TAKING BEHAVIORALISM SERIOUSLY:

THE PROBLEM OF MARKET

MANIPULATION

JON D. HANSON* & DOUGLAS A. KYSAR**

In recent years, legal scholars dissatisfied with the behavioral assumptions of the
rationalactormodel have increasingly turned to the findings of cognitive psycholo-
gists and decision theorists to enhance the accuracy of efficiency analysis. Jon
Hanson and Douglas Kysar review those findings in this Article, concluding that
scholars have been well justified in incorporating the behaviorallst account of
human behavior into law and economics. Nevertheless, Hanson and Kysar argue
that those scholars simultaneously have failed to take the findings of behavioral
researchto their logical conclusion. Using the scholarly applicationof behavioral-
ist insights to products liability theory as an example, the authors demonstrate that
legal scholars thus far have treated cognitive anomalies as relatively fixed and
independent influences on individualdecisionmaking. Rather than such an exoge-
nous analysis, Hanson and Kysar advocate an endogenous examination of behavi-
oralistfindings that allows for internal, dynamic effects of cognitive biases within
the decisionmaking model. By recognizing that cognitive anomalies influence not
only the behavior of biased decisionmakers but also the incentives of other eco-
nomic actors, Hanson and Kysar reveal the possibility of market manipulation-
that is, the possibility that market outcomes can be influenced, if not determined, by
the ability of one actor to control the format of information, the framing andpres-
entation of choices, and, more generally, the setting within which market transac-
tions occur. Again using the field of products liability theory as an example, the
authors argue that such market manipulation will come to characterize consumer
product markets; powerful economic incentives will drive manufacturersto engage
in practices that, whether consciously or not, utilize non-rationalconsumer tenden-
cies to influence consumerpreferences andperceptionsfor gain. The authors con-
clude by previewing the evidence from their companion article that demonstrates
the seriousness of the problem of market manipulationand the need for corrective
legal devices such as enterprise liability.

* Professor of Law, Harvard University. B.A., 1986, Rice University; J.D., 1990, Yale
University.
** Law Clerk, Hon. William G. Young, United States District Court for the District of
Massachusetts. B.A., 1995, Indiana University; J.D., 1998, Harvard University.
We are grateful to Carol Igoe for outstanding secretarial assistance and to Carl Bogus,
Grant Dixton, Ana C. Reyes, and Paul Slovic for helpful comments. In addition, this Arti-
cle has benefited greatly from conversations with Daniel Kahnemann and Neil Weinstein,
extended conversations with Baruch Fischhoff and Paul Slovic, and years of conversations
and collaboration with Steven Croley and Kyle Logue. We also gratefully acknowledge the
Harvard Law School Summer Research Program for funding portions of our research. Fi-
nally, we want to extend our deepest thanks to Kathleen, Emily, Erin, and Ian Hanson and
to Vicki Kysar, all of whom disprove the self-interest assumption on a daily basis.

630

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

Introduction .................................................... 633

I. A Brief Overview of Behavioral Research .............. 640
A. Expected Utility Theory and the Origins of
Behavioral Research ................................ 640
B. A Review of the Behavioral Literature ............. 643
1. Manipulability of Scientific and Probabilistic
Judgments ...................................... 645
a. The Formation and Influence of Personal
Hypotheses ................................. 646
i. Belief Perseverance ..................... 646
ii. Confirmatory Bias ...................... 647
iii. Hypothesis-Based Filtering ............. 650
iv. Entity Effect ............................ 650
v. Motivated Reasoning ................... 653
b. False Self-Confidence ....................... 654
i. Optimistic Bias ......................... 654
ii. Cognitive Dissonance ................... 658
iii. The Illusion of Control ................. 658
iv. Hindsight Bias .......................... 659
v. The Surprising Effect of Reasoning ..... 660
c. Bad Statisticians ............................ 662
i. Availability ............................. 662
ii. Representativeness ..................... 664
iii. Anchoring and Adjustment ............. 667
d. Experiential Thinking, Affect, and the
Perception of Risk .......................... 669
2. Manipulability of Preferences ................... 672
a. The Status Quo Bias and Endowment
Effect ....................................... 673
b. Context Effects and the Effect of Irrelevant
Options ..................................... 676
c. Elastic Justification ......................... 677
d. Time-Variant Preferences ................... 678
e. Reciprocity and Attribution ................ 680
L Preference-Trumping Effect of Visceral
Factors ...................................... 682
g. Framing Effects ............................. 684
C. Critiques ............................................ 687
II. The Academic Debate over the Significance of
Behavioral Research for Products Liability Law ........ 693
A. Under-Estimators ................................... 696
1. Robert Prentice and Mark Roszkowski ......... 697
2. Howard Latin ................................... 699

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

B. Over-Estim ators .................................... 704

1. W . Kip Viscusi .................................. 704
2. Alan Schwartz .................................. 710
C. The ALI Reporters' Study on Enterprise
Responsibility for Personal Injury:
A Comprehensive View? ........ . . . . . . . . . . . . . . . . . . . 715
1. The Reporters' Recommendations Based on a
Survey of "Evidence About Risk
Perceptions" .......... ......................... 716
2. Criticisms of the Reporters' Methodology and
Conclusions ..................................... 718
III. The Problem of Manipulation .......................... 721
A. The Indeterminacy of the Behavioral Research ..... 722
B. The Irrelevancy of Indeterminacy and the
Endogenous Nature of Perceptions ................. 723
C. Manufacturer Manipulation of Consumers.......... 724
1. Manipulation of Scientific and Probabilistic
Judgm ents ...................................... 727
a. The Formation and Influence of Personal
Hypotheses ................................. 727
b. False Self-Confidence ....................... 729
c. Bad Statisticians ............................ 731
d. Experiential Thinking, Affect, and the
Perception of Risk .......................... 732
2. Manipulation of Preferences .................... 733
a. The Status-Quo Bias and Endowment
Effect ....................................... 733
b. Context Effects and the Effect of Irrelevant
Options ..................................... 734
c. Elastic Justification ......................... 735
d. Time-Variant Preferences ................... 735
e. Reciprocity and Attribution ................ 736
f. Preference-Trumping Effect of Visceral
Factors ...................................... 738
g. Framing Effects ............................. 739
3. Summ ary ....................................... 743
D. Answering the Critiques of Behavioral Research ... 743
C onclusion ..................................................... 745

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

INTRODUCTION
Do we move ourselves, or are moved by an unseen hand... ?
-Tennyson 1
For the past few decades, cognitive psychologists and behavioral
researchers have been steadily uncovering evidence that human deci-
sionmaking processes are prone to nonrational, yet systematic, ten-
dencies. 2 These researchers claim not merely that we sometimes fail
to abide by rules of logic, but that we fail to do so in predictable ways.
In the words of Daniel Kahneman and the late Amos Tversky, two
pioneering researchers in the field, we are susceptible to illusions of
the mind which are "'neither rational, nor capricious.'" 3 These cogni-
tive illusions-sometimes referred to as biases 4 -are not limited to
the uneducated or unintelligent,5 and they are not readily capable of
being unlearned. 6 Instead, they affect us all with uncanny consistency
and unflappable persistence.
With a few notable exceptions, 7 implications of this research for
legal institutions were slow in reaching the academic literature.
1 Alfred Tennyson, Maud 13 (1874).
2 This is a wide, interdisciplinary area of research which could be (and has been) given
a variety of names. Following recent commentators, we will refer primarily to behavioral
research. See Christine Jols, Cass R. Sunstein, & Richard Thaler, A Behavioral Approach
to Law and Economics, 50 Stan. L Rev. 1471, 1473 (1998) (approaching economic analysis
through conception of choice that reflects improved understanding of human behavior);
Cass R. Sunstein, Behavioral Analysis of Law, 64 U. Chi. L Rev. 1175,1175 (1997) (theo-
rizing that future of economic analysis depends upon new understandings of decisionmak-
ing developed through behavioral research). We note, however, that although the term
"behavioralism" or "behavioral economics" seems destined to become dominant in the
legal academic literature, economists must share credit for the project with cognitive psy-
chologists and statisticians, among others.
3 Massimo Piattelli-Palmarini, Probability Blindness: Neither Rational nor Capri-
cious, Bostonia, Mar.-Apr. 1991, at 28,28 (quoting Amos Tversky and Daniel Kahneman).
4 See generally Amos Tversky & Daniel Kahneman, Judgment Under Uncertainty.
Heuristics and Biases, in Judgment Under Uncertainty. Heuristics and Biases 3 (Daniel
Kalmeman, Paul Slovic, & Amos Tversky eds., 1982) [hereinafter Judgment Under Uncer-
tainty] (enumerating biases in assessment of probabilities and prediction of values).
5 See id. at 7-8 (reviewing evidence of erroneous belief in "law of small numbers"
among experienced research psychologists). This article also appears in Judgment and De-
cision Making 38 (Hal R. Arkes & Kenneth R. Hammond eds., 1986).
6 See Amos Tversky & Daniel Kahneman, Rational Choice and the Framing of Deci-
sions, in Rational Choice: The Contrast Between Economics and Psychology 67, 90-91
(Robin M. Hogarth & Melvin W. Reder eds., 1987) [hereinafter Rational Choice].
7 Important early works include Robert C. Ellickson, Bringing Culture and Human
Frailty to Rational Actors: A Critique of Classical Law and Economics, 65 Chi.-Kent L
Rev. 23, 35-55 (1989) (describing potential contribution of psychology and sociology to
economic analysis); Ward Edwards & Detlof von Winterfeldt, Cognitive Illusions and
Their Implications for the Law, 59 S. Cal. L Rev. 225, 269-76 (1986) (speculating on rele-
vance of error analysis to legal practitioners); Elizabeth Hoffman & Matthew L Spitzer,
Willingness to Pay vs. Willingness to Accept: Legal and Economic Implications, 71 Wash.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

Within the last few years, however, we have seen an outpouring of

scholarship addressing the impact of behavioral research over a wide
range of legal topics.8 Indeed, one might predict that the current be-
havioral movement eventually will have an influence on legal scholar-
ship matched only by its predecessor, the law and economics
movement. Ultimately, any legal concept that relies in some sense on
a notion of reasonableness or that is premised on the existence of a
reasonable or rational decisionmaker will need to be reassessed in

U. L.Q. 59, 98-114 (1993) (describing divergence between willingness to accept and willing-
ness to pay as applied to law and economics analysis); Howard A. Latin, "Good" Warn-
ings, Bad Products, and Cognitive Limitations, 41 UCLA L. Rev. 1193, 1194-95 (1994)
[hereinafter Latin, "Good" Warnings] (examining numerous informational processing and
decisionmaking problems that reduce efficacy of disclosure requirements); Howard A.
Latin, Problem-Solving Behavior and Theories of Tort Liability, 73 Cal. L. Rev. 677,745-46
(1985) (using psychological and economic analysis to challenge rational maximizing behav-
ior assumption of tort liability and proposing situational analysis that accounts for actor
characteristics, category of risk, and corresponding legal rules).
8 See, e.g., Gary L. Blasi, What Lawyers Know: Lawyering Expertise, Cognitive Sci-
ence, and the Functions of Theory, 45 J. Legal Educ. 313, 329-54 (1995) (arguing that ad-
vances in cognitive psychology have made possible more systematic and precise study of
legal decisionmaking and problem solving than decisionmaking paradigm); Melvin Aron
Eisenberg, The Limits of Cognition and the Limits of Contract, 47 Stan. L. Rev. 211, 211-
13 (1995) (arguing that understanding of psychological constraints on decisionmaking abil-
ity should influence development of contract law); Christine Jolls, Contracts as Bilateral
Commitments: A New Perspective on Contract Modification, 26 J. Legal Stud. 203, 219-24
(1987) (discussing impact of intertemporal preference biases on contract modifications);
Christine Jolls, Hands-Tying and the Age Discrimination in Employment Act, 74 Tex. L.
Rev. 1813, 1826-28, 1840-45 (1996) (arguing that similar temporal biases justify age discrim-
ination laws in employment as "hands-tying" strategy); Russell Korobkin & Chris Guthrie,
Psychology, Economics, and Settlement: A New Look at the Role of the Lawyer, 76 Tex.
L. Rev. 77, 113-21 (1997) (uncovering evidence that lawyers as class can aid clients in over-
coming effect of heuristics and biases in litigation settlement negotiations); Linda Hamilton
Krieger, The Content of Our Categories: A Cognitive Bias Approach to Discrimination
and Equal Employment Opportunity, 47 Stan. L. Rev. 1161, 1216-17 (1995) (arguing that
discriminatory impact often can result from unintentional cognitive biases rather than from
discriminatory motive or intent); Donald C. Langevoort, Selling Hope, Selling Risk: Some
Lessons for Law from Behavioral Economics About Stockbrokers and Sophisticated Cus-
tomers, 84 Cal. L. Rev. 627, 669-91 (1996) (utilizing behavioral economics research to pro-
vide model of consumer-broker securities sales transactions); Edward J. McCaffery, Daniel
J. Kahneman, & Matthew L. Spitzer, Framing the Jury: Cognitive Perspectives on Pain
and Suffering Awards, 81 Va. L. Rev. 1341, 1354-73 (1995) (using experimental model to
assess impact of cognitive psychology research on framing of jury instructions and ensuing
pain and suffering damages awards); Jason Ross Penzer, Note, Grading the Report Card:
Lessons from Cognitive Psychology, Marketing, and the Law of Information Disclosure for
Quality Assessment in Health Care Reform, 12 Yale J. on Reg. 207, 233-54 (1995) (ques-
tioning effectiveness of consumer-directed health care information disclosures in light of
consumer cognitive heuristics and biases). For an overview of the "diffusion" of behavioral
research into legal scholarship over the past twenty years, see Donald C. Langevoort, Be-
havioral Theories of Judgment and Decision Making in Legal Scholarship: A Literature
Review, 51 Vand. L. Rev. 1499 (1998).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

light of the mounting evidence that a human is "a reasoning rather

than a reasonable animal." 9
This Article contributes to that reassessment by focusing on the
problem of manipulability. Our central contention is that the pres-
ence of unyielding cognitive biases makes individual decisionmakers
susceptible to manipulation by those able to influence the context in
which decisions are made. More particularly, we believe that market
outcomes frequently will be heavily influenced, if not determined, by
the ability of one actor to control the format of information, the pres-
entation of choices, and, in general, the setting within which market
transactions occur. Once one accepts that individuals systematically
behave in nonrational ways, it follows from an economic perspective
that others will exploit those tendencies for gain.
That possibility of manipulation has a variety of implications for
legal policy analysis that have heretofore gone unrecognized. We sus-
pect that the failure of scholars to recognize this risk of manipulation
stems from at least two factors. First, the primary researchers in the
field have devoted little attention to the possibility of market manipu-
lation simply because theirs is a descriptive project. Researchers such
as Kahneman and Tversky have devoted most of their work to identi-
fying the existence and causes of cognitive anomalies, not to spelling
out the implications of their findings for other disciplines.' 0 Second,
legal scholars who have examined the behavioral research have done
so with an eye toward incorporating the findings into the classical law
and economics model, and in so doing they have remained somewhat
wedded to that classical model. For instance, Professors Jolls,
Sunstein, and Thaler, in their recent and important blueprint for be-
havioral economic analysis of law, are explicit in that approach: "The
unifying idea in our analysis is that behavioral economics allows us to
model and predict behavior relevant to law with the tools of traditional
economic analysis, but with more accurate assumptions about human
behavior, and more accurate predictions and prescriptions about
1
law."
One important result of this sympathy to the classical economic
model is that commentators tend to view cognitive biases as exoge-
nous influences on individual behavior; that is, legal scholars typically
assume that the biases will have a fixed influence on behavior, unaf-
fected by other factors within the model. This supposed stability of
influence allows those scholars to utilize the classical model of eco-

9 Alexander Hamilton, quoted in Laurence J. Peter, Peter's Quotations: Ideas for Our
Tune 315 (1977).
10 For a discussion of some exceptions, see infra Part mI.C
11 JoUs et al., supra note 2, at 1474 (emphasis added).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

nomic behavior with only the (fixed) additional input of a particular

cognitive bias. Below we review the work of several scholars who im-
plicitly assume that consumers' risk perceptions are independent of
manufacturers' and marketers' efforts to influence them. And even
the most sophisticated analyses of the implications of behavioralism
for legal rulemaking dramatically understate the extent to which con-
sumers' risk perceptions are endogenous-that is, determined or al-
tered by the market contexts being analyzed. For example, Jolls,
Sunstein, and Thaler discuss the shortcomings of information disclo-
sure, a common regulatory solution to market failures.12 The authors
note, for example, that with respect to product hazard warnings, indi-
viduals will often perceive risks differently depending on the manner
in which risk information is presented. 13 They also note that "[prod-
uct manufacturers] will often have an interest in providing the least
scary, most pallid version of the information possible. ' 14 To circum-
vent such perceptual variances, the authors recommend that regula-
tors utilize a variety of bias-specific procedures such as "framing
consequences in terms of losses rather than gains,"'15 "tak[ing] account
of the fact that vivid and personal information will often be more ef-
fective than statistical evidence,"'1 6 and "[a]void[ing] the pitfalls of
overoptimism."'1 7 Although Jolls, Sunstein, and Thaler recognize
manufacturers' incentives to downplay risks, theirs is still mostly an
exogenous application of behavioral findings: Because individuals dis-
play certain fixed and identifiable cognitive features, regulators need
only alter hazard warning requirements to suit those features and,
consequently, overcome the cognitive limitations of individual actors.
Of course, such an approach is a welcome improvement over classical
economic analysis of law, which would have failed to acknowledge
that different formats for presenting information could have varying
degrees of success, much less attempt to identify which formats would
be most likely to succeed.
Nevertheless, in our view, the full significance of the behavioral
research can only be seen when limitations on decisionmaking abilities
are treated as endogenous to the economic model-that is, when it is
recognized that not only are cognitive biases influencing the behavior
of individuals, but other factors within the model are influencing the
presence and force of cognitive biases. This possibility arises from the

12 See id. at 1533-37.

13 See id. at 1534.
14 Id. at 1535.
15 Id. at 1536.
16 Id. at 1537.
17 Id. (emphasis removed).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

very same aspect of cognitive biases that has made other legal scholars
quick to incorporate them into the classical model: They are "system-
atic and can be modeled.""' Just as the predictability of biases makes
them easy to assimilate into economic models, it also makes them easy
to exploit by those who seek to influence the behavior of economic
actors. Thus, while Jolls, Sunstein, and Thaler recognize that individu-
als react differently to product risk information depending on the
manner in which it is presented, they do not fully explore the implica-
tions of that anomaly for the market behavior of other actors, namely
product manufacturers. Given that it is possible for government regu-
lators "to shape people's behavior in desired directions,"1 9 it is inevi-
table that product manufacturers will attempt to alter consumer risk
perceptions. The key difference is that they will attempt to elicit
lower than accurate perceptions of risk rather than accurate ones.
It is not simply that manufacturers would prefer "the least scary,
most pallid version" of a product warning when and if it is required by
the government. Rather, it is that manufacturers have incentives to
utilize cognitive biases actively to shape consumer perceptions
throughout the product purchasing context and independently of gov-
ernment requirements. Advertising, promotion, and price setting all
become means of altering consumer risk perceptions, regardless of
mandated hazard warnings. This is what we mean by manipulation-
the utilization of cognitive biases to influence peoples' perceptions
and, in turn, behavior. Again, the reluctance of earlier scholars to ac-
knowledge this possibility of manipulation appears to stem from their
allegiance to the classical model and desire to remain constructive
with respect to it, even when the behavioral research indicates that a
more drastic restructuring is required.
The problem of market manipulation has implications for a broad
range of legal issues. For at least three reasons, however, we will fo-
cus in this Article primarily on the field of products liability law. First,
products liability scholarship is dominated by law and economics the-
ory z° which is premised on a rational actor model of human behav-
ior.2 1 If behavioral economics represents an improvement on classical

18 Id. at 1475.
19 Id. at 1536.
20 See George L. Priest, The Inevitability of Tort Reform, 26 Val. U. L Rev. 701,704-05
(1992):
[T]here are few articles within the last ten years and no articles of importance
within the last five years written about modem tort law that have not ad-
dressed, either as the principal thesis or as the subject to which the thesis of the
article is responding, the functional economic analysis [of tort law].
21See Bailey Kulkin, The Gaps Between the Fingers of the Invisible Hand, 58 Brook.
L. Rev. 835, 837-39 (1992) (describing importance of assumptions of rationality to classical

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

law and economics theory, as many commentators claim, 22 then prod-

ucts liability scholarship will need to be reexamined accordingly. Sec-
ond, the issue of whether, and to what extent, consumers are able to
acquire and comprehend product risk information has assumed a posi-
tion of utmost importance to products liability theory.3 Because the
behavioral findings pertain largely to the manner in which individuals
process information, application of those findings to the issue of con-
sumer risk perception has the potential to make significant advances
in the products liability literature. Finally, because consumer product
purchases generally are the most frequent and familiar market trans-
actions that any of us experience, an examination of the consumer
product context provides a relevant, immediate, and accessible way to
explore our more general thesis about market manipulation. Indeed,
when reviewing the behavioral research findings summarized in this
Article, we invite the reader to draw from her experience as a con-
sumer and consider how such cognitive phenomena might play out in
the typical product context.
The Article is organized as follows: Part I utilizes the behavioral
research to outline a new model of the human decisionmaker, one
susceptible to a range of nonrational tendencies. Although other
surveys of the behavioral research are available, 24 we have attempted
to organize the findings in a manner that departs significantly from the
now standard law review summary of the literature. That is, we seek
to do more than merely catalogue a set of observed behavioral quirks.
By reviewing new sets of literature that begin to unravel the motiva-
tions and impulses lying behind such behavioral phenomena, we hope
to provide a decisionmaker model that is relevant to a broader and
more realistic range of behavioral contexts than the traditional eco-
nomic actor, or even the newly enhanced economic actor (with biases)
offered by Jolls, Sunstein, and Thaler.

law and economics); W. Kip Viscusi, Individual Rationality, Hazard Warnings, and the
Foundations of Tort Law, 48 Rutgers L. Rev. 625, 636 (1996) (noting that "foundation of
economic analysis of choice is based on the rationality of individual decision making").
22 See, e.g., Jolls et al., supra note 2, at 1473 ("Our goal in this article is to advance an
approach to the economic analysis of law that is informed by a more accurate conception
of choice, one that reflects a better understanding of human behavior and its well-
springs."); Sunstein, supra note 2, at 1175 ("The future of economic analysis of law lies in
new and better understandings of decision and choice.").
23 See, e.g., Enterprise Responsibility for Personal Injury, 1 A.L.I. 230 (1991) ("The
principle assumption in the literature supporting a role for legal liability is that consumers
underestimate product defect risks and, as a consequence, put insufficient market pressure
on firms to produce safety.").
24 See, e.g., Jolls et al., supra note 2, at 1473-75; Latin, "Good" Warnings, supra note 7,
at 1220-57; Sunstein, supra note 2, at 1177-79. A particularly helpful summary can be
found in Matthew Rabin, Psychology and Economics, 36 J. Econ. Lit. 11, 11-13 (1998).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

Moreover, because we shed some light on the psychological

processes that actually drive observed behavior, our literature survey
helps illuminate the endogenous nature of risk perceptions: that is, an
analysis that takes account of the dynamic effects that multiple actors
can exert upon each other within the decisionmaking context. Indeed,
the framework we employ in our review of behavioral literature en-
ables us to generate predictions about the way in which manufacturers
and consumers xvill behave in the product context. In short, our re-
view of behavioral research describes many findings from the litera-
ture that previously have received little or no attention from legal
scholars and presents them in a manner that yields new insight into
products liability and other legal questions.
Part II then examines the ways in which products liability schol-
ars have thus far viewed the behavioral research and its impact on
consumer risk perception. More specifically, we note that commenta-
tors tend to fall into one of two camps: those who find in the behav-
ioral literature evidence that consumers generally overestimate risk,
and those who find evidence that consumers generally underestimate
risk. We also review in Part II the ALI Reporters' Study on Enter-
prise Liability and Personal Injury. The ALI Study provides an im-
portant first attempt at organizing the behavioral research into a form
capable of yielding tort law policy implications. Nevertheless, we ar-
gue that the Study relies on a non-representative sample of behavioral
research and that much of the excluded evidence casts doubt on the
Study's policy conclusions. Moreover, we argue that even the evi-
dence relied upon in the Study fails to support its policy conclusions.
We turn in Part Ill to our own view of the significance of behav-
ioral research. We argue that previous commentators have focused
unwarranted attention on the largely unanswerable question of
whether consumers, before they interact with market forces, tend to
systematically over- or underestimate product risks. In our view, the
question is only a small part of the story, because whatever their risk
perceptions might be independent of market influences, consumers
will be susceptible to manipulation by manufacturers within the mar-
ket context. Thus, scholars attempting to categorize the relevant
product risks as ones that would be over- or underestimated ignore
the more fundamental message of the cognitive bias literature that
perceptions are manipulable. It is on this manipulability of consumer
perceptions and on the strong market forces that will lead manufac-
turers to manipulate that we focus in this Article.
We conclude by previewing two other articles that are part of this
project: Taking Behavioralism Seriously: Some Evidence of Market

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74-630

Manipulation, 25 and Taking Behavioralism Seriously: A Potential Re-

sponse to Market Manipulation. 26 In those articles, we will offer em-
pirical support for our thesis that manufacturers manipulate consumer
perceptions of products, and we will argue that a products liability
standard of enterprise liability may provide the best response to this
problem of market manipulation.
I
A BRIEF OVERVIEW OF BEHAVIORAL RESEARCH
During the last three decades, while law faculties across the na-
tion were succumbing to the brilliant simplicity of the Coase theorem
and the analytical force of the law and economics movement, 27 social
scientists in other departments were discovering evidence that should
have given pause to even the most ardent legal economic positivist.
Those scientists-cognitive psychologists, behavioral researchers,
probability theorists, and others-were discovering powerful evidence
that the rational actor model, upon which the law and economics pro-
ject depends, is significantly flawed. In place of the rational actor
model, those scientists were developing a human decisionmaker
model replete with heuristics and biases, unwarranted self-confidence,
a notable ineptitude for probability, and a host of other nonrational
cognitive features.

A. Expected Utility Theory and the Origins of Behavioral Research

In 1944, mathematician John von Neumann and economist Oskar
Morgenstern published a book entitled Theory of Games and Eco-
nomic Behavior.28 The importance of this work can hardly be over-
29
stated. In addition to launching the entire field of game theory, it
has also been credited with reviving the notion of utility and
25 Jon D. Hanson & Douglas A. Kysar, Taking Behavioralism Seriously: Some Evi-
dence of Market Manipulation, 112 Harv. L. Rev. 1420 (1999) [hereinafter TBS II].
26 Jon D. Hanson & Douglas A. Kysar, Taking Behavioralism Seriously: A Potential
Response to Market Manipulation, 6 Roger Williams L. Rev. (forthcoming Fall 2000)
[hereinafter TBS III].
27 See Ellickson, supra note 7, at 26-30 (describing historical growth of law and eco-
nomics movement in American legal academia).
28 John von Neumann & Oskar Morgenstem, Theory of Games and Economic Behav-
ior 555-86 (1944) (applying mathematical game theory to economic problems).
29 See William M. Goldstein & Robin M. Hogarth, Judgment and Decision Research:
Some Historical Context, in Research on Judgment and Decision Making 4 (William M.
Goldstein & Robin M. Hogarth eds., 1997). At least one commentator has speculated that
the awarding of Nobel Prizes to three game theorists in 1994-the fiftieth anniversary of
the publication of von Neumann and Morgenstern's seminal book-was more than mere
coincidence. See The Games Economists Play, Economist, Oct. 15, 1994, at 96. For a
useful introduction to game theory, see generally Ian Ayres, Playing Games with the Law,

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

"ma[king] subjectivity respectable again in economics.' Those ac-

complishments were related for von Neumann and Morgenstern: In
order to develop a theory of optimal economic decisionmaking under
situations of imperfect information, the authors needed a concept of
utility that was not woodenly restricted to monetary values over single
states of being. Von Neumann and Morgenstern therefore posited a
theory of decisionmaking according to a principle of maximizing ex-
pected utility. A somewhat oversimplified example helps make their
insight clear: If someone faces a 10% chance of winning $100, her
expected utility is .10 * $100 = $10. Using this concept, von Neumann
and Morgenstern were able to talk meaningfully about optimal eco-
nomic decisionmaking under situations of uncertainty. For instance,
when faced with a choice between a 25% chance of winning $100 and
a 10% chance of winning $1000, an individual should ordinarily select
the latter option to maximize expected utility (because .10 * $1000 =
$100 while .25 * $100 = $25).31
Much of von Neumann and Morgenstern's work then was de-
voted to deriving rules of optimal decisionmaking in similar, though
far more complicated, problem settings. That work, and the work of
other game theorists, led to the following sorts of decisionmaking
principles or axioms that rational actors could be expected to honor: 32
(i) Ordering. Between two objects, a person must either prefer one
to the other or be indifferent to both. In other words, under ex-
pected utility theory, apples and oranges are entirely susceptible to
comparison. Similarly, all preferences are transitive, such that if

42 Stan. L Rev. 1291 (1990); Stephen W. Salant & Theodore S. Sims, Game Theory and
the Law- Ready for Prime Time?, 94 Mich. L Rev. 1839, 1840-46 (1996) (book review).
30 Detlof Von Vinterfeldt & Ward Edwards, Decision Analysis and Behavioral Re-
search 561 (1986).
31 It should be noted that von Neumann and Morgenstem did not strictly equate maxi-
mum expected utility with maximum monetary payoffs. See Robyn M. Dawes, Rational
Choice in an Uncertain World 11 (1988) (reviewing modem origins of rationality theory).
Thus, for instance, someone down on her luck and simply in need of money to eat may
prefer a 25% chance of winning some amount of money to a 10% chance of winning a
greater amount. One would say that .25 times her expected utility for S10D is greater than
.10 times her expected utility for $1000. "The catch is that by specifying the theory in terms
of utility rather than monetary value, it is almost always possible to assume that some sort
of maximization principles work and then define utilities accordingly." Id. at 13 (emphasis
in original). Perhaps in order to avoid this tautology (and provide a more quantifiable
framework), legal economists often equate utility maximization with monetary wealth
maximization.
32 This list of "crucial axioms" is derived from Colin Camerer, Individual Decision
Making, in The Handbook of Experimental Economics 587,618 (John H. Kagel & Al.in E.
Roth eds., 1995) [hereinafter Handbook of Experimental Economics]. There have been
several different formulations of these axioms and much debate over their precise content.
One widely used summary is provided in R. Duncan Luce & Howard Raiffa, Games and
Decisions: Introduction and Critical Survey 25-31 (1957).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

one prefers A to B, and B to C, we would expect her to prefer A to

33
C as well.
(ii) Continuity. Every prize B is indifferent to some lottery involv-
ing just A and C. For instance, suppose a person prefers A to B,
and B to C. If there is a lottery in which the probability of winning
A is p and the probability of winning C is I-p, there must be some
value for p in which the person is indifferent between the lottery
and B.3 4 In other words, if the odds are right, a person will always
gamble.
(iii) Independence. A person's preferences between two objects
should remain unchanged when the objects are substituted into
identical lotteries. Thus, if one prefers A to B when the objects are
viewed alone, she should prefer a fifty-fifty chance at winning A to a
fifty-fifty chance at winning B. Similarly, if one is indifferent be-
tween A and B, she should also be indifferent between two lotteries
offering A and B on identical terms-"the other possible alterna-
tives [to the lotteries] must be irrelevant to the decision that they
''35
are indifferent.
(iv) Invariance. Kahneman and Tversky have identified a final ax-
iom so basic that it is often tacitly assumed by decision theorists
rather than explicitly offered as a testable axiom. This principle,
invariance, refers to the requirement that "different representations
' 36
of the same choice problem should yield the same preference.
Thus, where alternative descriptions of the same outcome are for-
mulated, players should express the same preferences regardless of
which description is presented.
The ultimate conclusion to be derived from these axioms of deci-
sionmaking is that players will act in the manner that maximizes their
expected utility. For behavioral research, the significance of Theory
of Games and Economic Behavior lies less with this theoretical propo-
sition than with the fact that economists began to assume its descrip-
tive validity. Rational behavior came to be synonymous with
expected-utility-maximizing behavior. Broad economic contexts, such
as capital and labor markets, were empirically shown to operate in a
manner consistent with the new rationality. Policy choices began to be
made on the assumption that humans, whether individually or through

33 Similarly, if she were indifferent between A and B, and B and C, she should be
indifferent between A and C as well.
34 To see this, note that if p is 1, then the lottery is equivalent to A, in which case the
lottery will be preferred to B. Ifp is 0, then the lottery is equivalent to C, in which case B
is preferred to the lottery. At some point in between, therefore, the person must be indif-
ferent between B and the lottery.
35 Luce & Raiffa, supra note 32, at 27.
36 Tversky & Kahneman, supra note 6, at 69.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

groups, behaved according to the principle of expected utility max-

imization. 37 Robert Ellickson's terse summary of economic modeling
reveals the wide influence of von Neumann and Morgenstern's theo-
retical work:
The economists' model, in its purest form, is based on elegantly sim-
ple propositions about both cognitive capacities and motivations.
The model assumes that a person can perfectly process available
information about alternative courses of action, and can rank possi-
ble outcomes in order of expected utility. The model also assumes
that an actor will choose the course of action that will maximize...
personal expected utility... 38
This powerful conception of rationality, largely accepted as both
a normative and empirical matter by the economic community, 39
prompted a far different reaction among psychologists: "A (seem-
ingly) defensible theory of rational behavior was too intriguing to re-
sist, and empirical tests . . . were soon conducted to see if people
actually behaved in the manner prescribed by expected utility the-
ory."'' 4 Because the axioms of decisionmaking described above were
said to lead to a rational strategy of maximizing expected utility, an
obvious research strategy for psychologists was to test whether people
actually behaved according to the required axioms. In that challenge,
41
the field of behavioral research was born.

