BARGAINING IMPEDIMENTS AND

SETTLEMENT BEHAVIOR
 
 

Samuel Issacharoff, Charles Silver, and Kent D. Syverud

---

INTRODUCTION

There is considerable popular support for two propositions: too few disputes settle, and too many that do settle drag on for too long. There is likewise a widespread tendency to hold lawyers largely, if not exclusively, responsible for both problems. The popular belief - reflected in the Republican Contract With America, the Attorney Accountability Act, the Manhattan Institute proposal to reform contingent fees, the pronouncements of conservative commentators, and the remarks of then Vice-President Quayle to the American Bar Association - is that lawyers foment controversy and prolong litigation because they make money by doing so. The logic is simple and, for many politicians and voters, it is compelling (Galanter 1993; Margolick 1991; Mullenix 1994).

Although academics may be less inclined than the general population to blame lawyers for these problems, some, especially those involved in the development and implementation of alternative dispute resolution techniques, endorse the popular wisdom about the protractedness of litigation and the frequency of trials. For these scholars, it is almost an article of faith that litigation occurs too often, lasts too long, and costs too much. The numbers enable them to make a credible case. For example, an insurance industry publication indicates that 12 percent of every dollar of earned premium from private passenger automobile insurance is spent on lawyers' fees, with plaintiffs' attorneys and defense lawyers dividing the spoils in nearly equal shares (Mooney 1994). Another striking example can be found in a famous Rand study of asbestos litigation, which determined that claimants netted only 37 cents of every dollar spent by both sides, the remaining 63 cents being consumed as litigation costs (Hensler 1985; Kakalik 1983.)

The numbers are not entirely one-sided, however. Studies suggest that trials occur in only a tiny fraction of tort disputes, that the majority of disputes that make their way into the legal system settle fairly quickly and with little or no discovery, and that exceptional cases in which trials occur or extensive discovery is undertaken tend to be complicated, high-dollar, multiparty affairs (Insurance 1994; Keilitz 1993; Mullenix 1994; Smith 1995). These studies suggest that an alternative economic logic is at work, one linking the level of resources devoted to litigation to the amount in controversy and the complexity of a dispute (Galmiter 1983).

Economiists of law have also considered why lawsuits proceed to trial. According to the most "widely-accepted economic model, litigation is a negative-sum game for litigants (Gould 1973; Miller 1987; Posner 1992). The longer play continues, the less the participants aggregate wealth because they must expend on litigation money they could save by settling. Even if five percent or fewer of all controversies that makeit into the courts ever get to trial, it is a puzzle to explain why so many disputes reach the point of litigation and why so many lawsuits are tried.

Economists and others who have applied the model offer a variety of explanations for the frequency of protracted litigation and trials. One view is that trials occur because disputants make errors when estimating their chances of winning, especially when the alleged wrongdoer's level of culpability is near the level at which liability will be imposed (Priest 1984). Although this view continues to have supporters, an empirical study of tried cases casts its validity in doubt (Gross 1991). Another possibility is that because the legal system systematicallv underproduces information about legal norms, parties make mistakes identifying applicable legal standards and are unable to agree on likely trial outcomes as a result (Silver 1996).

Economists have also supplied an answer that echoes the popular wisdom about lawyers. Some scholars argue that protracted litigation and trials occur because lawyers are compensated on the basis of the amount of time they spend on a case (Clermont 1978; Coffee 1987; Utan 1992; Mnookin 1993). This opinion puts at least some of the blame onopportunistic lawyers. A different view that holds lawyers responsible but not blameworthy is that litigation involves parties in economically wasteful "Prisoner's Dilemmas" (Ashenfelter 1993). Each party finds it rational to employ a lawyer-agent because doing so improves a party's expected outcome in litigation. However, when both sides are represented, neither gains a net advantage because the efforts of the opposing lawyers cancel out. This explanation is delightfully perverse. In contrast to the preceding explanation, which argues that litigation costs are high because lawyers disserve their clients, this explanation suggests that lawyers are responsible because they serve their clients too well. If clients gained no advantage from having lawyers, the dominant strategy of obtaining representation would break down and the problem of high litigation costs would disappear.

In all likelihood, trials and protracted litigation occur for many reasons, only some of which have been touched upon. In this chapter we will offer our views on certian aspects of the forces that lead to litigation and summarize the results of some ongoing empirical work on settlement. Our purpose will be to shine light into the black box that is the economic model of the decision to settle or sue and to suggest ways in which the model may need to be made more complex. The model leads one to wonder why protracted lawsuits and trials occur partly because it addresses neither the way participants in litigation process information about expected trial outcomes nor the dynamics of the bargaining games participants must play before settlements can be worked out. Empirical evidence, including experiments performed by one of the authors, suggests that participants in lawsuits often use information in faulty, biased ways, so that their estimates of trial outcomes make settlement less likely. Economic reasoning suggests that settlement games can have complex dynamics even in relatively simple disputes. It is easy to see that lawsuits may often become protracted and perhaps fail to settle when these dynamics are understood. Empirical study of settlement practices sheds further light on why settlements sometimes fail to occur.  When discussing the dynamics of settlement, we will also suggest that lawyers are as much responsible for the high rate of settlement that prevails as they are for the frequency of trials and protracted litigation. Our experience studying disputes and the legal profession leads us to think that it is at least as accurate to describe litigating lawyers as settlement engineers or "friends of the deal" as it is to characterize them as pit bulls. Each of us is familiar with many, many instances where lawyers worked hard to settle complicated, multiparty disputes that posed extremely difficult questions of liability and damages. They did not always succeed, but their willingness to work hard for sefflement is worth noting. A real difficulty is accounting for their failure. If lawyers are at least partially motivated to settle, and if litigation is a negative-sum game for litigants, how can settlement not occur? We hope to shed light on this and related questions below.

In the first section of this chapter, we will summarize evidence bearing on the manner in which people process infomiation about issues in contested litigation. The economic model bases the likelihood of settlement on the expectations of the parties. The processes by which these expectations form are more complicated than one might suspect, and they are worth studying in depth. In the second section, we talk in general terms about the dynamics of settlement games. These games must be played even when parties' expectations align. They can be quite involved, even in relatively simple disputes. By discussing a single example in some detail, we hope to convey a sense of that complexity and to suggest that lawsuits may often become protracted, despite the participants' willingness to settle, because settlement games are difficult to plav. Finally, in the last section, we will sumniarize the results of an empirical study of settlement practices that tests the accuracy of the economic model of settlement.

BIASED JUDGMENTS OF FAIRNESS AND THE BARGAINING PROCESS
The Priest-Klein Hypothesis

The most commonly cited and frequently tested explanation for nonsettlement in the law and economics literature is the "selection hypothesis" (Priest 1984; 1985). [Fn 1] This model posits that parties possess imperfect information about some aspect of the case that causes them to estimate its value with error. [Fn 2] The source of the imperfect information may be underdeveloped legal norms governing the conduct inquestion, or misperceptions about the facts of the case. Alternatively, the source may be idiosyncratic views of the litigants on the merits of the dispute. The elegance of the model stems from two sources. First, the litigation process itself helps to resolve uncertainty that impedes tfie ability of parties to settle. Both pretrial motions to dismiss and motions for summary judgment provide joint clarification of the legal standards that will apply in any dispute. This process is extended by the publication of opinions that clarify the operative legal rules for future litigants. In addition, discovery creates a joint body of factual information that should also facilitate a convergence of views about the likely outcome should, the case proceed to trial.

Second, the process of facilitating a convergence of assessments of the facts and the operable law means that cases that proceed to trial do so because of real uncertainty or because of perverse litigant behavior. Cases that fail to settle are those where the plaintiff overestimates and/or the defendant underestimates the expected value of going to trial. Adhering to conventional assumptions in economic modeling, the selection hypothesis assumes that estimation errors are random - that is, they equal zero on average. Cases that go to trial are simply the unlucky ones in which the plaintiffs and defendant's errors compound to eliminate the settlement range that would otherwise arise from the risks and costs associated with trial. [Fn 3] Accordingly, the Priest-Klein model predicts that plaintiff success rates at trial should be about 50 percent.

