Law and Economics Research Papers Series

Research Paper No. 00-01

September, 2000

Law and Behavioral Science: Removing the Rationality Assumption from Law and Economics
Russell B. Korobkin Thomas S. Ulen

California Law Review, Vol. 88, 2000

This paper can be downloaded without charge from the Social Science Research Network Electronic Paper Collection: http://papers.ssrn.com/paper.taf?abstract_id=229937

Electronic copy available at: http://ssrn.com/abstract=229937

California Law
VOL.88 JULY 2000

Review
No. 4

Copyright? 2000 by CaliforniaLaw Review, Inc.

Law and Behavioral Science: Removing the Rationality Assumption from Law and Economics
RussellB. Korobkint ThomasS. Ulent
Introduction............................................................................................. 1053
I. The Uses and Shortcomingsof RationalChoice Theory ................. 1060 A. Conceptionsof RationalChoice Theory . .................................. 1060 1. The DefinitionalVersion ....................................................1061 2. The ExpectedUtility Version ............................................ 1062 3. The Self-InterestVersion....................................................1064 4. The WealthMaximizationVersion.....................................1066

Copyright? 2000 Russell B. Korobkinand Thomas S. Ulen Associate Professor,Universityof Illinois College of Law and University of Illinois Institute t of Governmentand Public Affairs. Visiting Professor, UCLA School of Law (2000-01). J.D., B.A., StanfordUniversity. Alumni Distinguished Professor, University of Illinois College of Law and Professor, $ University of Illinois Institute of Governmentand Public Affairs. Ph.D., StanfordUniversity, M.A., OxfordUniversity,B.A., Dartmouth College. We would like to thank the faculty and staff of the Max Planck Institute for Research into Economic Systems in Jena, Germany,for providingsupportand a stimulatingenvironmentin which to complete an early draftof this work. We also thankRobertAshford, Stephen Bainbridge,Tom Cotter, Dan Farber, Christine Jolls, Owen Jones, Douglas Kysar, Richard Markovits, Eric Posner, and Jeff Rachlinski for their thoughtfulcomments on earlier drafts, along with participantsat the Society for Socio-Economics Annual Meeting Vienna, Austria, and participants in faculty workshops at the Copenhagen Business School, Miami University, and the law schools at Case Western Reserve University, UCLA, the University of Hamburg, the University of Texas, Trinity College Dublin, Katholieke Universiteit Leuven, the University of Ghent, the University of Ljubljana, and the University of Kentucky. We also thank Patrick Elder and Michael Hazy for their research and assistance.

1051 Electronic copy available at: http://ssrn.com/abstract=229937

1052

CALIFORNIA LAWREVIEW

[Vol.88:1051

B. Limitationsof RationalChoice Theoryin Legal Analysis ..................................................................................... 1066 1. The Inadequacyof Thin Versions of Rational Choice Theory.....................................................................1067 2. The Implausibilityof Thin and Thick Versions.................. 1069 C. The Responses of RationalChoice Theoryto Criticisms............... ............. ........... .......... ........1070 D. Modifying the BehavioralPredictionsof Rational Choice Theory ..........................................................................1074 II. BoundedRationalityand the Use of Heuristics............................... 1075 A. Decision-makingStrategiesThatDo Not Maximize ExpectedUtility........................................................................1076 1. Complexity..................................................................... 1077 2. Ambiguity ...........................................................................1083 B. Decision-makingHeuristicsand Biases . ................................... 1084 1. Availabilityand Representativeness ...................................1085 2. Overconfidenceand Self-ServingBiases............................ 1091 3. HindsightBias..................................................................... 1095 4. Anchoringand Adjustment ................................................. 1100 III. The Importance of Context ..............................................................1102 A. ReferencePoints and the FramingEffect . ................................. 1104 B. The EndowmentEffect and the StatusQuo Bias....................... 1107 C. Habits,Traditions,Addictions,and Cravings . ........................... 1113 D. Time Inconsistenciesand the Multiple-SelvesProblem ............1119 E. Sunk Costs................................................ 1124 IV. Deviations from Self-Interest .......................................................... 1126 A. Social Norms .............................................................................1127 B. Fairness......................................................................................1135 C. Collective Action.......................................................................1138 Conclusion .......................................................................................... 1143

Electronic copy available at: http://ssrn.com/abstract=229937

2000]

LAWAND BEHAVIORAL SCIENCE

1053

Law and Behavioral Science: Removing the Rationality Assumption from Law and Economics
Russell B. Korobkin Thomas S. Ulen

As law and economics turnsforty years old, its continued vitality is threatened by its unrealistic core behavioral assumption: that people subject to the law act rationally.Professors Korobkinand Ulen argue that law and economics can reinvigorate itself by replacing the rationality assumption with a more nuanced understandingof human behavior that draws on cognitive psychology, sociology, and other behavioral sciences, thus creating a new scholarly paradigm called "law and behavioral science." This article provides an early blueprint for research in this paradigm. The authorsfirst explain the various ways the rationalityassumption is used in legal scholarship and why it leads to unsatisfyingpolicy prescriptions. Theythen systematicallyexamine the empirical evidence inconsistent with the rationality assumption and, drawing on a wide range of substantiveareas of law, explain how normativepolicy conclusions of law and economics will change and improve under the law-and-behavioralscience approach.
INTRODUCTION

The law-and-economicsmovement has sufferedfrom the truthfulness of one of its most importantpostulates: the law of diminishing marginal returns.Although law and economics was once viewed as a revolutionary approachto legal scholarshipthat appliedthe principlesof microeconomic price theory to the analysis of legal rules, the value of its new insights is gradually diminishing.1The movement's vast initial successes were so
1. This is a theme in recent stocktakingby the profession. See, e.g., Douglas G. Bairdet al., The Future of Law and Economics: Looking Forward, 64 U. CHI. L. REV. 1129 (1997); Richard A. Epstein, Law and Economics: Its Glorious Past and Cloudy Future,64 U. CHI.L. REV. 1167 (1997). A decade ago, Robert Ellickson observed that "[T]he first generation of law and economics scholars has essentially accomplishedthe straightforward applicationsof the basic economic model in virtually every legal field. Currentscholarshipis more technical and interstitial."RobertC. Ellickson, Bringing Culture and Human Frailty to Rational Actors: A Critique of Classical Law and Economics, 65 CHI.-KENT L. REV.23, 24 (1989).

1054

CALIFORNIA LAWREVIEW

[Vol. 88:1051

sweeping that the currentpliers of the tradehave been forced to searchfor more narrowniches to fill. As a result, the discipline often seems to be devolving into a subdiscipline of applied economics that happens to focus What began as a form of legal analysis that substantivelyon legal matters.2 a economics as tool is now too often economic analysis that uses employed law as a target. Mathematicalelegance often becomes the primarygoal,3 with usefulness in the realm of law, that combines logic with humanexperience,4a mere afterthought.5 The seminal insight that economics provides to the analysis of law is thatpeople respondto incentives-a generalizedstatementof price theory.6 From this insight, two importantcorollariesfollow. First, the law can serve as a powerful tool to encourage socially desirableconduct and discourage undesirableconduct. In the hands of skillful policymakers,the law can be used to subsidize some behaviors and to tax others. Second, the law has efficiency consequencesas well as distributiveconsequences.Intentionally or unintentionally,legal rules can encourage or discourage the production of social resourcesand the efficient allocationof those resources.Although efficiency need not be the sole or primarygoal of legal policy,7economic analysis of law teaches that policymakers ignore the efficiency implications of theiractions at society's peril. Legal rightsthat are unobjectionable in the abstractare not free but rathermust be measured against their opportunitycosts.8
2. This is not to say that there are not some ongoing, exciting research projects in law and economics. One such projectis the study of the legal and economic aspects of social norms, a literature that we investigate in Part IV.A. below. Another is the emerging body of literatureon game theory, a branch of microeconomics largely neglected during the first two decades of the law-and-economics movement that has only recently been broughtto bear on legal analysis.See, e.g., DOUGLAS G. BAIRD ET AL., GAME THEORYAND THE LAW xi (1994) (arguing that game theory explains how laws affect behavior); Ian Ayres, ThreeApproaches to Modeling Corporate Games: Some Observations,60 U. CIN.L. REV.419 (1991); Eric Talley, InterdisciplinaryGap Filling: Game Theoryand the Law, 22 L. & Soc. INQUIRY 1055 (1997). 3. This criticism is often made of microeconomics research conducted in economics departmentsas well. See, e.g., Mark Blaug, The Disease of Formalism in the Economics, or Bad Games That Economists Play 8, Jena Lectures, ISSN-Nr. 0947-1561 (1998) (calling "the worship of the idol of... mathematical rigor"the "'originalsin' in economic methodology"). 4. See, e.g., OLIVER WENDELL THECOMMON LAW 5 (MarkDeWolfe Howe ed., 1963) HOLMES, ("Thelife of the law has not been logic: it has been experience."). 5. Cf. Ellickson, supra note 1, at 32-33 (calling law and economics a "technicalsideshow" in which scholarswith only a modest amountof technicaltraininghave difficulty makingcontributions). 6. See, e.g., STEVENE. LANDSBURG, The ARMCHAIR ECONOMIST: ECONOMICS AND EVERYDAY LIFE 1-9 (1993) (explaining that the heart of economics is that "[p]eople respond to incentives" and using the example that safer cars may induce people to drive more carelessly). 7. Cf Russell B. Korobkin & Thomas S. Ulen, Efficiency and Equity: What Can Be Gained L. REV.329 (1998) (suggesting ways that legal rules from CombiningCoase and Rawls?, 73 WASH. can take accountof efficiency and equity concernssimultaneously). 8. One need not conclude that law and economics leads to a wholesale overthrow of settled legal learning. Sometimes it does, as, for example, when it suggests that parties to contracts should have greater latitude to stipulate damages than the law currently allows. See ROBERT COOTER & THOMAS LAWANDECONOMICS 235-37 (3d ed. 2000). But just as frequently,law and economics ULEN,

2000]

SCIENCE LAWAND BEHAVIORAL

1055

lies its andtherein Lawandeconomics is, at root,a behavioral theory, truepower.The concernof law and economicswith how actorsin and to legal directives to the legal systemrespond (andwouldrespond subject to hypothesizedchangesin those directives)now permeatesthe mainof scholarship farbeyondtheboundaries of legalacademic stream thought, Thiscontradition. of the law-and-economics thatis self-consciously part but movement cernwasby no means"invented" by thelaw-and-economics and it is so it. universalized Indeed, widelyacknowledged acby certainly in the modernlegal academythat bears that it mentioning hardly cepted behavit hasrealeffectson private law does not exist in a vacuum; rather, for whenexamandaccounted ior, andthoseeffectsshouldbe considered legalregimes. iningalternative of viewinglaw as a seof the legal implications To speakcoherently aboutthe conseries of incentives,analystshave to make assumptions to the to the incentives of those legal system.To people subject quences from ecoscholarsimported satisfy this need, early law-and-economics abouthow peoplerespondto incentives, nomicsa series of assumptions debate Thereis considerable choicetheory." as "rational knowngenerally about communities andlaw-and-economics withinboththe economics preit is not. As and is is choice rational what implicitly applied theory cisely it is understood or explicitlyin the law-and-economics however, literature, that or as a relativelyweak, "thin," presumption individuals alternatively act to maximizetheirexpectedutility,howeverthey define this, or as a act to maximize thatindividuals presumption strong,or "thick," relatively theirself-interest.9 thebestseriesof whatwas,no doubt, choicetheory Rational provided of pricetheory the to to which begin develop application assumptions upon to legal rules. The use of rationalchoice theoryenabledthe law-andin advances in its earlydays,to achievesignificant economicsmovement, that But now rules and between the interaction society. legal understanding the rationality intellectual hasreached the movement assumption maturity, Thereis simplytoo scholarlydevelopment. severelylimits its continued act in evidence that individualsfrequently much credibleexperimental theof rational choice with the that are assumptions incompatible ways of rules effects It follows that the analysisof the incentive legal ory.10

provides an alternative justification for prevailing legal rules and institutions,as, for example, when it concludes that there is a deep economic logic to the basic structureof tort liability. See id. at 300-28; LAW THEECONOMIC STRUCTURE OFTORT M. LANDES & RICHARD A. POSNER, see also WILLIAM LAW(1987). ECONOMIC ANALYSIS OFACCIDENT SHAVELL, (1987); STEVEN 9. We discuss alternativedefinitionsof rationalchoice theoryin PartI.A, infra. 10. A good deal of the evidence is discussed in PartsIII-V, infra.

1056

LAWREVIEW CALIFORNIA

[Vol. 88:1051

based on such implausiblebehavioralassumptionscannotpossibly result in 1 efficacious legal policy, at least not in all circumstances. After the law-and-economics movement had stretched its academic legs, it might well have concerned itself with developing a more complex and realistic theory (or theories) of behavior, one (or ones) that, perhaps, would be less amenable to formal modeling but more relevant to creating legal policy.12But the movement has not yet fully begun to do this, alThe longer though we sense that it is beginning to move in this direction.13 that the field delays in elaborating this richer theory of behavior-the longer, that is, that it fails to take the "law"part of "law and economics" seriously-the more it engages in a deep irony: although the movement has become more deeply entrenchedin the legal academy'4and continues to gain adherentsand scholarlypractitioners of its arts,15 it has become less relevant to the making of legal policy, one of the ultimate ends of legal scholarship.16 As the rate of increase in the breadth and usefulness of law-andeconomics scholarship has declined, the shortcomings of rational choice theory have become more apparent.Behavioralanomalies and puzzles that rationalchoice theory (at least relatively strongversions) cannot explainonce little noticed because of the considerableutility rationalchoice theory did have-began to appear more significant as the economic analysis of law gained influence within the legal community.In the wake of increasing
11. See Jon Elster, WhenRationalityFails, in THELIMITS OFRATIONALITY 19 (KarenSchweers Cook & Margaret Levi eds., 1990). 12. Law and economics was not universallywelcomed among legal scholars or practitionersas a liberatorfrom a darknight of legal studies. Rather,a common criticism of law and economics was that its assumptionof rationalbehavior by legal decision makers was preposterous.See generally Thomas S. Ulen, Rational Choice and the Economic Analysis of Law, 19 L. & Soc. INQUIRY 487, 488 (1994) (noting widespread criticism of the law-and-economics movement based on the rational actor assumption). 13. See, e.g., BEHAVIORAL LAWANDECONOMICS (Cass R. Sunstein, ed., CambridgeUniversity Press 2000); ChristineJolls et al., A Behavioral Approach to Law and Economics, 50 STAN.L. REV. 1471 (1998); Cass Sunstein, Behavioral Law and Economics: A Progress Report, 1 AM. L. & ECON. REV.115 (1999). 14. The two primarylaw-and-economicsjournals, the Journal of Legal Studies and the Journal of Law and Economics, are among the most prestigious and widely cited among law professors. See Colleen M. Cullen & S. Randall Kalberg, Chicago-KentLaw Review Faculty Scholarship Survey,70 CHI.-KENT L. REV. 1445, 1453 (1995) (rankingthe Journal of Legal Studies tenth and the Journal of Law and Economics fifteenthby numberof citationsin otherlaw reviews). 15. The most recent issue of the Association of American Law Schools Directory lists 164 law OFLAW professors as members of its section on law and economics. See THE AALS DIRECTORY TEACHERS 1998-99, at 1175-76 (1998). No doubt this number far understates the number of law professorswho conduct scholarlyresearchwithin the law-and-economicsparadigm. 16. This may be due to the fact that some commentators perceive a widening gap between the bar and the legal academy and speculate that the rise of law and economics is at least partiallyresponsible for this gap. See, e.g., ANTHONY T. KRONMAN,THELOST LAWYER: FAILING IDEALS OFTHE LEGAL PROFESSION 225-40 (1993); HarryT. Edwards,The GrowingDisjunctionbetweenLegal Educationand the Legal Profession, 91 MICH. L. REV.34 (1992).

2000]

LAW AND BEHAVIORAL SCIENCE

1057

questions about the sanctity of the rational choice assumptions, the proponentsof rationalchoice theory retrenched,as defenders of criticized and developed more sophisticatedways to paperover paradigmsoften do,17 its empirical shortcomingsand to denounce its critics as overly concerned of human with minor details not truly importantto a generalunderstanding behavior or to the critical analysis of law. 1 As a result, its proponentsargue that rationalchoice theory is a still-viable descriptionof human decision making.19 Most, though by no means all, of the leading members of the law-and-economics movement in the legal academy are among these
defenders.20

In response to these trends, a new movement is emerging in the legal academy that builds on the core insights of law-and-economicsscholarship but takes seriously the shortcomings of rational choice theory.21This movement, which we call "law and behavioral science," lacks a single, coherenttheory of behavior.Although such a general theory may someday develop and would be welcome, the movement's currentlack of concern aboutthis shortcomingidentifies law and behavioralscience as a species of As we argue in this Article, one can analyze the legal pragmatism.22
17. See THOMAS THESTRUCTURE OFSCIENTIFIC REVOLUTIONS 77 (3d ed. 1996). KUHN, 18. In economics there have always been those critical of the assumptionof rationalconsumers, and other economic actors.The principalresponse has been thatprovidedby the suppliers,bureaucrats, Nobel laureate Milton Friedman. Friedman argued that in any positive analysis the reality of assumption was secondary to the primaryimportanceof having falsifiable predictions. See MILTON ESSAYS IN POSITIVE ECONOMICS FRIEDMAN, (1953). It might be argued, for example, that while the behavior of some individuals may deviate from that predicted by rational choice theory, those deviations are symmetricallydistributed(and, by implication,without a great deal of variance)around the average behavior posited by that theory. Professor Gary Becker has also suggested that, even if a significant number of decision makers are irrationaland the deviations are not always symmetrical, microeconomictheory. See Gary aggregatemarketbehaviorwill still follow the predictionsof standard 1 (1962). ECON. Becker, IrrationalBehaviorand EconomicTheory,70 J. POL. 19. This defense suggests that one should take the criticisms of rationalchoice theory seriously only to the extent that they describe systematicdeviations. This is a fair point. As we show below, the experimental evidence establishes that the deviations are, indeed, systematic and not randomly distributedarounda (rationalactor)mean. 20. See, e.g., RichardA. Posner,Rational Choice, Behavioral Economics,and the Law, 50 STAN. L. REV.1551 (1998) (commentingon Jolls et al., supranote 13, and arguingthat behavioraleconomics has no theory, is antitheoretical, and that rational choice is both descriptively and normatively superior). 21. As Donald Langevoort observes, elements of the relevant behavioral research have been employed in legal scholarshipfrom time to time for a numberof years, although the attemptto bring togetherthe various insights from this researchinto a coherentmovementis a very recentdevelopment. See Donald C. Langevoort, Behavioral Theories of Judgment and Decision Making in Legal Scholarship: A Literature Review, 51 VAND.L. REV. 1499, 1502 (1998). For examples of recent efforts to addressthe limitationsof rationalchoice theory, see sources cited supra note 13. 22. See Thomas F. Cotter, Legal Pragmatism and the Law and Economics Movement 84 GEO. L.J. 2071, 2072 (1996) (describinglegal pragmatistsas rejectingthe notion that legal doctrinemust be based on a grand theory). Cotter argues that law and economics itself is not necessarily inconsistent with pragmatism,but he concedes that much of the work of law-and-economicsscholarsis inconsistent with pragmatismsince it relies on strictfoundationalassumptionsabouthumannature.See id.

1058

CALIFORNIA LAWREVIEW

[Vol. 88:1051

appropriatelegal command in any given circumstance without a grand, overarching theory of behavior so long as one has a due regard for the relevant decision-makingcapabilities of the actors in that specific setting. By borrowingfrom psychological and socioculturaltheories in additionto economics, the law-and-behavioral-science approachconsciously chooses to emphasizeits externalusefulness in analyzinglegal problemsratherthan either its internal elegance or universal applicability. Its ultimate goal, the incentive effects of law betterthan modquite simply, is to understand ern law and economics is able to do by enlisting more sophisticatedunderstandings of both the ends of those governed by law and the means by which they attemptto achieve theirends. Applying behavioral models more nuanced and sophisticated than rational choice theory to legal rules and systems will require a broader range of academic forms than is traditionallyfound in legal scholarship,in additionto a broadertheoreticalbase. In the early stages of the movement, legal scholars have been able, by and large, to make importantstrides by hypothesizingthat empiricaland experimentalfindings publishedby social science researchersapply to actors subject to legal commands.23 To progress beyond the currentinitial stage of scholarship,legal scholarswill have to conduct more empirical and experimental work of their own to test whetherthese hypotheses are in fact true in the particularized settings they study. To use one example, findings by cognitive psychologists that student researchsubjects make exchanges in a certainway seems a fair place to begin an inquiryinto the incentive effects of commerciallaw but an insufficient foundation on which to base a proposal for amending the Uniform Commercial Code. Before such legal reform proposals will be taken seriously outside the academy, legal scholars will have to develop tangible evidence that commercialactors in commercial settings are likely to respondto incentives in the same way as do studentsubjects.24
23. For a good overview of the range of legal literaturethat applies, to some degree, to social science research on behavior that is inconsistent with rational choice theory, see Langevoort, supra note 21, at 1506-19. 24. Our call for a new scholarshipin law based on behavioralscience does not precisely follow the traditionalpatternof emendationin the sciences, but there are parallels.A typical process by which the demand for a new theory arises is that significant anomalies appear in the application of the reigning paradigm,as Thomas Kuhn calls it, to empirical phenomena.When those anomalies become numerous, Kuhn hypothesizes that the scholarly community recognizes the desirability of a new paradigmthat encompassesboth the phenomenaexplainedby the old paradigmand the anomalies. See KUHN, supra note 17, at 77. The patternthat gives rise to a demand for behavioral science in law is more complicated. First, law and economics has not yet reached the stage in which empirical studies have uncovered so many legal anomalies that there is a clearly felt need for a paradigmto replace rationalchoice theory. But second, those working in other social sciences have found many empiricalanomalies in the predictions of rational choice theory. In a notable column in the Journal of Economic Perspectives called "Anomalies,"RichardThaler and colleagues collected examples of these results. See, e.g., Robyn M. Dawes & Richard H. Thaler, Anomalies: Cooperation, 2 J. ECON.PERSP. 187 (1988); Kenneth A.

2000]

AND BEHAVIORAL SCIENCE LAW

1059

This Article provides an early blueprint for research in "law and which we hope will help guide the emerging scholarbehavioralscience,'"5 in area. In Part I, we describe in more detail the applications and ship this shortcomingsof rational choice theory as a behavioralexplanationuseful to legal analysis. Parts II through IV survey a range of research findings from the behavioralsciences that collectively articulatea more subtle and of humanbehavior realistic (althoughconsiderablyless tidy) understanding in rationalchoice theory, and each partpresentsexamthan that articulated ples of how these findings are useful to the analysis of legal policy. PartII of rationalchoice theory contends that, contraryto the usual understanding persons subjectto the legal system are by law-and-economicspractitioners, seldom ruthlessoptimizersof theirutility;rather,they often rely on a range of decision-makingshortcutsand heuristics.PartIII argues that the preferences of persons subject to law are not exogenous to the context in which the decision makers find themselves but are situationallydependent.Part IV contends that persons subject to the legal system routinely make specific choices that do not optimally serve their immediate self-interest.We conclude with our vision of how this Article might serve as a basis for a substantialresearchagendafor the next generationof scholarsinterestedin the confluence of law and behavioralscience.

Froot & Richard H. Thaler, Anomalies: Foreign Exchange, 4 J. ECON.PERSP. 179 (1990); Daniel Kahnemanet al., Anomalies: The EndowmentEffect, Loss Aversion, and Status Quo Bias, 5 J. ECON. PERSP. 193 (1991); CharlesM. C. Lee et al., Anomalies: Closed-EndMutual Funds,4 J. ECON. PERSP. 153 (1990); George Loewenstein & RichardH. Thaler,Anomalies: IntertemporalChoice, 3 J. ECON. PERSP.181 (1989); RichardH. Thaler,Anomalies: Interindustry PERSP. WageDifferentials, 3 J. ECON. 181 (1989); Richard H. Thaler, Anomalies: Saving, Fungibility, and Mental Accounts, 4 J. ECON. PERSP. 193 (1990); Richard H. Thaler, Anomalies: Seasonal Movements in Security Prices II: Weekend,Holiday, Turnof the Month,and IntradayEffects, 1 J. ECON. PERSP. 169 (1987); Richard H. Thaler, Anomalies: The UltimatumGame, 2 J. ECON. PERSP. 195 (1988) [hereinafterThaler, The UltimatumGame];RichardH. Thaler,Anomalies: The Winner'sCurse,2 J. ECON. PERSP. 191 (1988); Richard H. Thaler & William T. Ziemba, Anomalies: ParimutuelBetting Markets: Racetracksand 161 (1988); Amos Tversky & RichardH. Thaler, Anomalies: Preference PERSP. Lotteries, 2J. ECON. PERSP. 201 (1990). We posit in this Article that these anomalies in other social Reversals, 4 J. ECON. science fields would have been found in empirical legal studies if such studies were to occur. (Some have, and we reporton them below.) Moreover,as we have stressedand will stress again, we are not in a position to develop a new paradigm to replace rational choice theory. Our aim is simply to the wide-rangingexperimentalresults from other social sciences into law and economics so incorporate as to refine the connection between predicted human behavior and the attainmentof the goals of the legal system. 25. For other examples, see sources cited supra note 13; see also Langevoort, supra note 21; Cass R. Sunstein,BehavioralAnalysis of Law, 64 U. CHI.L. REV.1175 (1997).

1060

CALIFORNIA LAW REVIEW

[Vol.88:1051

I
THE USES AND SHORTCOMINGS OF RATIONAL CHOICE THEORY

Rational choicetheory is theheartof modemmicroeconomic theory.26 It is sucha powerful, andusefulconstruct that straightforward, compelling, scholarsin a wide rangeof disciplinescontiguous to economics,such as relations,sociology, financeand politicalscience, history,international of have rational choice theoryas and, course,law, accountancy, adopted theircentralaccountof humandecisionmaking.27 for the Unfortunately of there is no definition purposes preciseanalysis, single,widelyaccepted of rational choice theory.28 Although the use of the that actors behave rationallyis pervasiveamong law-andassumption economicsscholars,the assumption is most often implicit.As a result, thereis rarelya discussion in the legal literature aboutwhat,exactly,constitutesrational In actuality, behavior. thereareprobably nearlyas many differentconceptions of rational choice theoryas thereare scholarswho implicitly employit in theirwork. The varietyof conceptions of rational choicetheorymakescritiquing the theorysomething akinto shooting at a movingtarget. We will simplify the task somewhat that versions of the by positing theorycan be aligned a and fourspecificpointsalongthe specalong spectrum thenaddressing trumrepresenting the most commonconceptionsof the theory,as it is in legalscholarship. employed
A. Conceptionsof Rational Choice Theory

Thedifferent of rational choicetheorycanbe understood conceptions as pointsalonga continuum of how specificandprecisethe predictions of the theoryare.On the left side of the spectrum are "thin" of conceptions rationalchoice theory-that is, conceptionsin which the theoryis relaandin whichit is relativelyeasy for the behavior of tively undemanding actorsto be consistent with the theory.On the rightside of the spectrum are "thick"conceptionsof the theory-that is, conceptionswith more robustbehavioral thataremoreeasily falsifiableby empirical predictions
26. See, e.g., RICHARD A. POSNER, ECONOMIC ANALYSIS OFLAW3 (5th ed. 1998) ("Thetask of economics ... is to explore the implicationsof assumingthatman is a rationalmaximizerof his ends in life."). 27. Of course, rationalchoice theory has not gained a dominantmarketsharein these disciplines withouta fight. The level of objectionto rationalchoice theorycan be seen as a measureof the theory's success in the academy. Cf DONALD P. GREEN & IANSHAPIRO, PATHOLOGIES OFRATIONAL CHOICE choice theoryin political science). 28. Nonetheless, as we shall see there is a core set of understandingsof what rational choice theory is. See Thomas S. Ulen, Rational Choice Theoryin Law and Economics, in ENCYCLOPEDIA OF LAWAND ECONOMICS & SHAPIRO, (BoudwijnBockaert& GerritDe Geest eds., 1999); see also GREEN supra note 27, at 13.
THEORY: A CRITIQUE OF APPLICATIONS IN POLITICAL SCIENCE (1994) (attacking the use of rational

2000]

LAW ANDBEHAVIORAL SCIENCE

1061

evidence.29 Figure 1 illustrates the construct, with the dominant conceptions of the theoryin bold:
RATIONAL CHOICE THEORY SPECTRUM

"Thin" Conceptions

"Thick" Conceptions

<-I
Definitional Version
FIGURE 1

Maximizing ExpectedUtility Self-Interest Wealth Version Version Version

1. TheDefinitional Version In its thinnestversion, rationalchoice theory is definitional in nature. This conception postulates only that, as RichardPosner has written, "man is a rational maximizer of his ends,"30 without providing any predictions what an ends individual regarding might attempt to maximize or what means he might employ in such an effort. On this account, rationality is understood as suiting means to ends, but no normative theory of either means or ends is assumed.3' One can conceive of a situation in which the prediction of this thin theory (that economic decision makers choose means well suited to achieving theirends) would not hold-namely, when an individualchooses an action contraryto his goals. But in practice the definitional version of rational choice theory is nonfalsifiable, because both the means and the ends of behavior are defined by observing the behavior itself. Put another way, rationalbehavior is understoodas "the way people act." Because we assume a priori that people "rationally maximize their ends" (or their "utility"),we characterizeas rationalany behaviorthat we see and any behavior that can be justified as "rational"merely by noting its existence. Even behavior that seems anomalous, such as clucking like a chicken
29. The thin/thickdichotomy is borrowedfrom GREEN & SHAPIRO, supra note 27, at 17-18; see also John Ferejohn, Rationalityand Interpretation: ParliamentaryElections in Early StuartEngland, in THE ECONOMIC APPROACH TO POLITICS:A CRITICAL REASSESSMENT OF THETHEORY OF RATIONAL ACTION (KristenRenwick Monroe ed., 1991). 30. RichardA. Posner, Are We One Self or Multiple Selves?: Implicationsfor Law and Public
23, 24 (1997). Policy, 3 LEGALTHEORY

31. The distinction between rational means and rational ends deserves far more attention than economists have given it. Moreover, economists should also pay particularattentionto rationalbelief. The connectionbetween action and belief is straightforward: presumablyindividualsact in accordwith their beliefs, and individuals seek to have rational beliefs, those that are consistent and held for
64-106 (1993). Many other supportable reasons. See ROBERTNOZICK,THE NATUREOF RATIONALITY

commentatorshave remarkedon the thin descriptionof human motives contained in the economist's (thin) definition of rationality. For a representative view, see Martha C. Nussbaum, Flawed Foundations: The Philosophical Critique of (a Particular Type of) Economics, 64 U. CHI. L. REV. 1197 (1997).

1062

LAW REVIEW CALIFORNIA

[Vol.88:1051

is presumed to be well suitedto achieving theends whenever a bell sounds, of theclucker. confirms the rationality of beOn this thinunderstanding, everything refutesit. As Arthur Leff notedin an earlycritique of havior,andnothing this use of rational choice theorysubthe law-and-economics movement, for a normative or empirical stitutes a definition proposition.32
2. The Expected Utility Version

The next conceptionof rationalchoice theory, and the one most in modemmicroeconomics,33 is often termed"expected dominant utility in the This like is "thin" the definitional version, theory."34 conception, or goals decisionmakers sense thatit does not specify whatpreferences in will pursue.35 As before,the particular or tastescontained preferences the actors'utilityfunctions areexogenous,thatis, given fromoutsidethe Theexpected maximization problem. utilityversionis, however,"thicker" thanthe definitional versionbecauseit does specifythe means(or at least some of the means)by which actorswill seek to satisfytheirgoals and preferences. The assumption thateconomicdecisionmakingcan be shownto be of expected to constraints, is the resultof the maximization utility,subject a thicker modelof human in thatit positsa moreformalmodelof behavior individual decisionmaking thandoes the versionof rational choicetheory in the previousSection.Consider considered expectedutilitytheory'saccount of decisionmakingunderuncertainty. Supposethat an individual must choose betweena certainand an uncertain course of action-for
32. See ArthurAllen Leff, EconomicAnalysis of Law: Some RealismAboutNominalism,60 VA. L. REV.451, 458 (1974) (reviewing the first edition of RICHARD A. POSNER, OF ECONOMIC ANALYSIS LAW (1973)). Leff's strong doubts about the usefulness of law and economics as an approachto legal scholarship seemed largely based on his understandingthat the movement rested on a tautological definition of rational behavior. See id. at 482 (criticizing Posner for "substitutingdefinitions for both facts and values");id at 478-79 ("In such a system whateveris, is. If you do not 'buy' something,you are unwillingto do so."). 33. See GREEN & SHAPIRO, supra note 27, at 18. 34. See, e.g., Geoffrey Brennan,Comment,WhatMight RationalityFail to Do, in THELIMITS OF RATIONALITY 51, 52 (Karen Schweers Cook & Margaret Levi eds., 1990) (calling the "standard" version of rationalchoice theory "no more than the assumptionthat the agent is utility-maximizing"). The precise parametersof the expected utility theory version of rational choice theory, like rational choice theory itself, resist consensus. See, e.g., SCOTT THEPSYCHOLOGY OFJUDGMENT AND PLOUS, DECISIONMAKING83 (1993) (noting that "expected utility theory is actually a family of theories"). Economists sometimes refer to subjective expected utility theory or SEU. The addition of the word "subjective"merely allows for the probabilities by which the decision maker weighs the utilities of uncertainoutcomes to be subjective,ratherthan objective. Whetherthose probabilitiesare objective or subjective,they must obey the probabilitycalculus-for example, the sum of the probabilitiesof all the possible outcomes must sum to one. 35. See Elster,supra note 11, at 20 (statingthat rationalchoice theory "tells us what we ought to do in orderto achieve our aims as well as possible. It does not, in the standard version, tell us what our aims ought to be").

