Damage Caps and Settlement

Damage Caps and Settlement: A Behavioral Approach Author(s): Linda Babcock and Greg Pogarsky Source: The Journal of Legal Studies, Vol. 28, No. 2 (June 1999), pp. 341-370 Published by: The University of Chicago Press for The University of Chicago Law School Stable URL: http://www.jstor.org/stable/10.1086/468054 Accessed: 11-01-2017 12:45 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms

The University of Chicago Law School, The University of Chicago Press are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Legal Studies

This content downloaded from 169.226.11.193 on Wed, 11 Jan 2017 12:45:13 UTC All use subject to http://about.jstor.org/terms

DAMAGE CAPS AND SETTLEMENT: A BEHAVIORAL APPROACH LINDA BABCOCK and GREG POGARSKY*

Abstract Since 1970, at least 30 states have enacted legislation capping the damages plaintiffs can recover in a lawsuit. Guided by previous research on suit and settlement, we outline a simple framework for examining the effects of a cap on litigant judgments and the pretrial settlement rate. We then introduce several refinements to the simple framework that are based on findings from behavioral economics and psychology. In particular, we recognize litigants may possess ‘‘role-specific biases’’ in information processing and may be subject to nonpecuniary influences during pretrial bargaining. Finally, the paper presents experimental evidence for the effects of a damage cap on litigants’ judgments and the settlement rate. The results show that behavioral and psychological theories can greatly expand the explanatory power of models that depict the pretrial bargaining environment. Everybody in America knows, at least most everybody knows, that our Civil Justice System is not working well. We do not think anybody really can stand up and defend the status quo of the litigation system in America. The average person on the street—we stop them in Hartford, New Haven, Bridgeport— knows that lawsuits take too long; that people do not get justice in a timely fashion; that too much money goes to lawyers. (Senator Joseph Leiberman, during Senate debate over product liability reform, March 21, 1996)

There is widespread popular sentiment that litigation is out of control.

Excessive litigation has been blamed for many social ills, among them high

* Linda Babcock is professor of economics at the H. John Heinz III School of Public Policy and Management, Carnegie Mellon University. Greg Pogarsky is assistant professor at the College of Business and Public Administration, University of Arizona. The research was supported by grant SBR9730348 from the National Science Foundation. The authors gratefully acknowledge generous insights from Colin Camerer, Shane Frederick, Samuel Issacharoff, Russel Korobkin, George Loewenstein, Lisa Ordo´n˜ez, Lowell Taylor, Richard Thaler, participants in the Russel Sage Foundation’s 1996 Summer Workshop in Behavioral Economics, attendees of the Russel Sage Foundation’s Institute for Behavioral Economics 1997 conference, and attendees of the 1997 conference in Behavioral Law and Economics, sponsored jointly by the Russel Sage Foundation and the National Bureau of Economic Research. [ Journal of Legal Studies, vol. XXVIII ( June 1999)]  1999 by The University of Chicago. All rights reserved. 0047-2530/99/2802-0003$01.50

341 This content downloaded from 169.226.11.193 on Wed, 11 Jan 2017 12:45:13 UTC All use subject to http://about.jstor.org/terms

342

THE JOURNAL OF LEGAL STUDIES

insurance premiums, reduced technological innovation, higher prices, and unemployment. During the past several decades, many states have enacted tort reform legislation in an attempt to curb litigation. Some of these reforms have limited the damages plaintiffs can recover in a lawsuit.1 This paper reports results from experiments that test the effect of damage caps on pretrial bargaining between civil disputants. The extent of the alleged litigation ‘‘explosion’’ is debatable. Data from the National Center for State Courts indicate that for a sample of 16 states, tort filings have declined since 1986, and contract filings have declined since 1990.2 Yet heightened media coverage of a few, select cases 3 and increasingly vitriolic public discourse about our courts help sustain pressure for change. In 1990, Congress passed the Civil Justice Reform Act (CJRA), which required each federal district to adopt an ‘‘expense and delay reduction plan’’ for civil litigation in the federal courts.4 On December 1, 1993, the mandatory disclosure provisions of amended Rule 26 of the Federal Rules of Civil Procedure became effective. The new rule eliminated the need for a formal discovery request for many categories of information 5 and acceler1

Hereafter, a ‘‘damage cap’’ or ‘‘cap.’’ The study concludes that the raw volume of tort and contract filings has fallen, not merely the per capita rate. See Brian J. Ostrom & Neal B. Kauder, Examining the Work of State Courts, 1995: A National Perspective from the Court Statistics Project 26 (report from Nat’l Center for State Courts 1996) (concluding, ‘‘[T]he bottom line is that there is no evidence of an ‘explosion’ in the volume of tort filings’’). 3 We refer here to two, widely criticized recent cases. In the first, a New Mexico woman won a $2.9 million jury award after spilling a cup of McDonald’s coffee in her lap, and in the second, an Alabama doctor was awarded $4 million by a jury because BMW had clandestinely ‘‘retouched’’ the paint job on his new car. Emphasizing several salient cases can exaggerate perceptions of problems within the civil justice system. Memorable examples of events can cause people to overestimate their prevalence. See Paul Slovic, Baruch Fischoff, & Sarah Lichtenstein, Facts versus Fears: Understanding Perceived Risk, in Judgement under Uncertainty: Heuristics and Biases (Daniel Kahneman, Paul Slovic, & Amos Tversky, eds. 1982). At least one study concludes the media exaggerate litigation rates and damage awards. See Stephen Daniels, The Question of Jury Competence and the Politics of Civil Justice Reform: Symbols, Rhetoric, and Agenda-Building, 52 L. & Contemp. Probs. 269 (1989). After a ‘‘content analysis’’ of newsmagazine coverage of the civil justice system, Bailis & MacCoun found that the media overrepresent product liability and medical malpractice cases, exaggerate plaintiff win rates, and distort the actual distribution of jury awards (Daniel S. Bailis & Robert J. MacCoun, Estimating Liability Risks with the Media as Your Guide: A Content Analysis of Media Coverage of Tort Litigation, 20 L. & Hum. Behav. 419 (1996)). 4 See 28 U.S.C. 471 et seq. Under the act, each federal district was required to adopt a plan by December 31, 1993. However, districts opting to enact a plan before December 31, 1991, were designated ‘‘early implementation’’ districts and received additional federal funding. 5 Thus once the lawsuit commences, each side must identify the names of all persons with relevant information, disclose the content of that information, furnish copies or descriptions of relevant documents, compute all damages suffered, and furnish copies of all insurance agreements that could underwrite any eventual judgment (Fed. R. Civ. P. 26(a)(1)). 2


DAMAGE CAPS

343

ated the time limits for discovery compliance.6 Congress enacted these provisions despite substantial opposition,7 and (by some accounts) without careful analysis of their potential impacts.8 During the advent of the CJRA and mandatory disclosure, President Bush convened the President’s Council on Competitiveness, chaired by Dan Quayle. In 1991, the council produced a widely publicized report containing proposals to reform the civil justice process.9 The proposals were ‘‘geared toward reducing excessive and unnecessary litigation and decreasing the costs and time associated with resolving disputes.’’ 10 One of the 6 Under the new rule, information covered under mandatory disclosure must be produced within 10 days after a preliminary conference, which should be held as soon as practicable after the case has commenced (Fed. R. Civ. P. 26(f )). 7 See, for example, Griffin B. Bell, Chilton Davis Varner, & Hugh Q. Gottschalk, Automatic Disclosure in Discovery—the Rush to Reform, 27 Ga. L. Rev. 1, 4–5 (1992) (arguing that ‘‘although reform of the discovery process is clearly necessary, the authors believe that the automatic disclosure process would be counterproductive. It would increase the cost and inefficiency of the pretrial process and create a new welter of problems, including’’ confusion and acrimony in the pretrial process, increased costs, delay, and burden on the court system, overproduction of documents, and potential confusion about ethical obligations). 8 Much criticism resulted from the failure to study the experiences in pilot districts. See Linda S. Mullenix, Hope over Experience: Mandatory Informal Discovery and the Politics of Rulemaking, 69 N.C. L. Rev. 795 (1991) (recounting the experiences in pilot districts); Carl Tobias, In Defense of Experimentation with Automatic Disclosure, 27 Ga. L. Rev. 665 (1993) (arguing against mandatory disclosure before substantial experimentation). See also Amendments to the Federal Rules of Civil Procedure, 146 F.R.D. 402, 511 (1993) (J. Scalia dissenting: ‘‘It seems . . . most imprudent to embrace such a radical alteration that has not . . . been subjected to any significant testing on the local level’’). Well-intended policies can have unexpected impacts. Issacharoff and Loewenstein demonstrate this by recognizing first that ‘‘the net effect of mandatory disclosure on litigation costs will depend not only on its impact on discovery costs when litigation occurs, but also on the rate of settlement prior to discovery’’ (Samuel Issacharoff & George Loewenstein, Unintended Consequences of Mandatory Disclosure, 73 Tex. L. Rev. 753, 759 (1995)). The authors then observe that mandatory disclosure bifurcates the early settlement period into two periods: one prior to mandatory disclosure and one afterward. Assuming litigants are myopic—that is, when deciding whether to settle they factor in only discovery costs that would be imminent absent a settlement— such bifurcation makes failing to settle less costly. Further, the cost of impasse, which consists mostly of the costs of complying with the disclosure requirements, will be borne almost entirely by defendants, thereby encouraging ‘‘strike’’ or ‘‘nuisance’’ suits. There have been other recent efforts to examine the behavioral or psychological aspects of legal institutions. See Russel Korobkin, The Status Quo Bias and Contract Default Rules, 83 Cornell L. Rev. 608 (1998); Russel Korobkin & Chris Guthrie, Psychology, Economics, and Settlement: A New Look at the Role of the Lawyer, 76 Tex. L. Rev. 77 (1997); Jeffrey J. Rachlinski, A Positive Theory of Judging in Hindsight, 65 U. Chi. L. Rev. 571 (1998); and Jeffrey J. Rachlinski, Gains, Losses and the Psychology of Litigation, 70 S. Cal. L. Rev. 113 (1996). 9 President’s Council on Competitiveness, Agenda for Civil Justice Reform in America (August 1991); see Dan Quayle, Civil Justice Reform, 41 Am. U. L. Rev. 559, 561 (1992), for a summary. Among the proposals were mandatory settlement conferences, restricting expert testimony, and implementing the ‘‘English Rule’’ requiring the ‘‘loser’’ to pay both parties’ attorney fees. 10 Quayle, supra note 9, at 561.