The Priest-Klein model has been tested with field data and experimentally. [Fn 4] One study done byLinda Stanley and Donald Coursey, for example, had subjects bargain over the division of 100 valuable tokens (Stanley 1990). If the subjects were unable to determine a division during the allotted negotiation period, all the tokens were given to one subject or the other depending on whether an urn containing 100 red and white chips contained more red or more white chips. Nonsettlement costs were imposed on both parties. Before negotiating, each subject privately sampled a certain number of chips from the urn. By manipulating two variables - the number of chips each subject sampled and the magnitude of nonsettlement costs-it was possible to test the Priest-Klein model's prediction that settlement rates will increase with the number of chips drawn from the urn because the variance of subjects' predictions is larger with a smaller sampling of the chips.[Fn 5] While this portion of the Priest-Klein hypothesis was borne out, the study by Stanley and Coursey uncovered a failure of the parties to behave as economic models would have predicted (Stanley 1990). The study failed to confrom the expectation that settlement rates would increase with legal costs. Stanley and Coursey discussed a variety of possible explanations for this result, including the possibility that increasing the settlement range, which would increase the scope for settlement, also increases the range of indeterminacy and the scope for bargaining. However, they also speculated that although "the experiment above was not designed to study the psychological effects of the negotiation process ... these effects may detemine, in part, the outcome of the negotiations" (Stanley 1990).

By focusing on the psychological or behavioral factors involved in dispute resolution, it is possible to give a far more robust picture of why disputes fail to settle. Whereas the common assumption in economics is that errors of judgment will not be systematically biased [Fn 6],  such biases have been the active focus of recent research in psychology. [Fn 7] It has been documented that some biases in probability judgments are not eliminated by incentives for accuracy or feedback (Camerer 1987; 1989). Such biases include overweighing of vivid information; revision of probability estimates more radically in some situations, and less radically in others, than is called for by Bayes' theorems [Fn 8]; failure to expect regression to the mean (and misattribution when it occurs); misapplication of the "law of large numbers" to small samples; and, a tendency to think that plausible conjunctions of two events are more probable than is either event alone. These biases are widely believed to result from the use of judgmental heuristic - cognitive rules of thumb which are naturally adapted to limited human information-processing capabilities - instead of optimal statistical rules.

Self-Serving Bias and Fairness

Of particular concern are those biases that are likely to alter systematically the expectations of parties in such fashion as to compromise or defeat the capacity of parties to reach economically efficient bargains. These fall primarily into the category of what are termed "egocentric" or "self-serving" estimation biases. These are readily observed in a number of settings. For instance, self-serving bias is evident when seminar participants overestimate the amount of time they speak relative to estimates by other participants. In two-person discussions, both people typically believe they speak more than half the time. Another example can be found in the estimates each spouse in marriage has about how much work they do around the household. When asked to guess the fraction of various household tasks they are responsible for, married couples routinely give estimates which add up to more than 100 percent (Ross 1982).

Egocentric biases have also been observed in dispute settings. A classic study measured student perceptions of a contentious football game between Princeton and Dartmouth (Hastorf 1954). Students from both schools watched a film of the game and rated the number of penalties committed by both teams. Princeton students saw the Dartmouth team commit twice as many flagrant penalties and three times as many mild penalties as their own team. Dartmouth students, on the other hand, recorded an approximately equal number of penalties by both teams. Team allegiance influenced the students' perceptions of penalties. It is not difficult to speculate that such biases were at least partially responsible for the marked divergence in reaction of blacks and whites to the acquittal of 0. J. Simpson. A second source of deviation from economically-expected behavior is captured by the conception of fairness that emerges when parties make comparative assessments of rewards. Two social psychologists, David Messick and Keith Scnfis, demonstrated self-serving interpretations of fairness in a work context (Messick 1979). [Fn 9] Subjects were asked to specify the fair rate for two people who had worked either ten or seven hours at a task. The person who worked seven hours was always paid $25; subjects were asked how much the person who worked ten hours should be paid. When they had worked seven hours, they thought the other person should be paid $30.29 for ten hours of work. But when they had worked ten hours, they thought they deserved $35.24. The difference between $30.29 and $35.24-$4.95-was cited as evidence of a self-serving bias in wage demands.

Numerous experimental studies have demonstrated that the concern of disputants for fairness and unfairness atfects bargaining.[Fn 10] The classic example is "ultimatum bargaining." In this type of bargaining, one person divides an amount between herself and another person. If the other person rejects the proposed division, both people get nothing. Self-interested bargainers who are neither altruistic nor envious should offer the smallest possible amount (usually a penny) and the offer should be accepted. But in most studies, people keep an average of 60 percent of the amount for themselves and offer 40 percent to the other person. When less than 20 percent is offered, the offer is typically rejected (Thaler 1988). Of course, even if a person is not fair-minded, she might offer 40 percent to the other person simply because she fears the other person will reject a lower offer. True concern for fairness can be separated from this kind of "enlightened self-interest," which masquerades as concern for fairness, by modifying the game so that the second person must accept die first person's offer.[Fn 11] In such "dictator games," roughly equal numbers of subjects offer zero or offer an equal split of the total amount (Forsythe, Forthcoming). Taken together, the evidence from ultimatum and dictator games shows that some people simply prefer to make fair offers and others make fair offers only because their bargaining partners will reject offers that are unfair.

Surprisingly high offers in the ultimatum game led experimenters to study more complicated multi-stage bargaining. In these more complicated situations, the amount being divided is reduced at each stage to reflect the costs of delay in reaching an agreement. The final stage becomes an ultimatum game. In most studies, offers were somewhere between an equal split of the amount being divided and the offer predicted by a game-theoretic analysis assuming self-interest (and assuming foresight about the outcome of bargaining in future rounds) (Bolton 1991; Johnson 1991; Ochs 1989). The most curious observation is that 80 percent of the offers that are rejected at one stage of bargaining are followed by counteroffers that give less money to the person who rejected the earlier offer. An example is a typical subject who turns down an offer of $1.75 out of $5. After several stages, the $5 will be reduced significantly because of the costs associated with the delay. By the end,the $5 will have been reduced to $2.50. At this point, the subject is likely to accept an equal split of the $2.50 as an adequate recoverv. This will leave the subject with $1.25 after all is done. The acceptance of the $1.25 amount shows a 50 cent reduction from what the subject could have received if he or she accepted the first offer, which did not have a substantial penalty for the costs of delay. In essence, the final $2.50 amount and the equal split becomes a "disadvantageous counteroffer." These "disadvantageous counteroffers" are clear evidence that some people sacrifice money for the sake of fairness.

Experimental Evidence

While large numbers of psychology studies show self-serving assessments and many game-theory experiments show that assessments of fairness play a role in bargaining, only a few studies have combined the two phenomena to show that self-serving assessments of the fairness of different bargaining outcomes can cause disagreements that result in inefficiencies.[Fn 12] A recent series of experiments, involving one of the authors of this chapter, used a stylized version of a simple legal dispute to test the effects of self-serving biases on the ability of parties to efficiently settle disputes (Babcock 1995; Babcock manuscript on file; Loewenstein 19!)3). Consistent with the methodology of behavioral economics, these experiments attempted to first confirm the existence of self-serving biases, then to manipulate the biases, and finally to see whether they could be overcome.

In these experiments, pairs of subjects were assigned the roles of plaintiff and defendant. [Fn 13] They then attempted to negotiate the settlement of a tort case arising from the collision of an automobile and a motorcycle. The injured plaintiff (the motorcyclist) was suing the driver of the automobile for $100,000. Both subjects were given precisely the same case materials and were informed that the informationthey were given was identical. The subjects received 27 pages of actual testimony abstracted from a real case in Texas. The materials included witness testimony, police reports, maps, and the testimony of theparties. Subjects were informed that the case materials had been given to a Texas judge who had reached a decision regarding compensation to the plaintiff.[Fn 14] In the initial set of experiments, participants were assigned the role of plaintiff or defendant, and attempted to negotiate a settlement. Delays in reaching a settlement resulted in substantial financial penalties to the parties/participants. If the negotiators ulfimately failed to settle, the award to the plaintiff was determined by an impartial judge who had earlier read exactly the same case materials.

The critical feature of these experiments is that the assignment of participants into the role of plaintiff and defendant was completely random. There were no pre-existing biases or wealth effects that would distinguish the two camps. In addition, the disputants were assured that the materials provided were the complete universe of information, thereby removing any uncertainty about asymmetric access to information. The experiments were structured so that the parties were induced to be unbiased, both by offering cash rewards for guessing the judgment that had been rendered in the case by a neutral arbiter and by penalizing failure to settle quickly by taxing costs against the amount to be distributed among the parties. In sum, the incentives in the experiments all ran toward settlement. Any observed disparities between the parties could only be attributed to systemic biases arising from thier role in the dispute.