2000]

LAW ANDBEHAVIORAL SCIENCE

1063

andanuncertain return on an investment betweena certain return example, on a more speculativeinvestment.Suppose,further,that outcome01 butthatoutcomes occurswithcertainty 02,..., Onareprobabilistic: only one of them will eventuate,but they are all possible. How would the thatthe actorwill decidewhichinvestment rational choicetheorist predict attacha utilityto eachpossibleoutto choose?The actorwill presumably of each outand so forth, U(02), come-U(O1), alongwith a probability the valueof the certain come'soccurring- p1,P2, andso on. Calculating to occur, Because01 is certain courseof actionis straightforward. pl = 1, so thatthe expectedutilityto the actorof 01 equalsU(O1).The uncertain to evaluate becausethe uncertain investment moreof a challenge presents
outcomes are mutually exclusive, P2 + .. ing: + Pn = 1. The decision maker + PnU(On).

can reducethe multiplepossibilitiesto a single expectedutilityby solvEU(uncertainaction) = P2U(02) +...

The rationalconsumer then compares the expectedutilityof the certain courseof actionwiththe expected courseof action, utilityof the uncertain andselectstheone withthehigher value.36 the basic requirement of adornments, Strippedof its mathematical conductan explicitor imexpectedutilitytheoryis thatdecisionmakers plicit cost-benefitanalysisof competingoptionsand select the optimal method of achieving their goals (that is, the method that maximizes expected benefits and minimizes expected costs, or maximizes net to external constraints.37 benefits), expected subject If an actormakes a decisionthat does not maximizenet expected benefitsto him,thenhe violatesthe behavioral of theexpected predictions version of rational choice But because it is utility theory. impossibleto know what choices are optimalfor a particular decisionmakerwithout of his utilityfunction, andbecauseutilityfunctions knowingthe contours aredifficultto elicit,38 thebehavioral of expected predictions utilitytheory
36. An importantelement of expected utility theory is the decision maker's attitudetowardrisk. There are three possible attitudes: risk-aversion,risk-neutrality, and risk-preferring. If a person is riskaverse, then he or she prefers a certaintyof, say, $100 to an uncertainprospect of $100 (for example, the compoundpossibility of receiving either $100,000 with probabilityof 0.001 or $0 with probability then he is indifferentbetween a certaintyof $100 and an uncertainprospect 0.999). If he is risk-neutral, of $100. If he is risk-preferring (also termedrisk-seeking),he prefersan uncertainprospectof $100 to a & ULEN, certaintyof $100. For a more formaldiscussion, see COOTER supra note 8, at 46-49. 37. Jon Elster conceives of three"optimality conditions"of rationalbehavior: the action must be the best way for the actor to satisfy his preferences given his beliefs, the beliefs must be the best he could form given his information,and the amountof informationcollected must be optimal given the strengthof his preferences.Elster,supranote 11, at 21. 38. It is possible to elicit utility functions. For example, to test whetherconsumersmade product choices that maximizedtheir utility, one groupof experimentersasked experimentalsubjectsa series of questions about their preferencesfor differentproductattributesbefore giving them a choice problem. Whether consumers made utility-maximizing decisions could then be determined by examining whether choices made conformed with the subjects' prior statements about their preferences. See

1064

CALIFORNIA LAWREVIEW

[Vol.88:1051

often are not directly verifiable or falsifiable. Fortunately,there are some necessary (but not sufficient) conditions of rational behavior under the expected utility model that analysts can observe. These are usually understood to include the following: actors should be able to comparethe utility (1) Commensurability: consequencesof all alternativesto each other; (2) Transitivity:if an actorpreferschoice A to choice B and choice B to choice C, he shouldthen preferchoice A to choice C; (3) Invariance: the preferencebetween two or more choices should not depend on how the choice is presentedor structured, so long as the outcome possibilities are constant; (4) Cancellation: a choice between options should not depend on featuresof the options that are identical;and (5) Dominance: an actor should never choose an option in which every featureis only as good as the featuresof a competing option, and at least one featureis not as good.39 If an actor fails to follow one or more of these principles, he cannot be makingdecisions consistent with the expected utility model. Consequently, the predictionsof the model are testable,at least at some minimumlevel. 3. The Self-InterestVersion Still thicker versions of rational choice theory start from expected utility theory's predictions about the mannerin which actors will attempt to achieve their utility, and add predictions about the actors' goals and preferences-that is, about the content of the actor's utility function. Perhaps the most common assumption about ends, that actors will seek to maximize what is in their self-interest, can be traced to Adam Smith's famous statement: It is not from the benevolence of the butcher, the brewer, or the bakerthat we expect our dinner,but from their regardto their own interest. We address ourselves, not to their humanity but to their self-love, and never talk to them of our necessities but of their
advantages.40

The implicationis that if we can figure out what course of action will most profit the decision maker, we will be able to predict his course of action. This is an advance over the thin conceptions of rational choice in that it suggests falsifiable predictions about substantive behaviors, not just predictions aboutdecision-makingprocedures.

Naresh K. Malhotra,Information Load and ConsumerDecision Making, 8 J. CONSUMER RES. 419, 42227 (1982). 39. See PLOUS, supra note 34, at 81-82.
40.
ADAM SMITH, AN INQUIRY INTO THE NATURE AND CAUSES OF THE WEALTH OF NATIONS 15

(JamesE. ThoroldRogers ed., London,ClarendonPress 1869).

2000]

AND BEHAVIORAL LAW SCIENCE

1065

Such thickversionsof rational choice theorydominate the law-andeconomicsliterature,41 the is almost although assumption alwaysimplicit rather thanexplicit.Consider, for example,the simpleprediction thatif thereis no punishment for littering, peoplewill litter(the cost to an individualof disposing of his litterin a lawfulmanner exceedsthe cost to him of observing his individual litteron the ground).42 Thisprediction implicon relies the that individuals are concerned with punishitly assumption mentstheymightreceive,thedisutility thattheywill sufferfromlookingat theirown litter,andthetimeandenergyit takesto disposeof litter,butnot withthe disutility otherswill sufferfromlookingat theirlitter.Orconsider the prediction that if punitivedamageswere to be abolishedor capped, moredefectiveproducts wouldbe produced.3 Thisprediction relieson the that manufacturers are concerned with theirown botassumption product tom line, andare concerned withthe healthandsafetyof theircustomers affectthatbottomline.44 only to the extentthatthoseissuesdemonstrably Unlikethe thinversionsof rational choicetheory,the self-interested versioncanleadto thecreation of directly falsifiable behavioral predictions. defenders of rational choice theorymightat Suspectingan ambush, thispointobservethatit is plausible to hypothesize thata decisionmaker's "self-interest" of the well-being of othmightbe servedby takingaccount ers, notjust his or herown wantsanddesires.45 Unfortunately, expanding the conception of "self-interest" to includeother-regarding in preferences addition to selfishones wouldrobthe notionof "self-interest" of all of its value.46 We wouldno longerbe ableto predict thatpeoplewould predictive litter if doing so risks no punishment, or that productswould be more
41. Cf. Jennifer Arlen, Comment: The Future of Behavioral Economic Analysis of Law, 51 VAND.L. REV.1765, 1766 (1998) (observing that "[c]onventionallaw and economics assumes ... that people are self-interested... ."); Jeffrey L. Harrison,Egoism, Altruism, and Market Illusions: The Limits of Law and Economics, 33 UCLA L. REV. 1309, 1320 (1986) ("My impression is that narrow self-interest... is the behavioral assumption most commonly employed by those applying economic analysis to law."). 42. Cf. POSNER, supra note 26, at 416-17 (claiming that fines for litteringwould have to be very heavy to be a significantdeterrent given the high cost of apprehending litterers). 43. See COOTER & ULEN, supra note 8, at 352-53. 44. There is a great deal of debate about these mattersin the law-and-economicsliterature.See, e.g., A. Mitchell Polinsky & Steven Shavell, PunitiveDamages: An EconomicAnalysis, 111 HARV. L. REV. 870 (1998) (arguing for the use of punitive damages as a method of creating efficient incentives when victims fail to bring actions against their injurers);W. Kip Viscusi, The precautionary Social Costs of Punitive Damages Against Corporationsin Environmentaland Safety Torts, 87 GEO. L.J. 285 (1998) (arguingthat punitive damages ought not to be availablebecause they fail to deter and are subject to grave error in assessment, with inefficient consequences); Theodore Eisenberg, Measuring the Deterrent Effect of Punitive Damages, 85 GEO.L.J. 347 (1998) (criticizing Viscusi's statisticaltests of the effects of punitivedamages). 45. See POSNER, supra note 26, at 4 ("self-interestshould not be confused with selfishness; the happiness(or for thatmatterthe misery) of otherpeople may be partof one's satisfactions"). 46. See also Jolls et al., supra note 13, at 1488-89 (arguing that such a flexible notion of rationalityrobs the theory "of any real predictivepower").

1066

LAW CALIFORNIA REVIEW

[Vol.88:1051

withoutproducts dangerous liabilitylaw, becausepeople walkingin the parkmay or may not gain utility from keeping the parktidy for their andmanufacturers mayor maynot gainutilityfrompreventing neighbors, AdamSmith'sprediction wouldhave to be reharmto theircustomers. written to read: the brewer, or the It is fromeitherthe benevolence of the butcher, to theirown bakerthatwe expectourdinner,or fromtheirregard We address to theirhumanity and/ortheirselfinterest. ourselves, love, and talk to them of our necessities and/or of their advantages.47 Thepointis not thatwe believeactorsseekto fulfillonly selfishprefbelow.48 Butif "self-interest" erences-in fact,we arguequitetheopposite for the decision thatproducessatisfaction is definedto includeanything choicetheoryis no differthenthe self-interest versionof rational maker, entfromthethinexpected version.49 utility
4. The WealthMaximizationVersion

Thethickest of rational choicetheoryprovide even more conceptions the ends of decision makers than does about the selfspecificpredictions interestversion.The most commonof these very thick conceptionsis that actorswill attempt to maxi"wealthmaximization": the prediction mize theirfinancial or situation. all law-andwell-being monetary Nearly economicsliterature on businessorganizations, followingthe neoclassical is builton the explicitor implicitassumption economictheoryof firms,50 thatfirms seek to maximizeprofits.51 And muchlaw-and-economics liton individual erature behaviormakesan analogousassumption (usually at leastin circumstances in whichmoneyis at stake.52 implicitly),
B. Limitationsof Rational Choice Theoryin Legal Analysis

None of the conceptions of rational choicetheorydiscussedaboveis how actorswill respondto the optimalfor the purposeof understanding
47. Comparewith supra text accompanyingnote 40. 48. See infra PartIV. 49. Cf. Harrison, supra note 41, at 1311 (observing that broad definitions of self-interest are impossible to disprove). 50. See GREEN & SHAPIRO, supra note 27, at 18. 51. See, e.g., COOTER & ULEN, AN INTRODUCTION POLINSKY, supra note 8, at 26; A. MITCHELL TOLAWAND ECONOMICS 10 (2d ed. 1989). 52. The term "wealth maximization" is used as a normative principle of policymaking, and criticized as such. See, e.g., Gregory S. Crespi, Does the Chicago School Need to Expand Its 149 (1997) (arguing that wealth maximization is an incoherent Curriculum?,22 L. & Soc. INQUIRY and incomplete theory of law); Thomas S. Ulen, Professor Crespi on Chicago, 22 L. & Soc. INQUIRY 191 (1997) (arguing that despite its problems there is a sound practical reason for counting only market-mediatedpreferences in choosing among alternativepolicies). Here we use the term only to referto a positive theoryof individualor firm behavior.

2000]

AND BEHAVIORAL LAW SCIENCE

1067

incentivesthatthe law creates.The deficienciesof the different conceptheirinadequacy in pretionsarisefromone or bothof two shortcomings: of theirpredictions. ThisPart andtheimplausibility behavior dictingfuture versions of the affectthefourprimary will explainhow theseshortcomings The the Article will introduced above. remainder of modifitheory suggest to developan to rational choicetheory cationsandenhancements necessary in the formuof that is more useful human decision making understanding lationof legalpolicy.
1. TheInadequacyof Thin Versionsof Rational Choice Theory

Thinconceptions of rational choicetheorycan serveas a usefultool for socialscientists to behavior attempting explain, classify,or labelhuman froman ex post perspective. that actors will seek to maximize By positing theirutility,scholarscan observebehavior andthenreasonbackwards to of preferences an understanding Thinver(ends)andstrategies (means).53 sionsof rational thuscanbe viewedas usefulforrecognizing choicetheory thatpeopleusuallyactthe waytheydo for a reasonand,consequently, that observableactions provideclues for understanding these unobservable "reasons." theseadvantages areof littleuse to the makers of leHowever, who needto chooseamongcompeting gal policy, legalrules.Policymakers need to predictfuturebehavior undervariouslegal scenarios, not merely understand actions in past hindsight. The inadequacy of rational choicetheoryis mostobviousin its thinat all. nest, definitionalversion,which offers no behavioral predictions as an example, the parable of Buridan's Consider, ass, who faceda choice betweentwo equidistant as a sourceof nourishment.54 The dehaystacks finitionalversionof rational choicetheorycouldofferno prediction as to whichhaystack the ass wouldchoose,or whether he wouldchooseneither. Becausethis versionof the theoryinterprets all acts as rational, it would for the ass to standstill anddie of conclude,ex post, thatit was rational starvation.55 Becausehe believedthis versionof rational choice theoryto be the behavioral the law-and-economics movement, premiseunderlying Leff disparagingly referredto law and economics as "American legal
nominalism."56

Theexpected choicetheory, it rises utilityversionof rational although abovemeretautology, is scarcely moreadequate for generating behavioral
53. See, e.g., LANDSBURG, supra note 6, at 3. In a famous formulation,Paul Samuelson spoke of "revealedpreferences"-revealed, that is, by actual marketchoices and arguedthat not all the axioms of rational choice theory were necessary for the law of demand. See PAULANTHONY SAMUELSON,
FOUNDATIONS OF ECONOMIC ANALYSIS90-113 (1947).

54. 55. 56.

See, e.g., MargaretLevi et. al., Introduction: The Limits of Rationality, in THELIMITSOF See id. Leff, supra note 32, at 459.

RATIONALITY 1, 7 (KarenSchweers Cook & Margaret Levi eds., 1990).

1068

LAWREVIEW CALIFORNIA

[Vol.88:1051

predictions ex ante. In virtually any imaginable situationof importanceto predictions; legal policy, expected utility theory alone yields indeterminate not only does the theory fail to yield a single, unique behavioralprediction (the result that would be of most use to a policymaker),but it also fails to eliminate many conceivable possible actions.57This version of rational choice theory, like the definitionalversion, is helpless in predictingthe action of Buridan's ass, because it provides no theory or account of the content of the ass's utility function.If we knew, for example, that survivalwas the ass's sole preference, then expected utility theory would lead to the predictionthat the ass would choose one of the two haystacks.On the other hand, if we knew that the ass placed a higher value on avoiding difficult decisions than on remaining alive, expected utility theory would predict that the ass would stand still and perish. But without some theory aboutthe content of the ass's preferences,either standingstill or choosing a haystack could be rational.5 Expected utility theory can yield only the prediction that the ass will not violate one of the observable necessary conditions of expected utility theory. For example, it would lead to a predictionthat the ass will not announcethat he cannot comparethe value of choosing a haystack with the value of standing still and dying, because to do so would violate the principleof commensurability.59 To use an example of direct relevance to legal policy, consider the following question: if society wishes to reduce crime, assuming all other policy decisions are held constant,should it increase or decrease the length of prison sentences?Note that, althoughthe formerappearsto be the obvious answer, it is only correct if we assume that most people prefer to live outside of prison than to be incarcerated.But recall that thin versions of rationalchoice theory cannot aid us in creating this assumption.Virtually no predictionscan be made about decisions or behaviors without thickening the conception of rational choice theory to include some predictions aboutthe contentof preferences.60
57. Cf. Elster, supra note 11, at 24 (noting that the implications of a theory should be not only but also unique). determinate, 58. See Brennan,supra note 34, at 53 (noting that underthe "standard version"of rationalchoice theory, in which there are no substantiverestrictionson the actor's preferences,"it is at least arguable that there is no action that [rationalchoice theory] rules out. That is, for virtually any action, there exists some purposefor which the action is best"). 59. See supra Part I.A.2 for a list of the observable behavioral predictions associated with expected utility theory, includingcommensurability. 60. Cf. Brennan, supra note 34, at 53 ("Some restrictionon the agent's ends... is requiredto give rationalityany predictivebite."). In a famous attemptto amend rationalchoice theory precisely to avoid the problems to which we have referred, George Stigler and Gary Becker posited that a more productive assumption for economics was not that tastes differed substantiallyamong people but that they were relatively the same. On this assumption,differences in behavior arise because the relative prices, incomes, and other constraints faced by decision makers are different. See George J. Stigler & Gary S. Becker, De GustibusNon Est Disputandum,67 AM.ECON. REV.76, 76 (1977).

2000]

ANDBEHAVIORAL LAW SCIENCE

1069

2. TheImplausibilityof Thinand ThickVersions Thin conceptions of rational choice theory, such as expected utility theory, are not only inadequatelyspecified to be valuable to policymakers concernedwith the incentive effects of law, but the behavioralpredictions that they do make are also implausibleas predictionsof general applicability: that is, they have been demonstrated,as an empirical matter, to be incorrect,at least undersome conditions. substantially Thick conceptions of rational choice theory substantially avoid the inadequacyproblem. They can be used to generatepredictionsabout how actors will respond to alternativelegal regimes and, thus, can serve as a useful tool for policymakers.A thick conception of rationalchoice theory, for example, could enable us to predictthat Buridan'sass would value life sufficiently that it would be utility-maximizingto choose one of the haystacks, even if that meant suffering the pain of making a difficult choice. Thick conceptions, however, tradethe inadequacyproblemof thin conceptions for even more pronounced implausibility problems than those that plague thin conceptions. The predictions that they make about the preferences of actors (in additionto predictionsabout the means actors will use to satisfy their preferences) are demonstrablyincorrect, at least in many circumstances.Buridan's ass, of course, did not choose either of the haythus falsifying the predictionthat would be producedby any thick stacks,61 conceptionof rationalchoice theory. Thin and thick conceptions of rationalchoice theory share two types of implausibility problems. First, actors often fail to maximize their expected utility, but instead make suboptimalchoices among competing options given a set of preferences and use a range of heuristics-rules of thumb62-ratherthan complex cost-benefit analysis. This "boundedrationality" results from the high cost of processing information,the cognitive limitations of human beings, or a combination of the two. Evidence of these phenomenaare detailedin PartII below. Second, while all but the thinnest versions of rational choice theory assume that decision makersconduct a cost-benefit analysis that is invariant to factors externalto their choice, context is quite importantto behavior. How individualswill respondto legal rules depends on what reference points individuals recognize when making decisions. Decisions and behaviors are often affected by such factors as how an actor perceives choices relative to the status quo, whetherchoices are consistent or inconsistent with an actor's habits or traditions,and the temporaldistance of the rule's effects. In Part III, we describe experimentalevidence from the behavioralsciences thatmakes these and relatedpoints.
61. See Levi et al., supra note 54, at 7. 62. For a general discussion of heuristics, see IRWINP. LEVIN & JAMES V. HINRICHS, EXPERIMENTAL PSYCHOLOGY: CONTEMPORARY METHODS AND APPLICATIONS 246-48 (1995).

1070

LAW CALIFORNIA REVIEW

[Vol.88:1051

Thickversions of rational choicetheory face yet an additional implauThere is substantial evidence that decision often makers sibilityproblem. behavein ways thatare inconsistent with theirdirectself-interest.63 Indinot merelyaccording to theirsubjective vidualsformpreferences conceptions of well-being, but also in accordancewith norms that are often andsometimes Someprefersociallyconstructed evolutionarily adaptive. enceswell established in thebehavioral such as a commitment to sciences, notionsof fairnessor socialjusticeeven at a cost to the self, mightbe ingrainedby social or genetic forces, or by both. We discussevidenceof in PartIV. thesepreferences
C. TheResponsesof Rational Choice Theoryto Criticisms

withthe sortsof of rational whenconfronted Defenders choicetheory, criticismsthatwe have voiced, generallyrespondin one of threeways, noneof whichis satisfying. The first defense is an evolutionaryaccountof decision making, whofail to makerational decisions cannotsurvive whichasserts thatactors in a competitive world. In the competition for resources, utilityThe argubehaviorwill driveout nonmaximizing behavior. maximizing mentis mostcompelling in the contextof businessenterprises. Organizationswill seek to maximize profits,andthosethatfail to do so will be put out of businessby a lackof customers, Evenwhencomcapital,or both.64 or would to make or unimpossible plexity ambiguity optimization appear even will those that to survive, duly costly, by chance, manage optimize, andtheirstrategies will be emulated by competitors. We findthis account as it pertains to individuals. With unconvincing rare exception, individualswho fail to maximize their utility are not "driven out of the market." not necessarily) They might(although enjoy less happiness thantheirmorerational but they will live to counterparts, makedecisions another day.65

63. Of course, it can be arguedthat all actions taken by an individualare at least indirectlyin his self-interest.But to put the point this way is merely to adopt the definitionalversion of rationalchoice theory: thatis, to claim that we know all actions are self-interestedbecause individualsonly act in their self-interest. 64. See, e.g., Donald C. Langevoort, Organized Illusions: A Behavioral Theory of Why CorporationsMislead Stock MarketInvestors (and Cause Other Social Harms), 146 U. PA. L. REV. 101, 101-02 (1997) (noting that, with minor modifications, the literatureon corporationsassumes that firms that "depart too far"from maximizing shareholder wealth will be drivenout of business);Edward L. Rubin, Putting Rational Actors in TheirPlace: Economics and Phenomenology,51 VAND. L. REV. 1705, 1715 (1998) ("Neoclassical economists believe that the competitive marketwill induce optimal behavioramong firms by shapingthose that are adaptable,and eliminatingthose that are not."). 65. As one group of authors put the point: "Being a bad criminal is rarely fatal, and except possibly for organizedcrime, there is little opportunityfor 'hostile takeovers."'Jolls et al., supra note 13, at 1486.

2000]

LAW AND BEHAVIORAL SCIENCE

1071

Even with regard to businessenterprises, for whichthe evolutionary force of economicrationality most the accountis unbe forceful, might in to true most be circumstances. thatmakeclearly firms likely Certainly, decisionsroutinelyor in extremelyimportant circumstances suboptimal in product or capitalmarmaybe drivenout of business,butcompetition kets is rarelyso perfectthata firmthatoccasionally makesdecisionsthat fail to maximize or be takenoverby firmsthat profitswill facebankruptcy do maximize.66 In any event,if it weretruethatcompetition drivesimperrational behavior out of business such results wouldnot markets, fectly occurinstantaneously, andat anygivenmoment in timea substantial number of participants in markets wouldlikely be imperfectly rational actors whohavenotyet learned theirlessons.67 A weakerversionof this defenseconcedesthatmanyindividual and even some firmmistakesare not fatal,but stressesthatactorslearnfrom andcorrect failuresof rationality.68 This argument experience doubtlessly has validity,in the sensethattherearecircumstances in whichcorrectives arepossiblewithintherational choicetheory. Butwe do notthinkthatthis merits the that defenders on point weight place it. First,manyof the beII through haviors thatwe surveyin Parts IV arepersistent, evenwhenexof theirsubjects. Second,learning perimenters bringthemto the attention fromexperience is a long-term andtherewill alwaysbe a substrategy, stantialnumberof imperfectly rationalactorswho have not yet learned theirlessons. The secondargument offeredby defenders of rational choice theory when confronted with criticismis thatrationalchoice theory,while not behavioral Thetheory'sdefendperfect,is thebestsingleavailable theory. erscontend areaccurate, to a that,evenif thecriticisms theydo notamount coherent of decision that does a better of a theory making job predicting wide rangeof humanbehavior thandoes rational choice theory.69 As we
66. Cf. Langevoort,supra note 64, at 104 ("Empirical case studies aboundof systemic decisionmaking flaws, with many of the examples drawn from companies hardly destined for Darwinian extinction."). Further,it might be adaptive for firms to suffer from decision-making biases if such biases are closely linked to othertraits-confidence, optimism,forcefulness,for example-that provide a competitive advantage.See id. at 153-56. 67. See, e.g., id. at 148-49 (even if competitive forces eventually "weed out" poor decision makers, there will always be firms that make poor decisions that have not been weeded out yet); cf. Alexander Rosenberg, Does Evolutionary Theory Give Comfort or Inspiration to Economics?, in NATURALIMAGES IN ECONOMIC THOUGHT384 (Philip Mirowski ed., 1994) (arguingthat the theory of natural selection does not hold promise for helping modem microeconomics to explain or predict economic phenomena). 68. See, e.g., Arlen, supra note 41, at 1768 (citing RobertaRomano,A Commenton Information Overload, Cognitive Illusions, and Their Implicationsfor Public Policy, 59 S. CAL. L. REV. 313 (1986)). Later,Arlen also cites marketforces as a correctivefor individualerrors.See id. at 1777. 69. See, e.g., id. ("Behavioral economic analysis of law cannot serve as the basis for broad normative policy conclusions because it cannot provide a coherent alternative model of human behaviorcapable of generatingtestable predictionsand policy conclusions in a wide range of areas.");

1072

LAWREVIEW CALIFORNIA

[Vol.88:1051

shall see, imperfectly rational actors undoubtedly utilize many different decision-making strategies, no single one of which provides a general beIn a comhavioral descriptionmore realistic than rationalchoice theory.70 "It takes a heard around the academy, theory to beat a monly phrase even if do not amount to a theory. and these criticisms, correct, theory,"71 This rejoinderto the criticisms is sound, as far as it goes, but not particularly relevantto the project of developing a more nuancedunderstanding of behavior for use by legal policymakers. Rational choice theory is descriptively and prescriptivelyaccuratemore often than any other single theoryof behavior,or so even its critics generallybelieve. But the elegance and parsimonythat a single, universal theory of behavior such as rational choice can provide is of far less importance,if it is of any importanceat all, to legal policymakersthan to economists. The reason is that most laws are geared towardspecific portionsof the populationor to people who play specific roles. To be useful for legal policy, behavioraltheories need to predict (with reasonable success) the likely responses to legal rules of the particular classes of actors to whom the rules are geared, whether or not the responses of other classes of actors would likely be identical. For instance, if policymakers are considering revising products liability law, they need a on one hand, and product conprediction of how productmanufacturers, sumers, on the other, will likely respondto competing proposals.The predicted responses of these two groups need not be identical, nor need they be the same as they would be for other groups operatingin different contexts. Of course, it is possible thatthe predictedresponsesof the consumers of heavy machinery to a given phenomenon would be different from the predictedresponses of consumersof toasters to that same phenomenon.If this is in fact the case, ratherthan default to a reliance on rationalchoice theory, policymakers should consider whether it would be efficacious for the law to approachproducts liability for industrialgoods, which have a more limited, sophisticated market, differently than for mass consumer goods.
Crespi, supra note 52, at 167 (defending the "Chicago School's" version of economics against critics due to the lack of "alternativeexplanatory models of comparable scope and power");Posner, supra note 20, at 1559-60 (criticizing behavioral economics on the ground that it is "undertheorized": "Describing,specifying, and classifying the empirical failures of a theory [that is, rationalchoice theory] is a valid and important scholarlyactivity. But it is not an alternativetheory"). 70. As Richard Thaler has observed, the term often used to describe research that contradicts rationalchoice theory, "behavioral decision theory,"is somewhatof a misnomerbecause it is made up of more than one theory. Richard H. Thaler, The Psychology of Choice and the Assumptions of RATIONAL ECONOMICS Economics, in QUASI 137, 138 n.1 (1991). 71. Langevoort,supra note 21, at 1500 n. 1 ("[T]he notion has arisen ... that 'it takes a theory to beat a theory ...."'); cf TheodoreJ. St. Antoine, How the WagnerAct Came to Be: A Prospectus,96 MICH. L. REV. 2201, 2208 (1998) ("[T]he severest critic of the New Deal's labor policy would have it that 'it takes a theoryto beat a theory ...."').

2000]

LAW AND BEHAVIORAL SCIENCE

1073

Thereis no doubtthata single, universally theoryof beapplicable Butif universality is inconsistent andhighlydesirable. havioris convenient withsophistication andrealism, arebetter off foregoing legalpolicymakers a collection of and, instead, situation-specific creating universality minitheories usefulin the analysisof discretelegal problems.7 Moreover, there is a greatdeal to be said for an incremental to theoryapproach onjust enoughtheoryto dealwiththe matter at building-that is, grafting to construct a universal handrather thanattempting Cass Sunstein theory. has recently argued that the common law process tends to produce theorized "incompletely arguments," by whichhe meansthatthedecisions rendered law courts to resolve common by privatedisputescontainjust cover not to the instant facts and the overarching enoughtheory theoryor coherence thatalternative institutions (suchas a legislature) mightseek to achieveas justification.73 To try to do moreby, for example,reachinga andcomprehensive disthorough theory,Sunsteinargues,wouldproduce cord.Thisview of the sometime of common law adjudication is superiority to the the of law-and-behavioral-science analogous pragmatic superiority thatwe advocate here. approach The thirddefenseis to respond to evidenceof a purported failingof rational choice theoryby explaininghow the anomalyin questioncould possiblybe consistentwith the theoryif viewed froma different angle.74 This strategyavoidsthe problemof directlyrefutingor provingrational choice theorywrong,but in doing so the defender to a implicitlyretreats of rational choice that is so thin that it can have no conception theory predictivevalue. as an example,the contention thatrational individuals will Consider, not pay attention to sunkcosts when choosingbetweenoptionsbecause suchcosts cannotaffectthe marginal utilityto be derivedfromfutureactivities.75 Thisleadsto theprediction thatan individual who has boughtan ticket to the will be no more thanan expensive symphony likely to attend individualwho has been given a free ticket, assumingthe two have an for classicalmusic and that all otherthings are also equal appreciation There is equal. ampleevidencethatmanypeople wouldin fact be more
72. Cf MarkKelman, Behavioral Economics as Part of a RhetoricalDuet: A Response to Jolls, L. REV.1577, 1586 (1998) (observingthat behavioraleconomics does Sunstein,and Thaler, 50 STAN. not suggest an alternative general theory of behavior to contrast with rational choice theory, but claiming thatthis fact does not "fatallywound [the] enterprise"). 73. See generally Cass R. Sunstein, Incompletely TheorizedAgreements, 108 HARV.L. REV. 1733 (1995). 74. For example, the new edition of RichardPosner's EconomicAnalysis of Law makes note of a few of the phenomenathat we discuss infra, but claims thatthese "apparent from rationality departures may be explicable in rational-choiceterms."POSNER, supra note 26, at 20. 75. "Sunk costs" (sometimes equated with "fixed costs") are costs that have already been incurredand do not vary with one's subsequentactions. An example is the annualsubscription cost of a magazine.For a detaileddiscussion of this point, see infra PartIII.E.

1074

CALIFORNIA LAWREVIEW

[Vol.88:1051

if theyhavepaidfor the ticket,reasoning the performance likelyto attend that"themoneywouldbe wastedif I don'tgo."76 A clever proponent of rationalchoice theorymightrespondto this the violation of thatrational peoplearefully theoryby asserting apparent laterto attend a particular awareof thefactthattheymaybe reluctant concert,and,knowingthatthey will bothenjoythe concertif theyattendand thatthey will feel guilty if they allow an expensiveticketto go unused, a ticketin advanceas a precommitment to ensure they purchase strategy theirattendance. In otherwords,pre-purchase is usedas a meansto control to be painful butarereally one's tendency notto do thingsthatmayappear
desirablein the long run.77

behaviorex Responseslike this providea useful way to understand the ex post account buttheyalso post (although mightwell be inaccurate), the problem demonstrate withrelyingon rational choicetheoryto inform future-oriented makes it impossible to use raTheresponse policymaking. tionalchoice theoryto predictwhether a personwill be morelikely to attend the symphonyif he has purchaseda ticket in advance(and has, "sunk" therefore, ticket)thanif he weregiventhe moneyintoa symphony ticketfor free.Thisdefenseof rational choicetheory,in fact,demonstrates ourbasiccontention thatthe theorycanprovidea plausible of description behavior when the is in such a thin form that it has only theory adopted littleprescriptive value.78
D. Modifyingthe BehavioralPredictionsof Rational Choice Theory

To claim that rationalchoice theory is an insufficientbehavioral modelon whichto baselegalpolicyis notto arguethatindividuals behave do in some circumstances). it Rather, irrationally (although they certainly is to assertthatlegal scholars to understand the incentive effects of seeking law in orderto proposeefficaciouslegal policy shouldnot be limitedto rational choice theory.The goal of the law-and-behavioral-science movementis not, at leastat this stage,to replacerational choicetheorywithan inconsistent but to modifythe implausible elementsof rational paradigm

76. See infra note 297 and accompanyingtext. 77. These sorts of explanations have been a particular theme of the work of game theorist Thomas Schelling. See, e.g., Thomas C. Schelling, Self-Commandin Practice, in Policy, and in a REV. 1, 1 (1984) (discussing the phenomenonthat people Theory of Rational Choice, 74 AM. ECON. will rationallyseek to prevent,compel, or alter their futurebehavior, "to restrict[their]own options in violation of what [they] know[] will be [their]preferenceat the time the behavioris to take place"). 78. Thus, we agree with the observation of Jolls, Sunstein, and Thaler that Richard Posner's contention that actors' demonstrated deviations from rational choice theory are not irrational demonstratesmore of an agreementthan a disagreementwith the behavioralistagenda. See Christine Jolls et al., Theoriesand Tropes: A Reply to Posner and Kelman,50 STAN. L. REV. 1593, 1594 (1998).

