viscusi, 1991). Rubin and Bailey argue that tort lawyers are organized in exactly the way one would expect an interest group interested in shaping the law to be organized. The Association of Trial Lawyers of America (ATLA) provides several private goods to its members which are exactly related to its law-changing mission (e.g. information about techniques, data and methods of litigation). The pooling of information that occurs through the auspices of the ATLA both increases the chances of each lawyer winning a particular case and helps create precedents favorable to other lawyers working in the same area (see Rubin, 2005b).
Rubin and Bailey argue that the forces they identify will lead to increased damages payments, reduced consideration given to private agreements, and increased uncertainty and unpredictability in the law. All of these benefit lawyers by increasing the amount of litigation. This can explain the replacement of contract by tort in the context of products liability and other examples of courts’ unwillingness to enforce freely agreed on contract terms (e.g. the treatment of “unconscionability”), the weakened use of custom as a defense in cases where there is a contractual relationship between potential injurers and potential victims, the movement toward allowing payments for non-pecuniary losses, the replacement of contributory negligence with comparative negligence and the replacement of the rule of no contribution among joint tortfeasors with a rule of contribution. The authors also discuss some statistical evidence in support of their hypothesis. First, states with more attorneys per capita were likely to adopt comparative negligence sooner, with the number of lawyers leading the decision to adopt (see Curran, 1992). Second, after re-examining some results from Landes and Posner (1987), Rubin and Bailey find that states with more lawyers per capita were likely to reject privity sooner. Third, lawyers have benefited from the legal changes. The Association of Trial Lawyers of America was founded in 1946 and grew considerably in the next decades. Pashigian (1977) found that real earnings of lawyers increased markedly from the middle 1950s and that the increase in demand for lawyers cannot be explained by government regulation. According to Rubin and Bailey, the demand for lawyers will continue to grow as long as plaintiffs’ attorneys are successful in continuing to change legal doctrine.
3. Judicial Bias
3.1. Judicial Bias and Plaintiff Case Selection
Fon and Parisi (2003) provide an adverse selection model of legal evolution to explain the gradual expansion of legal remedies, and the resulting tendency of legal systems to grant ever-increasing levels of relief for plaintiffs’ claims. The main assumptions of the model are the following. The parties face symmetric stakes and cases are filed whenever the expected value from the case is positive. p. 135Plaintiffs thus decide whether to file suit based on the likelihood of success in a specific court. The expected value of a case depends on the current state of the law. For some marginal cases, which are generally outside the domain of past rules and emerge out of novel or borderline cases, expected value is also affected by the ideological views of the judge. Judges are assumed to have different policy perspectives and different propensities to expand the scope of legal remedies and causes of action. Although judges recognize the binding force of past precedents, their policy views can affect outcomes in new or borderline cases. Litigants form rational estimates of likely decisions. Plaintiffs (and their lawyers) have estimates of judges’ biases and take these ideological components into account when evaluating the expected value of their cases. From these assumptions, it follows that conservative and liberal judges are asymmetrically predisposed toward marginal cases. When given an opportunity to do so, all judges, both liberal and conservative, would push the threshold of liability closer to their ideal point. However, liberal judges have a greater opportunity to create pro-plaintiff precedents than conservative judges have to create pro-defendant precedents. Given rational choices by plaintiffs, marginal cases that are not supported by current case law are unlikely to be brought before a conservative court, but may be filed before a liberal court. The marginal cases, once adjudicated, may then affect the future state of the law. Subsequent cases are adjudicated according to new standards set forth by past marginal precedents. The biased flow of precedents in favor of plaintiffs will over time affect the interpretive standards of future judges, liberal or conservative. In the long run, this changes the liability threshold necessary for a plaintiff to obtain judicial relief. Whether the liability threshold will stabilize at a given point depends on whether ideology is absolute or is itself shifting over time. If liberal judges do not become more liberal, at some point no new precedents will be set. If ideology is not absolute, the interpretive standards of liberal judges may become more liberal over time and this will lead to an ongoing monotonic increase of remedial protection in the legal system.