At the most basic level, the results provided striking confirmation of the propensity for self-serving bias among disputants. The differences between the plaintiffs' and defendants' predictions of the judge's determination and between their perceived fair settlement points provided strong evidence for self-serving interpretations of fairness. Plaintiffs' predictions of the judge's determination were, on average, $14,527 higher than defendants'. Mean plaintiffs' fair settlement values were $17,709 higher than defendants'.[Fn 15] Equally significant was the relationship between the magnitude of the self-serving bias and settlement rates. An examination of the parties' assessments, broken down by the pairs that voluntarily settled and by the pairs that did not, revealed that the assessments of what was a fair settlement and of what a judge would award by the parties that settled out of court were closer together than their counterparts, the parties who could not settle. For example, among the 59 pairs who settled, the mean difference between the plaintiffs' and defendants' predictions of the judge's award was $9,050. For the 21 pairs who did not settle, the average difference was $29,917.

This first experiment provided strong evidence for the existence of a self-serving bias. The magnitude of the bias in terms of the monetary assessments is large - approximately one-half the value of the judge's actual settlement value whether measured by predictions of the judge's determination or perceived fair settlement points. This in turn means that one of the central assumption of the Priest-Klein model, that errors in the estimation of potential award amounts are random, was not substantiated in the experimental setting.

Moreover, the statistical link between the bias and nonsettlement is strong. The magnitude of the bias, although also present among pairs who settled, was over three times greater for pairs who did not. The strong correlation between the magnitude of the bias in a particular bargaining pair and nonsettlement supports the self-serving bias explanation for nonsettlement. The experimental setting further eliminated the possibility that some third factor, such as a particular character trait on the part of the negotiators, influenced either the interpretations of fairness or the negotiating behavior. By eliminating these, the experiment raised the possibility of the creation of a statistical association in the absence of a causal relationship. In a subsequent experiment, this possibility was eliminated by manipulating self-serving interpretations of fairness and examining the impact of the manipulation on negotiator behavior and outcomes. This was accomplished by assigning some subjects their roles (plaintiff and defendant) before reading the case materials and assigning others their roles after reading the case materials but before negotiating. Whereas 94 percent of the subjects in the latter group settled out of court, in the condition where subjects knew their roles when they read the case materials, 28 percent of pairs failed to settle out of court.

Subsequent experiments attempted to reduce the magnitude of the bias by either having subjects write an essay arguing the other side's point of view or informing them of the bias. Neither of these interventions had a measurable eftect. Subjects consistently believed that the judge's decision would be in their own material interest. Only in one experimental setting where subjects were both informed of the bias and made to write an essay substantiating the other side's case was the effect of the bias mitigated. That subjects were unable to rid themselves of the bias when rewarded for doing so and of their belief that they are not subject to the bias demonstrate clearly that the self-serving bias is not a deliberate strategy.

Other findings from the same series of experiments point to biased assimilation of information as the likely psychological mechanism underlying the selfserving bias. Subjects were presented with eight arguments favoring the side they had been assigned (plaintiff or defendant) and eight arguments favoring the other side. They were asked to rate the importance of these arguments perceived by a neutral third party. There was a strong tendency to view the arguments as supporting one's own position as more convincing than those supporting the other side, suggesting that the bias operates by distorting one's interpretation of evidence.

These studies reveal the shakiness of the assumption that parties process information revealed to them in an unbiased fashion so as to achieve settlement. Even with perfectly shared information and a complete absence of disputed legal issues, self-serving biases can cause inefficient impasses. On a theoretical plane, this study points to a view that litigants may not be seeking to maximize their own payoff, as assumed by analyses focusing on settlement ranges. Instead, they simply may be trying to obtain what they deem fair. For example, in the above experiments the parties seem to have based their decisions on whether to settle upon how close they could get to a specific monetary amount. The subjects of these experiments appear to have decided how much money they should receive or pay for the injury, based on fairness or on a self-serving assessment of the problem. The parties seem to have overlooked the penalty for delaying settlement in hopes of getting more money in the settlement, although in reality they received less money because of the penalties associated with the delay. If this is the case, the settlement range would not be expected to make much difference causally, although assessments of fairness would no doubt influence the settlement range.

In addition, if nonsettlement is driven by systematic bias instead of random error or a lack of information, the practical ramifications are numerous. First, it indicates that exchanges of information are not in themselves necessarily conducive to settlement, but must be analyzed in terms of how they interact with preexisting biases. Second, it suggests that effective alternative dispute resolution mechanisms should be, at least in part, directed at debiasing parties rather than simply facilitating information exchange.

SETTLEMENT GAMES The occurrence of protracted litigatoin and trials seems a quandary worth explaining for reasons already identified above. On the most widely-employed economic model, litigation is a negative-sum game for parties who can maximize their joint wealth by settling. In this model, Pp, equals the probability of a proplaintiff verdict as estimated by the plaintiff; Pd equals the defendant's estimate of the same probability; Jp, equals the plaintiff's estimate of the judgment at trial; Jd equals the defendant's estimate of the same qualifity; Cp, equals the expected cost to the plaintiff of trying the case; Cd equals the defendant's expected trial cost; Sp equals the cost to the plaintiff of settling (assumed to be less than Cp); and Sd equals the cost to the defendant of settling (assumed to be less than Cd). The model posits that a plaintiff will decline to settle unless offered an amount greater than:

(Pp) (Jp) - Cp + Sp,

and that a defendant, in order to settle a lawsuit, will offer no more than

(Pd) (Jd) + Cd - Sd.

The model thus predicts settlement when

(Pd) (Jd) + Cd - Sd > (Pp)(Jp) - Cp + Sp. When the defendant is willing to offer more than the plaintiff demands in order to settle the case, a range exists in which, by settling, the parties can bargain to a mutually advantageous result.

Concerns about protracted litigation - lawsuits that last a long time and consume large quantities of resources before settling - focus on the amount of money spent gathering and processing information about the values of claims or negotiating an agreement within a settlement range. The concern is that the amount expanded up to the point of settling is too large. In other words, the perception is that plaintiffs and defendants invest too heavily in protracted lawsuits, despite the fact that, by settling at all, they avoid the even greater cost of taking a lawsuit to trial.

The concern about protracted litigation raises at least two questions. Why do some disputes become protracted? And, how much money should reasonably be spent on a lawsuit that ultimately seffles? The second question can be rephrased in a helpful way: When is one justified in describing a dispute as "protracted" because the principals invested an unreasonably large amount of money in the dispute before settling,? The first question calls for a descriptive answer. The second calls for a normative one. It requires a normative theory concerning the level of investment that is appropriate in a dispute.

Having studied the actual conduct of litigation, we find it easier to explain than to define protracted litigation. However, we have a strong intuition that a proper definition will reflect the reasons for excessive expenditures. If principals, meaning decision makers whose resources are actually consumed in litigation, are bottom-line oriented, then their decisions to invest in litigation should generally be economically rational from their point of view. [Fn 16]
They should invest in litigation when the expected marginal return exceeds the marginal cost. Presumptively, a normative economic theory of litigation investments would therefore encourage principals to continue acting as they generally do. Absent an argument showing that individually rational actions yield collectively irrational results - an argument like Orley Ashenfelter and David Brooks' Prisoner's Dilemma account (Ashenfelter 1993) or like Leo Katz's suggestion that civil litigation resembles a dollar auction (Katz 1987) - a normative economic theory should encourage principals to continue acting as they do. Moreover, if there is a perverse economic logic to litigation, a descriptive account of investments in heavily litigated cases should be of interest to normative thinkers. It is usually easier to encourage desirable conduct when one understands why particular undesirable conduct occurs under existing institutional arrangements.

Although, as mentioned above, politicians and pundits seem bent on maintaining the impression that, in the lawsuit-crazed United States, every grievance is an occasion for litigation, the reality is considerably less bleak. Scholars universally acknowledge that people bring only a small fraction of their complaints to lawyers (Galanter 1983; Miller 1981). In the overwhelming majority of instances, complainants either work out informal solutions with their opposing numbers or 'lump it,' taking no action at all. It is likewise true that lawyers formally invoke the legal process only a fraction of the time. Lawyers decline lots of requests for representation, and they settle many claims without filing complaints. Lawyers also give up on claims a fair part of the time after agreeing to press them. A lawyer may withdraw when it becomes clear that quick settlement is unlikely, when an opposing party has neither insurance nor personal assets sufficient to justify litigation, or when an investigation reveals that a case lacks the necessary level of factual or legal support.

The act of formally invoking the legal system is thus anextraordinary event that occurs in a tiny fraction of all disagreements. Moreover, recent evidence suggests that most lawsuits end quickly (by settlement or abandonment) with little or no formal discovery and without trials. For example, a recent insurance industry publication states that although about 19 percent of bodily injury claims involve the filing of a lawsuit, only about six-tenths of 1 percent of all bodily injury claims are decided by a judge or jury. The rest are settled or abandoned (Insurance 1994). Because discovery and trials contribute heavily to the expense of litigation, the fact that discovery occurs in fewer than half of all lawsuits and that trials are extraordinarily rare suggests a desire to economize, even in cases deemed weighty enough to merit the filing of a complaint.