2000]

AND BEHAVIORAL LAW SCIENCE

1075

the inadequate elementsin orderto createa choicetheoryandsupplement tool withmorepredictive powerin specificsituations.79 choice theoryareinadeIf lightlyspecified,thinversionsof rational because tools for make too few predictions, they legal policymakers quate andmorefully specified,thickversionsof the theoryprovidemoresatisare often incorrect,the obvious but these predictions fying predictions withoutsubtake to how can theorists is steps solve one problem question the other?The answer,we believe,is to enhancerastantially worsening a moresubtle,textured tionalchoice theoryby developing understanding on of how actorsmakedecisionsin varioussituations, relyingextensively will likelynot research data.Theend pointof thisbroad agenda empirical be a singleunifiedtheorydesignedto explainor predictthe full realmof it is morelikely to be a praghumandecision-making behavior.Rather, maticcollectionof situation-specific insightsthatcan assistpolicymakers dealing with relevantproblems.What the law-and-behavioral-science to rational anduniversality, relative choice gives up in parsimony approach for by its increased usefulnessin the taskof theory,will be compensated developing situation-specific legalpolicy. II
BOUNDED RATIONALITY AND THE USE OF HEURISTICS

Whereas both thin and thick versions of rational choice theory presume that individualsact so as to maximize their expected utility, research in decision making, along with common experience, suggests that this prediction fails in many circumstances. "Bounded rationality," the term coined by HerbertSimon,80 capturesthe insight that actors often take short cuts in makingdecisions thatfrequentlyresult in choices that fail to satisfy the utility-maximization prediction. Actors may make boundedlyrationaldecisions for two somewhatdifferent reasons. In some cases, actors faced with a decision might aim to make a satisfactorychoice-one that meets a specified aspirationlevelratherthan one that maximizes their utility. For the decision maker, such intentional"satisficing"8l behavior is often quite sensible in light of both the costs of obtaining and processing the informationnecessary to make
79. Thus, we do not necessarily disagree with RichardPosner's contention that the "majorfruit of behavioraleconomics will be the stimulus it provides to new and betterrational-choicetheorizing," Posner, supra note 20, at 1567, although we think it is questionablewhether a significantly expanded and nuanced approachto "rational-choicetheorizing" would still be appropriatelylabeled "rational choice theory." 80. HerbertA. Simon, Rational Choice and the Structure of the Environment,in MODELSOF
MAN: SOCIALAND RATIONAL261, 270-71 (1957) ("Since the organism ... has neither the senses nor

the wits to discover an 'optimal' path ... we are concernedonly with finding a choice mechanismthat will lead it to pursuea 'satisficing' path, a path that will permit satisfactionat some specified level of all its needs."). 81. This term was coined by HerbertSimon. See id.

1076

CALIFORNIA LAWREVIEW

[Vol.88:1051

choicesandthe cognitivelimitations of human beingsthatofmaximizing ten render In other words,we utility-maximization physicallyimpossible. behavior canbe rational in a "global" sense.Or, mightsay thatsatisficing it into the rational choice we that the costs of framework, fitting mightsay or those createa information, processing by cognitivelimitations, imposed on maximizing constraint behavior-just as incomeandtime imposeconstraints on the feasibleset of choices.Onemightthenexplorethe systematic variationsin maximizingbehavior that might result from these andmaketestablepredictions in light of these constraints aboutbehavior variations. behavior be so as Despitethefactthatsatisficing might trimmed to fit withintherational it nonetheless choiceframework, violatesboththin andthickversions of rational choicetheory,at leastas theyareusuallyimin legal (andnonlegal)scholarship.82 This is because plicitlyunderstood causes actors fail to in to maximize their satisficing utility the particular situation athand. decision-making In othercircumstances, rational decisionmakingis an unboundedly intentional an of unconscious use of heuristics in judgment consequence and decision-making tasks.The widespread use of heuristics,at least in without manycases, is no doubta quiteuseful evolutionary adaptation; suchmentalshortcuts, the taskof makingeven relatively decisions simple wouldbecomeso complexthatdailylife wouldalmostcertainly grindto a halt. But the use of heuristics failureof surelyresultsin the widespread decisionmakersto maximizetheirexpectedutilityin particular decision situations.83 This Partreviews a varietyof causes of boundedly rational decisionmakinganddiscussestheirsignificance for the analysisof legal rules.
A. Decision-MakingStrategiesthat Do Not MaximizeExpected Utility

Commonexperiencesuggeststhatthereare some decisionsthatseor thatseem impervious to our attempts to reverelytax ourcapabilities duce them to manageable dimensions.In this section we considertwo aspectsof decisionmaking-complexityand ambiguity-that lead most humans to makechoicesthatdo notmaximize expected utility.

82. Cf. Blaug, supra note 3, at 11 (observing that game theorists are "exclusively concerned" with how actors will behave with perfect informationand unboundedcalculating abilities ratherthan how they will behave in the face of imperfectinformationand limited cognitive abilities). 83. There is, however, disagreementwithin the judgment and decision-makingcommunity as to how close (or how distant) oft-used heuristics can come to approximatingutility maximization.See, e.g., Gerd Gigerenzer & Daniel G. Goldstein, Reasoning the Fast and Frugal Way: Models of Bounded Rationality, 103 PSYCHOL. REV.650, 662 (1996) (arguing that decision-making heuristics often lead to performancein real world tasks that approachesthe performanceof "rational" decisionmakingapproaches).

2000]

AND BEHAVIORAL LAW SCIENCE

1077

1. Complexity The limits of human cognitive ability make utility-maximizing behavior physically impossible in some situations.HerbertSimon illustrates the problem with the example of the game of chess.84 From one perspective,chess is a simple and uninterestinggame.85 Players have a sinthe gle, clearly specified goal (checkmate opponent's king), and all possible moves are known to both players.86 In theory,a player should be able to determinethe optimal strategyto reach the goal by constructinga decision tree with branchesfor each possible move of each player, look to the bottom of the tree to find a branch that leads to checkmate, and follow the branchback up to the beginningof the tree.87 The problemis that while chess is simple in one sense, in that there is a relatively small numberof pieces and their allowable moves on the board are fixed,88 it is, in anothersense, an extremely complicated game, in that there are approximately10120possible combinationsof moves in a game.89 In light of this complexity, when it is time for a chess player to make a move, he must by necessity concede his inability to select a move guaranteed to maximize his expected utility and instead adopt a more simplified decision-makingstrategy(what Simon calls a "stoppingrule").9 Complexitybeyond humancognitive capacityis a sufficient condition for an actor to substitutea simplified decision strategyfor a complete expected utility calculation, but it is not a necessary condition. Even if a choice is not too complex for an actor to process physically, she might choose to limit her search for informationor considerationof the decision short of reaching a utility-maximizingdecision.9'The decision to adopt a
84. See Herbert A. Simon, Theories of Bounded Rationality, in 2 MODELS OF BOUNDED RATIONALITY: BEHAVIORAL ECONOMICS AND BUSINESSORGANIZATION 408 (1982). 85. The famous game theorists Von Neumann and Morgenstem said, "[I]f the theory of Chess were really fully known, there would be nothing left to play." JOHNVON NEUMANN & OSKAR
THEORY OF GAMESAND ECONOMIC BEHAVIOR 125 (3d ed. 1953). MORGENSTERN,

86. In game theory terms, chess is a game of complete information.See, e.g., Simon, supra note 84, at 416. 87. Game theorists call this problem-solvingstrategy"backwards induction."See BAIRD ETAL., supra note 2 (defining backwards induction as a "solution concept applicable to an extensiveform game made up of informationsets consisting of single node's); AVINASH K. DIXIT& BARRY J.
NALEBUFF, THINKING STRATEGICALLY: THE COMPETITIVE EDGE IN BUSINESS, POLITICS, AND

EVERYDAY LIFE 40-44 (1991) (explainingthe theory and practicaldifficulties of applyingsuch a theory to chess).
88.

chess providesno practicalhelp to the player). 89. This figure assumes there are about thirty possible moves each turn and about forty total turnsper game. See Simon, supra note 84, at 413. 90. Id. at 415. Chess tournamentsforce competitors to adopt a stopping rule by placing a time limit within which each competitormust make a move.
91.

Cf VON NEUMANN & MORGENSTERN, supra note 85, at 125 (noting that the triviality of

a theory in which decision makers,while using a variety of differentstrategies,generallytry to balance effort and accuracy).

See generally JOHN W. PAYNE ET AL., THE ADAPTIVEDECISION MAKER(1993) (articulating

1078

CALIFORNIA LAWREVIEW

[Vol.88:1051

simplified strategymight be sensible given the marginalbenefits and costs of making an optimal decision relative to a satisfactory one; in other words, the decision not to maximize utility when solving a single problem But althougha decision might in fact maximize the actor's overall utility.92 not to optimize in a particularsituationmight be globally "rational," such behavioris rarelycontemplatedby scholarsemploying rationalchoice thedecision. ory in the analysis of a particular Decision researchershave identified the complexity of a decision as a from the type of complete cost-benefit analysis leading cause of departures of decision options predicted by expected utility theory. Acting consistently with expected utility theory requires a substantialamountof cognitive effort. As the problem becomes more complex, either because there are more options from which to select or because each option has more attributesassociated with it,93actors might attemptto minimize effort by adopting simplified strategies,thus violating the proceduralpredictionsof rational choice theory.94 For example, one study found that experimental subjects were less likely to select a house that maximized their utility (defined by questions the subjects were earlier asked about their preferences) from among five alternativesas the numberof attributespresented to the subjectswas increasedbeyond ten.95 Decision theorists have suggested a number of decision-making approaches that actors employ in certain situations and that fall short of maximizing the expected utility of a particulardecision. The difference between satisficing and optimizing, as Melvin Eisenberg puts the point, is the difference between searching for the sharpest needle in the haystack and searchingfor a needle that is sharpenough for sewing.96 To pursuethe chess example, a player might satisfice in his choice of moves by selecting the first move that places one of the opponent's pieces in jeopardy. It should be clear that this strategymay not maximize the player's expected utility (assuminghis goal is to win the game), because a move that fails to place the opponent's pieces in immediatedangermight place the player in a more advantageousposition in the long run.
92. Recall our example of Buridan'sass, see supra text accompanyingnote 54. The ass might be said to have starved because he searched for a global optimum rather than merely satisficing by choosing a local optimum. 93. See JamesR. Bettmanet al., ConstructiveConsumerChoice Processes, 25 J. CONSUMER RES. 187, 189 (1998). 94. See, e.g., id. at 192 (noting that actors must compromise between their desires to make optimal decisions and to minimize effort); Peter Wright, ConsumerChoice Strategies: Simplifyingvs. RES. 60, 62 (1975) (noting that actors must "compromise between Optimizing, 12 J. MARKETING optimizing ... and reducingthe strainsof decision making"). 95. See Naresh K. Malhotra,InformationLoad and ConsumerDecision Making, 8 J. CONSUMER RES.419, 422-23 & tbl.l (1982). 96. Melvin Aron Eisenberg, The Limits of Cognition and the Limits of Contract, 47 STAN.L. REV.211, 214 (1995).

2000]

LAW AND BEHAVIORAL SCIENCE

1079

an Using the moreformal"lexicographic" decision-making strategy, actoridentifies the attribute of eachof herdecisionchoicesthatis themost and selectsthe alternative with the highestvalueon thatattribimportant ute.97 Forexample,a restaurant entree,or patron mightselectthe cheapest she mightselect the entreeshe thinkswill tastethe best. Notice thatthis strategymightcause the actorto select a choice with a lower expected valuethana competing ranked choice,for example,if the former slightly than the latter on the most attribute but the latterwas suhigher important to the former on severalotherattributes, or if the latter was substanperior to the former on even a singleless important attribute. More tiallysuperior choice is inconsistent with simply,lexicographic expectedutilitytheory becausethe actorviolates the principleof commensurability-the high scoresthattheoptimal decisionchoiceachieveson attributes otherthanthe most important one are not directlycompared to the high score thatthe decision choice achieves on themostimportant attribute.98 suboptimal "Elimination features of the"satisficing" (EBA)combines by aspects" and"lexicographic" Thisapproach callsforthe actorto examine strategies. how alternatives rateon the most important attribute and eliminatefrom consideration all alternatives thatdo not meeta threshold level of valueon thatattribute.9 If onlyone choiceremains, the actorselectsit. If morethan one choice remains(or no choices),the actorselectsthe next most desirableattribute, andcontinues thepattern untila singlechoiceremains.'? For decide that is the most example,ourrestaurant patron might price importantto her, and thus excludeall choices thatcost morethan$10. Then, withinthe choicesthatremain, she mightselecttheentreeshebelieveswill tastethe best. Like the "lexicographic" EBA violatesthe comapproach, that its use will often leadto choicesthat mensurability principle, meaning do notmaximize theactor'sexpected utility. In anygiven choicesituation, actorsmightuse a nearlyinfinitenumber of decision-making thatdeviatefromthe predictions of exstrategies In a decisionstrategies, pectedutilitytheory. leadingbook on individual the authors concludefroma wealthof empirical evidencethatactorsare morelikelyto combine elements of different thanto use a "pure" strategies
97. See Bettmanet al., supra note 93, at 190. 98. Consider the following illustrative example: In one study, researchersasked experimental subjects to list factors that would be relevant to selecting a cancer treatment (for example, cost, likelihood of survival, and so on) and then describe how they would go about choosing between alternativeswith differentscores on these differentattributes.See BarbaraE. Kahn & JonathanBaron, An ExploratoryStudy of Choice Rules Favoredfor High-StakesDecisions, 4 J. CONSUMER PSYCHOL. 305, 312 (1995). Approximatelyone-thirdof subjectsreportedthatthey would select the approachwith the highest value on the most importantattribute(that is, they would use a lexicographic strategy).See id. 99. See Bettman,et. al. supra note 93, at 190. 100. See id.

1080

CALIFORNIA LAWREVIEW

[Vol. 88:1051

For example, a chess playermight examine every legal move he strategy.'10 could make and every possible response to that move, and then determine which choice will leave him in the best position afterhe moves and his opponent responds. This strategyis, of course, not globally optimal, because being in what appears to be the strongest position after any single move does not necessarily maximize a player's likelihood of checkmating his opponent. It can be viewed, instead, as settling for a local optimum,or, in other words, a very approximateoptimization. Similarly, a player might initially select a few possible moves and attemptto follow the impact that each would have (makingcertainassumptionsabouthow the opponentwill make decisions) as many moves into the future as possible, and select the move that would appearto leave him in the strongestposition at the end of the string.There is some evidence thatmost chess players are considerably less methodicalthan either of these imperfectoptions and that they instead search potential moves until they find one that meets a minimum satisfactory condition and then choose that move.'02Regardless of the method a player uses for limiting his explorationof options, it is certainthateven the best chess players can evaluate only a small numberof moves relative to the total numberof possible combinations.'03 One importantareain which complexity and ambiguityunderminethe policy prescriptionsof traditionallaw-and-economics analysis is contract law. Law and economics posits that the law should enforce all agreements between informed and competentprivate parties, assuming that the agreement has no (or relatively low) externalitiesthat cause harm to third parties.'4 This theory of contractlaw is firmly rooted in the expected utility version of rational choice theory. If all parties wish to enter into an enforceable agreement,each must have determinedthat the benefits of being partyto the contractoutweigh its costs, and that its utility will be higher if the contractis enforced than if it is not. In addition, if a different bargain
101. See PAYNE ETAL.,supra note 91, at 28-29. 102. See id. (citing De Groot, 65). 103. Precisely because humancomputational abilities are limited but those of computersare not, it has long been thoughtthat an appropriately programmed computercould beat even the best humanin a game of chess. Indeed,that happenedin February,1996, when in Philadelphiathe IBM program"Deep Blue" defeated the world chess champion, Gary Kasparov,in a series of games. See Bruce Weber, In Feb. 11, 1996, at A32; see also NT serverfrom Upset, ComputerBeats Chess Champion,N.Y. TIMES, Siemens challenges the world'sfastest chess player, Viswanathan Anand, M2 Presswire,June 9, 1998 (1998 WL 12973670) (reportingplans for the program "Fritz,"which beat "Deep Blue" at the last world chess championshipin 1995, to challenge Anand, the world's numbertwo-rankedplayer, at the Frankfurt Chess Classic 98). Of course, not every decision that humanbeings must make is as complex as planning a chess move. But we believe that the complexities of our chess example pervade many humandecision-makingtasks, and probablymost that interestlegal scholars. 104. See, e.g., COOTER & ULEN, supra note 8, at 207-09; see also Melvin Aron Eisenberg, The L. REV.741 (1982). An "externality" Bargain Principle and Its Limits,95 HARV. is an unbargained-for cost or benefit that the utility- or profit-maximizingactivities of one person impose on others. Secondhand smoke and waterpollution areexamples.

2000]

LAWAND BEHAVIORAL SCIENCE

1081

would have increasedthe combined utility of the parties,it is assumed that the parties would have adopted different contractualterms. Consequently, the agreed-uponterms must maximize the parties'joint expected utility. If the goal of contractlaw is to help partiesmaximize the joint utility of their agreements,the traditionallaw-and-economicstheoryneeds to be modified in situationsin which complexity and ambiguitycreate substantialbarriers to optimizingbehavior.105 Consider the following timely policy issue: in the 1990s, thousands of laws were introduced in state legislatures"06 (and hundreds enacted107) that would mandate certain terms of contracts between health insurance providers and their customers. Many of these laws took the form of requiringthat all insuranceprovidersmust include in their policies coverage for specific treatmentsunder certain conditions. For instance, in response to constituent anger at "drive through deliveries," at least forty-one have legislated a higher states,'08 along with the federal government,'09 minimum coverage for postnatal hospital stays than most health maintenance organizations were theretoforewilling to provide. Expected utility theory suggests that such state interferencewith private contracts would encourage an inefficiently large amount of resources to be devoted to medical care. If health insurance consumers make optimal cost-benefit tradeoffs, insurance providers would have a market incentive to provide enhanced maternitybenefits if customers were willing to pay the higher premiums necessary to cover the marginalcost of such benefits. The absence of such coverage in many or all insurance agreements is evidence that consumers do not wish to pay for it.110 According to expected utility it follows that laws the theory, mandating provision of specific maternity benefits create an inefficient allocationof social resources.

105. The role that contractlaw can play in facilitatingthese transactionsis to allow the partiesto save on the transactioncosts of concluding bargainsor gifts by providingdefault and mandatoryterms that, among other things, alert the parties to the pitfalls of fraud, strategic behavior, asymmetric & ULEN, information,and other sources of bargainfrustration.See COOTER supra note 8, at 199-04, 207-12. 106. See Thomas Bodenheimer,The HMO Backlash - Righteousor Reactionary?,335 NEWENG. J. MED. 1601, 1601 (1996) ("In 1996 alone, 1,000 pieces of legislation attempting to regulate or weaken HMOs were introducedin state legislatures...."); Families U.S.A., HMO Consumersat Risk, States to the Rescue (visited Mar. 16, 2000) <http://www.familiesusa.org> (noting that more than 1,000 managedcare bills were introducedin 1995). 107. See Families U.S.A., Hit & Miss: State Managed Care Laws (visited Mar. 16, 2000) <http://www.familiesusa.org>(asserting that 49 states had enacted at least 1 of 13 managed care patientprotectionlaws thatthe reportstudied). 108. See Molly Stauffer,Inpatient Care After Childbirth,National Council of State LegislaturesHealthPolicy TrackingService, ISSUE 1997 AnnualEdition,Dec. 31, 1997. BRIEFS, 109. See Newborns' and Mothers' Health ProtectionAct of 1996, Pub. L. No. 104-204, 110 Stat. 2935 (1996). 110. For a succinct version of this argument,see PatriciaM. Danzon, TortLiability: A Minefield STUD. 491 (1997). for Managed Care?, 26 J. LEGAL

1082

REVIEW CALIFORNIA LAW

[Vol.88:1051

The complexityof the healthcare choices thatconsumers (or their insurance mustmakeamongcompeting however, products, employers11) The renders this efficiencyconclusion problematic. possiblepermutations of coverages thatcould,in principle, be provided by a given healthinsurandthe features of eachpolicythatareactually ancepolicyarenumerous, and promotional materials can overwhelm even describedin advertising that As a it is consumers. consumers, result, highlyunlikely sophisticated or even agentsof consumers who mightbe makingpurchasing decisions, feature of the will carefully and competing policies analyze compare every It is far decision. in a waythatwouldmaximize theutilityof theirpurchase will use a simplified morelikelythat,to reduce effort,consumers cognitive between healthinsurwhen competing strategy choosing decision-making will not necessarily market anceplans.If this is the case, the unregulated withtheefficientlevel of coverage.12 health insurance customers provide of boundedly rationalconsumer Underquiteplausibleassumptions mandated benefitterms choice in a very complexdecision-making task, contracts could actuallyincreasethe efficiencyof insurance by requiring thatconsumers arewillingto pay for butarenot insurers to sell coverages Whether or not policymarket behavior.13 likely to insist uponthrough of resources the efficientallocation should makerswho wish to promote will ultimately on the level of confifavorsuchbenefitmandates depend bodies to identifywhich dence they have in the abilityof governmental contract termsare inefficiently absentfrommost insurance contracts and whichtermsareefficientlyabsent.But policymakers arebetterservedby abouthow healthcareconsumers arelikelyto makepurchasing predictions decisionsthataremoreplausible thanrational choicetheoryis ableto provide. The policyrelevance of evidencethatactorssubstitute desimplified cisionstrategies forthefull cost-benefit analysis positedby expected utility can be from the of care In theory generalized example managed regulation. in situation which decisions are relative to the any extremelycomplex value of the resourcesinvolvedor to the capacitiesof those makingthe choice, decisionmakersare likely to makechoices thatfail to maximize theirexpectedutility.If efficiencyis the goal of the legal policy at issue, shouldat leastconsider whether intervention in policymakers government
111. Approximately 90% of privately insured Americans receive their health coverage through their employers, see Melinda L. Schriver& Grace-MarieAmett, UninsuredRates Rise Dramatically in States with StrictestHealth InsuranceRegulations,HERITAGE FOUNDATION REP.,Aug. 1998, at 8-9 & tbl.2, and employers pay roughly 80% of the costs of private health insurance.See KatherineR. Levit et al., National Health Spending Trends in 1996, 17 HEALTH AFFAIRS 35, 46 (ex. 6) (1998). Approximately48% of employees are offered a single health plan (no choice) by their employers. See AFFAIRS Lynn Etheredgeet al., Whatis Driving Health SystemChange?, 15 HEALTH 93, 94 (1996). 112. See Russell Korobkin, The Efficiency of Managed Care "Patient Protection" Laws: IncompleteContracts,BoundedRationality,and MarketFailure, 85 CORNELL L. REV.1 (1999). 113. For a more detailedversion of this argument,see id. at 62-74.

2000]

LAW AND BEHAVIORAL SCIENCE

1083

the marketcan be used to enhancethe efficient allocationof social resources.

2. Ambiguity

the consequences In addition to complexity, of ambiguity concerning decision can necessitate from decisionalternatives the suboptimal making arisesin two verydifferof expected utilitytheory. Ambiguity perspective for rational ent types of situation, choice only one of which is troubling In first the does not know what outcome the actor will situation, theory. resultfromhis possiblechoices,buthe knowsthe possibledistribution of outcomes.Consider, for example,an actorwho will win $100 if he cora coin calls calls it. In this cirrectly flip andlose $100 if he incorrectly the value of is heads but of cumstance, uncertain, the probability choosing each result,given the choice of heads,is knownprecisely.Ambiguity of this sortdoes notprevent the actorfrommaking choicesthatmaximize his expected utility.'14 The secondandtroubling thatdecisionmakers oftype of ambiguity the ten face is ambiguity content of decision Conalternatives. concerning of home insurancewho must decide sider, for example,the purchaser whether to pay an additional annual sumfor a riderthatwouldprotect him in case of earthquake If the could estimate the damage. purchaser probwill destroyhis home,he could maximizehis abilitythatan earthquake thecost of theadded to theexpected insurance benefit utilityby comparing it wouldprovide. Butif thepurchaser canneither estimate thelikelihood of nor obtain information that wouldallowhimto do so, earthquake damage it becomesimpossible for himto makethetypeof optimizing decisionthat rational choicetheory predicts.15 The implications of the ambiguity can be seen through anproblem otherfeatureof contract law thatrunscontrary to the free market bias of rationalchoice theory: the rule regardingstipulateddamagesclauses. Hornbookcontractlaw provides that contractual provisionsrequiring in the case of a breach of contract areunenspecifiedliquidated damages forceable unlessit is difficultfor actualdamages to be ascertained andthe are a reasonable of actualdamages.116 specifieddamages approximation
114. Expected utility theory, see supra PartI.A.2, explains how one ought to take account of this ambiguity.We know thatthe expected value of the bet, assumingthe coin to be a fair one, is 0.5 ($100) + 0.5 (-$100) = 0. The individualcan, by presumption,compute the expected utility of these outcomes in order to compare this alternativeto others. We consider evidence that actors often fail to assess probabilitiesaccuratelyin PartIII.B ("Heuristicsand Biases"), infra. 115. In addition to the effect cited, it appears that individuals dislike ambiguity, preferring a precisely understood probability to an uncertainprobability of equal expected value. Thus, "people would ratherface a known chance of an adverseconsequenceof 2/1,000 ratherthan a 50/50 chance that the risk is either 1/1,000 or 3/1,000." Viscusi, supra note 44, at 331. 116. See, e.g., Lake River Corp. v. Carborundum Co., 769 F.2d 1284 (7th Cir. 1985); Board of Trustees v. Johnson, 507 So. 2d 887 (Miss. 1987); Robbins v. Finlay, 645 P.2d 623 (Utah 1982); E.

1084

CALIFORNIA LAWREVIEW

[Vol.88:1051

This rule provokes consternationfrom devotees of rationalchoice theory, who assume that competentpartieswould never agree to a stipulateddamages term, no matterhow one-sided it may appearto a thirdparty,unless it made both signatoriesbetteroff and maximizedtheirjoint utility."7 As Melvin Eisenberg has observed, however, specifying liquidated damages for breach of contractis often made quite difficult by ambiguity problems. In many (but certainly not all) contractualsituations, the type and extent of damages that one party will suffer in the event of a breachis difficult to predict. Often, partieswill not even be able to anticipatecertain possible damages scenarios, making a reasonable prediction about the likelihood of such scenariosimpossible. The consequenceis that liquidated damages provisions, to the extent that parties intend them to serve as a proxy for expected actualdamages, are likely to be quite erratic.18 The policy implicationof this observation,accordingto Eisenberg, is that it is proper for courts to refuse to enforce stipulated damages provisions that diverge widely from actual damages suffered in the event of We are less sanguine about this conclusion, because it implicitly breach."9 assumes that parties are better served by accuracyin damages assessments than by ex ante certaintyas to what damages will be if a breachoccurs-a proposition that is open to debate. But regardless of the ultimately "correct"stipulateddamages rule, we think sound policymaking can only be enhancedby the subtlety of Eisenberg's analysis of boundedrationality in this particular context, as comparedto the more elegant but less realistic of rational choice theory. predictions Decision-MakingHeuristicsand Biases In a world in which the consequences of most decisions are, to some degree, uncertain,actorscan maximize the expected utility of a given decision only if theirjudgments are based on accurateperceptionsof the likelihood that specific choices will lead to various possible outcomes.120 To take a simple example, consider an actoroffered a gamble in which he will receive $10 if a coin toss comes up heads, and he will lose $15 if the toss comes up tails. The actor will maximize his expected utility by declining the gamble, but he is certainto make this utility-maximizingdecision only if he knows that there is a 50% probability that a fair coin will turn up
ALLAN FARNSWORTH ONCONTRACTS 939 (2d ed. 1990) (stating the general principle FARNSWORTH, that liquidated damages provision sustained when loss resulting from breach will be "uncertainin amountand difficult to prove"). 117. See, e.g., COOTER & ULEN, supra note 8, at 235-37. 118. See Eisenberg,supra note 96, at 227-28. 119. See id. at 230. 120. Cf. Matthew Rabin, Psychology and Economics, 36 J. ECON. LITERATURE 11, 24 (1998) (noting that "[e]conomists traditionally have assumed that, when faced with uncertainty, people correctlyform their subjectiveprobabilisticassessmentsaccordingto the laws of probability").

B.

2000]

LAW ANDBEHAVIORAL SCIENCE

1085

heads.If the actormistakenly believesthattheprobability thatthetoss will come up headsis 66%,he mightacceptthe gamble,thusfailingto maximizehis expected utility. in thebehavioral Research scienceshas demonstrated thatindividuals are systematically biased in theirpredictions of the probable resultsof variousevents. This line of research, by psychologistsAmos pioneered andDanielKahneman,'2does not necessarily leadto the concluTversky sion thatindividuals are "bad" decisionmakers.Often,systematic errors arise fromthe use of decision-making heuristicsthat simplifydecisionthe costs of information makingtasks,thussignificantly reducing processit possibleto operate in an ining anddecisionmaking, thereby rendering In some cases, systematicdecision-making creasinglycomplexworld.'22 errorsmightbe the resultof perceptual biases thatmay be, on balance, But whether or not the well-documented collecevolutionarily adaptive.'23 tionof heuristics andbiasesarerational in a adaptations globalsense,they havethe consequence of causingactorsto makedecisionsthatviolatethe of rational choicetheoryin individual circumstances. Thissecpredictions tion bothreviewsthe evidenceof these failuresof rational choice theory andconsiders theirimplications fortheanalysis of legalrules.
1. Availabilityand Representativeness

To accurately of futureevents, actorsmust predictthe probability considerthe statistical that an event will occurand "update" probability this"base rate" withanyavailable information about (adjust) particularized a specific situation.'24 Thereis significant that actors evidence,however, underuse baserateswhenmaking or systematically probability predictions, Two relatedphenomena-the"representativeness ignorethemaltogether. heuristic" andthe "availability heuristic"'25-describe the ways in which actors oftenerrin making thesepredictions.
121. For an analysis of the broad impact of the work of Tversky and Kahneman, see David Laibson & RichardZeckhauser, Amos Tverskyand the Ascent of Behavioral Economics, 16 J. RISK& UNCERTAINTY 7 (1998). 122. See Amos Tversky & Daniel Kahneman, Judgment Under Uncertainty: Heuristics and Biases, 185 SCIENCE 1124, 1124 (1974). 123. See, e.g., THEADAPTED MIND: EVOLUTIONARY PSYCHOLOGY ANDTHEGENERATION OF CULTURE(JeromeH. Barkowet al. eds., 1992); STEVENPINKER,HOW THE MIND WORKS(1997). 124. This method of calculating probabilities is known as "Bayes' Law." ROBERT V. HOGG & ALLEN T. CRAIG,INTRODUCTION TO MATHEMATICAL STATISTICS 208-09 (3d ed. 1970) (showing that in application of the law, one "updates"a prior probability assessment with new information, in a mannerdefined by the law, in orderto get a posteriorprobabilityestimate). 125. For a more detailed (but still brief) discussion of the representativenessand availability heuristics, see PLOUS, id. at 121-30 (availability).More supra note 34, at 109-20 (representativeness); detailed treatmentsare contained in JUDGMENT UNDER UNCERTAINTY: HEURISTICS ANDBIASES 23100 (Daniel Kahneman et al. eds., 1982) [hereinafter JUDGMENT UNDERUNCERTAINTY] (representativeness);id at 163-210 (availability); Timur Kuran & Cass R. Sunstein, Availability Cascades and Risk Regulation, 51 STAN. L. REV.683 (1999).

1086

LAWREVIEW CALIFORNIA

[Vol. 88:1051

The "representativeness heuristic" refersto the tendency of actorsto base rates and overestimate the correlation between what ignore something to be andwhatsomething is.126As an example of this tenappears actually "bankteller"problemused by Tversky dency,considerthe now-famous and Kahneman. Experimental subjects were given a descriptionof a with a number of characteristics thatappeared woman,Linda, representative of someone who is a feminist. werethenaskedwhether it was Subjects morelikely thatLindawas (a) a banktelleror (b) a banktelleractivein the feministmovement.127 chose b,128 a choice Nearly90%of respondents thatis logicallyimpossible becauseeverypersondescribed by choiceb is also described the reverseis not true.129 by choicea, although Subjects ignoredthe baserate(therearemorebanktellersthanfeministbanktellers) becausethe description of Lindaappeared more "representative" of the latter thanof theformer.130 The pervasiveness of the representativeness heuristic can helpjustify a set of rules of evidencelaw frustrating to rationalchoice theory.The Federal Rules of Evidence provide that "characterevidence" is inadmissible in a trialto provethe truthof the matter unlessone asserted,
126. See Amos Tversky & Daniel Kahneman, Judgments of and by Representativeness, in

JUDGMENT UNDERUNCERTAINTY, supra note 125, at 84.