344


Quayle proposals called for federal legislation to limit punitive damages to the amount of compensatory damages awarded.11 In May of 1996, Congress passed the Commonsense Legal Reform Act. The legislation capped punitive damages in product liability cases at the larger of $250,000 or two times compensatory damages for large businesses, and the lesser of the two for small businesses or individuals.12 The act also imposed a uniform statute of repose for durable goods of 15 years, eliminated joint and several liability for noneconomic damages, made recovery more difficult for suppliers of raw materials for life-sustaining medical devices, and preempted state product liability law. President Clinton immediately vetoed the bill. Though the 1996 veto quelled momentum toward a federal product liability cap, approximately 30 states currently employ some form of liability limit. States began to enact damage-cap legislation during the medical malpractice insurance crisis of the mid-1970s. A second wave of cap statutes arrived during the mid-1980s. While most of the earlier statutes pertained to medical malpractice cases, many later statutes were broader, some applying to entire categories of damages in all cases.13 Currently, there exist as many different cap schemes as states that employ them, ranging from Georgia’s straightforward cap, which limits punitive damages to $250,000,14 to elaborate attempts to tailor punitive damages to the assets of the defendant and the degree that the defendant benefited from its tortious conduct.15 Thus, while some states limit punitive damages,16 others limit noneconomic, com-

11

Id. at 565. The act defined a ‘‘small’’ business as one that either had a net worth below $500,000 or fewer than 25 full-time employees. 13 See, for example, Ala. Code 6-11-21 (1993) (capping punitive damages at $250,00); W. Va. Code 55-7B-8 (1994) (limiting noneconomic recovery to $1,000,000). 14 Ga. Code Ann. 51-12-5.1(g). 15 By far the most complicated and fraught with contingencies is Kansas. Pursuant to Kan. Stat. Ann. 60-3701: ‘‘(e) Except as provided by subsection (f), no award of exemplary or punitive damages pursuant to this section shall exceed the lesser of: (1) The annual gross income earned by the defendant, as determined by the court bases upon the defendant’s highest gross annual income earned for any one of the five years immediately before the act for which such damages are awarded, unless the court determines such amount is clearly inadequate to penalize the defendant then the court may award up to 50 percent of the net worth of the defendant, as determined by the court; or (2) $5 million. (f) In lieu of the limitation provided by subsection (e), if the court finds that the profitability of the defendant’s misconduct exceeds or is expected to exceed the limitation of subsection (e), the limitation on the amount of exemplary or punitive damages which the court may award shall be an amount equal to [one and one-half] times the amount of profit which the defendant gained or is expected to gain as a result of the defendant’s misconduct.’’ 16 See, for example, Ala. Code 6-11-21 (1993) (capping punitive damages at $250,00); N.D. Cent. Code 32-03.2-11(4) (Supp. 1994) (limiting punitive damages to $250,000 or twice compensatory damages, whichever is greater). 12


DAMAGE CAPS

345

pensatory damages.17 States employ a flat dollar cap,18 a ‘‘multiplier,’’ 19 or some combination of both.20 Some caps pertain to all civil cases,21 while others apply to certain classes of actions, such as medical malpractice 22 or product liability.23 Though diverse, the variety of statutory damage limitations share a common feature—they circumscribe a previously unbounded array of potential trial outcomes.24 Despite the flurry of reform activity, there is little empirical evidence about the effects of various tort reforms on litigation. This scarcity partly reflects the practical difficulties of conducting field research in this area. Not all states collect extensive data about the functioning of their courts. Those that do lack a uniform system of caseload reporting, which impedes cross-state comparison of litigation trends. Further, structural changes in state court systems can hamper interpretation of time series data.25 17 See, for example, Cal. Civ. Code 3333.2 (Bancroft & Whitney 1995) (placing a $250,000 cap on ‘‘noneconomic’’ damages, which are defined as ‘‘compensation for pain, suffering, inconvenience, physical impairment, disfigurement, and other nonpecuniary injury’’); Idaho Code 6-1603 (1994) (limiting noneconomic damages to $400,000 beginning in 1994 and adjusting it upward each year thereafter, but excluding punitive damages from the cap). 18 See, for example, W. Va. Code 55-7B-8 (1994) (limiting noneconomic recovery to $1,000,000); Va. Code Ann. 8.01-581.15 (Michie 1984) (capping all punitive damage awards at $350,000). 19 See, for example, Conn. Gen. Stat. Ann. 52-240b (West 1989) (limiting punitive damages to twice actual damages in product liability actions). 20 See, for example, N.D. Cent. Code 32-03.2-11(4) (Supp. 1994); Nev. Rev. Stat. Ann. 42.005 (Michie Supp. 1993) (limiting punitive damages to three times compensatory damages when compensatory damages exceed $100,000; otherwise punitive damages capped at $300,000). 21 See, for example, N.C. Gen. Stat. 1D-25 (1995) (limiting punitive damages in all civil cases to three times compensatory damages or $250,000, whichever is greater). 22 See, for example, Colo. Rev. Stat. 13-64-302 (1994) (limiting damages for medical malpractice against a hospital or physician to $1,000,000 per patient); Md. Code Ann., Cts. & Jud. Proc. 11–108 (1991) (imposing a $350,000 limit on noneconomic damages for any personal injury claim for medical malpractice). 23 See, for example, Conn. Gen. Stat. Ann. 52-240b (West 1989). 24 There are two exceptions. First are the ‘‘mandatory contribution’’ regimes where, for example, a portion of any punitive damages award is paid to a state fund. See, for example, Mo. Ann. Stat. 537.675(2) (Vernon 1988) (mandating that 50 percent of all punitive damages awards are payable to the Tort Victims’ Compensation fund); N.Y. Civ. Prac. L. & R. 8701 (McKinney Supp. 1995) (ordering 20 percent of punitive damages awards payable to the state). Second are states that have increased the burden of proof for punitive damages claims. See, for example, Alaska Stat. 09.17.020 (1994) (mandating no punitive damages unless proven by clear and convincing evidence). Neither of these reform measures make particular dollar recoveries impossible, only more difficult. We acknowledge that all methods for limiting recovery might not affect pretrial bargaining in the same way. In this paper, we employed a straightforward cap, one that limited plaintiff ’s recovery to a specific dollar amount. Future research might explore whether different capping methods produce different results. 25 ‘‘Roving jurisdictional limits’’ can complicate interpretation of trend data. For example, Alaska has a general jurisdiction trial court (the superior court) and a court with, inter alia,