A sample of seriously litigated cases, those having long life spans and requiring lots of resources, seems to consist disproportionately of high-dollar business disputes and multiparty cases involving complicated legal issues and large stakes (Keilitz 1993; Smith 1995). The settlement model might have led one to predict that large stakes would be associated with trials and protracted litigation. In the model, trials occur when plaintiff and defendant are both optimistic about their chances of winning at trial. However, whether a particular level of mutual optimism will erase the settlement range depends partly on Jd and Jp the amount at stake. Assuming mutual optimism and holding other things constant, the settlement range decreases in size and ultimately disappears as Jd and Jp grows.

Common sense also suggests that protracted litigation and trials are more likely when the stakes are great. Why invest heavily in litigation unless one can expect a sizable return? If the principals (be they parties or attorneys advancing time and litigation costs) are economically rational, they will expend heavily on litigation only when they stand to recover a large judgment or to reduce greatly an expected loss.

A further implication of common sense is that the larger the number of persons who share economic interests in and control of a lawsuit, the more protracted and difficult to settle the lawsuit will be. To settle a lawsuit completely, one must build a grand coalition of all participants who are in a position to keep any part of the lawsuit alive. The larger the number of participants and the more divergent theirinterests, the more difficult that is to do. It is generally and obviously easier to build smaller coalitions, which is why partial settlements are common in protracted, multiparty cases. However, partial settlements can also make final settlements more difficult to arrange by destabilizing coalitions and changing the stakes for parties who remain in the litigation.

To show how complex the settlement dynamics of even garden-variety lawsuits can be, we will describe in some detail a dispute about defects in a home that ultimately became a lawsuit over insurance coverage between the real estate agent who orchestrated the sale of the home and the carrier that provided the agent's errors and emissions coverage. We will use the case to show that simple disputes are often embedded in intricate preexisting webs of legal and economic relationships, that these relationships create complex settlement games (including games within games) in which the participants have mixed motives, that opportunities to strategize make it difficult to build grand settlement coalitions, that one need not attribute bad motives to participants (including lawyers) to understand why lawsuits become protracted, and that lawyers - as architects of settlement coalitions deserve much of the credit for settling disputes.

The facts we will recite are based loosely on the case of John Daugherty Realtors, Inc. v. Acceleration National Insurance Company, an insurance coverage dispute pending in the United States District Court for the Southern District of Texas. The events that later led to the lawsuit occurred in 1989, when Pon and his company, Foremost Development Corporation (Foremost), built and sold a luxury home to the Durdins for $1.6 million. John Daugherty Realtors (JDR) represented Pon and Foremost in the sale. According to the Durdins, the home contained numerous undisclosed defects, including marble floors that failed under normal home traffic pressure, inadequate lighting, faulty plumbing, improperly constructed windows, defective carpet, cracked paint, warped doors, damage due to water penetration, and other shortcomings. Many of these defects had been discovered by Deal, an inspector hired by another potential purchaser of the home who gave a copy of his report to Pon and JDR. However, neither Pon nor JDR showed the Durdins a copy of Deal's report. To make matters worse, Pon allegedly masked many of the defects, so that subsequent inspectors would not find them and falsely told JDR that he effected full and complete repairs. When Pon failed to satisfy the Durdins by putting the property in proper condition, they sued him, his wife (also an owner of Foremost), Foremost, JDR, and several other potentially liable parties to recover compensatory damages, punitive damages, and attorneys' fees. The formal party structure of the Durdin's lawsuit, the list of names appearing on the complaint as shown in Table 1, suggests the complexity of the legal and economic relationships that were embedded in their dispute.

The Durdins alleged that the Defendants were jointly and severally liable to them for all losses they incurred as a result of defects in the house. The theories of liability included conspiracy, breach of express and implied warranties of fitness, negligence, breach of the duty of good faith and fair dealing, negligent and intentional misrepresentafion, and fraud. In turn, the defendants alleged that the Third-Party Defendants (TPDs) ought to reimburse them for all sums they paid the Durdins by reason of defects in particular products used in the construction of the house. The defendants did not seek to hold the TPDs jointly and severally liable for the entire amount paid to the Durdins. They wanted each TPD to indemnify them for the portion of the total loss attributable to that TPD's product. The Durdins also asserted claims directly against certain TPDS.

Table 1. Party Alignment in  Durdin V. Pon

Plaintiffs	Defendants	Third-Party Defendants
(1) Mr. Durdin	(1) Mr. Pon	(1) Terrazzo & Marble Co., Inc.
(2) Mrs. Durdin	(2) Mrs. Pon	(2) Bac-Trac Plumbing co., Inc.
	(3) Foremost Development Corp.	(3) Bison Building Materials, Inc.
	(4) John Daughtery Realtors, Inc.	(4) Buffelen Woodworking Co., Inc.
		(5)Walker and Zander West Coast, Ltd.
		(6) Oregon Glass Co.

Table 2. Lines of Control and Economic Interest in Durdin V. Pon

Plaintiffs	Defendants	Third-Party Defendants
(1) Mr. Durdin	(1) Mr. Pon	(1) Terrazzo & Marble Co., Inc.
(2) Mrs. Durdin	(2) Mrs. Pon	Insurance Company
Attorney 1 (for 1 & 2)	(3) Foremost Development Corp.	Attorney (for TMC & Ins. Co.)
Attorney 2 (for 1 & 2)	Attorney (for 1- 3)	(2) Bac-Trac Plumbing co., Inc.
	(4) John Daughtery Realtors, Inc.	Attorney (for B-TP & Ins. Co.)
	(5) Acceleration National Ins. Co.	(3) Bison Building Materials, Inc.
	Attorney 1 (JDR & Accel.)	Insurance Company
		Attorney (for BBM & Ins. Co.)
	Attorney 2 (JDR only)	(4) Buffelen Woodworking Co., Inc.
		Insurance Company
		Attorney (for BW & Ins. Co.)
		(5)Walker and Zander West Coast, Ltd.
		Insurance Company
		Attorney (for W&ZWC & Ins. Co.)
		(6) Oregon Glass Co.
		Insurance Company
		Attorney (for OG & Ins. Co.)

As shown in Table 2, the economic and decision making structure of the dispute was even more complicated than the formal party structure.

Table 2 conveys a sense of the lines along which control of and economic interest in decisions relating to the litigation were allocated. Onthe plaintiffs' side, a basic fact was that there were two Durdins who shared control and economic interest. They may have cooperated well or poorly. They may have been unequally experienced with or educated about legal matters. They may have had different or similar attitudes toward risk. The Durdins also shared control and economic interest with Attorney 1 and Attorney 2, whom they hired to replace Attorney 1. Presumably, the Durdins gave Attorney 1 and Attorney 2 economic stakes in the outcome of the litigation by entitling them to contingent percentage fees plus a contingent entitlement toward the reimbursement of expenses. The Durdins probably also ceded the attorneys day-to-day control of the litigation, although as a matter of law they must have retained control of the decision to settle themselves (Rule 1.2 1995; Restatement 1992). As a practical matter, though, the attorneys would have had significant input into the latter decision as well. The Durdins probably wanted the attorneys' advice on settlement and, given the attorneys economic interest and sophistication, the attorneys would have wanted to voice their opinions as well. Finally, whether a financial relationship existed between Attorney 1 and Attorney 2 is also unknown, but it is likely one did. The practice in Texas is that a receiving attorney promises a referring attorney a contingent percentage interest in the fee.

The implications of the allocation of economic interest and control on the Durdins' side for investment in the lawsuit and for settlement are vast and can only be suggested here. By retaining counsel on a contingent percentage basis, the Durdins acquired legal services they needed. They also reduced the size of a large risk they could not otherwise insure or readily diversify. Presumably, the exchange both increased their expected return and reduced the variance thereof, making them unambiguously better off and, probably also less risk-averse. The exchange thus put them in a better position to bargain for settlement.

Attorney 1 and Attorney 2 were also sources of strength because they could diversify the risk of loss associated with the litigation better than the Durdins. Plaintiffs' attorneys are managers who maintain portfolios of risky lawsuits. Although their diversification is imperfect because the risks they maintain are not independently distributed, they can tolerate risk better than most plaintiffs can. Moreover, the fee to be earned in a given matter usually constitutes a small part of a plantiffs' attorney's expected income stream. That is especially likely to be true when an attorney shares fees with other members of a firm and with other lawyers who are recipients of referral fees. Consequently, a plaintiffs' attorney can bargain harder in settlement than a plaintiff can.