B. A Review of the BehavioralLiterature

In a well known experiment, Amos Tversky and Daniel

Kahneman asked subjects to imagine that they must select between
alternative vaccine programs to combat an unusual disease that will

37 See Goldstein & Hogarth, supra note 29, at 11 ("Economics, and significant parts of
political science and sociology, largely accepted utility theory as normatively binding and
descriptively valid and took it as a cornerstone of the theoretical edifice." (citations
omitted)).
38 Elickson, supra note 7, at 23.
39 This should not be overstated. Some of the most important early criticisms of ex-
pected utility theory, involving decision problem "paradoxes" which consistently elicited
behavior at odds with expected utility predictions, came from economists. See Goldstein
and Hogarth, supra note 29, at 12 (describing Allais and EUsberg paradoxes).
40 Id. at 5.
41 See Robin M. Hogarth & Melvin W. Reder, Introduction: Perspectives from Eco-
nomics and Psychology, in Rational Choice, supra note 6, at 1, 5 ("[A]Ithough decision
making has become an increasingly important topic in psychology in recent years, this in-
terest is fairly recent, dating from the 1950s, when psychologists first began systematic in-
vestigations of human choice behavior, investigations that, incidentally, were originally
stimulated by economic models." (citation omitted)).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

kill 600 people if nothing is done.42 The first group of subjects was
presented with these choices:
If program A is adopted, 200 people will be saved.
If program B is adopted, there is a 1/3 probability that 600 people
will be saved, and a 2/3 probability that none of the 600 people will
be saved.
The second group was asked to select from these choices:
If program C is adopted, 400 people will die.
If program D is adopted, there is a 1/3 probability that nobody will
die, and a 2/3 probability that 600 people will die.
Under von Neumann and Morgenstern's expected utility theory,
one might predict that the subjects would be indifferent between all
four programs because each presented the same expected result: 200
people will live and 400 will die. One might even attempt to enrich
the model by, say, predicting a higher utility for those programs which
present a possibility of escaping with zero lives lost.43 Thus, one might
expect that the subjects would prefer programs B and D, respectively.
The way Kahneman and Tversky's subjects actually responded, how-
ever, could not have been predicted with expected utility theory, at
least not without violating its most basic axioms. In the first scenario,
seventy-two percent of subjects chose program A; in the second scena-
rio, seventy-eight percent chose program D. 44 Yet, programs A and C
are logically identical, as are programs B and D; they both present the
same consequences but inspire vastly different reactions. Put differ-
ently, the options offered to each group of subjects were identical, yet
the subjects' choice between the two options appears to depend
largely on how the options were described. Such inconsistency calls
into question the central principle of expected utility theory-that in-
dividuals seek to maximize their expected utility. If individuals be-
have differently when presented with identical possible consequences,
then they must be operating under some other influence.
Kahneman and Tversky's study provides an example of what cog-
nitive psychologists call framing effects. 45 People prefer program A

42 See Daniel Kahneman & Amos Tversky, Choices, Values, and Frames, 39 Am. Psy-
chologist 341, 343 (1984).
43 This would be based on a prediction that the subjects' utility for avoiding all casual-
ties was not strictly linear with avoiding some positive number of casualties; that is, sub-
jects would experience a utility far greater in proportion from saving all 600 lives than
simply saving 200. Again, von Neumann and Morgenstern's conception of expected utility
allowed for such variations. See von Neumann & Morgenstem, supra note 28, at 30.
44 See Kahneman & Tversky, supra note 42, at 343.
45 See generally Amos Tversky & Daniel Kahneman, The Framing of Decisions and the
Psychology of Choice, 211 Science 453 (1981) (finding significant changes in preferences
based on seemingly inconsequential changes in formulation of choice problems).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

because it is framed in terms of guaranteed lives saved; people reject

program C because it is framed in terms of guaranteed lives lost.
Framing effects are representative of the type of phenomena that be-
havioral researchers encountered when they began testing the veracity
of economic predictions about human behavior. Similar nonrational
mental processes and representations have steadily been identified
over the past two decades such that recent commentators can now cite
"a whole range of systematic errors and biases. ' 46 Largely because of
the nature of the biases, the discoveries have been slow in coming and
subject to considerable challenge. 47 At this point, however, as two of
the pioneers in the field of cognitive biases have concluded, "the devi-
ations of actual behavior from the normative model are too wide-
spread to be ignored, too systematic to be dismissed as random error,
and too fundamental to be accommodated by relaxing the normative
system." 48 These discoveries have not only found their way into the
law review literature, they have given rise to a cascade of articles and
to what might well turn out to be the most significant conceptual de-
velopment in legal theory since the emergence of law and economics.
This Section will review those deviations and cognitive illusions,
dividing them into two rough categories: those affecting the way in
which individuals attempt to make "scientific" and probabilistic judg-
ments and those affecting the way in which individuals determine and
exhibit their preferences. 49 We will discuss some important examples
of both, bearing in mind that "[behavioral] research is moving rapidly,
and it is a daunting task to comprehend its scope."50
1. Manipulability of Scientific and ProbabilisticJudgments
In this subsection, we will review a series of cognitive biases that
impact the way in which individuals make scientific and probabilistic
judgments. As wili be shown, these biases do not merely represent
erroneous or imprecise applications of reasoning-they do not repre-
sent reasoning in the classical sense at all. Instead, these biases reveal
46 J. St B.T. Evans, Bias and Rationality, in Rationality. Psychological and Philosophi-
cal Perspectives 6, 6 (Y-I. Manktelow & D.E. Over eds., 1993) (discussing paradox of indi-
vidual rationality and bias).
47 Cf. Piattelli-Palmarini, supra note 3, at 31-32-
[W]e are intimately persuaded that we are able to cope with uncertainty, at
least in the main. In hindsight, given the pervasiveness and insidiousness of
these cognitive illusions, it is hardly surprising that it took so long to discover
them, and that even today these stunning discoveries do not attract the public
attention they deserve.
48 Amos Tversky & Daniel Kahneman, Rational Choice and the Framing of Decisions,
59 J. Bus. S251, S252 (1986).
49 See Goldstein & Hogarth, supra note 29, at 11-26 (adopting this diision).
50 Id. at 35.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

a deep misunderstanding of the nature of scientific judgment and a

series of mental crutches awkwardly constructed to take up the slack.
Their cumulative effect is to impair our appreciation of risk and uncer-
tainty so severely that one commentator has termed the condition
probability blindness.5 ' Whether blind or just impaired, people's
sense of scientific and probabilistic judgment is inarguably of doubtful
acuity.

a. The Formation and Influence of Personal Hypothe-

ses. Considerable research indicates that individuals act, in effect, as
lay scientists, adopting and testing hypotheses, or explanations for un-
derstanding the world. People have a strong tendency to construct
"theories" to account for events or relationships that they find sali-
ent.5 2 Such theories enable individuals to organize and, perhaps, un-
derstand the world around them. Unfortunately, however, lay
scientists, much like professionals, often fail to practice "good sci-
ence." Put differently, people's constructed theories often can be a
source of confusion and misestimation rather than clarification and
understanding.

i. Belief Perseverance. One of our most basic flaws as lay

scientists, for example, is that after constructing a hypothesis or expla-
nation, we tend to disregard evidence that contradicts that hypothesis
and exaggerate evidence that confirms it. There is, in other words,
strong support for the common-sense view that first impressions carry
disproportionate weight. One of the better known experiments dem-
onstrating this tendency was a 1964 study in which subjects were
shown blurred pictures that were gradually brought into focus. 5 3 Dif-
ferent subjects were introduced to the photos at different stages of
sharpened focus, but the pace of the sharpening process and the final
degree of focus was identical for all subjects. Interestingly, the more

51 See Piattelli-Palmarini, supra note 3, at 28 (describing common cognitive traps).

52 See generally Harold H. Kelley, The Processes of Causal Attribution, 28 Am. Psy-
chologist 107 (1973). By "theories" we do not necessarily mean full-blown scientific causal
accounts. Rather, we mean informal social theories that merely attempt to explain, to the
satisfaction of the thinker, a relationship between two or more observations. These theo-
ries may have their origins in a variety of places: "Some theories or beliefs about particu-
lar relationships may thus be based at least partially on data. Others may be based on
deductions from broader beliefs about the world, conventional folk wisdom, semantic as-
sociations, or persuasive communications by family, friends, or the mass media." Dennis
L. Jennings, Teresa M. Amabile, & Lee Ross, Informal Covariation Assessment: Data-
Based Versus Theory-Based Judgments, in Judgment Under Uncertainty, supra note 4, at
211, 227.
53 See Jerome S. Bruner & Mary C. Potter, Interference in Visual Recognition, 144 Sci.
424 (1964).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

of the sharpening process that subjects viewed, the less likely they
were to be able to identify the object in the photograph. That is, if a
subject began looking at the photo when it was severely blurred, that
subject had a more difficult time identifying the object than a subject
who began viewing the photo when it was less severely blurred. As
the researchers concluded, "[I]nterference may be accounted for
partly by the difficulty of rejecting incorrect hypotheses based on sub-
standard cues." 4 Put differently, if individuals construct their initial
hypothesis at a time when their basis for, or ability to, make such a
judgment is particularly weak, they may be unable to interpret cor-
rectly subsequent better information that is inconsistent with that
hypothesis. 55

iL Confirmatory Bias. Our inadequacies as scientists run

deeper than mere cognitive stubbornness. As Matthew Rabin re-
cently expressed the point, "[p]sychological evidence reveals a
stronger and more provocative phenomenon: People tend to misread
evidence as additional support for initial hypotheses."5 6 Numerous
experiments have demonstrated, for example, that the same ambigu-
ous information can further polarize people who already hold differ-
ing beliefs on certain topics. In one such experiment, researchers
asked forty-eight undergraduate students to review evidence on the
deterrent effect of capital punishment.57 Half of those students had
indicated in an earlier questionnaire that they supported capital pun-
ishment and believed that it had a deterrent effect and that research
supported their view; half indicated that they opposed capital punish-
ment and believed that it had little or no deterrent effect and that
research supported their view. The subjects were then exposed to ran-
domly selected studies on the deterrent effect of capital punishment
and asked to assess both whether a given study supported or discred-
ited their views and whether the study had affected their views in any

54 Id. at 424.
55 As will become clear from the examples provided below, this sort of perseverance
bias goes well beyond theories regarding optical uncertainties.
56 Rabin, supra note 24, at 26 (discussing people's reluctance to adapt beliefs to new
information).
57 See Charles G. Lord, Lee Ross, & Mark R. Lepper, Biased Assimilation and Atti-
tude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence, 37
J. Personality & Soc. Psychol. 2098,2099-2100 (1979); see also John M. Darley & Paget H.
Gross, A Hypothesis-Confirming Bias in Labeling Effects, 44 J. Personality & Soec. Psychol.
20, 22-25 (1983) (demonstrating similarly striking polarization effect in study regarding as-
sessment of students' skill levels); S. Pious, Biases in the Assimilation of Technological
Breakdowns: Do Accidents Make Us Safer?, 21 J. Applied Soc. Psychol. 1058, 1078-SO
(1991) (replicating result in study regarding judgment about safety of nuclear technology).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

way. 58 The researchers found that the subjects' attitudes were po-
larized at high confidence levels; that is, proponents of capital punish-
ment became more in favor of it and believed more in its deterrent
effect, while opponents of capital punishment became less in favor of
59
it and believed less in its deterrent effect.
Research of this sort also suggests that the strength of the confir-
matory bias may depend in part on the type of evidence an individual
views. The more ambiguous and complex the evidence, the more that
evidence seems to be susceptible to the confirmatory bias.60 As the
authors of the experiment regarding capital punishment put it: "With
confirming evidence, we suspect that both lay and professional scien-
tists rapidly reduce the complexity of the information and remember
only a few well-chosen supportive impressions." '6' With disconfirming
evidence, they continue to reflect upon any information that suggests
less damaging "alternative interpretations. '62 Indeed, lay scientists
may even view ambiguities or conceptual weaknesses in data opposing
their theory as somehow supporting their theory's fundamental verac-
ity (rather than simply serving to discredit any other theories that
63
would have relied on the flawed data).
In a slightly different study of the effects of confirming and dis-
confirming data, Nancy Pennington and Reid Hastie found that mock
jurors remember different facts depending upon the stories that they
construct to decide their verdicts.6a Not only do they forget facts that
are incongruent with their stories, they "remember" facts that were

58 See Lord et al., supra note 57, at 2100-01.

59 See id. at 2101-02.
60 See, e.g., Dale Griffin & Amos Tversky, The Weighing of Evidence and the Determi-
nants of Confidence, 24 Cognitive Psychol. 411,426-29 (1992) (describing "difficulty effect"
whereby subjects exhibit more pronounced overconfidence when presented with more dif-
ficult judgment tasks); Gideon Keren, Facing Uncertainty in the Game of Bridge: A Cali-
bration Study, 39 Organizational Behav. & Hum. Decision Processes 98, 113 (1987)
(finding that amateur bridge players exhibit significant overconfidence in predicting game
outcomes and proposing that "extremely complex" nature of bridge helps explain overcon-
fidence); Gideon Keren, On the Ability of Monitoring Non-Veridical Perceptions and Un-
certain Knowledge: Some Calibration Studies, 67 Acta Psychologica 95, 115-17 (1988)
(concluding that visual evidence is not particularly subject to bias and that bias may be
significant only where some abstraction and some need for interpretation exists); Rabin,
supra note 24, at 28 (noting that "[a]mbiguity of evidence is widely recognized to be an
important mediating factor in... confirmatory bias" (citations omitted)).
61 Lord et al., supra note 57, at 2099.
62 Id.
63 See id.
64 See Nancy Pennington & Reid Hastie, Evidence Evaluation in Complex Decision
Making, 51 J. Personality & Soc. Psychol. 242, 251 (1986) (concluding that "jurors who
choose different verdicts have different interpretations of the evidence, that is, different
stories"); Nancy Pennington & Reid Hastie, Explanation-Based Decision Making: Effects
of Memory Structure on Judgment, 14 J. Experimental Psychol.: Learning, Memory and

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

not in evidence. This type of "self-serving bias" 65 may be just as pro-

nounced among the litigants themselves. Consider, for example, sev-
eral studies involving an actual Texas tort case in which subjects were
randomly allocated the role of plaintiff or defendant in settlement ne-
gotiations. 66 The subjects received a case summary along with twenty-
seven pages of materials from the trial and were told that a judge had
awarded an amount between $0 and $100,000.6 Prior to the settle-
ment exercise, the subjects were asked to guess the amount that the
judge awarded and to state what they believed was a "fair" amount
for the plaintiff to receive in settlement.6 For the judge's award,
plaintiffs guessed an average $14,527 higher than defendants; for the
fair settlement amount, plaintiffs stated an average $17,709 higher
than defendants. 69 Thus, not only do individuals selectively recall evi-
dence to suit a particular story that they have constructed, but they
weigh the merits of an entire body of evidence wildly differently de-
pending on something as simple as which side they have been assigned
in a role-playing exercise. 70
Finally, according to behavioralists, one of the most significant
effects of the confirmatory bias is the tendency to exaggerate a corre-
lation when doing so confirms one's hypothesis or to underestimate a
correlation when one does not subscribe to a hypothesis or theory that
might explain the correlation.71 As an initial matter, Loren J. Chap-
man and Jean P. Chapman's seminal work on illusory correlations
demonstrated that clinicians and laypersons often perceive correla-
tions between variables based on their preconceived notions of the
relationship that "should" exist, rather than any actual correlations
between the variables. 72 Later researchers extended this concept to

Cognition, 521, 523-28 (1988) (describing how jurors considered facts confirming their ver-
sion of story to be stronger and more believable than facts opposing their version of story).
65 See Jolls et al., supra note 2, at 1501.
66 See id. at 1503-04.
67 See id at 1503.
6s 'See id.
69 See id. at 1503 (citing Linda Babcock, George Loewenstein, Samuel Issacharoff, &
Colin Camerer, Biased Judgments of Fairness in Bargaining, 85 Am. Econ. Rev. 1337
(1995), and George Loewenstein, Samuel Issacharoff, Colin Camerer, & Linda Babcock,
Self-Serving Assessments of Fairness and Pretrial Bargaining, 223. Legal Stud. 135 (1993)).
70 See id. at 1504.
71 See, e.g., Jennings et al., supra note 52, at 227-30 (noting "an actor's beliefs or expec-
tations can constitute a self-fulfilling prophecy" because of need for data to be consistent
with actor's theories).
72 For instance, clinicians were shown to perceive illusory correlations between draw-
ings and the personality traits of those who drew them. See Loren J. Chapman, Illusory
Correlation in Observational Report, 6 J. Verbal Learning & Verbal Behav. 151, 151-52
(1967) (noting that four studies failed to confirm clinicians' observation that paranoid pa-
tients drew figures with particularly elaborate eyes); Loren J. Chapman & Jean P. Chap-

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

cases in which the individual has no preconceived notion or "theory"

to explain observed data, and in which the individual does have a the-
ory but the data itself is imprecise or nonexistent. 73 These researchers
found that individuals underestimate correlation when they have no
theory to explain it, and imagine or exaggerate correlation when they
do have a preconceived theory to explain it.74 In short, the lay scien-
tist appears to be driven by theory rather than data when making
judgments and decisions-an approach dramatically at odds with the
classical scientific model.

iii. Hypothesis-Based Filtering. There is a related type of con-

firmatory bias, which Matthew Rabin has dubbed "hypothesis-based
filtering." 75 As Rabin explains, although it may be "sensible to inter-
pret ambiguous data according to current hypotheses, people tend to
use the consequent 'filtered' evidence as further evidence for these
hypotheses. ''76 Put differently, the confirmatory bias seems to have a
self-reinforcing and escalating quality: An individual interprets am-
biguous evidence as consistent with her initial hypothesis and then
views that evidence, as interpreted, as further confirmation of her hy-
pothesis. That confirmation in turn strengthens her faith in the initial
hypothesis and makes her even more willing to interpret future am-
biguous evidence as consistent with it. This process continues in circu-
lar fashion, further solidifying the initial "theories" or beliefs of the
lay scientist.

iv. Entity Effect. Finally, perhaps the most striking finding is

that people's hypotheses often take on a life of their own-that is,
they persevere even when the evidence that initially gave rise to them
is thoroughly and completely discredited. For example, it seems that
people's impressions of their own abilities (or the abilities of their
peers) often survive even after the evidence upon which the initial

man, Genesis of Popular but Erroneous Psychodiagnostic Observations, 72 J. Abnormal

Psychol. 193, 198-200 (1967) (observing significant tendency among clinicians to make illu-
sory correlations between personality traits and physical body parts typically associated
with them); Loren J. Chapman & Jean P. Chapman, Illusory Correlation as an Obstacle to
the Use of Valid Psychodiagnostic Signs, 74 J. Abnormal Psychol. 271, 274 (1969) (sug-
gesting that clinicians' correlation of traits with homosexuality is determined primarily by
strength of verbal associative connection with trait rather than by objective reality).
73 See Jennings et al., supra note 52, at 224 ("The theories we hold apparently lead us to
expect and predict stronger empirical relationships than actually exist; and many of the
empirical relationships that do exist, even ones of consequential magnitude, are apt to go
undetected unless we already expected to find them.").
74 See id.
75 Rabin, supra note 24, at 28 (emphasis removed).
76 Id.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

impression is based is invalidated. In one study, subjects, who had

been told that their social sensitivity and empathetic ability were be-
ing assessed, received fabricated feedback indicating their ability to
distinguish authentic from inauthentic suicide notes.77 Subjects were
then thoroughly "debriefed" in a way that made very clear that the
feedback they had received had been bogus. Subjects' subsequent
predictions of future task success and ratings of their own abilities
were nevertheless heavily influenced by the earlier (but discredited)
feedback.78 The same sort of perseverance effects were observable in
the predictions of observers who witnessed the subjects' original feed-
back and debriefings. 9
More recent studies have continued to demonstrate the same
phenomenon and, furthermore, have begun to shed light on the way in
which individuals' causal theories can take on their own momentum.
In one pair of experiments, for instance, subjects were provided a
small amount of evidence to suggest that either a positive or negative
empirical relationship existed between a trainee's preference for
risk-as measured by a paper and pencil test-and his or her subse-
quent success as a firefighter. 80 The fictitious evidence consisted of
background information, such as age, marital status, and hobbies, of
one successful firefighter and one unsuccessful firefighter along with
each firefighter's scores on the risk-preference test. Some of the sub-
jects were asked to provide an explanation for any relationship that
they discovered when reviewing that evidence.81 Then, as with the
previous study, some subjects were "debriefed"-that is, they were

77 See Lee Ross, Mark R. Lepper, & Michael Hubbard, Perseverance in Sef-Percep-
tion and Social Perception: Biased Attributional Processes in the Debriefing Paradigm, 32
J. Personality & Soc. Psychol. 880, 882-83 (1975).
78 See id. at 883-84 ("[E]ven after debriefing procedures... subjects continued to as-
sess their performances and abilities as if these test results still possessed some validity.").
79 See id. at 885-88. For additional evidence and discussion of this sort of phenomenon,
see Lee Ross & Craig A. Anderson, Shortcomings in the Attribution Process: On the
Origins and Maintenance of Erroneous Social Assessments, in Judgment Under Uncer-
tainty, supra note 4, at 129, 144-51 (observing that beliefs are "remarkably resilient in the
face of empirical challenges that seem logically devastating"); Lee Ross, Mark Lepper,
Fritz Strack, & Julia Steinmetz, Social Explanation and Social Expectation: Effects of Real
and Hypothetical Explanations on Subjective Likelihood, 35 J. Personality & Soc. Psychol.
817, 817 (1977) (noting that false feedback continued to influence actors' self-perceptions
even after that feedback was totally discredited through extensive debriefing procedure).
80 See Craig A. Anderson, Mark R.Lepper, & Lee Ross, Perseverance of Social Theo-
ries: The Role of Explanation in the Persistence of Discredited Information, 39 J. Person-
ality & Soc. Psychol. 1037, 1039-40 (1980).
81 By prompting deliberation in this manner, researchers may have tapped into the
false-confidence effect that individuals sometimes experience when they go through a pro-
cess of reasoning. See infra text accompanying notes 128-32 (describing unwarranted con-
fidence that often results from reasoning).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

informed that the evidence they were shown suggesting a relationship

between risk preferences and success as a firefighter was fictitious and
of absolutely no probative value. The researchers then examined
whether the subjects would continue after the debriefings to be influ-
enced by the theories they constructed before the debriefings. The
results were striking:
[A]lIthough the "data" to which subjects had been initially exposed
were objectively quite weak (consisting of only two cases) and in a
domain of little personal relevance, this initial ostensible evidence
clearly exerted a strong effect on subjects' theories about the true
relationship between the two variables. Thus, in the no-debriefing
conditions, subjects exposed to a positive relationship saw risky re-
sponses as highly diagnostic of later success, whereas subjects ex-
posed to an apparent negative relationship believed the opposite to
be true.... [T]he total discrediting of the evidence on which sub-
jects' initial theories had been based had only a minimal impact on
subjects' beliefs concerning the relationship between risk preference
and firefighting ability. Within the debriefing conditions, subjects
initially exposed to data indicative of a positive relationship contin-
ued to believe that a positive relationship existed, whereas subjects
in the negative relationship condition continued to believe in a neg-
ative relationship .... In fact, the slight decrease in the strength of
subjects' beliefs following debriefing . . . was not statistically
significant .... 82

The authors concluded that the studies provide support for three
general conclusions. 83 First, the studies provide further evidence for
the basic hypothesis that individuals persist in their beliefs even when
such persistence is neither normatively nor logically warranted. Sec-
ond, the studies illustrate that earlier research may have underesti-
mated the tenacity of belief perseverance: "[I]nitial beliefs may
persevere in the face of a subsequent invalidation of the evidence on
which they are based, even when this initial evidence is itself as weak
and inconclusive as a single pair of dubiously representative cases.18 4
Finally, the studies provide support for the hypothesis that belief per-
severance is exacerbated when an individual generates her own causal

82 Anderson et al., supra note 80, at 1041-42.

83 See id. at 1045.
84 Id. at 1045 (emphasis added). "That subjects will persevere in beliefs with such weak
empirical grounding in the face of a complete refutation of the formative evidence for
those beliefs seems eloquent testimony to the pervasiveness of our propensity to resist
changing our attitudes or beliefs." Id. at 1046. Notably, "[i]n everyday experience our
intuitive theories and beliefs are sometimes based on just such inconclusive data, but chal-
lenges to such beliefs and the formative data for those beliefs are rarely as decisive as the
discrediting procedures employed in this study." Id. at 1042.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

explanations or scenarios to imply the correctness of her initial beliefs,

even if the original data underlying those beliefs is later discredited. 8 5
It bears noting that the subjects in this experiment seemed to un-
derstand the debriefing procedure that thoroughly discredited the
original evidence. Nevertheless, subjects also seemed to believe that
their hypotheses were independent of the invalidated evidence and
were worthy of independent deference. As the authors of the study
explained, subjects "simply felt that the relationship they had ex-
amined-whether positive or negative-appeared to be the correct
one and that the discrediting of the evidential value of the initial cases
was largely irrelevant to their personal beliefs concerning the 'true'
relationship existing between these variables."8 6 If these studies pro-
vide an accurate indicator, then clearly the perseverance bias is pow-
erful and pervasive; attitudes and beliefs do not change easily. New
data or the refutation of old data does not necessarily influence our
beliefs because the hypotheses that we begin with are "largely autono-
mous" and they "remain available and continue to imply the existence
of particular relationships or outcomes even if the data on which they
were initially based subsequently prove to be completely devoid of
evidential value." 87

v. Motivated Reasoning. Belief perseverance, the confirmatory

bias, hypothesis-based filtering, and the entity effect may all be under-
stood as examples of a more general phenomenon known as motivated
reasoning.88 This refers to the tendency for individuals to utilize a
variety of cognitive mechanisms to arrive, through a process of appar-
ently unbiased reasoning, at the conclusion they privately desired to
arrive at all along. As Ziva Kunda has noted, "when one wants to
draw a particular conclusion, one feels obligated to construct a justifi-
cation for that conclusion that would be plausible to a dispassionate
observer. '89 In order to ensure that those justifications still reach the

85 The subjects who were "explicitly induced to explain the evidence [that] they ha[d]
been shown" exhibited this final result. See id. at 1047. The authors attribute this phe-
nomenon to two features of explanatory accounts: "First, such explanations are, by defini-
tion, selectively constructed to fit the evidence or outcome observed. Second, once
created, such explanations become largely autonomous of the initial data that led to their
postulation." Id.
86 Id. at 1045.
87 Id. at 1047.
88 See Ziva Kunda, The Case for Motivated Reasoning, 103 Psychol. Bull. 480, 480
(1990) ("[Motivation may affect reasoning through reliance on a biased set of cognitive
processes: strategies for accessing, construing, and evaluating beliefs.... [M]otivation can
be construed as affecting the process of reasoning- forming impressions, determining one's
beliefs and attitudes, evaluating evidence, and making decisions.").
89 Id. at 493.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

desired conclusion, "one accesses only a biased subset of the relevant

beliefs and rules." 90 Thus, not only does the lay scientist perceive and
interpret evidence in a manner designed to confirm initial hypotheses,
but she constructs the initial hypotheses themselves through biased
cognitive processes designed to "reason" toward a desired conclusion.
This selective belief construction can have unfortunate conse-
quences. As Kunda points out:
[P]eople who play down the seriousness of early symptoms of se-
vere diseases such as skin cancer and people who see only weak-
nesses in research pointing to the dangers of drugs such as caffeine
or of behaviors such as drunken driving may literally pay with their
lives for their motivated reasoning. 9 '
Despite the possibility for such regrettable consequences, the practice
of motivated reasoning appears to be a universal and, perhaps, immu-
table characteristic of human nature.

b. False Self-Confidence. The next set of biases, which involve

individuals' own self-assessments, are often treated by scholars as
independent from the biases described above. Whether or not they
are in fact independent, they can be usefully understood simply as
manifestations of the motivated reasoning biases. To see how, imag-
ine that individuals begin with the following hypotheses about them-
selves: "I'm more talented, intelligent, and agile than most; I'm a more
competent lay scientist than most; and I'm less likely to be injured
than most." g With that sort of self-confidence or theory of the self,
the following can be viewed as the perseverance and confirmatory bi-
ases, among others, at work in those people's assessment of the risks
that they themselves face. This section will summarize evidence that
people are indeed overoptimistic or overconfident regarding their
own susceptibility to risks.

i. Optimistic Bias.93 An old saying defines a pessimist as "an

informed optimist." Behavioral researchers might change that defini-
90 Id.
91
Id. at 496.
92 As Massimo Piattelli-Palmarini describes: "What is so worrisome is that we are vastly
overconfident in these intuitive judgments. We do not know, but we think we know. We
are ready to bet, not just money, but in certain cases our life, or the lives of others." Piat-
telli-Palmarini, supra note 3, at 31.
93 This bias is sometimes referred to as "overconfidence" or "unrealistic optimism." In
another article, one of us (with Kyle D. Logue) dubbed a closely related phenomenon the
"third-person effect"-the tendency for people to regard others as more susceptible to
exogenous risks than they themselves are. See Jon D. Hanson & Kyle D. Logue, The Costs
of Cigarettes: The Economic Case for Ex Post Incentive-Based Regulation, 107 Yale L.J.
1163, 1186-88 (1998). Overoptimism might also be called the Lake Wobegon effect

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

tion to "a debiased optimist," for there is growing evidence that we

are naively and stubbornly optimistic at heart, regardless of how well
informed we are. For instance, respondents in one study, although
correctly estimating that fifty percent of American couples end up in
divorce, estimated their own chance of divorce at zero.9 Similarly,
college students are six times more likely to think they will have above
average job satisfaction than below average. 95 They are also about six
times more likely to think they will own their own homes, two times
more likely to think they will have a mentally gifted child, seven times
less likely to think they will have drinking problems, and nine and
one-half times less likely to think they will divorce soon after mar-
riage. 96 Overoptimism such as this is present not just among freshman
7
and honeymooners; it exists even in the world of professionals.
More generally, it is not limited to any particular age, sex, education
level, or occupational group.98 In short, it appears that most people
are overconfident with regard to future life events, even when they
understand the actuarial probabilities of such events. 99

("where the women are strong, the men are good looking, and all of the children are above
average").
94 See Lynn A. Baker & Robert E. Emery, When Every Relationship is Above Aver-
age: Perceptions and Expectations of Divorce at the Tune of Marriage, 17 Law & Hum.
Behav. 439,443 (1993). Similarly, respondents were overly optimistic about likely results if
their marriage did fail. See id. (providing further evidence of such optimism with respect
to issues such as alimony and child custody in event of divorce).
95 See Neil D. Weinstein, Unrealistic Optimism about Future Life Events, 39 J. Person-
ality & Soc. Psychol. 806, 810 (1980).
96 See id.
97 See Larry T. Garvin, Adequate Assurance of Performance: Of Risk, Duress, and
Cognition, 69 U. Colo.L. Rev. 71, 149 (1998) (summarizing study results and literature);
see also Colin F. Camerer & Howard Kunreuther, Decision Processes for Low Probability
Events: Policy Implications, 8 J. Pol'y Analysis & Mgmt. 565, 569 (1989) (summarizing
empirical evidence).
98 See Neil D. Weinstein, Optimistic Biases About Personal Risks, 246 Science 1232,
1232 (1989) [hereinafter Weinstein, Optimistic Biases] ("[Optimistic bias] is robust and
widespread. It appears with diverse hazards and samples and with different questions used
to elicit the personal risks ratings ....Pessimistic biases are... rare."); Neil D. Weinstein,
Unrealistic Optimism About Susceptibility to Health Problems: Conclusions from a Com-
munity-Wide Sample, 10 J.Behav. Med. 481, 494-96 (1987) [hereinafter Weinstein, Conclu-
sions from a Community-Wide Sample] (determining from demographically diverse
sample that "optimistic biases are largely unrelated to age, sex, level of education, or occu-
pational prestige").
99 Research also indicates that those people who are not overconfident in their ability
to make assessments in general are nevertheless often overconfident in specific cases. See
Rabin, supra note 24, at 31 (citing Griffin & Tversky, supra note 60, and Andrea 0.
Baumann, Raisa B. Deber, & Gail G. Thompson, Overconfidence Among Physicians and
Nurses: The "Micro-Certainty, Macro-Uncertainty" Phenomenon, 32 Soc. Se. & Med. 167
(1991)). "Even if people learn the relevant statistical truths of their environment, they may
continue to make errors in their judgments and decision making in every single case."

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

One particular manifestation of this bias is the tendency of people

to underestimate their own chance of suffering some adverse outcome
even when they accurately state or even overstate everyone else's
chance of suffering that same outcome. In other words, we tend
to exhibit the "it can't happen to me" syndrome or, more proper-
ly, the "it's less likely to happen to me than the average person"
syndrome. As a result, ninety percent of drivers consider them-
selves to be above-average drivers, 100 and ninety-seven percent
of consumers believe that they are average or above average in
their ability to avoid bicycle or power-mower accidents. 01 Simi-
larly, more people than is logically possible perceive themselves
to be less likely than average to suffer negative experiences such
as disease, 10 2 divorce, 103 criminal victimization,104 or unemploy-
ment.105

Rabin, supra note 24, at 31. Kahneman and Tversky call such mistakes "errors of applica-
tion." Id. (emphasis removed).
It bears noting that such naive optimism may have its advantages: "[T]he only group
that seems consistently to get it right-to get subjective probabilities to mirror objective
probabilities-is the clinically depressed." See Garvin, supra note 97, at 149 (citing Lauren
B. Alloy & Lyn Y. Abramson, Judgment of Contingency in Depressed and Nondepressed
Students: Sadder but Wiser?, 108 J. Experimental Psychol. 441 (1979), and Benjamin M.
Dykman, Lyn Y. Abramson, & Jeanne S. Albright, Effects of Ascending and Descending
Patterns of Success upon Dysphoric and Nondysphoric Subjects' Encoding, Recall, and
Predictions of Future Success, 15 Cognitive Therapy & Res. 179 (1991)); see also
Weinstein, Optimistic Biases, supra note 98, at 1232 (summarizing studies suggesting that
optimism may be associated with less depression, greater willingness to work hard to make
optimism self-fulfilling, and increased physical health).
100 See Ola Svenson, Are We All Less Risky and More Skillful Than Our Fellow Driv-
ers?, 47 Acta Psychologica 143, 146 (1981).
101 See W. Kip Viscusi & Wesley A. Magat, Learning About Risk: Consumer and
Worker Responses to Hazard Information 95 (1987).
102 See James A. Kulik & Heike I.M. Mahler, Health Status, Perceptions of Risk, and
Prevention Interest for Health and Nonhealth Problems, 6 Health Psychol. 15, 24 (1987);
Arthur G. Miller, W.A. Ashton, J.W. McHoskey, & Joel Gimbel, What Price Attractive-
ness? Stereotype and Risk Factors in Suntanning Behavior, 20 J. Applied Soc. Psychol.
1272, 1298 (1990); Weinstein, Conclusions from a Community-Wide Sample, supra note 98,
at 486; Neil D. Weinstein, Unrealistic Optimism About Susceptibility to Health Problems,
5 J. Behav. Med. 441, 447 (1982) [hereinafter Weinstein, Unrealistic Optimism]; Neil D.
Weinstein, Why It Won't Happen to Me: Perceptions of Risk Factors and Susceptibility, 3
Health Psychol. 431, 436 (1984) [hereinafter Weinstein, Why It Won't Happen to Me].
103 See Baker & Emery, supra note 94, at 446-48.
104 See Weinstein, supra note 95, at 810; Weinstein, Why It Won't Happen to Me, supra
note 102, at 436.
105 See Weinstein, supra note 95, at 810. Perhaps the area most focused on by social
scientists studying this type of optimistic bias has been individual perceptions of the health
hazards of smoking. See, e.g., Ralf Schwarzer, Optimism, Vulnerability, and Self-Beliefs as
Health-Related Cognitions: A Systematic Overview, 9 Psychol. & Health 161, 162-63
(1994) (noting that "social comparison bias" can give rise to such flawed beliefs as, "My
fellow smokers might get lung cancer one day, but it is less likely that this would happen to
me"); Joop Van Der Pligt, Risk Perception and Self-Protective Behavior, 1 Eur. Psycholo-

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

Overoptimism stems from a general tendency to use past experi-

ence to estimate future susceptibility. As a result, the risks most likely
to cause naive optimism are those in which the individual believes that
if the problem has not happened yet, it will not happen at all.10 6
Other factors strongly correlated with the operation of the optimistic
bias include the perception that a risk is preventable by individual ac-
tion,10 7 the perception that a risk is of low frequency,"03 and the lack
of personal experience with a fisk.' 09 For policymakers, a final impor-
tant feature of the optimistic bias is that it is largely resistant to
"debiasing interventions." 110 In a study of the effects of the provision
of risk-factor information, Neil Weinstein and William Klein con-
cluded that "[flour studies testing a variety of approaches and using a
variety of health hazards were unsuccessful in reducing optimistic bi-
ases about familiar health hazards." ' In short, it seems that the opti-
mistic bias is an indiscriminate and indefatigable cognitive feature,
causing individuals to underestimate the extent to which a threat ap-

gist 34, 36 (1996) (same); see also Hanson & Kysar, TBS H, supra note 25, at 1511-16
(providing detailed review of studies demonstrating optimism among smokers).
106 See Weinstein, Conclusions from a Community-Wide Sample, supra note 98, at 496
("For many hazards, an optimistic bias is introduced into comparative risk judgments when
people extrapolate from their past experience (not having experienced a problem) to con-
clude that their future vulnerability is relatively low.").
107 The perception need not have any basis in truth. For instance, when researchers
asked gay men to rate the riskiness of their own sexual conduct for contracting AIDS, most
who engaged in high-risk activity did not rate their own risk as high. Their optimism was
apparently based on the false confidence they placed in their own risk-countering prac-
tices, such as inspecting their sexual partners for lesions or showering after sex. See Laurie
J. Bauman & Karolynn Siegel, Misperception Among Gay Men of the Risk for AIDS
Associated with Their Sexual Behavior, 17 J. Applied Soc. Psychol. 329, 34445 (19s7).
108 See id. at 331.
109 See id. Another explanation for the optimism bias is, as noted above, that people
want to maintain a self-conception of being above average. See supra text accompanying
notes 100-05. As Weinstein states, "[a]dmitting that peers are less susceptible to harm can
threaten our feelings of competence and self-worth." Weinstein, Optimistic Biases, supra
note 98, at 1232. And such threats are likely particularly strong in circumstances where a
risk is preventable or controllable, which explains the "strong optimism-controllability cor-
relation that exists." Id.; see also Kunda, supra note 88, at 483 (stating that people are
motivated to believe that their conduct is not harmful in order to preserve their own self-
esteem, but also that "people will come to believe what they want to believe only to the
extent that reason permits"); Shelley E. Taylor, Positive Illusions: Creative Self-Deception
and the Healthy Mind 243 (1989) ("[Tlhe idea that positive illusions are in the service of
self-esteem virtually requires that they stay in check. If one develops substantially unreal-
istic expectations regarding the future that greatly exceed what one is actually able to ac-
complish, then one is set up for failure and disappointment, leading to lower self-esteem.");
infra text accompanying notes 114-21 (describing "illusion of control" bias).
110 Neil D. Weinstein & William M. Klein, Resistance of Personal Risk Perceptions to
Debiasing Interventions, 14 Health Psychol. 132, 132 (1995).
111 Id. at 138.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

plies to them even when they can recognize the severity it poses to
others.