346


There have nevertheless been several efforts to study the effect of litigation reforms. A recent study by Campbell, Kessler, and Shepherd examined the effect of liability reforms on the economic performance of industry, using a panel data set with industry-level data by state for the years 1969– 90.26 They divided eight potential reforms into two categories, those likely to increase industry liability and those likely to decrease it. While they found a positive association between the number of liability-decreasing reforms (of which damage caps were one) and industry performance, they did not isolate the effect attributable to damage caps. Since early cap laws were a direct response to spiraling medical malpractice insurance premiums, research has primarily considered the effect of caps on the value of claims paid by medical malpractice insurance carriers. Danzon found that caps reduced the average value of paid medical malpractice claims by 23 percent during 1975–84.27 Sloan, Mergenhagen, and Bovbjerg found similar effects for the same time period.28 However, a 1991 study by Gronfein and Kinney found that the value of paid medical malpractice claims in Indiana increased once a cap was in place, relative to Ohio and Michigan where there were no caps.29 While this study has several limitations,30 its primary finding is consistent with recent limited jurisdiction over tort and contract cases (the district court). In 1986, Alaska instituted a $500,000 cap on noneconomic damages in personal injury cases, and between 1984 and 1988, superior court civil filings plummeted. However, during the period 1975–95, the jurisdictional limit for the district court rose from $10,000 to $50,000. In this instance, focusing only on general jurisdiction filings could be misleading since it is likely the increasing jurisdictional limit of the lower court siphoned civil filings from the higher court. It can therefore be difficult to disentangle the effects of the cap from the effects caused by litigants taking increasing advantage of the less formal litigating forum. Another hurdle for field analyses arises when states either abolish a court of limited jurisdiction or create one anew. Limited jurisdiction litigation generally involves vastly reduced discovery and a setting more conducive to pro se litigation, factors that promote the more expeditious resolution of cases. If a state combines their limited and general jurisdiction courts into a unified court, comparisons across years become difficult. Litigants who would have filed claims in the lower court employing expedited procedures may be dissuaded from entering the unified system because the costs (time, expense, aggravation) have become prohibitive. This casts doubt on any claim that total filings across the 2 years declined due to some policy intervention. 26 Thomas Campbell, Daniel P. Kessler, & George B. Shepherd, The Causes and Effects of Liability Reform: Some Empirical Evidence (Working Paper No. 4989, Nat’l Bur. Econ. Res. 1995). 27 Patricia M. Danzon, The Frequency and Severity of Medical Malpractice Claims: New Evidence, 49 L. Contemp. Probs. 57, 76–77 (1986). 28 Frank A. Sloan, Paula M. Mergenhagen, & Randall R. Bovbjerg, Effects of Tort Reform on the Value of Closed Medical Malpractice Claims: A Microanalysis, 14 J. Health Pol. Pol’y & L. 663 (1989). 29 William P. Gronfein & Eleanor DeArman Kinney, Controlling Large Malpractice Claims: The Unexpected Impact of Damage Caps, 16 J. Health Pol. Pol’y & L. 441 (1991). 30 First, the authors did not employ data for the period preceding institution of the cap. This precludes separation of ‘‘trend’’ effects from those genuinely attributable to the policy


DAMAGE CAPS

347

experimental research on jury awards. Saks et al. asked ‘‘mock jurors’’ to provide jury awards for three factual scenarios, in which the injuries were minor, moderate, or serious.31 There were two experimental treatments, one with no damage cap and one in which subjects were informed of a $250,000 damage cap. While the average trial award for minor injuries was $3,895 in the uncapped treatment, it was $15,718 in the capped treatment.32 Saks et al. suggest the damage cap served as an ‘‘anchor,’’ attracting subjects’ judgments upward. Beyond settlement amounts and jury awards, damage caps may also affect the duration and success rate of settlement discussions. Sloan, Mergenhagen, and Bovbjerg examined the relationship between caps and case duration.33 Using data on closed medical malpractice claims for 1975– 78 and 1984, they found that damage caps reduced two distinct time periods: from the injury to the filing of a claim and from filing to termination. We attempt here to further this line of research by investigating how caps affect the proportion of settled cases. However, our work departs from the strict application of the economic framework prevalent in much research on this topic. We conceptualize litigant behavior more broadly, integrating relevant behavioral and psychological theories into the law and economics approach. In Section I, we outline a simple theoretical framework, guided by previous research in law and economics, that under several assumptions yields a clear prediction about the effect of a damage cap on the pretrial settlement rate. We then enrich this model by recognizing litigants may possess rolespecific biases in information processing and may allow nonpecuniary motivations to affect their bargaining positions. Finally, we explore the predicted effect of a damage cap on settlements using this enriched framework. In Section II, we describe our experiment in which subjects assigned the role of plaintiff or defendant in a dispute attempt to negotiate a pretrial setintervention. More troublesome, however, is the unique manner of processing medical malpractice claims in Indiana following the reforms. To obtain protection, medical service providers were required to obtain insurance up to $100,000. They also paid a surcharge for a Patient Compensation Fund (PCF). Once a private insurance provider settled a claim for its maximum exposure of $100,000, the PCF assumed defense of the claim. For such claims, liability was not an issue, the only question being the extent of damages. Further, there was some indication that the lawyers who administered the fund were young and inexperienced, and consequently overmatched by more savvy plaintiffs attorneys. See id. at 458–59. These factors likely elevated the mean value of claims in Indiana during the study period, making it difficult to disentangle the pure effect of the damage cap from the other procedural changes. 31 Michael J. Saks et al., Reducing Variability in Civil Jury Awards, 21 L. & Hum. Behav. 243 (1977). 32 Id. at 251. 33 See supra note 28.


348


tlement. In Section III, we examine the effect of a damage cap on the likelihood the subjects are able to reach a pretrial settlement. We also test the predictions of our enriched framework and investigate the importance of these ‘‘behavioral’’ modifications for explaining subject behavior. Finally, Section IV concludes. I. Theories about Impasse Pretrial bargaining can be viewed as a two-stage process: the first in which litigants develop initial judgments about the trial outcome and the second during which actual bargaining occurs. Most of the previous literature has focused on the second stage, which models the strategic interactions that occur after the parties have made judgments about their case.34 This paper concerns the first stage, in which initial litigant judgments help create the negotiation environment. Such prefatory judgments can affect the parties’ ability to reach a pretrial settlement. For example, the plaintiff and defendant may view a case so dissimilarly that there is no single agreement that both parties prefer to a trial. Alternatively, opposing litigants can form coincident assessments that generate a wide range of acceptable agreements and, hence, a high probability of settlement. In order to explore the effect of a cap on the negotiation environment, we begin with a simple model. A. A Simple Model of the Negotiation Environment This section outlines a simple model of the pretrial negotiation environment based on previous rational choice research in law and economics.35 In order to evaluate settlement proposals during pretrial bargaining, litigants estimate the monetary outcome if their case were tried: Tp ⫽ T ⫹ e p ,

(1)

Td ⫽ T ⫹ e d ,

(2)

where Tp and Td are plaintiffs’ and defendants’ estimated trial outcomes,36 T is the actual trial outcome, and ep and ed are estimation errors that arise 34 This large literature, based upon game-theoretic models of bargaining, is summarized in Andrew Daugherty, Settlement, in Encyclopedia of Law and Economics (B. Bocjaert & G. De Geest eds., forthcoming 1999). 35 See William M. Landes, An Economic Analysis of the Courts, 14 J. Law & Econ. 61 (1971); Richard A. Posner, An Economic Approach to Legal Procedure, 2 J. Legal Stud. 399 (1973); and George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Stud. 1 (1984). 36 Previous research has employed two, distinct determinants of T: P, representing the estimated probability of recovery, and J, the size of the potential judgment. See Priest & Klein, supra note 35, at 12–13; Posner, supra note 35, at 555–56. In our discussion, T ⫽ P ⫻ J.


DAMAGE CAPS

349

because litigants are unable to predict the trial outcome with perfect accuracy. In each lawsuit, we assume the parties draw their estimation errors, and therefore their estimated trial outcomes, from the same probability distribution.37 As a result, the plaintiff and defendant, on average, expect the same trial result. Though divergence in any one case is possible, it is just as likely that the plaintiff ’s estimate will exceed the defendant’s estimate as it is that the defendant’s will exceed the plaintiff ’s.38 The model represents each litigant’s ‘‘bottom line’’ bargaining position with a reservation value. The plaintiff ’s reservation value is the minimum dollar amount that is preferable to a trial. The defendant’s reservation value is the maximum acceptable payment in lieu of trial. Reservation values are determined by the expected trial outcome and anticipated litigation costs. A plaintiff should settle for at least the expected trial award minus the litigation costs the plaintiff will save by settling. Defendants should settle by paying the plaintiff at most the expected trial award plus the costs of defense.39 The parties’ reservation values can be expressed as follows: RVp ⫽ Tp ⫺ C p ,

(3)

RVd ⫽ Td ⫹ C d ,

(4)

where Tp and Td are the estimated trial outcomes from above, and C p and C d are litigation costs. If the maximum the defendant will pay to avoid trial exceeds the plaintiff’s minimum demand, settlement is possible. Thus, RVp ⬍ RVd

(5)

is a necessary condition for settlement. The amount by which RVd exceeds RVp is a contract or settlement zone, since it represents a range of possible payments from defendant to plaintiff that both parties prefer to a trial. Substituting equations (3) and (4) into equation (5) yields an alternative expression for the existence of a contract zone: Tp ⫺ Td ⬍ C p ⫹ C d .

(6)

Under equation (6), if Td ⬎ Tp , there must be a contract zone since C p and C d are nonnegative. If Tp ⬎ Td , a contract zone exists when aggregate litigation costs exceed the difference in predicted trial outcomes. Thus, whether 37 This model assumes the parties have imperfect (there is uncertainty about the trial outcome) but symmetric information (neither side holds private information about the likely trial outcome). 38 See Priest & Klein, supra note 35, at 9–10. 39 This simple model assumes risk neutrality. For a summary of settlement models that incorporate risk aversion, see Daugherty, supra note 34.


350


Figure 1.—Hypothetical belief distributions

condition (6) is satisfied relates inversely to the amount by which the plaintiff predicts a larger trial result than the defendant. Figure 1 contains two hypothetical probability distributions for the parties’ expectations about the trial outcome, one with a higher variance. Regardless of the distribution, settlement is possible if total costs exceed (Tp ⫺ Td ). However, the tighter distribution in the bottom panel of Figure 1 renders large disparities in beliefs less likely, thus (ceteris paribus) increasing the probability that condition (6) is met.