The Durdins also benefitted from the contingent percentage arrangement by freeing themselves from the risk of loss associated with investments in the litigation. Attorney 1 and Attorney 2 advanced their costs and their time, and they bore the risk of nonpayment. This arrangement encouraged the attorneys to invest in litigation only when the expected marginal return exceeded the marginal cost. When that was true, the Durdins stood to gain something from an outlay of time or funds. The arrangement thereby reduced the burden of monitoring upon the Durdins. However, because the attorneys would personally receive only a portion of the expected marginal return on their investments of time and other resources (the portion represented by the fee), they might predictably fail to make some investments based on cost-benefit grounds even though these investments, from the perspective of the plaintiffs' team as a whole, might be justified.

Fracturing the interest on the plaintiffs' side therefore gave the defendants a potential bargaining advantage. Because a defendant who bore one-hundred percent of the loss would find it rational to invest more in the lawsuit than a plaintiffs' attorney whose interest was only, say, 40 percent, the defendant could make a credible threat to outspend the attorney unless the attorney agreed to settle the case for less than its worth. Attorney 1's efforts to monitor Attorney 2 and Attorney 2's reputational interest as a receiving attorney would remedy the problem somewhat (Spurr 1988), but a defendant advantage would predictably remain.

A look at the roster of defendants and TPDs reveals, however, that none was exposed to 100 percent of the cost to be incurred by the coalition of all defendants and TPDs. The TPDs were alleged to be liable for only losses stemming from the particular products they manufactured. The TPDs were also insured for at least some of their liability and defense costs. True, the four defendants were asserted to be severally liable for the entire loss the Durdins incurred. However, three of them, the Pons and Foremost Development Corp., were judgement-proof against losses inexcess of an amount that, although undisclosed, appeared to be considerably less than the more than $1 million in damages the Durdins claimed. The Pons were also insulated by Texas debtor protection law, which put their personal assets of significant value beyond the Durdins' reach. The Pons and Foremost Development also had claims against JDR and the TPDs for contribution or indemnification that substantially reduced the net amount they stood to lose. The exposure of the remaining defendant, JDR, was reduced too. JDR was insured for more than $200,000 in defense and liability costs, and JDR had claims against the Pons, Foremost Development, and the TPDs as well. Thus, JDR could not make a credible threat to outspend the Durdins' lawyers. Nor could JDR's insurer, Acceleration National Insurance Company (AN) make that threat. Although it controlled the defense, its policy limits were considerably lower than even the liability JDR faced. Its policy was also self-liquidating. Every dollar spent on defending the lawsuit reduced by a dollar the amount AN had to pay to indemnify JDR. Consequently, AN's marginal incentive to invest in the lawsuit was always diminishing.

The contribution and indemnity claims running between and among the defendants and TPDs gave rise to a complicated allocation game that further weakened the position of these parties relative to the plaintiffs. In theory, the defendants and TPDs could have bargained for the greatest possible concessions from the plaintiffs by presenting a united front in negotiations. In practice, it was difficult for the defendants and TPDs to act collectively, for two reasons. First, each defendant and TPD wanted to minimize its own loss by shifting as large a part of the settlement burden as possible to others. In other words, each wanted to acquire for itself the largest possible share of the joint savings settlement confered upon the defendants and TPDs as a group. In this environment, coalitions were unstable. Second, when coalitions did form, the plaintiffs undermined them by settling directly with "small-fry" TPDs. These partial settlements effectively underwrote the plaintiff's litigation costs and required the remaining defendants and TPDs to bargain anew over the allocation of settlement costs. They also caused each remaining defendant and TPD to fear that no grand coalition could remain stable for long.

The defense side was further weakened by insurance coverage limitations that afflicted the relationships between the defendants and TPDs and their insurers. Consider the plight of JDR, the only insured defendant. It had limits far lower than the damages sought. Consequently, it faced the prospect of having to pay a large sum out of its own pocket unless the case was won at trial or settled within the policy limits. JDR also faced the possibility of having no insurance at all for the loss. The Durdins' charged JDR with acts like negligence that were covered and with acts like fraud and intentional misconduct that were not. If a trial determined that JDR committed fraud, its insurer, AN, might disclaim coverage and leave JDR uncovered for the judgment. To make matters worse, the AN policy was self-liquidating, as already noted. JDR therefore had a strong interest in a quick sefflement at or below the policy limits. By contrast, AN was presumably indifferent between settling now for $X, where X could be any amount up to the policy limits, and bearing any combination of defense and settlement costs equal to $X.

For the reasons given, the alliance between JDR and AN was an uneasy one. Although they shared an interest in minimizing theloss on the claim, JDR's desire to settle was considerably stronger than AN's, and JDR's interests on coverage questions were directly opposed to AN's. These conflicts at least partly explained JDR's decision to hire independent counsel whose job was to monitor the activities of the lawyer AN retained to defend the lawsuit and to pressure AN to settle. Matters came to a head when, after several failed attempts by the parties to negotiate a global settlement of the dispute, the Durdins settled with all the defendants and TPDs other than JDR. Suddenly, JDR found itself the only remaining defendant in a lawsuit where fraud was alleged and punitive damages were claimed, where the existence of insurance was questionable, where the amount of insurance was likely to be far smaller than the eventual judgment, and where trial was scheduled to begin in just a few days. To make matters worse, the plaintiffs, sensing JDR's predicament, increased their sefflement demand threefold. JDR reacted to the situation by taking its fate into its own hands. It settled with the Durdins for $500,000, roughly $320,000 more than the unspent policy limits and $375,000 more than the highest amount offered the Durdins on behalf of JDR in prior negotiations. It then sued AN, alleging wrongful failure to settle and bad faith. Thus did the lawsuit that began Durdin v. Pon become John Daugherty Realtors, Inc. v. Acceleration National Insurance Co., a dispute still pending years after the Durdins' complaints were resolved.

The history of John Daugherty Realtors, Inc. v. Acceleration National Insurance Company is interesting for many reasons. We recite it here to show how complicated the control structure and the economic dynamics of a garden-variety dispute can be. The Durdins' complaint (the grievance, not the legal document) arose in a context characterized by preexisting, intricate legal and economic relationships. When the Durdins decided to voice that complaint (instead of lumping it), they activated those relationships and created a complex bargaining situation that housed many smaller bargaining games within it. Neither they nor anyone else chose to create so difficult a situation, and probably neither they nor anyone else was happy to be caught up in it. The documents from the many lawyers involved suggest that everyone desired to settle the dispute and that everyone worked hard for settlement as well. In all likelihood, the lawsuit became protracted, in the ordinary sense of the word, for innocent reasons, despite the participants' desire to settle: the stakes were significant, control of decision making and economic interest were fractured along intricate lines, and there was no obviously correct or prominent solution to the problem of allocating the Durdins' loss among the defendants and TPDs.

In high-dollar, multiparty litigation, settlement can be and usuallv is a difficult game to play. The economic model of the decision to settle or sue obscures the difficulty because it says nothing about the structure or dynamics of the settlement game. It also says nothing about the role lawyers play in that game. One of the great miracles of litigation is that complex lawsuits settle as often and as quickly as they do. Lawyers deserve considerable credit for that achievement, credit they rarely receive. One of the most significant services litigating lawyers provide, for their clients and for society at large, is that of structuring deals that end litigation.

BARGAINING IMEPEDIMENTS AND SETTLEMENT BEHAVIOR The American civil justice system is explicitly designed, through rules of procedure and the practice of American courts and lawyers, to produce settlement and other non-adjudicatory outcomes. Despite the continued fascination of our popular culture with trials, fewer than ten percent of cases filed in state courts of general jurisdiction ever get to trial. Procedural devices ranging from the pretrial conference, to the class action, to the offer of judgment have been reformed in efforts to squeeze an ever larger fraction of cases out of the system short of trial.

Legal scholars still debate whether the civil justice system should be designed to encourage or coerce settlement. Trials can be a significant public good for many reasons. Trials can be a source of education and information about law to the public as a whole and to individual actors, and they canprovide an opportunity to vindicate public goals. Trials also can be horrendously expensive, and some of that expense is traditionally borne by the public through taxation to support the court system. By and large, the last twenty-five years have seen efforts in most states and the federal system to reduce that public expense by encouraging litigants to settle earlier and more often. We want to prevent trials, not encourage them.