ii. Cognitive Dissonance. In addition to identifying countless

manifestations of overconfidence such as those discussed above,
behavioralists have attempted to explain the phenomenon using the
concept of cognitive dissonance. One aspect of this concept is the ten-
dency to reject or downplay information that contradicts other, more
favorable views about oneself. In light of this tendency, it is perhaps
easy to understand the basic source of people's optimistic bias: Peo-
ple prefer to believe that they are intelligent and are not subjecting
themselves to a substantial risk. In the face of a known risk, there-
fore, individuals come readily to the opinion that they themselves-
unlike the average person-are relatively immune, 112 and they hold
113
onto these optimistic assessments tenaciously.

iii. The Illusion of Control. Scholars have also identified a bias

known as the illusion of control,1 4 referring to the human tendency to
"treat chance events as if they involve skill and hence are controlla-
ble."' 1 5 A simple example of this tendency can be seen as gamblers
throw dice harder when they need high numbers and softer when they
need low numbers. The tendency was even more evident in a series of
experiments conducted by Ellen Langer to test the subjective values
people placed on lottery tickets." 6 The experiment involved a lottery
in which each participant was given a card containing the name and
picture of a National Football League player; an identical card was
112 See Weinstein, Unrealistic Optimism, supra note 102, at 456-57 (explaining that "op-
timism arises because people give themselves credit for factors in their favor... and be-
lieve that their self-protective actions are more extensive or effective than actions taken by
others"); Schwarzer, supra note 105, at 163 (describing "optimistic bias" that causes people
to believe that while their peers are at risk of falling victim to negative event, such event
will not happen to them); see also George A. Akerlof & William T. Dickens, The Eco-
nomic Consequences of Cognitive Dissonance, 72 Am. Econ. Rev. 307, 308-09 (1982):
In practice most cognitive dissonance reactions stem from peoples' view of
themselves as "smart, nice people." Information that conflicts with this image
tends to be ignored, rejected, or accommodated by changes in other beliefs....
Cognitive dissonance theory would suggest that persons in dangerous jobs
must decide between two conflicting cognitions. According to one cognition,
ego is a smart person who would not choose to work in an unsafe place. If the
worker continues to work in the dangerous job, he will try to reject the cogni-
tion that the job is dangerous.
113 See Weinstein & Klein, supra note 110, at 138-39.
114 See Ellen J. Langer, The Illusion of Control, 32 J. Personality & Soc. Psychol. 311,
311 (1975) (defining illusion of control as expectation of personal success that is inappro-
priately higher than objective probability of success).
115 Dawes, supra note 31, at 256 (discussing how individuals respond to uncertainty).
116 See Langer, supra note 114, at 316.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

placed in a bag and the person whose card was chosen from the bag
won the lottery. In one trial, cards were randomly allocated while in
another subjects were allowed to select their cards. Prior to the draw-
ing, subjects were offered a chance to sell their cards to an experimen-
tal collaborator. Subjects who had selected their own player cards
demanded more than four times as much money to sell their cards as
117
did those with randomly allocated cards.
In another experiment, Yale undergraduates were convinced that
they were better or worse than the average person at predicting the
outcome of coin tosses. 118 By giving subjects feedback on their pre-
dictions that was rigged to show either much greater or much less than
50% accuracy, researchers led subjects to view themselves, not as
lucky or unlucky coin-toss predictors, but as good or bad predictors.
Indeed, a significant portion of the subjects reported that their per-
formance would be hampered by distraction and that they would im-
prove with practice!' 19 Finally, Amos Tversky and Ward Edwards
uncovered a similar overconfidence in subjects' tactics for predicting
the results of random selections from a bag in which 70% of chips
were blue and 30% were red.12o The maximizing strategy is to predict
blue every time, leading to an eventual success rate of 70%. Tversky
and Edwards' subjects, however, attempted to match probabilities-
they predicted blue only 70% of the time in hopes of hitting 100% of
the results. That strategy, however, only leads to a 58% success rate
((.70 x .70) + (.30 x .30) = .58). Despite being paid for the accuracy of
their predictions and despite receiving feedback through thousands of
trials, most subjects refused to adopt the maximizing strategy of al-
ways predicting blue-the strategy which, of course, admits that skill
and control are irrelevant to the task.Ul
iv. Hindsight Bias. In circumstances in which individuals are
confronted with unambiguous evidence of a past outcome, they often
construct a hypothesis from which they claim that they could have and
would have predicted (and perhaps did predict) that outcome. In es-
sence, most of us believe ourselves to be better than average at pre-
dicting outcomes and we use known outcomes as our means of
confirming our own "scientific" acumen. This widely recognized and
117 See id.
118 See Ellen J. Langer & Jane Roth, Heads I Win, Tails It's Chance: The Illusion of
Control as a Function of the Sequence of Outcomes in a Purely Chance Task, 32 J. Person-
ality & Soc. Psychol. 951 (1975).
119 See id. at 954.
120 See Amos Tversky & Daniel Kahneman, Judgment Under Uncertainty. Heuristics
and Biases, 185 Science 1124, 1129 (1974).
121 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

researched cognitive tendency is known as the hindsight bias.122 In

1975, one of the founders of the cognitive-bias literature, Baruch
Fischhoff, observed that "[r]eporting an outcome's occurrence in-
creases its perceived probability of occurrence; and .. people who
have received outcome knowledge are largely unaware of its having
changed their perceptions ...."12 Apparently, the mere fact that a
probabilistic outcome eventuates alters our perception of how proba-
bilistic it was. Indeed, subjects even tend to overestimate the accuracy
of their own estimations. That is, not only do subjects overestimate
the likelihood of an event after learning of its occurrence, but they
also incorrectly recall their own earlier predictions as being more ac-
curate than they actually turned out to be. 124 This tendency to exag-
gerate the extent to which our beliefs before an event would have
been our beliefs (and that others' beliefs before the event should have
been their beliefs'2) is widely acknowledged even outside of the
scholarly literature; hence, the expressions "20/20 hindsight" and
"Monday-morning quarterbacking.' 26 That we may be aware of the
bias seems not to eliminate it.127

v. The Surprising Effect of Reasoning. It is sometimes sug-

gested that individuals become more realistic when they have experi-
ence with a certain risk. To be sure, it is generally the case that some
experts have a reasonably correct sense of the accuracy of their own
predictions. But that tends to be the case primarily in the limited cir-
cumstances in which "feedback takes the form of unambiguous statis-
122 See, e.g., Hal R. Arkes, Thomas J. Guilmette, David Faust, & Kathleen Hart, Elimi-
nating the Hindsight Bias, 73 J. Applied Psychol. 305 (1988).
123 Baruch Fischhoff, Hindsight * Foresight: The Effect of Outcome Knowledge on
Judgment Under Uncertainty, 1 J. Experimental Psychol.: Hum. Perception & Perform-
ance 288, 288 (1975).
124 See Dawes, supra note 31, at 119-20 (discussing "hindsight bias").
125 See Scott A. Hawkins & Reid Hastie, Hindsight: Biased Judgments of Past Events
After the Outcomes Are Known, 107 Psychol. Bull. 311, 319 (1990); Rabin, supra note 24,
at 30 n.25 (noting that "the evidence suggests that we have a tendency to think that other
people 'should have known"').
126 The legal system has also recognized this bias in some ways-consider, for example,
one of the rationales for corporate law's "business judgment rule." See, e.g., Ski
Roundtop, Inc. v. Hall, 658 P.2d 1071, 1083 (Mont. 1983) (Morrison, J., dissenting) ("The
'business judgment rule' is employed to immunize corporate officers and directors where
they act in good faith, but, as a matter of hindsight, it can be determined that they did not
take the course of action which hindsight shows to have been the best selection.").
127 For more recent and thorough discussions of the hindsight bias, see Hawkins &
Hastie, supra note 125 (reviewing empirical research concerning hindsight phenomena);
Jay JJ. Christensen-Szalanski & Cynthia Fobian Willham, The Hindsight Bias: A Meta-
Analysis, 48 Org. Behav. & Hum. Decision Processes 147 (1991) (reviewing 122 studies
revealing hindsight bias and noting moderation of its effect depending on subject's famili-
arity with task and type of outcome information presented).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

tical evidence."' In contrast, when the pertinent events are not

easily predictable and the feedback is not unambiguous, experts tend
to be even more overconfident than laypersons. The explanation
sometimes offered for that finding is closely related to the "entity ef-
fect" of causal theories described above. 12 9 Specifically, because ex-
perts tend to develop elaborate causal theories with which to generate
predictions, the independent weight of those theories can yield espe-
cially robust overconfidence. As Griffin and Tversky explain,
If the future state of a mental patient, the Russian economy, or the
stock market cannot be predicted from present data, then experts
who have rich models of the system in question are more likely to
exhibit overconfidence than lay people who have a very limited un-
derstanding of these systems. Studies of clinical psychologists ...
130
and stock market analysts... are consistent with this hypothesis.
The counterintuitive lesson is this: Elaborate theories and careful rea-
soning may foster, rather than mitigate, overconfidence.
It should be noted that this phenomenon, although most pro-
nounced in experts, is not limited to them. Several studies of layper-
sons have demonstrated that the quality of decisionmaking sometimes
declines when individuals are prompted to deliberate and offer rea-
sons for their decision.1 3 ' Because the individuals' confidence levels
do not similarly decline, the result is an overconfidence fostered by
deliberation.
Timothy 'Wilsonand Suzanne LaFleur recently conducted an ex-
periment that illustrates this false-confidence-building effect of rea-
soning.1 32 In this experiment, members of six sororities at the
University of Virginia were asked to predict whether they would en-
gage in each of six different behaviors toward fellow sorority members
during the upcoming semester. Randomly, some subjects were asked
to list reasons why they might or might not perform each of the behav-
iors. At the end of the semester, subjects were asked whether they

128Rabin, supra note 24, at 32.

129See supra notes 77-87 and accompanying text.
130Griffin & Tversky, supra note 60, at 430 (citing Stuart Oskamp, Overconfidence in
Case-Study Judgments, 29 J. Consulting Psychol. 261 (1965), and Frank J. Yates, Judgment
and Decision Making (1990)).
131 For instance, in a study of undergraduate students, Timothy D. Wilson and Jonathan
W. Schooler demonstrated that student assessments of the overall quality of strawberry
jams comported far less with the assessments of trained sensory experts when the students
were asked to provide reasons for their judgments than when they were left to their own
devices. See Timothy D. Wilson & Jonathan NV. Schooler, Thinking Too Much: Introspec-
tion Can Reduce the Quality of Preferences and Decisions, 60 J. Personality & Soc.
Psychol. 181 (1991).
132 See Timothy D. Wilson & Suzanne J. LaFleur, Knowing What You'll Do: Effects of
Analyzing Reasons on Self-Prediction, 68 J. Personality & Soc. Psychol. 21 (1995).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [V/ol. 74:630

had performed each of the six behaviors. While both groups of sub-
jects-those who reasoned about their predictions and those who did
not-had stated roughly the same confidence level in their predictions
at the beginning of the semester, those who reasoned were in fact sig-
nificantly less accurate than the control group in their predictions.
Thus, both laypeople and experts exhibit a heightened degree of over-
confidence when their predictions and decisions are characterized by a
process of reasoning.

c. Bad Statisticians. Special attention must be given to the sub-

ject of probabilistic judgment, for this is an area where researchers
have uncovered a veritable fool's gold mine of nonrational cognitive
133
anomalies.
The seminal work in the field remains Tversky and Kahneman's
Judgment Under Uncertainty: Heuristics and Biases.1 34 In it, they de-
scribe the way in which individuals rely on heuristics, or mental rules
of thumb, to "reduce the complex tasks of assessing probabilities and
predicting values to simpler judgmental operations.' ' 135 These heuris-
tics provide a rough and ready means for approximating the results of
complex computational tasks; however, they can lead to biases, or "se-
vere and systematic errors"13 6 in probabilistic judgment.

i. Availability. In making decisions about uncertain future

events, people tend to ignore statistical data in favor of evidence that
is particularly salient, vivid, or easily "available" to them. 137 That is,
their probability judgment is driven by the ease with which they can
recall previous occurrences of the event or the ease with which they
can imagine the event occurring in the future.' 38 Much of the time a
person's experience or available memories can provide a fairly sound

133 See, e.g., Paul Slovic, Judgment, Choice and Societal Risk Taking, in Judgment and
Decision in Public Policy Formation 98, 99-100 (Kenneth Hammond ed., 1978) ("By and
large .. research indicates that intelligent people systematically violate the principles of
rational decision making when judging probabilities, making predictions, or otherwise at-
tempting to cope with probabilistic tasks.").
134 See Tversky & Kahneman, supra note 4.
135 Id. at 3.
136 Id.
137 See generally Susan T. Fiske & Shelley E. Taylor, Social Cognition 142-79, 245-94 (2d
ed. 1991) (providing general review of availability heuristic in contexts of social schemas
and social encoding).
138 In the words of Tversky and Kahneman, who coined the term, "a person... em-
ploy[s] the availabilityheuristic whenever he estimates frequency or probability by the ease
with which instances or associations could be brought to mind." Amos Tversky & Daniel
Kahneman, Availability: A Heuristic for Judging Frequency and Probability, 5 Cognitive
Psychol. 207, 208 (1973) (emphasis added).

Imaged with the Permission of N.Y.U. Law Review

June 1999) TAKING BEHAVIORALISM SERIOUSLY

sample on which to base a judgment. At other times, however, the

1 39
availability heuristic biases an individual's estimate.
For instance, in one experiment Tversky and Kahneman asked
subjects to state whether words that begin with "r" are more or less
common in the English language than those that have an "r" as the
third letter. 14° A significant majority of the subjects said that words
that begin with an "r" are more likely. The statistical fact is the re-
verse, but words that begin with "r" are easier to recall or generate
than words that have an "r" in the third place.1 4 1 In another of their
experiments, subjects listened to recorded lists of the names of thirty-
nine people, some of whom could be described only as well known,
the remainder of whom could be described as famous.1 42 Some lists
included nineteen women and more famous women than men, some
included nineteen men and more famous men than famous women.
When asked whether men or women predominated on the list they
had heard, eighty of ninety-nine subjects chose the sex that involved
143
the larger number of famous individuals.
The availability heuristic can be quite powerful in our everyday
decisions. In making risk assessments, individuals will often allow
"available" evidence to trump much more probative statistical infor-
mation of which they are aware. As Matthew Rabin describes:
[O]ur assessment of a given city's crime rate is likely to be too influ-
enced by whether we personally know somebody who has been as-
saulted, even if we are familiar with much more relevant general
statistics. Likewise, dramatic stories by people we know about diffi-
culties with a brand of car are likely to be overly influential even if
we are familiar, via Consumer Reports, with general statistics of the
reliability of different brands. 144

139 In general, use of the availability heuristic will yield acceptable results when a par-
son's experience and memory of observed events corresponds well with actual event fre-
quencies. It is likely to lead to overestimation if recall or imagination is enhanced (e.g.,
easily imagined scenario, recent experience, dramatic or salient event), and underestima-
tion if recall or imagination is difficult (e.g., scenario is not easily imagined, no recent
experience, no other knowledge of event). See Richard Nisbett & Lee Ross, Human Infer-
ence: Strategies and Shortcomings of Social Judgment 18-23 (1980).
140 See id. at 21.
141 For those who have their doubts about just exactly where the letter "r" is more com-
monly located, consider another of Tversky and Kahneman's findings. Subjects asked
about the percentage of words ending in "ing" give a much higher estimate than subjects
asked about words in which the next-to-last letter is "n"-though it is clear that there must
be more of the latter. See Amos Tversky & Daniel Kahneman, Extensional Versus Intui-
tive Reasoning: The Conjunction Fallacy in Probability Judgment, 90 Psychol. Rev. 293,
295 (1983).
142 See id. at 208.
143 See id.
144 Rabin, supra note 24, at 30.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

That is, we use availability as a substitute for rigorous scientific or

probabilistic analysis not only when we lack information on a given
subject, but also when we have information but lack the cognitive will-
power to utilize it properly.

ii. Representativeness. There is another well-known mental

rule of thumb, which scholars have dubbed the representativenessheu-
ristic.1 45 This refers to the tendency of individuals to judge the fre-
quency or likelihood of something by the degree to which the
individual believes it resembles something else (i.e., its class). One
consequence of representativeness is that people often pay too much
attention to specific details (which may or may not contain relevant
information), while ignoring or paying too little attention to back-
ground information that is relevant. To understand this heuristic, con-
146
sider a famous experiment conducted by Tversky and Kahneman.
Subjects were given the following information and asked to respond
to a set of questions in an anonymous survey:
A panel of psychologists have interviewed and administered person-
ality tests to 30 engineers and 70 lawyers, all successful in their re-
spective fields. On the basis of this information, thumbnail
descriptions of the 30 engineers and 70 lawyers have been writ-
ten.... For each description, please indicate your probability that
the person described is an engineer, on a scale from 0 to 100....
Jack is a 45-year-old man. He is married and has four children. He
is generally conservative, careful, and ambitious. He shows no in-
terest in political and social issues and spends most of his free time
on his many hobbies which include home carpentry, sailing, and
mathematical puzzles.
The probability that Jack is one of the 30 engineers in the sample of
100 is _%.
Dick is a 30-year-old man. He is married with no children. A man
of high ability and high motivation, he promises to be quite success-
ful in his field. He is well liked by his colleagues.
The probability that Dick is one of the 30 engineers in the sample of
100 is _%.

145 See Tversky & Kahneman, supra note 4, at 4:

[P]eople typically rely on the representativeness heuristic, in which probabili-
ties are evaluated by the degree to which A is representative of B, that is, by
the degree to which A resembles B. This approach to the judgment of
probability leads to serious errors, because similarity, or representativeness, is
not influenced by several factors that should affect judgments of probability.
146 See Daniel Kahneman & Amos Tversky, On the Psychology of Prediction, 80
Psychol. Rev. 237 (1973).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

Suppose now that you are given no information whatsoever about

an individual chosen at random from the sample. The probability
that this man is one of the 30 engineers in the sample of 100 is

Tversky and Kahneman found that responses to the first two ques-
tions were often based on how much the described person was judged
to sound like an engineer or lawyer without regard to the 30:70 ratio.
For example, the description of Jack was judged with very high
probability to involve an engineer. Of course, only three in ten will be
engineers. But that fact seems to have been forgotten by many. The
description of Bill was intended to provide no information. Tversky
and Katneman found that Bill was judged at 50:50, not 30:70. As they
explained, most subjects recognized that the description said nothing
relevant about a lawyer-engineer distinction, but apparently forgot the
base rate information. If subjects believed that the description did not
help them determine the occupation of the person, then they should
have gone back to base rates and estimated thirty percent. Finally,
with respect to the third question, Tversky and Kahueman found that
subjects correctly gave 30:70 odds, indicating that they did know how
147
to use the base rate information.
From their many studies of this sort, Tversky and Katneman con-
clude that "[e]vidently, people respond differently when given no evi-
dence and when given worthless evidence. When no specific evidence
is given, prior probabilities are properly utilized; when worthless evi-
dence is given, prior probabilities are ignored."1 48 Worthless informa-
tion trumps more relevant evidence, resulting in a failure to honor
base rates.
Tversky and Kahneman observed a related misconception in a
study of experienced research psychologists who honored the falla-
cious law of smallnumbers. This is the belief that "even small samples
are highly representative of the populations from which they are
drawn."'1 49 As a result of this bias, the psychologists in the experi-
ments, who were asked to make predictions from a limited set of evi-
dence and assess their confidence levels in the predictions, afforded
too much credence to results from small samples and "grossly overes-
timated the replicability of such results."' 50
In another demonstration of this effect, Tversky and Kahneman
asked undergraduate students the following question:

147 See Tversky & Kahneman, supra note 4, at 5 (summarizing original study).
148 Id.
149 Id. at 7.
150 Id. at 8.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW (Vol. 74:630

A certain town is served by two hospitals. In the larger hospital

about 45 babies are born each day, and in the smaller hospital about
15 babies are born each day. As you know, about 50 percent of all
babies are boys. However, the exact percentage varies from day to
day. Sometimes it may be higher than 50 percent, sometimes lower.
For a period of 1 year, each hospital recorded the days on which
more than 60 percent of the babies born were boys. Which hospital
do you think recorded more such days? 151
The same percentage-twenty-two--of respondents estimated
that either the large or the small hospital recorded more. The rest-
fifty-six percent-estimated that it was about the same for both hospi-
tals. Only the twenty-two percent that responded that the small hos-
pital recorded more were correct; that is, seventy-eight percent were
wrong. The law of large numbers-which, unlike the law of small
numbers, is not fallacious-tells us that the larger hospital with more
births per day on average would be more likely to replicate the overall
distribution. Subjects obviously did not understand that, nor did they
understand that the small hospital would be less likely to replicate the
overall distribution. 152
Individuals experience an inverse bias to the law of small num-
bers whenever they honor the gambler's fallacy, the belief that the
roulette wheel is "due" to hit red after a long streak of black.
"Chance is commonly viewed as a self-correcting process in which a
deviation in one direction induces a deviation in the opposite direc-
tion to restore the equilibrium."'1 53 Thus, just as Tversky and
Kahneman's psychologist subjects believed that a global pattern could
be predicted from a local sample, gamblers believe that a local sample
must replicate the global pattern. In truth, neither supposition is
valid, for probabilities only become meaningful over sufficiently large
54
sample sizes.
A final manifestation of the law of small numbers is that people
expect too few lengthy streaks in sequences of random events. 55 In
the face of such a streak, people tend to invent spurious explanations
for the seeming deviation from randomness, when, in fact, there has
been no such deviation. For example, basketball coaches, players, and
spectators have long believed in something called a "hot hand"-
when a basketball player has made a series of shots and when that
streak seems too long to be random. Psychologists have shown, how-

151 See id. at 6.

152 See id.
153 Id. at 7.
154 See id. at 7-8.
155 See id.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

ever, that a seemingly hot-handed player is in fact no more likely to

make his or her next shot than at other times. Put differently, there
appears to be no such thing as a truly "hot hand," only random streaks
15 6
of success that appear to us too long to be truly random.
The danger in using representativeness as shorthand for proper
probabilistic decisionmaking is significant. The erroneous belief or
desire to see patterns' 5 7 in random events causes decisionmakers to
have unrealistic expectations about the stability of observed patterns
and the replicability of prior beneficial experiences. Likewise, the
failure to honor base rates when making predictions can lead to esti-
mates bearing little or no relationship to actual probabilities. In either
case, decisionmakers may also tend to have undue confidence in their
assessments, believing that the world is much more ordered and deter-
ministic than it actually is.

iii Anchoring and Adjustment. In yet another series of their

famous experiments, Tversky and Kahneman identified the heuristics
of anchoring and adjustment.l5 8 For reasons that should become
clear, this heuristic can be understood as a combination of framing
and perseverance. When people are asked to generate an estimate,
they frequently anchor on an obvious or convenient number (e.g., the
mean or the mode) and then adjust upward or downward from that
anchor if there is reason to believe that the correct number should be
moved in either direction. This procedure naturally leads to estima-
tions which are skewed toward the initial value.
For instance, in one of Tversky and Kahneman's experiments,
subjects were asked to estimate the number of African countries in
the United Nations as a percentage of total membership.15 9 Before
the subjects responded, however, a large wheel of chance was spun in
the subjects' presence. Though the wheel contained numbers from
one to one hundred, it was rigged to land either on ten or sixty-five.
When the wheel landed on ten, subjects estimated that African coun-
tries comprised twenty-five percent of the United Nations; when the
wheel landed on sixty-five, the estimation rose to forty-five percent. 16°

156 See Amos Tversky & Thomas Gilovich, The "Hot Hand": Statistical Reality or Cog-
nitive Illusion?, 2 Chance 31, 31 (1989); Thomas Gilovich, Robert Vallone, & Amos
Tversky, The Hot Hand in Basketball: On the Misperception of Random Sequences, 17
Cognitive Psychol. 295, 296 (1985); Amos Tversky & Thomas Gilovich, The Cold Facts
About the "Hot Hand" in Basketball, 2 Chance 16, 16 (1989).
157 This desire may be connected to the more general tendency to construct causal theo-
ries. See supra note 52 and accompanying text.
158 See Tversky & Kahneman, supra note 4, at 14.
159 See id.
160 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

What is striking about this demonstration is that the anchorprovided

to the subjects was overtly random and irrelevant,yet still it had a sig-
nificant impact on the subjects' intuitive judgments.
For clarity, consider another demonstration of the anchoring ef-
fect. Subjects were given ten seconds to compute the result of the
following multiplication:
2x3x4x5x6x7x8
Kahneman and Tversky found that the average response given
was 512. Then a second group of subjects was given the same prob-
lem, only arranged in reverse order:
8x7x6x5x4x3x2
This time the average response was 2,250.161 Thus, the subjects
anchored on the result of the first few steps of multiplication and
came up with vastly different estimates of products which, by the com-
mutative property, must be identical.
This disparity is quite understandable: Actual calculation of the
products would have been impractical given the short time allowed,
and subjects naturally begin the estimation by reading left to right
along the series. This tendency to read left-to-fight results in esti-
mates that are anchored near the product obtained by multiplying the
first few numbers in the series, thereby making the order of the series
a key factor in the size of the subjects' estimates. It is for this reason
that the order of the factors matters-another framing effect. The ef-
fect occurs even though subjects know that they read left to right; they
simply do not correct for it. But that is not all. The correct answer is
40,320. When subjects are told that they have underestimated the
product, their self-correcting guesses are systematically lower than the
true result, and always remain anchored to their initial answers. 162
The lower the initial figure, the lower the final result, after correction.
As one scholar explains:
It is extremely rare that a subject decides to invert the order of cal-
culation and do it right-to-left. It does not help to know that the
product does not depend on the order of factors. We know this, but
our intuitive estimates cannot use this piece of knowledge. This is
what often happens to us: one corner of our mind is unable to use
what another corner of our mind knows full well.
...[O]nce we have made an intuitive estimate, even if we are
told that we are wrong, we still keep the initial rough estimate as an

161 See id. at 15.

162 See Piattelli-Palmarini, supra note 3, at 29 (describing anchoring effect).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

implicit baseline. We are anchored16to3 it. We are unwilling to ne-

glect it completely and start afresh.

d. Experiential Thinking, Affect, and the Perception of

Risk. There is wide consensus among psychologists that individuals
process information through two parallel mechanisms: a rational sys-
tem and an emotionally driven experiential system. 16 The former is
logical, deliberate, and abstract. As a consequence, it is also ineffi-
cient and ill-suited for the majority of everyday judgments and deci-
sions. Instead, individuals typically rely on experiential thinking that
is largely automatic and removed from consciousness. This type of
information processing is associated with intuitive judgments, emo-
tional responses, and other subtle, nonconscious reactions to external
stimuli. These reactions are referred to collectively in the literature as
affect. 165
Significantly, "because [experiential thinking] is generally associ-
ated with affect, it is apt to be experienced as more compelling than is
dispassionate logical thinking." 166 That is, individuals often will be
swayed by the force of their affective responses to events and deci-
sions, regardless of whether their rational, sequential, analytical sys-
tem would opt for a different course. Moreover, "because the
influence is usually outside of awareness, the rational system fails to
control it because the person does not know there is anything to
control."' 67
Reliance on experiential thinking has important consequences for
individual perceptions of risk. In contrast to the expected utility maxi-
mizer of the economist's model-who would assess all risks based
solely on a probabilistic analysis of costs and benefits-the experien-
tial thinker can be expected to view risks as multidimensional con-
cepts entailing a range of beliefs, prejudices, and predispositions.
Because most information processing occurs automatically and effort-
lessly "outside of awareness,"'' 6 the individual's perception of a given
risk is likely to be heavily influenced, if not determined, by the affect
163 Id. (first emphasis added); see also Paul Slovic & Sarah Lichtenstein, Comparison of
Bayesian and Regression Approaches to the Study of Information Processing in Judgment,
6 Org. Behav. & Hum. Performance 649, 712-16 (1971) (demonstrating that people's ad-
justments from sometimes arbitrary anchors in making numerical estimates of uncertain
quantities are usually insufficient).
164 See Seymour Epstein, Integration of the Cognitive and the Psychodynamic Uncon-
scious, 49 Am. Psychologist 709 (1994) (reviewing wide variety of psychological theories of
cognition that support this bifurcated view).
165 See id. at 710-13.
166 Id. at 716.
167 Id.
168 Id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

associated with that risk. 169 To give just one example of how this type
of processing might work, assume that individuals are empirically
more likely to associate a feeling of dread with technologically im-
posed risks such as nuclear power than naturally imposed risks such as
earthquakes. If that were the case, individuals could be expected to
perceive the magnitude of a technologically imposed risk to be greater
than a probabilistically equal natural risk, solely because of the indi-
vidual's negative affective response to the idea of technological
hazards.
In an early and influential series of articles, Paul Slovic, Baruch
Fischhoff, and Sarah Lichtenstein began to explore seriously this po-
tential for qualitative, affective responses to bias individual risk per-
ceptions. 170 Utilizing a psychometric study of the determinants of risk
perception, these researchers were able to identify a range of charac-
teristics that heavily influence the way in which individuals perceive a
given risk. For instance, as described above, a feeling of "dread" can
have powerful influences over the risk that individuals associate with a
given activity or technology: "The higher a hazard's score [on a
'dread' impression scale] the more people want to see its current risks
reduced, and the more they want to see strict regulation employed to
achieve the desired reduction in risk.' 171 Conversely, the more
hazards are viewed as "controllable," "voluntary," and "well-known,"
the less likely individuals will view the hazard as excessive or undesir-
able. 172 Other relevant attributes include whether the hazard is ob-
servable, whether its effect is immediate or delayed, whether it is a
"new" risk, whether its consequences are fatal, whether it is globally

169 See Melissa L. Fmucane, Ali Alhakami, & Paul Slovic, The Affect Heuristic in Judg-
ments of Risks and Benefits 3 (Sept. 29, 1998) (unpublished manuscript, on file with the
New York UniversityLaw Review) (reporting studies showing that people often use "affect
heuristic," which improves judgmental efficiency by deriving risk and benefit evaluations
from common source).
170 See P. Slovic, B. Fischhoff, & S. Lichtenstein, Accident Probabilities and Seat Belt
Usage: A Psychological Perspective, 10 Accident Analysis & Prevention 281 (1978); Paul
Slovic, Baruch Fischhoff, & Sarah Lichtenstein, Characterizing Perceived Risk, in Perilous
Progress: Managing the Hazards of Technology 91 (Robert W. Kates et al. eds., 1985);
Paul Slovic, Baruch Fischhoff, & Sarah Lichtenstein, Cognitive Processes and Societal Risk
Taking, in Cognition and Social Behavior 165 (John S. Caroll & John W. Payne eds., 1976);
Paul Slovic, Baruch Fischhoff, & Sarah Lichtenstein, Facts and Fears: Understanding Per-
ceived Risk, in Societal Risk Assessment: How Safe is Safe Enough? 181 (Richard C.
Schwing & Walter A. Albers, Jr. eds., 1980); P. Slovic, B. Fischhoff, & Sarah Lichtenstein,
Perceived Risk: Psychological Factors and Social Implications, 376 Proc. Royal Soc'y
London 17 (1981); Paul Slovic, Baruch ischhoff, & Sarah Lichtenstein, Rating the Risks,
21 Environment 14 (1979).
171 Paul Slovic, Perception of Risk: Reflections on the Psychometric Paradigm, in Social
Theories of Risk 117, 121 (Sheldon Krimsky & Dominic Golding eds., 1992).
172 Id.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

catastrophic, and whether it poses high risks to future generations.1 r

What is significant is that none of these attributes are relevant under a
strict probabilistic assessment of expected costs and benefits. Thus,
the principal determinants of risk perception are affective and experi-
ential, not deliberate and rational.
While early researchers focused on the risk characteristics that
tend to lead toward positive or negative evaluative feelings about cer-
tain risks, more recent studies have focused on the actual mechanism
whereby affective responses influence and bias risk perceptions. 174
The most robust finding from the literature is that perceptions of the
costs and benefits posed by risks are inversely related. That is, where
a risk is perceived as posing high costs, it tends also to be perceived as
posing low benefits, and vice versa.175 In a series of experiments
designed to explore this relationship, Ali Siddiq Alhakami and Paul
Slovic found that affect significantly influences the degree and direc-
tion of individual risk perceptions:
When the attitude [about an activity or technology] is favorable, the
activity or technology being judged may be seen as having high ben-
efit and low risk. On the other hand, when the item being evaluated
is viewed unfavorably, with negative affect, it may be seen as having
low benefit and high risk. Our general attitudes176or affective states
may thus "confound" the risk/benefit judgment.
Thus, individuals consistently fail to assess costs and benefits sepa-
rately, as an analytic approach to decisionmaking would require, and
instead seem to allow their overall impression of the risk to bias both
judgments. In that manner, and consistent with the early findings of
Slovic, Fschoff, and Lichtenstein, affect becomes a crucial determi-
nant of individual risk perceptions. 177
As one can see, the manner in which individuals actually perceive
and understand uncertainty departs significantly from the actuarial as-
sessments of the scientist or the expected utility functions of the econ-
omist. We wish to emphasize that this does not necessarily mean that
human understanding of risk is misguided or naive simply because it
does not comport with the probabilistic understanding of risk. As
Paul Slovic has eloquently put it, "[r]isk does not exist 'out there,'
independent of our minds and cultures, waiting to be measured.

173 See Paul Slovic, Perception of Risk, 236 Science 282, 282-83 (1987).
174 See, e.g., All Siddiq Alhakami & Paul Slovic, A Psychological Study of the Inverse
Relationship Between Perceived Risk and Perceived Benefit, 14 Risk Analysis 1035
(1994); Fmucane et al., supra note 169, at 8.
175 See Alhakami & Slovic, supra note 174, at 1085.
176 Id. at 1088.
177 See supra notes 170-73 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

Human beings have invented the concept 'risk' to help them under-
stand and cope with the dangers and uncertainties of life. There is no
such thing as 'real risk' or 'objective risk."1 78 Instead, "real risk" and
"objective risk" merely represent the central concepts of an alterna-
tive theory of how to perceive and comprehend uncertainty in the
world, a theory that is necessarily laden with the subjective values and
beliefs of the individuals who constructed and perpetuate it. Our
point is simply that there is no reason, standing alone, to give primacy
to probabilistic judgments of risk, especially when the overwhelming
experience of individuals bears little or no relationship to strict proba-
bilistic assessments.