127. See id. at 92-93. 128. See id. at 93. 129. More specifically, this is an example of the conjunctionfallacy. The ConjunctionRule says, "The probabilityof some event A occurringcannot be less than the probabilityof A and some other event B both occurring."EDWARD GOOD REASON: THERATIONALITY DEBATE IN STEIN,WITHOUT PHILOSOPHY AND COGNITIVESCIENCE6-7 (1996). As a furtherexample, suppose that I tell you that there is a 50% chance that it will rain and be at least 70 degrees Fahrenheittomorrow. What is the probabilitythatit will raintomorrow?It cannotbe less than 50%. 130. Consideranotherfamous example of how individualsoften fail to adequatelyconsider "base rates"when makingprobabilitypredictions,providedby Kahnemanand Tversky. Supposethat 85% of the cabs in town are green and 15%are blue. A cab is involved in a hit-and-run accident,but the victim did not see the color of the cab. An eyewitness says that the offending cab was blue. How much credence should we give to the eyewitness's account?Suppose that the court, in an attemptto probethe worthiness of the eyewitness, creates an experiment in which he is shown a number of cabs under exactly the same conditions as those prevailingat the time of the accident and that he gets the color of the cab correct80% of the time. When asked, "Whatis the likelihood of the offending cab's being blue, given that the eyewitness said it was blue?", most people estimate it to be 80% likely. See Amos UNDER Tversky & Daniel Kahneman,EvidentialImpact of Base Rates, in JUDGMENT UNCERTAINTY, supra note 125, at 151, 156-58. But that is wrong. In point of fact, the probabilitythat the cab is blue, given that the witness says that it is blue, is only about 40%. To see why, suppose that the eye witness had been shown 100 cabs of which 85 were green and 15 were blue. Of the 85 green cabs, he would have correctly identified 80% of them, or 68, as being green, but he would have incorrectlyidentified 17 of the green cabs as being blue. Of the 15 blue cabs, he would have correctlyidentified 12 of them as blue, but he would have incorrectlyidentified 3 of the blue cabs as being green. All togetherhe said that 29 of the cabs that he saw were blue (17 green cabs mistakenlyidentified as blue plus 12 blue cabs correctly identified). Of those 29, only 12 were actually blue. The probabilitythat the offending cab was blue, given that the eyewitness said that it was blue, is 12/29 or approximately40%. Because most people do not take accountof the base rate in makingprobabilityestimates,they are proneto the sort of errorof this example: here, they give twice as much credenceto the eyewitness as they ought to.

2000]

LAW AND BEHAVIORAL SCIENCE

1087

of a seriesof exceptionsto the rule applies.131For example,a prosecutor evidencethata criminal defendant withmurder charged maynot introduce for armed has a previous recordof conviction In most circumrobbery.132 cannoteven introduceevidence that a defendant stances, a prosecutor withmurder hasa previous murder conviction.133 charged seek the positiverelationship that to maximize beAssuming jurors tweentheirverdictanda criminal defendant's this of guilt, type character evidenceis quiteusefulin enabling to maximize theirutility.Armed jurors robbers aremorelikelythanthe average to commita murder, so the person character evidenceis quiterelevant. Therational should consider the juror in thepopulation baserateof murderers thatprob(quitesmall)andupdate that the particular defendant has been conability with the information victed of armedrobbery.Preventingthe presentation of the character evidencereducesthe likelihoodthata rational jurorwill reachthe approconclusion about thedefendant's priate guilt. The law correctlyexcludescharacter evidence from consideration, however,if jurorsarelikely to ignorethe base rate-that is, to ignorethe factthatmostarmed robbers arenot murderers-and basetheirconclusion aboutthe defendant's on whether or not abouthim guilt specificfeatures look like stereotypical features of a murderer. Usingtherepresentativeness are to conclude that because thedefendant has heuristic, many jurors likely the appearance of a criminal he there(in thathe has a felonyconviction), foremusthavecommitted thecrimeforwhichhe is charged. Whenactorsoverestimate the relevance of salientor memorable incidents at the expense of base rates,they employ the "availability heurisThis mentalshortcut tic."'34 can often lead to estimatesthatapproximate statistical as memorable eventscan be memorable probabilities, precisely becausetheyarecommon.135 for the sakeof precision, howUnfortunately eventscan also be memorable for reasons ever,memorable havingnothing to do with theirgeneralprevalence-for example,becausethey arevivid, well publicized, or moreprevalent actor'sfriendsand amonga particular
131. FED.R. EvID.404(a). 132. See, e.g., 1 CHRISTOPHER B. MULLER & LAIRD C. KIRKPATRICK, FEDERAL EVIDENCE 536-37 (2d ed. 1994) ("[T]he basic rule of exclusion is of fundamental importance. It implements the philosophy that a defendantshould not be convicted because he is an unsavoryperson, nor because of crime charged."). past misdeeds, but only because of his guilt of the particular 133. See id. at 539 n. 1. 134. See Amos Tversky & Daniel Kahneman,Availability: A Heuristicfor Judging Frequency and Probability, in JUDGMENT UNDER UNCERTAINTY, supra note 125, at 163, 164 ("A person is said to employ the availability heuristic whenever he estimates frequency or probability by the ease with which instancesor associationscould be broughtto mind."). 135. Note that, so long as available incidents are representative of base rates, relying on available anecdotes ratherthan statistical probabilitieswill not lead to sub-optimaldecision making. See, e.g., Posner, supra note 69, at 1572 ("It is entirely rationalfor people to rely on anecdotal evidence in the absence of betterevidence ....").

1088

CALIFORNIA LAWREVIEW

[Vol.88:1051

acquaintances.Actors often estimate these available events as being much more common thanthey actuallyare. For example, most people believe that words beginning with the letter "k"are more prevalentthan words in which "k"is the thirdletter, although the latter set of words is actually twice as large as the former.136 Presumably, this is because it is easier to bring to mind words that begin with the letter "k" (such as "knave"and "kite") than those that have "k" as their third letter (such as "ark"and "ankle").137 Similarly, most people believe that homicides and car accidents kill more Americans than diabetes and stomach cancer, presumablybecause of the greater media coverage provided to the formertwo, althoughthe two diseases kill far more people.'38 Edward McCaffery hypothesizes that most states favor sales taxes over income and propertytaxes even though income and propertytaxes have a lower net cost to residents (they are deductible from federal income tax, while sales taxes are not) in part because the latter taxes are more prominent to taxpayersthanare the former.139 That people are biased by availability can have a widespreadimpact on legal policy in a variety of specific situations. Consider the potential impact of the availabilityheuristic on the policy prescriptionsthat rational choice theory provides for deterringcrime. According to conventionalrational choice analysis, potential criminals maximize their utility by committing crimes only if the expected benefits exceed the expected costs.'40 The expected costs of crime are determinedby multiplyingthe (monetized) severity of punishmentby the likelihood that the criminal will be arrested and convicted.'4 If criminals are rational utility maximizers, society can deter crime by raising these expected costs above the expected benefits of crime. And society can achieve this end with equal success by increasing the severity of punishmentor the frequencyof arrest,conviction, and punishment. Because increasingseverity (that is, lengtheningjail terms or imposing monetaryfines) is generally thought to be cheaper than increasing frequency (that is, increasing the number of police, prosecutors, and

136. See Tversky& Kahneman, supra note 134, at 166-68. 137. See Jolls et al., supra note 13, at 1518 (describing a study in which subjects guessed that there are more words ending with "ing"thanwords in which the second-to-lastletteris "n"). 138. See PLOUS, supra note 34, at 121-22. 139. See Edward J. McCaffery, Cognitive Theory and Tax, 41 UCLA L. REV. 1861, 1901-04 (1994). 140. See, e.g., COOTER & ULEN,supra note 8, at 436-39; Gary S. Becker, Crime and Punishment: An EconomicApproach, 76 J. POL.ECON. 169 (1968); Steven Shavell, CriminalLaw and the OptimalUse of NonmonetarySanctionsas a Deterrent, 85 COLUM. L. REV.1232, 1235 (1985). 141. See COOTER & ULEN,supra note 8, at 447 ("When the probability of punishment is multipliedby its severity, the resultis the expected punishment.").

2000]

LAW AND BEHAVIORAL SCIENCE

1089

judges), it is generally assumed that increasing severity is the more efficient deterrencestrategy.142 If criminals are biased by availability when calculating (no doubt in an informal sort of way) the anticipatedcosts of crime, this analysis could change radically depending on what types of events are more salient to potentialcriminals.In orderto determinewhich deterrencemechanismwill be most efficient, policymakersneed to understandwhether criminals are likely to over- or underestimatethe frequency and the severity of punishment that is actually meted out. If punishments are so severe that some sentences become shocking and well publicized, increasing severity could be the more efficient deterrence strategy. Otherwise, increasing the frequency of punishmentis likely to be more efficient, underthe assumption that if a criminalknows or knows of someone who has been imprisonedfor a particularcrime, this informationis likely to be available and to cause him to overestimatethe likelihood that he will be arrestedand convicted if he commits the same crime. When allocating resources, policymakers should also be concerned with what types of law enforcementefforts are most available (thatis, visible) to potentialcriminals.The sharpoverall decline in crime in New York City in the 1990s is often creditedto police crackdownson "publicorder" offenses that are highly visible.'43Such observable evidence of law enforcement activity might affect the cost-benefit calculations of would-be lawbreakerscontemplatingunrelatedcriminal acts by causing them to believe the frequency of apprehensionis greater than it actually is.'44The point, generally, is that criminals may not use objectively verifiable evidence of frequency and severity, as rational choice theory implicitly assumes thatthey do.145 The availability heuristic also sheds light on a long-runningcontroversy in the medical community over the use of clinical practice guidelines, which, based on statistical evidence, instruct physicians on how to treat patients who exhibit particular sets of symptoms.l46 Defenders of
142. See, e.g., POSNER, supra note 26, at 244; Dan M. Kahan,Social Influence, Social Meaning, and Deterrence, 83 VA. L. REV.349, 351-52, 378 (1997) (calling this a "foundationalinsight of the standard economic conception"). 143. See Kahan,supra note 142, at 368-69. 144. See, e.g., Robert J. Sampson & Jacqueline Cohen, Deterrent Effects of the Police on Crime: A Replication and Theoretical Extension, 22 L. & Soc'Y REV. 163 (1988) (finding that increasedpolice attentionto disorderlyconduct offenses is correlatedwith a decreasein robberyrates). 145. There is empirical evidence that criminals systematically underestimatethe likelihood of their being caught, convicted, and punished. See, e.g., James Q. Wilson & Alan Abrahamse, Does CrimePay?, 9 JUST. Q. 359, 373 tbl.10 (1992). 146. See, e.g., Mark Kadzielski et al., Peer Review and Practice Guidelines Under Health Care L. REV. 157, 158 (1995) (Proponentsof guidelines argue that they reduce health Reform, 16 WHITTIER care spending, serve as a basis for defining a basic package of health care benefits, protectphysicians in malpracticelitigation, help physicians and patients make better health care decisions, and increase the quality of patient care. Physicians, on the other hand, fear that guidelines will become rigid

1090

LAWREVIEW CALIFORNIA

[Vol.88:1051

decision-making autonomy in the medical profession often argue that doctors should be free to deviate from the prescriptionsof such guidelines based on their clinical experience. After all, each patient is unique, and treatmentsspecified in practice guidelines, derived from base rates of the effectiveness of various treatments,may not best serve every patient. Although this argumentis hardto dispute in the abstract,the availabilityheuristic suggests that deviations from practice guidelines by individual doctors are more likely to be based on the experiences of the physician than on the unique characteristicsof the patient.That is, a doctor who has treatment successfully used an "alternative" regime in the past and found it successful is highly likely to ignore the practiceguidelines, althoughthose guidelines, based on statisticalevidence, will be effective more often than the vivid but anecdotal data derived from a particularphysician's experience. One policy implication of this might be for the legal system to privitreatment that follows practiceguidelines. As long as each humanbelege is ing unique, deviations from practiceguidelines will be desirablein some thatphysicians adhereto such guidelines cases, so regulatoryrequirements would likely be an unfortunate overreactionto the problem.But it might be sensible to require physicians who deviate from guidelines to document their reasons for doing so, thus ensuring that such deviations are at least carefully considered. Alternatively, tort law could provide that adherence to practice guidelines by a physician creates a rebuttablepresumptionof nonnegligence in malpracticelawsuits or even an affirmativedefense to a chargeof malpractice. Finally, an understandingof the availability heuristic can provide a caution to policymakers tempted to enact new regulatory regimes in response to highly available informationconcerning health or safety risks. Jolls, Sunstein, and Thaler argue persuasively that the enactment of "Superfund" legislation in 1980 was largely a result of publicity concerning the coverage of environmental contamination in Love Canal, New York, beginning in 1978.147 It is at least arguablethat this attemptto solve the problemof toxic waste dumpsdivertedscarce public and congressional attention from far more dangerous and pervasive environmental health risks.

standards used for reimbursement and for other legal purposes, will impede advances in new technology and will quickly become outdated. Physicians also criticize the fact that guidelines are developed by academics who have no real world experience in medicine.); cf Thomas M. Burton,An HMO Checks Up on its Doctors' Care and is Disturbed Itself, WALL ST. J., July 8, 1998, at Al (reportingthat an HMO's internal audit found that doctors routinely fail to follow accepted clinical guidelines). 147. See Jolls et al., supranote 13, at 1520-22; see also Kuran& Sunstein, supra note 125, at 69198 (arguingthatthe publicity surrounding Love Canalmay have createdan "availability error").

2000]

LAWAND BEHAVIORAL SCIENCE

1091

2. Overconfidence and Self-ServingBiases

of a parEven when actorsknow the actualprobability distribution ticularevent, their predictionsas to the likelihoodthat that event will bias": the belief happento them are susceptibleto the "overconfidence that good things are more likely than averageto happento us and bad to happen to us.148 thata thingsareless likely thanaverage Demonstrating individual is overconfident is difficultto do, becausethe indiparticular vidualmightwell differfromthe statistically averagepersonin positive For a student who believes he will scoreabovethe mean ways. example, on an exam mightbe overconfident of his ability.On the otherhand,he be smarter a or more studier thanhis peers,making the premight diligent dictionquitereasonable. But the pervasiveness of the overconfidence bias has been demonstrated in experiments that elicit opinions persuasively fromall members of a groupabouthow theyarelikelyto compare to other members of the group.149 Garrison Keillor's all the Notwithstanding report, children in LakeWobegon cannot reallybe aboveaverage.'50 Theoverconfidence biasis well demonstrated who by Neil Weinstein, askeda sampleof students at Rutgers to estimate whether or not University eachof a seriesof eventswas moreor less likely (andhow muchmoreor less likely)to happen to themthanto theirclassmates.15' Of eighteenpositive events,rangingfromowningtheirown home to avoidinga hospital the nextfive years,the meanrespondent estimated his chances stayduring to be greater thanthe averagefor his peers-that is, the othersurveyreIn contrast, on average bespondents-in fifteenof the events.152 subjects lievedthatnegative fromsuffering a divorce to losinga job events,ranging to contracting to themthanto their lungcancer,wereless likelyto happen events.153 peers in twenty-twoof twenty-four Similarly,in a survey of license, BakerandEmery Virginiaresidentswho appliedfor a marriage foundthat,although mostrespondents knewthatclose to half of all marthe likelihood thattheirmarriagesend in divorce,whenaskedto predict themodalresponse waszero.154 riagewouldendin divorce,
148. See generally ChristineJolls, Behavioral Economics Analysis of RedistributiveLegal Rules, 51 VAND.L. REV. 1653, 1659 & n.22 (1998) (noting that nearly two hundred studies support this descriptiveclaim). 149. See generally sources cited in Neil D. Weinstein, Unrealistic OptimismAbout Future Life & SOC. PSYCHOL. Events,39 J. PERSONALITY 806, 806 (1980). 150. GARRISON LEAVING xvii (1987) ("'That's the news from Lake Wobegon, HOME KEILLOR, where all the women are strong, the men are good-looking, and all the children are above average'...."). 151. See Weinstein, supra note 149, at 809. 152. See id. at 810 tbl.1. 153. See id. 154. See Lynn A. Baker & Robert E. Emery, When Every Relationship Is Above Average: Perceptions and Expectationsof Divorce at the Timeof Marriage, 17 LAW& HUM.BEHAV. 439 (1993).

1092

CALIFORNIA LAWREVIEW

[Vol.88:1051

The overconfidence bias could have a wide-rangingimpact on deterrence policy in a variety of areas of law. Policymakersrarelywish to deter 100% of even undesirable conduct, because the costs of doing so would For any type of conduct that the state wishes to dislikely be too great.155 courage, from criminalbehaviorto carelessness likely to lead to a tort, rational choice theory advises policymakers to set the penalty for the undesirableconduct such that the desired fraction of the population-say, potential injurers-will calculate that the expected costs of the conduct exceed the expected benefits to them. Where the targets of such policies exhibit overconfidence, however, policymakerswill have to set the penalties higher, sometimes substantiallyso, than they would in a world of utility-maximizing actors who are not systematically overconfident. If bank robbersbelieve that they are less likely to be apprehended than theirpeers, if absentmindeddrivers believe they are less likely to cause an accident than other drivers, or if some physicians believe they are less likely to be found liable for malpracticethan other physicians, penalties for the undesirablebehaviorwill have to be higher than policymakerswould otherwise thinknecessaryto achieve the desiredlevel of deterrence.156 For policymakersto be able to make effective use of the insights provided by the overconfidence bias, more empirical research needs to be done on which groups and in what situationsoverconfidenceis likely to be most severe.'57 Currently,one useful conclusion can be drawnfrom the literature: at least for events perceived to be negative, actors apparentlytend to be more overconfident when the event in question is perceived to be controllablethan when it is perceived not to be controllable.58 This finding suggests that an understanding of the overconfidence bias is, in fact,

155. See COOTER & ULEN,supra note 8, at 21-24 (demonstratingthat social optimality occurs where social marginalbenefit and social marginalcost are equal and that this rarely occurs at a zero quantityof a bad thing). 156. See Atul Gawande, WhyDoctors Make Mistakes,THENEWYORKER, February1, 1999, at 40, 40-52 (arguing that most medical mistakes, even those by very good physicians, are simply effect oversights, and that exposure to malpracticeliability does not have an additionalprecautionary on physicians). 157. One plausible,but as yet unproven,hypothesisis thatolder people are less overconfidentthan A. POSNER, AGING ANDOLDAGE 104 (1995). Available data does younger people. See RICHARD serve as subjects in suggest, however, that the bias is not limited to the young, who disproportionately psychology experiments. See, e.g., Andrew Guppy, Subjective Probability of Accident and ANALYSIS & Apprehensionin Relation to Self-OtherBias, Age, and ReportedBehavior, 25 ACCIDENT PREVENTION 375, 377-78 & tbl.1 (1993); Neil D. Weinstein, Unrealistic Optimism About Susceptibility to Health Problems: Conclusionsfrom a Community-Wide MED.481, 487-89 Sample, 10 J. BEHAV. (1987). 158. See Weinstein, supra note 149, at 814. Weinstein concluded that, when subjectsperceived an event to be controllable, they tended to compare themselves with the stereotypical victim of the negative event, leading to overconfident predictions. In contrast, when events were perceived as uncontrollable,subjects did not perceive a stereotype of a victim with whom to compare themselves. See id.

2000]

LAWAND BEHAVIORAL SCIENCE

1093

particularlynecessary to develop optimal deterrencepolicies, since such policies are targetedat controllableevents. Related to the overconfidence bias is the "confirmatory"or "selfserving" bias, the term to describe the observation that actors often interpretinformationin ways that serve their interestsor preconceivednotions. For example, Lord and his colleagues found that when experimental subjects were given factual evidence about the effects of the death penalty, subjects identified as proponentsof capital punishmentsaid the evidence reinforced their prior beliefs, while subjects identified as opponents of capital punishment said that the information reinforced their prior beliefs.'59 In a series of papers, Loewenstein and his colleagues'60 and Babcock and her colleagues'6'found a similar effect of informationin the litigation A group of law student subjects were provided with factual incontext.'62 formation about a dispute in litigation. Despite being given identical information, subjects who were told to imagine that they were the attorney representingthe plaintiff interpretedthe facts as favorable to the plaintiff, while subjects told to imagine that they.were the attorneyrepresentingthe defendantinterpreted the facts as favorableto the defendant.163 The conventionallaw-and-economicsapproachto trial and settlement, based on rational choice theory, predicts that, because trials are more costly than out-of-courtsettlement, lawsuits will settle out of court unless the parties have substantiallydifferent predictions about the likely results of trial.'6 Plaintiffs and defendants may reach different predictions about the likely outcome of a trial, but differences in predictionsare presumedto be in both directions-that is, where the predictions of plaintiffs and defendants diverge, half of the time plaintiffs will believe their prospects are
159. See Charles G. Lord et al., Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on SubsequentlyConsidered Evidence, 37 J. PERSONALITY & Soc. PSYCHOL. 2098, 2102 (1979). 160. See generally George Loewenstein et al, Self-ServingAssessments of Fairness and Pretrial STUD.135 (1993). Bargaining, 22 J. LEGAL 161. See generally Linda Babcock et al., Biased Judgments of Fairness in Bargaining, 85 AM. ECON. REV. 1337 (1995) [hereinafter Babcock et al., Biased Judgments]; see also LindaBabcock et al., Choosing the WrongPond: Social Comparisonsin Negotiations that Reflect a Self-ServingBias, 111 1 (1996) [hereinafter Babcock et al., Choosing the WrongPond]. Q. J. ECON. 162. See also Jolls et al., supra note 13, at 1503 (discussing the Loewenstein and Babcock studies). 163. See Loewenstein et al., supra note 160, at 151-52; Babcock et al., Biased Judgments,supra note 161, at 1340. Investigatorshave identified a closely relatedeffect called "cognitive dissonance." This is a form of selective perception in which actors give greater weight to evidence that confirms beliefs they alreadyhold and lesser weight to contradictory evidence. See LEON A THEORY FESTINGER, OFCOGNITIVE DISSONANCE (1957); PLOUS, supra note 34, at 22-30. 164. See COOTER & ULEN,supra note 8, at 377-84 and 390-94; George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. LEGAL STUD. 1, 12 (1984). For an excellent review of the literature, see Robert D. Cooter & Daniel Rubinfeld, Economic Analysis of Legal LITERATURE 1067 (1989). Disputes and TheirResolution, 3 J. ECON.

1094

LAWREVIEW CALIFORNIA

[Vol.88:1051

An implication worse thandefendants of this conventional anticipate.165 will be less likely whenthe parties accountis thatbecauselitigation have of the likelihoodof prevailing, moreaccurate estimates that anything imor betterlegal suchas expanded discovery provesthoseestimates, pretrial is to be favored. representation, Evidenceof the self-serving biasin the analysisof lawsuitssuggests, in contrast, thatplaintiffs(anddefendants) will systematically anticipate as beingbetter thandefendants theirtrialprospects believe. (andplaintiffs) The consequence of this is moretrialsthanwouldbe predicted by the rationalchoicemodel,unlessstepsaretakento mitigate the parties'evaluative biases.Moreinformation-provided, in theformof expanded perhaps, to be effective as we have seen in because, pretrial discovery-is unlikely the experiments cited aboveconcerning the deathpenaltycontroversy,166 evidenceto solidifytheirviews, rather than peopleseemto use additional to alterthem. thepolicyimplications of thisarenotobvious,a strong case Although can be madethatevidenceof the self-serving biasprovides for lesupport that the structures to view of a facts gal require litigatingparties dispute the eyes of theiropponents. The moder trendin civil litigation through towardmandatory settlementconferences,167 court-ordered mediation,'68 andnonbinding arbitration169 makelittle sense froma rational choice perunder which are to make (accurate) spective, parties presumed maximizing decisionsaboutwhetherto settleor even whether to negotiatewith their In a rational adversaries. choiceworld,mandated interaction wouldmerely increasetransaction costs for no useful purpose.But these requirements seemquitedefensible if the self-serving biasis present andpervasive. The implications of self-servingbias for reexamining conventional law-and-economics wisdomof legalrulespotentially farbeyondthe carries realmof litigation versussettlement. Forexample,thelaw-and-economics law is basedlargelyon the assumption of the Coase theoryof property Theorem thatproperty will be traded when the holder of sucha right rights
165. See Priest & Klein, supra note 164, at 10-11 & fig.3; Babcock et al., Biased Judgments, supra note 161, at 1337. 166. See supra note 159 and accompanyingtext. 167. See, e.g., FLA. STAT. ANN. ? 766.108 (West 1987) (requiring a mandatory settlement conference in all medical malpracticeactions); CAL.R. CT. 222 (West 1996) (requiringa mandatory settlement conference in all "long cause matters"); HAW.ST. USDCT Civ., RULE235 (Michie 1995) (requiringa mandatorysettlementconference in every civil action); MICH.RULE3 CIR.,RULE2.401 settlementconferencein all civil actions). (West 1998) (requiringa mandatory 168. See, e.g., U.S. DIST.CT. RULES E.D. PA., CIVIL RULE53.2.1 (West 1998) (orderingall oddnumberedcases to participatein an experimentalmediationprogram). 169. See, e.g., FLA.STAT.ANN. ? 44.103 (1993) (allowing courts to refer certain civil actions to nonbinding arbitration);HAW.REV. STAT.? 601-20 (1998) (establishing a program of mandatory nonbinding arbitration for all civil matters under $150,000); NEV. REV. STAT. ? 38.250 (1999) for all civil actions under$40,000). (requiringmandatory nonbindingarbitration

2000]

LAW ANDBEHAVIORAL SCIENCE

1095

values it less thandoes another costs.'70 actor,assuminglow transaction that are to estimate Butthe self-serving bias suggests individuals likely the value of propertyrights differentlydependingon which side of the transaction they find themselves.This could resultin rightsbeing more "sticky"(resistantto exchange)thanrationalchoice theorywould predict-a pointto which we will return in a slightlydifferentguise in our effect"andtherelated discussion of the"endowment "status quobias."71 An additional comes in the context,wheremanagexample corporate ers withundueconfidence in theirfirms'abilityto overcome obstacles and a self-serving of information that perception mightobjectivelysignalfutureproblems could potentially misleadthose who wouldinvestin their firms' securities.DonaldLangevoort has suggestedthat this likelihood a for securities law rules, such as "duediligence" provides justification for and accountants under the Securities Act of 1933, requirements lawyers thatrequire thirdparties who arepotentially less likelyto sufferfromsuch biasesto verifythetruthfulness of information thatthefirmprovides to the
marketplace.172 3. HindsightBias

the most-studied in probabilistic assessment is Perhaps shortcoming the "hindsight the tendency of actorsto overbias,"thetermthatdescribes estimatethe ex ante prediction thatthey hadconcerning the likelihoodof an event'soccurrence afterlearning thatit actually did occur.173 In whatis arguablythe most famous of the many hindsightbias studies, Baruch Fischhoffgave five groupsof subjectsa passageto readdescribing the eventsleadingup to a military confrontation betweenthe Britisharmyand the Gurkas in Nepalin the nineteenth andaskedthem,on thebasis century of that information alone, to specify the likelihoodthat four specified outcomes wouldhaveresulted.74 Eachof fourgroups wastoldthat military a different outcome of thefourspecified outcomes while occurred, actually the fifth group(thecontrolgroup)was given no information on the actual in eachof the groupsto whomthe investigators outcome.'75 Subjects gave an outcome anex anteprediction of thatoutcome thatwas considreported for thatoutcomemadeby the subjects in erablyhigherthanthe prediction
170. See generally RonaldH. Coase, TheProblemof Social Cost, 3 J. L. & ECON. 1 (1960). 171. See infra PartIII.B. 172. See Langevoort,supra note 64, at 158-59. 173. For a review of well over 100 studies of the hindsightbias, see Jay J.J. Christensen-Szalanski & CynthiaFobian Willham, TheHindsightBias: A Meta-Analysis,48 ORG. BEHAV. & HUM. DECISION
PROCESSES 147 (1991).

174. (1975). 175.

Baruch Fischhoff, Hindsight - Foresight: The Effect of Outcome Knowledge on Judgment See id. at 289.

Under Uncertainty, 1 J. EXPERIMENTAL PSYCHOL.:HUM. PERCEPTION & PERFORMANCE 288, 289-90

1096

CALIFORNIA LAW REVIEW

[Vol.88:1051

the controlgroup.'76 In otherwords,information aboutwhatactuallyoccurredapparently influencesourjudgments whatwe thought concerning would occurbeforewe knew the outcome.Eventsthathave actuallyoccurred can seem,through the lens of hindsight, to havebeenalmostinevitable. KaminandRachlinski demonstrated the effect thatthe hindsight bias canhaveon theassignment of tortliability.'77 basedon Usinga factpattern asked a group of subjectsto play the role of jurors and to determine whether a company a bridgehadbeennegligentin its failureto operating takeprecautions thatwouldhaveprevented flood damage to thirdparties.
to They instructedsubjects to use the famous "LearnedHand formula"'79 the famous case, In re: Kinsman Transit Company,'7the experimenters

determine or not an act was negligent-that is, to assessliability whether if the costs of to thebridgecompany wouldhavebeenless only precaution thanthe expectedcosts of flood damageto thirdparties(giventhe uncerthatwouldresultfromnot takingtheprecautaintyof a flood's occurring) While 57% of the juror subjectswould have found the bridge tion.l80 in a control company negligentunderthis standard, only 24%of subjects who hadthe sameinformation condition, exceptthatthe flood hadin fact notoccurred, believedthatthecost of precaution wouldhavebeenjustified the risks of not the Traditional law andeconomics by taking precautions.'81 that is efficientwhen its benefitsoutweighits costs suggests precaution andthattortlaw shouldimposeliabilityfor negligence whenan actorfails to takeefficientprecaution butnot whenhe fails to takeinefficient precaution.'82 But if jurorsare subjectto the hindsight will be bias, defendants foundnegligentin situationsin which they acted in a socially efficient manner butwerestruck (failedto takeinefficient precautions) by badluck.

176. See id. at 290. The experimentis discussed, among other places, in Rabin, supra note 120, at 30, and Jeffrey J. Rachlinski,A Positive Psychological Theoryof Judging in Hindsight, 65 U. CHI. L. REV.571, 576 (1998). 177. See Kim A. Kamin & Jeffrey J. Rachlinski, Ex Post - Ex Ante: Determining Liability in Hindsight, 19 LAW & HUM. BEHAV. 89 (1995); see also Susan J. LaBine & Gary LaBine, Determinationsof Negligence and the Hindsight Bias, 20 LAW.& HUM.BEHAV. 501 (1996) (finding that a large minorityof subjectsplaying the role of jurors would find a psychiatristnegligent for failing to prevent a patient's violence, even when the psychiatrist followed a widely approved standardof behaviorand the violent act was unpredictable). 178. 338 F.2d 708 (2d Cir. 1964), cert. denied, 380 U.S. 944 (1965). 179. The formula was made famous in Judge Hand's opinion in United States. v. Carroll Towing Co., 159 F.2d 169 (2d Cir. 1947). "[I]nalgebraicterms: if the probabilitybe called P; the injuryL; and the burdenB; liability depends upon whether B is less than L multiplied by P: i.e., whetherB is less thanPL." Id. at 173. 180. See Kamin& Rachlinski,supra note 177, at 96. 181. See id. at 98. 182. See, e.g., COOTER & ULEN,supra note 8, at 313-16; LANDES & POSNER, supra note 8, at 5860; POSNER, supra note 26, at 180-83; SHAVELL, supra note 8, at 78.

2000]

LAW AND BEHAVIORAL SCIENCE

1097

This, in turn,could provideactorswith a privateincentiveto take an of precaution. inefficiently highamount As is truewithmanyof the deviations fromthe predictions of rational choicetheory, thelegalimplications of thisbehavioral arenot phenomenon clearcut.If juryinstructions couldobviatethe hindsight bias,sucha stratin a varietyof situations in whicheffiegy wouldclearlybe appropriate cient deterrencerequiresjurors to impose liability based on ex ante aboutthelikelihood of certain events'occurring. judgments Unfortunately, haveyet to find a methodof eliminating the hindsight bias, psychologists andevenreducing its effecthasproven difficult.183 quite Some analystshave suggestedthatthe bias mightbe countered by of proofnecessary to find a defendant raisingthe standard negligent-for of the evidence"standard to example,fromthe existing"preponderance one of "clearandconvincing Jeff Rachlinski has argued evidence."'84 that this is one way to understand the businessjudgmentrule in corporate law:'85 corporateofficers and directorsare held liable for, in effect, theircompaniesonly when thereis evidenceof "negligently" operating Theproblem withthisapproach, of course,is its lack "gross negligence."186 of precision.If the hindsight bias is strong,raisingthe standard of proof not eliminate overdeterrence. the On other the standard hand,raising might of proof could swampthe bias, leadingto underdeterrence. This might happenif the hindsightbias effect, in the event of litigation,is not so under the "clear andconvincing strongthatliabilityis difficultto establish evidence"standard takefewerprecauand,as a result,potentialinjurers tions thanthey oughtto becausethe likelihoodof being held liable has fallensignificantly. Jolls, Sunstein,and Thalerhave suggestedthat the bias might be avoidedby shielding whatactionthe dejuriesfromevidenceconcerning fendantactuallytook until afterjurorshave determined what decision would have been reasonable ex ante.'87 To illustrate how this mightwork, the authorsgive the exampleof a food processingcompanythatmight have subjected its customers to contaminated food if it chose not to use a certain and themto carcinogenic chemipreservative mighthavesubjected cals if it choseto use thepreservative. Jurors wouldthenbe askedto assess
183. See Rachlinski, supra note 176, at 603 (concluding that any procedurethat might mitigate againstthe hindsightbias is "so intrusive[thatit] may not be suitablefor the courtroom"). 184. See id. at 606; Jolls et al, supra note 13, at 1529-32. 185. Cf Rachlinski, supra note 176, at 574 (suggesting that the business judgment rule can be viewed as a rule of "no liability," which might have fewer consequences than the alternative of negligencejudged in hindsight). 186. See, e.g., Aronson v. Lewis, 473 A.2d 805 (Del. 1984) (stating the "gross negligence" standard).Of course, there are other ways of justifying this rule. For instance, a standardargumentis that shareholderswould want managersto have wide discretionto run the company without having to face repeatedchallenges to their decisions on a close standard like "preponderance of the evidence." 187. See Jolls et al., supra note 13, at 1527-29.