Given a legal process systematically designed to produce settlement, what explains the residue of cases that fail to settle and produce trials? As mentioned, some economists explain trials in terms of party optimism, assuming that litigants make demands and offers solely because of what they expect courts to do, not because of how they expect opponents to respond inbargaining (Gould 1973; Landes 1971; Posner 1973.) Others emphasize these strategic calculations, suggesting that some trials are caused by bargaining tactics designed to force opponents to surrender as large as possible a fraction of the gain to the parties from seffling rather than proceeding to trial (Cooter 1982; Mnookin 1979). The array of potenfial strategic ploys is virtually unlimited. However, the ploys can periodically fail, producing trial in a case where both parties would seem to be better off through settlement.

Some scholars have looked beyond strategic and non-strategic bargaining models to the interests of agents - including attorneys and insurance companies - in the decision to try or settle a case. These repeat players may have broader stakes in the litigation which, combined with a legal and practical ability to control the decision whether to settle, determine what gets tried (Galanter 1974; Syverud 1994). Thus, a plaintiff's personal injury attorney or a defendant's insurance company, each of whom may be simultaneously handling many lawsuits, may choose which cases from the portfolio to try on grounds quite distinct from those that would inspire individual plaintiffs or defendants (Coffee 1986).

Finally, some legal scholars have suggested that trials are or should be driven by the desire of some litigants to obtain a day in court, a hearing on the merits of the dispute, a public rendering of justice through the application of law to the specific facts, and an embodiment of that jusfice in a legal judgment (Fiss 1984). This view implies either that at least some portion of American trials are driven by a desire to obtain the one thing settlement rarely provides, public vindication, or that vindication is a goal the legal system should actively promote.

So what explains trials? Are they driven by party optimism, strategic ploys that misfire, the inceritives of insurers and attorneys, by desires for vindication, by some combination of these impediments to bargaining, or by none of the above?

We can obtain some qualified insights into the answer by asking attorneys. In a forthcoming study, one of the authors of this chapter, incollaboration with Professor Samuel Gross of the University of Michigan Law School, reports the results of a study of a sample of 363 civil cases tried to civil jury verdicts in California Superior Courts from 1990 to 1991. The sample was selected from those reported in Jury Verdicts Weekly, a publication which attempts to report comprehesively on all California trials and the settlement negotiations that precede them. As part of the study, the authors conducted telephone interviews of the attorneys for all parties in the trials. At least one plaintiff's attorney was interviewed in 92 percent of the cases, and at least one defendant's attorney in 95 percent of the cases. The interviews included questions about attomey's fee and insurance arrangements, as well as an open ended question about why, in the attorney's opinion, the case went to trial rather than settling (Gross 1995, 1991).

The attorneys' answers to the question of why the case went to trial ranged so widely as to defy any effort to prove or disprove theories of settlement. There are nevertheless at least four interesting aspects to the answers, aspects that require further testing in efforts to determine why cases go to trial.

First, a desire for vindication figured in attorney explanations for trial in no more than three of the hundreds of attorney interviews. No attorney said a case was tried because a party demanded her day in court; only a few attributed trials even in part to the desire of a client for a hearing or a public judgment.

Why did vindication not shine out among attorney explanations for trial? There are several possibilities. One is that attorneys as agents undervalue their clients' desire for vindication, and focus more on their clients' (and their own) economic interests in the litigation. They therefore understate the role of vindication in trial-seeking behavior. Another possibility is that the clients in most mine-run litigation in California courts don't care much about a day in court, that they are indifferent between private settlement and public adjudication except on economic grounds. A third possibility is that in most cases plaintiffs and defendants, regardless of their preferences for public vindication through trial rather than private settlement, do not have the power to act on that preference and to force trial. The insurance company and the plaintiff s attorney working on a contingent fee may usually control the settlement decision, with the result that few of cases taken to trial are attributable to anyone's desire for vindication.

Second, plaintiffs' attorneys frequently attribute trials to the refusal of the defendant to make any offer in settlement. Defendants offered nothing in settlement, at any point in the pretrial proceedings, in 27 percent of the cases in the sample. This "zero-offer" rate was particularly high in medical malpractice trials, where in 59 percent of the cases the defendant had never made a settlement offer prior to trial. The rate was lowest - 11 percent - in vehicular negligence trials.

The existence of zero-offers in a significant fraction of trials is largely inconsistent with settlement models that assume parties will make offers and demands that take into account the legal costs of proceeding to trial. Previous research has suggested that zero-offer cases afford significant evidence of strategic bargaining. Defendants' refusal to bargain in these cases is not a forecast of their expected trial costs, but rather an attempt to influence the behavior of opponents. For example, defendants might be attempting to induce some plaintiffs to dismiss their cases by imposing trial costs upon them, even though in the process the defendants themselves would incur costs greater than potentially acceptable settlement offers (Gross 1991).

The high rate of zero-offers in medical malpractice cases probably has a different explanation. Both the jury verdict reports and the interviews suggested that the lack of offers in medical malpractice cases was attributable to the defendant physician's refusal to consent to an offer. Most physician malpractice insurance policies sold in California contain a "consent to settle" clause which requires the permission of the doctor to any non-zero settlement negotiated by the insurer. The insurance company remains responsible for any defense costs that might be caused by a trial that results when consent to settle is withheld. Thus, the defendant doctor can usually reject nuisance settlements without risking personal responsibility for defense costs or for any judgment within policy limits. The data suggest that doctors are engaging in just such behavior in a large fraction of all medical malpractice trials.

Why are doctors withholding consent? We believe it is because they desire vindication for financial, professional, and personal reasons. Also, unlike other litigants, they have negotiated insurance contracts that give them the authority to act on that desire. In most civil trials, the insurance company controls the decision to settle, and insurance arrangements do not require the consent of the insured before a settlement can be effectuated (Syverud 1994). However, doctors can insist on trial, and are most likely to so where they are convinced they behaved in a professionally responsible manner and have had their self-esteem and reputation damaged by the complaint's allegations that they committed malpractice. A trial provides the opportunity for vindication. This may explain why medical malpractice trial rates are considerably higher across the nation than most other categories of litigation, and why doctors win defense verdicts in more than 90 percent of the cases in which there is no settlement offer at any point in the litigation (Gross 1991). It appears that, at least in one type of litigation where reputation and vindication are significant interests for defendants, they are able to order their private relationships with their insurance companies in a way that protects that interest.

Third, most attorneys explain their failure to settle a lawsuit by blaming the other side. Fifty-four percent of the plaintiffs' attorneys blamed trial on the defendant or the defendant's attorney. Forty-one percent of defendants blame trial on the plaintiff or the plaintiffs' attorney. The tendency to blame the other side is largely unaffected by whether an attorney won or lost at trial. Plaintiffs' attorneys, for example, are as likely to say defendants, caused trial when plaintiffs win big as when there is a defense verdict.

The need to blame someone else for a trial suggests that, at least for attorneys, trials are something to be excused rather than celebrated. Although popular culture glorifies trials, the attorneys in our sample appeared largely embarrassed by them. Perhaps the design of our civil justice system, which consciously seeks to produce settlements rather than trials, has come to suffuse the values of attorneys in ways that scholars and the public do not fully appreciate. It certainly appears that for most attorneys pride about taking a case to trial cannot be properly viewed as an impediment to settlement.

Fourth, no other attorney explanation for why cases failed to settle stood out in a clear pattern. Emotion, including spite, figured in some explanations, seemingly more often in commercial cases than elsewhere. Only a few attorneys blamed trials on ambiguities in legal doctrine. And only a few attorneys attributed trials to a defendant's institutional policy against settlement or a dispute between a defendant and its insurer over insurance coverage. Trials may be driven by bargaining impediments created by emotion, legal ambiguity, settlement policy, or insurance arrangements, but these interviews do not suggest attorneys deem these impediments to be important in any significant fraction of cases.

CONCLUSION

Dispute resolution is a complex enterprise that, not suprisingly, consumes vast resources in many cases. Certainly, lawyer avarice and self-interest contribute to the escalation of costs in some degree, but attempts to blame the litigation explosion solely on lawyer misconduct and the adversary process are utterly simplistic. There are inherent difficulties in the resolution of conflict in a world of complicated interactions, counterposing interests, and fragmented control of decisions. That is the world we live in, a world in which intricate social, commercial, and legal relations undergird everyday conduct and influence the conduct of many disputes.

In this chapter we have attempted to set out some of the sources of complication in dispute resolution by looking to how real people respond to conflict in litigation and experimental simulations. What we find is in some ways quite heartening. Despite the overwhelming pressures toward conflict escalation, the current system channels most disputes toward resolution well short of trial. We can document the role of cognitive biases and strategic conflict in complicating the settlement process and in undermining even clearly efficient settlements. We can further demonstrate the vast array of factors trial prompts different types of cases to settle at different rates with varying degrees of concern for collateral consequences of thee particular dispute. Yet for all the foibles of humans and their social networks, the system grudgingly does move cases toward some form of resolufion. The system may be imperfect; it is by no means clear that it is any more imperfect than the actual human beings that come before it.