2. Manipulability of Preferences
The preceding discussion should underscore the fact that we are
far from flawless assessors of scientific and probabilistic judgments.
Unlike the classical economic actor who "can perfectly process avail-
able information about alternative courses of action, and can rank
possible outcomes in order of expected utility,179 human individuals
display a startling ineptitude for comprehending causality and
probability. We mistake familiarity with numerosity, we allow our
predictions to be skewed by something as arbitrary and irrelevant as
the spinning of a wheel, and we report that with a little more practice
and concentration, we might do a better job at predicting coin tosses.
Apart from these biases, however, behavioralists have also identi-
fied numerous ways inwhich preferences-holding probability esti-
mates constant-can vary by context.180 We offer descriptions of
these biases to emphasize that the way in which humans think about
and exert their preferences is complicated. Von Neumann and
Morgenstern's elegantly simple conception of expected utility was just
that: simple. Behavioral researchers have shown that we do not order
our preferences according to simple maximization axioms, but rather
our preferences are shaped at least in part by the manner in which
they are elicited: "Alternative descriptions of the same choice
problems lead to systematically different preferences; strategically
equivalent elicitation procedures give rise to different choices; and the
preference between x and y often depends on the choice set in which
they are embedded."'' In this manner, for the past three decades,
178 Slovic, supra note 171, at 119.
179 Ellickson, supra note 7, at 23.
180 For an overview of these findings, see Paul Slovic, The Construction of Preference,
50 Am. Psychologist 364 (1995).
181 Amos Tversky, Rational Theory and Constructive Choice, in The Rational Founda-
tions of Economic Behaviour 185, 186 (Kenneth J. Arrow et al. eds., 1996).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

behavioral researchers have been offering a compelling criticism of

expected utility theory; the time now has come to incorporate their
findings into legal analysis so that we may have a fuller, more textured
model of human decisionmaking. 18

a. The Status Quo Bias and Endowment Effect. There is what

one economist has called "[o]verwhelming evidence show[ing] that
humans are often more sensitive to how their current situation differs
from some reference level than to the absolute characteristics of the
situation. 18 3 Put differently, individuals have a tendency to prefer the
state of affairs which they perceive as the status quo rather than
switching to an alternative state, other things being equal. That ten-
dency is sometimes referred to as the status quo bias.184
Russell Korobkin has demonstrated the status quo bias in an ex-
periment involving the reaction of first year law students to a hypo-
thetical contractual setting in which they were asked to advise a
shipping company during negotiations with a commercial customer. 1'
By changing the applicable contract default rule between subjects,
Korobkin found that "subjects exhibited a largeand statisticallysignifi-
cant bias in favor of the term embedded in the default nde-the status
quo term. 18 6 Thus, for example, subjects in one group were asked to
recommend a per-package dollar amount (on a scale of one dollar to
ten dollars) that the shipping company should pay to contract around
a default rule of full consequential damages liability; subjects in a sec-
ond group were asked to recommend an amount that the company
should demand in order to accept full liability beyond a statutory de-

182 A similar call made by Herbert Simon to the economic community more generally is
worth noting:
Economics without psychological and sociological research to determine the
givens of the decision-making situation, the focus of attention, the problem
representation, and the processes used to identify alternatives, estimate conse-
quences, and choose among possibilities-such economics is a one-bladed scis-
sors. Let us replace it with an instrument capable of cutting through our
ignorance about rational human behavior.
Herbert A. Simon, Rationality in Psychology and Economics, in Rational Choice, supra
note 6, at 25, 39-40.
183 Rabin, supra note 24, at 13 (citation omitted); see also Jalls et al., supra note 2, at
1535 ("[P]eople evaluate outcomes based on the change they represent from an initial ref-
erence point, rather than based on the nature of the outcome itself ....").
184 The status quo bias has been known to legal scholars for some time under such
names as the offer/asking problem or the willingness to pay versus willingness to accept
problem. See, e.g., Hoffman & Spitzer, supra note 7; Duncan Kennedy, Cost-Benefit
Analysis of Entitlement Problems: A Critique, 33 Stan. L Rev. 387, 401 (1981).
185 See Russell Korobkin, he Status Quo Bias and Contract Default Rules, 83 Cornell
L. Rev. 608, 634-41 (1998).
186 Id. at 639 (emphasis added).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

fault of limited liability. 187 Subjects in the first group recommended

paying a maximum of $4.46 to limit consequential damages liability,
while subjects in the second group recommended a minimum asking
price of $6.96 to accept full liability. 18 8 Similarly, when faced with a
choice between "pay your own fees" or "loser pays" attorney fee pro-
visions, subjects overwhelmingly preferred whichever provision was
designated as the state default rule, despite a clear admonition that,
"if the parties wish to specify [otherwise] by contract.., such a provi-
sion would be fully enforceable."' 89 These two findings, along with
others provided by Korobkin,190 vividly demonstrate for legal scholars
the manner in which the status quo bias affects how individuals regis-
ter their preferences.
Many behavioral tendencies related to the status quo bias have
been identified. For instance, when it comes to choosing between op-
tions that represent deviations from the status quo, people are sub-
stantially more averse to losses from some given reference point than
they are attracted to same-sized gains.1 91 In addition to this loss aver-
sion, behavioralists have also identified an endowment effect. 192 Once
an individual comes to possess an item, she instantaneously (or nearly
so) values that item more than she did prior to possessing it. In one
well-known study, for example, several scholars randomly allocated to
subjects either a coffee mug or six dollars. 193 Mug holders were then
187 See id. at 637-41.
188 See id. at 639.
189 Id. at 646. Seventy-two percent of subjects presented with a "loser pays" default rule
preferred to retain it for the parties' agreement, while 59% of subjects presented with a
"pay your own fees" default opposed switching to a "loser pays" rule. See id. at 64647.
190 Korobkin provides several fascinating permutations of the experiment, including sce-
narios involving a recent legislative reversal of the default rule (to rule out substantive
attachment to the 'traditional' rule) and an industry custom of contracting around the de-
fault (to clarify what exactly subjects would perceive as the default rule). After demon-
strating the operation of the status quo bias in the context of contract default rules,
Korobkin goes on to challenge the conventional law and economics wisdom that the selec-
tion of default rules matters only when parties face high transaction costs or asymmetrical
information. See id. at 664-75.
191 See Daniel Kahneman, Jack L. Knetsch, & Richard H. Thaler, The Endowment Ef-
fect, Loss Aversion, and Status Quo Bias, 5 J. Econ. Persp. 193, 197-201 (1991) (explaining
how preference for status quo derives in part from finding that losses and gains of equal
amount do not have identical impact on decisions but rather that losses generally exert
stronger impact).
192 One may view the endowment effect as a particular instance of the status quo bias:
It is the tendency of individuals to value items which they already own higher than an
identical item which they do not own. See Korobkin, supra note 185, at 625-30 (describing
distinction in significantly more detail); see also Cass R. Sunstein, Endogenous Prefer-
ences, Environmental Law, 22 J. Legal Stud. 217, 230 (1993) (describing endowment effect
as "special instance" of status quo bias).
193 See Daniel Kahneman, Jack L. Knetsch, & Richard H. Thaler, Experimental Tests of
the Endowment Effect and the Coase Theorem, 98 J. Pol. Econ. 1325, 1330 (1990).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

asked to state the minimum amount they would be willing to accept to

sell their mugs, while cash holders were asked to state the maximum
amount they would be willing to pay for a mug. Despite the fact that
mugs were distributed randomly, mug holders as a group turned out
to value mugs at approximately twice the amount that cash holders
did.194 Scholars now attribute this peculiar behavior, which has been
replicated in many studies, to the endowment effect: The initial allo-
cation of mug ownership affected the value which subjects attached to
mugs.1 95
In a recent study by George Loewenstein and Daniel Adler, sub-
jects were asked to "[i]magine that we gave you a mug exactly like the
one you can see, and that we gave you the opportunity to keep it or
trade it for some money.' 96 Minimal selling prices were then elicited
from subjects both before and after they actually received mugs. This
formulation of the now familiar mug study allowed Loewenstein and
Adler to test the ability of subjects to predict the operation of the
endowment effect. Prior to receiving the mugs, subjects on average
predicted that their minimal selling price would be $3.73.197 Once
they actually received the mugs, however, their minimal selling price
averaged $5.40.198 Thus, subjects significantly and systematically un-
derestimated the impact that the endowment effect would have on
their valuations. As with so many of these biasing or preference-alter-

194 See id. at 1332.

195 Another possible explanation would be that those subjects who happened to receive
mugs coincidentally had a higher utility for mugs even before the experiment began (per-
haps the random allocation disproportionately hit coffee rather than soda drinkers). Be-
sides being statistically unlikely, this explanation is negated by the fact that Kalmeman,
Knetsch, and T1haler repeated the experiment over several trials with the same subjects,
and attained similar results. See id.
The status quo bias is also exhibited in multiple-good choice problems. Loss aversion
predicts that individuals would prefer the status quo to changes that involve losses of some
goods even if compensated with other goods. In one study, for example, students were
randomly given either candy bars or mugs. Later, each student was offered the opportu-
nity to exchange her gift for the other one. Ninety percent of both groups chose not to
trade. See Jack L. Knetsch, The Endowment Effect and Evidence of Nonreversible Indif-
ference Curves, 79 Amer. Econ. Rev. 1277, 1278 (1989); Jack L. Knetsch & J. A. Sinden,
Willingness to Pay and Compensation Demanded: Experimental Evidence of an Unex-
pected Disparity in Measures of Value, 99 QJ. Econ. 507, 508 (1984) (explaining that stud-
ies demonstrated "the compensation demanded to give up an entitlement far exceeded the
respondents' indicated willingness to pay to maintain it").
196 George Loewenstein & Daniel Adler, A Bias in the Prediction of Tastes, 105 Econ. J.
929, 931 (1995).
197 See id. at 932.
198 See id. As Matthew Rabin has noted, Loewenstein and Adler's "procedure underes-
timates the true degree of misperception, because people don't like to contradict recently
expressed predictions of their own behavior." Rabin, supra note 24, at 36. 'Iis can be
seen in the fact that subjects who had made no prediction at all averaged a selling price of
$5.62. See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

ing phenomena, even when people appreciate their effect, they under-
estimate their magnitude. 199

b. Context Effects and the Effect of Irrelevant Op-

tions. Behavioralists have identified a variety of ways in which the
context of a choice influences the choice itself. 200 For example, re-
searchers have discovered that the addition of a new option to a set of
options may increase the proportion of individuals who choose one of
the initial options.201 This anomaly, an example of context effects, has
been demonstrated by student subjects asked to train rats with electric
shocks. In one trial, subjects could only select between "mild" and
"slightly painful" shocks; in other trials, a third option was present,
labeled either "moderately painful" or "extremely painful. ' '202 Sub-
jects were told not to use the more extreme option and none did; thus,
the researchers were able to observe whether an individual's prefer-
ence between A and B would change in the presence of an irrelevant
alternative C. While the "slightly painful" option was selected only
twenty-four percent of the time in the first trial, it was chosen twenty-
eight percent of the time when "moderately painful" was also present
and thirty-nine percent of the time when "extremely painful" was
present.203 In another experiment designed to test this phenomenon,
subjects were offered a choice of either six dollars or an attractive
Cross pen, and only thirty-six percent chose the pen. However, when
subjects were offered a three-way choice among the cash, the Cross
pen, and an inferior pen, forty-six percent chose the Cross pen.204
It is likely that context effects such as these originate from our
view of ourselves as "lay scientists." As "lay scientists," we want "rea-
sons" (or "theories" or "hypotheses") for our decisions or conduct.
The reasons need not actually be good; they just need to be good
enough. As Loewenstein writes:

199 See Loewenstein & Adler, supra note 196, at 935-36; see also supra notes 125-27
(discussing similar evidence); infra notes 217-18 (same); infra text accompanying notes 232-
33 (same); infra notes 269-75 (same).
200 See, e.g., Itamar Simonson, Choice Based on Reasons: The Case of Attraction and
Compromise Effects, 16 J. Consumer Res. 158, 170-71 (1989) (discussing how addition of
various types of options can influence decision to choose one option over another); Itamar
Simonson & Amos Tversky, Choice in Context: Tradeoff Contrast and Extremeness Aver-
sion, 29 J. Marketing Res. 281, 281-82 (1992) (explaining how tendency to choose one
option over another depends on factors such as tradeoffs implied in set of options and
whether outcome is below reference point (loss) or above reference point (gain)).
201 See id.
202 Michael Harrison & Albert Pepitone, Contrast Effect in the Use of Punishment, 23 J.
Personality & Soc. Psychol. 398, 400-01 (1972).
203 See id. at 400.
204 See Simonson & Tversky, supra note 200, at 287.

Imaged with the Permission of N.Y.U. Law Review

Junme 1999] TAKING BEHAVIORALISM SERIOUSLY

Most people experience their own actions as resulting from deci-

sions ...or at least as deliberate. However, it is questionable
whether these introspections represent veridical reports of underly-
ing decision processes, or ex post rationalizations of behavior. The
limitation of verbal reports is well established ....Trained to view
behavior as the result of attribute-based decisions... most people
in Western culture will almost inevitably interpret their own behav-
ior accordingly. 05
The point is that the addition of a new option can provide a "reason"
that otherwise was not available for choosing one of the initial op-
tions. As such, the findings "suggest[ ] that a simple axiom of conven-
tional economic theory-involving the irrelevance of added, unchosen
alternatives-is wrong."206

c. Elastic Justification. Christopher Hsee has recently uncov-

ered another way in which people's preferences can be altered
through decision-irrelevant factors. 20 7 By focusing on differences in
individuals' decisions when expected outcomes are expressed as an
uncertain range of estimates (a condition that Hsee refers to as "elas-
ticity") rather than a fixed point estimate, Hsee has found that indi-
viduals construct "reasons" to justify conclusions that they are already
privately motivated to reach.20 8 Specifically, this process, which Hsee
has termed elastic justification, refers to the tendency for individuals
to use the fact that an option presents a range of outcomes as a justifi-
cation for choosing among options in a way that, in the absence of
elasticity, the individual would not find justifiable.
This bias may be best understood through an example:
A salesman is asked by his company to take a trip to find as many
buyers for a certain product as possible. He has a choice of two
destination cities: He can find 60 buyers in one city and 40 in the
other but the city with fewer buyers is more enjoyable.2 0 9
As formulated, the decision should be clear: The salesman
should travel to the first city because the number of buyers (the A
factor) is a factor relevant to his task, while the enjoyability of the city
(the B factor) is not. Hsee theorizes, and supports through empirical
experiments, that if the decision is reformulated such that the first city
will produce somewhere between thirty and ninety buyers, the sales-

205 George Loewenstein, Out of Control: Visceral Influences on Behavior, 65 J. Org.

Behav. & Hum. Decision Processes 272, 276 (1996) (citations omitted).
206 Sunstein, supra note 2, at 1182.
2X See Christopher Y. Hsee, Elastic Justification: How Tempting but Task-Irrelevant
Factors Influence Decisions, 62 J. Org. Behav. & Hum. Decision Processes 330,330 (1995).
208 See id.
209 Id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

man will be more likely to select the second, more enjoyable city.
This is so regardless of the fact that the first city still presents an ex-
pected number of buyers equal to the first formulation of the decision,
sixty.
The mental process at work in preference reversals such as these
is simple:
[E]iasticity in the A factor allows the decision maker to bias his or
her view of the A values of the options in such a way that the B-
superior option may seem not as inferior on the A factor as it origi-
nally is, and this biased view makes it more justifiable for the deci-
sion maker to choose the B-superior option.210
In other words, the salesman will view the expected number of buyers
in the less enjoyable city as being toward the low end of the range
(say, thirty), thereby making it seem more justifiable to travel to the
more enjoyable city. Conversely, if the first city still presented a fixed
sixty buyers, but the second city presented a range of between ten and
seventy buyers, the salesman would view the expected number of buy-
ers in the second city as being toward the high end of the range (say,
seventy). In that manner, regardless of which option is reformulated
to have an elastic number of buyers, the salesman will find a way to
"justify" selecting the option that is less attractive from the perspec-
tive of the only relevant criterion: the expected number of buyers.
This bias of decisionmaking parallels the mechanisms described
earlier in connection with motivated reasoning. Both elastic justifica-
tion and motivated reasoning capture ways in which individuals pur-
port to "justify" or "reason through" their beliefs and decisions, when
in actuality their initial impulses have biased the process all along.
d. Time-Variant Preferences. There is a great deal of evidence
that, other things being equal, individuals will choose an activity that
will deliver immediate benefits and delay any perceived costs.2 1 ' We
prefer, in other words, immediate gratification and delayed dissatis-
faction. But as this evidence suggests, there is a problem with that
pair of preferences. Specifically, people's short-term willingness to
delay rewards in exchange for higher returns in the future is often less
than people's long-term willingness to delay those same rewards. For
instance, most people say they would rather have a prize of a $100
check available immediately rather than a $200 certified check that
could not be cashed for two years, but do not prefer a $100 check that
could be cashed in six years to a $200 certified check that could be
210 Id.
211 See Hanson & Logue, supra note 93, at 1203-05 (discussing evidence for problem of
myopic discounting).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

cashed in eight years,12 even though this is the same choice seen at six
2
years greater delay
In a recent experiment, Kris Kirby and Richard Herrnstein asked
subjects to choose among a series of pairs of smaller, earlier rewards
and larger, later rewards. 213 In two experiments offering subjects ac-
tual monetary rewards, twenty-three of twenty-four subjects "consist-
ently reversed their choices from the [smaller, earlier reward] to the 14
[larger, later reward] as the delay to both rewards increased."2
When the moment of choice is relatively far away, we tend to give
proper weight to the later consequence; however, when the moment
of choice arrives, the smaller but earlier consequence overshadows the
later one, causing an "impulsive" reversal of the original preference.
In short, our willingness to delay gratification (or encounter risks) var-
ies over time.2 15 Today we believe that we should stop smoking or
diet tomorrow, but tomorrow we feel we should continue smoking or
overeating, at least for another day.
In part because of that sort of intertemporal variation, econo-
mists have begun to model individuals as a collection of selves-
"multiple selves"-each self with its own set of preferences.2 16 Schol-
ars have employed the multiple-selves model to explain many sorts of
common behavior that otherwise seem inexplicable within the con-
ventional, single-self, rational-actor model. Greek mythology teaches
us of Ulysses, who had himself tied to the mast in order to hear but
withstand the song of the Sirens. Researchers today have identified
many analogous self-commitment (sometimes called "pre-commit-
ment" or "hands-tying") strategies that many of us employ in our
daily lives, such as setting up automatic monthly transfers to invest-

212 See George Ainslie & Nick Haslam, Hyperbolic Discounting, in Choice over Time
57, 69 (George Loewenstein & Jon Elster eds., 1992).
213 Kris N. Kirby & R.I. Hermstein, Preference Reversals Due to Myopic Discounting
of Delayed Reward, 6 Psychol. Sci. 83, 84 (1995).
214 Id. at 85.
215 For additional treatments of this phenomenon, see Philip J. Hilts, Smoke Screen:
The Truth Behind the Tobacco Industry Cover-Up 51-52 (1996) (summarizing Dr. Dunn's
study finding that question of who does or does not smoke is really "a question about who
is unable to resist the quick gain of nicotine now versus the later catastrophic conse-
quences" and that "[i]n simplified form, this means those who have many other small and
large pleasures in their lives are likely to be better at resisting nicotine's lure"); Hanson &
Logue, supra note 93, at 1203-05 (discussing how temporal separation of smoking's short-
term benefits and long-term costs helps explain why people who want to quit smoking do
not quit); Jolls et al., supra note 2, at 1479 (calling this "bounded wilpower," and referring
"to the fact that human beings often take actions that they know to be in conflict vith their
own long-term interests"); Rabin, supra note 24, at 38-41 (reviewing studies).
216 See Hanson & Logue, supra note 93, at 1205-09 (summarizing Thomas Schelling's
groundbreaking work on multiple selves); Rabin, supra note 24, at 39-40 (noting that multi-
ple selves hypothesis is "ready-made" for use by economic theorists).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

ment accounts, entering alcohol treatment programs or diet clinics

that restrict exit, or devising internal rules of conduct such as "only
smoke outdoors. '217 As with many of the other cognitive-biasing,
preference-changing phenomena, individuals seem, at some level, to
understand this source of preference shifting. Nevertheless they do
not always fully appreciate the extent of deviation between the prefer-
ence sets, and, in any case, cannot always manage to locate or create
218
successful self-commitment (or pre-commitment) strategies.

e. Reciprocity and Attribution. Most economic models assume

that preferences are defined solely over outcomes. But that assump-
tion is clearly inaccurate. As reviewed above, for instance, choices are
often determined in significant part by the extent or direction of
deviation of the options from some reference point or by the presence
or absence of certain options. In addition, there is considerable evi-
dence indicating that preferences are also a function of an actor's per-
ceived conduct or intentions to which a particular outcome is
attributed. 21 9 "The same people who are altruistic toward deserving
people are often indifferent to the plight of undeserving people, and
'220
motivated to hurt those whom they believe to have misbehaved.
Put differently, there is considerable evidence to suggest that
preferences are "reciprocal." 22 ' Reciprocity norms manifest them-

217 See Rabin, supra note 24, at 40 (discussing other examples of self-commitment strat-
egies); see also Hanson & Logue, supra note 93, at 1206-08 (same); Jolls et al., supra note 2,
at 1479 (same).
218 See Hanson & Logue, supra note 93, at 1208-09 (discussing evidence that 70% of
current smokers say they would like to quit, and suggesting that real world purchasing
contexts make self-commitment strategies likely to fail); Loewenstein, supra note 205, at
272, 274-76 (explaining how visceral factors create discrepancies between "perceived self-
interest and behavior"); Rabin, supra note 24, at 41 (discussing importance of self-control
for economic theory).
219 The distinction between outcomes and conduct attributed to the outcomes-be it
inevitable, accidental, reckless, malicious, or intentional-is a distinction recognized in the
law, albeit sometimes only implicitly. In tort law, for instance, these sorts of distinctions
play a significant role in the doctrinal divisions between intentional and unintentional torts
as well as in the doctrinal boundaries of consent and assumption of risk. Attribution the-
ory may also help explain punitive damages. Where an injury is attributed to an inten-
tional act, the harm is more pronounced than it would be where it is not attributed to an
intentional act. That is, there is the injury, plus the insult. For that reason, the expectation
of redress may be greater, thereby justifying higher damages. These thoughts are explored
more fully in Jon D. Hanson & Ana C. Reyes, Law and Attribution: Toward a New Posi-
tive Theory of Tort Law (Apr. 10, 1999) (working title for unpublished work in progress).
220 Rabin, supra note 24, at 21. See also Jolls et al., supra note 2, at 1479 ("In many
market and bargaining settings ... people care about being treated fairly and want to treat
others fairly if those others are themselves behaving fairly.").
221 See Rachel T.A. Croson, Theories of Altruism and Reciprocity: Evidence from Lin-
ear Public Goods Games (May 1998) (unpublished manuscript, on file with the New York
University Law Review) (reviewing literature); Robyn M. Dawes & Richard H. Thaler,

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

selves in numerous ways, most of which seem relevant to the market

context. For example, individuals are often more willing to cooperate
with those actors they feel are behaving cooperatively or fairly. 2m On
the other hand, individuals will often refuse to cooperate with others
who are being uncooperative. Moreover, individuals are often will-
hag to sacrifice to hurt others who are being unfair. A consumer may
refuse to buy a product sold by a monopolist at an "unfair" price, even
if she hurts herself by foregoing the product. 4
These sorts of phenomena have been repeatedly demonstrated in
experiments employing one or another version of what game theorists
call "the ultimatum game."' ' -5 In the most basic version of the game
(or experiment), two players are given a sum of money to split. One
of the two players, known as the Proposer, decides how the sum
should be split and announces that decision to the Responder, who is
left with the option of either accepting or rejecting that split. If the

Anomalies: Cooperation, J. Econ. Persp. 187, 190-92 (1988) (same); Richard E. Goranson
& Leonard Berkowitz, Reciprocity and Responsibility Reactions to Prior Help, 3 J. Per-
sonality & Soc. Psychol. 227,232 (1966) (providing classic experimental evidence); Rabin,
supra note 24, at 21-22 (reviewing literature). In a superb article, Rabin has explained
much of the evidence regarding the reciprocal nature of human behavior with the follow-
ing three assumptions:
(A) People are willing to sacrifice their own material well-being to help those
who are being kind.
(B) People are willing to sacrifice their own material well-being to punish
those who are being unkind.
(C) Both motivations (A) and (3) have greater effect on behavior as the ma-
terial costs of sacrificing becomes smaller.
Matthew Rabin, Incorporating Fairness into Game Theory and Economics, 83 Am. Econ.
Rev. 1281, 1282 (1993).
2n See generally Daniel Kahneman, Jack L Knetsch, & Richard H. Thaler, Fairness as a
Constraint on Profit Seeking: Entitlements in the Market, 76 Am. Econ. Rev. 728, 740
(1986) (describing results of household surveys of public opinions regarding what pricing
conduct is perceived as "fair" and "unfair," examining implications of those community
norms for market outcomes, and confirming earlier market evidence that fairness consider-
ations have significant market implications); Arthur M. Okun, Prices and Quantities. A
Macroeconomic Analysis 170 (1981) (describing some of ways in which fairness considera-
tions may influence pricing behavior in consumer markets); Eldar Shafir & Amos Tversky,
Thinking Through Uncertainty: Nonconsequential Reasoning and Choice, 24 Cognitive
Psych. 449, 454-55 (1992) (describing results of prisoner's dilemma study in which subjects
cooperated 16% of time when other player was known to have cooperated, compared with
3% of time when other player was known to have competed).
223 See supra note 221 (citing authorities); see also Rabin, supra note 24, at 21-22 (re-
viewing evidence).
224 See Kabneman et al., supra note 222, at 736 (describing willingness to drive out of
one's way to shop at store not engaged in unfair practices); Rabin, supra note 24, at 22
(discussing psychology of reciprocity); see also Jolls et al., supra note 2, at 1480, 1489-93
(explaining that parties "will be willing to punish unfair behavior even at personal financial
cost").
225 See generally Colin Camerer & Richard H. Thaler, Anomalies: Ultimatums, Dicta-
tors and Manners, 9 J. Econ. Persp. 209 (1995) (reviewing ultimatum game literature).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

Responder rejects the offer, neither of the two players gets any of the
money, and the game is over. For most economic or game-theoretic
models, which assume self-interest on the part of all actors, the out-
come was extremely easy to predict: Proposers would split the sum
(e.g., ten dollars) disproportionately such that the Responder would
get a nominal amount (e.g., one penny) and the Proposer would get
the rest. That is the prediction, but "[i]nstead, offers typically average
about thirty to forty percent of the total, with a fifty-fifty split often
the mode. Offers of less than twenty percent frequently are re-
jected.122 6 Self-interest clearly does not capture the complete psycho-
logical picture of findings such as these.
In a related manner, people tend to care about whether an in-
jurer's act was intentional or volitional or whether, instead, the injury
was, in some sense, unpreventable. Thus, where the injurer's act is
volitional, the victim is far more likely to reciprocate or retaliate with
negative actions. To get some sense of how this sort of reaction might
play a role in bargaining or market relationships, consider a recent
ultimatum-game experiment conducted by Sally Blount, in which she
asked three groups of subjects to state their willingness to accept take-
it-or-leave-it offers made by anonymous parties regarding how to split
ten dollars.2 27 The first group was told that the offer was coming from
anonymous students and that their response would affect the division
between them and the anonymous students. The second group was
told that a third party would determine the offer and that the third
party would not be affected by their response. The final group was
told that the offer would be produced randomly by computer. In one
study, subjects accepted average minimal offers of $2.91, $2.08, and
$1.20, respectively. Thus, subjects were more willing to accept low of-
fers when those offers were not perceived as the result of volition by
the person who would be hurt by rejecting the offer. 228

ft Preference-TrumpingEffect of Visceral Factors. While many

of the cognitive biases discussed in the behavioral literature address
ways in which people incorrectly perceive their self-interest, George
Loewenstein has recently offered a compelling theory which seeks to
account for cases in which people correctly perceive their self-interest,

226 Id. at 210.

227 See Sally Blount, When Social Outcomes Aren't Fair: The Effect of Causal Attribu-
tions on Preferences, 63 J. Org. Behav. & Hum. Decision Processes 131, 134-36 (1995).
2z See id. Interestingly, subjects seemed less offended by low offers when those offers
purportedly came from a computer rather than a third party, even though the computer
and the third party played identical roles in the experiment.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

yet still behave in contradiction to that interest.2 He argues that dis-

junctions between behavior and perceived self-interest result from vis-
ceral factors such as the feelings associated with drug addiction, drive
states like hunger, thirst and sexual desire, moods and emotions, and
physical pain. "At sufficient levels of intensity, these, and most other
visceral factors, cause people to behave contrary to their own long-
term self-interest, often with full awareness that they are doing so."230
The most obvious example of such self-destructive behavior can
be seen in the actions of drug addicts, who often consume drugs to the
exclusion of nourishment, sleep, and even the desire to survive.P 1
The visceral factor-drug addiction-causes the impacted individual
to narrow her focus almost exclusively to the object of that factor,
drugs. In a similar manner, visceral factors can cause the individual to
experience "a good-specific collapsing of one's time-perspective to-
ward the present.' ' 232 Thus, a hungry person is likely to make short-
sighted tradeoffs between food now and food delayed, even if that
person knows that tomorrow's hunger will be equally intense. Loew-
enstein explains such behavior by identifying two central premises of
the visceral factor theory. First, visceral factors that are experienced
now tend to dominate one's attention, crowding out all goals other
than satiating the drive behind the visceral factor. Second, visceral
factors that will be experienced in the future, have been experienced
in the past, or are experienced by other people tend to be discounted
by individuals; that is, individuals fail to predict, recall, or perceive in
others the force of visceral factors correctlyp3
Together, these two factors form the basis for a new account of
impulsivity. Rather than simply being prone to hyperbolic discount
functions, as some commentators have argued a2 4 individuals might be
suffering the "effect of visceral factors on the desirability of immediate
consumption. ' ' 235 That is, the impact of visceral factors such as hunger
might cause individuals both to overestimate the desirability of imme-
diate consumption and underestimate the desirability of delayed con-
sumption. As a result, they will behave impulsively.236 The advantage

229 See Loewenstein, supra note 205, at 289. In addition to his general account of vis-
ceral forces, Loewenstein has also applied his insights more specifically to the phenomenon
of addictive behavior. See George Loewenstein, A Visceral Account of Addiction, in Get-
ting Hooked 235 (Jon Elster & Ole Jorgen Skog eds., 1999).
230 Loewenstein, supra note 205, at 272-73.
231 See id. at 285-86.
232 Id. at 275.
233 See id. at 274-75.
234 See supra notes 211-18 and accompanying text.
235 Loewenstein, supra note 205, at 279.
236 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

of this account of impulsivity is that it predicts for which goods and in

which situations individuals will be most likely to behave impulsively
(i.e., when visceral factors such as hunger, thirst, physical pain, or
emotions are strong).
Additionally, the visceral factors theory is able to account for
"[o]ne of the most difficult patterns of behavior to subsume under a
conventional rational choice framework[:] ... the phenomenon of in-
trapersonal conflict and self-control. ' '1 37 While other theorists have
attempted to account for such human behavior according to the "mul-
tiple selves" models described above,231 Loewenstein describes the
behavior in terms of visceral factors. The experience of being "out of
control" is attributable to the effect of visceral factors. Because those
factors overwhelm an individual's self-interest, of which that individ-
ual is fully aware, one experiences the event as if another self were in
control of the behavior. For Loewenstein, this represents a more
plausible account of the widely recognized sense of intrapersonal con-
flict that individuals often feel: "The fact that impulsive selves never
promote one-another's behavior is not surprising if these selves are
not, in fact, coherent entities with consciousness and personal motives,
2' 3 9
but instead represent the motivational impact of visceral factors.
Loewenstein's account of visceral factors provides an important
insight heretofore absent from the behavioral account of human be-
havior: "[M]uch behavior is nonvolitional or only partly volitional-
even in situations characterizedby substantialdeliberation. '240 More-
over, individuals may experience irrationality, not just in the sense of
violating axioms of rational decisionmaking, but also in the sense of
engaging in "impulsive and self-destructive behavior and.., actions
that violate generally accepted norms about the relative importance of
'241
different goals."

g. FramingEffects. We have already introduced the concept of

framing effects through Kahneman and Tversky's famous vaccine ex-
ample at the outset of this Section. 242 We revisit it briefly here, how-
ever, to emphasize that framing effects are somewhat different from
the other cognitive anomalies that have been identified by behavioral
researchers. They are perhaps the most obviously exploitable of the
biases, capable, for instance, of causing dramatic preference reversals

237 Id. at 288.

238 See supra notes 216-18 and accompanying text.
239 Loewenstein, supra note 205, at 288-89.
240 Id. at 289 (emphasis added).
241 Id.
242 See supra notes 42-45 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

based on an entirely nonsubstantive shift in terminology. And that is

true inasmuch as they trigger or reflect the operation of other cogni-
tive biases. For instance, Kahneman and Tversky's vaccine example
relies not only on the-operation of framing effects, but also on loss
aversion in order to elicit its contradictory results. Whether some-
thing is coded as a loss, thus raising the possibility of loss aversion,
depends on how it is framed. In this respect, one may usefully con-
ceive of framing effects as a mechanism for eliciting other cognitive
biases-in other words, a mechanism for manipulating individual per-
ceptions and decisions.
Consider, for instance, a recent study of how framing affects the
allocation decisions of university staff employees between two retire-
ment savings funds with different risk-return attributes: bonds (the
safe fund) and stocks (the risky fund).243 Although the employees
were shown actual historical data on the returns of the two funds, the
data were framed in two different ways: One group of employees was
shown the distribution of one-year rates of return, while the second
was shown a simulated distribution of thirty-year rates of return. Em-
ployees shown the thirty-year returns chose to invest almost all of
their savings in stocks, while those shown the one-year returns in-
vested a majority of their funds in bonds.244 As Jolls, Sunstein, and
Thaler explain, this result highlights the fact that "the way firms de-
cide to describe and display information... will have a powerful influ-
ence on the choices" of those who receive the information. 245 In other
words, because firms are in a position to frame the choice, they can
alter the decision.
The cognitive influence of framing effects has been demonstrated
in numerous other studies, including at least one in which both sets of
options were given to the same group of subjects (rather than to sepa-
rate groups of subjects). 246 Not only did the subjects continue to
demonstrate the inconsistency, many persisted in their beliefs even af-
ter the logical incompatibility of their selections was pointed out to
them.247 Throughout these studies, subjects have shown a remarkable
proclivity to exhibit different preferences based solely on the manner
in which the options are presented. 24 As Kahneman and Tversky
243 See Shlomo Benartzi & Richard H. Thaler, Risk Aversion or Myopia? Choices in
Repeated Gambles and Retirement Investments (Nov. 8, 1997) (unpublished manuscript),
cited in JoUs et al., supra note 2, at 1534.
244 See Jolls et al.,supra note 2, at 1534.
245 Id.
246 See Dawes, supra note 31, at 36-37 (discussing study).
247 See id. at 37.
248 Public choice theorists have also studied a phenomenon akin to framing effects in the
"voting paradox," also known as the "Arrow Impossibility Theorem" or the "Cordocet

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

have commented, the significance of framing "is both pervasive and

robust . . . [and] as common among sophisticated respondents as
among naive ones .... In their stubborn appeal, framing effects re-
semble perceptual illusions more than computational errors.."249 We
believe that framing effects provide particularly strong evidence in

paradox." See, e.g., Richard H. Pildes & Elizabeth S. Anderson, Slinging Arrows at De-
mocracy: Social Choice Theory, Value Pluralism, and Democratic Politics, 90 Colum. L.
Rev. 2121, 2129-35 (1990) (describing above theorems). This refers to voting situations in
which "for any option under consideration some majority exists that would prefer one of
the other options." Id. at 2129. Thus, the outcome of a vote can depend on purely "proce-
dural" variables such as the order in which different pairs of options are considered and
narrowed or, in the terminology of the behavioralists, the manner in which the vote is
framed. Public choice theorists have also discovered how elicitation effects such as the
"voting paradox" can render elections manipulable by those in a position to set the agenda
for a vote. Michael Levine and Charles Plott, for instance, conducted a real world "experi-
ment" of such manipulation. See Michael E. Levine & Charles R. Plott, Agenda Influence
and Its Implications, 63 Va. L. Rev. 561, 571-72 (1977). One of the authors was asked by
his flying club to devise a fair voting method to determine what planes the club would
purchase for use by its members. The author, a social choice theorist, happily responded
by constructing a voting agenda that would lead to the outcome he personally favored. See
id.
249 Daniel Kalmeman & Amos Tversky, Choices, Values, and Frames, 39 Am. Psycholo-
gist 341, 343 (1984). As an example of the persistent and nondiscriminatory impact of
framing effects, Tversky and Kahneman give the following example, taken from a study of
medical decisions:
Respondents were given statistical information about the outcomes of two
treatments of lung cancer. The same statistics were presented to some respon-
dents in terms of mortality rates and to others in terms of survival rates. The
respondents then indicated their preferred treatment. The information was
presented as follows.
Problem 1 (Survival frame)
Surgery: Of 100 people having surgery 90 live through the post-operative
period, 68 are alive at the end of the first year and 34 are alive at the end of
five years.
Radiation Therapy: Of 100 people having radiation therapy all live through
the treatment, 77 are alive at the end of one year and 22 are alive at the end
of five years.
Problem 1 (Mortality frame)
Surgery: Of 100 people having surgery 10 die during surgery or the post-
operative period, 32 die by the end of the first year and 66 die by the end of
five years.
Radiation Therapy: Of 100 people having radiation therapy, none die dur-
ing treatment, 23 die by the end of one year and 78 die by the end of five
years.
The inconsequential difference in formulation produced a marked effect. The
overall percentage of respondents who favored radiation therapy rose from
18% in the survival frame (N = 247) to 44% in the mortality frame (N = 336).
The advantage of radiation therapy over surgery evidently looms larger when
stated as a reduction of the risk of immediate death from 10% to 0% rather
than as an increase from 90% to 100% in the rate of survival. The framing