DAMAGE CAPS

351

Various factors can affect uncertainty about the trial outcome. Cases of first impression should more readily produce large errors in judgments about the trial outcome, while parties to a case with ample precedents should more likely expect comparable outcomes. The characteristics of the jury pool may also affect the uncertainty about the trial outcome. Greater homogeneity of candidate jurors should better allow litigants to predict the characteristics of an eventual jury and the resulting verdict. The Effect of a Damage Cap. A cap can reduce uncertainty by narrowing the range of possible trial outcomes. The degree of uncertainty reduction will depend on the location of the cap along the distribution of predicted trial outcomes. In the top panel of Figure 1, the cap lies in the upper tail of the belief distribution. We refer to a cap that precludes at least one trial outcome the litigants would have thought possible as a ‘‘binding’’ cap. The cap is ‘‘nonbinding’’ for the distribution depicted in the bottom panel since it exceeds what both litigants believe to be the highest possible trial award. Suppose the top panel of Figure 1 represents the parties’ precap beliefs. When a cap of D is imposed, trial outcomes above D are no longer possible. What happens to the precluded portion of the belief distribution? When a jury award exceeds an applicable damage cap, the court reduces the award to the maximum allowable amount. Litigants may simply mimic this process and truncate the distribution of predicted trial outcomes, producing a spike at D.40 Litigants seeking accurate predicted trial outcomes should most readily ‘‘truncate’’ when they believe their jury will be ignorant of the cap. A majority of cap statutes address this issue directly,41 though some remain silent.

40

This implies that the cap does not effect probability mass below the cap value. And of statutes with jury-notification provisions, a majority prohibit instruction about the cap. See Mo. Ann. Stat. 538.210 (‘‘where the trier of fact is a jury, such jury shall not be instructed by the court with respect to the limitation on an award of noneconomic damages, nor shall counsel for any party or any person providing testimony during such proceeding in any way inform the jury or potential jurors of such limitation’’); Md. Code Ann., Cts. & Jud. Proc. 11-108(d)(1) (‘‘In a jury trial, the jury may not be informed of the limitation established under section (b) of this section’’); and Ala. Code 6-5-544(b) (‘‘During the trial of any action neither the court nor any party shall advise or infer to the jury that it may not return an award for noneconomic losses in excess of an amount specified herein; in the event the jury is so advised or such inference is made, the trial court, upon motion of an opposing party, shall immediately declare a mistrial’’); but see Wis. Stat. Ann. 893.55(4)(c) (‘‘If an action is before a jury, the jury shall make a finding as to noneconomic damages without regard to the limit under par. (d). If the jury finds that noneconomic damages exceed the limit, the jury shall make any reduction required’’); and Mass. Gen. Laws Ann. ch. 231, 60h (‘‘the court shall instruct the jury that in the event they find the defendant liable, they shall not award the plaintiff more than five hundred thousand dollars for pain and suffering, loss of companionship, embarrassment, and other items’’). 41


352


Of course, a statutory prohibition cannot guarantee juror ignorance, since other sources exist for the information, including previous experience with the courts, private conversations, and newspapers. Finally, despite juror awareness, litigants may still truncate if they believe the jury will do so to calculate its award. If both litigants truncate the belief distribution, a damage cap increases the likelihood of a contract zone and the prospects for settlement.42 Since the modified distribution has a smaller variance than its untruncated counterpart, the cap renders extreme disparities in predicted trial outcomes less likely. This increases the probability that the necessary condition for a contract zone, condition (6) above, is satisfied. We next incorporate insights from psychology and behavioral economics into the simple framework outlined here. We further present alternative methods for litigants to modify the belief distribution when they believe the jury knows of a cap. The next section proposes three refinements to this basic model, each providing richer, and sometimes conflicting, predictions for the effect of a cap on the existence of a contract zone. B.

Role-Specific Biases

Consistent with the law and economics literature, the preceding analysis acknowledges only the pecuniary motives of litigants and attributes the absence of a settlement zone to random estimation errors. We next introduce a richer conception of litigants, allowing the presumed desire for a favorable trial result to generate predictable, ‘‘nonrandom’’ estimation errors and for considerations of fairness to influence judgments. We propose three refinements to the previous model, each related to a party’s role as plaintiff or defendant in the litigation. Each refinement supplies a prediction that is either more detailed or antithetic to the predictions from our earlier discussion. Precap Judgments. The simple framework presented above assumes opposing litigants’ beliefs about the trial outcome contain random estimation errors. As a result, while trial estimates for any pair of litigants might diverge, the average disparity across all pairs should be zero. This view neglects the abundant empirical evidence that individuals consistently exhibit

42 Throughout this paper, we assume the probability of settlement relates positively to both the likelihood and size of the contract zone. See Linda Babcock, George Loewenstein, & Xianghong Wang, The Relationship between Uncertainty, the Contract Zone, and Bargaining Efficiency, 22 J. Econ. Behav. & Org. 475 (1995). For dissenting views, see David Bloom, Is Arbitration Really Compatible with Bargaining? 20 Indus. Rel. 233 (1981); and Vince Crawford, A Theory of Disagreement in Bargaining, 50 Econometrica 607 (1982).


DAMAGE CAPS

353

‘‘self-serving biases’’ during negotiations—which, in a trial setting, means that plaintiffs tend to predict higher trial outcomes than defendants.43 In a series of studies, Babcock et al. have demonstrated that subjects consistently give optimistic predictions of trial outcomes. In these experiments, subjects were randomly assigned the role of plaintiff or defendant and given detailed materials outlining a personal injury lawsuit. Before settlement discussions, each subject predicted the trial outcome after being assured their prediction would not be shared with their adversary. Although plaintiff and defendant subjects received identical case materials, plaintiffs’ estimates exceeded defendants’ estimates by a substantial margin.44,45 The existence of self-serving biases suggests the jointly held probability distribution in Figure 1 inadequately depicts the parties’ precap beliefs. Instead, plaintiffs’ beliefs are likely to be distributed across higher amounts than defendants’ beliefs, consistent with the hypothetical distributions in Figure 2. 43 See Linda Babcock et al., Biased Judgments of Fairness in Bargaining, 85 Am. Econ. Rev. 1337 (1995); George Loewenstein et al., Self-Serving Assessments of Fairness and Pretrial Bargaining, 22 J. Legal Stud. 135 (1993); Linda Babcock & George Loewenstein, Explaining Bargaining Impasse: The Role of Self-Serving Biases, 11 J. Econ. Persp. 109 (1997). The evidence for self-serving biases in litigation is consistent with the large psychology literature on self-serving biases in other domains. These biases have been found in selfassessments of skill. See Ola Svenson, Are We All Less Risky and More Skillful than Our Fellow Drivers? 97 Acta Psychol. 143 (1981) (for driving); R. Baumhart, An Honest Profit (1968) (for ethics); L. Larwood & W. Wittakher, Managerial Myopia: Self-Serving Biases in Organizational Planning, 62 J. Applied Psychol. 194 (1977) (for managerial skills); and P. Cross, Not Can but Will College Teaching Be Improved, 17 New Directions Higher Educ. 1 (1977) (for productivity). People also overestimate their contribution to joint tasks. When spouses estimate the proportion of all household tasks they perform, the estimates for a couple typically sum to more than one. See Michael Ross & Fiore Sicoly, Egocentric Biases in Availability and Attribution, 37 J. Personality & Soc. Psychol. 322 (1979). 44 Babcock et al., supra note 43, at 1340. These results squarely contradict rational choice models of settlement. Daugherty, supra note 35, at 20, notes that ‘‘for the consistent application of rational choice, differences in assessments must reflect differences in private information, not differences in world views. Presented with the same information, conflicts in assessments would disappear.’’ 45 Researchers have examined the reason for this bias. See D. Dunning, J. A. Meyerowitz, & A. D. Holzberg, Ambiguity and Self-Evaluation: The Role of Idiosyncratic Trait Definitions in Self-Serving Assessments of Ability, 57 J. Personality Soc. Psychol. 1082 (1989); John Darley & Paget Gross, A Hypothesis-Confirming Bias in Labeling Effects, 44 J. Personality Soc. Psychol. 20 (1983); and Rasyid Sanitioso, Ziva Kunda, & Geoffrey Fong, Motivated Recruitment of Autobiographical Memories, 59 J. Personality Soc. Psychol. 229 (1990). The prevailing view attributes the bias to selective information processing. Thus, ‘‘people attempt to construct a rational justification for the conclusions that they want to draw. To that end, they search through memory for relevant information, but the search is biased in favor of information that is consistent with the desired conclusions. If they succeed in finding a preponderance of such consistent information, they are able to draw the desired conclusion while maintaining an illusion of objectivity’’ (Sanitioso, Kunda, & Fong, supra, at 229).