NOTES

1. We leave out of our consideration the additional effect of strategic bargaining. (Mnookin 1993.) It may be in the interest of bargainers to commit to established bargaining positions in order to negotiaite more favorable agreements (see Schelling 1956).  Bargainers who are able to commit credibly shift the range of agreements intheir favor (see Crawford 1982).  This is particularly the case with institutional defendants who are repeat players in litigaiton and who may be able to dissuade potential plaintiffs from filing suit by the proscpect of high investment levels prior to potential recovery.  As expressed by Priest and Klein, in situations invvolving repeat players, "the stakes will almost surely differ between the parties, because the alternative costs of their future activities are unlikely to be equal" (Priest 1984, p. 28).

   However, commitments are credible only if bargainers face large costs to backing down.  This scenario occurs, for example, in labor-management negotiations in which union negotiators stake their reputations on achieving a certain wage increase.  This strategy can be optimal even when it risks a positive probability of impasse. Impasse is inevitable if the bargainers commit to incompatible bargaining positions whichcannot be reversed (Schelling 1956, p. 287).

2. In the economics literature, impasses have also been attributed to uncertainty about the bargaining opponent's reservation value (see Babcock 1991).  A ready example arises in labor-management negotiations, where there is uncertainty about the firm's profitability, the consequences of a strike to both paties, arbitrator preferences, and the opponent's beliefs about arbitrator preferences (see Hayes 1984; Tracy 1987; Babcock 1992).

3. A second source of information asymmetry may be attributed to principal-agent relationships (McCall 1990).  Negotiators (agents) may prefer to use costly dispute resolution procedures in environments which are unfavorable to their constituents (principals).  Negotiators will avoid settlement if the principal cannot distinguish between agent shirking and an unfavorable bargaining environment.  This type of relationship may exist within labor unions in contract negotiations or between a lawyer and a client pretrial bargaining (Miller 1987).

   Problems of agency are particualrly acute in contingent fee cases (Mitchell 1972; Clermont 1978), and where the recovery of the plaintiff can be played off against payment of plaintiff's counsel.  For example, in Evans v. Jeff D. the  Supreme Court approved a settlement offer made contingent upon waiver of attorneys' fees by the plaintiffs' counsel (Evans 1986).  The facts of the case made the agency issue more dramatic, since plaintiffs' counsel was representing a class of institutionalized mentally retarded minors - a group with which he was unable to contract around the problems of attorney nonrecovery in a fee-waiver settlement.  But the problems of tension between principal and agent may lead to nonoptimal decisions by plaintif under a variety of circumstances, as Professor Geoffrey Miller was shown (Miller 1987, p. 209).

4. The principal area of empirical assessment of the Priest-Klein selection hypothesis has involved evaluation of the likelihood of a plaintiff's judgement in a litigated acse.  According to Priest and Klein, where the stakes in litigation are roughly comparable and where both parties have equal abilities of predicting the outcome of cases, plaintiffs should be expected to prevail at trial roughly 50 percent of the time (see Priest 1984, pp 5, 17-22, and footnote 42, 1985, p. 219).  The efforts to test the Priest-Klein hypothesis have primarily involved empirical assessment of plaintiff success rates in cases that proceed to trial (see Priest 1984, p. 31; Ramseyer 1989; Eisenberg 1989).

5. One might expect subjects to lower their reservation values to take account of the high variance when only a small number of chips are drawn.  However, Stanley and Coursey did not discuss this complication (Stanley 1990).

6. Although this assumption is well-known, some authors have noted that there are factors that can cause an error in judgement to become more systemically bias.  Those factors include a party's optimism, experience, persuasive ability and pecuniary resources. (Anderson 1994).  These factors can, in general, cause a situation where bargaining failure could be seen as more systematically biased.  For an excellent discussion on these factors, see Anderson (1994).

7. An excellent introduction to this line of research is found in Judgement Under Uncertainty: Heuristics and Biases (Kahneman 1982).

8. Bayes theorem is a method of calculating the probability of the validity of a proposition on the basis of a prior estimate of its probability and new relevant evidence (see Britannica 1985).

9. A doctoral student of Messick's obtained very similar results in a study involving real payoffs (Messick 1979).

10. For an exmaple, see Selten (1987).

11. This modification was first suggested by Daniel Kahneman, Jack L. Knetsch, and Richard H. Thaler (Kahneman 1986).

12. For example, in studies by Professors Alvin Roth and J. Keith Murnigham, pairs of players bargained over 100 chips which determined their chances of winning a dollar prize (e.g., 37 chips gave a 37 percent chance of winning). One player's prize was $20 and the other's prize was $5. Notice that there are two ways to split the chips equally: give 50 chips to each (equal chance of winning) or give 20 chips to the $20-prize player and 80 chips to the $5-prize player (equal expected dollar winnings). When neither player knew the prize amounts, they agreed to divide the chips about equally; only about 10 percent of the pairs disagreed and ended up with nothing. However, when only the $5-prize players knew both prize amounts, they insisted on getting 80 chips (to equalize dollar winnings). Since the $20-prize players wanted 50 chips to equalize their chances of winning, about 30 percent of the pairs disagreed. Thus, the tendency to focus self-servingly on the kind of equal split which favored oneself created disagreement (Roth 1982).

    In another studv, pairs of MBA students negotiated the hypothetical sale of a piece of land, knowing only the value of the land (if they were buying it) or its cost (if they were selling it). All pairs agreed on a sale price. Trouble began when students negotiated a second time, with the same partner, after finding out the value or cost of the land to their partner. Students who had struck a good deal in the first negotiation thought that selling the land at the same price was fair; their partners often disagreed, preferring to split the difference between the land's cost and its value or use some price they thought was more fair than the first-round price. While none of the pairs disagreed in the first negotiation, 20 percent disagreed in the second negotiation when self-serving assessments of fair prices influenced the bargaining (Camerer Forthcoming).

13. The subjects were undergraduates from the Univeristy of Chicago, law students from the Univeristy of Texas, students from Carnegie Mellon University, arid Wharton.

14. We told the subjects that the judge had seen the same case materials for two reasons. First, we wanted the subjects to know that an independent arbiter had read exactly the same materials they were reading. Second, we wanted them to know we had not selected the case based on the judgment awarded by the judge, but instead had first selected the case and then solicited a judgement on it. If the judgment had been simply chosen from an actual trial, subjects might have anticipated that we would choose a case with an award amount lying within a particular desired range. The actual trial figure was computed by a law professor at die University of' Texas (our "judge") specializing in civil litigation issues.

15. Both of these differences are statistically different from zero at the p=.0001 level. Fairness assessments and predictions of the judge were positively correlated across subjects in the same role (r =.73 for plaintiffs; r = .71 for defendants). These high correlations are consistent with the notion that parties are using their biased assessment of fairness to predict what a judge would award.

16. For purposes of this discussion, we will ignore the biased assessments of information discussed previously.

    REFERENCES

    Anderson, D. A. 1994. "Improving Settlement Devices: Rule 68 arid Beyond." Journal of Legal Studies 23(l): 225.

    Ashenfelter, O. and D. Bloom, 1993.  "Lawyers as Agents of the Devil in a Prisoner's Dilemma Game." National Bureau of Economics Research, Ic. Working Paper No. 4447.

    Babcock, L. C. 1991. "Impasses in Contract Negotiations: Why Do BAraginers Use Arbitration?" (manuscript on file with the authors).

    Babcock, L. C., mid C. A. Olson. 1992. "'The Causes of Impasses in Labor Disputes." Industiral Relations 31(2): 348.

    Babcock, L.C. G. Loewenstein, S. Issacharoff, and C. Camerer. "Negotiation Impasses: Causes and Cures." (manuscript on file with the authors).

    ____. 1995. "Biased Judgements of Fairness in Bargaining." American Economic Review 85:1337.

    Bolton, G. E. 1991. "A Comparative Model of Bargaining: Theory and Evidence."  American Economic Review 81:1096.

    Camerer, C. 1987. "Do Biases in Probability Judgement Matter in Markets? Experimental Evidence." American Economic Review 77: 981.

    Camerer, C. and G. Loewenstein, Forthcoming "Information, Fairness, and Efficiency in Bargaining."

    Camerer, C., G. Loewenstein, and M. Weber. 1989. "The Curse of Knowledge in Economic, Settings: An Experimental Analysis." Journal of Political Economics 97: 1232.

    Clermont, K. M. and J.D. Currivan. 1978. "Improving on the Contingent Fee." Cornell Law Review 63: 529.