1098

LAW REVIEW CALIFORNIA

[Vol. 88:1051

the costs and benefitsof using and not using the preservative, without the or what choice defendant made the knowing consequencesof that
choice.188

Thisis a creative butit canworkonly in the limitednumber solution, of cases in whichit is plausible thatbothactingandnot actingcouldsubto a lawsuit.Butmany,if notmost,casesdo not involve ject the defendant defendants who wereforcedto choosebetweenpotential torts.Rather, the choicemostdefendants or notto takea precaution face is whether thatwill reducethe riskof an accident or injury. In this typeof case,hindsight bias will be unavoidable becauseit will be clearto jurorsfromthe fact thata lawsuithas beenfiled anda jury impaneled thata particular type of accidentoccurred as creativeas the JollsTherefore, (or allegedlyoccurred). Sunstein-Thaler it not does as a generalsoluseem is, proposal practicable tion to the problem thatthe hindsight bias createsin assessingnegligence liability. we believethatthe presence of the hindsight bias arguesfor Instead, two otherreforms of accidentregulation, one withinthe realmof private tortlaw, the otherwithinpubliclaw. The firstis the wideruse of strictlia morerestricted abilityand,consequently, scopefor negligenceliability. The standard, rational-choice-based law-and-economics analysisof the selectionof a liabilitystandard focuses on two aspectsof accidents:the of and the betweentheunderlying risktechnology precaution relationship and the of harm. When the of creating activity scope technology precaution is bilateral victimandthe potential (in the sensethatboththe potential inactionthatwill reducethe probability or sejurercan takeprecautionary thenlaw andeconomicsarguesfor the negligence verityof an accident), standard.189 in thetechnology of precaution is unilateral When, contrast, (in the sense thatonly the potential can realistically takeactionto reinjurer ducethe probability or severityof an accident), thenstrictliabilityis more 9 Finally,where,in additionto the precaution efficientthannegligence.
188. See id. at 1527-28. 189. This argument is a good example of the thrust of rational choice theory in the economic analysis of law. A rationalpotentialinjurerwill reason that, if complying with the due care standardis cheaperthanexpected liability costs, he should comply with the legal duty and therebyescape liability. A rationalpotentialvictim will reason that if she is injuredin an accident, she will surely be injuredby a rationalinjurer,who will, no doubt, have reasonedthat he should comply with the legal duty of care and will thereforenot be liable for the victim's losses. That means that the potential victim must bear her own losses in the event of an accident. Faced with this "residualliability," the rationalpotential victim will take as much care as necessary to minimize her expected costs arising from an accident. The argumentalso works when decision makers are not certain ex ante an accident whether they will be an injureror a victim. The real insight of this way of thinking about the negligence standardis its recognition that the standardinduces both parties, not just the potential injurer, to take care. See COOTER & ULEN, supra note 8, at 304-11 for a basic explanationand 338-45 for extensions of the basic theory. 190. There is no point in imposing residualliability on the potentialvictim if there is nothing that she can do to reducethe probabilityor severity of an accident.See id. at 272-75.

2000]

LAW AND BEHAVIORAL SCIENCE

1099

or "activity level effect"(an inexternality exterality, thereis a quantity effect thatthe level of the underlying activityhas risk-creating dependent of an accident-for example,the moremiles one drives, on the likelihood the morelikely an accidentbecomes),strictliabilityis superior to negligence.191

The pervasiveevidenceof the hindsight bias providesan additional for a strict The hindsight liabilitystandard. pragmatic argument selecting biascastsdoubton the abilityof juriesto reachproper determinegligence that could have nationsbecausejuries are likely to believe precautions been takenwould have been more cost-effectivethanthey actuallyapno similarimpediments to the pearedto be ex ante. But the bias presents of to reach determinations under a strict liability juries properliability where the need determine that the accident abilityregime, jury only alleged occurred andwascausedby thedefendant in order forliability to attach. The secondreformdesignedto overcomethe problems of hindsight bias in adjudication is to downplayex post adjudication as a methodof the socialcostsof accidents in favorof broader ex anteregulaminimizing tionof safetyby administrative Administrative has,of agencies. regulation a host of of its own: from course, problems inflexibility, political pressures interest bureaucratic andthe like.192 But, well-organized groups, bumbling, all other evidence of the bias constant, holding things hindsight strengthens the argument for protecting adminispublicsafetyprospectively (through trativeregulation) as opposedto retrospectively (through privatetortactions). Thepredictions of rational choicetheoryoftenleadto the conclusion andhigh transaction that,in the absenceof externalities costs, bargaining will achieveefficiency,andthatbureaucratic intervention into privateoris undesirable.'93 behavioral deviations fromthepredering Consequently, dictions of rationalchoice theory often provide a basis for plausible in favorof government activism.The hindsight bias,however, arguments also carriesthe seeds of an argument for morerestrained at government, leastin somecircumstances. Justas jurorscanbe subject to thisbiaswhen of makingnegligencedeterminations (so thatthey skew theirassessment ex anteprobabilities), can administrative decision makers fall so, too, prey

191. See Landes & Posner, supra note 8, at 70; Steven Shavell, Strict Liabilityversus Negligence, 9 J. LEGAL STUD.1 (1980). 192. For a useful overview of these and related problems, see STEPHEN REGULATION BREYER, ANDITSREFORM A. FARBER & PHILIP P. FRICKEY, LAWANDPUBLIC CHOICE: A (1982), and DANIEL CRITICAL INTRODUCTION (1991). 193. See COOTER & ULEN,supra note 8, at 85 ("Whentransactioncosts are zero, an efficient use of resources results from private bargaining, regardless of the legal assignment of... rights.");see generally Coase, supra note 170 (arguing that resources will inevitably move to their highest-valued use when transaction costs are trivial).

1100

CALIFORNIA LAWREVIEW

[Vol.88:1051

to the samebias. Thus,government officialsshouldbe cautiouswhenenincident hasoccurred.'94 actingnewregulatory regimesafteranundesirable For example,regulations sufficientto guarantee with nearcertainty thattherewill neverbe an airlinefatalityarealmostcertainly too costlyto be efficient.The strengthening of air safetyregulations followinga fatal disaster is justifiedif the circumstances the accident surrounding provided new information to the agencythatcausedit to readjust its cost-benefit But new regulations wouldnot be analysisabouta particular regulation. of sucha crashwas recognized ex ante justifiedif the statistical possibility and acceptedas justifiedundera cost-benefitanalysis.In otherwords, bias does not causethemto agenciesneedto be carefulthatthe hindsight conflatetheoccurrence of an unfortunate eventwiththe determination that cost-benefit were flawed. previous analyses
4. Anchoringand Adjustment

Researchon the phenomenon of anchoring and adjustment demonstrates thatprobabilistic assessments areoftenflawedbecauseactorsfail to theirassessments frompreexisting adjustsufficiently cognitiveanchors.195 In one demonstration of thisphenomenon, askedone group experimenters of professional accountants whether fraudocthey believedmanagement in morethanten companies curred out of eachthousand audited by major firmsandaskedanother whether accounting groupof accountants theybelieved fraudoccurred in morethantwo hundred of every thousand such Whenthenaskedto estimate the actualnumber of instances companies.196 of fraud the accountants who wereexposedto the perthousand companies, anchorof two hundred provided,on average,a significantlyhigherrewho were exposed to the anchorof ten.'97 sponse than the accountants Therewas no logicalreasonfor subjects to believethatthereference numberused in the firstquestionwas relatedto the correct answerto the second question. Nonetheless, the subjects apparentlyfailed to adjust theirestimate of thenumber of casesof fraud fromthe number sufficiently to whichtheywereoriginally exposed.198

194. This, of course, assumes that regulators' sole concern is to balance the costs of prevention against the costs of accidents. It is an interesting question whether regulators' cost-benefit analysis should "count"the utility that citizens might derive from knowing that their government has taken decisive (if perhapswrongheaded)action aftera tragic incident. 195. See Amos Tversky & Daniel Kahneman,Rational Choice and the Framing of Decisions, 59 J. Bus. S251, S251-54 (1986). 196. See EdwardJ. Joyce & GaryC. Biddle, Anchoringand Adjustment in Probabilistic Inference in Auditing, 19 J. ACCT. RES 120, 122-23 (1981). 197. See id. at 125. Subjects given the higher anchorestimated,on average, that fraudoccurredin more than 43 companies per 1000, while subjects given the lower anchor estimated, on average, that the numberwas just over 16. Id. 198. For otherexamples, see PLOUS, supra note 34, at 149-51.

2000]

LAWANDBEHAVIORAL SCIENCE

1101

In two articles,Korobkin and Guthrie demonstrated experimentally how the anchoring andadjustment bias can affectthe settlement of litigaTwo groupsof subjects wereaskedto play the role of the plaintiff tion.199 in a litigationscenarioand to choose between (a) acceptinga "final" settlement offer madeby the defendant and (b) optingfor a trial,with a chanceto recovermore money but a risk of recoveringnothing.20 One hadmadea relativelylow "initial" setgroupwas told thatthe defendant tlementoffer andthenraisedit substantially to reachthe final offer.The othergroupwas informed thatthe defendant had madea relativelyhigh initialofferandraisedit only slightlyto reachthe samefinaloffer.201 The actualbenefitof settlement andthe risk-adjusted benefit of trial expected were identicalfor subjectsin both groups.Subjectsin the formergroup, weresignificantly morelikelyto acceptthefinalsettlement however, offer, that their the value of the case had suggesting expectations concerning beenanchored at a lowerlevel thanthe expectations of subjects in the lattergroup.202 Whilethis findingdoes not obviouslysuggestspecificlegal policies, it does suggestan interesting andcounterintuitive for individimplication ual attorneys, and who others are in promotinterested mediators, judges, resolution. to rational choice which would ing dispute Contrary theory, that characterized initialoffersfollowedby suggest bargaining by extreme is inefficientbecauseof its high time-consuming positionalbargaining transaction costsandpotential fordelay,203 evidenceof anchoring effectsin suchcostsmightactually be an efficient bargaining suggeststhatincurring becausethesecostlyactivities increase the chances way to resolvedisputes of out-of-court settlements. science literature on anchoring and adjustAlthoughthe behavioral ment emphasizesthe way in which numbers can anchorthe probability of actors,it is not difficultto see how verbalcues could have prediction similar effectsin waysthatarerelevant to the law. Tortlaw, for example, manufacturers of potentially hazardous to avoidliability permits products

199. See Russell Korobkin& ChrisGuthrie,OpeningOffersand Out of CourtSettlement:A Little ModerationMight Not Go a Long Way, 10 OHIO ST. J. DISP.RES. 1 (1994) [hereinafterKorobkin& Guthrie, Opening Offers]; Russell Korobkin & Chris Guthrie, Psychological Barriers to Litigation Settlement: An Experimental L. REV.107, 139-42 (1994) [hereinafterKorobkin Approach, 93 MICH. & Guthrie,Psychological Barriers]. 200. Korobkin& Guthrie,Psychological Barriers,supra note 199, at 140. 201. See id. at 141. 202. See Korobkin& Guthrie,OpeningOffers, supra note 199, at 19. 203. Cf. Robert H. Gertner& Geoffrey P. Miller, SettlementEscrows, 24 J. LEGAL STUD.87, 87 (1995) (proposing "settlementescrows" to reduce delays in litigation bargaining);Geoffrey P. Miller, The Legal-EconomicAnalysis of ComparativeCivil Procedure,45 AM.J. COMP. L. 905, 915-16 (1997) (proposing a technique that would avoid the problem of settlements being delayed or never reached because one or both partiesrefuse to make legitimate settlementoffers for fear of appearingweak).

1102

CALIFORNIA LAWREVIEW

[Vol.88:1051

of the product.204 by providinga suitablehazardwarningto purchasers Such a rule appearssocially efficient, given the predictions of rational consumers can balancethe risksof usinga prodchoicetheory:informed uct againstits benefitsanddecidewhether usingor not usingthe product theirutility.205 will maximize But the efficiency of such a rule is called into questionif product manufacturers market theirproducts in sucha way as to anchor consumers' of the safetyof theproduct beforeproviding suitable warnings. perceptions for example,an advertisement for a sportutilityvehiclethatdeConsider, the vehicleat highspeedson irregular Anterrain. pictsa driver operating thatsuchmarketing choringandadjustment theoryleadsto the hypothesis consumer thatit is safeto operate the vehiploysmightanchor perceptions cle in sucha manner, it difficult for the same consumers to making adjust theirsafetyestimates whentheypurchase the vehicleandread sufficiently the manufacturer's detailedwarnings thatit can be dangerous to operate the vehicle at high speedsundersuchconditions.20 If this is true,even a well-designedwarningmay not lead to optimallyefficient consumer choices.Onepossiblelegalremedy wouldbe fortortlaw to denyor reduce for manufacturers of dangerous who provide liabilityprotection products hazard if they advertise in ways thatare theirproducts adequate warnings inconsistent withthosewarnings. III
THE IMPORTANCE OFCONTEXT

As a thought considerthe two versionsof the following experiment,

well-known hypothetical,originallydevised by Tversky& Kahneman: (1) You are on your way to see a play for which you do not have a ticket. Tickets cost $10. You realize that you have lost a $10 bill from your wallet. Will you still purchasea ticket for the play? (2) You are on the way to see a play for which you have purchased a ticket for $10. You realize you have lost the ticket. Will you purchase another ticket for $10?207

Expected utility theory predicts that actors will answer both questions In both cases the actor must comparethe benefit of attending identically.208
204. See, e.g., Moss v. Parks Corp., 985 F.2d 736 (4th Cir. 1993) (affirming grant of summary whose productwarninglabel complied with federal judgment in favor of mineral spirits manufacturer labeling requirements). 205. See generally W. Kip Viscusi, IndividualRationality,Hazard Warnings,and the Foundations L. REV.625, 629 (1996) (exploring the cognitive limitationsthat individuals of TortLaw, 48 RUTGERS have in incorporating warnings and suggesting how regulatory policy and tort law ought to take accountof those limitations). 206. We thankRetha Stotts for this insight into the possible interactionof advertisingand product warnings. 207. Amos Tversky & Daniel Kahneman, The Framing of Decisions and the Psychology of 453 (1981). Choice, 211 SCIENCE

2000]

LAWAND BEHAVIORAL SCIENCE

1103

the play with its marginal cost of $10. In bothcases, the actorfindshe is than he believed.This mightaffecthis cost-benefit $10 poorer previously in whichthe sum was lost shouldnot lead to a analysis,but the manner different of situation re(1) thanof situation (2). The empirical analysis sultsarecontrary to thisprediction. Whilenearlyall experimental subjects askedthe firstquestion answer a majority of thoseaskedthe affirmatively, secondquestion answer negatively.29 Thisexperiment is oftenusedto illustrate thepointthatindividuals do not considerall moneyfungibleand,instead,establishdifferent"mental accounts" fordifferent classesof goodsandservices.Havingalready spent $10 on a theater ticket,subjects exposedto the secondhypothetical might find their "theater account"empty, whereassubjectsexposed to situation (1) do not havethe sameproblem.210 We use this experiment to illustrate a far broaderpoint that forms the basis for this Part of the Article:despiterational choicetheory'simplicit to thecontrary, prediction context in decisionmaking. matters Rationalchoice theoryassumesthatwhen confronted with two options-$10 or a theater ticket,for example-the actorwill have a single, inherent valuefor each optionthathe can utilizeas the basis of a utilitycost-benefit In reality,a largenumber calculation. of external maximizing circumstances can affectthe utilitythathe will receivefromeach option, not simplywhetherhe had alreadypurchased (andlost) a theaterticket. Does he normally attendthe theater on Saturday night?Will his peersattendthe theater? Is the only otheroptionhe has considered stayinghome, or has he also considered a concert for which tickets cost $40? attending Will the play takeplace tonight,or next month?Answersto all of these it turnsout,arelikelyto affectthe actor'sdecisionin systematic questions, ways. Thesecontextual effectsareall inconsistent with the expectedutility versionof rational choicetheorybecausethey violatethe invariance prinin whicha choiceis presented should ciple, whichpositsthatthe manner not affectthe selectionan actormakesso long as thevariation in presentation does not affect the outcomesof the choices.211 Put another way, the of anygivenactorfor A or B shouldbe invariant to the means preferences aninvestigator usesto elicitthesubject's preferences. In thisPart,we reviewsomeof the evidenceof the effectsof context, in variousguises,on decisionmaking,anddescribesomeof the potential of the evidencefor legal policy analysis.First,however,a implications
208. See, e.g., Jolls, supra note 148, at 1669 (noting that "standard economic theory predicts no differencebetween these two situations"). 209. See Tversky & Kahneman,supra note 207, at 457. 210. See, e.g., Jolls, supra note 148, at 1669-70. 211. See supra PartI.A.2.

1104

LAW REVIEW CALIFORNIA

[Vol.88:1051

is in orderon the similarities anddifferences briefcomment betweendeof rationalchoice theorydiscussedin this viationsfrom the predictions in the previous Part.Whenpolicymakers wish to Partandthosediscussed a pre-established use law as a meansof achieving the goal, understanding of contextto decisionmakers,like understanding boundedly importance can enablepolicymakers a closerfit rational decisionmaking, to establish choicetheorywouldpermit. But betweenthe meansandendsthanrational to law as means of wish use a whenpolicymakers promoting efficiency, of contextaremore the policy implications thatflow fromthe importance The consethan those that flow from boundedrationality. complicated decisions of is that individuals make bounded particular quence rationality thetimeand in waysthatarenot utilitymaximizing for them(eventhough effort saved by using heuristicsmight enable them to maximizetheir globalutility).To the extentthatthe law can be used as a tool to help actors makedecisionsthatbettermaximizetheirutilityin those particular if an actorselects law can improve circumstances, efficiency.In contrast, in a particular buthe A"because "choice thechoicesarepresented context, wouldhaveotherwise selected"choiceB,"it is oftendifficultto determine whetherthe law will enhanceefficiencyby reinforcing "choiceA," enB" in or "choice of the context, changingthe contextso couraging spite thatthe actorwill select "choiceB." This difficultywill arisethroughout in various thisPart,andwe suggestsomepossiblesolutions circumstances whereappropriate.
A. ReferencePoints and the FramingEffect

In 1979,Kahneman andTversky a descriptive proposed theoryof decisionmakingthattheycalled"prospect The theory."212 theorywas meant to serveas a morerealisticalternative to expectedutilitytheoryin thatit some empirical accommodated findingsaboutdecisionmakingunderunthat did not fit into One fundamental certainty expectedutilitytheory.213 of known as the effect, is thatactors' insight prospecttheory, "framing" choices underconditions of uncertainty dependnot only on the absolute values of the butalsoon thedirection in which expected competing options thoseoptionsdeviatefroma baseline,or reference When decision point.214 as "gains" relative to thereference optionsareperceived point,individuals areriskaverse; thatis, theyprefer morecertain withthe optionsto gambles sameexpectedvalue.But whendecisionoptionsareperceived as "losses" relativeto the reference will be risk-seeking; point,the sameindividuals
212. Daniel Kahneman& Amos Tversky, Prospect Theory: An Analysis of Decision UnderRisk, 47 ECONOMETRICA 263 (1979). 213. See id. at 267. 214. See Daniel Kahneman, Reference Points, Anchors, Norms, and Mixed Feelings, 51 ORG. BEHAV. & HUM. DECISION PROCESSES 296 (1992); Kahneman& Tverksy, supra note 212.

2000]

LAWAND BEHAVIORAL SCIENCE

1105

that is, they will prefer a gamble to the certain option when both have the same expected value.215 For example, when experimental subjects were asked to choose between a certain payment of $240 and a 25% chance to receive $1,000, 84% preferredthe certain payment (although this option has a slightly lower expected value thanthe gamble-$240 versus $250).216 But when asked to choose between a certain loss of $750 and a 75% chance of losing $1000, 87% preferred the gamble (although the two choices have the same expected value).217 Considerable additional research indicates that frames-that is, whether options are perceived as gains or losses-are fluid and easily manipulable, at least in some situations. Consider the following experiment: when subjectswere told to imagine that they had been given $1,000 and asked to choose between a certain gain of an additional $500 and a 50% chance of gaining an additional $1,000, 84% preferredthe certain payment, therebyexhibiting risk aversion. But when subjects were told to imagine that they had been given $2,000 and then asked to choose between a certain loss of $500 and a 50% chance of losing $1,000, 70% preferred the risky alternative,thereby exhibiting risk preference. Notice that both groups of subjects were asked, in effect, whetherthey would prefer a certain $1,500 or a 50% chance of $2,000 coupled with a 50% chance of $1,000. Merely by dividing the presentationof the choice into two parts, the experimenterswere able to alter the referencepoint that many subjects used to evaluate the decision, and this, in turn, affected the subjects' revealed preferences.218 How frames can affect decision makingin the context of litigationhas been explored in three sets of experiments,includingone by Korobkinand Guthrie.219 In those experiments,subjects playing the role of plaintiffs in a litigation matterwere asked to choose between a certainsettlementoffer of $21,000 and a trial that could potentially yield $28,000 but had a riskadjusted expected value of $19,000. Predictably, and in accordance with the predictions of rational choice theory, subjects chose the safe choice with the higher expected value when both the settlement and the most desirable trial verdict would leave them betteroff than they were priorto the accidentthat gave rise to the litigation-that is, when the losses sufferedin the accident were less than $21,000.220 But subjects facing the same choice
215. See Kahneman,supra note 214, at 298. 216. See Tversky & Kahneman,supra note 195, at 255. 217. See id. 218. Note that these robustexperimentalresults imply violation of the invarianceprinciple,one of the background conditionsfor the operationof expected utility theory.See supra PartI.A.2. 219. See Korobkin& Guthrie,Psychological Barriers,supranote 199; see also Linda Babcock et al., Forming Beliefs about Adjudicated Outcomes: Perceptions of Risk and Reservation Values, 15 INT'L REV. L. & ECON. 289 (1995); Jeffrey J. Rachlinski, Gains, Losses, and the Psychology of Litigation, 70 S. CAL.L. REV.113 (1996). 220. See Korobkin& Guthrie,Psychological Barriers,supra note 199, at 130-33.

1106

CALIFORNIA LAWREVIEW

[Vol.88:1051

settlement offerof $21,000andthe riskytrialprospect betweenthe certain were significantly morelikely to opt for trialif acceptingthe settlement wouldhaveleft themworseoff thanthey werepriorto the accident-that in the accident exceeded$21,000.221 is, whenthe losses suffered Although facedthe samechoice,in absolute bothgroupsof subjects terms,subjects in the firstgroupcouldonly perceivetheirchoiceas one between"gains," while subjectsin the secondgroupcould have perceivedtheirchoice as if theyviewedit froma "pre-accident" one between"losses" rather thana reference Consistent with the predictions of pros"post-accident" point.222 who couldplausibly view the decisionto settleor go pect theory,subjects to trial as a choice between "losses"were more likely to exhibit riskseekingbehavior. If litigants oftenperceive thechoicebetweensettlement andtrialto be a choicebetweenlosses, the framing like the effect, bias,can self-serving for why lawsuitssometimes fail to settle providea plausibleexplanation out of courtdespitetheenormous costsof litigating to a finaladjudication. Rachlinski has that Indeed,Jeffrey argued plaintiffsarelikely to perceive since they standto receivemoney,whereas litigationoptionsas "gains," defendants arelikely to perceivetheiroptionsas "losses," sincetheygenstand to As a be more erally may pay money.223 result,plaintiffs willingto settle thanpreviously morewilling to litigate.If thought,anddefendants this is correct,thenthe legal systemmay wish to focus its effortson enthanplaintiffs)to settle.Moreover, if legal defendants (rather couraging believe it is appropriate for the legal systemto encourage policymakers low-costdisputeresolution, evidenceof the framing effect providessupfor semicoercive resolution such as mandatory measures, port dispute pretrialmediation, or court-supervised settlement or arbitration, conferences, even morecoercivemeasures, suchas requirements thatplaintiffs who fail to recoverat trialan amountgreater thanthe defendant's most generous
settlementoffer be held liable for the defendant'strialcosts.224

As a moregeneral evidenceof theframing effectsuggeststwo matter, of for who wish to encourage a possibletypes strategies legalpolicymakers certain behavior. to extent the that frames canbe manipulated (in that First, as gainsor losses),thegovernment canattempt to optionscanbe presented actors with information framed in a to the desired present way encourage
221. See id. 222. See Russell Korobkin& Chris Guthrie,Psychology, Economics,and Settlement:A New Look at the Role of the Lawyer,76 TEX.L. REV.77, 99 (1997). 223. Rachlinski,supra note 219, at 128 (presentingexperimentalevidence that subjects playing the role of plaintiff were more likely to favor settlementrelative to a trial with a given expected value than subjectsplaying the role of defendant). 224. See, e.g., IND.CODE ANN. ? 34-50-1-6 (West 1998) (awardingattorney'sfees if the plaintiff refuses a settlementoffer and the final judgment is less favorableto the plaintiff than the settlement); OKLA. STAT. ANN. tit. 36, ? 3629 (West 1998) (awardingcosts and attorneyfees in insurancematters to the insuredif the judgmentexceeds the highest settlementoffer and to the insurerotherwise).

2000]

ANDBEHAVIORAL LAW SCIENCE

1107

behavior.If risk-aversebehavior is desirable,informationshould be in a way to encourage actorsto see theiroptionsas choicesamong framed is desirable, information shouldbe framed in a way gains;if riskybehavior to see theiroptionsas choicesamonglosses. to encourage actors andthis Second,to the extentthatframesaredifficultto manipulate shouldconsider behavior, difficultymay lead to undesirable government choicesfromthe realmof individual decisionmaking. As an exremoving assume that the decidesthat,for the purpose of autoample, government mobilesafety,it is desirable to increasethe use of seatbelts.Peoplewho are accustomed to not wearinga seatbelt, however,are likely to view the decisionof whether or not to do so as a choicebetweenlosses. Wearing a seatbelt will likelybe perceived as causinga smallbutcertain loss of freedom andcomfort. Not wearinga seatbelt, in contrast, can be seen as subthe actor to a small of a loss (in the formof jecting possibility very large increased if an accident of occurs),coupledwitha largepossibility injuries no loss at all (if thereis no accident). Of course,peoplewho view the inconvenience of wearing a seatbelt as minorandtherisksof possibleinjury as catastrophic will choosethe safe optionof buckling up, even whenthe is viewed as a choice betweena certainloss and a probabilistic problem loss. But viewingthe problem as one of competing losses cancauseactors on the marginto engagein risk-seeking in this case, to choose behavior; not to wearthe seatbelt. The government the respond might by removing decision from individual actors. For seatbelt safety example,mandatory laws can be imposed,225 or automobile manufacturers can be required to in safetywithout relyingon thechoicesof drivers.
B. install air bags in their cars,226 thus providing some of the desired increase TheEndowment Effectand the Status Quo Bias

Underanyof the usualconceptions of rational choicetheory,an actor facedwith a choicebetweentwo items-say, for example,betweena loaf of bread anda glass of wine-should compare the inherent valueto himof eachandselectthe itemwiththehighervalue.227 Whether he ownstheloaf of bread,the glass of wine, or neither,shouldmakeno differenceas to whichitemhe wouldprefer. An unusually richbodyof behavioral science
225. See, e.g., ARK.CODE ANN. ? 27-37-702(a) (LEXIS 1999); D.C. CODE ANN. ? 40-1602(a) (Michie 1998); HAW.REV. STAT.ANN. ? 291-11.6(a)(1) (LEXIS 1998); N.C. GEN. STAT.? 20STAT. ANN. tit. 47, ? 12-417A (1998); S.C. CODE ANN. ? 56-5-6520 135.2A(a) (LEXIS 1999); OKLA. (Law Co-op. 1998). 226. See 49 U.S.C.A. ? 30127(b)(C) (1998) (mandatingair bags in every new car beginning in 1997). 227. See, e.g., Samuel Issacharoff,Can ThereBe a Behavioral Law and Economics?, 51 VAND. L. REV. 1729, 1735 (1998) (explaining that law and economics presumes that value is "source independent"); Murray B. Rutherford et al., Assessing Environmental Losses: Judgments of ENVTL. L. REV.51, 61 (1998) (noting that "traditional Importanceand Damage Schedules, 22 HARV. economic theory"assertsthat valuationsof gains and losses are the same).

1108

CALIFORNIA LAWREVIEW

[Vol.88:1051

literature demonstrates, however, that individuals often place a higher monetaryvalue on items they own than on those that they do not own. Because the consequence of this effect is that people place a higher value on their endowments than on other items, this phenomenon is often Because the effect also results in referredto as the "endowmenteffect."228 actors' placing a higher dollar value on goods they are selling than on or goods they are buying, it is also referredto as the "offer/askinggap"229 the "Willingness to Accept (WTA)/ Willingness to Pay (WTP) gap."230 However labeled, the effect is a consequence of the fact that individuals tend to value losses more highly than equivalent gains (and thus wish to avoid "losing"things more than they desire "gaining"things of an equivalent value). This phenomenon,known descriptivelyas "loss aversion,"is a close relative of the framingeffect,231 discussed above. The most famous examples of the endowmenteffect come from a series of experimentsconcerningmugs and lotterytickets. In one experiment, Kahnemanand his colleagues providedeach memberof one group of subjects with a coffee mug and each memberof anothergroupwith $6.232 They elicited from the first group ("sellers")the minimumprice that the subjects would demandto give up the mug. From the second group("buyers"), they elicited the maximum amountof money that the subjects would pay to acBoth groups were told that the experimenters quire one of the mugs.233 would take this information, calculate the market-clearingprice for the mugs, and reallocate and execute trades between the mug holders who would prefer cash to their mug at the market price and the cash holders who would prefera mug to cash at that price.234 Surprisingly,from the perof rational choice sellers valued the mug at roughly spective theory,235 twice the price of buyers, and very few trades took place, even when
228. See, e.g., RichardThaler,Towarda Positive Theoryof ConsumerChoice, 1 J. ECON. BEHAV. & ORG. 39,44(1980). 229. See, e.g., Duncan Kennedy, Cost-BenefitAnalysis of EntitlementProblems: A Critique,33 STAN. L. REV. 387, 401 (1981) (referring to the "offer/asking" problem); Russell Korobkin, Policymakingand the Offer/AskingPrice Gap: Towarda Theoryof Efficient EntitlementAllocation, 46 STAN. L. REV.663, 664-65 (1994). 230. See, e.g., Elizabeth Hoffman & Matthew L. Spitzer, Willingness to Pay vs. Willingness to U. L. Q. 59 (1993). Accept: Legal and EconomicImplications,71 WASH. 231. See, e.g., Kahneman & Tversky, supra note 212, at 279 (discussing loss aversion as an element of prospecttheory). 232. See Daniel Kahneman et al., Experimental Tests of the EndowmentEffect and the Coase Theorem, 98 J. POL.ECON.1325 (1990); see also Jolls et al., supra note 13, at 1483-84 (discussing Kahneman'sexperiment). 233. See Daniel Kahnemanet al., supra note 232, at 1330-31. Note that the experimentersgave the "buyers" cash to avoid the problemof "sellers " being wealthierthanbuyers, on accountof owning the mug. 234. See id. at 1331. 235. The reason the result is surprisingto rational choice theorists is that there is no particular reason not to imagine that tastes for mugs versus cash are symmetricallydistributedamong any random group of subjectsso that abouthalf of mugs and cash shouldchange hands.