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

support of the basic thesis of this Article-that the numerous biases to

which people are subject can be tapped into by those who set the
frame. This follows from our "tendency to accept problem formula-
tions as they are given ...[to] remain, so to speak, mental prisoners of
the frame provided to us by the experimentalist, or by the 'expert,' or
by a certain situation."' so

C. Critiques
As researchers assembled this impressive collection of observed
nonrational phenomena, critics of the behavioral project began to
emerge. "In particular, the research fields have come under attack by
those who appear to regard rationality as axiomatic.'2 1 We offer a
brief review of those critiques, along with the rejoinders of behavioral
researchers. Our aim is to show that, even in spite of attracting criti-
cism from economists and logicians, behavioral research remains a ro-
bust field with serious implications for legal analysis, particularly for
products liability.
An early and frequent criticism of behavioral research assumed
the basic form, "It takes a theory to beat a theory." - Von Neumann
and Morgenstern had given theorists an immensely flexible concept of
human behavior: The simple principle of expected utility maximiza-
tion could be used to generate prescriptive recommendations and de-
scriptive accounts for all manner of human transactions.2 The
findings of researchers such as Kahneman and Tversky, a2 4 on the
other hand, took the form of a hodgepodge of cognitive anomalies-
effect was not smaller for experienced physicians or for statistically sophisti-
cated business students than for a group of clinic patients.
Daniel Kahneman & Amos Tversky, Rational Choice and the Framing of Decisions, 59 J.
Bus. S251, S254-55 (1986). For a summary of examples of this phenomenon in commerce,
see Garvin, supra note 97, at 157.
250 Piattelli-Palmarini, supra note 3, at 30.
251 Evans, supra note 46, at 6.
252 See, e.g., Robert E. Scott, Error and Rationality in Individual Decisionmaking: An
Essay on the Relationship Between Cognitive Illusions and the Management of Choices,
59 S. Cal. L. Rev. 329,334 (1986) ("[A] legal theorist is struck by the atheoretical quality of
[behavioral research] taken as a whole. No general theories have been advanced linking
the separate processes of searching for information, forming judgments and making
choices."); Alan Schwartz, Proposals for Products Liability Reform: A Theoretical Synthe-
sis, 97 Yale LJ.353, 380 (1988):
If the psychologists had a general theory about how people make decisions,
and the theory generated predictions about what people will do in various cir-
cumstances, their experiments could be regarded as testing these predic-
tions.... Psychologists lack such a theory, however. They have instead a large
set of observations about how experimental subjects behave.
253 See von Neumann & Morgenstern, supra note 28, at 15-31 (describing their theory
for utility).
Z4 See supra Part I.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

predictable, persistent, and pandemic, but lacking any overall struc-

ture or systematization that would yield theoretical tractability.2 55
Admittedly, each individual anomaly could be explained by the re-
searchers through processes such as framing effects or anchoring, but
a meta-explanation tying together the various processes was more elu-
sive. Thus, before von Neumann and Morgenstern's model of the ra-
tional expected utility maximizer could be displaced, behavioral
researchers would have to come up with a unifying theory to explain
the various phenomena they were observing.
Katneman and Tversky responded to this challenge in a paper
entitled Prospect Theory: An Analysis of Decision Under Risk. 256 In
it, they first proposed their descriptive theory of human behavior,
prospect theory, which attempted to unify such phenomena as framing
and anchoring into a coherent account of decisionmaking. Howard
Latin has provided a useful summary of Kalneman and Tversky's
theory:
The central elements in this theory are: (i) people make most
choices on the basis of relative gains or losses from an initial refer-
ence point, not on absolute states of wealth; (ii) people are gener-
ally risk averse for gains and risk seeking for losses, with the latter
effect more pronounced; (iii) people assign nonlinear weights to po-
tential gains and losses (for example, the change from $100 to $200
is usually accorded greater subjective weight than the change from
$1100 to $1200); and (iv) sure outcomes are usually overweighted in
comparison with uncertain outcomes (for example, "an increase
from 0% to 5% appears to have a larger effect than an increase
from 30% to 35%, which also appears smaller than an increase from
257
95% to 100%").
To see the explanatory capabilities of prospect theory, we might
revisit Kahneman and Tversky's experiment involving four differently
framed vaccine programs for a threatened disease. 258 Recall that sub-

255 See Hogarth & Reder, supra note 41, at 20 ("Psychologists have been quick to point
out deficiencies in economic reasoning and have amassed much experimental evidence on
so-called anomalies. However, psychologists have been slow to propose alternative models
that economists might use to overcome the noted descriptive deficiencies.").
256 Daniel Kahneman & Amos Tversky, Prospect Theory: An Analysis of Decision
Under Risk, 47 Econometrica 263 (1979). Of course, a great many other explanatory theo-
ries of behavior have been offered from both inside and outside of the behavioral research.
For instance, in this Article, we also touch upon multiple selves theory and attribution
theory, both of which seek to account for human decisionmaking in the way that
Kahneman and Tversky's prospect theory does. We focus here on prospect theory, how-
ever, because it is most directly responsive to the criticisms raised by defenders of expected
utility theory.
257 Latin, "Good" Warnings, supra note 7, at 1238 (footnotes omitted) (quoting
Kahneman and Tversky, Choices, Values, and Frames, 39 Am. Psychologist 341 (1984)).
258 See supra notes 42-45 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

jects simultaneously preferred a sure 200 lives saved out of 600 to a

chance of saving all 600 lives, and a chance of losing no lives to a sure
400 out of 600 lives lost. Prospect theory helps us unravel the various
cognitive processes at work in these apparently conflicting prefer-
ences. We know that people are generally risk averse for gains and
risk seeking for losses,2 9 "but whether an event 'codes' as a loss or a
gain depends not on simple facts but on a range of contextual factors,
including how the event is framed."260 Thus, describing an alternative
as resulting in 200 lives saved creates a far different impression than
one resulting in 400 lives lost: In the first scenario, the 200 lives code
as a gain whereas in the second scenario, the 400 lives code as a loss.
Because of loss aversion, people are willing to take a risk in the sec-
ond scenario that they are not willing to take in the first. They will
accept a risk of losing 600 lives for a chance of losing none if the sure
option appears as a loss ("If program C is adopted, 400 people will
die."). Conversely, they will avoid the risky alternative when the sure
option appears to be a gain ("If program A is adopted, 200 people will
be saved.").
Prospect theory has won the support of many cognitive psycholo-
gists.261 It has yet, however, to conquer other departments of the
academy in the way that von Neumann and Morgenstern's expected
utility theory did. Some fault may lie with the relative complexity and
inelegance of the descriptive model. Katmeman and Tversky's deci-
sionmaker is subject to a litany of cognitive influences, whereas von
Neumann and Morgenstern's rational maximizer follows only one ba-
sic mantra. Indeed, Kahneman and Tversky's decisionmaker may be
subject to conflicting biases: For instance, the endowment effect may
cause one to want to retain an asset that is dropping in market value
while the maxim of loss aversion counsels abandonment. In short, the
model provided by behavioral research, in addition to being theoreti-
cally rich and descriptively textured, is somewhat unruly.
This difficulty of application, however, does not mean that the
model should be rejected, for a complex model ith realistic predic-
tive capabilities is far preferable to a simplified model that bears little
relationship to actual behavior. Moreover, we have attempted in our
own summary of the behavioral research to organize the findings into

259 This effect, known as loss aversion, has been extensively documented. See Latin,
"Good" Warnings, supra note 7, at 1238 n.197 (citing sources); see also Kahneman et al.,
supra note 191, at 197-98 (discussing loss aversion).
260 Sunstein, supra note 2, at 1180.
261 See, e.g., Dawes, supra note 31, at 44 ("Prospect theory is a successful descriptor...
not just because it incorporates irrationality, but because it predicts the direction of irra-
tionality when it occurs.").

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

a more tractable depiction of the individual decisionmaker without

sacrificing the integrity of our behavioral assumptions. We hope this
depiction represents a first step toward achieving the type of realistic
yet manageable model that many legal scholars have been calling for
since the advent of law and economics.
Another persistent criticism of behavioral research is that it
judges subjects against a norm of logic and probabilistic theory that is
computationally impossible to apply to problems of real world com-
plexity.2 62 These critics seem to be saying that researchers should not
be surprised when subjects fail to apply Bayes's Theorem 263 perfectly
or to compute large sums precisely in their heads: Humans are crea-
tures of finite cognitive capacity and should not be compared to an
unrealistic rational ideal. Instead, given limits on cognitive capacities,
humans should be viewed as conforming to norms of logic and ration-
ality as best they possibly can. This criticism draws from Herbert
Simon's well-known conception of bounded rationality:264 "People act
in such a way as to maximise benefit to themselves, within the con-
265
straints of their cognitive processing capacity."
This criticism, however, fails to give full credit to the behavioral
project. Behavioral researchers have acknowledged this notion of
bounded rationality-at least implicitly-all along.266 By focusing on
problems that are comparatively simple in form yet still induce error,
researchers have uncovered behavior that must either support their
notions of cognitive heuristic processes, or support a notion of

262 See, e.g., Mike Oaksford & Nick Chater, Bounded Rationality in Taking Risks and
Drawing Inferences, 2 Theory and Psychol. 225, 226 (1992) ("Since the mind/brain is a
limited information processor, the processes of risky decisionmaking cannot be based on
optimal, algorithmic procedures."); W. Kip Viscusi & Wesley A. Magat, Information
Processing and Individual Decisions, in Learning About Risk 1, 5 (W. Kip Viscusi & Wes-
ley A. Magat eds. 1987) ("Some studies of individual rationality have used a model of
perfect information as the reference point, rather than an optimal learning model in the
presence of imperfect information.").
26 For an introduction to Bayes's Theorem and its difficulties of application by jurors,
see Laurence H. Tribe, Trial by Mathematics: Precision and Ritual in the Legal Process, 84
Harv. L. Rev. 1329, 1356-59 (1971) (stating that "mathematical proof, far from providing
any clear benefit, may in fact decrease the likelihood of accurate outcomes" in jury trials).
264 See Herbert A. Simon, Models of Man 198 (1975) (explaining that "the capacity of
the human mind for formulating and solving complex problems is very small compared
with the size of the problems whose solution is required for objectively rational behavior in
the real world") (emphasis removed); Herbert A. Simon, Rationality as Process and as
Product of Thought, 68 Am. Econ. Rev. 1, 14 (1978) (same); Herbert A. Simon, supra note
182, at 39 (same).
265 Evans, supra note 46, at 10.
266 They have also done so explicitly. See, e.g., Judgment Under Uncertainty, supra
note 4, at xii (acknowledging debt to Simon for use of notion of bounded rationality);
Evans, supra note 46, at 16-20 (defending bias research against charges that it utilizes un-
realistic norm of complete rationality rather than bounded rationality).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

bounded rationality so "bounded" as no longer to represent rational-

ity at all. Thus, rather than being in tension with Simon's notion, be-
havioral researchers have embraced it. They have accepted that
humans operate within a limited sphere of rational behavior and they
have attempted to flesh out the area beyond that sphere. 267 As
Tversky and Kabneman themselves have noted, "[the behavioral re-
search results] are consistent with the conception of bounded rational-
ity originally presented by Herbert Simon. Indeed, prospect theory is
an attempt to articulate some of the principles of perception and judg-
ment that limit the rationality of choice. ' '268 Thus, behavioral research
should be seen as an important complement to, rather than a contra-
diction of, bounded rationality theory.
Another frequent argument against the behavioral research and
its depictions of nonrational behavior is that if such behavior occurred
in real market settings, it would be driven out through an evolutionary
process 269 While it is possible to see how irrational behavior might be
driven out in, say, a well-functioning stock market,270 in many other
contexts the requirements for such arbitrage will not be met. For in-
stance, for a great variety of personal decisions, one's irrationality wil
not present an opportunity for profit by other actors because entry
will not be feasible. 271 Likewise, for irrational behavior to be driven
from a market in the long run, there must be conditions conducive to
the spreading of rational capabilities. In other words, people must be
able to learn.
As Tversky and Kahneman have argued, "[e]ffective learning
takes place only under certain conditions: [I]t requires accurate and
immediate feedback about the relation between situational conditions

26 See Evans, supra note 46, at 10 ("It is arguable... that human intelligence must rely
on the kind of heuristics that Kahneman and Tversky propose-unreliable and error prone
though they may be-given our cognitive restraints.").
268 Tversky & Kahneman, supra note 6, at 88-89 (citations omitted).
269 See Richard Zeckhauser, Comments: Behavioral Versus Rational Economics: What
You See Is What You Conquer, in Rational Choice, supra note 6, at 251, 260 ("Even if
there are many nonrational participants [in the market], they will in effect stan'e."); see
also infra notes 538-42 and accompanying text.
270 But see Shyam Sunder, Experimental Asset Markets: A Survey, in Handbook of
Experimental Economics, supra note 32, at 445, 445-500 (reviewing evidence which shows
that even in well-functioning asset markets, "[ilnformation dissemination and aggregation
can occur, but does not occur under all conditions [and w]hen it does occur, it is rarely
instantaneous or perfect").
271 See Zeckhauser, supra note 269, at 260 ("[Ihe aggregate outcome of a cluster of
personal... decisions-such as insuring appropriately, planning for retirement, or having
babies in or out of wedlock-will be more heavily influenced by a behavioral component
[than by the possibility of arbitrage].").

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

and the appropriate response. 2 72 These requirements often are not

met because:
(i) outcomes are commonly delayed and not easily attributable to a
particular action; (ii) variability in the environment degrades the re-
liability of the feedback ...; (iii) there is often no information
about what the outcome would have been if another decision had
been taken; and (iv) most important decisions 2 73
are unique and there-
fore provide little opportunity for learning.
These and other necessary conditions for arbitrage to drive out
biases successfully have been studied by Thomas Russell and Richard
Thaler.2 74 By investigating the operation of competitive markets in
which some agents are fully rational and others are "quasi-rational,"
Russell and Thaler identify the conditions needed to guarantee that
such markets will achieve rational equilibria despite the presence of
quasi rational agents. In their analysis, "these conditions are quite re-
strictive and are unlikely to occur in any but the most efficient of fi-
nancial markets. In goods markets, a mistake by one individual will
generally not create an arbitrage or profit opportunity for someone
else." 275 Thus, it is not safe to assume that the nonrational behavior
identified by researchers such as Kahneman and Tversky will simply
be weeded out by an evolutionary process in the market.
Two final criticisms merit discussion. First, the so-called citation
bias refers to the alleged tendency of behavioral researchers to cite
only instances of bias while ignoring studies reporting good reason-
ing.2 76 Proponents of the citations bias contend that if a fuller descrip-
tion of cognitive behavior were provided by researchers, we would
view nonrational behavior as involving only insignificant aberra-
tions.277 Whether or not this contention is accurate, we will argue in
Part III that it is of little or no relevance to our discussion of products
liability law. We will argue that even if cognitive biases are compara-
tively rare and insignificant, they assume a special importance in the
consumer product context because consumers are rendered suscepti-

272 Tversky & Kahneman, supra note 6, at 90.

273 Id.
274 See Thomas Russell & Richard Thaler, The Relevance of Quasi Rationality in Com-
petitive Markets, 75 Am. Econ. Rev. 1071, 1071-82 (1985).
275 Richard H. Thaler, The Psychological and Economic Conference Handbook: Com-
ments on Simon, on Einhorn, and on Tversky & Kahneman, in Rational Choice, supra
note 6, at 95, 97.
276 See Jay J. J. Christensen-Szalanski & Lee Roy Beach, The Citation Bias: Fad and
Fashion in the Judgment and Decision Literature, 39 Am. Psychologist 75, 77 (1984) (not-
ing alleged tendency of behavioral literature to give too much emphasis to findings of
nonrationality).
277 See id. at 83.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

ble to bias manipulation by marketers and manufacturers, however

rare or isolated the consumer bias may be.278
Finally, it has sometimes been argued that cognitive anomalies
are simply a product of experimental design or laboratory conditions
and that the behavioral findings do not support conclusions about cog-
nitive performance in the real world. 2 79 "Psychologists are criticized
for extrapolating far too freely from their findings and creating a fash-
ionable view outside of the psychological world-in business schools,
for example-that people are irrational." 280 Like the citation bias, we
will argue in Part I that this criticism is irrelevant to our discussion.
The consumer product market, after all, in many ways resembles the
laboratory setting. We will argue that marketers and manufacturers
who devise product advertising schemes are engaged in a project very
similar to the scheme that behavioral researchers allegedly have been
undertaking: Both projects are intended to trigger cognitive imperfec-
tions in their subjects.

II
THE ACADEMIC DEBATE OVER THE SIGNIFICANCE
OF BEHAVIORAL RESEARCH FOR PRODUCTS
LIABILITY LAW
In an earlier article, one of us has argued (with Steven Croley)
that the evolution of the common law tort rule for product-caused
accidents from negligence toward enterprise liability was premised on
three primary assumptions: Consumers are imperfectly informed of
product risks, manufacturers exert exploitative market power over
consumers, and manufacturers are best able to insure against the costs
of product-caused accidents. 28 1 Contemporary scholars largely agree
that, if these assumptions are true, enterprise liability necessarily fol-

278 See infra Part m.D.

279 See, e.g., David C. Funder, Errors and Mistakes: Evaluating the Accuracy of Social
Judgments, 101 Psychol. Bulletin 75, 76 (1987) (arguing that studies of error have little
relevance to accuracy of social judgment); Lola L Lopes, The Rhetoric of Irrationality, 1
Theory & Psychol. 65, 67 (1991) (asserting that studies purportedly finding irrationality
have shed their experimental foundations and thus become popular); see also Viscusi &
Magat, supra note 262, at 5 (noting limitation of research to predict behavior in markets
where individuals have interest in outcomes and learn how to utilize available
information).
280 Evans, supra note 46, at 24.
281 See Steven P. Croley & Jon D. Hanson, Rescuing the Revolution: The Revived Case
for Enterprise Liability, 91 Mich. L. Rev. 683,706-12 (1993); see also George L Priest, The
Invention of Enterprise Liability: A Critical History of the Intellectual Foundations of
Modem Tort Law, 14 3. Legal Stud. 461, 517 (1985) (identifying similar list of principles as
"true foundation" for enterprise liability).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

lows as the optimal tort regime.282 The obvious strategy then for any-
one attempting to advocate enterprise liability is to argue the veracity
of those underlying assumptions; conversely, those who champion tort
reform and other liability-constricting measures must concentrate
their energies on debunking these same assumptions.
Legal journals have been filled for three decades with arguments
along these lines, producing a body of theory that is intellectually rich
and historically important.283 The indeterminacy of the subject and
the great variance of opinions regarding the basic assumptions of ab-
solute manufacturer liability have fueled a debate far outlasting its
humble origins in an exploding soda bottle. 284 We have seen a wealth
of deliberation over the extent and accuracy of consumer product risk
information,2 5 the presence and force of exploitative manufacturer
market power, 286 and the wisdom of providing product-caused injury

282 See Richard A. Epstein, The Unintended Revolution in Product Liability Law, 10
Cardozo L. Rev. 2193, 2205 (1989) (noting that enterprise liability "makes sense" if one
believes that consumers lack ability to prevent losses); Priest, supra note 281, at 527 ("The
unavoidable implication of the three presuppositions of manufacturer power, manufacturer
insurance, and internalization is absolute liability."); Alan Schwartz, The Case Against
Strict Liability, 60 Fordham L. Rev. 819, 832 (1992) (noting that "[s]trict liability may be
justified .. if the [risk information] assumption-that consumers know risks of harm-is
false").
283 For a survey of much of the leading scholarship, see Croley & Hanson, supra note
281, at 713-60.
284 See Priest, supra note 281, at 498 (describing Escola v. Coca Cola Bottling Co., 150
P.2d 436 (Cal. 1944), as "the most prominent antecedent of our modern regime"). Escola
involved an allegedly defective soda bottle which exploded in the plaintiff's hand. While
the majority applied traditional negligence and res ipsa loquitur principles to the case,
Justice Traynor in concurrence advocated a rule "that a manufacturer incurs an absolute
liability when an article that he has placed on the market, knowing that it is to be used
without inspection, proves to have a defect that causes injury to human beings." Escola,
150 P.2d at 440 (Traynor, J., concurring).
M See, e.g., William M. Landes & Richard A. Posner, The Economic Structure of Tort
Law 280-81 (1987) (arguing that contracting and information costs render inaccurate con-
sumers' estimations of products' risks); W. Kip Viscusi, Reforming Products Liability 64-66
(1991) (noting that "the chief inadequacy of the market is inadequate risk information"
and arguing that purpose of products liability is to remedy such inadequacies); Croley &
Hanson, supra note 281, at 770-76 (arguing that costs of obtaining consumer product infor-
mation are prohibitively high); Epstein, supra note 282, at 2203-05 (challenging view that
consumers have imperfect product risk information); Latin, "Good" Warnings, supra note
7, at 1240-41 (arguing that consumer cognitive limitations lead to systematic underap-
preciation of product risks); Schwartz, supra note 252, at 374-84 (arguing that available
evidence does not support view that consumers systematically underestimate product
risks).
286 See, e.g., Croley & Hanson, supra note 281, at 779-85 (providing renewed theoretical
arguments in support of exploitation assumption); George L. Priest, A Theory of the Con-
sumer Product Warranty, 90 Yale L.J. 1297, 1320-25 (1981) (arguing that exploitation as-
sumption is belied by empirical evidence that suggests that warranties allocate product
risks efficiently); Alan Schwartz & Louis L. Wilde, Imperfect Information in Markets for
Contract Terms: The Examples of Warranties and Security Interests, 69 Va. L. Rev. 1387,

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHFAVIORALISM SERIOUSLY

insurance through the tort system, especially pain-and-suffering dam-

ages insurance. 287
To most observers, champions of enterprise liability have suffered
smartly at the hands of tort law critics in recent years.288 Despite this
general perception of a "counterrevolution" among torts scholars,
however, at least one of the basic assumptions behind enterprise lia-
bility remains relatively robust: "The principle assumption in the
literature supporting a role for legal liability is that consumers under-
estimate product defect risks and, as a consequence, put insufficient
market pressure on firms to produce safety.' 289 Traditionally, com-
mentators have taken an information-costapproach to the question of
whether consumers are ill-informed of product risks.290 This approach
assumes that consumers are rational and responsive to product risk
information and that they would therefore learn of those risks for
which the marginal benefits exceed the marginal costs of becoming
informed. Under the information-cost approach, attention has fo-
cused primarily on the efficacy of product warnings given that the in-
formation they provide to consumers is costly to acquire 291 One of us
has argued (with Steven Croley) that the combination of high infor-
mation costs and low product accident probabilities make warranty
and warning information prohibitively costly to obtain. z
In recent years, however, scholars have begun to take a second,
infonnation-biasapproach to the debate over the extent and accuracy
of consumer product risk information.293 This approach responds to
what might be considered an obvious implication of the behavioral

1402-20 (1983) (arguing on theoretical grounds that manufacturers would not exercise mar-
ket power over consumers by reducing product or warranty quality).
287 Compare Steven P. Croley & Jon D. Hanson, The Nonpecuniary Costs of Accidents:
Pain-and-Suffering Damages in Tort Law, 108 Harv. L Rev. 1785, 1812-95 (1995) (arguing
that consumers prefer some insurance against nonpecuniary losses), with Schwartz, supra
note 252, at 362-67 (arguing that consumers would prefer not to receive pain-and-suffering
damages), and George L. Priest, The Current Insurance Crisis and Modem Tort La%., 96
Yale LJ. 1521, 1546-47 (1987) (same).
288 See Richard B. Stewart, Crisis in Tort Law? The Institutional Perspective, 54 U. Chi.
L. Rev. 184, 189 (1987) ("ITihe critics of enterprise liability have turned the system's own
premises against it. Tort law's new paradigm of social welfare has been turned into an
indictment of tort law and a justification for abandoning the system.").
2M9 1 A.LI., supra note 23, at 230.
290 See Croley & Hanson, supra note 281, at 770 n.356 (describing approaches of com-
mentators to question of whether consumers are informed of product risks).
291 Costly in the sense that consumers must expend time and effort to read and compre-
hend warnings.
292 See Croley & Hanson, supra note 281, at 770-79.
293 See, e.g., Schwartz, supra note 282, at 829-32 (considering impact of availability and
anchoring biases on consumer judgments); Viscusi, supra note 285, at 64-65 (discussing
consumer tendency to overestimate catastrophic, dramatic, and highly publicized risks
while underestimating mundane and underpublicized ones).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

research for products liability law: The demonstrated inability of indi-

viduals to follow axioms of rationality under certain conditions calls
into question the consumer's ability to comprehend product warnings,
regardless of the costs and benefits of reading those warnings. If con-
sumers as a class are cognitively unable to appreciate product risks
accurately, then manufacturers-with their superior risk informa-
tion-should bear liability for product-caused accidents, including
those which are avoidable by consumers at least cost.294
Contemporary commentators, however, are far from unanimous
on the question of whether consumers actually are unable to appreci-
ate product risks and, even if they are, whether they systematically
underestimate rather than overestimate product risks. Within the
literature, we have seen two general perspectives emerge. The first,
belonging to a group we have termed the "Under-Estimators," argues
from primarily theoretical grounds that consumers might systemati-
cally underestimate product risks. Therefore, increased manufacturer
liability might provide an important incentive for manufacturers to ad-
dress product risks that would otherwise be borne, futilely, by con-
sumers. The second perspective has been adopted by a group of
scholars who we refer to collectively as the "Over-Estimators." They
argue that the evidence suggests consumers systematically overesti-
mate product risks and are therefore overdeterred from product use.
Under this view, increased manufacturer liability would only serve to
exacerbate an already pernicious inefficiency.

A. Under-Estimators

The first group of scholars, the Under-Estimators, argue that cog-

nitive anomalies support the view that consumers systematically un,
derestimate product risks. Their position builds on the argument
developed early in products liability theory by Guido Calabresi:

294 Cf. William M. Landes & Richard A. Posner, A Positive Economic Analysis of Prod-
ucts Liability, 14 J. Legal Stud. 535, 549-50 (1985) ("If information on product injuries is
costly to obtain, we want to place liability (other things the same) on the party who has the
information or can obtain it at lower cost."). Although Landes & Posner were speaking
specifically about information costs that might impede consumers from obtaining perfect
product risk information, their reasoning applies equally to information biases that would
produce the same impediment. Indeed, it is possible to conceive of cognitive biases as
simply another instance of information costs, capable of being analyzed under the tradi-
tional theoretical framework. The difficulty with this conception, however, is that it fails to
account for the fact that cognitive biases are largely incapable of being unlearned and
therefore present a more permanent obstacle to efficiency than mere information costs.
See, e.g., supra notes 5, 110-11, 218 and accompanying text. For this reason, we continue to
speak of information biases as a phenomenon distinct and apart from information costs.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

[E]ven if individuals had adequate data for evaluating the risk [of a
product], they would be psychologically unable to do so. The con-
tention is that people cannot estimate rationally their chances of
suffering death or catastrophic injury. Such things always happen to
"the other guy," and no amount of statistical information can con-
vince an individual that they could happen to him. Whether people
know what is best for themselves in other areas is, therefore, irrele-
vant to the conclusion that they do not know what is best in decid-
ing between insuring themselves and bearing the risk of an
unspread accident cost.295
Scholars espousing this view call for expanded manufacturer lia-
bility because they believe that manufacturers are better than consum-
ers at responding rationally to the safety incentives provided by
296
products liability law.

1. Robert Prentice and Mark Roszkowski

Robert Prentice and Mark Roszkowski, for example, have de-
fended strict products liability on the basis that "consumers are lim-
ited in their ability to evaluate and calculate the risks that they
face.''297 In support of this contention, Prentice and Roszkowski rely
on a list of six factors which they argue in combination lead to the
conclusion that "consumers... understate their needs for injury pro-
tection";2 98 that is, consumers underestimate product risks.
First, "evaluation of risk is heavily biased by culture." 299 The au-
thors argue that consumer understanding of risk is strongly colored by
cultural attitudes and norms that may bear no relationship to objec-
tive fact.
Second, "people tend to exaggerate some dangers and minimize
others, usually by overassessing the risks of low-probability events and
underassessing the risks of high probability events."300 Here, Prentice
and Roszkowski draw from a large body of research that attempts to
quantify consumer risk perceptions, including the work of W. Kip

Guido Calabresi, The Cost of Accidents 56 (1970).

295
See infra notes 311-12 and accompanying text; see also Latin, "Good" Warnings,
296
supra note 7, at 1282-94.
297 Robert A. Prentice & Mark E. Roszkowski, "Tort Reform" and the Liability
"Revolution": Defending Strict Liability in Tort for Defective Products, 27 Gonz. L Rev.
251, 291 (1991-1992). The authors' views on behavioral research and its implications for
products liability law are merely part of a larger project arguing in favor of strict liability
along several grounds.
298 Id. at 297.
299 Id. at 292.
300 Id. (footnote omitted).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [V/ol. 74:630

Viscusi.30 1 Unlike Viscusi, 3° 2 however, the authors seem unwilling to

draw sweeping conclusions about consumer perceptions, venturing
only to say that consumers sometimes underestimate and sometimes
overestimate risks.
Third, Prentice and Roszkowski cite the work of Tversky,
Kahneman, and others for the general conclusion that "'people do not
follow the principles of probability theory in judging the likelihood of
uncertain events,' and are therefore frequently unable to gauge the
actual dangers that risks pose. '303 Several cognitive biases could be
said to impede consumers' ability to appreciate product risks. For ex-
ample, the authors cite the availability heuristic as tending to lead
people to overestimate the frequency of high-proffle events such as
earthquakes or tornadoes while underestimating the frequency of less-
publicized and less-dramatic risks such as asthma, emphysema, or
diabetes. 304
Fourth, Prentice and Roszkowski argue that the optimistic bias
mitigates against consumer appreciation of product risks.305 As noted
above, 30 6 this bias refers to the generic tendency of individuals to view
themselves as less vulnerable to a particular risk than others are.
Fifth, the authors cite the illusion of control.3 07 For Prentice and
Roszkowski, the fact that experimental subjects believed that they
could become better coin toss predictors with practice suggests that
consumers might not fully appreciate the product risks that face
them.308 Consumers might mistakenly overestimate the significance
of their own "control" over products, believing erroneously that their
personal input dampens the operation of chance.
Finally, Prentice and Roszkowski refer to the phenomenon of
cognitive dissonance to argue that consumers might disregard product
risk information because it tends to discredit the consumer's original

301 See infra notes 358-92 and accompanying text (discussing some of Viscusi's work in
this area).
302 See infra notes 368-70 and accompanying text (discussing Viscusi's view of the
evidence).
303 Prentice & Roszkowski, supra note 297, at 293 (footnote omitted) (quoting Daniel
Kahneman & Amos Tversky, Subjective Probability: A Judgment of Representativeness,
in Judgment Under Uncertainty, supra note 4, at 32, 32).
304 See Prentice & Roszkowski, supra note 297, at 294.
305 See id.
306 See supra notes 93-111 and accompanying text.
307 See Prentice & Roszkowski, supra note 297, at 294 ("[P]eople tend to believe that
they exert control over purely chance events, thus increasing their vulnerability, and lead-
ing them to conclude that their chances of avoiding injury in a dangerous situation are
inappropriately higher than the objective probability would warrant.") (internal quotation
marks and footnote omitted).
308 See supra notes 118-19 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

decision to purchase the product.309 That is, because people often ig-
nore information that casts doubt on their decisions,310 consumers
who make a purchase will be reluctant to process safety information
that conflicts with their sense of having selected a beneficial, risk-free
product.
For Prentice and Roszkowski, those factors in combination
strongly suggest that consumers generally underestimate product
risks. The implication of that conclusion is obvious: "'[Manufacturers]
have no incentive to offer compensation or incur prevention costs for
risks that consumers do not perceive." 31 ' Thus, liability should be
"aimed primarily at the deliberate decision-making of manufacturers
and suppliers."3 12
313
2. Howard Latin
Howard Latin has presented a more exhaustive review of the be- 314
havioral research and its implications for products liability law.
Latin focuses on the Comment J presumption from section 402A of
the Restatement (Second) of Torts (1965), 3 15 which courts have inter-
preted as precluding independent scrutiny of product design so long3 as
16
an adequate warning has been supplied by the manufacturer
Latin's selection of this subject is provocative, for the Comment J pre-
sumption rests on background assumptions about consumer behavior
which might be said to reflect von Neumann and Morgenstern's ex-
pected utility maximizer. Consumers are expected to read product
warning information, process the safety data precisely, and make effi-
cient preference selections based on the newfound knowledge. 317 For
309 See Prentice & Roszkowski, supra note 297, at 295-96; see also supra text accompa-
nying notes 112-13 (summarizing literature on cognitive dissonance).
310 See Akerlof & Dickens, supra note 112, at 309.
311 Prentice & Roszkowski, supra note 297, at 297 (quoting John E. Calfee & Clifford
Winston, Economic Aspects of Liability Insurance Rules and Liability Insurance, in Liabil-
ity. Perspectives and Policy 16,20 (Robert E. Litan & Clifford Winston eds., 1988)).
312 Id. at 301.
313 For another review of Latin's work in this area, see Kenneth Ian Weissman, A
"Comment J" Parry to Howard Latin's "Good" Warnings, Bad Products, and Cognitive
Limitations, 70 St. John's L Rev. 629 (1996).
314 See Latin, "Good" Warnings, supra note 7. Like Prentice and Roszkowski, Latin's
article reviews strict products liability on a variety of grounds beyond just behavioral re-
search. See id. at 1197-98.
315 The Comment reads: "Where [adequate] warning is given, the seller may reasonably
assume that it will be read and heeded; and a product bearing such a warning, which is safe
for use if it is followed, is not in defective condition nor is it unreasonably dangerous."
Restatement (Second) of Torts § 402A cmLt. j (1965).
316 See Latin, "Good" Warnings, supra note 7, at 1196 nn.8-9 (discussing cases).
317 "The central characteristics of the archetypal [rational actor] analysis are that it em-
phasizes consumer choice, not consumer safety, and that it assumes adequate disclosure

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

Latin, this type of expectation is indicative of a particular mode of

legal analysis, one which clearly evokes the normative underpinnings
of expected utility theory.
In contrast to that mode of legal analysis, which he terms the
"Rational Risk Calculator (RRC) Model, '318 Latin proposes an alter-
native conception of consumer behavior, the "Mistake and Momen-
tary Inattention (MMI) Model. '319 This MMI Model is enriched by a
variety of nonrational features that conspire to prevent the type of
consumer risk calculation which Comment J and expected utility the-
ory presuppose. Many of these features-illiteracy and functional il-
literacy, 320 predictably inattentive or incompetent groups such as
children or unskilled laborers, 321 reliance on learned intermediaries
such as physicians, 322 and information overload 323-are identified by
Latin as reasons why consumers would not read warnings at all, For
present purposes, we will focus on Latin's discussion of why consum-
ers would either fail to understand or fail to follow warnings, even if
32 4
read.
Latin begins his discussion by addressing some of the arguments
offered by Alan Schwartz in his products liability scholarship. 32- For
326
instance, Schwartz has argued that the representativeness heuristic
leads consumers experiencing a series of safe encounters with prod-
327
ucts to expect an unsafe encounter, just as the gambler's fallacy
leads people to expect a heads coin toss to follow a long series of
tails. 328 To Latin, this argument relies upon a flawed application of
Kahneman and Tversky's representativeness heuristic: Rather than

will enable people to choose whatever risk levels are consistent with their preferences."
Latin, "Good" Warnings, supra note 7, at 1201.
318 Id. at 1199.
319 Id.
320 See id. at 1207-08.
321 See id. at 1208.
322 See id. at 1209-10.
323 See id. at 1211-12.
324 Although we focus here on Latin's treatment of the behavioral research, Latin also
identifies a variety of reasons why consumers would fail to understand or follow warnings
from outside of behavioral research. For instance, Latin argues that drafting product warn-
ings necessarily involves imperfect tradeoffs among detail, clarity, and impact; warnings
often fail to inform consumers about the consequences of misuse; consumers must contend
with competing demands on time and attention; consumers lack perfect memory and do
not reread warnings with each product use; some accidents result from reflexive rather
than contemplated action; and some warnings are disregarded because manufacturers lack
credibility. See id. at 1220-48.
325 See infra notes 393-417 and accompanying text for a discussion of Schwartz's work in
this area.
326 See supra notes 145-48 and accompanying text.
327 See supra text accompanying notes 153-54.
328 See Schwartz & Wilde, supra note 286, at 1438-40.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

relying on (erroneous) pre-conceived notions of how random series of

events should "represent" themselves in small samples, consumers
rely on their experiential "input" to construct the "representative"
pattern.329 Because "[m]ost product uses entail small probabilities of
harm[,] ...[p]eople who generalize from their own experiences may
treat this limited sample as 'representative' of overall product risks
and therefore anticipate continued safety. '330 Latin bolsters this con-
clusion by pointing out that manufacturer advertising is more likely to
portray products as safe and reliable, further 3cementing the con-
31
sumer's conception of the products as innocuous.
Latin next turns to Schwartz's reliance on the availability heuris-
tic 3 32 Schwartz uses this heuristic to support the conclusion that vivid
images of product accidents will be more salient in consumers' minds
than mundane information about reliability; hence the availability
heuristic might lead consumers to systematically overestimate the
dangers posed by products.3 33 To Latin, this represents an overly
broad treatment of the availability heuristic. 3M Admittedly, people
may overestimate the severity of such high-profile threats as nuclear
power plant meltdowns or airplane crashes, but for the vast majority
of product-caused accidents, "safe usage will be far more typical and
hence more available. '335 Even for those product risks which do
achieve salience in the consumer's mind, "[v]ivid risks seldom remain
immutably vivid and available because newly publicized hazards grad-
ually supplant old ones. '336 In short, Latin believes that the vast ma-
jority of product risks will not be readily available to consumer
consciousness and, "[b]ecause people seldom think about unavailable
risks, the effect is precisely the same as if they were unduly optimistic
'337
about those product hazards.
Like Prentice and Roszkowski, Latin cites cognitive dissonance 338
as a force that might cause consumers to downplay information incon-
sistent with their beliefs and actions 3 39 Thus, Latin argues, someone
attracted by the style and power of a Corvette might undervalue the
329 See Latin, "Good" Warnings, supra note 7, at 1232.
330 Id. at 1231.
331 See id. at 1232.
332 See supra notes 137-44 and accompanying text.
333 See Schwartz, supra note 252, at 381; Schwartz & Wilde, supra note 286, at 1437-38.
334 See Latin, "Good" Warnings, supra note 7, at 1233 ("This argument treats product
risk in a highly generic manner;, in reality, people are exposed to countless hazards created
by thousands of discrete products.").
335 Id.
336 Id. at 1234.
337 Id.
338 See supra notes 112-13 and accompanying text.
339 See Latin, "Good" Warnings, supra note 7, at 1234.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

safety implications of driving a fast sports car.340 This problem might

be particularly acute in the frequent case in which product warnings
are only available in package inserts or instructions after the product
has already been purchased. By that point, consumers have already
made the purchase decision and formed a positive belief about the
product; warning information inconsistent with that belief may be dis-
counted. Furthermore, Latin argues, cognitive dissonance might oper-
ate synergistically with consumer overconfidence. Not only might
people tend to overestimate their own ability to avoid harmful risks,
but they also might tend to ignore information contrary to their in-
34 1
flated sense of competence.
Latin also discusses some complex applications of Kahneman and
Tversky's prospect theory342 that indicate limitations on consumers'
ability to process risk information. First, because of the anchoring ef-
fect,343 Latin believes that "information on risks often proves less in-
formative than it should be, and anchoring on varied preconceptions
344
leads people to derive different meanings from a given disclosure."
Not only will people arbitrarily fix on some initial estimate of a prod-
uct risk, but they will also fail to adjust sufficiently their estimations in
the face of new information, such as product warnings. Second, fram-
ing effects 345 can be used by manufacturers to blunt consumer risk
appreciation. 346 Third, Latin notes that "people greatly undervalue a
reduction in the probability of a hazard in comparison to the complete
elimination of that hazard" 347 and that warnings and instructions can
rarely result in a complete elimination of product risks. Thus, Latin
concludes, "users may underestimate the utility of complying with
partial risk-reduction measures recommended in warnings. 3 48 Fi-
nally, Latin observes that the prevailing legal standard for strict prod-
ucts liability requires an analysis of whether a product design or
warning is "unreasonably dangerous"-a standard which leads to ex
ante uncertainty for both manufacturers and consumers. Given that
people tend to be "risk seeking in the domain of losses," 3 49 both ac-

340 See id.

341 See id. at 1235.
342 See supra notes 256-61 and accompanying text.
343 See supra notes 158-63 and accompanying text.
344 Latin, "Good" Warnings, supra note 7, at 1238.
345 See supra notes 242-50 and accompanying text.
346 See Latin, "Good" Warnings, supra note 7, at 1241 (explaining that manufacturers
often choose to present warnings in manner conducive to sales thereby contributing to
consumer underestimation of product risks).
347 Id. at 1239 (quoting Kahneman & Tversky, supra note 42, at 346).
348 Id. at 1239.
349 Id. at 1240.