354


Figure 2.—Biased belief distributions

This role-specific difference in precap beliefs suggests a damage cap will produce an additional positive effect on the likelihood of settlement that is not captured by the model outlined in Section IA. As we assume in Section IA, if the two litigants share a probability distribution and make random estimation errors, the expected disparity in trial outcomes both before and after the cap is zero. However, if the plaintiff ’s distribution lies to the right of the defendant’s distribution, the average precap disparity in trial predictions is positive. Recall that a contract zone exists when Tp ⫺ Td ⬍ C p ⫹ C d. With a damage cap of D, the litigants must revise all estimates above D downward. A cap that is binding on at least the plaintiff ’s distribution will again reduce the variance of Tp ⫺ Td , increasing the likelihood of a contract zone. However, if beliefs are as depicted in Figure 2, there is greater downward revision for plaintiffs than for defendants. Such asymmetry reduces the average precap disparity in point estimates created by the parties’ self-serving biases. This decreases the likelihood of a large disparity in trial estimates, increasing the prospects for settlement. Thus, if the parties begin with self-serving expectations of the trial outcome, the cap’s effects extend beyond the tendency to reduce the uncertainty of verdict estimation. This first behavioral refinement therefore produces a more detailed prediction about the mechanism through which caps can affect settlement rates.


DAMAGE CAPS

355

Revising Judgments under a Cap. We discussed earlier how ‘‘capped’’ trial predictions may depend on the litigant’s assessment of whether their jury will know of the cap. Under the truncation approach, predictions above the cap are revised to the cap amount, and predictions below the cap value are unaffected. Litigants are likely to truncate in one of two circumstances. First, litigants who are satisfied their jury will be unaware of an applicable cap may simulate the process followed by the court. Second, litigants who believe the jury will be aware of the cap may truncate if they conclude their jury will do the same. There are, however, alternative heuristics for predicting trial outcomes in the presence of a cap that we illustrate via example. Assume litigants A and B are involved in separate, identical lawsuits and each would have expected a $500,000 trial outcome, but for a $250,000 cap. If litigant A ‘‘truncated,’’ she would predict a $250,000 award, reasoning the highest possible result is appropriate since the case would have generated a greater award, were it not for the cap. An alternative approach, termed ‘‘recalibration,’’ has the litigant estimate the relative severity of the civil infraction and then assign a corresponding dollar value on the new, bounded scale. For example, suppose litigant B placed the civil injury in the fiftieth percentile of possible transgressions. With a $250,000 cap, litigant B would predict a trial result of $125,000, precisely midway along the capped range. ‘‘Anchoring and adjustment’’ provides another alternative to truncation. In several classic studies, Kahneman and Tversky found individuals use ‘‘anchors’’ or ‘‘focal points’’ when making judgments and then adjust their estimate from the anchor. Since such adjustments tend to be insufficient,46 beliefs are biased toward the anchor. This phenomenon occurs even when potential anchors are ostensibly irrelevant to the judgment task.47 Rather than recalibrate, litigants may use the damage cap as a starting point and subtract an amount that yields their trial prediction. Prior psychological research suggests motivational biases related to a 46 See Paul Slovic & Sarah Lichtenstein, Comparison of Bayesian and Regression Approaches to the Study of Information Processing in Judgment, 6 Org. Behav. & Hum. Performance 649 (1971). 47 See Amos Tversky & Daniel Kahneman, Judgment under Uncertainty: Heuristics and Biases, 185 Science 1124 (1974). Tversky and Kahneman asked subjects to estimate the percentage of African countries in the United Nations. The authors employed a wheel numbered 1 to 100, which was rigged to land on 10 for half the subjects and 65 for the other half. After the spin, the experiments asked subjects whether their estimates were higher or lower than the result of the spin. Then they asked subjects to estimate the percentage. Subjects for whom the wheel landed on 10 gave an average estimate of 25 percent, while subjects for whom the wheel landed on 65 gave an average estimate of 45 percent.


356


party’s role as plaintiff or defendant can cause ‘‘self-serving’’ choices of capped judgment heuristics.48 In a series of studies, Messick and Sentis presented each subject with hypothetical scenarios in which they had completed a project with another student. In some scenarios, the subject worked for 10 hours and their partner worked for 7 hours, while in other scenarios, subjects were told the opposite. Subjects told they had worked 10 hours preferred payment in proportion to hours worked, while subjects informed they had worked fewer hours than their partner opted for equal payments.49 The Messick and Sentis work suggests that when individuals can choose among multiple, credible judgment criteria, they select the criterion yielding the greatest personal benefit.50 In the Messick and Sentis experiments, the competing criteria are proportionality and equality. In the above posed hypothetical, truncation and recalibration make up one of several pairs of decision methods. When a cap is binding, plaintiffs may disproportionately truncate, allowing the estimate to remain as high as possible. Truncation best comports with plaintiffs’ presumed desire to achieve the highest possible trial outcome. Conversely, defendants may disproportionately employ a criterion, such as recalibration or anchoring and adjustment, that yields a lower trial prediction than would truncation. If, as hypothesized here, plaintiffs and defendants account for the existence of a cap in motivationally biased ways, a cap can widen the disparity in trial outcomes, reducing the likelihood of a contract zone. Continuing with the hypothetical case, absent the cap, both parties predict a trial result of $500,000, ensuring a contract zone. When a cap is imposed, the plaintiff truncates the trial prediction to $250,000, while the defendant recalibrates the trial prediction to $125,000, thus creating a large judgmental disparity and reducing the likelihood of a contract zone. 48 Motivational biases arise from a desire to reach a specific conclusion or hold a particular belief. Such biases are believed to operate subconsciously. See David J. Schneider, Albert H. Hastorf, & Phoebe C. Ellsworth, Person Perception 105 (2d ed. 1979); and Ziva Kunda, The Case for Motivated Reasoning, 108 Psychol. Bull. 480 (1990). 49 David Messick & Keith Sentis, Fairness and Preference, 15 J. Experimental Soc. Psychol. 418 (1979). 50 See also Alvin Roth & Keith Murnighan, The Role of Information in Bargaining: An Experimental Study, 50 Econometrica 1123 (1982). In their experiment, two subjects divide 100 lottery tickets. One subject is to receive $5 and the other $20 for winning the lottery. The allocation of prize money suggests two, alternative methods for splitting the tickets: an 80/20 split (it equalizes expected payoffs) and a 50/50 split. When subjects were unaware of who had which prize, only 12 percent of pairs could not agree on an allocation of tickets between them. When subjects knew one another’s payoffs, 22 percent of pairs could not agree. The authors attribute the increased impasse rate to different views about splitting the tickets. In the complete information condition, the $5 subject likely considered an 80/20 split fair (because it equalized expected payoffs), while the $20 subject likely viewed an equal split as fair.


DAMAGE CAPS

357

Incorporating Fairness into Bargaining Positions. We have assumed thus far that litigants base settlement decisions exclusively on monetary considerations.51 However, there exists a large body of evidence demonstrating that fairness considerations enter into a wide range of economic decisions.52 In this section, we extend our basic framework to allow perceptions of fairness to influence judgment formation during pretrial bargaining. We now represent the litigants’ reservation values as follows: RVp ⫽ f [α p Fp , β p Tp ] ⫺ C p

(7)

RVd ⫽ g[α d Fd , β d Td ] ⫹ C d ,

(8)

and where Fp and Fd are the litigants’ assessments of a fair outcome in the case and α and β are the weights afforded the fairness assessment and predicted trial outcome, respectively. Under the simple model in Section IA, α p and α d are zero, while β p and β d are one. Each litigant must determine how much to weight the fairness assessment relative to the predicted trial outcome. The Messick and Sentis work on motivational biases suggests litigants will place greater weight on the dimension most compatible with their underlying monetary objective. For example, suppose the plaintiff estimates an adjudicated outcome of $200,000 but 51 See, for example, Priest & Klein, supra note 35, at 4 (‘‘According to our model the determinants of settlement and litigation are solely economic, including the expected costs to parties of favorable or adverse decisions, the information that parties possess about the likelihood of success at trial, and the direct costs of litigation and settlement’’); Posner, supra note 35, at 555 (‘‘Each party’s best settlement offer will depend on how he expects to fare in the litigation’’). 52 The dictator and ultimatum game studies from the early 1980s help illustrate the role of fairness in economic decision making. In the ultimatum game, a subject (the offeror) is given some money, for example, $10, and instructed to divide it between herself and another (the offeree). The offeree can either accept the proposed division or reject it, in which case both subjects receive nothing. Under the economic assumption of wealth maximization, the offeree should accept any positive offer, since the alternative is to receive nothing. Yet in these experiments, offers below 25 percent are typically rejected while the modal offer is often 50 percent. See Werner Guth, Rolf Schmittberger, & Bernd Schwarze, An Experimental Analysis of Ultimatum Bargaining, 3 J. Econ. Behav. & Org. 367 (1982); and Daniel Kahneman, Jack L. Knetsch, & Richard H. Thaler, Fairness and the Assumptions of Economics, 59 J. Bus. S285 (1986); Colin Camerer & Richard H. Thaler, Ultimatums, Dictators and Manners, 9 J. Econ. Persp. 209 (1995). The dictator game was developed to respond to the criticism that, fairness aside, offers were generous so they would be accepted. In the dictator game, where offerees must accept whatever is given, 76 percent of offerors nevertheless offered an even split. See Kahneman, Knetsch, & Thaler, Fairness and the Assumptions of Economics, supra. Survey evidence further shows that individuals are willing to make substantial economic sacrifices to punish unfair decisions. See Daniel Kahneman, Jack L. Knetsch, & Richard H. Thaler, Fairness as a Constraint on Profit Seeking: Entitlements in the Market, 76 Am. Econ. Rev. 728 (1986).