2000]

LAWAND BEHAVIORAL SCIENCE

1109

multiple iterationsof the experimentwere conductedto allow participants to learnfrom experience.236 In anotherexperiment, Knetsch and Sinden gave some experimental subjects lottery tickets, which entitled the holder to the chance to win a sum of money, and gave others $3, then offered to buy the tickets from ticket holdersfor $3 and offered to sell tickets to cash holdersfor $3.237 The majorityof ticket holders elected to keep their tickets, while the majority of cash holders elected to keep their cash,238 again suggesting that ownership affected the value that subjectsplaced on the items. The most obvious policy implications of the endowment effect come in the area of propertylaw. Much of the law-and-economics approachto propertylaw is based on the insight of the Coase Theoremthat the assignment of property rights has no efficiency impact if transactioncosts are low.239 In such circumstances,any individual who has a higher value for a given entitlementthan the original owner will purchaseit from the owner. From this follows the general principle that property law should seek to minimize transactioncosts, as well as other, more specific corollaries, including the following: (1) property rights should be clearly delineated, because cloudy title increases transactioncosts;240 (2) injunctionsare the remedies to unless transaction costs are very high, bepreferred damages cause injunctionsclarify propertyrights (again, facilitatingbargaining)and court-determined and (3) propertyrights damages can be unpredictable;241 should often be assigned to the claimant who can transferthe right most cheaply.242 Evidence of the endowmenteffect calls into questionthe efficiency of each of these law-and-economicsprescriptions.The endowmenteffect acts similarly to transactioncosts: it presents a barrierto the reallocation of propertyrights from an owner to anotherpartywho can put those rights to
236. See Kahneman,supra note 232, at 1332. 237. See Jack L. Knetsch & J.A. Sinden, Willingness to Pay and Compensation Demanded: ExperimentalEvidence of an UnexpectedDisparity in Measures of Value,99 Q. J. ECON. 507, 512 (1984). 238. See id. at 513. 239. See Coase, supra note 170, at 6; see also COOTER & ULEN,supra note 8, at 85 (summarizing the Coase theorem). 240. See, e.g., COOTER & ULEN,supra note 8, at 93 ("Lowering transactioncosts 'lubricates' bargaining.One importantway for the law to do this is by defining simple and clear propertyrights.") and 107 ("[T]he law can achieve the efficient allocation of private goods by, for example, lowering bargainingcosts by assigning clear and simple ownershiprights."). 241. The original insight comes from Guido Calabresi & A. Douglas Melamed, PropertyRules, Liability Rules, and Inalienability: One View of the Cathedral, 85 HARV.L. REV. 1089, 1116-17 & ULEN,supra note 8, at 102-03; James E. Krier & Stewart J. Schwab, (1972). See also COOTER Property Rules and Liability Rules: The Cathedral in Another Light, 70 N.Y.U. L. REV. 440, 445 (1995). 242. See, e.g., MARK A GUIDE TOCRITICAL LEGAL STUDIES KELMAN, 123-24 (1987) (describing the "marketfacilitating" approach to entitlement allocation); Korobkin & Ulen, supra note 7, at 336-38; see also Coase, supra note 170, at 17-18.

1110

CALIFORNIA LAWREVIEW

[Vol.88:1051

a more valuableuse. Consequently, in some situationsit mightbe effito leave in an effortto unclear, ciencyenhancing property rightssomewhat If the endowment an endowment effect fromtakinghold.243 effect prevent causes lower-valuingowners to refuse to sell entitlements to highervaluingpurchasers, damageremediesmay be moreefficientthaninjuncthe higher-valuing tive reliefbecausethe former to take purchaser permits theentitlement by payingthemarket pricefor it.24Finally,theendowment effect suggeststhat,even when transaction costs are low, policymakers to allocate withefficiencyshouldattempt concerned property rightsto their likelihood of efficientreallocations. mostefficientuserdueto thereduced to highlight animportant coman opportunity Thislastpointprovides for in interested that the endowment effect raises policymakers plication user. rightsefficiently-that is, to theirhighest-valuing allocating property value on the in the resources which claimant Determining places highest more than the of rational can be much assumptions question complex choice theorywould suggest. Considerthe following two hypothetical in ascending order of difficulty: described situations, (1) Cain would be willing to pay (WTP)$5,000 for a piece of if he didnotown it, butwouldaccept(WTA)no less than property $7,000to sell it if he did own it. Abelhas a WTPof $4,000for the samepieceof property anda WTAof $6,000. a WTP of anda WTA Cain has $3,000 for a piece of property (2) of $7,000. Abel has a WTP of $4,000 for the same piece of anda WTAof $6,000. property We summarize thesecasesin thefollowingfigure: Situation(1)
Cain Abel

Situation(2)
Cain Abel

WTP
WTA

$5,000
$7,000

$4,000
$6,000

WTP
WTA
FIGURE 2

$3,000
$7,000

$4,000
$6,000

243. Ian Ayres and Eric Talley have arguedfor cloudy or divided propertyrights from within the rationalchoice paradigm,claiming that uncertaintyreduces rational strategic behavior that can block efficient entitlement transactions. See Ian Ayres & Eric Talley, Solomonic Bargaining: Dividing a Legal Entitlementto Facilitate Coasean Trade, 104 YALEL.J. 1027, 1029-30 (1995) ("[W]hen two partieshave privateinformationabout how much they value an entitlement,endowing each partywith a partial claim to the entitlement can reduce the incentive to behave strategicallyduring bargaining, thereby enhancing economic efficiency."). But see Louis Kaplow & Steven Shavell, Property Rules Versus Liability Rules: An Economic Analysis, 109 HARV.L. REV.713, 779-88, 790 nn.69 & 71 (1996) (criticizing the Ayres & Talley article and arguing generally for liability rules to deter externalitiesand for propertyrules to protectholdings in things). 244. Jolls, Sunstein, and Thalerhave hypothesizedthat parties who litigate contested entitlements and are actually awarded injunctive relief by a court are particularlyunlikely to bargain away their endowment, even when the opposing party places a higher value on it, because they are likely to believe they have earned and/orare particularlyentitled to the endowment. See Jolls et al., supra note 13, at 1497-1501.

2000]

LAWAND BEHAVIORAL SCIENCE

1111

In both situations, the initialallocation decisionis of criticalimporwill not be exchanged:if Cainoriginally tancebecausethe property owns theproperty, thenhe woulddemand morefortheproperty thanAbelwould be willingto pay for it, andif Abel originally ownedthe property, thenhe woulddemand morethanCainwouldbe willingto pay. eitherinitialallocation wouldbe Pareto mostobAlthough optimal,245 serverswouldprobably (1), Cainis the moreeffiagreethatin situation cient user of the property, since both his WTPandWTA values exceed Abel's equivalentvalues. In situation(2), the identityof the efficient owneris farless clear,becauseCainhas a higher WTAthanAbelbutAbel has a higherWTPthandoes Cain.246 how to allocate thepropDetermining most in this of set circumstances a more nuanced erty efficiently requires of the reasonsthatCainandAbel mighthaveasymmetrical understanding differences betweentheirWTA andWTPvalues,a complicated question thatis farbeyondthe scope of this Article.247 But the exampleaptlyillustratesthatevidenceof the endowment effectcalls for a moresophisticated of property analysisof how to assurethe efficientdistribution rightsthan rational choicetheory requires. of theendowment Discussions effectusuallyassume thatthedisparity oftenobserved betweenan actor'sWTPandWTAvaluesis a function of But related evidence shows that the endowment ownership. experimental effect is betterunderstood as a specificapplication of a moregeneral pheoftenknownas the "status theremightbe nomenon, quobias."248 Although some uniqueattributes of ownership thatmakeactorsespeciallyreluctant to partwith itemsin whichthey enjoyproperty scirights,the behavioral ence literature indicates thatactors tendto placea higher valueon anystate of affairsthatthey consider to be the statusquo thanthey wouldplaceon thatsamestateof affairs if it werenotthe status of whether quo,regardless is implicated.249 property ownership Thisbroader statusquo bias has implications for legal policy farbethe confines of law. Considersome implications for the yond property of contract law rules.Traditional law-and-economics promulgation analysis of contract "default" rules-that is, rulesthatgovernthe relationship betweencontracting partiesonly if the partiesdo not explicitlyagreeto differentterms-posits that (1) unless transaction costs are unusually
245. An allocation is Pareto optimal if it is impossible to reallocate goods and services so as to make one party better off without making anotherparty (in this case, the other party) worse off. See COOTER & ULEN, supra note 8, at 12. 246. Indeed, one of the problems that this analysis highlights is the ambiguity in the concept of user." "highest-valuing 247. For a theoreticalanalysis of how policymakersconcernedwith efficient propertyallocations might solve this problem,see Korobkin,supra note 229. 248. See William Samuelson & Richard Zeckhauser,Status Quo Bias in Decision Making, 1 J.
RISK& UNCERTAINTY 7 (1988).

249.

For a series of examples, see id.

1112

CALIFORNIA LAWREVIEW

[Vol. 88:1051

high, the choice of default rules will have little effect on contract terms because wealth-maximizingparties will contractaroundinefficient default and (2) default terms should mirrorthe terms that the majorityof terms,250 defaults) to minimize the contractingparties would choose ("majoritarian" transaction costs when contractingaroundinefficient defaults.251 Evidence of the status quo bias suggests that revisions to both elements of the conventional wisdom are appropriate.In a recent article, Korobkinhas shown experimentallythat default rules are more difficult to contract aroundthan rational choice theory explanations suggest. This is because contracting parties are likely to see default terms as part of the status quo and, consequently, prefer them to alternative terms, all other If this is correct, default terms will be "sticky," and the things equal.252 choice of defaults may determinethe terms that the parties adopt in many cases. Even if majoritarianterms are selected as defaults, this stickiness will cause some of the parties in the "minority"to adopt inefficient contract terms. To avoid this inefficiency, default terms tailored to specific partiesand determinedby courts at the time disputes arise might be preferable to well-specified majoritariandefaults. The former, by leaving the content of default rules at the time of contractingunresolved, should reduce the opportunitiesfor parties to become biased in favor of the status quo terms.253

250. See, e.g., ROBERT A. HILLMAN, THE RICHNESS OF CONTRACT LAW: AN ANALYSIS AND OFCONTEMPORARY OFCONTRACT THEORIES LAW225 (1997); see also BernardS. Black, Is CRITIQUE Corporate Law Trivial?: A Political and Economic Analysis, 84 Nw. U. L. REV.542, 557 (1990) because partiescan contractout of (claiming that defaultrules in corporatelaw "aren'tvery important" them at low cost). 251. See, e.g., HILLMAN, supra note 250, at 225 ("theefficient ... defaultrule is what most parties would want");POSNER, supra note 26, at 105 (explainingthat in interpreting contracts,courts establish default rules that "imagine how the parties would have provided for the contingency if they had decided to do so"); CharlesJ. Goetz & Robert E. Scott, The Mitigation Principle: Towarda General Theoryof ContractualObligation,69 VA. L. REV.967, 971 (1983) (default rules should be createdby asking "what arrangementswould most bargainersprefer?").Against the majoritarian interpretation, Ayres and Gertnerhave argued in favor of what they call "penalty"default rules, which would not in favor of the least-advantaged reproducewhat most partiesdesire but would be structured partyso as to create an incentive for the more-advantaged party to divulge information.See Ian Ayres & Robert Gertner,Filling Gaps in Incomplete Contracts: An Economic Theory of Default Rules, 99 YALEL.J. 87 (1989) (giving as an example of a penalty default the Uniform CommercialCode's rule voiding a contractfor failing to specify a quantityterm). But see Jason Scott Johnston,StrategicBargaining and the Economic Theory of Contract Default Rules, 100 YALEL.J. 615 (1990) (arguing that a penalty default may not induce partieswith superiorinformationto divulge that informationif doing so leaves them open to exploitationby the otherparty). 252. See Russell Korobkin,The Status Quo Bias and ContractDefault Rules, 83 CORNELL L. REV. 608, 637-47 (1998) [hereinafterKorobkin,Status Quo Bias]. For additionalevidence of the effect of the statusquo bias on the negotiationof contractterms, see Russell Korobkin,Inertiaand Preference in ContractNegotiation: The Psychological Power of Default Rules and Form Terms,51 VAND.L. REV. 1583 (1998) [hereinafter Korobkin,Inertia]. 253. See Korobkin,StatusQuo Bias, supra note 252, at 670-73.

2000]

ANDBEHAVIORAL LAW SCIENCE

1113

determine thatmajoritarian defaultsareapproEven if policymakers the bias of the status the bias, highlightsthe quo priatedespite presence that face in to determine whichterms attempting difficulty policymakers the employof contracting the majority partieswouldfavor.Forinstance, ment law academyhas long debatedwhetherthe properdefaultrule for contracts shouldbe thatemployersmay dismissemployees employment In theUnitedStates,the "atwill"rulehas "atwill"or onlyfor "just cause." default.254 Law-and-economics scholars been the dominant operating long have often cited as evidencefor the withinthe rational choice paradigm fact thatrelativelyfew partiesapparefficiencyof this rulethe empirical contract around this default.255 The statusquo bias revealsthatthis ently is fallacious. The fact that around contract the "atwill" logic parties rarely is efficientformostparties, rulemightmeanthat"atwill"employment but it also mightmeanthatthe statusquobias swampsa preference manyparties wouldotherwisehave for a "justcause"term.256 At a minimum, the statusquo bias demandsthat lawmakers seekingto promulgate majoritarian defaulttermslook for evidenceotherthanwhattermsareadopted in a market with an existentdefaultfor indications as to whattermsthe majoritywouldprefer.
C. Habits, Traditions, Addictions,and Cravings

and behaviorare affectednot only by the relationship of Judgment decisionoptionsto baselinereference pointsbut also by theirrelationship to temporally behaviors. ThisSectionexamines the implications separated of the effect thatpast behaviors, andthe relatedphenomenon of visceral choices.The followingSectionexaminesthe drives,can have on current effect thatthe temporal distance betweenthe effectsof different optionsthe distance between the presentdecisionand futureconsespecifically, quences-can haveon choices. Becauseeconomistsemployingrationalchoice theorygenerallyassume preferences are determined of behaviors,a common independent
254. See J. Hoult Verkerke, EmploymentContractLaw, in THE NEW PALGRAVE DICTIONARY OF ECONOMICS ANDTHELAW47 (PeterNewman ed., 1998). 255. See RichardA. Epstein, In Defense of the Contractat Will, 51 U. CHI.L. REV.947, 951-52 is "freelywaivable by a joint expressionof contrary (1984) (statingthatthe contractat will requirement intention,"yet remains (and should remain) the most common employment relationship);RichardA. Epstein,StandingFirm, on ForbiddenGrounds,31 SAN DIEGOL. REV.1 (1994) (continuingto defend the contract-at-willdoctrineagainsta host of critics of his book-lengthdefense, ForbiddenGrounds);J. Hoult Verkerke,An Empirical Perspective on Indefinite TermEmploymentContracts: Resolving the Just Cause Debate, 1995 Wis. L. REV. 837, 867-75 (1995). 256. Of course, anotherpossibility is that a "justcause" term would be efficient for many parties but employees do not wish to bargain for it because they fear such a request would be perceived by employers as a sign that they were likely to shirk their responsibilities.See, e.g., Samuel Isaacharoff, Contracting for Employment:TheLimitedReturnof the CommonLaw, 74 TEX.L. REV.1783, 1794-95 (1996).

1114

LAW CALIFORNIA REVIEW

[Vol.88:1051

is thatchoicesan actorhas madein the pastwill not affecthis assumption current For example,if Abel had been asked to structure.257 preference choose betweenan appleand an orangeyesterday and chose the orange, this wouldnot affect how likely he wouldbe to choose an orangetoday it wouldcertainly a prediction thathe wouldbe likelyto (although support chooseanorange choicetheorists believe today).To theextentthatrational thatyesterday's choice will affecttoday's,the correlation is usuallypresumedto be negative,an application of the conceptof declining marginal Even someone who loves orangeswill desire his first orange utility.258 morethanhis second,andso on. If Abel ate too manyoranges yesterday, the utilityhe wouldderivefromeatingyet another orange todaymightdip belowtheutilityhe wouldderivefromhis firstapple. The problem for the conventional of rational choicetheconceptions is that the between and actions is oftenposiory relationship past present tive rather thannegative.Thatis, the fact thatan actoractedin a certain the likelihood thathe will actin the sameway way in thepastcanincrease in the future. Thispositiveeffectof pastchoiceson current ones canresult fromat leastthreesomewhat different which we will defineas phenomena, and"addictions." "habits," "traditions," Actorsoftenrepeat behaviors choosethe samegoodor (orrepeatedly out of a as of the costs of decisionmaking. "habit," service) way reducing of behavior device.We may is, in this way, used as a heuristic Repetition drivethe sameway to worktodaynotbecausewe determine whenwe start the ignitionthatthis is the mostefficient(or utility-maximizing) routebut becausewe presume thatit is likelyto be efficientbecausewe haveusedit before.This calculation often will be unconscious. Like otherheuristics, decisionmakingvia habitis oftenquiterational in a globalsense,because it permitsus to approximate behaviorat a reasonable utility-maximizing on ourseatbelt cost, as whenwe developthe habitof alwaysputting or of both before into the street. But habitual behavior looking ways stepping can havethe consequence of causingactorsto makesuboptimal decisions in particular circumstances. Whereas behaviors thatresultfromforce of habitcan be viewed as heuristicbased,behaviorsthat are drivenby "tradition" can be seen as to the statusquobias.Consider, for example,a personwho closelyrelated andwould,therefore, gets negativeutilityfromeatingfruitcake normally decline even a free slice, but eats fruitcakewith abandon every winter results from holidayseason.Likethe status quobias,thepowerof tradition
257. See, e.g., Gary S. Becker, Habits, Addictions, and Traditions,45 KYKLOS 327, 327 (1992) ("The usual assumptionin most discussions of behaviorover time is that choices today are not directly dependenton choices in the past."). 258. See MICHAEL L. KATZ& HARVEY S. ROSEN, MICROECONOMICS 57 (1991) (increasing the consumptionof a good decreasesthe marginalutility of additionalunits of the good consumed).

2000]

ANDBEHAVIORAL LAW SCIENCE

1115

the utility that individualsderive from conformingto a sharedfamily, rather thanfromthe inherent valueof a bepractice, group,or community value of a traditional havior.Of course,the inherent activityis far from Individuals are to beside the unlikely engagein trulydespised being point. behaviors Forexample,a personmadeill by merelybecauseof tradition. fruitcake is unlikely to eatit evenat Christmas thelong-time despite family tradition to do so. Theyare,however, morelikelyto makea givendecision or engagein a given behavior if they have a tradition of doingso thanif in our taxonomy, do not. Unlike the heuristic of habit, "traditions," they havea conscious effecton individual preferences.259 like traditions, resultwhen the fact that an actorhas "Addictions," in the pastmakeshimmorelikelyto engagein that engagedin a behavior in behavior the futurebecausethe past behavior makescurrent behavior morepleasant. Fromone vantagepoint,an addiction is just a particularly But the physicalor chemicalcompulsion powerfulspeciesof tradition.26 thatmotivates an "addict," underthe commonunderstanding of the term, aredifferent in kind,notmerely suggeststo us thatthe effectsof addiction in degree,fromthe effects of tradition andthat,therefore, very different A personwhose policyimplications mightflow fromthe two phenomena. desireto drink,smoke,or gamble,is motivated the fact thathe has enby in those activities we in a different before, believe,belongs gaged category fromthatof a personwhosedesireto eat fruitcake is motivated by the fact thathe does so everyChristmas without fail. Thepositiverelationship between behaviors as reprepastandpresent sentedby habits,traditions, andaddictions, has two potentially broadimplications for legal policy. The first, and relatively uncontroversial, to encourageor discourage implicationis thatpolicymakers attempting certain behavior shouldrealizethatbehaviors motivated by habit,tradition, or addiction, are likely to be muchmoredifficultto manipulate thanrationalchoicetheory wouldpredict. Whenpastbehavior, rather thanmerely inherent behavvalue,drivesthe utilitythatan actorreceivesfromcurrent of a behavior with ior, laws that try to offset the inherent correutility costs will not create a sufficient deterrence. To use a stark sponding this is federal criminal for posexample, probably why draconian penalties sessionof even a relativelysmallquantity of crackcocainehavefailedto

259. Traditions,when shared,may be a means of solidifying communitybonds throughrituals.To the extent that traditions serve this purpose, they resemble social norms, which we examine in Part IV.A below. 260. Cf. Alan Schwartz, Views of Addiction and the Duty to Warn,75 VA. L. REV.509 (1989) (arguing that addicts continue to engage in addictive behavior only when the benefits of doing so exceed the costs to them).

1116

CALIFORNIA LAWREVIEW

[Vol.88:1051

Even if the risk-adjustedexpected cost of using put an end to drug use.261 an illegal substanceexceeds the inherentbenefit to an individual of doing so, the addict is unlikely to go straight.At least in such extreme situations, policymakers are likely to have to find measures other than merely increasingthe price of the undesirablebehaviorif they hope to eliminateit. The more controversial policy question is to what extent the law habshould paternalisticallyattemptto protect individualsfrom "harmful" an actor's total its, traditions,and addictions(those that reduce utility over time), or to encourage "beneficial"habits, traditions,and addictions(those that increase futureutility).262 High "sin"taxes on addictive substancescan be seen as an example of the former approach.The cost-benefit tradeoff involved in smoking, for example, might appearmore favorable initially than it will over the long run as the health risks increase and the ease of quitting decreases. The governmentmight tax such behavior to the point that only individuals who would actually maximize their utility over the Rational choice theorists long term by smoking will begin the habit.263 would arguethat even if actions in time period 1 motivatebehaviorin time period 2, a rational individual will equate long-term costs with benefits, taking into account the power of a phenomenon such as addiction, and choose to engage only in activities that they expect to be utility enhancing But there are many reasons that this prediction too over the long term.264 could fail-for example, if actors are overconfident about their ability to avoid the utility-decreasing effects of harmful traditions and addictions. Addictive behaviors seem likelier targetsfor this type of legal intervention than merely traditionalbehaviors, as there is a strongersense with the former that the future behavior lacks an element of free will that is present with the latter. In addition to governmentalprodding,private parties may create incentives for individualsto avoid activities with long-termadverse effects. For example, health and life insurersmay offer lower premiumsto those who refrainfrom smoking and drinking. Just as the governmentmight tax behaviors that motivate harmfulfuture behavior, it might actually or constructively subsidize behaviors that motivatefuturebeneficial behavior.For example, fasteninga seatbeltwhen driving might be a behavior that is initially costly, as the driver must rememberto fasten the belt and must suffer a loss of comfort.As fastening
261. See Timothy Egan, War on Crack Retreats, Still Taking Prisoners, N.Y. TIMES, Feb. 28, 1999, at Al, A20 (reportingthat studies show druguse has remainedconstantover the last decade even in the face of increasingcriminalsanctions). 262. The distinctionis drawnfrom Becker, supra note 257, at 328. 263. We set to one side the difficult task of specifying how this determination might be made. 264. See generally Becker, supra note 257. Becker notes that because of uncertainty an actor might develop a habit that turns out to have a negative net utility, but his argumentis premised on the belief that an actor will accurately perceive the likelihood of various future consequences of current behavior.See id.

2000]

LAWAND BEHAVIORAL SCIENCE

1117

the seatbeltbecomes a habit, however,the costs of doing so decrease is replaced greatly,andit is possiblethatthe feelingof inconvenience by a of comfort and the as behavior becomes habitual. Consefeeling security it is possiblethat,left to theirowndevices,actorswill underutilize quently, seatbeltsunlessthey are somehowprodded to changetheirhabits.Laws canprovidesuchprodding the failure to wearseatbelts or by recby fining a in seatbelt defense which cases victims in (under ognizing negligence automobile collisionscouldrecover thattheywouldhave only the damage sufferedhadtheybeen wearinga seatbelt). If theseeffortsproveunavailthe law can habit reinforcement for example, ing, go beyond by requiring, thatall carscomewithpassiverestraint systems,suchas airbags. Behaviors thataremotivated addictions beara strong resemblance by to behaviors thataremotivated visceral thetwo pheby cravings, although nomena havedifferent includehunger, thirst, etiologies.Visceralcravings sexual desire,sleep, and pain, to namea few examples.265 Unlike addictions, such cravingsare generallyevolved traitsratherthandesiresthat stemdirectlyfrompastindividual behaviors.266 But muchlike harmful adthem to act in dictions,visceralcravingscan overpower actors,causing that fail to maximizeutility.Hungercan makethe dieterovereat, ways even thoughhe would ratherlose weight than enjoy a fatteningmeal. Sleepinesscan makethe truckdriverfall asleepat the wheel,even though thecostsof doingso clearlyoutweigh thebenefits. Rational choice theoryleadsto the prediction thateventsthatcreate net will not occur. The actorcan use negative utility utility-maximizing to makesurethattemptations areout of reach precommitment strategies whencravings in potentially addictive arise,just as he canavoidindulging behaviors.267 The dietercan keep only celerystalksin the refrigerator, for trucker can stop for the nightat the first example,and the long-distance signof sleepiness.268 Rational choicecan fail as a behavioral in these situations prediction because actorstend systematically to underestimate the power of such visceral cravingsbefore they occur, which hampersrationaladvanced
265. For a very thoughtfultreatmentof the behavioralimpact of these and other visceral drives, see George Loewenstein, Out of Control: Visceral Influences on Behavior, 65 ORG. BEHAV. & HUM. DECISION PROCESSES 272, 272 (1996). 266. This distinction,while suitablefor the purposesof this article, is perhapsa bit too simplistic, as the intensityof universalvisceral cravingsmight be affected by individualbehaviors. 267. See Schelling, supra note 77. Contrastthe view in the text of the difficulty that actors have with visceral cravings and Schelling's view that individuals are at least rational enough to recognize some of their own limitations and to take actions (such as putting the alarm clock out of easy reach across the room) so as not to be put in a position of being at the mercy of this lack of self-control. 268. Alternatively, a driver might purchase a special device that must be squeezed around the steering wheel. If the driverrelaxes pressureon the device, as he might do if falling asleep, the device will emit a loud sound to startle the driver awake. We learned of this device from a talk by Thomas Schelling, "A Critiqueof RationalChoice Theory,"Society for the Advancementof Socio-Economics, Vienna, July 13, 1998.

1118

LAWREVIEW CALIFORNIA

[Vol.88:1051

Consider the following clever demonstrationof this by planning.269 wererecruited and andhis colleagues. Loewenstein subjects Experimental if they agreedto completea quiz that told they would be compensated would test their knowledge of history. Some subjects were asked to the quiz,whether choose,priorto completing they wouldbe compensated with a large chocolatebar or a specifiedcash payment.Othersubjects to wouldbe butwerenotrequired knewwhattheirchoiceof compensation the quiz. Subjectsin the latter makethe choice untilthey had completed morelikely to select the chocolatebarthansubgroupwere significantly in the first thatpeopleareoftennotverygoodat demonstrating jects group, is imminent. the of their beforea temptation cravings predicting power Thisinsightgoes a longway toward whygovernment poliexplaining to encourage thatallowsindividuals to cies thatattempt or coercebehavior avoidthe ill effects of visceralcravingsare oftenless successfulthanrationalchoicetheorywouldpredict. Facedwithevidencethatmassivepubto encourage condomuse to preventthe spreadof lic healthcampaigns diseaseshave enjoyedlimitedsucAIDS and othersexuallytransmitted of rational choice thecess, Richard Posner,writingfromthe perspective ory, concludesthat for manypeople the benefitsof engagingin unsafe A moreplausible sexualpractices the costs of doingso.270 outweigh explanationof the data,we believe, is thatmanypeople who wouldconclude thatcondomuse is utilitymaximizing for themfail to predict thepowerof sexualcravings.Consequently, are not for the prepared they sufficiently of the when the and arise, theyengagein unmagnitude cravings cravings safesex despitetheirbetter judgment.27 In situations in whichvisceralcravings, like addictions, areespecially in to cause individuals to act to their more considered, likely wayscontrary stablejudgments as to how to maximizetheirutility,an argument can be madeformoreaggressive If ex antegovernment rational choice regulation. sufficient of the would theoryprovided predictions behavior, government need to do no moreto combatthe spread of AIDS thanto makesurethat information the risksof unsafesex was madeavailable to the concerning But in a in world which visceral can overwhelm generalpublic. cravings considered it mightbe moreefficaciousto require the installajudgment, tion of condommachines in publicrestrooms did in or, as San Francisco

269. See Loewenstein,supra note 265, at 281-82. 270. RICHARD A. POSNER, SEXANDREASON 114 (1992) ("'Safe sex' is not a perfect substitutefor unsafe sex ...."). 271. Cf. Loewenstein, supra note 265, at 279 (predictingthat "[i]f food, pain killers, or sex have undesirableconsequences [actors]will plan to desist from these behaviors.When these visceral factors arise, however, and increase [their] momentaryvaluation of these activities,... [actors] will deviate from [their]priorplans.").

2000]

SCIENCE LAW ANDBEHAVIORAL

1119

the spread for encouraging notorious the early 1980s,to close bathhouses of thevirus.272
D. TimeInconsistenciesand the Multiple-SelvesProblem

to compareshort-term Manyof life's decisionsrequireindividuals to spend versuslong-term benefits.Forexample,we mustdecidewhether whether moneytodayor investit so we will havemoreto spendtomorrow; or laborlate into the nightin to leave workearlyto enjoy the afternoon to eat an ourjob securityor earninga raise;whether hopesof increasing or to makedo with ice creamcone afterdinner for immediate gratification in the future. carrot sticksin orderto be thinner Becauseof the timevalue to of money,rational actorswill discount future income.273 Thatis, in order morethana dollartogive up a dollartoday,theywill demand something In theorythese samecalculations shouldapplyto the futureenmorrow. joymentof goodsotherthanmoney,so thatthosefuture goodsshouldalso be discounted to present valuefor decision-making for purposes, although somewhatdifferentreasons.A day of vacationtoday shouldbe worth somewhat morethana day tomorrow, for example,sincethereis a chance thattomorrow never come. However,becausedifferentindividuals may havedifferent for present overfuture to (referred preferences consumption as their"rateof time preference"), different individuals have might very differentdiscountrates.If investments earneda risk-free10%annualrewouldinvest a largeportionof theirsalarywhile turn,some individuals otherswouldinvestnothing,andmembers of bothgroupscouldbe maxitheir in so utilities mizing doing.274 Althoughdifferentdiscountrates acrossindividualsare consistent with (at least thin conceptionsof) rationalchoice theory,275 that theory
272. See RANDYSHILTS, AND THE BAND PLAYEDON (1988) (discussing the controversy surroundingthe decision of the San Francisco Public Health Director to close public bath houses in 1980 to help preventthe spreadof AIDS and other STDs). 273. See KATZ& ROSEN, supra note 258, at 164 (stating that the present value of an amountof money, M, to be received t time periods in the futureis equal to the dollar amountof the futuremoney divided by one plus the discountrateraisedto the t power, PV = M / (1 + )t ). 274. Any discount rate that an individual might have is consistent with thin versions of rational choice theory, so long as the individual does not exhibit inconsistent discount rates. However, if we "thicken"rational choice theory just enough to predict that people will prefer more money to less money (see infra Part IV for a discussion of this mildly "thick"assumption)then a negative discount rate is irrational (even though a positive discount rate of any magnitude could be rational). Many taxpayers routinely have too much income tax withheld during the year so that they can receive a refund from the InternalRevenue Service after filing their tax returnsin the spring of the following year. This overwithholdingconstitutesan interest-freeloan to the federalgovernmentand suggests that the taxpayerhas a negative discount rate. The taxpayerwould have more money, and would suffer no offsetting losses in utility, if he reducedthe amountwithheldand invested thatextracash on handat the risk-freerate of return. 275. Thick versions of rational choice theory that presume wealth maximization predict that all individualswill have the same discount rate: the marketrate of return. See, e.g., George Loewenstein & RichardH. Thaler, Anomalies: IntertemporalChoice, 3 J. ECON. PERSP. 181, 191-92 (1989) ("The

1120

CALIFORNIA LAWREVIEW

[Vol.88:1051

predictsthat any individual'sdiscountrate will be invariantto the length of time for which he must wait for his money. For example, an individual who would prefer$10 today to $11 dollars in exactly one year should also prefer$10 five years from now to $11 six years from now. The interestrate in both cases is 10% per year, and any individual should have a stable preferencefor earning that rate of returnor enjoying earlier consumption. The behavioralscience literaturedemonstratesthatthis assumptionis often false. Rather, discount rates often decline as the date of the reward recedes.276

This means that, for many people, preferences between logically identical sets of choices may reverse in a predictabledirectionas the temporal context of the choice changes. Suppose that an individual is to choose between ProjectA, which will maturein nine years, and ProjectB, which will mature in ten years. Suppose, further,that an individual who compares the two projects across all their different dimensions prefers ProjectB to A. Now suppose that we bring the dates of maturityof the two projects forwardwhile maintainingthe one-year difference in their maturity dates. Because discount rates increase as maturitydates get closer, it is possible that the individual's preference will switch from Project B to ProjectA as the dates of maturitydecline (but preservingthe one-year difference). If discount rates were constant over time, this switching would not occur.277 An implication of time-inconsistent discount rates is that will "people always consume more in the present than called for by their An individualmight plan to save X percentof her salary previousplans."278 next year but then decide when she receives it that she prefers to spend it ratherthan save (thus making appropriate the cliche that money can "bur a hole" in one's pocket). How the law should respond to time inconsistencies depends on the philosophical question of whetherthe preferencesof an earlier or later period of time deserve more respect.279 Perhaps the most obvious legal

standardview is that the marketrate of interest,correctedfor tax distortions,representsan aggregation of individualtime preferences,and is the appropriate social rate of time discounting."). 276. See id.; see also Uric Benzion et al., Discount Rates Inferred from Decisions: An Experimental Study,35 MGMTSCI.270 (1989); RichardThaler,Some EmpiricalEvidenceon Dynamic LETTERS 201 (1981). Inconsistency,8 ECON. 277. See RICHARD THE WINNER'S CURSE 96-98 (1992). Another effect regularly THALER, observedin experimentswith choice over time is that the discountrate for gains is much largerthan the discount rate for losses, what Thalerrefers to as "debtaversion."Id. at 100. This finding is a close kin of the loss aversionnoted above. See supra note 231 and accompanyingtext. 278. Id. at 98. Robert Strotz first discussed this myopia, as he felicitously termed it. See R. H. Strotz,Myopiaand Inconsistencyin Dynamic UtilityMaximization,23 REV.ECON. STUD. 165 (1955). 279. Cf. Loewenstein & Thaler,supra note 275, at 186 ("The problem of dynamic inconsistency raises questions about consumer sovereignty. Who is sovereign, the self who sets the alarm clock to rise early, or the self who shuts it off the next morning and goes back to sleep?"). See also the thoughtfuldiscussion of the legal implications of these mattersin Posner, supra note 30 (showing the

2000]

LAWAND BEHAVIORAL SCIENCE

1121

implications of temporalinconsistencies come in the areas of tax and savings policy. In the United States, the Social Security system effectively forces workers to save a percentageof their income for retirement.In addition, the governmentprovides billions of dollars of tax incentives each year to encourage individuals to invest even more of their incomes in retirement savings accounts, such as IRAs, Keoghs, and 401(k) plans. From the perspectiveof rationalchoice theory, as traditionallyunderstood, this combinationof coercion and subsidies should lead to an inefficiently high amountof savings, because people who would prefercurrentto future consumptionare coerced or bribedto forestall that utility-maximizingcurrent consumption.280 However, if the forward-looking self that plans to invest is presumedto representthe more stable and consistent preferences of the individual than the present-orientedself that changes course and decides to spend, these policies can give the former self the advantageit needs to win the struggle against its less thoughtful counterpart.Under these assumptions,either forced savings or subsidies to encourage savings could enhance efficiency in the consumptionof resources.281 Two interesting twists on the problem of time-inconsistent discount rates deserve mention. First, like the length of the time delay, the size of the rewardat issue appearsto affect discountrates. The generalproblemis that people perceive the difference between $100 today and $150 in a year as greater than the difference between $10 today and $15 in a year, and thus will exhibit a higher discount rate in the latter case than in the forThat is, many people are willing to wait for the extra $50 in the first mer.282 instancebut not for the extra $5 in the second instance. Shefrin and Thaler have proposed that the explanation lies in how people take mental account of small and large windfalls. They hypothesize that small windfall gains are put into a mental account that allows for immediate consumption,while large windfall amountsare put into a separate mental account for which there is a much lower propensity for
"child"self and a legal ramificationsof supposingthat we each have an impulsive, short-term-focused longer-term,more contemplative"adult" self). 280. Cf. DeborahM. Weiss, PaternalisticPension Policy: Psychological Evidenceand Economic Theory, 58 U. CHI. L. REV. 1275, 1276 (noting that paternalisticconcerns that underlie government support for pension policies are "difficult to reconcile with the assumptions that underlie economic analysis"). Requiringor subsidizing savings could still be socially efficient under the assumptionsof rational choice theory if the external costs associated with immediate consumption are large. For example, if too much currentspending will lead to destituteretirees whom the state will be forced to support,it might be efficient to encourage savings even if the saving is not utility maximizing to the saver. This analysis might explain minimum savings requirements,such as with Social Security,but it seems implausibleto explain the extent to which the U.S. tax system subsidizesretirementsavings. 281. Cf id. at 1282-83 (noting that Congressional support for pension policies is based on the belief that people are unableto make wise savings decisions for themselves). 282. See Loewenstein & Thaler, supra note 275, at 187 (explaining that the difference may be attributable to the categorizationof small future gains as foregone consumptionbut large future gains as forgone interest).