Imaged with the Permission of N.Y.U. Law Review

June 19991 TAKING BEHAVIORALISM SERIOUSLY

tors may prefer to accept a chance of a large loss (i.e., personal injury
or tort damages) to a sure small loss (i.e., an investment in care). "In
effect, legal uncertainty can lead to inefficient risk-seeking
35 0
behavior."
Following his exposition of prospect theory, Latin discusses one
final cognitive feature which draws from the behavioral research-
overconfidence or optimism .35 ' "People may not respond properly to
many risks designated in warnings because they are unduly optimistic
about their ability to avoid these hazards." 35 2 Many people not only
fail to understand complex risks such as product hazards, but they also
fail to realize that they do not understand and hence fail to approach
product usage with the requisite caution. Even when they are aware
of product risks, however, they will often attribute "accidents and
other undesirable results to bad human judgments rather than varia-
ble environmental factors. ' 35 3 This in turn will cause them to discount
the risk to their own safety, because they judge themselves to be "un-
usually capable and careful." 35 Thus, both because they fail to under-
stand product risks and because they fail to believe them even when
they do understand, consumers might display a systematic overconfi-
355
dence in their own ability to avoid product harms.
Having thus introduced a series of heuristics, biases, and other
anomalies from the behavioral research, Latin attempts to tie the
threads together:
Although I have discussed common heuristics and biases in serial
fashion for expositional clarity, their effects on consumer choices
are interrelated and often cumulative. Some sources of error may
lead to overestimation of product risks in specific product-use con-
texts and other heuristics and biases will have the opposite effect.
Taken together, however, bounded rationality constraints, availabil-
ity, representativeness, cognitive dissonance, anchoring, and risk
seeking in the domain of potential losses all support the conclusion
that consumers tend to underestimate rather than overestimate most
35 6
product risks most of the time.
Thus, Latin concludes his rigorous treatment of behavioral re-
search and products liability law by citing a string of cognitive anoma-
lies which he believes, in the aggregate, lead consumers to
underestimate the prevalence of product risks systematically. By pro-

350 Id.
351 See supra notes 93-111 and accompanying text.
352 Latin, "Good" Warnings, supra note 7, at 1243.
353 Id.
354 Id. at 1244.
355 See id. at 1243.
356 Id. at 1240-41 (emphasis added).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

viding a thorough analysis of cognitive features heretofore largely ab-

sent from the legal literature, however, Latin's article should be
regarded as more than just an argument in favor of increased manu-
facturer liability. His work, along with an under-appreciated early ar-
ticle on similar issues, 357 should be seen as an important step in
bringing the wealth of behavioral research to legal scholarship.

B. Over-Estimators
A second group of scholars draws roughly the opposite conclu-
sion from the behavioral research. Rather than viewing Latin's list of
underestimation biases as predominant, these scholars find several
countering psychological effects that lead them to conclude that con-
sumers overestimate product risks. In their view, consumers are al-
ready inappropriately deterred from product consumption, and
manufacturers have greater than optimal incentives to invest in prod-
uct safety. Expanding manufacturer liability would therefore only
strengthen those undesirable tendencies. In short, the Over-Estima-
tors are opposed to increased manufacturer liability, at least when
premised on assumptions that cognitive heuristics and biases lead con-
sumers to underestimate product risks.

1. W. Kip Viscusi
In a recent article, Viscusi has responded to Latin's arguments
and offered his own thoughts on the behavioral research and its impli-
cations for consumer risk perception. 358 Like Latin, Viscusi's main
concern is with the efficacy of product warnings. Unlike Latin, how-
ever, Viscusi views warnings as an antidote to consumer cognitive lim-
itations. 359 According to Viscusi, the behavioral research should not
cause products liability theorists to abandon the prospect of warnings
having a beneficial impact on consumer risk information; instead, it
should spur theorists toward developing a fuller understanding of
product warnings and their operation within the context of bounded
consumer rationality. Viscusi's desire to preserve a role for product
warnings is premised on what he believes is the superior ability of

357 See Latin, Problem-Solving Behavior and Theories of Tort Liability, supra note 7.
358 See Viscusi, supra note 21. Viscusi's work in this area has been extensive. See Vis-
cusi, supra note 285; W. Kip Viscusi, Fatal Tradeoffs: Public and Private Responsibilities
for Risk (1992). We will focus on Individual Rationality, see supra note 21, because it
represents Viscusi's most recent and thorough treatment of the subject.
359 See Viscusi, supra note 21, at 628 ("The principal theme of this discussion is that
while cognitive limitations may be important and define the context in which warnings are
perceived, they do not radically alter how we should think about liability tests or the role
of warnings.").

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

warnings to allow the possibility for individual choice?'m' Thus, like

fellow Over-Estimator Alan Schwartz, Viscusi's views of the behav-
ioral research are accompanied by an underlying preference for the
individual autonomy associated with the rational actor model over any
361
competing norms.
Viscusi begins by somewhat oddly recharacterizing Latin's argu-
ment: "What I call Howard Latin's Law, which is a variant of Mur-
phy's Law, might aptly be summarized as: 'Everything goes
wrong." 362 Viscusi's point is that Latin's "extreme view" 363 may be
unduly pessimistic concerning individual capacity for rationality. Ad-
mittedly, cognitive limitations play a role in consumer processing of
risk information, but "we must identify which failures are important in
a particular warning context and, if the failures are consequential, de-
3
termine how much they impede the efficacy of the warnings."'6
Viscusi proceeds to agree with Latin that consumers frequently
fail to read product warnings; are often functionally illiterate; are sus-
ceptible to information overload; must often rely on intermediaries
who may be unreliable; lack sufficient expertise to make many prod-
uct decisions; vary across populations in a way that prevents optimal
warning construction; are widely susceptible to framing effects; and
display overconfidence in their own ability to avoid risks.3 65 Despite
agreeing with Latin that consumers are susceptible to such limita-
tions, 366 Viscusi refuses to accept Latin's conclusion. Rather, he con-
cludes that "[i]f a warning fails to convey information in a credible
manner, the resulting shortcoming should be judged a failure of the
warning itself rather than of warning policies more generally."3 67

360 See id. at 629 ("The flexibility of warnings enables those who are unwilling to incur
risks to take appropriate precautions or to avoid the risky activity, and also enables indi-
viduals who are willing to engage in risky behavior to do so.").
361 See infra note 416 and accompanying text (discussing Schwartz's work).
362 Viscusi, supra note 21, at 631 n.8.
363 Id. at 630.
364 Id. at 631.
365 See id. at 632-34.
366 Viscusi also notes that Latin "alludes to other forms of failures in decision making,
such as the representativeness heuristic, the availability heuristic, and cognitive disso-
nance." Id. at 634 n.17. Oddly, however, Viscusi refuses to address these behavioral find-
ings, stating that they are "shortcomings of decisions more generally" and are not specific
to hazard warnings. Id. A similar excuse could be offered for all of Latin's findings. Illit-
eracy, for example, is a shortcoming which impacts decisions in far more numerous and
more important settings than the product warning context. That does not, however, alter
the fact that for the afflicted, illiteracy poses a significant barrier to understanding product
warnings.
367 Id. at 634. In support of this argument against Latin's call for removing the exculpa-
tory effect of product warnings, Viscusi argues by analogy that for many important choices
such as "career, school, religion, or spouse" in which greater social costs may result from

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

On that upbeat note, Viscusi turns to the wide body of empirical

evidence relating to consumer risk perceptions and finds that a rough
pattern of risk perception errors emerges. First, "[s]ome hazards may
be hidden, and they may involve risk consequences that are not well
publicized. ' '368 Those hazards remain completely removed from con-
sumer consciousness. Second, "[1]ow probability risks, such as those
from botulism and tornadoes, tend to be overestimated. ' 369 Finally,
"very large risks, such as our lifetime risk of being killed from heart
disease or cancer, tend to be underestimated. '370 Thus, Viscusi ar-
gues, to the extent that product accidents pose low probability risks
(as opposed to hidden or high probability risks), they may be overesti-
mated by consumers.
In addition to that empirical evidence, Viscusi cites some cogni-
tive anomalies that he believes point toward consumer overestimation
of product risks on a theoretical level. First, alluding to the Ellsberg
Paradox, 371 Viscusi notes the tendency of individuals to react ad-
versely to ambiguity over outcomes. Thus, when consumers have
some awareness of a product risk, but not sufficient awareness of it,
they may react to the ambiguity by overestimating the threat of the
harm. Second, Viscusi argues that "consumers tend to overreact to
increases in risk as opposed to risk decreases. '372 Here, Viscusi relies
on experimental evidence gathered on consumer willingness to pay or
willingness to accept values for a series of increases or reductions in
the risk of injury from household chemicals. The evidence strongly
suggests, in his view, that "[p]erceived risk increases are likely to be

inefficient decisions, "there is little activity on the part of the courts to impose sanctions
that would reduce these failures." Id. at 636. We find this argument odd for several rea-
sons. First, it is unclear why the absence of regulation in completely disanalogous contexts
should weigh in favor of abstention in the consumer product context. It is not immediately
obvious that consumer product decisions should be shrouded with the same degree of pri-
vacy/autonomy values that we afford religious faith or marital decisions. Second, there
may be features of the consumer product context, such as its relative ease of accident
prevention or its relative unimportance to individuals vis-A-vis Viscusi's other examples,
which make it a particularly attractive candidate for legal intervention. Finally, Viscusi's
argument proves too much for it is hard to know how he would justify requiring warnings,
if the analogies were controlling.
368 Id. at 639.
369 Id.
370 Id.
371 The Ellsberg Paradox refers to a model that reveals individuals' tendency to be
averse to ambiguity as between probabilistically identical outcomes. In it, contestants are
given the choice between selecting from an urn containing a 50/50 mix of red and white
balls and an urn in which the precise mix of red and white balls is randomly determined.
Most individuals prefer the certain 50/50 probability of the first urn, despite the fact that
the second urn presents the same probability of success. See id. at 640-41.
372 Id. at 642.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

viewed with alarm, resulting in a significant danger of overreaction to

warnings with respect to new risks that consumers did not view as an
' 373
accustomed characteristic of the product.
Finally, Viscusi cites studies indicating that smokers overestimate
the health risks of their smoking for the proposition that "highly pub-
licized risks tend to be overestimated. ' 374 Here, Viscusi is arguing
that media coverage or other imagery of product accidents can result
in overestimation of the harm through the availability heuristic.375 In-
deed, Viscusi believes that "[h]azard warnings [themselves] ...may
create excessive perception of small risks. ' 376 For instance, Viscusi
argues that the wording of California's Proposition 65 3,n would have
resulted in consumers receiving "a warning for Liquid Paper that indi-
cated a risk comparable to that of cigarettes." 378 Thus, in Viscusi's
view, ambiguity aversion, overreaction to risk increases, and the avail-
ability heuristic might all lead consumers to overestimate product
risks.
Beyond those theoretical arguments, Viscusi also assembles em-
pirical evidence relating directly to the effectiveness of warning labels.
For instance, he notes that U.S. per capita cigarette consumption has
fallen considerably in the wake of post-World War II warning ef-
forts3 79 Likewise, product warnings and prominent media attention
of the hazards of saccharin caused a marked decrease in sales of
saccharin-containing products. 3 0 Finally, Viscusi describes one exper-
imental study which found that a high proportion of consumers (up to
thirty-three percent in the study) could be "informed" through prod-
uct warnings about the risks of using household bleach.381 Thus, while

373 Id. at 645-46.

374 Id. at 646.
375 For a more complete summary of Viscusi's opinions regarding consumer knowledge
of the risks of smoking, see Hanson & Kysar, TBS 11, supra note 25, at 1503-05.
376 Viscusi, supra note 21, at 648.
377 Proposition 65 would have required a wide variety of products to carry this warning:
"WARNING: The State of California has determined that this product is dangerous to
your health." Id. at 648.
378 Id. at 650. In support of this claim, Viscusi offers an experimental survey in which
respondents were asked to make comparisons between the perceived severity of differently
worded warnings, some of which were modeled after federally mandated cigarette warn.
ings. See id. at 648-50. We note, however, that survey respondents were not told which
products the warnings were from, but instead were asked to judge the severity of the varn-
ings in isolation. See id. at 649. Consumers might display different responses when actu-
ally viewing the warnings in a product-specific context. That is, other product information
besides that contained on the warning may influence consumer perception of the warning
itself.
379 See id. at 653-54.
3S0 See id. at 655.
381 Id. at 658-59.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

only sixteen percent of respondents were aware that bleach should not
be mixed with toilet bowl cleaner when no warning was given, twenty-
three percent were aware of the precaution after receiving the warn-
ing provided on the national bleach brand, Clorox. 38 2 We note, how-
ever, that while Viscusi believes that this "evidence suggests that these
types of warnings are potentially effective in alerting consumers to the
presence of [a] risk and the need to take precautions, ' 383 he does not
acknowledge that even after receiving the Clorox warning, less than
one-quarter of respondents actually processed and understood the
information.
Perhaps in recognition of this overall failure of the Clorox warn-
ing (despite its small incremental success), Viscusi devotes a substan-
tial section of his paper to describing the research which has been
undertaken to develop a theory of how warnings should be con-
structed in order to be most effective. More particularly, he argues
that "[t]he salience of the information and the manner in which it is
presented do influence, at least to a degree, the ability of consumers
to process the information. ' '384 For instance, label clutter refers to the
tendency to include more information on a warning label than con-
sumers can process effectively. Similarly, label proliferation-the
growing appearance of warning labels on a wide variety of products-
can result in decreased consumer risk appreciation by "distort[ing]
relative product comparisons" or "dilut[ing] the warnings for the real
hazards that should be identified. ' ' 385 In light of these potential fail-
ings of product warnings, Viscusi proposes a uniform format and
structure for all warnings, as well as a national hazard warning
vocabulary. 386
Such a plan would help alleviate the difficulties of comparing risk
information across products-it would, for example, distinguish be-
tween the hazards of cigarette smoking and Liquid Paper use. Not
only would Viscusi adopt such a uniform system, he would make com-
pliance exculpatory. At present, Viscusi believes manufacturers have
an incentive to overwarn because they are held liable only for un-
derwarning about product risks, not overwarning. Thus, consumers
face a state of label clutter and label proliferation prompted by the
legal system. 387 In contrast, "[tihe establishment of a uniform vocabu-
lary coupled with a regulatory compliance defense for warnings that

382 See id. at 659 tbl.8.

383 Id. at 658.
384 Id. at 661.
385 Id. at 665-66.
386 See id. at 666.
387 See id. at 665.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

adhere to this vocabulary would ensure that firms could warn consist-
ently across product lines and that the warning language that was se-
lected would have a consistent meaning to the consumers who process
3as
the information."
Viscusi proposes this national warning system in part because he
fears an autonomy loss that he believes results from Latin's alterna-
tive of putting product risk responsibility on manufacturerss 9 At
heart, Viscusi's quarrel with Latin appears to be over the intuitive re-
sponse to one question: What proportion of product-caused accidents
are caused by consumer cognitive limitations, and what proportion by
deliberate cost-benefit calculations? For Viscusi, the answer is clear:
"[W]arnings should remain the desired policy approach even though
direct control of the product risk characteristics might protect the
39 0
small segment of consumers who do not make sound decisions.1
Thus, although consumers do suffer from cognitive limitations, "un-
derestimation [of product risks] is not the norm."391 Instead, Viscusi
believes that "for very small risks, there is a consistent pattern of over-
estimation of the hazards," 392 such that, if anything, consumers are
underutfiizing products. The proper legal response is therefore not to
increase manufacturer liability for product design but to institute a
national system of warnings to minimize the impact of consumer cog-
nitive failings. Rather than flee the sinking ship of the rational actor
model, Viscusi would try to shore up the leaks.

388 Id. at 666.

389 See id. at 668 ("[O]ne must assess whether the benefits from avoiding mistaken deci-
sions achieved through mandating safety characteristics are sufficient to offset the welfare
loss incurred both through the additional cost of providing the safety as well as by the
elimination of choices in the marketplace."). We note that Viscusi appears to be arguing
against both strict products liability and an alternative of providing substantive product
standards and safety requirements through government regulation. See id. at 667
("How... should we assess the desirability of warnings as compared with other more
direct interventions, such as either government regulation or tort liability?"). However, his
fear of welfare loss associated with "mandating safety characteristics" is overstated with
respect to the possible effect of tort liability. Under strict products liability, manufacturers
could continue to offer Viscusi's "choice[ I in the marketplace" so long as the benefits
associated with such choice outweighed the concomitant accident costs. It is not necessar-
ily the case that expanded tort liability would entail "mandat[ory] safety characteristics" or
"the elimination of choices in the marketplace"; indeed, arguably such results would occur
only when they were socially optimal. Manufacturers would compare the benefits of ad-
ding safety features or pulling product options (lowered liability costs) to the costs of those
safer features or product options (foregone profits), and decide accordingly. In short, lia-
bility for losses is not tantamount to a mandate that such losses be eliminated.
390 Id. at 668 (emphasis added).
391 Id. at 671.
392 Id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74,630

2. Alan Schwartz
Alan Schwartz has addressed the issue of cognitive biases and
consumer risk perception in two prominent articles. 393 For Schwartz,
the stakes are clear: "Strict liability may be justified .. if the [risk
information] assumption-that consumers know risks of harm-is
false. '394 That is, when consumers underestimate the risks of prod-
ucts, "[s]trict liability is necessary to get the right contracts and the
right safety level, as measured by the preferences of well-informed
consumers. '395 Because he is arguing "The Case Against Strict Liabil-
ity," it is crucial that Schwartz show why the behavioral research does
not establish a convincing case that consumers are optimistic with re-
396
gard to product risks.
Schwartz begins by establishing an analytical framework that dis-
tinguishes between consumer perception of product risks and con-
sumer perception of safety improvements. For both of these
perceptions, consumers either can overestimate, underestimate, or
correctly estimate the risk of the product or the effect of a safety im-
provement. This produces the somewhat counterintuitive result that
when a consumer overestimates the effect of a safety improvement
(for example, she thinks that the innovation will lower risk levels
more than it actually does), the consumer is concomitantly underesti-
mating the actual risk of the product. Schwartz views these two fac-
tors as operating independently, such that one cognitive bias could
cause a consumer to overestimate the risk of a product ex ante, while
another bias could cause the same consumer to overestimate the ben-
efits of a subsequent safety improvement to the product. Needless to
say, it is important (and difficult) to assess which effect will be more
pronounced on consumers' overall perceptions.3 97

393 See Schwartz, supra note 282; Schwartz, supra note 252. Although his works address
a full spectrum of issues from products liability theory, we will focus on his treatment of
the behavioral research. More specifically, we will focus on his defense of the consumer
sovereignty norm against behavioral charges of irrationality and flawed risk perception.
394 Schwartz, supra note 252, at 374.
395 Schwartz, supra note 282, at 823.
396 We note from the outset, however, that Schwartz does not view the behavioral re-
search as clearly decisive in either direction. See infra note 415 and accompanying text.
We have placed him with the Over-Estimators more because of his views on strict liability
than his specific views on the behavioral research.
397 Schwartz provides an instructive example: Assume that consumers overestimate the
risk of medical drugs but also underestimate the benefits of drug safety improvements.
Because safety improvements are rare in the medical drug industry, they would have a less
pronounced effect on perceptions than the consumer's general view of the product's risk
level. "Thus, whether there would be too few or too many accidents without strict liabil-
ity... is context dependent." Schwartz, supra note 282, at 833.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

Using these categories, Schwartz divides the world of consumer

risk perception into nine possible states 3 98

Risk Level
0,0 O,P O,C
Safety
P,O P,P P,C
Improvement
C,O C,P C,C

In this model, "0" represents consumer optimism, "P" represents

consumer pessimism, and "C" represents a correct assessment of
product risks.399 Consumer perception of safety improvements are re-
ported first, followed by consumer perception of risk levels. Thus, the
O,P cell would represent a state in which consumers overestimate the
beneficial effect of safety improvements but also overestimate the cur-
rent risk of a product.
With that framework, Schwartz attempts to discern from the be-
havioral research which of the nine cells is a viable description of ac-
tual consumer perceptions. He begins by observing, "The weight of
evidence shows that consumers do not underestimate risk levels.' 400
Here, Schwartz explicitly relies on Viscusi's summary of consumer
perception evidence. 4 1 He adds, however, that from a theoretical
perspective, consumers would also be likely to overestimate product
accident rates given the vividness of accidents and the relative stale-
ness of a statistically good safety record. Here, Schwartz is presuma-
bly arguing that the availability of product accident imagery,
combined with the more general failure of people to appreciate prob-
abilistic data, would cause consumers to overestimate the frequency of
accidents. 4o2 As Latin argued, however, Schwartz does not explain
why product accident rates would not be dulled by the same failure to
appreciate probabilistic data, or why product safety experiences would
not be overestimated via the availability heuristic given their relatively
greater familiarity. 4°3 Nevertheless, Schwartz believes that the combi-
nation of Viscusi's empirical data and the theoretical possibility of the

398 See id. at 829.

399 See id.
400 Id.
401 See supra notes 368-70 and accompanying text.
4M See Schwartz, supra note 282, at 829; see also Schwartz, supra note 252, at 381 ("Be-
cause negative information about products is often alarming, it has high salience. This
psychological claim implies, and supporting evidence predicts, that consumers attach dis-
proportionate weight to negative data, and thus overreact to product-related risks.").
403 See supra notes 137-44 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

availability bias increasing product accident salience makes consumer

4°4
underestimation of product risks unlikely.
Schwartz goes on to note that another cognitive feature, the
anchoring effect, might result in insufficient adjustments of consumer
risk perceptions in the face of product safety improvements.405 Coun-
tering this effect, however, is the apparent possibility that consumers
might pay more for safety improvements than they are actually worth,
thus implying that consumers overestimated the risk of a product ex
ante.40 6 Although Schwartz does not cite any empirical support for
this possibility, he does note a study which found that an automobile's
high crash-worthiness rating strongly influenced consumer purchase
decisions. 407 We are asked then to assume that, given consumers' ten-
dency to use safety as a purchasing criterion, consumers might pay
more for safety improvements than is actually warranted by the im-
provement, which might suggest that consumers initially overesti-
mated the dangerousness of the product. By leading the reader
through that inferential odyssey, Schwartz erects what we consider to
be a straw man in favor of strict liability: "It is difficult to know
whether anchoring... is more significant than the effect of pessimism
respecting risk levels. To make the best case for strict liability, assume
that the anchoring effect is at least as important as the overpayment
effect." 408
By hypothesizing the overpayment effect as a point in favor of
strict liability and then trading it for the anchoring effect, Schwartz
appears analytically generous to strict liability. However, as Latin has
argued, the anchoring effect may actually point to consumer underesti-
mation of product risks rather than overestimation as Schwartz has

404 See Schwartz, supra note 282, at 830 ("Because both evidence and theory imply that
consumers are not optimists respecting risk levels, the first column of the matrix should be
eliminated (all cells that have 0 as the second letter).").
405 See id. ("[I]f consumers think that a product is 'just this dangerous,' they may under-
value evidence that the improved version is materially safer.").
406 Schwartz's phrasing is less explicit: "On the other hand, consumers may be pessimis-
tic about risk levels, which implies that consumers will overpay for safety improvements."
Id.
407 See id. at 830 n.22 (citing Patrick S. McCarthy, Consumer Demand for Vehicle
Safety: An Empirical Study, 28 Econ. Inquiry 530, 541 (1990)).
408 Schwartz, supra note 282, at 830. Schwartz's terminology may have gotten the best
of him here. The anchoring effect, as used by Schwartz, is a point against strict liability in
that it suggests consumers underestimate the benefits of a safety improvement and there-
fore overestimate the product's risk. Schwartz's view of the overpayment effect, on the
other hand, is that it provides evidence that consumers overestimate the benefits of safety
improvements and therefore underestimate the product's risk. Thus, to make the best case
for strict liability, it is the overpayment effect which must be at least as important as the
anchoring effect, not vice versa.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

suggested.4 9 Consumers anchor on their original perception of the

product at the time of purchase, which is likely to be one of a safe
product. Whether through marketing efforts of manufacturers or con-
sumer willingness to ignore data that contradict their purchasing im-
pulse, the original purchase decision is unlikely to involve significant
awareness of product risks. Thus, any adjustments from this original
anchor will insufficiently account for information concerning the prod-
uct's levels.
Despite these arguments, Schwartz treats the anchoring effect as
tending to prove consumer overestimation of product risks and, as
shown above, he argues that the anchoring effect and the overpay-
ment effect should be considered a wash. It is difficult to see, then,
how he could conclude that "the view that consumers are optimists 410
respecting the effect of product improvements must be rejected.
Having just argued that the cases for consumer optimism and pessi-
mism respecting safety improvements are equally compelling (or non-
compelling), it is not clear why Schwartz chooses only to disregard
possibilities in which consumers behave optimistically. It would be
more appropriate in our view to retain both sets of possibilities in rec-
ognition that the data is indeterminate. 41 ' Schwartz was too quick to
eliminate consumer optimism respecting safety improvements from
his analysis, both because he understated the case for optimism and
because his decision to exclude optimism as a possibility rested on
unconvincing logic.
Based on that questionable reasoning, Schwartz removes con-
sumer optimism about safety improvements from consideration, leav-
ing him with the following four possible states of consumer risk
perception: (i) PP; (ii) CP; (iii) P,C; (iv) C,C.412 Of those options, 413
Schwartz believes that only the first and fourth cells are stable.
With the second option, if consumers correctly perceive the effect of
safety improvements, eventually their underlying perception of the
product's risk level will become correct as well. With the third option,
if consumers incorrectly perceive the effect of safety improvements,
eventually their initially accurate perception of the product's risk level

409 See supra note 344 and accompanying text.

410 Schwartz, supra note 282, at 830.
411 Conceivably, Schwartz could have argued that both consumer optimism and pessi-
mism respecting safety improvements should be eliminated as possibilities. That position,
however, would depend on the unlikely coincidence that the anchoring effect and the over-
payment effect operated in tandem to produce correct consumer estimates of safety
improvements.
412 See Schwartz, supra note 282, at 831.
413 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

will become corrupted by the incorrect information gained from safety

improvements.
Thus, the second and third options will merge with the fourth and
first, respectively, leaving only those two options as viable possibilities
for the state of consumer risk perception. In Schwartz's view, then,
either consumers accurately perceive both risk levels and safety im-
provements (the C,C cell), or they view risk levels and safety improve-
ments pessimistically (the P,P cell). In the latter case, products will be
less safe than they should be (because manufacturers have blunted
incentives to make safety improvements given consumers' uncharita-
ble perception of them), but there will also be less products purchased
than there should be (because consumers view the product risk level
as worse than it really is). To Schwartz, determining which one of
414
those trends will predominate is unknowable a priori.
Those arguments constitute the bulk of Schwartz's treatment of
the behavioral research, a treatment that he acknowledges is incom-
plete: "This analysis does not provide a conclusive case for rejecting
the assumption that consumers misperceive risks.... [T]here is sub-
stantial uncertainty respecting whether consumers do misperceive
risks and what the effect of likely misperceptions are. '415 Given this
uncertainty, Schwartz devotes the rest of his article to making a "sec-
ond order case" against strict liability. 416 That is, accepting that we do
not know for certain whether consumers misperceive product risks,
Schwartz argues that the best legal regime under such uncertainty is
not strict liability.
Although we share his sense that the behavioral research is
largely indeterminate, we do not feel that this renders a first order
case for strict liability impossible.41 7 In Part III of this Article and in
our companion article, we will take on Schwartz's challenge, attempt-
ing to establish on theoretical grounds a first order case for enterprise
liability as the most appropriate public response to manufacturer ma-
nipulation of consumer risk perceptions.

414 See id. at 832.

415 Id. at 832-33.
416 Similarly, in his earlier article, Schwartz argued that, given the uncertainty of the
psychological evidence, the strict liability debate devolves to a choice among two compet-
ing "tiebreaker" rules: the desire to compensate victims, and the desire to uphold free
contracting and the consumer sovereignty norm. Of these choices, Schwartz presented an
argument in favor of the latter. See Schwartz, supra note 252, at 382-84.
417 "[A] first order case for strict liability would establish the existence and effect of
imperfect information and then show why strict liability is the appropriate public re-
sponse." Schwartz, supra note 282, at 835.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

C. The ALI Reporters' Study on EnterpriseResponsibility for

PersonalInjury: A Comprehensive View?

The articles addressed thus far have shared the same basic meth-
odology: Each author appears to have surveyed the vast array of be-
havioral research in serial fashion, placing study after study in
respective baskets of over- and underestimation depending on the di-
rection each particular cognitive anomaly pulls. Then, after this allo-
cation is concluded, a final tally is taken, and the author concludes
that, in the aggregate, there appear to be more over- than underesti-
mation anomalies, or vice versa. 4 18
We believe there is a very simple explanation for this somewhat
rudimentary approach to the behavioral research: The research itself
is presented in this serial fashion. Despite the best efforts of thinkers
like Kahneman and Tversky,4 19 the behavioral research remains a
somewhat haphazard collection of seemingly unrelated cognitive
quirks. Thus, for anyone attempting to draw legal implications from
the studies, the natural approach is to address each heuristic and bias
individually. Drawing legal conclusions for a topic like consumer risk
perception then becomes primarily a game of "who has the most
anomalies wins." While we understand the necessity for this ap-
proach, we believe that ultimately a more systematic approach will be
required before we place confidence in any legal policy conclusions
that are drawn from the behavioral research. One attempt at such a
treatment, a chapter from the American Lav Institute Reporter's
Study on Enterprise Responsibility for Personal Injury, is described in
this section.