Although Fon and Parisi mainly focus on expansion of legal remedies, they stress that their model is agnostic on the actual direction of legal evolution. Depending on the initial environmental conditions, the adverse selection mechanism may bias legal evolution toward expansion or contraction. In the case where defendants have better control of the jurisdiction, a gradual shrinking of remedial protection may be observed.
Luppi and Parisi (2010) discuss some effects of anonymous judicial decision-making, including the effect of the possibility of adverse selection on the selection of disputes brought for litigation. More anonymous decision-making will thwart adverse case selection and weaken the tendency of legal remedies to expand in number and scope of application. All else equal, adverse case selection should be weaker in civil law jurisdictions than in common law jurisdictions. p. 136The reason is that several decision drafting practices in civil law jurisdictions make it difficult to discern the ideology of individual judges. While signed opinions and concurring and dissenting opinions characterize common law adjudication in appellate and supreme court cases, in most civil law countries panel decisions are generally taken on a simple majority vote and dissenting and concurring opinions are generally not admitted.
3.2. Judicial Bias without Plaintiff Case Selection
Some articles have modeled the impact of judicial bias on legal change, without relying on an adverse selection mechanism. Whitman (2000) provides a model of self-interested decision-making by judges. In the model, an endless succession of cases arise in sequence within a certain area of the law. For each new case, a judge is selected at random from a pool of judges. This judge takes as the relevant precedent the rule announced by the last judge to decide such a case. The current judge must decide whether to follow the precedent or to announce an alternative rule, which will then constitute the precedent faced by the next judge. If a judge chooses to break from precedent, she experiences “preference satisfaction” (see Kuran, 1990) and reaps a reputational gain if the next judge upholds her decision or a reputational loss otherwise. Judges are taken to be expected utility maximizers. There are two rules competing for the attention of judges. A certain proportion of judges favor rule 1, the others favor rule 2. When a substantial majority of judges favor one rule over the other, the division of opinion is said to be low. When the two rules command more or less equal support among judges, there is greater division of opinion. The ratio of the greatest potential gain from breaking precedent to the total utility at stake in the gamble of breaking precedent is treated as a measure of the strength of a judge’s activist impulses. Whitman finds that if division of opinion among judges is low relative to the strength of judges’ activist tendencies, then the system will converge on a single rule. The reason is that even activist judges who disagree with a precedent will follow the precedent anyway, for fear of rejection by subsequent judges. If division of opinion is high, however, oscillation between the two rules will take place. Whitman also revises his model by adding a third, compromise rule into the mix. No judge considers it the best rule, but all judges consider it better than the alternative. Whitman finds that the legal system is most likely to converge on the compromise rule when opinion is divided. Note that the judges need not explicitly agree to announce only the compromise rule. The compromise emerges out of a decentralized process of judge-by-judge decision-making and does not need an external enforcement mechanism.
Gennaioli and Schleifer (2007a and 2007b) assume that judges are principally efficiency-seeking, but that they may be biased in favor of one or the other side in a dispute. They examine the direction of legal change when judges are bound by precedent (in which case legal change can only occur if judges distinguish their decision from existing precedent) and when judges are not bound by p. 137precedent (they can change the law directly by overruling previous decisions). In the first case, judicial bias distorts the law away from efficiency, but it can improve the precision of law in the long run. In the second case, they find that selective litigation cannot counteract the detrimental effect of judicial bias when the bias is strong enough. Miceli (2010) criticizes this result, however, because Gennaioli and Shleifer assume that the divergence of beliefs by litigants about trial is unrelated to the nature of judicial bias. This treatment of the impact of selective litigation does not allow a careful examination of its interaction with judicial bias in determining the direction of legal change.