358


believes that $300,000 is fair.53 Since the ‘‘fair’’ outcome better comports with the plaintiff ’s preference for a large settlement, the plaintiff will heavily weight the fairness assessment when developing a reservation value. Conversely, defendants, who prefer a lower trial outcome, will place more weight on the dimension that is smaller. Suppose the defendant estimates an adjudicated outcome of $150,000 but believes that $100,000 is fair. Since the assessment of fairness is more consistent with the defendant’s preference for a small settlement, the defendant will more heavily weight this dimension when determining a reservation value. A damage cap may also affect the parties’ reservation values by altering α and β. We hypothesize that a cap will change α and β in predictable ways related to a subject’s role as plaintiff or defendant. If the predicted trial outcome or fairness assessment becomes more favorable once a cap is imposed, that dimension should receive increased weight under a capped system. For defendants, who prefer lower settlements, the predicted trial outcome should become relatively more favorable (compared with their fairness assessment) under a cap. While the trial award is directly constrained by a damage cap, the fairness assessment may emanate from a number of sources (for example, beliefs about justice or emotions), none of which pertain to the liability limit. Thus, defendants litigating in a capped environment should place greater weight on the predicted trial outcome than defendants in an uncapped environment. As a result, defendants’ reservation values will be lower than they would have been had their weights not changed with the introduction of a cap.54 Analogously, dimensions less favorable under a cap should receive less weight. Since plaintiffs prefer higher settlements, the predicted trial outcome should become relatively less favorable than the assessment of fairness when a cap is imposed. In a capped system, plaintiffs should decrease the weight afforded the predicted trial outcome and increase the weight

53 The ‘‘hostile media’’ effect can cause persons to consider impartial third parties biased. Robert Vallone, Lee Ross, & Mark Lepper, The Hostile Media Phenomenon: Biased Perception and Perceptions of Media Bias in Coverage of the Beirut Massacre, 49 J. Personality & Soc. Psychol. 577, 584 (1985), finds that individuals ‘‘view media coverage of controversial events as unfairly biased and hostile to the position they advocate.’’ In this context, plaintiffs will likely view the trial outcome as less than fair, while defendants should view the outcome as more than fair. 54 The following example is illustrative. Suppose that, absent a cap, defendants determine their reservation values as follows: RVd ⫽ .6T ⫹ .4F ⫹ C, where T is the predicted trial outcome, F is the fairness assessment, and C denotes costs. If a cap affects the weighting process in the self-serving manner described above, defendants should increase the weight placed on T and decrease the weighting of F after a cap is imposed. Since the cap directly constrains T but not F, defendants will reduce their reservation values by more than they would have had the weights not changed.


DAMAGE CAPS

359

placed on the assessment of fairness. This reduces the downward revision of reservation values that should have occurred had plaintiffs not altered their weighting. The process described above can reduce the probability that a contract zone exists. This occurs because the defendant’s reservation value falls by a great deal (because the expected trial outcome falls and the defendant increases the weight placed on the trial outcome) and the plaintiff ’s reservation value falls by less (even though the expected trial outcome falls, the weight on the trial outcome falls as well). This increases the likely disparity in reservation values, reducing the likelihood of settlement. We next propose a bargaining experiment to investigate the effect of a cap on settlement. We use this experiment to test the predictions of both the simple and enriched models and to investigate the importance of the ‘‘behavioral’’ modifications for explaining subject behavior. II. Experimental Procedure A. The Experiment One hundred and sixty-six MBA students from the University of Chicago and 54 masters students in public policy from Carnegie Mellon University participated in this study. There were three phases of data collection: first, at the University of Chicago (Chicago I); next, at Carnegie Mellon (CMU); and again at Chicago (Chicago II). We presented subjects with materials outlining a personal injury lawsuit in which the only unresolved claim was the amount of damages the plaintiff should receive for pain and suffering. Subjects were randomly assigned the role of plaintiff or defendant and then conducted pretrial settlement discussions. We imposed a judge’s award on all pairs that did not reach voluntary settlement. There were two experimental conditions, one with no liability limit (‘‘uncapped’’) and the other with a $250,000 damage cap (‘‘capped’’). About 1 week before the negotiation, we randomly distributed case materials and a cover sheet to each subject. The cover sheet indicated whether the subject was designated plaintiff or defendant, gave a brief overview of the lawsuit, and outlined the negotiating procedure in detail. The cover sheet also informed plaintiffs and defendants randomly assigned to the capped condition about the cap.55 That part of the instructions read, ‘‘Recently, Congress has passed legislation limiting the amount of damages a judge or jury can award to plaintiffs for physical pain and suffering and mental anguish to $250,000. That law applies in this case.’’ 55

Subjects assigned to the uncapped condition received no information about a cap.


360


The case materials contained about four pages of narrative and 14 pages of excerpts from the deposition testimony of several witnesses.56 The witnesses included the plaintiff, Bradley Madsen, and defendant, Ed Clayton, president of Clayton Fencing and Roofing. In the case, the plaintiff fell through a street vent manufactured by the defendant. Each party’s legal position had potential strengths and weaknesses.57 For instance, the plaintiff was under the influence of marijuana at the time of the accident, and there was some suggestion that gang members previously loosened bolts on street vents as part of an initiation ritual. On the other hand, the vent covering was manufactured and installed by Clayton only 3 years before the accident as one of its first jobs since expanding the company to include ventcovering services. About 1 year before the incident, someone informed Clayton about a loose bolt on the grate in question. Some Clayton workers inspected the grate and decided not to replace it. Other possible defendants were eliminated from consideration, as was the issue of insurance. The sole defendant, Clayton, owned the property on which the accident occurred and had manufactured and installed the grating. The plaintiff had been fully compensated through his employer’s healthinsurance plan for all medical expenses and lost wages. Such compensation was expected to continue into the future if the need arose. Thus, all issues had been resolved except the amount, if any, the plaintiff should recover from the defendant for pain and suffering and mental anguish. For this, the plaintiff was seeking $1,000,000.58

56 We informed plaintiff and defendant subjects that their case information was identical to that of their counterpart. 57 We sought to prevent straightforward predictions about the outcome for several reasons. First, many actual lawsuits involve considerable ambiguity about the likely result. Second, we required sufficient variation in the rate of settlement, our dependent measure, in order to produce meaningful empirical results. Third, self-serving beliefs are most likely to arise when there is ambiguity or uncertainty in the decision environment (Babcock & Loewenstein, supra note 43). Such ambiguity was essential to a fair test for our psychological claims. 58 Admittedly, plaintiffs do not always plead a specific amount of damages, or if they do, it is a mere formality. However, in pretesting the case materials, subjects, who in general were neither exposed to legal training nor personally involved in lawsuits, had difficulty predicting a specific monetary outcome with no guidance. We included the $1,000,000 feature to eliminate this dissonance. However, both conditions had to receive the ‘‘suit for $1,000,000’’ language to obviate any confound between the presence of a cap and the absence of the $1,000,000 clause. Therefore, uncapped subjects were informed that the plaintiff was seeking $1,000,000. Capped subjects were informed there had been a debate surrounding the damage-cap legislation. When the suit was filed, the plaintiff ’s attorney advised the plaintiff to sue for $1,000,000 in case the damage cap was repealed. Subjects were then informed that the law would not be repealed and that the $250,000 damage cap would apply to their case.


DAMAGE CAPS

361

B. Procedure We randomly paired ‘‘plaintiffs’’ with ‘‘defendants’’ to conduct a pretrial negotiation. The negotiation consisted of four bargaining periods, each lasting 5 minutes. Following each period, the subjects privately wrote down and exchanged offers of settlement. If the amount the defendant offered exceeded the amount the plaintiff demanded, the case was settled at the midpoint of the two amounts. If not, the parties continued to negotiate. If they reached the end of the fourth period without settling, the judge’s award was imposed. Subjects were informed that we had given the case materials to an actual trial judge who had rendered a decision.59 Both litigants were assessed legal fees of $10,000 to enter the next negotiation period. If subjects settled in the first period, they paid no fees. The maximum legal fee for either side was therefore $30,000. Subjects were graded by their monetary outcomes relative to other similarly situated subjects.60 For example, if an uncapped subject pair settled for $400,000 after 15 minutes, the defendant would have to pay $420,000 (adding $20,000 in legal fees), and the plaintiff would receive $380,000 (subtracting $20,000 in legal fees). Uncapped defendant subjects were graded according to how much they paid relative to other uncapped defendants. Lower amounts meant higher grades. Uncapped plaintiff subjects were graded according to how much they obtained relative to other uncapped plaintiffs. For plaintiffs, the higher the monetary outcome, the higher the grade. Uncapped pairs who failed to settle received the judge’s award of $770,000 to the plaintiff. For these pairs, the plaintiff ’s final outcome was $740,000 (accounting for $30,000 in legal fees), and the defendant’s was $800,000. We employed the same procedure for capped subjects, except we truncated the judge’s award at $250,000 in the event of nonsettlement.61 Before negotiating, subjects provided answers to three questions. First, we asked them to estimate the monetary outcome were the case to be tried. Any subject whose estimate was within $25,000 of the actual award ($770,000 in the uncapped condition or $250,000 in the capped condition) received three lottery tickets. Second, we asked that subjects provide their 59 Unknown to the subjects, the award was $770,000, which was truncated to $250,000 for capped litigants who went to trial. 60 All subjects were students in a negotiation class in which performance on weekly classroom negotiations represented a portion of their semester grade. They were regularly exposed to this method of calculating payoffs and grades. 61 The classroom discussion following the negotiation produced some revealing information about the manner of belief formation. Some capped subjects objected to our procedure for determining the judge’s award in the capped condition, which was simply to truncate the $770,00 judgment at $250,000. They believed that had the judge been aware of a cap when ruling, the award would have been less than $250,000.