418 Latin, for instance, is explicit in this analysis:"Taken together, however, bounded
rationality constraints, availability, representativeness, cognitive dissonance, anchoring,
and risk seeking in the domain of potential losses all support the conclusion that consumers
tend to underestimate rather tan overestimate most productrisks most of the time." Latin,
"Good" Warnings, supra note 7, at 1240-41. Schwartz also adopts this methodology,
though he draws the opposite conclusion from Latin: After distinguishing between evi-
dence of consumer "'pessimism' [which] creates insufficient incentives for firms to improve
safety [and] 'optimism' [which] creates excessive incentives," Schwartz, supra note 282 at
827, he concludes that "[t]he weight of evidence shows that consumers do not underesti-
mate risk levels." Id. at 829. The remaining authors display similar methodologies. See
Prentice & Roszkowski, supra note 297, at 297 (noting that "the above [six] factors, in
combination, lead consumers to understate their needs for injury protection"); Viscusi,
supra note 21, at 630-50 (contrasting "inventory of consumer failings with respect to varn-
ings" with "the potential rationality of individual decisions" in order to develop "patterns
of risk perception errors").
419 See supra notes 256-61 and accompanying text (describing Kahneman and Tversky's
prospect theory).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

1. The Reporters' Recommendations Based on a Survey of

"Evidence About Risk Perceptions"
The Reporters begin with a much-needed cautionary statement:
"[T]he degree to which there is overestimation or underestimation of
injury risk varies with the character of the risk and the particular mar-
ket context. Empirical evidence indicates that there are tendencies to
both overestimate and underestimate the risk in different circum-
stances .... ",420 This welcome disclaimer is followed by a summary of
what the Reporters appear to regard as the central finding of their
research: "One result that is definitely borne out by our review of this
evidence is that there is no general support for the widely held as-
sumption that individuals always tend to underestimate product risks
when they misperceive the chance of injury. '' 421 Instead, the Report-
ers find that people display a complex, lock-step awareness of risk
levels. Roughly speaking, the Reporters argue that events of very low
probability tend to be completely ignored, events of low probability
are overestimated, and events of higher probability are
underestimated. 422
To compile this schedule, the Reporters begin by noting that "[i]n
some instances individuals simply ignore events of very low
probability. ' '423 Citing "[s]tudies of natural disasters such as earth-
' 424
quakes and floods, the Reporters note that for at least some very
low probability risks, people completely fail to register the probabilis-
tic threat. In such cases, people have no information concerning the
risk and therefore implicitly set the probability at zero for that activity
or event. When people do have some information about a risk, how-
ever, they display a different set of miscalculations: They "tend to
overestimate low-probability risks and to underestimate the more sub-
stantial risks they face. '425 Thus, the Reporters argue that people
think their chances of being struck by lightning or killed by a tornado
are higher than they actually are; conversely, people underestimate
their risks of dying from a heart attack or stroke. After developing
this schedule, the Reporters argue that for cases "in which the risks
posed by a product are actually very small, psychology tells us that

1 A.L.I., supra note 23, at 223.

420
Id. at 223-24.
421
422 See id. at 224-25. A similar delineation is utilized by Viscusi. See supra notes 368-70
and accompanying text.
423 1 A.L.I., supra note 23, at 224.
424 Id. (citing Howard Kunreuther et al., Disaster Insurance Protection: Public Policy
Lessons 19-41 (1978)).
425 1 A.L.I., supra note 23, at 225.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

consumers will systematically overreact to any such hazard of which

42 6
they are aware or which is drawn to their attention."
The crucial question then becomes whether typical product safety
risks actually are "very small" such that they will be overestimated by
consumers. Without much elaboration, the Reporters answer yes. 4 7
The Reporters argue first that most product-caused accidents do not
fall into the "hidden" or "very low probability" category; thus, con-
sumers do not typically completely fail to recognize a product risk.
They maintain that for those rare accidents that do fall into this cate-
gory, they may be "hidden" from manufacturers as well, implying that
holding manufacturers liable would not have any beneficial deter-
rence effect. 42z Even if manufacturers are aware of risks that are "hid-
den" to consumers, the Reporters believe that a warning requirement
is sufficient to bring the risk into consumer consciousness, such that it
is no longer "hidden." A warning transforms the risk from a very low
probability "hidden" risk to a low probability risk which tends, if any-
thing, to be overestimated.
It is this second class of risks, the low probability risks, which the
Reporters believe represents the typical consumer product accident.
Here, because consumers tend to overestimate the severity of these
types of risks, safety incentives generated for firms will be too strong
rather than too weak. Consumers will demand too much safety, buy
too few products, and take too much care during product use.429 This
is a problem which the Reporters believe tort law is incapable of alle-
viating. 430 Thus, imperfect consumer risk information might lead to
inefficiencies in product markets (just as Calabresi predicted) 4 31 but
they will lead to the opposite inefficiencies: They will cause under-
consumption as consumers incorrectly perceive product risks as
greater than they truly are. Finally, the Reporters discount the impor-
tance of the third category of risk-the high probability risk which
tends to be underestimated. Such risks in their view do not typify the
consumer product accident context.
In light of that analysis, the Reporters advise against a move to-
ward greater manufacturer liability. Instead, they believe that "a sen-

426Id.
427See id. at 230 ("The evidence concerning risk perceptions suggests that the most
widespread phenomenon is consumer pessimism.").
428 See id.
429 See id. ("[C]onsumers will have excessive price reactions to [the risk]-that is, the
drop in consumer demand in response to a small risk is greater than is warranted by the
expected accident costs.").
430 See id. at 223 ("Although this outcome is inefficient... it is not the outcome that
strict liability was designed to avoid.").
431 See supra text accompanying note 295.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

sible warning policy" 432 should be enacted to ensure that consumers

have accurate information about product risk levels and safe usage.
"When consumers are so informed, courts can have greater assurance
about playing the more limited role.., of creating default rules and
'433
enforcing the parties' contracts.

2. Criticisms of the Reporters' Methodology and Conclusions

While we view the Reporters' Study as a significant early attempt
to order the behavioral research into a useable form for products lia-
bility theorists, we are nonetheless skeptical of the Reporters' recom-
mendations. In this subsection, we will argue first that the Study does
not provide support for the policy conclusion it draws, even assuming
that it accurately summarizes the behavioral research. Second, we will
attempt to show that the Study's analysis of the behavioral research
ignores relevant cognitive phenomena that may have drastically al-
tered the Reporters' conclusions.
In response to the contention that some product risks will be hid-
den to consumers, the Reporters recommend a warning require-
ment.434 However, if one accepts the Reporters' summary of the
behavioral research, it is unclear why warning requirements would
produce a desirable result. In the best case scenario, consumers would
read the warning and, assuming the risk was of the very low
probability type, develop a new-found irrational fear of the product's
risks. That is, providing a product warning would cause the risk to
become salient in the consumer's mind, transforming it from a very
low probability "hidden" risk to a low probability "overestimated"
risk. All that a warning policy would do is replace one inefficiency
(underestimation) with another (overestimation). Moreover, of the
two inefficiencies, it is only the former that the Reporters believe can
be addressed by tort law. Although they acknowledge that enterprise
liability can respond to situations in which consumers underestimate
product risks, 435 they view consumer overestimation as a market inef-
ficiency that tort law is ill-equipped to address. The question, then, is
why adopt a policy that will replace a reversible inefficiency with an
irreversible one? For those reasons, the Reporters' policy conclusion

432 1 A.L.I., supra note 23, at 232.

433 Id.
434 See id. ("Adequate warnings will inform consumers respecting risk levels and use
technologies.").
435 See id. at 226-27 (arguing that where consumers appreciate risks posed by generic
product category but not specific products within category, "[a] strict liability system re-
sponds directly to this potential market imperfection... [by allowing] better or worse
accident records [to be] directly reflected in the prices [of products]").

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

regarding warning requirements does not follow inevitably from their

summary of the behavioral research.
Our remaining criticisms of the Reporters' Study center mainly
on the manner in which they gathered and interpreted the behavioral
research. To be fair, we wish to emphasize that some of the weak-
nesses of the study likely reflect, at least in part, the fact that the be-
havioral research reflects a still loosely formed theoretical base.
Nevertheless, we believe that the Reporters' ordering of the behav-
ioral research is significantly less precise than they suggest. Their
summary, again, is that (a) consumers ignore risk probabilities that
are too low, (b) they overestimate risks with probabilities that are low
but not too low, and (c) they underestimate risks with probabilities
that are higher than low. A problem with the framework is that it
does not tell us which risks are too low, which are merely low, and
which are higher than low. There is no numerical rubric to accom-
pany these categories, and the exemplars provided seem unhelpful.
For instance, the Reporters cite earthquakes as a very low probability
risk, while tornadoes and lightning are only low probability risks. We
find those risks difficult to distinguish, and yet under the Reporters'
typology, they are separated by a crucial theoretical chasm: that be-
tween consumer underestimation and overestimation.
Without a fuller foundation to the Reporters' risk schedule, it is
difficult to see how they reach their conclusion that most product risks
are low probability risks. We find it intuitively plausible that most
consumer products could fall into the Reporters' other two categories;
that is, many seemingly innocuous products such as kitchen appliances
or office supplies could fall into the very low probability, "hidden"
risk category, while other more infamous products such as cigarettes
or all-terrain vehicles could fall into the high probability, underesti-
mated risk category. 43 6 Of course, our intuition could be far from the
mark, but the important point is that the Reporters' intuition could be
equally inexact. We believe it is unwise to base tort law policies on an
analytical rubric that draws crucial distinctions, without elaboration,
between earthquakes and tornadoes-especially when we believe it is
arguable that consumer products are more like very low probability
earthquakes than low probability tornadoes.
Apart from this built-in ambiguity, the Reporters' schedule of
risk perception also suffers from a non-representative sample of the
behavioral research. We would certainly not fault the Reporters for
failing to treat all identified cognitive biases, for given the vast array

436 Indeed, in our companion article, we argue that consumers do appear to underesti-
mate the risks of smoking. See Hanson & Kysar, TBS II, supra note 25, at 1505-27.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

of sometimes ill-defined, sometimes overlapping, and sometimes con-

flicting biases, such a task would be unwieldy if not impossible. Nev-
ertheless, the Reporters' Study appears to have omitted many of the
behavioral phenomena that would tend to suggest that consumers sys-
tematically underestimate rather than overestimate product risks.
For instance, the Reporters' Study fails to consider the effect that
cognitive dissonance might have on consumer risk perception. As
Latin, Prentice, and Rozskowski pointed out, cognitive dissonance
might cause consumers to fail to heed important product safety infor-
mation.437 That is, consumers might ignore or downplay their own
information that they face a risk in order to avoid the dissonance that
results from such contemplation. Similarly, the Reporters' Study ig-
nores the optimistic bias, which leads individuals to regard others as
more susceptible to exogenous influences than they. 438 Because of
this bias, even if people overestimate generalized risks, they might still
underestimate their own personal incidence of that risk.
In light of the evidence suggesting that the optimistic bias is a
robust phenomenon, 439 it would be a serious error for the Reporters'
Study to rely on generalized rather than personal perceptions of risk.
The truly misleading aspect of both the optimistic bias and cognitive
dissonance is that, in one sense, they appear to validate the Reporters'
schedule of risk perceptions. That is, people might report that they
perceive a certain risk to be a certain level even if their behavior does
not correspond to that report. It is important to remember, then, that
consumers may be placed on a separate schedule of actual behavior
toward risk in contrast to the Reporters' schedule of reported risk per-
ceptions. In this alternative schedule, consumers might be said to be-
have in a manner demonstrating an underestimation of product risks.
That is, consumers might behave optimistically even if they think or
respond to surveys pessimistically. The significant point, of course, is
that it is consumers' actual behavior that determines the efficiency of
product markets, not their generalized perceptions of risks.
In addition to these biases, the Reporters also ignored many of
the phenomena identified by Latin, Prentice, and Roszkowski as pos-
sible sources of consumer underestimation of product risks. For in-
stance, because of the availability heuristic, 440 people tend to ignore
statistical data in favor of evidence that seems relevant to them or is
easily recalled when making decisions about uncertain future events.
As the overwhelming majority of consumer experiences with products
437 See supra notes 309-10, 339-41 and accompanying text.
438 See supra notes 93-111 and accompanying text.
439 See supra notes 110-11 and accompanying text.
440 See supra notes 137-44 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

are safe, consumers might base their estimates of products' risks on

that easily recallable perception of products as safe. Similarly, the il-
lusion of control 441 might cause consumers to underestimate certain
product risks based on the misperception that the product is under
their control. People exhibit a strong tendency to overestimate their
influence over the likelihood of events: Consumers might easily over-
estimate their chances of avoiding a product risk through some combi-
nation of their own skill and luck. Finally, the Reporters ignored
Latin's identification of the many ways in which prospect theory il-
luminates consumer risk perception. For example, because of anchor-
ing,442 people might adjust insufficiently their initial assessments of a
product risk in light of new information such as product warnings.
We are not arguing that any of these biases is individually suffi-
cient to carry the day for the Under-Estimators. We are arguing, how-
ever, that a treatment of the behavioral research that fails to examine
any of the above biases is suspect. Taken together, the various under-
estimation biases indicate that consumer risk perception may be far
more complicated than the Reporters' schedule would suggest.
Rather than a neat tri-level awareness of product risk, consumers
might exhibit widely varying attitudes toward risks depending on con-
text and circumstances. For example, consumers might systematically
underestimate the risk of Product A because they control it, while
overestimating the risk of Product B because it has achieved salience
through high-profile media coverage. Even the risk of Product A
alone may be simultaneously underestimated and overestimated de-
pending on the consumer's frame. In short, we are dealing here with a
highly textured, complex question about the nature of human cogni-
tion. In their effort to order the behavioral research into a form capa-
ble of yielding policy conclusions for tort law, the ALI Reporters may
have unwittingly given short shrift to research that contradicted their
schedule of risk perception. That discounted research, if incorporated
into the analysis, might have led to vastly different policy conclusions
than those eventually reached by the Reporters.

mf
TIE PROBLEM OF MANIPULATION

In this Section, we begin discussion of our own view of the behav-

ioral research. We depart from previous commentators in two signifi-
cant respects. First, we accept that the behavioral research is simply
too indeterminate to draw broad conclusions about whether consum-
441 See supra notes 114-21 and accompanying text.
442 See supra notes 158-63 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [V/ol. 74:630

ers underestimate or overestimate product risks. Second, we consider

our policy analysis to be largely unaffected by that indeterminacy, be-
cause the behavioral research supports a far different and far more
important implication for products liability law. Consumers, regard-
less of whether they would otherwise systematically over- or underes-
timate product risks, are susceptible to manipulation by
manufacturers due to their cognitive anomalies. This susceptibility to
manipulation produces an opportunity for exploitation that no profit-
maximizing manufacturer can ignore. We view this latter point as the
single most significant implication to be drawn from the behavioral
research for scholars of products liability law.

A. The Indeterminacy of the BehavioralResearch

Unlike the Over- and Under-Estimators, we are unwilling to gen-
eralize from the behavioral research on the question of consumer
product risk perceptions. Although we recognize the power of any
individual cognitive bias to influence consumer behavior, we do not
view the behavioral research as capable of generating an overall pre-
diction about the manner in which consumers will perceive product
risks. 4" 3 Indeed, the mere fact that there are Over- and Under-Esti-
mator camps suggests that the behavioral research is incapable of easy
or incontestable application to products liability.
We hold that view for a variety of reasons. First, as noted above,
it seems likely that different cognitive anomalies may simultaneously
bias consumer risk perceptions in different directions. For instance,
consumers might fail to realize the significance of product safety
warnings when presented in statistical fashion but also overreact to
similar information when presented in a vivid television news account.
Second, it appears impossible to tell, ex ante, which cognitive effect
will be most pronounced in any given context. That is, when analyzing
a consumer product setting, it is impossible to know as a theoretical
matter whether representativeness, anchoring, dissonance, overconfi-
dence, or any other cognitive bias will exact the greatest influence on
consumer perceptions. Even if the existence of their effect can be pre-
dicted, the relative impact cannot.
Finally, the behavioral research, however intriguing and persis-
tent its findings, lacks an organizing principle that would render it ca-
443 As George L. Priest has noted, "consumer perceptions are very difficult to identify
or to measure. As a consequence, hypotheses concerning the relationship between percep-
tions and [product information] are highly speculative and essentially nonfalsifiable."
Priest, supra note 286, at 1298; see also Schwartz, supra note 282, at 833 ("In short, there is
substantial uncertainty respecting whether consumers do misperceive risks and what the
effect of likely misperceptions are.").

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

pable of answering the question of whether consumer risk perceptions

are accurate. The behavioral research remains a field of loosely inter-
related cognitive anomalies lacking a unifying principle. To generalize
from the behavioral research about whether consumers systematically
over- or underestimate product risks would therefore require a coher-
ence and tractability which the behavioral research currently lacks.

B. The Irrelevancy of Indeterminacy and the Endogenous Nature

of Perceptions
Despite those sources of indeterminacy, the behavioral research
still presents an invaluable lesson for scholars of products liability law:
Consumers are subject to a host of cognitive biases which, particularly
when taken together, appear to render them vulnerable to manipula-
tion. Thus, our view of the behavioral research complements our criti-
cism of the methodology employed by earlier torts scholars.44 Earlier
scholars reviewed the behavioral research to arrive at a generalization
about consumer risk perceptions and then simply inserted that gener-
alization into the familiar market model. In essence, they treated cog-
nitive biases as exogenous factors that exert a significant but fixed
influence on consumer risk perceptions. Such a strategy is not only
futile, given the indeterminate nature of the behavioral research, but
it also ignores the most significant finding of the behavioral research:
Consumers do exhibit systematic and predictable cognitive biases that
the careful psychological researcher or the clever marketer can in-
duce. Put differently, individual perceptions can be studied, isolated,
and manipulated by those in a position to influence the individual's
perceptual context.
In light of the behavioral research, therefore, consumer percep-
tions should be treated as factors endogenous to the market model,
such that the behavior of others within that model may influence con-
sumers' perceptions. Our point may be illustrated through a simple
example. Suppose, rather than modeling economic behavior, our ex-
ercise is to model the behavior of a cognitive psychologist and a group
of experimental subjects. For instance, suppose we are attempting to
predict whether a particular survey will reveal the influence of fram-
ing effects on the subjects. The approach of earlier scholars simply
has been to theorize, based on no additional information about the
nature of the survey, that the subjects either will or will not exhibit the
influence of framing effects. In large part, the determination has
seemed to come down to the scholar's intuitive sense about whether
cognitive biases are or are not a robust phenomenon. This approach

444 See supra notes 417-19 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

is misguided, for without far more extensive information on the nature

of the survey, one cannot predict the results of the experiment by in-
dependently focusing on the biases of the subjects. Instead, we be-
lieve that one should consider additionally the motivations, biases,
and beliefs of the other actor in the model, the cognitive psychologist,
for it is this researcher who is in control of the substance, form, and
frame of the survey. Our prediction would then be that the subjects
will exhibit the influence of framing effects in the survey. This predic-
tion is not based on a chauvinistic position either for or against cogni-
tive biases or on an ill-informed guess about the nature of the
experimental survey. Rather, it is based on a prediction about the
motivations of the researcher who, being an interested academic, we
surmise has constructed the survey in a manner likely to yield interest-
ing cognitive phenomena.
Of course, this prediction depends on the academic having suffi-
cient knowledge of cognitive biases and training in experimental de-
sign to demonstrate reliably the existence of framing effects through
the survey. One might take issue with that aspect of our prediction, at
least insofar as we are implicitly comparing consumer product manu-
facturers to cognitive psychologists. In the next Section, however, we
will argue that there is no such requirement of knowledge and training
in the market for consumer products. Instead, we will argue that the
competitive forces of the market will lead, in an evolutionary fashion,
to a state in which manufacturers behave as if they had the requisite
knowledge and training to manipulate consumer perceptions, whether
or not they actually do.

C. Manufacturer Manipulation of Consumers

Applying these insights to the consumer product market leads to
the following conclusion: Once it is acknowledged that consumer risk
perceptions may be affected by, for instance, the manner in which in-
formation is framed, then it becomes inevitable that manufacturers
will exploit those framing effects in a way that maximizes manufac-
turer profits. Other things being equal, it is in the manufacturer's in-
terest for consumers to have the lowest estimate of product risks
possible: The lower the consumer's risk estimate, the more consumers
will be willing to pay for the product, leading to greater sales and in-
creased profits for manufacturers. Generating consumer underesti-
mation of product risks in this manner is simply another means of cost
externalization, a practice that manufacturers have every incentive to
pursue. Manipulation goes further than just minimizing perceived
costs, however. Manufacturers can also attempt to shape consumer

Imaged with the Permission of N.Y.U. Law Review

June 19991 TAKING BEHAVIORALISM SERIOUSLY

views of product benefits. That is, manufacturers may also elevate

consumer willingness to pay by manipulating the view that consumers
have of a product's benefits (as opposed to its costs). In either case,
consumer failure to perceive product attributes accurately can lead to
undesirable levels of consumption.
Some scholars have noted that risk assessments are manipulable
in this fashion, but have failed to see that manufacturers can take ad-
vantage of such manipulability. That failing is evident, for example, in
Viscusi's work on risk perceptions. He attempts to discredit studies
that suggest that consumers underestimate risks by arguing that the
designers of the studies manipulated survey respondents through
framing effects: "Some observed biases in past studies may be due to
the manner in which the risk perception question is framed rather
than to any underlying shortcoming in individual behavior."'* 5 Here,
Viscusi recognizes that survey respondents, like consumers, are sub-
ject to manipulation, but he later fails to see how that vulnerability
will play out in the retail product context. Similarly, Viscusi's own
proposal for reforming products liability law-imposing a national,
uniform warning system-at least implicitly acknowledges that con-
sumer risk perceptions can be manipulated or shaped by the manner
in which information is presented. Nevertheless, Viscusi fails to rec-
ognize that similar manipulations can be perpetrated by manufactur-
ers for less benign or paternalistic reasons than Viscusi's proposal
entails.
Finally, Viscusi argues in one of his more famous articles that the
addition of child-safety caps on medications did not lower the inci-
dence of child poisonings in the expected fashion." 6 Instead, through
something Viscusi termed the lulling effect, consumers overestimated
the efficacy of safety caps and responded by lowering their own paren-
tal care levels. Put differently, the fairly insignificant design change of
safety caps had the effect of substantially altering consumer risk esti-
mates. We believe that if government regulators can have an opti-
mism-producing effect that is both unintentional and contrary to their

445 Viscusi, supra note 358, at 102. An example of what Viscusi means by this can be
seen in a study that he performed involving worker willingness to accept values for occupa-
tional exposure to various chemical hazards. See Viscusi, supra note 21, at 651-54. In that
study, workers were shown warnings containing risk information for various hazards such
as TNT or asbestos and were subsequently asked to evaluate how much of a wage increase
they would require to accept occupational exposure to the hazard. See id. Viscusi's meth-
odology inadvertently may have heightened subject awareness of the hazards simply by
virtue of the nature of surveys: While survey respondents are instructed to read and ana-
lyze product warnings in an isolated setting where no other task confronts them, workers
on the job likely confront risk information in a far different and less accessible manner.
446 Viscusi's work in this area is synthesized in Fatal Tradeoffs, supra note 358, at 234-42.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

goal, then certainly manufacturers can generate similar effects when

they act deliberately in pursuit of profit.
Even Prentice, Roszkowski, and Latin, who each recognize that
consumers may be subject to manipulation by manufacturers, 447 do
not fully realize the significance of this vulnerability. They do not ap-
pear to realize that manipulation of consumers by manufacturers is
not simply a possibility in light of the behavioral research but that it is
an inevitable result of the competitive market. Cognitive biases pres-
ent profit-maximizing opportunities that manufacturers must take ad-
vantage of in order to stay apace with competition. Whether by
design or not, the market will evolve to a state in which only firms that
capitalize on consumer cognitive anomalies survive.448 Thus, even if
consumers initially approach a product greatly overestimating its
safety risks, manufacturers will attempt to counteract that mispercep-
tion through calculated marketing techniques or other bias manipula-
tions. Even if the overwhelming majority of cognitive biases points
toward overestimation of product risks, manufacturers will selectively
target only those biases that lead to underestimation of risks. Indeed,
even if the behavioral researchers themselves have failed to discover a
particular bias, the forces of the market will lead manufacturers to
exploit it; that is, manufacturers will behave as if they know the be-
havioral literature and then some. They have all the incentives to dis-
cover the biases even if they do so unwittingly and even if they and we
cannot name them. Again, not only can manufacturers achieve these
effects, but the hidden hand of market forces requires that they do so
in order to remain competitive.
Next we explore a few of the possible ways in which manufactur-
ers might accomplish such manipulation, using our review of the be-
havioral literature from Part I as a frame for the discussion. We do
not offer these descriptions as strong predictions of actual manufac-
turer behavior; rather, we intend to present a few exemplars of the
type of manipulative conduct that one might expect to find in products
markets, given the existence of consumer cognitive limitations. 449 In

447 See Latin, "Good" Warnings, supra note 7, at 1239 ("An important conclusion of
prospect theory is that consumer responses to risk information can be deliberately or unin-
tentionally manipulated by the ways in which manufacturers and other actors with superior
information frame their disclosures."); Prentice & Roszkowski, supra note 297, at 297 (not-
ing that consumers "have great difficulty determining the risk of a product and are exposed
to producer manipulation").
448 Cf. Milton Friedman, The Methodology of Positive Economics, in Essays in Positive
Economics 3 (1953) (arguing that assumptions posited by economic models are often not
accurate but can be helpful for predictive purposes).
449 We hesitate to offer precise predictions of manufacturer behavior for at least two
reasons. First, we believe that it is difficult if not impossible to determine ex ante which

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

essence, we are engaging in a thought-project similar to that of the

Over- and Under-Estimators, only instead of theorizing about how
cognitive biases will influence consumer perceptions, we are theo-
rizing about how manufacturers will behave given that consumers are
subject to those biases.

1. Manipulation of Scientific and ProbabilisticJudgments

As noted in Part I, individuals tend to act as lay scientists, adopt-
ing and testing hypotheses about the world in a manner that often fails
to comport with rationalist ideals. While understandable given indi-
vidual cognitive limitations, this departure from models of scientific
and statistical reasoning nevertheless creates an opportunity for man-
ufacturers to shape, influence, and exploit consumer perceptions.

a. The Formation and Influence of Personal Hypothe-

ses. Strong evidence suggests that individuals generate, revise, and
sustain personal hypotheses in ways designed to provide "support" for
desired conclusions. The cart, essentially, is before the horse, a cogni-
tive condition captured well by Ziva Kunda's term motivated reason-
ing.450 This self-serving reasoning process is accomplished through a
variety of biases such as the belief perseverance bias or the confirma-
tory bias. Each of those cognitive mechanisms, in addition to support-
ing the process of motivated reasoning, opens the door for
manufacturers to influence the perceptions that consumers have of a
product's risks. For instance, given that individuals tend to disregard
evidence that contradicts their initial hypotheses, 45 ' manufacturers
might attempt to create a favorable impression in consumers as soon
as possible so that their initial views create interference with any unfa-
vorable information that subsequently comes to light. Conversely, be-
cause salient public announcements regarding a product's risks may
lead one generation of consumers to view the product with added sus-
picion, manufacturers might seek to attract the next generation that
was not exposed to the salient event. Some scholars have suggested
that "fresh thinkers" (such as a next generation) may more readily
generate alternate hypotheses (such as non-risky views of a product),

cognitive features will dominate in any given situation. See supra notes 443-46 and accom-
panying text. Second, we believe that regardless of how well-developed the behavioral
research becomes, the powerful incentives provided by consumer products markets vil
drive manufacturers to be more sophisticated than researchers when it comes to uncover-
ing cognitive anomalies. See supra notes 447-48 and accompanying text.
450 See supra notes 88-91 and accompanying text.
451 See supra note 61-63 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

inasmuch as they do not suffer the interference of the initial

452
hypothesis.
Once the favorable product impression is established, manufac-
turers might seek to take advantage of the confirmatory bias by pro-
viding consumers with any and all evidence that does not
unambiguously support an opposing, negative view of the product.
Because of the confirmatory bias, any such evidence will tend to be
misread by the consumer as additional support for the initial product
impression. This possibility of confirmation through ambiguous evi-
dence might explain the omnipresent practice of marketing products
with gee-whiz statistical data, authoritative surveys of physicians, and
imposing bar graphs, all of which, upon inspection, turn out to contain
nothing of informative value. After all, researchers have discovered
that the more ambiguous and complex the evidence, the more that
453
evidence seems to be susceptible to the confirmatory bias.
In a similar fashion, the confirmatory bias appears to cause lay
scientists to view disconfirming evidence with an unjustifiably jaun-
diced eye. Any aspect of disconfirming evidence that appears capable
of a less damaging, alternate explanation will be latched onto by indi-
viduals eager to maintain their initial hypotheses. 4 54 Hence, manufac-
turers of risky products might seek to offer "alternate" explanations
for any evidence that suggests that their products are dangerous. Due
to the confirmatory bias, individuals also appear to gain more convic-
tion in their original hypotheses merely by hearing of methodological
or conceptual flaws in evidence that contradicts their hypotheses.4 55
The natural strategy for manufacturers of risky products, therefore, is
to identify for consumers any and all problems in evidence that is
damaging to product impressions.
The benefit to manufacturers of establishing a favorable product
impression in this manner is that the impression becomes to a large
extent entrenched. The practice of hypothesis-based filtering causes
an individual to interpret ambiguous evidence in light of her initial
hypothesis, which in turn reinforces her conviction in the hypothesis
and makes her even more willing to interpret further ambiguous evi-
dence as consistent with it.456 Additionally, the entity effect can cause
an individual to retain belief in a hypothesis even when it has been
thoroughly discredited; that is, an individual's belief in a hypothesis
may be completely independent of the veracity of the data underlying

452 See Rabin, supra note 24, at 26 (citing D.N. Perkins, The Mind's Best Work (1981)).
453 See supra note 60 and accompanying text.
454 See supra note 62 and accompanying text.
455 See supra note 63 and accompanying text.
456 See supra note 76 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

457
it. For those reasons, manufacturers that create a favorable first
impression of a product hold a significant advantage in later debates
over the safety or value of that product.

b. False Self-Confidence. Apart from orchestrating the forma-

tion and perseverance of product impressions in this manner, manu-
facturers can also capitalize on the many ways in which individuals
display an unwarranted sense of self-confidence. For instance, the op-
timistic bias might make individuals unduly responsive to product ap-
peals that emphasize the ability to avoid negative product aspects.
Just as researchers were able to demonstrate that ninety-seven percent
of consumers believe they are average or above average in their abil-
ity to avoid bicycle or power-mower accidents, 458 manufacturers of
other products might be able to convince a similar preponderance of
consumers that negative product experiences will not befall them.
Significantly, many of the factors identified by researchers as
leading to naive optimism-past experiences free from harm, a per-
ception that risks are of low frequency, and a perception that the harm
is preventable by human action 459-wiUi already characterize con-
sumer perceptions of product harms. As products do generally pose
low-risk harms, the vast majority of consumer experiences with prod-
ucts will be innocuous. Manufacturers might employ ubiquitous por-
trayals of carefree product experiences to reinforce this perception.
Research by Neil Weinstein and William Klein46° points to more
proactive ways in which manufacturers may exploit the optimistic
bias. While studying the debiasing effects of various public health in-
formation campaigns, Weinstein and Klein discovered that certain tac-
tics might actually increase rather than decrease the effect of optimism
on individual risk estimates. Three warning strategies in particular
were identifid as having this exacerbating effect: one in which sub-
jects were given a list of worst-case risk factors, one inwhich subjects
were shown an image of a high-risk individual as a comparison, and
one in which subjects were given a list of personal risk-decreasing fac-
tors.461 The first two approaches caused an increase in individual opti-
mism because individuals "exaggerate the health-threatening
behaviors and underestimate the health-enhancing behaviors of their
peers, arriving at comparative-risk judgments that are optimistically

457 See supra notes 77-87 and accompanying text.

458 See supra note 101 and accompanying text.
459 See supra notes 107-09 and accompanying text.
460 See Weinstein & Klein, supra note 110.
461 See id. at 138.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

biased by using an inappropriate standard of comparison. ' 462 Thus,

by giving subjects a "worst-case" list of risk factors or a "high-risk"
comparison target, the researchers provided them with further "sup-
port" for their biased self-conception. Likewise, with the third warn-
ing strategy, researchers gave subjects an opportunity to "pay greater
attention to their own risk-decreasing factors than to their risk-in-
creasing factors or selectively misremember how often they engage in
relevant behaviors. '4 63 What is significant is that while these results
may be achieved unwittingly (and with great dismay) by public health
advocates, they may also be achieved (wittingly or unwittingly but
with far less dismay) by manufacturers. That is, manufacturers of
risky products may provide consumers with lists of risk-decreasing fac-
tors or a particularly high-risk consumer as a comparison example so
that consumers will process the information in a self-serving manner
and thereby exacerbate the optimistic bias.
Cognitive dissonance may provide an opportunity for manufac-
turers to further solidify these effects. The tendency of individuals to
seek cognitive consistency by disregarding or downplaying informa-
tion that contradicts their favorable views of themselves 464 might
make consumers especially receptive to advertising that helps rein-
force that cognitive consistency. Thus, manufacturers of, say, alcohol
and fatty foods might use advertisements featuring healthy, fit, thin
models. Consumers viewing the ads will fix on the image portrayed by
the models to help resolve the dissonance they feel when engaging in
unhealthy behavior.
Finally, the tendency of individuals to treat random events as if
they are controllable by human skill 465 indicates another way in which
manufacturers might manipulate consumer overconfidence. A recent
advertising campaign for a four-wheel drive vehicle illustrates the type
4 66
of manipulation that may be possible due to the illusion of control.
The campaign features a driver calmly and expertly averting a series
of near-disasters involving an overturned semi-trailer, an avalanche,
and a herd of elk, all on the driver's way home from work. The im-
plicit suggestion is that all owners of this particular vehicle display a
similar ability to avoid accidents, regardless of how random and un-
controllable those accidents may appear. Recalling that a key factor
in the operation of the optimistic bias is a perception that a given risk

462 Id. at 133.

463 Id.
464 See supra notes 112-13 and accompanying text.
465 See supra notes 114-21 and accompanying text.
466 Ironically, this campaign may not have been ultimately effective, as the authors can-
not remember the name of the car manufacturer that waged it.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

is preventable by individual action, 467 the benefits to the manufacturer

of fostering this illusion of control are manifest.

c. Bad Statisticians. Many product attributes are probabilistic.

Parts and equipment failure rates, as well as accidental injury rates,
are all probabilistic outcomes. Moreover, many products-such as
smoke alarms, extended service contracts, auto safety features, and
some pharmaceutical drugs-are designed to reduce the impact of
probabilistic losses. This point is not lost on manufacturers:
"[Managers of products with probabilistic attributes must consider
how consumers' valuations are likely to differ from expected valua-
tions." 468 By studying this disparity, manufacturers can better under-
stand how to shape consumers' valuations in an effort to reduce
consumer sensitivity to product risks or to exaggerate consumers' per-
ceptions of risks that the product is said to reduce.
For instance, the availability heuristic469 can be easily tapped into
by simply maximizing the frequency and intensity of advertisements.
A moment's introspection about the relative amounts of air time
spent on jubilant beer advertisements versus alcoholism awareness
campaigns suggests the power of this heuristic. Alternatively, the
human desire to observe patterns in random events-expressed in
both the law of small numbers and the gambler's fallacy470 -might be
utilized by manufacturers to generate false images of what behavior or
characteristics are representative of users of a product. The late actor-
comedian George Bums, with his longevity and omnipresent cigar,
comes to mind as an excellent spokesperson for tobacco products.
"See," the implicit message would go, "smokers do live long, happy
lives." Similarly, the willingness of individuals to ignore prior
probabilities in favor of new, worthless information471 suggests a clear
method for manufacturers to obscure product risk attributes: con-
found the consumer audience with meaningless information. Finally,
Kalmeman and Tversky's phenomenon of anchoring and adjustment
presents another instance for manufacturer influence over consumer
perceptions. Because adjustment insufficiently accounts for new evi-
dence, perceptions tend to be biased toward the anchor, or first assess-
ment of a probabilistic event. 472 Thus, manufacturers may treat

467 See supra note 107 and accompanying text.

468 Gerald E. Smith & Thomas T. Nagle, Frames of Reference and Buyers' Perceptions
of Price and Value, 38 Cal. Mgmt. Rev. 98, 112 (1995).
469 See supra notes 137-44 and accompanying text.
470 See supra notes 149-56 and accompanying text.
471 See supra note 148 and accompanying text.
472 See supra notes 158-63 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

anchoring as a valuable mechanism for fixing consumer risk percep-

tions during the initial product purchase context, when all products
are safe and attractively packaged; and they may rely on adjustment
from that initial anchor to account insufficiently for the enclosed risk
information which is read, if at all, well after the product has been
purchased.
Inasmuch as consumer risk assessments have been the touchstone
for selecting among liability standards and determining the legal effect
of product warnings, 473 these behavioral phenomena potentially have
dramatic implications for products liability law. If humans systemati-
cally falter whenever "beliefs concerning uncertain events are ex-
pressed in numerical form as odds or subjective probabilities," 474 then
any products liability theory that depends on consumers possessing
adequate product risk information will need to be re-examined.
d. Experiential Thinking, Affect, and the Perception of
Risk. The impact of experiential thinking in the consumer context
has been stated nicely by an early proponent of the significance of
affect to decisionmaking:
We sometimes delude ourselves that we proceed in a rational man-
ner and weigh all the pros and cons of the various alternatives. But
this is probably seldom the case. Quite often, 'I decided in favor of
X' is no more than 'I liked X.'... We buy the cars we 'like,' choose
the jobs and houses we find 'attractive,' and then justify these
choices by various reasons .... 475
In other words, our affective responses to products more often
than not determine the purchasing decision, regardless of whether we
experience the decision as having resulted from "reasons."
Manufacturers clearly stand to gain from cultivating positive af-
fect in relation to their products. Perhaps this explains the onnipres-
ent practice of feel-good advertising that carries little if any
information about the product being pitched, but plenty of gushing
views of the happiness, wealth, and beauty that can allegedly be
gained from its consumption. Indeed, the decades-long "Marlboro
Man" campaign of Philip Morris might be considered the ultimate in
such "lifestyle advertising." By continually offering depictions of the
free and natural cowboy smoker, Phillip Morris may have succeeded
473 See, e.g., Viscusi, supra note 285, at 146 ("[I]f hazard warnings are effective in fully
communicating the [product] risk, then any additional examination of whether a product
design is defective is redundant since the market has already performed such a risk-utility
test.").
474 Tversky & Kahneman, supra note 4, at 3.
475 R. B. Zajonc, Feeling and Thinking: Preferences Need No Inferences, 35 Am. Psy-
chologist 151, 155 (1980).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

in instilling in many smokers a positive affective association with the

product. The significance of this accomplishment is that subsequent
negative information about smoking will be viewed by the consumer
through the bias of experiential thinking. As Seymour Epstein
explains:
Cigarette advertising agencies and their clients are willing to bet
millions of dollars in advertising costs that the visual appeal of their
messages to the experiential system will prevail over the verbal
message of the surgeon general that smoking476 can endanger one's
life, an appeal directed at the rational system.
Given the durability of such practices throughout this century, the bet
appears to be a good one.
Practices such as these find significant theoretical support in Al-
hakami and Slovic's finding that risk and benefit are often confounded
in people's minds. Because of this failure to segregate analysis of risk
and benefit, individuals may be susceptible to manipulation. 477 For
instance, it has been shown that providing information designed to
increase the perceived benefits of various technologies can lead to a
decrease in the perceived risks of those technologies. 478 Again, find-
ings such as this appear to explain a great deal of product advertising,
particularly for notoriously harmful products such as cigarettes. By
developing a positive affect around the product, manufacturers can
not only bias the consumer's interpretation of conflicting evidence but
also actually decrease the consumer's existing perception of the prod-
uct's risk.