Miceli (2010) focuses on the interaction between judicial bias and the (Rubin-Priest) selective litigation effect.2 He considers a unilateral care accident model in which potential injurers and victims interact a fixed number of times over a given time interval and each interaction is a potential accident. Injurers choose a level of care in anticipation of each interaction. There are two rules for allocating liability: strict liability and no liability. Initially, both rules exist within the population of legal rules in some arbitrary proportion. Each potential accident involves just one of these rules. Both parties know with certainty the prevailing rule that applies to their particular interaction. Judges have biases (either pro-plaintiff or pro-defendant) and are imperfectly bound by precedent, so they will occasionally overturn laws based on their preferences. The probability that a judge will overrule a precedent is exogenously determined. Since there are only two rules, overruling means replacing strict liability with no liability and vice versa. Consequently, as time unfolds, the process of litigation will cause the distribution of legal rules to evolve as cases come before the court to be adjudicated. One of Miceli’s main goals is to find the ultimate distribution of legal rules (more formally, the steady state equilibrium distribution).3 Miceli finds that the ultimate proportion of strict liability rules depends on the relative number of trials that arise under each of the two rules, and on the distribution of judicial bias. Logically, a larger fraction of pro-defendant judges leads to a smaller ultimate proportion of strict liability rules. Likewise, a larger relative number of trials under strict liability leads to a smaller ultimate proportion of strict liability rules.4 Intuitively, the more often a rule comes to trial, the more p. 138chances it has to be overturned by a judge and replaced by another rule. The strength of precedent affects the rate of legal change (how long it takes to reach the steady state), but not the long-run distribution of legal rules. Only in some special cases will the law converge to a single rule (to strict liability when all judges are pro-plaintiff and to no liability when all judges are pro-defendant). Generally, the common law will embody multiple rules, reflecting the biases of judges and the selection of cases for trial.
4. Judicial Path Dependence
Fon, Parisi and Depoorter (2005) examine the effect of judicial path dependence on the consolidation and contraction of liability rules and legal remedies. They highlight the importance of an often overlooked condition for the emergence of a relevant legal dispute: cases that may lead to litigation must have a positive expected net return. This condition may create a bias in the evolution of case law. This result does not, in contrast to the one in Fon and Parisi (2003), rest on judges’ ideological decision-making but complements the results reached in the earlier literature. The authors use the following assumptions. First, a judge does not consider himself bound by a single decision in a single previous instance, but considerable authoritative force stems from a consolidated trend of decisions on a certain point (jurisprudence constante; see e.g. Lambert and Wasserman, 1929; Dennis, 1993; Dainow 1974; MacCormick and Summers, 1997). Consequently, if the rate of positive judgments with respect to some new legal issue or interpretation of existing causes of action falls above a critical institutionally determined threshold, the recognition of such legal claims in future disputes will be facilitated by the presence of legal authority. Positive precedents are assumed to have equal weight with negative precedents. Second, potential litigants form rational estimates of the probability of success in litigation and take them into account when evaluating the expected value of their cases. The expectations of the parties are unbiased, but have some margin of error (which explains why some disputes are litigated). Third, the plaintiff obtains a judicial award (W) when he wins in court and suffers a prejudice (L) in case of a verdict in favor of the defendant. The prejudice can be seen as an immediate cost imposed on plaintiffs or any other liability imposed by the court in case of unsuccessful action by court sanctions or defendants’ counterclaims, or as the net present value of the loss from litigation in similar cases in the future (as in Rubin, 1977). Plaintiffs also face direct litigation costs (C). Fourth, plaintiffs are rational in deciding whether to pursue litigation. Thus, the expected net judicial award should be positive and should exceed the settlement amount offered by the defendant. Based on these assumptions, the process of legal evolution can be described as follows. With relatively high win-loss ratios (W/L), cases can be rationally filed even when the probability of success is small. Consequently, a large number of negative precedents may p. 139be produced. When the percentage of positive precedents falls below the level of support that the legal system considers necessary before widespread judicial recognition occurs, an initial wave of filing may be followed by a gradual implosion. Conversely, a small fraction of early favorable decisions could lead to wider acceptance and eventually be consolidated into a binding doctrine. It’s easy to see that an increase in the win-loss ratio and a decrease in litigation costs, given a fixed win-loss ratio, increase the scope for gradual contraction of remedies, since they render smaller probability cases worthy of pursuit. A change in the institutional weight of past precedents may have a substantial impact on the stability of remedies and on the direction of the process of legal evolution. For example, an increase in the level of case consistency required for an emerging jurisprudential trend to become binding case law may reduce the likelihood of gradual consolidation and undermine conditions for legal stability.