    Coffee, J. C., Jr. l987. "Rethinking the Class Action: A Policy, Primer on Reform." Indiana Law Journal 62: 625.

    ____. 1986. "Understanding the Plaintiff's Attorney: The Implications of Economic Theory for Private Enforcement of Law through Class and Derivative Actions. Columbia Law Review 86: 669.

    Cooter, R., S. Marks, and R.H. Mnookin. 1982. "Baragining in the Shadow of the law: A 'I'estable Model of Strategic Behavior." Journal of Legal Studies 11: 225.

    Crawford, V.P. 1982. "A Theory of Disagreement in Bargaining." Econometrica 50:609.

    Eisenberg, T. 1989. "Litigation Models and Trial Outcomes in Civil Rights and Prisoner Cases." Georgetown Law Journal 77:1567.

    Evans v. Jeff D., 475 U.S. 717 (1986).

    Fiss, O. 1984. "Against Settlement." Yale Law Journal 93: 1073.

    Forsythe, R., J. Horowitz, N. E. Savin, mid M. Sefton. 1994. "Fairness in Simple Bargaining Experiments." Games and Economic Behavior 6: 347.

    Galanter, M. 1993. "News from Nowhere: The Debased Debate on Civil Jusfice." Denver Univeristy Law Review 71: 77.

    ____. 1983. "Reading the Landscape of Disputes: What We Know and Don't Know (And Think We Know) About Our Allegedly Contentious and Litgious Society." University of California at Los Angeles Law Review. 31: 1.

    _____. 1974. "Why the Haves Come Out Ahead: Specualtions on the Limits of Legal Change." Law & Society Review 9:95.

    Gould, J. P. 1973. "The Economics of Legal Conflicts."Journal of Legal Studies 2: 279.

    Gross, S. R., and K.D. Syverud. 1995. "Don't Try: Civil Jury Verdicts in a System Geared to Settlement" (manuscript on file with the authors).

    _____. 1991. "Getting to No: A Study of Settlement Negotiations and the Selection of Cases for Trial." Michigan Law Review 90: 319.

    Hastorf, A. H., and H. Cantril. 1954. "They Saw a Game: A Case Study." Journal of Abnormal and Social Psychology 49: 129.

    Hayes,B. 1984."Unions and Strikes With Assymmetric Information."  Journal of Labor Economics 2:57.

    Hensler, D. R. 1985. Asbestos in the Courts: The Challenge of Mass Toxic Torts. Santa Monica, CA: Institute for Civil Justice.

    Insurance Research Council. 1994. Auto Injuries: Claiming Behavior and Its Impact on Insurance Costs. Oak Brook, Ill. (manuscript on file with the authors.)

    Johnson, E., C. Camerer, S. Sen, and T. Rymon. 1991. "Behavior and Cognition in Sequential Bargaining." (manuscript on file with the authors.)

    Kahneman, D., J. L. Knetsch, and R. H. Thaler. 1986. "Fairness and the Assumpstions of Economics" Journal of Business 59(S): 285.

    Kahneman, D., P. Slovick, and A. Tversky (eds.). 1982 Judgement Under Uncertainty: Heuristics and Biases. New York: Cambridge.

    Kakalik, J., P. Ebener, W. Felstiner, and M. Shanely. 1983. Cost of Asbestos Litigation. Santa Monica, CA: Institute for Civil Justice.

    Katz, L. 1987. Bad Acts and Guilty Minds. Chicago: University of Chicago.

    Keilitz, S., R. Hanson, and H. W. K. Daley. 1993.  "Is Civil Discovery in State Trial COurts Out of OCntrol?" State Court Journal 17:8.

    Landes, W. M. 1971. "An Economic Analysis of the Courts." Journal of Law & Economics 14:61.

    Litan, R. E., and S.C. Salop. 1992. "More Value for the Legal Dollar: A New Look at Attorney-Client Fees and Relationships." Value Billing and Gaining a Competitive Advantage in the Legal Marketplace. Chicago, Ill.: AMerican ar Association (Section of Litigation).

    Loewentstein, G. S. Issacharoff, C. Camerer, and L. Babcock. 1993. "Self-Serving Assessments of Fairness and Pretrial Bargaining," Journal of Legal Studies 22: 135.

    Margolick, D. 1991.  "Address by Quayle on Justice Proposals Irks Bar Association." New York Times. (August 14). p. 1.

    McCall, B. P. l990. "Interest Arbitration and the Incentive to Bargain." Journal of Conflict Resolution 1990: 151.

    Messick, D. M., and K. Sentis. 1979. "Fairness and Preference." Journal of Experimental Social Psychology  15: 418.

    Miller, G. 1987. "Some Agency Problems Settlement."Journalof Legal Studies 16: 189.

    Miller, R. E. and A. Sarat. 1981. "Greivances, Claims, and Disputes: Assessing the Adevsary Culture." Law and Society Review 3-4:525.

    Mitchell, D. J. B., and M. L. Schwartz. l972. "'Theoretical Implications of Contigent Legal Fees." Quaterly Review of Economics and Business 12: 69.

    Mnookin, R. H. 1993. "Why Negotiations Fail: an Exploration of Barriers to the Resolution of Conflict."  Ohio State Journal on Dispute Resolution 8: 235.

    Mnookin, R. H. and L. Kornhauser. 1979. "Bargaining in the Shadow of the Law: The Case of Divorce." Yale Law Journal 88: 950.

    Mooney, S. F. 1994. "Where the Auto Premium Dollar." Insurnace Information Institute, Special Report:1.

    Mullenix, Linda S. 1994. "Discovery in Disarray: The Pervasive Myth of Discovery Abuse and the Consequences for Unfounded Rulemaking." Stanford Law Review 46: 1393.

    New Encyclopedia Britannica, 15th ed., S. V. "Bayes, Thomas."

    Ochs, J., and A. E. Roth. 1989. "An Experimental Study of Sequential Bargaining." American Economic Review, 79: 355.

    Posner, R. A. 1992. Economic Analysis of Law, 4th ed. Boston: Little, Brown.

    _____. 1973. "An Economic Approach to Legal Procedure and Judicial Administration." Journal of Legal Studies 2: 399.

    Priest, George L. 1985. "Reexamining the Selection Hypothesis." Journal of Legal Studies. 14: 215.

    Priest, G. L.,and B. Klien. 1984. "The Selection of Disputes for Litigation." Journal of Legal Studies, 13: 1.

    Ramseyer, J. M., and M. Nakazato. 1989. "The Rational Litigant: Settlement Amounts and Verdict Rates in Japan." Journal of Legal Studies 18: 263.

    Restatement (Third) of the Law Governing Lawyers, § 33. 1992. (March 16). Tentative Draft No. 5.

    Ross, M., and F. Sicoly. 1982. "Egocentric Biases in Availability and Attribution." P. 179 in Judgement Under Uncertainty: Heuristics and Biases, edited by D. Kahneman, P. Slovick, and A. Tuersky. New York: Cambridge.

    Roth, A. E., and J. K. Murnighan. 1982. "The Role of Information in Bargaining: An Expermental Study." Econometrica. 50: 1123.

    Rule 1.2. 1995. Model Rules of Professional Conduct. Chicago, IL: American Bar Association.

    Schelling, T. C. 1956. "An Essay on Bargaining." American Economic Review 46: 287.

    Selten, R. 1987. "Equity and Coalition Bargaining in Experimental Games." In Laboratory Experimentation in Economics: Six Points of View, edited by A. E. Roth. New York: Cambridge.

    Silver, C. 1996 "Do We Know Enough About Legal Norms?" In Social Rules: Origin; Character; Logic; Change, edited by D. Bratybrooke. Boulder, Co.:  Westview Press.

    Smith, S. K., C. J. DeFrances, P. A. Langan, and J. Goerdt. 1995. "Tort Cases in Large Counties." U.S. Department of Justice: Bureau of Justice Statistics Special Report. NCJ-153177.

    Spurr, S.J. 1988. "Referral Practices Among Lawyers: A Theoretical and Empirical Analysis."  Law and Social Inquiry 13:87.

    Stanley, L. R. and D. R. Coursey. 1990. "Empirical Evidence on the Selection Hypothesis and the Decision to Litigate or Settle."  Journal of Legal Studies 19: 145.

    Syverud, K. D. 1994.  "On the Demand for Liability Insurance."  Texas Law Review 72: 1629.

    Thaler, R.H. 1988.  "Anomalies:  The Ultimatum Game."  Journal of Economic Perspectives.   2: 195.

    Tracy, J.S. 1987.  "An Empirical Test of an Asymmetric Information Model of Stirkes."  Journal of Labor Economics 5: 149.