1122

CALIFORNIA LAWREVIEW

[Vol.88:1051

cost of waitingfor a small immediate Thus,the opportunity consumption. windfallmay be perceived to be foregoneconsumption. But the opportuwill be interest of for a windfall or investcost nity waiting large foregone is more difficultto resist thanforegone ment. If foregoneconsumption or investment, interest thatwouldexplainthe observed effectof the size of the award rate.283 If ShefrinandThaler's causinga declinein the discount is correct,it suggeststhatthe forward-looking motivational explanation selvesof the wealthy aremorelikelyto prevail in thebattlewiththeirpresent selves thanthe forward-looking selves of the poor.Thiscouldcounsel for differentially focusingsubsidiesfor savingson the poor and middle class. of the statusquo bias,individuSecond,in an interesting application als' discount rateshave also been shownto varywhenexperimenters mathe reference from which versus nipulate point consumption savings choices areevaluated. In one experiment, for example,Loewenstein provided subjectswith gift certificatesto a recordstore that could be redeemedat one of a varietyof specifiedfuture dates.He thenelicitedfrom the subjectshow muchthe value of the certificate wouldhave to be enhancedfor themto agreeto waitlongerto redeem it, andhow muchof the valueof the certificate theywouldbe willingto give up in orderto be permittedto redeemit earlier.24 No matter how distant the initialredemption morethantwice as muchto delaythatdatethan date,subjectsdemanded dateby the same they were willing to foregoto speedup the redemption amount of time.28 One generalconsequence of this is that,if it so desires,the governmentcanprobably the of benefitsit distributes encourage conservation by them as redeemable at a certain future framing pricein a specified year,or at an increased thana givenpriceat the eartime,rather priceat an earlier lier time or a decreased in a future price period.For example,the status bias in decision quo temporal makingsuggeststhatconservation mightbe betterservedby the government's valuablerightsto harvest transferring timberon federallandsto a lumber in the formof (1) a certain company number of acreseach year or (2) fewer acresif the companyprefersto harvestthe treessooner,rather thanin the formof a (1) totalnumber of acresor (2) moreacresif the company to out its agrees spread harvesting overtime. The problems raisedby time-inconsistent discountratescan be seen as a specific exampleof a largerissue, sometimescalled the "multiple

283. 284. 285.

See Hersh M. Shefrin & RichardH. Thaler,The Behavioral Life-CycleHypothesis, 26 ECON. See Loewenstein & Thaler,supra note 275, at 187-88. See id. at 188 tbl.1.

INQUIRY 609 (1988).

2000]

LAW ANDBEHAVIORAL SCIENCE

1123

selves"problem.286 Eachindividual, at any given pointof time, mightnot be the unitary, coherentset of preference orderings imaginedby rational choice theory.Rather,each individual be viewed as a collectionof may If so, thentheremaybe a collectiveaction competing preference orderings. in the of thesemultiple problem aggregating contemporaneous preferences
selves.287

Whileinnate andage mayplayimportant roles judgment, experience, in achieving coherence the contriblaw, selves, too, amongmultiple might ute. Forinstance, wherethereis general aboutwhichaspectsof agreement the competing withinan individual preference orderings oughtto be enthen law can construct that the more desirable couraged, policies give "self' an upper handin determining the individual's courseof action.The is potentially withregard to everydecimultiple-selves problem troubling of thechoicebetween and consion,butwe use theexample present future to illustrate the issue. A stiff tax on to take an obvious sumption cigarettes, self in its battlewith example,can be viewedas aidingthe future-oriented a morepresent-oriented self thatvaluesimmediate overlonggratification termhealth.288 Rulesthatpromote or impedeeffortsof individuals to make future commitments areanother meansby whichlawmakers can lendsup"self."Today'sself can attempt to makecommitments portto a particular thateitherwill completely bindtomorrow's self or, at least,raisethe cost of takingactionthattoday'sself wishesto avoid.A parent mightwish to establisha spendthrift trustfor a child as a methodof stiflinga potential future A healthy urgeto spendthemoneyon immediate consumption. person mightwish to pen an "advanced directive" thatspecifiesthe type of treatment he wishes to receive in the event of futureillness. Thomas thata drugclinicin Denveruses "self-blackmail" as part Schellingreports of its rehabilitationof drug addiction.289 The client writes a selfaddressees if letter,whichthe clinic will mail to designated incriminating
286. See generally Posner, supra note 30 (describing a host of policy problems presented by conceiving the individual as a collection of personal preferences rather than as a unitary set of preferences). 287. A "collective action problem" arises when the high costs of coordinating action among a group of people preventthem from achieving some collective goal. For example, everyone might agree that they would be collectively better off if everyone reduced their stock of weapons. But few individuals are willing to dispose of their arms unless everyone else does so, too. These individuals may need the help of an outside party such as the governmentto craft an agreementthat can compel and enforce the arms reduction. Robert Frank describes another interesting example of a collective action problem: the race to purchase more conspicuous consumer goods so as to signal one's social status, a race that he argues can only be stopped by a consumption tax. See generally ROBERT H. LUXURY FEVER: WHYMONEY FAILS TOSATISFY IN AN ERAOFEXCESS FRANK, (1999). We discuss these issues furtherinfra PartIV.C. 288. See Posner, supra note 30, at 20 (explaining that rational actors may overvalue present pleasure with respect to futurepain by viewing individuals as a series of rationalselves, each with its own interests). 289. See Schelling, supra note 77, at 7 n.4.

1124

CALIFORNIA LAWREVIEW

[Vol.88:1051

the client fails in his rehabilitativeefforts. For instance, a physician might write such a letter to the State Board of Medical Examinersconfessing that he has violated state law and professional ethics by using cocaine and deserves to lose his medical license.29 Should the law enforce trusts or advanced directives that the actor wishes to make "irrevocable" at the time of establishment if he later clinic face liability if it refuses changes his mind? Should the rehabilitation a requestfrom a relapseddrug addictto returnhis self-incriminating letter? The proper answer to these questions might rest, at least in part, on whether we believe that the present self or the future self best represents the preferencesof the single individual. In all likelihood, this will have to be a situation-specificjudgment. We might believe, for instance, that tomorrow's self should prevail over today's self when the issue is health care choices. This is because tomorrow'sself will have more available,relevant informationthan today's self, since today's self is healthy. But it is possible that ourjudgment might be differentwhen we consider the parentwho wishes to form a spendthrift trustfor his child. E. SunkCosts

One of the most famous dictates of expected utility theory is that irreversible past actions should not influence currentchoices.291 Since actors are expected to make choices based on their net expected outcomes, and since irreversiblepast actions cannotaffect thatcalculus, "bygones"should be "bygones."For instance, a supplier,in making a decision about whether or not to expand productiontoday, should focus solely on the profit potential less the variablecosts of that expansion and ignore any fixed costs previously incurred.A person should not stop at the gym on the way home from work merely because he has paid a hefty annualmembershipfee; he should stop only if he expects that the utility derived from a workout will exceed the utility derivedfrom getting home earlier. Notwithstandingeconomic wisdom to the contrary,people routinely cite sunk costs as a reason for pursuinga particularcourse of action. People choose to attendthe theateror an athleticevent when they would prefer not to on the ground that the sunk cost of the ticket would otherwise be wasted. Consider the following famous experiment by Richard Thaler.292 Thaler advertised a $3 "all-you-can-eat"lunch at a local pizza restaurant near Cornell University in Ithaca,New York. He then collected the admission price from enough customers to fill every table in the restaurantand
290. See id. 291. See, e.g., POSNER, supra note 26, at 7-8. 292. See RICHARD H. THALER, RATIONAL ECONOMICS 11-13 and especially 12 n.8 (1991); QUASI H. FRANK, Thaler,supra note 228, at 39; see also ROBERT MICROECONOMICS ANDBEHAVIOR 13, 22627, 231-32 (1991).

2000]

LAWAND BEHAVIORAL SCIENCE

1125

remittedthe $3 admissionprice to half the customers,selected randomly.293 Rationalchoice theory would predictno significant difference between the amounts of pizza consumed by those who did and those who did not receive a refund, but Thaler observed a significant difference. Diners who received a refund ate less,294 presumablybecause they had no need to jusof the cost the meal. tify In some instances, an actor's failure to ignore sunk costs might constitute what could be called a decision-making "error"-the actor fails to recognize that as a factual mattersunk costs don't affect relative outcomes of choices-but such behaviormight also serve as a useful heuristicor sensible precommitmentdevice. People often pay heed to sunk costs because And while the desire for consistency can be they want to act consistently.295 foolish in the face of changed circumstances,it can also be quite sensible when our past actions are based on reliable evaluations of costs and benefits.296 By purchasinga ticket to the opera in advance,we might send a signal to our future self that we would gain a great amount of utility from attending the performance,obviating the need to conduct a cost-benefit analysis on the day of the performance, when we might be tired, overworked, or distracted.In either case, the failure to ignore sunk costs can cause an actor in particularcases to make choices that do not maximize expected utility, such as when he uses his prepaidopera ticket even though he would derive more satisfaction from pursuing a different activity that evening. The propensity of actors not to ignore sunk costs could serve as a valuable asset to legal policymakers seeking to encourage socially desirable conduct. One principlemight requireprepaidcopaymentsfor government-subsidized services with positive externalities. For example, Medicaid benefits might be conditioned on a small prepaymentby beneficiaries for preventive medical services with public health implications, such as immunizations, thus using the sunk cost principle to encourage beneficiaries to take advantageof subsidized immunizations.Or payments for educationmight be paid priorto a course of study, ratherthan on a periodic basis, so as to induce the payerto feel compelled to "get his money's
worth."297

293. See Thaler,supra note 228, at 48 n.8. 294. See id. 295. See ROBERT B. CIALDINI, INFLUENCE: SCIENCE AND PRACTICE 50-93 (3d ed. 1993). 296. See id. at 52-53 (observingthat consistency "allows us a convenient,relatively effortless, and efficient methodfor dealing with the complexities of daily life that make severe demandson our mental energies and capacities"). 297. It should be obvious that these devices are also available to, and indeed, frequentlyused by, private parties, as demonstrated by the examples of opera tickets, fitness center memberships,and the like.

1126

LAW REVIEW CALIFORNIA

[Vol.88:1051

The sunkcost fallacymightalso haveprivatelaw implications. Sellcontracts such thatthe ers, especiallyof consumer goods, often structure to make but can is stop makingpaymonthlypayments buyer obligated at anytime.Thesedeals,we believe,are themerchandise mentsandreturn to takeadvantage of commitments to sunkcosts. oftenconsciousattempts is made,the purchaser is unlikelyto disOncethe firstmonthlypayment the merchandise, even if themarginal andreturn cost continue the contract benefithe receives is higherthanthe marginal of keepingthe merchandise resultin believethatsuchcontracts fromit. To the extentthatlawmakers to those lawmakers consumers' maximize their utility, might failing many contracts restrictions on the way consumer consider maybe implementing structured. IV
DEVIATIONSFROMSELF-INTEREST

evidencethateven thin PartsII andIII of this articlehave presented of rational choice as by the expectedutilconceptions theory, exemplified of generalapplicaas behavioral descriptions ity version,areimplausible as been thin of rational has versions choicetheory But, mentioned, bility. contain no face an additional because of they theory the goalsor problem: of a decisionmaker, value.Evenif preferences they have littlepredictive actorsdid make all decisions consistentwith the completecost-benefit wouldhave no analysisimpliedby expectedutilitytheory,policymakers basis for predictingwhat behaviorswould resultfrom changesin legal choicetheoryusefulfor legalpolicy,it mustconpolicy.To makerational tainsomeprediction aboutactors'"ends," notmerelypredictions aboutthe "means" theywill use to achievethem. The most mild, andgenerally aboutacunobjectionable, predictions tors'substantive are whether actors will view a certain itemas preferences a "good" or a "bad." Thesepredictions are,in fact, so mild andunobjectionablethatthey are almostalwaysimplicitandusuallynot even recognized as predictions aboutpreference structures at all. Forexample,lawand-economics analysisalwaysassumesthatpeoplewill view moneyas a whichmeansthattheywill prefermoreof it to less. The opposite "good," is madeaboutactors'perception of time in prison:this is alprediction understood to be a bad,meaningthatactorswill preferless prison ways timeto more. itemsas "goods" or "bads" Classifying providesenough"thickness" to rational choicetheoryto permit somesimplepredictions to be made.For leadsto the prediction thatraising example,assuming moneyis a "good" taxeson anitemwill resultin reduced sales.Assuming thatprison timeis a "bad" leads to the prediction thatincreasingthe severityof prisonsentenceswill causea reduction in crime.Unfortunately, a "thicker" theoryof

2000]

LAWAND BEHAVIORAL SCIENCE

1127

for rational is necessary choicetheoryto leadto moreprecise preferences and sophisticated In an effortto generatebehavioral predicpredictions. in law-and-economics tions, rationalchoice theoryis often understood to predict thatactors'utilityfunctions will be basedentirely on scholarship whatis in theirself-interest. This statement canincorporate all actionsima tautological definition of "self-interest" as beinganyact ply by adopting or thingthatgives well-being to theactor.Onthisreading, thereareno acts to self-interest. Even a seeminglyselfless act is, whenexamined contrary critically, reallyan actthatconfersbenefitto the actorprecisely by conferbenefit to it is another else done in (or ringuncompensated hopesof getin Understood this the "self-interest" version ting something return). way, of rationalchoice theoryis no differentfromthe "expected utility"version: thatis, it contains no theory of theactors'preferences or goals,andit canleadto few if anypredictions about future behavior. Forthe "self-interest" versionof rational choicetheoryto achievethe "thickness" to have it mustgo out on a limb value, necessary predictive and specify certainsubstantive that behaviors,capableof observation, would be inconsistent with the theory.For the sake of predictivevalue, rational choicetheoryas applied to law oftenis basedon the prediction, be it implicitor explicit,thatactorswill seek to maximizetheirselfishinterests. Although we applaud to makepredictions attempts by legal scholars abouthow actorssubject to the laws will respond to legalrulesandinstituversionof rationalchoice theorymakespredictions, the "self-interest" tionsthataregrosslyimplausible as a general matter. ThisPartexamines a few reasonswhy the predictions are implausible-becausepeople act in accordance with social norms,298 becausethey value fairness,299 and becausetheycontribute to publicgoods300-and examinessomeof the implicationsof thesefindings forlegalpolicy.
A. Social Norms

It shouldnot be controversial to assertthatindividuals' behavioris often informedby social norms,definedas social attitudes that specify whatbehaviors an actoroughtto exhibit.301 Forthis statement to contradict the thickversionof the rational choicetheory,it mustbe the case thatin some circumstances the behavior thatresultsfromcompliance with social normsdiffersfromthe behaviorthatrationalself-interest would dictate. Whentwopedestrians eachotherheadon, theyeachmaycomply approach witha socialnormof stepping to theright,butcomplying withthe normis
298. See infra PartIV.A. 299. See infra PartIV.B. 300. See infra PartIV.C. 301. See Cass R. Sunstein, Social Norms and Social Roles, 96 COLUM. L. REV.903, 914 (1996) (defining social norms as "social attitudes of approval and disapproval, specifying what ought to be done and what ought not to be done");see also ERIC A. POSNER, LAWANDSOCIAL NORMS (2000).

1128

CALIFORNIA LAWREVIEW

[Vol. 88:1051

consistent with self-interestedbehavior. Consider, however, the following three examples of norm-conforming behaviors that are problematic for thick versions of rationalchoice theory. a man might First, to borrow a comparisonmade by RobertCooter,302 remove his hat when he enters a furnace room and when he enters a church, but the two acts are different in an importantway. In the furnace room, the man prefersto take off his hat ratherthan keep it on. By removing it, he acts in accordancewith his self-interest. In the church, the man presumablyprefers to leave his hat on (at least we can assume as much if he leaves his hat on inside other buildings), but he removes it in deference to a social normthatmen do not wear hats in churches. Second, consider how a self-interested decision maker would approach the decision of whetheror not to tip his server in a restaurant. Tips are usually left after the meal, and are presumablywithheld until that time to create an incentive for the server to performwell.303 A diner may enjoy good service without leaving a tip, but this carries some risks to even the self-interesteddinerwho might somedayreturnto the restaurant (whatkind of service could he expect then?).304 a self-interested diner However, should certainly not leave a tip if he never intends to visit the restaurant The overwhelmingweight of the evidence, however, suggests that again.305 in least the United States) almost all diners do tip, even when they are (at traveling far from their hometowns and are extremely unlikely ever to returnto the establishment,30 thus creatinga significantanomalyfor the selfinterestversion of rationalchoice theory.307
302. See RobertD. Cooter,DecentralizedLawfor a ComplexEconomy: TheStructuralApproach to Adjudicatingthe New Law Merchant, 144 U. PA. L. REV.1643, 1656 (1996) [hereinafter Cooter,The New Law Merchant];RobertCooter, NormativeFailure Theoryof Law, 82 CORNELL L. REV.947, 954 (1997) [hereinafter Cooter, NormativeFailure]. 303. Notice that a part of this rational choice view of tipping is that, in the absence of ex post paymentin the form of a tip, there would be little incentive for the serverto performwell. The price of the meal, that is, would be invariant to the quality of the service, unless there were some other mechanism (as, say, between the restaurantowner and the employees) to compel quality service. A humorous cartoon in the New Yorkermagazine suggests an even better method of inducing good performance: on the table in front of a solitary diner is a "plate with a few coins on it and a small placardreading, 'Yourtip so far."' 304. See generally William Grimes, Tips: CheckYourInsecurityat the Door, N.Y. TIMES, Feb. 3, 1999, at B1, col. 3 (describingscholarly researchon how people behave in situationswhere tipping is expected or appropriate). 305. In an attemptto explain puzzling (thatis, nonrational) behavior,economist Steven Landsburg concedes, "I do not know why people leave anonymoustips in restaurants." LANDSBURG, supra note 6, at 19; see also Jolls et al., supra note 13, at 1492-93 (discussing the phenomenon of out-of-town tipping). 306. One might also speculate about the behavior of the rational server in these circumstances. Presumably,a rationalserverwould recognize that an out-of-townguest would be far less likely to be a that might occur repeatcustomer and, therefore,far more likely not to tip. To prevent disappointment to the server who works diligently for a customerin the hope of handsome ex post compensationonly to find that the customeris not going to so compensatehim, a rationalservermight greet each customer in some artfulway designed to discover whetherthe patronis from town or out of town. The very fact

2000]

LAWAND BEHAVIORAL SCIENCE

1129

studyof actualpracticesin Finally,RobertEllickson'snow-famous ShastaCounty, for dealingwithdamages fromwanderCalifornia, arising at provides an look the that social normscan cattle,308 ing in-depth power have on behavior.Ellicksonfoundthatranchers and farmers conformed theirinteractions to a strongsocialnormof "neighborliness."309 Thisnorm, as understood a farmer onto whoseproperty a rancher's locally,required cattlehadwandered to notifythe ownerandto carefor the cattleuntilthe ownerwas able to retrievethem,even if this meanthousingandfeeding the cattlefor months.310 the socialnormdiscouraged the farmer Moreover, fromaskingthe rancher him for the expenseshe incurred to reimburse in for the cattle some ranchers reimbursement voluncaring (although paid To evaluate whether the observed behavior was consistent witha tarily).311 farmer's one wouldhaveto know something aboutthe legal self-interest, who werevictimized cattle.Themostsurrightsof farmers by trespassing prising finding of Ellickson's study was that ranchersand farmersin Shasta hadlittleknowledge aboutthe law of wandering cattle,and, County moreimportantly, had no in interest about theirlethey virtually learning A personwho caresfor his neighbor's chattels without knowgal rights.312 whether he has a to do so or a to reimbursement, ing legal obligation right or even havingany interest in knowinghis legalrights,can hardly be said to be seekingto maximize his directself-interest. Two basictheories aremostoftenofferedto explainwhypeopleobey social norms,bothof whichcan be seen as rational in a globalsense.On one account,people value not only the inherent qualitiesof actionsthat of others.313 Comthey mighttakebutalso the esteem,or socialapproval, with social norms earns the actor whereas of violation soesteem, pliance cial norms costs her esteem. In economic terms, social norms can be of as providing a subsidy (in theformof positiveesteem)for some thought

that this does not happensuggests that patronsdo not generally intendto "stiff' serversand that servers do not generallyexpect to be "stiffed."
307. See ROBERT H. FRANK, PASSIONS WITHIN REASON: THE STRATEGIC ROLE OF THE EMOTIONS

16-19 (1988). 308. Robert C. Ellickson, Of Coase and Cattle: Dispute Resolution Among Neighbors in Shasta
County, 38 STAN. L. REV. 623 (1986). The study is expanded in ROBERT C. ELLICKSON, ORDER
WITHOUT LAW: How NEIGHBORS SETTLE DISPUTES (1991).

309. Ellickson,supra note 308, at 673 (explainingthat trespassvictims do not view errantcattle as trespassers). 310. See id. (explaining that a telephone call was the customaryresponse to someone else's cattle strayingonto one's property). 311. See id. at 674-75 (stating that many ranchershad never heardof anyone's charging or being chargedfor such boarding). 312. See id. at 668-70 (explaining that most ranchers had a limited and oversimplified view of trespasslaw). 313. See RichardH. McAdams, The Origin, Development,and Regulationof Norms, 96 MICH. L. REV.338, 355-56 (1997) (assumingthat people seek the esteem of others).

1130

CALIFORNIA LAWREVIEW

[Vol.88:1051

behaviors while imposinga tax (in the formof negativeesteem)for others.314

we mightinterThus,whenwe see a manremovehis hat in church, from his hat the he his behavior as evidence that utility enjoys wearing pret is outweighed by the negativeutilityhe wouldsuffer(in the formof the withsocialconloss of esteem)shouldhe fail to removehis hatconsistent vention.And whenwe see a dinerleave a tip in a restaurant to whichhe will never return,we mightinterpret her behavioras evidencethat she fearsthe loss of esteemshe wouldsuffershouldherfriendsandneighbors Thistheorysuggeststhatactors'relilearnthatshe failedto tip herserver. ance on social norms bears a certainrelationshipto the problemsof boundedrationalitydiscussed in Part II: obeying social norms, like adoptingsimplifieddecision strategiesand relying on decision-making rational buthas the consequence heuristics, mightbe a "globally" strategy of causing individuals to makedecisions thatarenotlocallyoptimal. A competing view positsthatpeopleobey socialnormsthatareconbecauseactorsinternalize the normsof their traryto theirdirectinterests communities.315 to this view, the cost of violatingsocialnorms According is not loss of esteem in the eyes of peers but guilt or shamefor doing the actorexperiences as "wrong" (thebenefitto be gainedfrom something with social normscan be referred to as "pride"). The costs of compliance social norms are not but the violator himviolating imposed by society by self.316 If a man takes off his hat in an emptychurchas readilyas in a crowded one, his behavior mightbe betterexplained by the internalization than the esteem if a dinertips her explanation by explanation. Similarly, serverin a restaurant in a community in which she knows no one and whichis manymiles awayfromherhometown, we mightsuspectthatinternalization is thebetter behavioral Theinternalization explanation. theory that social norms bears a to the contextual suggests obeying relationship effectson decisionmakingdiscussed in PartIII: existingnormsmaybe a contextualfactor that affects individuals'construction of preferences,

314. See generally Sunstein,supra note 301, at 935 (statingthat norms"helpidentify the costs and benefits of actions"). 315. See, e.g., Robert Cooter, Expressive Law and Economics, 27 J. LEGAL STUD.585, 585-86 (1998); Cooter, The New Law Merchant supra note 302, at 1662; Robert C. Ellickson, Law and EconomicsDiscovers Social Norms, 27 J. LEGAL STUD.537, 539-40 (1998). 316. Cooter calls this the difference between "principled"conformity to a norm (conformity caused by internalization) and "adventitious" conformity(conformitycaused by the externalbenefits to be derived). See Cooter, The New Law Merchant, supra note 302, at 1667. Ellickson frames the difference as one between first-partyand third-party enforcementof norms. See Ellickson, supranote 315, at 547.

2000]

LAWAND BEHAVIORAL SCIENCE

1131

which rationalchoice theory presumes to be exogenously determinedand fixed.317 In many cases, a social norm might derive its power from both the That is, a man might desire for social approvaland from interalization.318 remove his hat in a churchand a diner might leave the server a tip because they would fear the imposition of social sanctions and feelings of guilt should they act otherwise. This could make social norms doubly powerful as a determiningcause of behavioralchoices. When individual behavior motivated by social norms differs from what would be in the actors' direct self-interest,there are a numberof potential implications for legal policymakers.First, the existence of a social norm supportingor undermininga particulardesired behavior can affect whetherand to what extent policymakersneed to employ law to encourage the behavior. Some law-and-economics analysis has concluded that laws in employmentare unnecessary, prohibitingracial or genderdiscrimination because in a competitive marketindividualsand firms have a strongprofit motive to hire the best workers available-in other words, direct selfinterest will prevent discrimination, at least in the long run.319 But in a world in which there is a social norm supportiveof racial discrimination, strong legislation imposing penalties for discriminating on such bases, which can be viewed as a discrimination "tax," might be necessary to eradicatethis behavior.320 In other words, law might be used to encourage individualsto violate inefficient or undesirablesocial norms.321 Second, rather than attempting to support or impede social norms, policymakers might attemptto shape social norms throughlaw and other forms of public policy.322 The primarydeterrenteffect many laws have on undesirable behavior might not be the direct increase in the price of the behaviorthroughthe threatof fines, civil liability, or jail sentences,but the
317. Cf. Sunstein, supra note 301, at 913 (assertingthat "preferencesare constructed,ratherthan elicited, by social situations, in the sense that they are very much a function of the setting and the prevailingnorms"). 318. See Cass R. Sunstein, On the Expressive Function of Law, 144 U. PA. L. REV.2021, 2031 (1996) (positing that choice depends on the intrinsicutility of the options, the reputational utility of the options, and the effects of choice on the actor's self-conception). 319. Gary Becker has arguedfor allowing changing social realities to cause a discriminatory norm to erode over time. In his model, those who exercise an unwarranted taste againstcertaingroupsbear a cost for doing so and, if there is enough competition in the marketplacein which they are operating, those costs will eventually become too much to bear. Real forces of profit, loss, and utility will cause social conventionsto change. 320. Cf Cooter,NormativeFailure, supra note 302, at 977-78 (statingthat the law can destabilize discriminatory practicesby protectingpartieswho cease to follow a group's discriminatory norms). 321. See, e.g., Eric A. Posner,Law, Economics, and InefficientNorms, 144 U. PA. L. REV.1697, 1728 (1996). 322. Cass Sunsteinrefers to this as "normmanagement." Sunstein, supra note 301, at 907. For an insightful discussion of the difference between law's direct effects on behavior and its indirecteffects on behavior throughits impact on social norms (and other inputs to behavior), see Lawrence Lessig, The New Chicago School, 27 J. LEGAL STUD. 661, 662-72 (1998).

1132

CALIFORNIA LAWREVIEW

[Vol.88:1051

of a social norm againstthe activity. For example,by encouragement the of use a criminal raisesthe offense,the government making marijuana suchuse on the mardecreases use, whichundoubtedly priceof marijuana it but also create norm or reinforce a social gin, might againstmarijuana use thathas an even largereffect on behavior. the selling of By banning babies,the sellingof bodilyorgans,or the sellingof sex, the government a socialnormthatneither the bodynorits partsshouldbe mightreinforce commodified.323 tax a evasion crime rather thana civil violation By making to a fine, the government a socialnormsupportive subject mightcultivate of taxpaying.324 Note thattheseapproaches canbe successfulonly in communitiesgenerallyproneto respectthe law.325 If a subgroup is hostileto statements of law, attempts to manage socialnormsmighthave expressive have a just the oppositeeffect. For example,if teenagersas a subgroup social norm that favors the of general defyingauthority, legal prohibition marijuana might paradoxically encourage marijuana use among
teenagers.326

Justas law mightencourage the development of a normopposedto a of a socially undesirable activity, it might encouragethe development norm supportiveof a desirableactivity. Tax subsidies for Individual Retirement Accountsand401(k)plansmightbe enactedtodayto encourthe desired amount of privateretirement age savings.They also mightbe enacted in an effortto encourage a stronger socialnormin favorof saving forretirement, whichcouldreduce theneedforthe subsidy in thefuture. Perhapsless controversially, policymakersmight use the state's thanits powerto legislateto encourage the develpowerto educaterather of certain norms. Governmental to disseminate informaopment attempts tionon thehealth risksof tobacco the consumption mighthaveencouraged of a social normagainstsmoking.327 development Government-sponsored tobacco education is certainly less intrusive thananexplicitlegalcommand tax on cigarettes, andmayhave as much againstsmokingor a prohibitive
323. See generally MargaretJane Radin, Market-Inalienability, 100 HARV. L. REV. 1849 (1987) (arguing for a relatively wide scope for the legal policy of inalienability on the ground that this will preservedesirablesocial behavior). 324. See Dan M. Kahan,Social Meaning and the EconomicAnalysis of Crime,27 J. LEGAL STUD. 609, 609-10(1998). 325. See TOM R. TYLER, WHYPEOPLE OBEYTHELAW45 (1990) (citing a study where 82% of respondentsagreed with the statement,"People should obey the law even if it goes against what they thinkis right"). 326. See, e.g., Dan M. Kahan, Social Influence, Social Meaning, and Deterrence, 83 VA. L. REV. 349, 375 (1997) (noting that "delinquency is status-enhancing" among gang members because willingness to breakthe law is viewed as a signal of strengthand courage); Kahan,supra note 324, at 611-12 (suggesting that increasingthe severity of punishmentfor carryinga gun at school might have the opposite of its intended effect by "reinforcing the message of defiance" associated with gun possession); Sunstein, supra note 301, at 919 (observing the relevance to legal policy of the fact that "some people like to reject social norms"). 327. Governmentattemptsto portraysmoking as "uncool"or "dirty" might have a similareffect.