362

THE JOURNAL OF LEGAL STUDIES TABLE 1 Differences in Dependent Measures between Experimental Conditions Condition Uncapped Settlement rate Average negotiating time (for pairs that settled) Average legal fees paid Average settlement amount (for pairs that settled) Total number of pairs Number of pairs that settled

Capped

p-Value

.57 (.07)

.73 (.06)

.083

14.70 (.95) $47,778 (2,610)

14.50 (.82) $43,928 (2,700)

.900

$490,000 (27,650) 54 31

$182,000 (13,500) 56 41

.310 .000

Note.—SEs are given in parentheses; p-values represent the significance level for the differences in means between the conditions.

assessment of a fair outcome. Capped litigants were instructed that their fairness estimate need not be bound by the $250,000 liability limit. Third, we elicited subjects’ reservation values. We asked plaintiffs to assume that the defendant proposed a ‘‘take it or leave it’’ offer at the outset and that this offer would not improve before trial. We provided hypothetical offers increasing in $25,000 increments from $0 to $1,000,000. For each offer, the plaintiff indicated whether they would accept the offer or go to trial. The lowest offer the plaintiff agreed to accept rather than go to trial provided the subject’s reservation value. We followed a precisely analogous procedure to elicit defendants’ reservation values. Our design allows us to compare settlement rates, time to settlement, and legal fees assessed between the uncapped and capped experimental conditions. We will also analyze the subjects’ estimates of the trial outcome, their assessments of a fair outcome, and their reservation values in order to assess why settlement rates might vary across the two conditions and test the importance of the behavioral modifications to the basic model in Section I. III. Results There were no significant differences among the three subject populations (Chicago I, CMU, Chicago II) across any of the key variables, so we aggregated the data. Table 1 provides some summary measures. In the capped condition, 73 percent of subject pairs settled, compared with 57 per-


DAMAGE CAPS

363

cent in the uncapped condition.62 The mean settlement amount was $182,098 for capped pairs and $490,129 for uncapped pairs. For pairs that settled, negotiating time was nearly identical across conditions. However, we must account for the greater proportion of uncapped pairs that negotiated the full 20 minutes without settling. We use a censored regression in which the dependent variable is time to settlement, with pairs that did not settle censored at 20 minutes. The coefficient on condition (a dummy variable that equals one for capped pairs) is negative but not statistically significant (p ⫽ .16). Uncapped pairs incurred an average of $47,778 in legal fees, compared with $43,929 in the capped condition ( p ⫽ .31). Thus, while the sign of the regression coefficient and difference in mean legal fees suggest that caps reduced the depletion of negotiating resources, the absence of statistically significant findings prevents a firm conclusion on this point. One of our primary goals was to examine whether caps affect the settlement rate. Having discovered some evidence that caps encourage settlement, we now investigate why. Uncertainty about the Trial Outcome. The simple model outlined in Section IA suggests a cap will increase the settlement rate by reducing uncertainty about the predicted trial outcome. Such uncertainty reduction increases the probability that, for a given negotiating pair, aggregate litigation costs exceed any disparity in predicted trial outcomes. This increases the likelihood that a contract zone will exist and improves prospects for settlement. The results provide some support for this account of behavior. First, the variance in predicted trial outcomes is markedly lower for capped subjects ($66,688, compared with $240,904 for uncapped subjects).63 Second, there is a higher percentage of contract zones in the capped condition (43 percent) than in the uncapped condition (26 percent).64 Finally, there is a statistically significant, positive relationship between the existence of a contract zone and the likelihood of settlement.65 We next investigate several psychological refinements to the simple framework outlined in Section IA. Self-Serving Precap Judgments. Our results clearly comport with previous findings that litigants hold self-serving beliefs about the trial outcome. In the uncapped condition, the average predicted trial outcome for plaintiffs was $562,222, compared with $400,611 for defendants. For capped plaintiffs, the average predicted trial outcome was $199,420, while for capped 62 63 64 65

p ⫽ .083 for chi-square test. F(107, 111) ⫽ 13.05, p ⫽ .000. t(108) ⫽ 1.88, p ⫽.06. χ 2 (1) ⫽ 4.67, p ⫽ .03.


364

THE JOURNAL OF LEGAL STUDIES TABLE 2 Logit Regressions (Dependent Variable Is 1 If Settled, 0 Otherwise) (1) Condition (1 if cap, 0 otherwise)

.7070 (.4084)

Difference in predicted trial awards (Tp ⫺ Td ) Log L

(2) ⫹

69.4

.1889 (.4675) ⫺.0041** (.0012) 61.9

Note.—SEs are given in parentheses. The sample size is 110 pairs for each regression. Both models include a constant term. Log L(0) ⫽ 70.9. ⫹ Statistical significance at the .10 level. ** Statistical significance at the .01 level.

defendants it was $151,982. All statistical tests for differences in means and differences in underlying distributions across roles are significant at the .001 level. Although subjects in both experimental conditions exhibited self-serving biases, the cap clearly reduced their magnitude. In each condition, the difference in mean predicted trial outcomes (Tp ⫺ Td ) indicates the magnitude of the bias. For uncapped pairs, plaintiffs’ trial predictions exceeded defendants’ trial predictions by an average of $161,611. The cap reduced this difference to $47,438.66 Finally, we find the reduction of the bias mediated the effect of the cap on settlement rates. Table 2 presents two logit regressions, each with the dependent variable equal to one for pairs that settled and zero otherwise. The first regression includes one explanatory variable, which equals one for capped pairs and zero for uncapped pairs. The estimated coefficient for this variable is positive (p ⫽ .083), confirming the marginally significant difference in settlement rates between conditions. In column 2, we add an independent variable measuring the extent of self-serving biases present in each negotiating pair (Tp ⫺ Td ). The coefficient for this difference is negative and significant ( p ⫽ .001), indicating the inverse relationship between the magnitude of the bias and the probability of settlement. Further, in regression (2), the p-value for the condition coefficient increases to .686, eliminating the effect of the cap. This suggests that one of the primary ways through which caps affect settlement is by diminishing the parties’ self-serving biases. Using the parameter estimates from column 2, when the bias is reduced from its mean level to 1 standard 66

t ⫽ 2.77, p ⫽ .006 (for difference in means); p ⫽ .067 (for difference in distributions).


DAMAGE CAPS

365

Figure 3.—Distributions of predicted trial outcomes

deviation below its mean, the probability of settlement increases from .67 to .84. Role-Specific Postcap Judgments. Here we examine how the cap affected subjects’ predictions of the trial outcome. Figure 3 shows six histograms of the parties’ trial predictions by role and experimental condition.


366


The first column contains distributions of defendants’ predictions, and the second column displays plaintiffs’ predictions. The first row compares the uncapped trial predictions of defendants and plaintiffs. For these subjects, 85 percent of plaintiffs and 67 percent of defendants predicted a trial outcome of $250,000 (the cap amount) or higher. If capped subjects truncated otherwise high trial predictions at the cap amount, the distributions of capped trial predictions should resemble those simulated in the second row of Figure 3. In those distributions, 85 percent of plaintiffs and 67 percent of defendants predict a trial outcome of exactly $250,000. The last row contains the actual capped distributions where, instead, only 25 percent of capped plaintiffs and 14 percent of capped defendants actually predicted a $250,000 trial outcome.67 The distributions of trial predictions by condition suggest that subjects accounted for the cap in motivationally biased ways. Our experimental design does not disclose the specific processes governing belief formation under a cap—that analysis would require more in-depth data, such as that produced by the elicitation of verbal protocols. The resulting histograms do suggest that, whatever the processes are, they appear different for plaintiffs and defendants. For capped defendants, beliefs are more evenly distributed throughout the entire $0–$250,000 range than are the beliefs of capped plaintiffs (relative to what is predicted from the uncapped beliefs). Consistent with our earlier discussion, this suggests greater use of judgment processes such as recalibration by defendants and greater use of judgment processes such as anchoring and adjustment or truncation by plaintiffs. Role-Specific Biases in Incorporating Fairness into Bargaining Positions. We find strong evidence that the subjects incorporate assessments of fairness into their bargaining positions and some evidence that they do so in a self-serving manner. Table 3 presents four regressions: one each for uncapped plaintiffs, capped plaintiffs, uncapped defendants, and capped defendants. In these regressions the dependent variable is the subject’s reservation value, while the independent variables are the subject’s predicted trial outcome and their assessment of a fair outcome.68 In three of four regressions, the fairness coefficient is significantly greater than zero. Thus, even controlling for the trial prediction, there is a positive and statistically significant association between fairness assessments and reservation values. 67 The percentage of plaintiffs that have beliefs of $250,000 or greater is statistically different across conditions (t(108) ⫽ 5.16, p ⫽ .000). Also, the percentage of defendants that have beliefs of $250,000 or greater is statistically different across conditions (t(108) ⫽ 8.67, p ⫽ .000). 68 In the regression, we also include a dummy variable indicating whether the subjects’ trial estimates are censored by the $250,000 cap (that is, are equal to $250,000). This is to avoid biased coefficients that would result from such censoring.