Building further on this framework, Luppi and Parisi (forthcoming) study the effect of the American and the English rule of cost allocation on the path of evolution of judge-made law. Under the American rule, each party is responsible for paying his own litigation costs. Under the English rule, the losing party pays (at least some) of the legal expenses of his winning counterpart. Since the American and the English rule affect the typology of cases filed for litigation, they will have different effects on legal evolution. Under the American rule, some cases with small probability of success are filed, although they would not be filed under the English rule. The reason is that the American rule allows losing plaintiffs to impose an uncompensated externality on the defendant. Over time, these low probability cases may result in the accumulation of negative precedent and lead to a decreased likelihood of success for similar cases in the future. This can generate bubbles of litigation and the possibility of contractionary trends under the American rule that would not be observed under the English rule. Conversely, some low-stake cases with a higher probability of success will be filed under the English rule, but not under the American rule. The filing and adjudication of these cases will expedite the consolidation of judge-made rules (when the higher probability of success hinges upon a question of law). The higher probability of success means that a larger flow of positive precedents will be generated, with a resulting path dependence in the evolution of case law.
5. Asymmetric Information
Hylton (2006) provides a general model that includes both the Priest-Klein model (see Priest and Klein, 1984) and asymmetric information models (e.g. Bebchuk, 1984)5 as special cases. His model shows that an apparent bias in the legal standard can occur and trends can develop favoring the better informed p. 140litigant (i.e. the litigant who makes a superior prediction of the case outcome) whose case is also meritorious. The model points out that the direction of the law is influenced by the “litigation likelihood ratio” of innocent to guilty litigants. In the Priest-Klein case, a challenge (of a practice as negligent) is most likely to go to judgment if the standard’s application to the particular practice is highly uncertain. Given that the rational components of the litigant’s predictions are the same, the plaintiff and the defendant appear equally likely to win and the litigation likelihood ratio is thus equal to one. Consequently, the law is equally likely to move in a direction favoring plaintiffs as it is to move in a direction favoring defendants (see also Priest, 1980). In cases involving asymmetric information, the relative frequency of litigation favors the party who is informed and has the strongest case. The reason is that informed and non-meritorious parties (e.g. guilty defendants) tend to settle, which leaves a relatively large share of informed and meritorious litigants in the pool of cases litigated all the way to judgment. In the asymmetric information setting, existing legal rules are thus shaped by the information provided by innocent litigants in court. To make it more concrete, Hylton considers the example of a medical malpractice claim. Suppose the doctor is far better informed than the plaintiff as to the doctor’s potential compliance with the legal standard (negligence). A priori, the negligence standard is somewhat ambiguous. The model implies that, over time, the negligence standard in medical malpractice is infused disproportionately by the information provided by innocent doctors. The case law will be “informationally biased”: it tends largely to identify specific types of non-negligent conduct.
Hylton describes his model as one of “short-run evolution” because it focuses on short-run changes in the information content of the legal standard. However, the informational bias could lead to “rule evolution” over time, as the information embodied in legal rules alters the nature of the rule itself. Returning to the medical malpractice example, negligence determinations are made predominantly by referring to the custom of the profession. The emergence and resilience of the custom rule may be because the case law, summing up so many specific types of conduct deemed to be non-negligent, has generated the custom rule to supplant the (relatively ambiguous) negligence test.
The information biasing that occurs in Hylton’s model seems consistent with the common law efficiency hypothesis: the party in a legal dispute who is likely to be in the best position to improve the efficiency of a legal rule is the party that is both informed and meritorious. However, as Hylton explains, the model does not suggest an unambiguous trend toward efficient legal rules.