2000]

LAWAND BEHAVIORAL SCIENCE

1133

or moreof an effect on socialnormsconcerning The statemight smoking. also shapenormsthrough its role as a participant in variousmarkets. By for implementing nondiscriminatory hiringpractices government employa moregeneralnondiscrimiees, for example,lawmakers mightencourage
nation norm.328

Third,althoughthereis no undisputed theoryof how social norms or some commentators have thatthe exisdevelop change,329 hypothesized tence of a normfavoringa certainbehaviormightbe evidencethatthis behavior is moreefficientthanalternatives,330 or at least more particular efficientthanalternatives thatare likely to be encouraged by lawmakers Unlesstheyhavea clearreasonto engagedin politicaldecisionmaking.331 believea prevailing normis notefficient,legalpolicymakers mightbe wise not to interfere withthe norm.332 Thisprinciple of non-interference canbe eithernot legislatingin a given field, thus operationalized by lawmakers' relying on norms to be self-enforcing,or by codifying the prevailing the prevailing normswith the enforcement norms,thussupporting power of the state.In the areaof commercial for law, KarlLlewellyn'ssupport the latterstrategyled to the creationof the UniformCommercial Code, to codifycommercial which,to a largeextent,attempts tradepractices.333 Lisa Bernsteinhas arguedon behalf of the formerstrategy,contending to ongoing,relational that,at leastwithregard it is wiserpolicy contracts, to leave the enforcementof commercial norms to the commercial
328. See, e.g., LawrenceLessig, The Regulationof Social Meaning, 62 U. CHI.L. REV.943, 100814 (1995); Eric A. Posner, supra note 321, at 1730. 329. Cf. Ellickson, supra note 315, at 550 (admittingthatmany normsscholars have "ducked" the challenge of creatinga theory of norm developmentand reform,"in effect relegatingnorm change to a black box"). 330. See, e.g., Cooter, TheNew Law Merchant,supra note 302, at 1677 (suggesting that efficient equilibria will become norms). Cooter concedes that some norms may develop even though they are inefficient: for example, if the norm of one community enables it to capture benefits while externalizingcosts on anothercommunity.See id. at 1684-85. 331. See id. at 1690 ("My view that failures are rarein business games and norms, and that rentseeking by lobbyists is common, lies behind my claim that much business law should be found, not made, by the state."). 332. In the text accompanyingsupra notes 249-51, we arguedthat courts ought not to assume that default contract provisions are efficient simply because parties continue to contract for them. This might seem to contradict our argument here that courts should defer to social norms, including, perhaps, contractual default norms. As will become clear in our conclusion to this section, the deference to social norms that we propose is a highly contingent one. Legislators and other legal decision makersmight begin from a presumptionin favor of deference,but it should not be difficult to overcome that presumption. Similarly, we would argue that our earlier argument in favor of not deferringto contractualnorms is also a presumptionthat can be easily surmounted.In the end, these two positions are muchcloser thanthey would otherwiseseem to be. 333. See, e.g., U.C.C. ? 2-202 cmt. 2 (proclaiming the importance of commercial practice in driving the rules of commercial law); Cooter, The New Law Merchant,supra note 302, at 1651-52 (noting that Llewellyn's goal in draftingthe U.C.C. was to codify best commercialpractices);cf Grant Gilmore, On the Difficulties of CodifyingCommercialLaw, 57 YALE L.J. 1341, 1341 (1948) (claiming the U.C.C.'s purposeis to codify commerciallaw, not to change the habitsof business people).

1134

CALIFORNIA LAWREVIEW

[Vol. 88:1051

themselvesand to invoke externallegal rules only in rare communities norms break down completely.334 cases when relationship-preserving point,butone thatis relatedin its assumption Makinga slightlydifferent RobertCooterhas argued thatnormsshouldbe respected by lawmakers, thatlaw shouldbe usedto fill gapsin socialnorms,yieldingto the norms wheretheyexist.335 normis likelyto be effiEvenif it is true,however, thatan emergent circumstances render once-efficient a norminefficient, changed might nature of normsmakesthemslow to changein cient.If the decentralized policymakers may play a valuable responseto changingcircumstances, role by enactingcounter-incentives andencouraging normchangein recircumstances.336 sponseto suchchanging Becausewe havetakento taskthe law-and-economics for community too to the of rational behavioral clinging tightly implausible assumptions choice theoryin its traditional evenhandedness that conceptions, requires we give creditwhereit has exhibited a moreflexiblereassessment of the In the last decadetherehas been an outpouring of rationality assumption. fromwithinthetraditional literature law-and-economics on community the of social normsin understanding humanbehavior,337 and we importance have drawnupon much of this work in this section. If the law-andeconomicscommunity embraces to the sameextentthe otherphenomena in this Article,it will be well on its way to developinga more described nuanced and insightful of economically mindedlegal scholargeneration ship.
334. Lisa Bernstein, Merchant Law in a Merchant Court: Rethinking the Code's Search for ImmanentBusiness Norms, 144 U. PA. L. REV. 1765, 1796-1815 (1996). Bernstein's claim is that commercial parties will act in accordance with certain norms in the conduct of ongoing business relationships,and that no legal coercion is needed to enforce these norms in such relationships.Legal rules are necessaryonly when the relationshipbetween contractingpartiesbreaksdown. In these cases, rules that are at odds with commercial norms might be more efficient than norms employed in very differentcircumstances.See id. 335. See Cooter, Normative Failure, supra note 302. In a similar vein, Eric Posner argues that "solidary"groups generally are capable of governing themselves throughsocial conventions and that legal intervention typically (but not always) does more harm than good. See Eric A. Posner, The Regulationof Groups: The Influenceof Legal and Nonlegal Sanctionson CollectiveAction, 63 U. CHI. L. REV.133, 136 (1996). 336. See Avery Katz, TakingPrivate OrderingSeriously, 144 U. PA. L. REV.1745, 1750 (1996) and, thus, dependenton historical accident); (noting that social norms might be highly path-dependent Posner, supra note 321, at 1713 (arguing that information lag affects diffuse groups more than legislatures and judges, suggesting that the latter may produce better rules than the norms created by the former). 337. See, e.g., Symposium, Law, Economics, & Norms, 144 U. PA. L. REV. 1643-2339 (1996); STUD. Symposium, Social Norms, Social Meaning, and the Economic Analysis of Law, 27 J. LEGAL 537-823 (1998). As David Charnyexplains, the pervasivenessof "all sorts of conduct that did not seem particularlyrational"caused legal economists to investigate how nonlegal sanctions that attachedto violations of norms affect behavior. David Charny,Illusions of a Spontaneous Order: "Norms" in ContractualRelationships, 144 U. PA.L. REV.1841, 1843-44 (1996).

2000]

AND BEHAVIORAL LAW SCIENCE

1135

B.

Fairness

Oneparticular socialnorm-fairness-merits a discussion of its own because it is so fundamentally contraryto the predictionsof the selfinterestversionof rationalchoice theory.To predictthat actorsbehave notionsof fairnesswill be ignored. selfishlyis to predictthatunenforced Thereis considerable evidenceto thecontrary. Consider howpeopleplaythemuch-studied "ultimatum The game."338 with two who do not know one another and are gamebegins participants not allowedto communicate. Player1 proposesa divisionof a monetary stakeprovided $10 or $20). Player2 thenhas by theexperimenter (perhaps two choices.He mayaccepttheproposed division,in whichcase theplayers receivetheirproposed sharesof the stake.Alternatively, he mayreject the division,in whichcase the stakerevertsto the experimenter andeach receives A self-interested it is player nothing. Player 1, alleged, should the smallest for Player proposea divisionthatprovides possibleincrement 2 andreservesthe lion's shareof the stakefor himself.BecausePlayer2 will realize somethingis betterthannothing,he shouldthen acceptthe division. proposed has demonstrated thatPlayer1 rarelyproposesthe preExperience numberof experidicted,one-sideddivision.Rather,in a wide-ranging mentsovermanyyearsandin manydifferent the modal(thatis, countries, most common)proposalis for a 50-50 split, and the meanproposalhas beenfor a 63-37 split.339 offera less evendivisionif Player1 will generally he is selectedto be Player1 on the basisof an apparently reason objective a preliminary rather than (suchas his havinganswered question correctly) as the result of a coin flip, or if the identitiesof the two playersare shielded fromeachother.However, evenin thesecircumstances Player1 is the mostlopsideddivisionpossibleconsistent withofunlikelyto propose 2 a nonzero feringPlayer payout.34 Thisresult,whileoftenpublicized, is notfatalfor the self-interest version of rational choicetheory. his expected Player1 mightmaximize profit from the game if he correctlyfears thatPlayer2 mightrejecta grossly division,leavingPlayer1 with nothing.Mosttroubling unequal proposed for rational choicetheoryarethe findingsconcerning Player2. Player2, it turnsout, rejectsalmost25% of the proposeddivisions,with divisions

BEHAV. & ORG. 367 (1982); Jolls et al., supra note 13, at 1489of UltimatumBargaining, 3 J. ECON. 93. 339. See THALER,supra note 277, at 22-25. 340. See id. at 27.

338. See, e.g., THALER,supranote 277, at 21-35; Thaler, The UltimatumGame, supra note 24, at 195; Alvin E. Roth, Bargaining Experiments, 282-92, 296-302, in THE HANDBOOK OF EXPERIMENTAL ECONOMICS (JohnH. Kagel & Alvin E. Roth eds., 1995); WernerGuthet al., An Experimental Analysis

1136

CALIFORNIA LAW REVIEW

[Vol. 88:1051

below an 80%-20% split being almostuniformly rejected.341 Havingthe finalmove,he will oftenchoose"nothing" rather than"something," if acmeansaccepting treatment. Thedesireto ceptingthe something inequitable be treated fairly-perhapsbetterlabeledas "pride"-it seems,can trump
self-interest.342

It appears thatthe desireto treatothersfairly can cause deviations fromself-interested behavior just as the desireto be treated fairlycan do the same. Resultsof the "dictator demonstrate this effect. In the game" dictator a divisionof a stakebetweenhimself game,Player1 mustpropose andPlayer2, just as in the ultimatum game.Unlikethe ultimatum game, 2 has no choice but to division.In the however, Player accepttheproposed dictator divisionthandoes game,the average Player1 offersa less-equal the average but he still offers,on average, Player1 in the ultimatum game, a significant of the stake to 2.343 The effect is reduced if percentage Player the contextis manipulated so that,for example, Player1 is toldthereis an thanPlayer2 has the rightto determine the objectivereasonthathe rather allocation,or the "social distance"between the two playersis emphabut it fails to disappear even if the players'identities remain hidsized,344 den from one another.345 It would be naive to supposethat transacting of a bargaining abovetheirown partiesalwaysplacefairtreatment partner but the evidence for profits, suggests that, many people, self-interest maximization canbe somewhat desireto treat tempered by the affirmative othersfairly. The powerof the fairnessnormhas clearimplications for any legal rulethatrelies on assumptions aboutprivatebargaining behavior. As one consider a contested issueof legalremedies:should entitlements example,
341. See Daniel Kahnemanet al., Fairness as a Constrainton Profit Seeking: Entitlementsin the REV.728 (1986). Market,76 AM.ECON. 342. Steve Lubet provides an excellent example of this phenomenonwhen he describes a trip his horse ride into the center of the bazaarwas priced at seven family took to Petra,Jordan.A round-trip Jordanian dinars.Touristswho wished to walk to the bazaarcould bargainwith horse driversfor a oneway returntrip at the end of the day. To Lubet's dismay, no matterhow thin the crowd at the end of the day, no matter how many idle horse drivers, and no matter how hard he bargained, he could not convince a single horse driver to accept less than four dinars for the one-way journey. See Steven Lubet,Notes on the BedouinHorse Tradeor "WhyWon'tthe MarketClear, Daddy?" 74 TEX.L. REV. 1039, 1039-42 (1996). Lubet reasoned that, because the variable cost of providing the trip was low, a horse driver faced little opportunitycost by accepting two or three dinars for a trip, and transactions were done in secret so other horse driverscould not impose formalor informalsanctionson a colleague who undercut the others on price-consequently, he was flummoxed. See id. at 1042-49. His son, however, pointed out that the drivers' pride would be hurt by accepting less than they believed their services were worth.See id. at 1050. 343. See Roth, supra note 338, at 298-302; see also Colin Camerer & Richard H. Thaler, Anomalies: Ultimatums, Dictators, and Manners, 9 J. ECON.PERSP.209, 213-14 (1995). 344. See Camerer& Thaler, supra note 343, at 213-14; Elizabeth Hoffman et al., Preferences, & ECON. BEHAV. PropertyRights, and Anonymityin Bargaining Games, 7 GAMES 346, 362-65, 370-72 (1994). 345. See Roth, supra note 338, at 299.

2000]

LAWAND BEHAVIORAL SCIENCE

1137

rules" (meaning the owner of an entitlementcan be protectedby "property invoke an injunction to protect against its taking) or "liability rules" (meaning that a nonowner can take an entitlementfrom its owner so long as he is willing to pay court-determineddamages)?346 Standardlaw-andeconomics analysis compares the costs of the two alternatives. Liability rules are problematicbecause they lead to more litigation (and its attendant costs), and because courts might set the "price"of a taking too low, thus encouraging inefficient takings, or too high, thus discouraging efficient On the other hand, propertyrules are problematicbecause transtakings.347 action costs (and, as we have seen, status quo bias) could preventefficient transactionsfrom occurring.One prominenttransactioncost is the risk of This negotiations failing due to the problem of "bilateralmonopoly."348 problem is best understood with an illustration: in a situation in which there is only one possible buyer of an entitlementand one possible seller, and the buyer is willing to pay $5 million for the entitlementand the seller is willing to sell the entitlement for $3 million, the sale may not occur (even though it would be efficient) if the buyer holds out for a price near $3 million and the seller holds out for a price near $5 million.349 The fairness norm hints that the problem of bilateral monopoly may be overstated in the traditional analysis. Reliance on the fairness norm might make the division of a fixed surpluseasier to accomplish in practice than traditionaltheory predicts. The end result might be to strengthenthe argumentfor protectingentitlementswith propertyrules ratherthan liability rules.350 Similarly, predictions about how private parties negotiate underlie law-and-economics analysis of whether the remedy of "specific performance" or "expectationdamages"for breachof contractwill lead to the efficient allocation of resources.351 Suppose that Abel has a contractto sell a carton of widgets to Cain, but a third party desires the same widgets.
346. See generally COOTER & ULEN,supra note 8, at 103-06; Calabresi& Melamed, supra note 241. 347. See COOTER & ULEN, supranote 8, at 103-05. 348. POSNER, of a potentially value-maximizing supra note 26, at 69 (stating that the "frustration exchange is the most dramaticconsequenceof bilateralmonopoly"). 349. The possibility that negotiations can fail in these circumstances is frequently attributedto STUD.1, strategicbehavior, and was first discussed in RobertCooter, The Cost of Coase, 11 J. LEGAL 28 (1982) (explainingthat "strategic behaviorsometimesresults in noncooperativeoutcomes"). 350. Note that, ironically, the fairness norm cuts opposite of the endowment effect, which providesan argumentin favor of liability rules over propertyrules. See supra PartV.B. 351. See generally COOTER & ULEN,supra note 8, at 226-35 (explaining the general equality of specific performanceand expectation damages in terms of efficiency); Alan Schwartz, The Case for L.J. 271, 275-76 (1979) (arguingthat specific performanceshould be Specific Performance,89 YALE the properremedy for breach of contractbecause the nonbreachingparty is never fully compensated); Thomas S. Ulen, The Efficiency of Specific Performance: Toward a Unified Theory of Contract L. REV.341 (1984) (arguingfor specific performanceas the routineremedy on the Remedies, 83 MICH. groundthat it will more efficiently protectthe innocentparty'sexpectation).

1138

LAWREVIEW CALIFORNIA

[Vol.88:1051

Court-determined moneydamagesthatwouldmakeCainpreciselyindifwill encourthe contract ferentto Abel's breaching (expectation damages) do so and him from to if it is efficient Abel to breach discourage age Abel will choose to breach if if is inefficient.352 it only theprofit breaching Theproblem is that fromdoingso is enoughto pay therequisite damages. inefficient courtsmighterr in settingthe damagesamount,encouraging efficientones (if breaches(if damagesare set too low) or discouraging areset too high).353 damages will encourage The remedyof specific performance just the right costs are low. Underthis remedy,Abel amountof breachif transaction withoutCain'spermission-thatis, an agreecannotbreachthe contract If Cainvaluesthe widgetsat $100 andthe mentto rescindthe contract.354 thirdpartyvaluesthemat $150, eitherAbel will pay Cainbetween$0 and $50 to rescindthe contract (allowingAbel to sell to the thirdparty)or the thirdpartywill pay Cainbetween$0 and$50 plus Cain'scosts for Cainto withspecificperformance is thatAbel and resellthe widgets.Oneproblem would and thethirdparty. Cainface a bilateral as Cain monopoly problem, would fail if neither set of parties could reach Specific performance on the between$0 and$50. If, howagreement preciselevel of payment the fairness norm reduces the thatAbel andCain,or Cain likelihood ever, andthe thirdparty,wouldfail to reachan agreement on how to dividethe valuecreated the to $50 surplus by allocating widgets the thirdparty,the for relativeto expectationdamagesbeargument specific performance comesstronger.
C. CollectiveAction

In the familiar"prisoner's dilemmagame,"355 two prisoners who are isolated from each other are given the choice between keeping silent for shorthand) or implicating theircoconspirator (knownas "cooperating," in a crime(knownas "defecting").356 If bothprisoners eachrecooperate, ceives a shortprisonsentence.If bothdefect,eachreceivesa long prison sentence.If one defectsandthe othercooperates, the defectoris released fromcustodyandthe cooperator receivesan extremelylong sentence.357 The interesting featureof the game is thateach prisoner is individually betteroff defecting (becauseeach player is betteroff defectingif his
352. See COOTER & ULEN, supra note 8, at 238-45. 353. See id. at 244 ("The subjectivevaluationof the buyer [on performance]is difficult for courts to estimate.").
354. See Thomas S. Ulen, Specific Performance, in NEW PALGRAVE DICTIONARY OF ECONOMICS AND THELAW, supra note 254, at 481.

355.

The usual cite for an outstandinganalysis of the game is ROBERT THEEVOLUTION AXELROD,

OF COOPERATION (1984). 356. See COOTER & ULEN, supra note 8, at 34-38.

357.

See id.

2000]

LAWAND BEHAVIORAL SCIENCE

1139

defectsandis also betteroff defectingif his coconspirator coconspirator (becauseboth yet they arecollectivelybetteroff cooperating cooperates), The result is that selfcooperatingis superiorto both defecting).358 will alwaysdefect,ensuring interested thateachreceivesa long prisoners rather the shorter sentence than could havereceivedif both sentence, they hadcooperated. The prisoner's to includemorethantwo players, dilemma, expanded in debates as the "collectiveaction"problem.359 emerges public policy Mostfrequently, theproblem thecollectiveprovision concerns of a service or good knownas a "publicgood."A "publicgood"is one thatexhibits nonrivalrous andforwhichthecoststo suppliers of excluding consumption beneficiaries are prohibitively Clean air and national nonpaying high.36 defenseareprototypical themsiexamples:manypeoplecan "consume" and it is not feasible to exclude from their benefits multaneously, enjoying thosewhorefuseto contribute to theirupkeep. Self-interested if left actors, to theirown devices,wouldrefuseto contribute to theprovision of a public in theprisoner's dilemma will good,for the samereasonthateachprisoner defect.If everyone else (ormostothers)contributes to the publicgood,the self-interested actorcan "free-ride"-that is, enjoyequalbenefitswithout of the costs. If no one else to (or few others)contributes any shouldering the publicgood,the self-interested actoris still betteroff not contributing, becausethe publicgood presumably does not have enoughvalue to the actorto makeit worthpayingthe cost himselfof providing it to all other If is all will refuse to leavcontribute, citizens.361everyone self-interested, worseoff thanif eachhadcontributed his orhershare.362 ing everyone The self-interested versionof rational choicetheoryleadsto the predictionthatno publicgoods will be produced as a resultof voluntary behavior.363 It follows that an important role of government is to use its coercivepowersto guarantee the production of publicgoods thatmake off thanit wouldbe without theirproduction. societybetter Thereis a good deal of experimental evidence,however,thatpeople do willinglyandvoluntarily contribute to theproduction of publicgoods.364
358. See id. 359. See generallyMANcuR OLSON, THELOGIC OFCOLLECTIVE ACTION: PUBLIC GOODS AND THE THEORY OFGROUPS (2d ed., 1971). 360. See COOTER & ULEN, supra note 8, at 42 (defining public goods as a commodity with the characteristics of nonrivalrous consumptionand nonexcludability). 361. See id. 362. See id.; OLSON, supra note 359, at 2 ("[E]ven if all of the individuals in a large group are rationaland self-interested,and would gain if, as a group, they acted to achieve their common interest or objective, they will still not voluntarilyact to achieve that common or groupinterest."). 363. See COOTER & ULEN, supra note 8, at 42-43. 364. See generally THALER, supranote 277, at 9-11; John 0. Ledyard,Public Goods: A Surveyof OFEXPERIMENTAL 111 (John H. Kagel & Alvin ECONOMICS ExperimentalResults, in THEHANDBOOK E. Roth eds., 1995).

1140

CALIFORNIA LAWREVIEW

[Vol.88:1051

Consideran experimentsometimes referredto as the "groupexchange."A group of people, usually college students,are broughttogether,and each is given the same sum of money. They are told that they can invest some, none, or all of that money in something called a "groupexchange." The decision to invest in the group exchange will be secret. That is, one does not know whether or not the other players have contributed.The group is also told that the game operatorwill multiply the total sum invested by the membersof the group by a numberthat is largerthan one but smallerthan the numberof people in the group and will then divide the resulting sum equally among all of the groupmembers,whetherthey have invested in the groupexchange or not.365These rules make the groupexchange into a public good. Because self-interestedplayers can maximize their profits by receiving an equal share of the sum generated by whatever contributions others make, the prediction of rational choice theory is that no one will
contribute.366

To see the logic of the experiment,suppose that there are five people in the group and that each of them is given $5. If no one contributesanything to the group exchange, then there is nothing for the game operatorto multiply and nothing, therefore,for the group to divide. But suppose that only one person contributesnothingand the other four people in our example contributetheir entire $5 to the group exchange. Further,suppose that the group operatordoubles the resulting $20 to $40 and then distributes that sum equally among all five players. Each, therefore,receives $8. The incrementalreturnto the four players who contributed$5 is $3, but that of the player who contributednothing is $8. This logic should be clear to all the subjects. As a result, no self-interested player should invest in the groupexchange.367 Although not everyone contributesto the group exchange in these experiments, a substantial number do. On average, subjects in the experiments contributedbetween 40% and 60% of their initial sum to the public good.368 Manipulating certain aspects of the exercise can affect the percentageof their money that subjects will invest in the public good. Increasing the returnon contributionsto the group exchange increases contributions, as does permitting subjects to communicate with one another
365. See, e.g., THALER, supra note 277, at 10-11 (describing the structureof group exchange experiments);Ledyard,supra note 364, at 111-112 (same). 366. See Ledyard,supra note 364 (observing that contributingnothing is a dominantstrategyfor all players). 367. See, e.g., Ulen, supra note 12, at 493 ("Thepredictionof the theory of rationalchoice is that no one will invest in the groupexchange."). 368. See, e.g., Gerald Marwell & Ruth E. Ames, Economists Free Ride, Does Anyone Else?: Experimentson the Provision of Public Goods, IV, 15 J. PUB. ECON. 295, 299 (1981) (finding subjects contributeapproximately50 percent of the sum they are initially given); Ledyard,supra note 364, at 172-73 (concluding from the examination of a range of public goods experiments that half of subjectsplay the game the way rationalchoice theorypredictsand half do not). approximately

2000]

AND BEHAVIORAL LAW SCIENCE

1141

decisions.369 The evidence is equivocal as priorto makingtheir contribution to whether increasing the number of subjects involved in a single group exchange game increases or decreases average contributionlevels?70Regardless of how the features of the game are altered, the fundamental finding persists: contributionsto the public good are well above what rational choice theory predicts. In one study, even economics students(who ought to know better) contributedto the group exchange, although their average contributionrate (20%) was significantly lower than the rate observed for otherplayers.371 One explanation of the results, not inconsistent with self-interest, is that novice experimentalsubjectsmay not understand the incentives of the game the first time they play it but will learnwith experience.In fact, there is evidence that when multiple rounds of the game are played, contributions decline over time, although they never reach zero.372 But James Andreoni found a surprisingtwist to this observation,which he called the "restart effect."373 When he told players that ten roundsof the game would be played, their contributionrates declined over the course of those ten trials. But when he later announced that the same players would play the game for an additional ten trials, the participationrate rose back to the standard 40-60% rangebefore declining again.374 The best explanation for this curious effect appearsto be that many subjects,perfectly aware of the incentive structures,are initially willing to contribute to public goods, assuming their colleagues will do the same. When others exhibit less-than-altruisticbehavior, the initial contributors reduce their contributions(althoughusually not to zero), but they are willing to optimisticallyassume again full cooperationamong their colleagues when a new game begins. Andreoni's results suggest that contributionsto public goods are not purely altruistic in nature-actors expect reciprocal contributionsfrom their colleagues375-but that actors will often assume that others will reciprocate,even in the face of experience that might suggest otherwise.
369. See Ledyard,supra note 364, at 149-51, 156-58. 370. See id. at 151-55. 371. See Marwell & Ames, supra note 368, at 306. 372. See Ledyard, supra note 364, at 135-41 (summarizingstudies); see also R. M. Isaac et al., Public Goods Provisions in an ExperimentalEnvironment,26 J. PUB.ECON. 51 (1985); Oliver Kim & MarkWalker, The Free Rider Problem: ExperimentalEvidence, 43 PUB.CHOICE 3 (1984). There are, however, contrary studies that find repetition of the game has no effect on contribution levels. See Ledyard,supra note 364, at 147 (citing studies). 373. James Andreoni,WhyFree Ride?: Strategiesand Learningin Public Goods Experiments,37 J. PUB.ECON. 291, 298 (1988). 374. See id. at 298. 375. This hypothesis is demonstratedby Rachel Croson. See Rachel T. A. Croson, "Theoriesof Altruism and Reciprocity: Evidence from LinearPublic Goods Games"(draft,Nov. 4, 1998) at 14-16 (demonstratingthat subjects' contributionsto the public good are highly correlatedwith their ex ante predictionsof how much other subjectswill contribute)(on file with authors).

1142

CALIFORNIA LAWREVIEW

[Vol. 88:1051

The evidencesuggeststhatthe collectiveactionproblem is less proon the other nounced thanrational choicetheorywouldpredict,although, still existsto a largedegree.376 Theimplications for legal hand,it certainly somewhat subtle.Theevidencecertainly doesnot policyare,consequently, that income should make taxes or with suggest government compliance control laws but it does that in cases pollution voluntary, suggest marginal the costs of coerciondesigned to ensurecollectiveactionmightnot be justified. for example,the problem of littering. CleanstreetsareanConsider, otherclassicpublicgood,as theycanbe enjoyedsimultaneously by a large numberof people and it is impossibleto exclude free ridersfrom the aesthetic benefitsof the streets'cleanliness. Without for littering, penalties self-interested actors to maylitterin publicplaces(thatis, fail to contribute thepublicgood by cleaningup theirtrash) becausethe benefitsof not carryingtheirlitterto a waste basketinureentirelyto themselveswhile the costsareexternalized ontoeveryone else whoutilizesthepublicspace.In a worldfilled with self-interested actors,largefines and strictenforcement of anti-littering laws wouldbe necessary to prevent publicspacesfrombecomingtrash heaps. In therealworld,somepeoplewill litter,butmanywill not,377 at least not very oftenor in very largequantities.378 As long as enoughpeopledo not litterso thatothersbelieve the principle of reciprocity them requires not to litter,thereis likely to be a stablemajority of the citizenry thatexhibitssuchcooperative behavior.379 This mightnot providea justification
376. See, e.g., Robin M. Dawes & RichardH. Thaler, Cooperation,2 J. ECON. PERSP. 187 (1988) (concluding that "[i]t is certainlytrue that there is a 'free riderproblem"... [but] the strong free rider predictionis clearly wrong"). 377. In one study, approximatelyone-thirdof drivers who had flyers placed on the windshield of their cars threw the flyers on the street, while the othertwo-thirdsdid not. See RobertB. Cialdiniet al., A Focus Theory of Normative Conduct: A Theoretical Refinementand Reevaluation of the Role of Norms in Human Behavior, in ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY 201, 221-23 (Mark P. Zannaed., 1991). 378. A caveat: experimental evidence suggests that people are more willing to contribute to a public good if they are part of the group that benefits from it. See Ledyard, supra note 364, at 164 (reviewing studies). Consequently,we might expect fewer people to litter in their own neighborhoods than in neighborhoodsthat they do not frequent. 379. That is, there may develop a social norm against littering. See supra Part IV.A. Note the similaritiesbetween the decision to litter and the decision to leave a tip. Specifically, people are more likely to litter in a distantplace in which they know no one than in their neighborhood.Similarly,there is a "tipping"phenomenonassociated with public goods and social norms in the social amenities of an urbanneighborhood.Social norms may argue for keeping one's propertyclean and in good repair.If the vast majorityof those in an urbanneighborhoodfollow this norm,then they enjoy the public good of high neighborhoodamenity and public order. But if that cooperation with the social norm should fail, then the public amenity of neighborhood cleanliness and the public order can quickly decline, bringingother social problems,such as streetcrime. James Q. Wilson and George Kelling have argued that one of the surest precursorsthat a neighborhoodis likely to experience rampantcrime was the failure to repair broken windows. See James Q. Wilson & George L. Kelling, Broken Windows, ATLANTICMONTHLY,March 1982, at 29. Some people may take the observable fact that private and

2000]

LAWAND BEHAVIORAL SCIENCE

1143

for repealing butit mightsuggestthatthe costs ordinances,380 anti-littering of strictlyenforcing thoseordinances (withpolice specifically assignedto the problem, for example)arenotjustifiedby the magnitude of the problem.
CONCLUSION

for actorssubject to the or disincentives Legalrulescreateincentives theseincentive legal systemto act.Thoughtful legalpolicymustrecognize effects andbe responsive to them.The law-and-economics movement has on these the and has conseforcefullyimpressed points legal academy in legal thinking. Further adquentlybeen the catalystof majoradvances vancesborneoutof thisparadigm, we believe,arelargelydependent on the of legal scholars withthe incentive effectsof law to concerned willingness rethink the behavioral of rational choicetheory,at leastin the assumptions formsthatthis theoryhas traditionally been understood in legal scholarship. In this Article,we have arguedthatthin versionsof rational choice basis on theory-for example,expectedutilitytheory-are an inadequate whichto rest legal policy becausethey have little or no predictive value. are as theories of for Further, they implausible generalapplicability two reasons.First,people are boundedly To save time, avoid comrational. and make with the of dailylife tractaplexity, generally dealing challenges actors often decision or heuristics thatleadto ble, adopt strategies employ decisionsthatfail to maximizetheirutility.Second,people'spreferences are affectedby context.Choicesarenot mademerelyby comparing decision outcomes; situational variables are criticalinputsinto decisionmaking. Thickversionsof rational choicetheory-for example,"self-interest" and "wealthmaximization"-adopt expectedutilitytheoryas theirbasis but addto it predictions aboutactors'substantive goals andpreferencesthatis, predictions about"ends" in addition to predictions about"means." These versionsof rational choice theoryreducethe inadequacy problem created by relyingon expectedutilitytheoryalone,but they addnew imto thosethatplagueexpected The selfplausibility problems utilitytheory. interest that will act in the face of commutheorypredicts people selfishly treat others when there is no cost to doing so, nity pressures, inequitably andfail to contribute to the provisionof publicgoods. Unfortunately for
public propertyare not being maintainedas a sign that other social normsmay be unenforced,too, and, therefore,begin to pursueanti-socialbehaviors. 380. The existence of such an ordinanceitself might, in fact, play a role in maintainingthe social norm supportiveof the observed cooperative behavior. See Sunstein, supra note 318, at 2032 (noting that littering laws that are lightly enforced might "have an importanteffect in signaling appropriate behavior").

1144

LAW REVIEW CALIFORNIA

[Vol.88:1051

the theory (but fortunatelyfor humanity),all of these predictionsare in some situations relevant false-not all the time, but,at a minimum, to legalpolicy. thattraditional of rational Ourcontention choice theory conceptions areflawedin important ways does not suggestthatwe believepeopleare "irrational." Most of the observeddeviations frombehavior predicted by rationalchoice theoryare quite sensibleand understandable, and many seem quiterationalin a "global" sense, although they resultin behavior thatviolatesthe predictions of rational choice theoryon the more"local" level on whichlegal scholarship we do generally operates. Consequently, not arguethat the edifice of rationalchoice theory,which underliesso muchof legal scholarship, be rippeddown.Rather, we suggestthatit be flawsin its structure thatunduly andunrevised,payingheedto important limit the of a more nuanced of how necessarily development understanding law affectssociety.381 In PartsII through IV of this Article,we describe morethana dozen different flaws in the mightyarmor of rational choicetheory,as traditionin legal analysis, andwe suggestpotential ally understood legalpolicyimplicationsof each one of those flaws. We presentnone of these policy as definitivesolutionsto legal problems. Rather, implications they areofferedas stakesin the ground we will serve as that, hope, starting pointsfor a new generation of legal scholarship,one that will see "law and economics" transformed into"lawandbehavioral science."

381. Cf. Ellickson, supra note 315, at 551 (arguing that evidence of the importance of social norms on behaviorhas "destabilized" the traditionallaw-and-economicsparadigmbut does not require scholarsto "throwover the rational-choicemodel").