367

DAMAGE CAPS TABLE 3 Ordinary Least Squares Regressions (Dependent Variable Is Reservation Value) Plaintiffs Assessment of fairness Predicted trial outcome R2

Defendants

No Cap

Cap

No Cap

Cap

.19** (.06) .49** (.10) .61

.30** (.09) .25 (.47) .27

.27* (.12) .36** (.12) .59

.01 (.05) .60** (.19) .48

Note.—SEs are given in parentheses. * Statistical significance at the .05 level. ** Statistical significance at the .01 level.

We also find evidence for a motivational bias in the alteration of weights between conditions: subjects increase the weight placed on the more favorable dimension. The cap causes defendants’ predicted trial outcomes to decrease by 62 percent (from $400,611 to $151,982), but their fairness assessments drop by only 32 percent (from $307,685 to $210,223). Defendants should therefore decrease their weight on the fairness assessment when a cap is imposed. This in fact occurs: the coefficient on the fairness assessment is .27 in the uncapped regression (col. 3) and .01 in the capped regression (col. 4).69 The cap causes plaintiffs’ trial estimates fall by 65 percent (from $562,222 to $199,420), but their fairness assessments decrease by only 25 percent (from $701,435 to $526,518). Plaintiffs should therefore increase their weight on the fairness assessment when a cap is imposed. The coefficient on the fairness assessment is .19 (col. 1) in the uncapped regression and .30 (col. 2) in the capped regression, though it is not statistically higher.70 IV.

Discussion

In this study, we begin with a simple model of how rational agents achieve settlement. We then enrich the model by including role-specific biases in information processing and allowing nonpecuniary motivations to affect the litigants’ bargaining positions. We use the simple and enriched model to predict the effect of a damage cap on the likelihood that litigants 69 The difference in coefficients of fairness between columns 3 and 4 is statistically significant ( p ⫽ .031). 70 The difference in coefficients on assessment of fairness between columns 1 and 2 is not statistically significant ( p ⫽ .34).


368


are able to reach pretrial settlements. Finally, we report an experiment showing that, compared with the simple model, the behavioral enrichments more thoroughly depict the effects of a cap on the pretrial bargaining environment. We find that a damage cap has a positive and marginally statistically significant (p ⫽ .083) effect on the likelihood of avoiding trial. By recording subjects’ perceptions preceding the negotiation, we identify several mechanisms through which the cap affects the settlement rate. First, we find strong evidence that a cap reduces uncertainty about the trial outcome, increasing the chance there is a contract zone. This increases the likelihood that opposing subjects will find agreements that both sides prefer to trial. Second, the subjects exhibited self-serving biases, causing plaintiffs to predict higher trial outcomes than defendants. These biases impede the ability of subjects to reach voluntary settlements. Beyond the invariable uncertainty reduction that accompanies a cap, our regression results identify another mechanism through which caps affect the settlement rate—caps diminish the self-serving biases present in negotiating pairs. Third, we find that subjects incorporate the cap into their judgments in predictably different ways, depending on their role as plaintiff or defendant. Defendants more often employed ‘‘recalibration’’ or some similar process, presumably to generate lower trial predictions. Plaintiffs disproportionately ‘‘truncated’’ or ‘‘anchored and adjusted,’’ in order to generate higher trial predictions than would another decision method. This asymmetry in postcap judgments likely prevented further improvement in settlement rates. Fourth, fairness considerations play an important role in determining the pretrial bargaining environment. Controlling for the trial predictions, subjects’ fairness assessments independently predict their reservation values. There was some, albeit weaker, evidence that subjects’ roles in the litigation determined how important they considered the fairness assessment. Further, when a cap was imposed, subjects shifted weight to the dimension of reservation values most consistent with their objective in the litigation. As with asymmetric postcap judgments, this process can widen the disparity in opposing parties’ beliefs, producing less settlement. In this study, however, reduction in the uncertainty of verdict estimation and the impact of selfserving biases outweighed the countervailing effects of other role-specific biases. Thus, the net effect of the cap was more settlement. External validity is often a concern in experimental research. Contrary to real world litigation, our subjects were not represented by dispassionate and experienced counsel. The question arises whether the behavioral regularities reported here are as pervasive when attorneys participate in judgment formation. While we advocate the use of experts and professionals in experimental research, existing findings already suggest that even experienced


DAMAGE CAPS

369

counsel are susceptible to the cognitive and motivational biases identified here.71 Further, litigation decisions often embody the combined input of both a litigant and counsel. Thus, were such biases even relegated only to lay judgments, they would still influence decisions during litigation. Another concern involves the artificiality of the experimental environment. However, for our purposes, experiments offer several advantages over naturally occurring data. Strict control over the data-producing environment and the random assignment of subjects permit more credible causal attributions. With field data, internal validity is invariably a concern, since econometric problems such as endogeneity, sample-selection bias, and omitted variables can cause incorrect inferences about the effect of a policy change. Further, experiments are useful for uncovering underlying behavioral processes because we can elicit subjects’ beliefs, something not often possible with field data. Understanding the intervening processes through which reforms affect behavior helps researchers determine the robustness of behavioral outcomes to differences in the reform environment. In our experiments, the cap was extremely binding—the overwhelming majority of subjects in the uncapped condition predicted trial outcomes greater than the cap value. Under these conditions, the cap produced a higher settlement rate. Consider, however, how a nonbinding cap could affect settlement. Imagine a minor case in a jurisdiction with a $1,000,000 cap. Under the simple model of settlement with no behavioral modifications, a nonbinding cap should not affect settlement. However, our results suggest the cap affects the negotiating environment in more complex ways. We suggested earlier that damage caps could act as an ‘‘anchor’’ or focal point in constructing judgments. We also showed how individuals preferentially select judgmental criteria that best comport with their underlying objectives. In a small-stakes case, a nonbinding cap might lure plaintiffs’ predicted trial outcomes upward to a larger degree than for defendants, therefore decreasing prospects for settlement. Thus, the net effect of a cap might well depend on its placement along the distribution of claim severity. Evidence for the intervening mechanisms through which policy changes 71 See Margaret Neale & Gregory Northcraft, Expert, Amateurs and Refrigerators: Comparing Expert and Amateur Negotiators in a Novel Task, 38 Org. Behav. & Hum. Decision Processes 305 (1986) (finding that the behavior of professional negotiators is influenced by how a problem is ‘‘framed,’’ even when the situations are objectively identical); see also Barbara J. McNeil et al., On the Elicitation of Preferences for Alternative Therapies, 306 New Eng. J. Med. 1259 (1982) (finding that physicians are also susceptible to framing manipulations in making recommendation for treatment of lung cancer). Even seasoned negotiators in public school teacher contract negotiations are vulnerable to self-serving biases (Linda Babcock, Xianghong Wang, & George Loewenstein, Choosing the Wrong Pond: Social Comparisons in Negotiations That Reflect a Self-Serving Bias, 111 Q. J. Econ. 1 (1996)).


370


affect behavior can also help predict the effect of combinations of policies. We found that the cap improved settlement rates, at least in part, by reducing the impact of the parties’ self-serving biases. Recently, Babcock, Issacharoff, and Loewenstein demonstrated that having each litigant identify weaknesses in their own legal position successfully ‘‘debiases’’ them and produces higher settlement rates.72 Incorporating effective debiasing techniques into pretrial mediation procedures would eliminate a major source through which caps improve settlement rates. To the extent several policies possess similar objectives, understanding the intervening processes that lead to impasse can help identify a source of redundancy in policy making. We undertook here an interdisciplinary examination of the effects of damage caps on pretrial bargaining. Not surprisingly, we cannot supply a definitive answer. The effects of this policy intervention appear more complicated than a straightforward application of simple, rational choice framework suggests, with competing mechanisms operating in different directions. Barring a more thorough understanding of the behavioral effects of damage caps, policy makers should proceed cautiously when considering large-scale alteration of our current dispute resolution procedures. 72 See Linda Babcock, Samuel Issacharoff, & George Loewenstein, Creating Convergence: Debiasing Biased Litigants, 22 L. & Soc. Inquiry 401 (1997).