6. The Feedback Effect of Civil Settlements
Depoorter (2010) highlights the influence that civil settlements have on legal change when (1) settlement outcomes do not simply reflect the expected p. 141outcome at trial and (2) information on settlements is not confined to the parties involved in the settlement. Regarding the first condition, if settlement outcomes simply reflect the expected outcome of trial, then settlements are not relevant as a source of law. However, it is now widely understood that various factors unrelated to the substantive merits of a legal dispute often influence negotiated outcomes (e.g. Mnookin and Kornhauser, 1979). These non-legal factors include the ability to bear litigation costs, diverse attitudes toward risk, asymmetric information, the principal–agent relationship between an attorney and his or her client, cognitive limitations and a party’s sensitivity to the negative public attention that a lawsuit generates. Such factors may cause the terms of private settlement agreements to be below, but also above, legally available awards. Consequently, settlement trends may provide remarkable departures from the conventional expectation of the available legal remedies. Regarding the second condition, Depoorter argues that despite the widespread use of non-disclosure agreements, information on settlements is available to those legal professionals for whom such information is most valuable. Information on settlements is distributed through various channels, reaching actors both inside and outside legal communities. The channels include specialized reporters, professional interest organizations, mass media coverage and the oral culture in legal communities. An incentive analysis of the parties involved leads Depoorter to conclude that information on novel settlements featuring high awards and novel remedies is especially widely distributed. For example, special interest groups focus on extravagant settlements that are most likely to induce legislative action, media reports highlight spectacular tort awards and professional organizations bring attention to novel settlements that are strategically important to lawyers.
In view of these observations, Depoorter argues that novel civil settlements have a feedback effect on the path of the law. First, individual settlement concessions make it more difficult for similarly situated defendants to deflect forthcoming claims. Ambitious trial lawyers will use innovative settlements as benchmarks, making plaintiffs in future cases more reluctant to accept settlements below what others have agreed to in prior, analogous cases. Second, settlements may frame the normative outlook on particular claims, due to their non-coercive nature. A novel legal claim is less likely to be perceived as outrageous if such a claim has been gratified by a prior concession in a settlement agreement. Judges may interpret settlement precedents as expressive statements regarding the appropriateness of compensation. A defendant who is unwilling to settle when the settlement proposal is comparable to concessions made by other defendants in similar cases risks receiving little sympathy from a judge or jury in subsequent proceedings. Prior settlement concessions may thus create sustained pressure towards higher awards and novel remedies, resulting in a gradual expansion of the legal system. Depoorter stresses, however, that in some instances, defendants might be able to drive down legally available awards over time. For example, in p. 142concentrated industries where tort settlements are handled by a few select lawyers or defense firms, we can expect that selective use will be made of information on settlements that are below the going rate at trial. The feedback effect does not work to the exclusive benefit of claimants and trial lawyers.
The law and economics literature has provided several reasons why judge-made law may become biased over time: the long-term interests of tort lawyers, judicial bias (with or without plaintiff case selection), judicial path dependence, asymmetric information and the feedback effect of settlement. Most of this literature is quite recent. There is still room for exponential growth in this field of academic research. Future studies could use empirical data to validate the predictions of the various models and to discover their relative importance. Theoretical extensions could examine the effect of a combination of adverse selection and asymmetric stakes, could study the interaction between ideological judicial intervention and path dependence in judicial action, could analyze the effect of complex doctrines of stare decisis and percolation theories, could examine in greater detail the influence of short-term information biasing on the long-term pressure toward inefficient rules when some parties are better able to devote resources to litigate in favor of their preferred rules etc.
For extensive reviews of the early evolutionary models and critical analyses of these models, I refer to the chapters by Paul Rubin and Todd Zywicki and Edward Stringham.
Note that the equilibrium distribution of rules is a description of the overall state of the legal system at a point in time, as seen by an outsider. The equilibrium does not have any relevance to litigants, who know the particular rule in place in their jurisdiction. The model can best be seen as a way of describing how an aggregate index of the legal system emerges from individual optimizing behavior.
Note that the number of trials may be larger or smaller under strict liability compared to no liability. Strict liability leads to fewer accidents, but results in a higher probability of trial given an accident (since strict liability makes trials more valuable).
To simplify the analysis, Hylton (2006) dispenses with modeling strategic behavior.