2. Manipulation of Preferences
A second set of ways in which manufacturers might manipulate
consumers is through an alteration of their preferences. Like the ma-
nipulation of scientific and probabilistic judgments, this type of ma-
nipulation can occur in ways relevant to products liability law.
Elevating the consumer's willingness to pay for the product is simply
the flip-side of reducing the consumer's perception of the product's
risks.

a. The Status-Quo Bias and Endowment Effect. For instance,

the endowment effect presents an opportunity for manufacturers to

476 Epstein, supra note 164, at 712.

4'n See Alhakami & Slovic, supra note 174, at 1096 ("lit might be possible to change
perceptions of risk by changing perceptions of benefit and to change perceptions of benefit
by changing perceptions of risk.").
478 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

increase the perceived utility of their products. 479 By simply getting

the product into the hands of the consumer, its value to that consumer
may be enhanced. Thus, manufacturers might use money-back guar-
antees, test drives, thirty-day no-risk trial periods, free samples, and
other marketing ploys, all of which are designed to create in the con-
sumer a sense of ownership. 480 Because of the endowment effect, the
sense of ownership by itself might lead the consumer to experience an
increased valuation of the product. Significantly, it has been demon-
strated that people are unable to predict the operation of the endow-
ment effect. 48 1 Thus, marketers utilizing an endowment effect
approach have the added advantage that consumers will fail to per-
ceive the risks of taking a product home-they will, in effect, really
perceive it as a "no-risk" offer.

b. Context Effects and the Effect of Irrelevant Op-

tions. Manufacturers may similarly increase the likelihood of
purchase by carefully controlling the context in which the purchase
choice is presented. For instance, the introduction of an irrelevant
third option has been shown to affect the preferences that some indi-
viduals have for original options.482 Manufacturers seeking to capital-
ize on this bias need only add new "irrelevant" options designed to
enhance the attractiveness of the original ones. For instance, it is folk
wisdom in the restaurant industry that every dessert menu should
have at least one excessively indulgent item to make the others appear
comparatively less indulgent. Similarly, manufacturers can make
products appear less expensive by adding a higher-priced option to
the product line. Itamar Simonson and Amos Tversky demonstrated
the wisdom of this marketing technique in an experiment asking sub-
jects to select a microwave oven based on various product features
and prices. 483 Fifty-seven percent of the subjects given a choice only
between a low-priced Emerson and a medium-priced Panasonic pre-
ferred the Emerson model. When a high-priced Panasonic model was
479 See supra notes 192-98 and accompanying text.
480 See Smith & Nagle, supra note 468, at 101:
[I]t is often better to decouple product acquisition and payment by first endow-
ing buyers with the product. If buyers can be persuaded to take the product
home, they will adjust their reference point to include the newly acquired as-
set. They will then be reluctant to return the product when payment is due,
since this will require that they incur a loss.
See also Richard H. Thaler, Toward a Positive Theory of Consumer Choice, 1 J. Econ.
Behav. & Org. 39, 46 (1980) (describing "two week trial period" as marketing strategy that
takes advantage of endowment effect).
481 See supra notes 196-98 and accompanying text.
482 See supra notes 202-04 and accompanying text.
483 See Simonson & Tversky, supra note 200, at 286-87.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

added to the selection, however, sixty percent of the subjects selected

the medium-priced Panasonic, thus dramatically increasing
Panasonic's share of the experimental market.

c. Elastic Justification. Manufacturers of risky products can

capitalize on the individual's penchant for elastic justification by en-
suring that a range of possible accounts of the product's hazards are
available. As Christopher Hsee has shown, when an attribute is ex-
pressed as a range of possible values rather than as a fixed value, indi-
viduals interpret the range in favor of the outcome they are
predisposed to select. Thus, if a consumer were predisposed to
purchase an item that provided short-term pleasure with long-term ill-
effects, the manufacturer could benefit significantly from generating
controversy over the exact nature of the long-term ill-effects. By en-
larging the range of possible outcomes to include more innocuous
ones than those urged by, say, public health or consumer advocates,
manufacturers can provide consumers with a "reason" or "justifica-
tion" for ignoring the warning messages and continuing to consume
the product. In this manner, elastic justification can serve as a power-
ful tool by which manufacturers minimize the damage caused by nega-
tive accounts of product hazards.

d Time-Variant Preferences. We noted earlier the marked ten-

dency for individuals to seek immediate gratification and delayed dis-
satisfaction, despite a recognition that such practices can lead to a net
loss of utility. 4 That individuals often struggle with intertemporal
self-control in this manner may have significant implications for con-
sumer behavior in numerous contexts. Any product which offers
short-term gratification at the cost of long-term adverse consequences
(e.g., alcohol, fatty foods, environmentally hazardous products) can be
marketed in a manner that exploits this well-recognized weakness of
human will. Also, manufacturers of fatty foods or addictive sub-
stances might package their wares in small sizes or amounts to capital-
ize on the tendency of individuals to seek self-control strategies for
resisting these choices. The individual, enticed by the promise of an

484 See supra notes 207-10 and accompanying text.

485 See supra notes 211-18 and accompanying text.
486 For a discussion of some of those implications, see David Laibson, Golden Eggs and
Hyperbolic Discounting, 112 QJ. Econ. 443 (1997); Richard H. Thaler & H.M. Shefrin, An
Economic Theory of Self-Control, 89 J. Pol. Econ. 392, 398-404 (1981); Deborah M. Weiss,
Paternalistic Pension Policy- Psychological Evidence and Economic Theory, 58 U. Chi. L
Rev. 1275 (1991); see also Hanson & Logue, supra note 93, at 1205-09 (describing some
ways in which smokers appear interested in, but unable easily to, precommit to quit their
habits).

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

instantaneous sugar rush, will be able to rationalize the purchase be-

cause the small package ensures only a limited degree of indulgence.
As at least one commentator has noted, "the role of self-control in
'487
purchasing decisions is well known among marketing experts.

e. Reciprocity and Attribution. According to attribution theo-

rists, consumers search for causes following a product failure, and the
cause to which they attribute the failure can have a significant effect
on their reaction to the failure.488 Attribution theorists have identi-
fied several dimensions along which causal inferences are made in the
consumer context.48 9 Valerie Folkes, for instance, has found that
three causal dimensions go a long way toward explaining or predicting
consumer reactions. 490 One such dimension, which she calls stability,
is a measure of whether the inferred cause is relatively permanent or
temporary.491 A second causal dimension, locus, has to do with
whether the cause of the product failure is perceived as located with
the manufacturer or the consumer. 492 The third dimension, which
Folkes refers to as controllability, has to do with the extent to which
493
the cause was the consequence of the manufacturer's volition.
The evidence suggests that consumer responses are significantly
influenced by these causal dimensions of stability, locus, and control-
lability. 494 For example, where consumers perceive a cause to be sta-
ble they are more apt to prefer a refund to an exchange because an
exchange would not redress the underlying problem. Where the man-
ufacturer is the locus of the failure, the consumer is perceived to be
entitled to redress, including a refund and an apology, but where fail-

487 Rabin, supra note 24, at 41.

488 See, e.g., S. Krishnan & Valerie A. Valle, Dissatisfaction Attributions and Consumer
Complaint Behavior, in 6 Advances in Consumer Research 445,448 (William L. Wilkie ed.,
1978); Marsha L. Richins, Negative Word-of-Mouth by Dissatisfied Consumers: A Pilot
Study, 47 J. Marketing 68, 72, 74 (1983) (correlating consumers' likelihood of complaining
with their attribution of product failure to causes external to them).
489 See Bernard Weiner, Human Motivation chs. 7 & 8 (1980) (reviewing relevant
literature).
490 See Valerie S. Folkes, Consumer Reactions to Product Failure: An Attributional
Approach, 10 J. Consumer Res. 398 (1984) (building on theoretical work of Bernard
Weiner).
491 See id. at 399 ("For example, a car might be poorly repaired because the mechanic
was sloppy just this once, or the mechanic might be a consistently sloppy worker.").
492 See id. ("For example, a set of bookshelves might collapse because the consumer
assembled them incorrectly or because the manufacturer made a defective product.").
493 See id. ("For example, when shoes are poorly repaired because the shop does not
train its personnel, the cause is under the shop's control; when the poor repair is due to an
externally set fire in the store, the cause is not under the shop's control.").
494 See id. at 400-07 (using above-described causal dimensions to interpret data from
studies).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

ure is consumer-related, the consumer is not perceived to be entitled

to redress. 495 Moreover, controllability and locus both influence an-
ger reactions. That is, consumers feel angry and vengeful (desiring to
hurt the firm's business) when they perceive that a product failure is
the consequence of the manufacturer's volitional acts 4 96 The evi-
dence also indicates that consumers, in response to that anger, are less
likely to repurchase a product from the same seller 497
This analysis offers many implications for the possibility of manu-
facturer manipulation. For instance, because individuals tend to pre-
fer cooperating with those they view as behaving cooperatively or
fairly, manufacturers will benefit from cultivating an appearance of
cooperation and fairness. 498 Thus, if a product is discovered to pose a
safety threat, manufacturers might choose to project an image of co-
operative concern, irrespective of actual manufacturer behavior.
Manufacturers might also attempt to shift the perceived locus of a
product failure to consumers, much as cigarette manufacturers have
invested a great deal of effort portraying the risks of smoking as a
consequence of consumer decisions.499 More generally, manufactur-
ers might do whatever they can to deny controllability.500 To the ex-
495 See id. at 399-400:
When product failure is firm-related, the inequity of the exchange upsets the
relationship between buyer and seller. the consumer has been wronged. Be-
sides giving a refund, damaged interpersonal relations must be repaired. By
means of an apology, the firm admits to and regrets not providing the prom-
ised product benefit.
496 See id. at 405; see also Valerie S. Folkes, The Availability Heuristic and Perceived
Risk, 15 J. Consumer Res. 13, 20 (1988) ("Consumers are angrier about product failure
when they believe the firm has control over the problem than when constraints force the
firm's actions."). For a slightly updated version of the more complex contributors to and
effects of anger, see Valerie S. Folkes, Susan Koletsky, & John L Graham, A Field Study of
Causal Inferences and Consumer Reaction: The View from the Airport, 13 J. Con-
sumer Res. 534, 534-35 (1987).
497 See, e.g., Folkes et al., supra note 496, at 534-35 (providing evidence of that effect
with respect to delayed airline passengers). Furthermore, according to Jolls, Sunstein, and
Thaler, "[t]his is the same behavior that drives boycotts, where consumers refrain from
buying something they normally enjoy in order to punish an offending party." Jolts et al.,
supra note 2, at 1495.
498 See Folkes, supra note 490, at 407-09 (discussing how manufacturers can alleviate
consumer anger by cooperating with consumer).
499 See Graham E. Kelder, Jr. & Richard A. Daynard, The Role of Litigation in the
Effective Control of the Sale and Use of Tobacco, 8 Stan. L. & Pol'y Rev. 63, 82 (1997)
("The tobacco companies won many of the first and second wave cases by asserting the
defenses of assumption of risk and contributory negligence, or by asserting that the
smoker's willfulness, not the industry's misbehavior, was the proximate cause of the plain-
tiff's smoking and consequent illness.").
500 Indeed, Folkes offers manufacturers just this advice. See Folkes, supra note 496, at
20 ("[Mvlarketers want to know what facilitates retrieval of successful product perform-
ances and hampers failure recall .... This may be accomplished by influencing consumers'
attributions for product failure. Consumers are angrier about product failure when they

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

tent that products do pose risks to consumers, they might be

characterized as natural or inevitable risks that manufacturers are
powerless to avoid. Because individuals tend to seek retribution less
against uncooperative actors whom they view as lacking volition, man-
ufacturers of risky products who characterize the risk as inevitable in
this manner will suffer less consumer ill will. Finally, where causal
attributions are extremely complex, consumers sometimes do not ar-
rive at a causal attribution at all. In those circumstances, consumers
often turn to the manufacturer for assistance in making the causal at-
tribution.50 As a result, manufacturers might attempt to make causal
attributions as complex as possible so that they can more powerfully
influence consumer perceptions.

f. Preference-TrumpingEffect of Visceral Factors. The prefer-

ence-trumping effect of visceral factors has obvious and powerful im-
plications for product manufacturers who seek to manipulate
consumer preferences and perceptions. For instance, Loewenstein
himself notes that cookie shops often vent baking smells into shopping
malls in order to trigger the visceral factor of hunger 5° 2 and that
"[d]istilled spirits ads opt for simple 'bottle, glass, and ice' depictions
to help consumers visualize the experience." 503 These are but two ex-
amples of a broader phenomenon: Because Loewenstein's theory
"predicts... that impulsive behavior will tend to occur when visceral
factors such as hunger, thirst, physical pain, sexual desire or emotions
are intense,"50 4 one should expect marketers to emphasize those same
visceral factors heavily in order to maximize impulsive (consumption)
behavior. Thus, appeals for food products might attempt to empha-
size hunger through vivid, close-up views of freshly baked pizza, or
through free potato chip samples at grocery stores. Makers of pain
relievers might seek to offer compelling metaphors of headache pain:
"It felt like a jackhammer going off inside my head." And it almost
goes without saying that manufacturers of virtually any product may
utilize sexual desire as a visceral factor that increases impulsive con-
sumption. These tactics have, of course, been a feature of consumer
products advertising since its inception. What Loewenstein's account
of visceral factors adds to the topic, however, is an understanding that

believe the firm has control over the problem than when constraints force the firm's
actions.").
501 See Folkes, supra note 490, at 407 (stating that consumers may turn to manufacturers
to facilitate causal attribution with complex products).
502 See Loewenstein & Adler, supra note 196, at 279.
503 Stephen J. Hoch & George Loewenstein, Tine-Inconsistent Preferences and Con-
sumer Self-Control, 17 J. Consumer Res. 492, 497 (1991).
504 Loewenstein & Adler, supra note 196, at 279.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

such appeals can result in behavior that is at odds with one's

preferences.505

g. FramingEffects. Richard Thaler has extensively studied the

way in which such psychological phenomena as framing effects can
illuminate the concept of consumer choice. Drawing from Kahneman
and Tversky's prospect theory, Thaler has developed a concept of
mental accounting, which forms the basis of a behaviorally based
model of consumer choice.506 A representative example of Thaler's
findings includes the rather surprising result that money, contrary to
standard economic assumptions, is not fungible. 5 7 This can be seen in
the fact that most individuals would not drive across town to save
twenty-five dollars on a car, but would to save twenty-five dollars on a
vacuum cleaner. Thus, twenty-five dollars is simultaneously worth
more and less than a trip across town. For our purposes, Thaler's the-
ory of mental accounting provides insight into the way in which manu-
facturers might utilize framing effects and prospect theory to
manipulate consumer perceptions. Because his work involves the ear-
liest and most influential application of behavioral research to the
study of consumers, we will consider it in some detail.
Thaler proposes replacing the utility function from economic the-
ory with the value function developed by Kahneman and Tversky as a
central aspect of their prospect theory.5 08 The value function could be
described as more psychologically textured than the utility function
for at least three reasons. First, it assesses perceived gains and losses
relative to some natural references point-a feature of the model
which takes account of the fact that "people appear to respond more
to perceived changes than to absolute levels." 509 Second, the value
function is assumed to be concave for gains and convex for losses.
This merely acknowledges the fact that the perception of quantities is
a relative process. Twenty-five dollars is a small percentage of the
price of an automobile but a significantly larger percentage of the
price of a vacuum cleaner. Third, the loss function is steeper than the
gain function, a feature which Thaler describes as capturing the en-
dowment effect. Losing something that one owns is perceived as more
pronounced than gaining the identical item, just as people "will de-

505 See id. at 289.

506 See Richard Thaler, Mental Accounting and Consumer Choice, 4 Marketing Sdi. 199
(1985); Thaler, supra note 480.
507 See Richard H. Thaler, Quasi Rational Economics 28 (1991).
508 See id.
5D9 Id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

mand more to sell an item they own than they would be willing to pay
to acquire the same item." 510
Utilizing this enriched model of consumer value perceptions,
Thaler examines the case in which an individual experiences the joint
outcome (x,y). Psychologically, the individual could perceive the re-
sult of such an experience as v(x + y) or as v(x) + v(y). The former
case Thaler refers to as integration; the latter, segregation. Whether
an individual prefers integration or segregation of outcomes x and y
will depend on their magnitude and whether they are gains or losses.
If both are gains, Thaler predicts that segregation is preferred because
the value function is concave for positive values. 511 If both are losses,
Thaler predicts that integration is preferred for the inverse reason.512
If gains and losses are mixed, Thaler predicts that the individual would
prefer integration where gains exceed losses.513 Where losses exceed
gains, Thaler predicts that the individual would prefer segregation of
the gain, especially where it is small relative to the loss (such small
gains Thaler refers to as "silver linings").
14
After verifying these predictions with experimental evidence,
Thaler discusses their implications for the theory of marketing: "The
results of [the joint outcome study] can be summarized by two princi-
ples: segregate gains and integrate losses. Each principle also has a
corollary: segregate 'silver linings'.., and integrate (or cancel) losses
when combined with larger gains." 51 5
Thus, manufacturers desiring to enhance the perceived value of
their products need only follow Thaler's concise instructions. For in-
stance, product features should be described separately so that gains
appear segregated. "The most vivid examples of this are the late-night
television advertisements for kitchen utensils. '516 But wait, there's
more. Product rebates can be used as an effective form of price pro-
motion because they seem to follow the silver lining principle: "A re-
bate strongly suggests segregating the saving. 517 Where losses are
concerned, "sellers have a distinct advantage in selling something if its
cost can be added to another larger purchase."518 Thus, individuals
can frequently be convinced to purchase optional stereo or security

510 Id.
511 See id. at 29 ("Moral: Don't wrap all the Christmas presents in one box.").
512 See id. ("For example, one desirable feature of credit cards is that they pool many
small losses into one larger loss and in so doing reduce the [perceived] total value lost.").
513 This is because the loss function is steeper than the gain function.
514 See Thaler, supra note 507, at 31-32.
515 Id. at 38.
516 Id.
517 Id. at 39.
518 Id.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

systems when buying a new car, despite the less than competitive price
car dealers typically charge for such options. Finally, Thaler notes, the
federal government appears to be following the principle of cancella-
tion (whereby smaller losses are integrated with larger gains) by with-
holding tax payments from wages as opposed to requiring payment in
one lump sum.
Another lesson to be gained from the mental accounting model
of consumer behavior stems from the significance of a reference out-
come to perceived value. Because consumers generally perceive value
in relation to some expected or "just" outcome, manufacturers must
be mindful not only of the value of the goods sold but also of the
manner in which they are sold (i.e., "the perceived merits of the
'deal'").519 Thus, Thaler notes, professional sports leagues, the manu-
facturers of the toy behind the latest Christmas season craze, and the
operators of the most popular restaurant in town all face a similar
dilemma-each could, at times when demand exceeds supply, charge
a higher than "normal" price and still sell out of their respective goods
and services. However, the ongoing nature of the seller-buyer rela-
tionship prevents such opportunism. Because the elevated price
would be judged in relation to the "normal" price, customers would
experience significant disutility from the nature of the transaction and
520
would be less likely to offer the manufacturer repeat business.
Thaler provides guidance for manufacturers who wish to circum-
vent such social norms. First, steps can be taken to raise the perceived
reference price. This can be accomplished by explicitly suggesting a
high reference price (for instance, by offering a "suggested retail
price") or by raising the perceived cost of the product (for instance, by
adding amenities that give the product an appearance of luxury,
whether or not the individual consumer herself actually values the ad-
ded amenities). 521 Second, manufacturers can tie the sale of the prod-
uct to something else, thus integrating the premium price for the
product into a larger purchase. Thaler notes that Super Bowl tickets
frequently are sold in packages that include airfare and hotel, thereby
masking the ticket premium to some degree. 522 Fmally, manufactur-
ers can offer the product in an unusual size or format, thus making it
difficult for consumers to conceive of an appropriate reference price.
As a prime example of this tactic, Thaler reminds the reader that
candy at movie theaters, in addition to being priced well above ordi-

519 Id. at 33.

520 See id. at 40.
521 See id. at 42.
522 See id.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

nary retail levels, is "typically sold only in large containers rarely seen
'5 3
in other circumstances. " 2
Other researchers have begun to follow Thaler's work in this area
by further examining the way in which framing effects can alter cus-
tomer perceptions. Gerald Smith and Thomas Nagle have found that
"[m]anagers can... influence purchase decisions by how they present,
or 'frame,' price and benefits relative to a reference point. 52 4 For
instance, it is preferable to begin with a higher reference point and
then offer a discount in appropriate circumstances, rather than begin
with a lower price and charge a premium. "If the explicit base price is
seen as the starting point, buyers paying a higher price will view their
failure to qualify for a discount as a gain denied. They will find that
less objectionable than being charged a premium, which they would
view as an explicit loss.' ' 5a 5 This disparity in perceived value from for-
mally identical outcomes stems from the fact that individuals perceive
out-of-pocket costs as weighted more heavily than opportunity
costs. 526 For the same reason, banks have been able to convince cus-
tomers to keep a minimum amount in their checking accounts in order
to avoid monthly fees, even where the fees would be less than the
5 27
(opportunity) cost of the foregone interest.
The importance of reference prices to consumer perceptions also
suggests significant implications for managerial price setting. For in-
stance, when introducing a new product such as an electronics good,
manufacturers might consider setting the initial price high in order to
establish an elevated reference price. In personal sales settings, "the
salesperson should begin by first framing customers' perceptions by
showing them products above their price range, even if customers may
not be in the market for such products. '52 8 Finally, because consum-
ers, as noted in Thaler's discussion of the value function, view price in
proportional terms rather than in absolute terms, manufacturers can
carefully calibrate price changes to minimize or maximize impact.
Price increases can be accomplished in small steps over time with little
or no effect on consumer perceptions.52 9 This is because the buyer's
implicit reference price will be elevated subtly throughout the gradual
price elevation.

523 Id.
524 Smith & Nagle, supra note 468, at 100.
525 Id. at 101.
526 See Thaler, supra note 480, at 43-47.
527 See Smith & Nagle, supra note 468, at 102.
528 Id. at 106.
529 See id. at 108.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

3. Summary
Although we have attempted in the foregoing discussion to de-
scribe some of the ways in which manufacturers may manipulate con-
sumers' risk perceptions and preferences, for reasons described
above 530 we are sure that our list is both over- and underinclusive.
Thus, we wish to emphasize that our central prediction regarding
manufacturer behavior remains far more general: Manufacturers Vill
respond to market incentives by manipulating consumer perceptions
in whatever manner maximizes profits. Our accompanying articles
provide market evidence of this possibility of manipulation, as well as
an argument that enterprise liability provides the best legal response
to it.5 31 Here, we conclude by explaining why two of the most persis-
tent criticisms of the behavioral research do not affect our understand-
ing of manufacturer manipulation of consumer risk perceptions.

D. Answering the Critiques of Behavioral Research

As noted in Part I, the behavioral research has been criticized on
the ground that researchers allegedly tend to cite instances of poor
analytical reasoning, while failing to credit the numerous situations in
which individuals act in accordance with principles of rationality. 2
This so-called citation bias, however valid as applied to behavioral re-
search generally, is irrelevant to our depiction of consumer product
markets. Regardless of whether instances of cognitive bias make up
only a small percentage of overall consumer thought processes, manu-
facturers have every incentive to isolate those instances and utilize
them for gain. Thus, for instance, manufacturers have the incentive
and ability to-exploit consumers by triggering instances of mental ac-
counting even if it can be shown that, in the absence of such exploita-
tion, consumers would only rarely rely on mental accounting. Put
another way, regardless of whether mental accounting is a relatively
infrequent cognitive phenomena for consumers generally, it will not
be infrequent when consumers purchase a new car-this is because
car salespeople know that many new car purchasers are easily per-
suaded to purchase high-margin options such as stereos. Thus, the ci-
tation bias, for our purposes, gives way to the exploitation bias: the
tendency of manufacturers and retailers to exploit only those con-
sumer cognitive phenomena that result in increased sales, higher prof-
its, and decreased perceptions of risk.
530 See supra note 449.
531 See Hanson & Kysar, TBS II, supra note 25; Hanson & Kysar, TBS I, supra note
26.
532 See supra note 276 and accompanying text.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

Similarly, our depiction of consumer product markets is unaf-

fected by the frequent criticism that the behavioral research bears lit-
tle relevance to "real world" situations. Schwartz has restated this
argument as follows:
The external validity of [the behavioral research] is now in contro-
versy ... [because] the tasks people are assigned in laboratories
sometimes seem too artificial to support a strong inference that per-
sons routinely misperform important tasks in their actual lives.
Consequently, it seems premature to make 533 this experimental data
the factual premise of important legal rules.
There are two possible objections to this argument. First, one
might argue in defense of the behavioral research that the researchers'
laboratory conditions effectively replicate real world situations and
that, if anything, the real world is far more complex and therefore
more likely to induce analytical reasoning errors. 534 Second, even if
cognitive anomalies are phenomena especially generated by labora-
tory settings, there are many reasons to suppose that they will be
equally present in the consumer product setting. This is because man-
ufacturers, through careful marketing strategies and control over
product information, are able to construct "laboratories" of con-
sumerism. That is, manufacturers have the incentive and ability to in-
duce anomalies in consumers in just the manner that behavioral
researchers do for laboratory subjects.
Both manufacturers and researchers have a large degree of con-
trol over the content and form of information conveyed to their sub-
jects, both have command of their subjects' attention whether through
a survey questionnaire or a television commercial or product package,
and both are able to determine precisely the nature of the decision
context-whether it is in a lab or in front of a carefully arranged store
display. Perhaps the only difference between manufacturers and be-
havioral researchers is that manufacturers are operating under much
stronger financial incentives and typically have more opportunities to
ensure that their efforts succeed. In short, even if cognitive anomalies
are less frequent in the "real world," there is no reason to suppose
that the typical consumer product context is representative of the
"real world" and therefore no reason to doubt the existence of cogni-
tive anomalies in consumer purchasing decisions.

533 Schwartz, supra note 252, at 380.

534 As one commentator put it: "If biases, errors, and mistakes are so easy to produce in
laboratory reasoning tasks, it beggars belief to suppose that these are easily avoided at all
other times." Evans, supra note 46, at 25.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

CONCLUSION

In recent decades legal economists have converged upon numer-

ous broad (if rarely articulated) tenets-tenets that we believe have
had a profound influence on the sort of policy analysis that legal econ-
omists conduct and the sort of conclusions that they reach. Perhaps
the most significant example is the nearly axiomatic view that there is
no force as powerful as marketforces-that is, the incentives that peo-
ple (or firms) have to exchange goods or money in order to increase
their own personal welfare. A second tenet is that, by happy coinci-
dence, well-functioning markets lead to desirable (that is, social-wel-
fare maximizing or efficient) outcomes. A third tenet follows from the
first two: Policymakers should not attempt to regulate or otherwise
"interfere" with markets absent some evidence of a market failure.53S
Finally, a fourth tenet is that there exists only a finite set of clearly
defined and reasonably tractable market failures-e.g., collective ac-
tion problems, externalities, informational asymmetries, and strategic
behavior.
With the first three tenets, we are, at least for purposes of this
Article, largely in agreement. It is the fourth that we seek to chal-
lenge. Conspicuously absent from the set of potential market failures
is the problem of irrational actors. That is, economists have tradition-
ally paid little mind to the possibility that consumers or firms cannot
or do not act in accord with the rational actor model on which most
efficiency analysis is premised. For that seeming oversight, they have
offered two general justifications. First, they assert that there is inade-
quate evidence that actors are biased in a way that threatens the basic
law and economics model. To be sure, it is easy to come up with anec-
dotal evidence of irrationality, but a great deal of policy-related schol-
arship is concerned with group behavior, not individual behavior.5 36
Furthermore, absent any reason to believe that actors are biased in
any particular way, the law of large numbers ensures that the errors of
individuals will wash out and the estimates of the group will be
roughly accurate.5 3 7 As the evidence of systematic bias has mounted,

535 Relatedly, legal economists generally assume that the goal of any regulation should
be to replicate the results that the hampered market would otherwise yield, if well-func-
tioning. It is for those sorts of reasons that there has emerged a consensus among econo-
mists that the best way to ensure that regulation is effective is to employ regulation that
relies on the same powerful incentives that drive well-functioning markets-so.called in-
centive-based regulation. See Hanson & Logue, supra note 93, at 1174 (arguing trend in
regulation is toward incentive-based regulation).
536 Cf. Richard G. Lipsey, Peter 0. Steiner, & Douglas D. Purvis, Economics 19 (8th ed.
1987) ("There are many situations in which group behavior can be predicted accurately
without certain knowledge of individual behavior.").
537 See id. at 21:

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

however, this justification for downplaying the problem of nonrational

actors has, as we detailed in Part II, given way to debate over the net
effect of many potentially conflicting biases. The second general type
of justification is that, even if there were some tendency of economic
actors to be biased, those tendencies would be largely quashed
through various forms of market discipline.5 38 This justification,
which depends in part on some of the economic tenets just described,
takes a variety of forms, most of which involve a more or less explicit
analogy to natural selection accounts of biological evolution.5 39 Prob-
ably the best known economic rendition of the point was made in a
now-classic essay by Milton Friedman. 540 As Friedman argued, eco-
nomic actors will, whether conscious of it or not, behave "as if" they
are attempting to maximize their own utility just as an expert pool
player will behave "as if" he or she is knowledgeable of the complex
physics of motion entailed in the game.5 41 Survival in the market-
place, just like success in billiards, requires as much. Current assess-
ments of the significance of behavioralism for law and economics
contain the same threads of "as if" rationality and Darwinian
discipline. 542

[Because of the law of large numbers,] we can assert with confidence that more
people will make small errors than will make large errors, that the larger the
error the fewer will be the number making it, that roughly the same number of
people will overstate as will understate the distance, and that the average error
of all individuals will be close to zero.
538 See, e.g., Peter Knez, Vernon L. Smith, & Arlington W. Williams, Individual Ration-
ality, Market Rationality, and Value Estimation, 75 Am. Econ. Rev.: Papers & Proc. 397,
401 (1985) (suggesting that individual irrational behavior decreases with increased market
exposure).
539 See, e.g., Jack Hirshleifer, Economics from a Biological Viewpoint, 20 J.L. & Econ.
1, 9-10 (1977) (using natural selection analogy to explain why successful firms pursue profit
maximization); Frank H. Easterbrook & Daniel R. Fischel, The Economic Structure of
Corporate Law 31 (1991) (finding that market weeds out practices that do not assist inves-
tors); Frank H. Easterbrook, The Supreme Court, 1983 Term-Foreword: The Court and
the Economic System, 98 Harv. L. Rev. 4, 8 (1984) (comparing evolutionary forces disci-
plining courts and those disciplining market participants and writing: "Often [business
managers] thrash about quite innocent of economic theory. No matter. Those who offer
what consumers want-by design or by accident-and produce it at low cost will prosper.
Rewards and punishments arise automatically in any market system.").
540 Friedman, supra note 448.
541 See id. at 21-22; see also id. at 15-16 (defending "as if" arguments and arguing that
utilitarian model is good for describing and predicting individual actions, even if individual
does not consciously engage in utility-maximizing choice); Richard A. Posner, Economic
Analysis of Law 4 (5th ed. 1998) ("Rational maximization should not be confused with
conscious calculation.").
542 See, e.g., Jennifer Arlen, Comment: The Future of Behavioral Economic Analysis of
Law, 51 Vand. L. Rev. 1765, 1782 (1998) (applying two strands to show that in futures
market those who incorrectly assess risks are eliminated and market as whole functions as
if individuals rationally assess risk).

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

In our view, this Article substantially undermines those justifica-

tions by showing that they, in effect, do not take behavioralism seri-
ously. The behavioralist literature reviewed here makes clear the
potential for a new sort of market failure, market manipulation: Be-
cause individuals are subject to a host of nonrational yet systematic
cognitive phenomena, any party who has control over a decisionmak-
ing context can influence the perceptions of the decisionmaker. When
a party to a transaction has the ability to assert this influence, the un-
derlying transaction will not necessarily yield an increase in social wel-
fare. Indeed, flipping Friedman's classic justification of the rational
actor model, one might say that the evolutionary forces of the market
will force the parties in the dominant position to behave "as if' they
know and understand how best to use the teachings of the behavioral
literature to manipulate other actors for gain.
For instance, to use the example from this Article, manufacturers
and marketers in consumer product markets can shape consumer risk
assessments by altering the way they manufacture, package, and mar-
ket their products. What is more, that possibility creates profit-en-
hancing opportunities for manufacturers and marketers by affording
them a way of increasing consumer willingness to pay. As a result, the
problem of market manipulation seems inescapable in an unregulated
consumer product market. Manufacturers, to survive, must behave
"as if" they are attempting to manipulate consumer risk perceptions.
And in light of the immense power of the market forces driving these
attempts, it seems highly doubtful that manufacturer strategies (be
they deliberate or accidental) will fail.
Of course, the reader might view our arguments as overstated.
Indeed, the reader might react to these arguments as Viscusi reacted
to Latin's arguments-by viewing them as unduly pessimistic, even
paranoid.543 Admittedly, there are overtones of pessimism in a theory
that suggests that "we fall into all sorts of cognitive traps at almost
every moment of our lives." 544 There may be a tinge of paranoia in a
theory that predicts massive corporations will behave as if they have
meticulously studied consumers' cognitive processes, all so that we
may be manipulated like marionettes in a multibillion-dollar game of
consumption. Surely, one might respond, we have mistaken a cloudy
sky for a falling one.
Initially, our reaction was the same: We hesitated to believe such
a pessimistic, even condescending, view of consumers and, by implica-

543 See supra notes 362-64 and accompanying text (describing Viscusi's characterization
of "Howard Latin's Law").
544 Piattelli-Palmarini, supra note 3, at 28.

Imaged with the Permission of N.Y.U. Law Review

 NEW YORK UNIVERSITY LAW REVIEW [Vol. 74:630

tion, ourselves. After learning, however, that the manufacturer of

Campbell's Soup knows, as an empirical fact, that placing soup cans
out of alphabetical order on store shelves will increase sales by exactly
six percent,545 our view began to change. After discovering that retail-
ers, studying such research as "ArousalHypotheses" and the Effects of
Music on PurchasingBehavior,546 can lower customer blink rates from
the normal average of thirty-two times a minute to a narcotic fourteen
blinks a minute, we began to suspect that our manipulation hypothesis
was not so far-fetched after all. And, finally, when we learned that
Anheuser Busch, coiners of the beer-marketing gem, "Put a little
weekend in your week," had relied on a decade-long psychological
study of drinking behavior to develop marketing proffles capable of
targeting those who use alcohol to escape their problems,5 47 we knew
that we were on to a full-fledged theory.
These examples provide a tiny sample of the empirical evidence
that we offer in our accompanying article.5 48 They are, of course, an-
ecdotal, and much of our evidence will be similarly modest. Our aim,
however, also will be modest: to show, as an empirical matter, that at
least some manufacturer manipulation of consumers is occurring in
the market consistent with our theoretical prediction, and to make
what sounds compelling in theory also seem plausible in practice. Af-
ter providing market evidence toward this end, we will consider the
implications of this evidence for the still raging products liability de-
bate regarding the most desirable liability standard.5 49 We will argue
that an enterprise liability regime provides the best available response

545 The reason, of course, is that by making it more difficult to find our desired can of
tomato soup, we are more likely to stumble across that attractive-looking can of clam
chowder. See Jack Hitt, The Theory of Supermarkets, N.Y. Times, Mar. 10, 1996, § 6
(Magazine), at 58 (describing strategy in supermarket product placement).
546 Patricia Cain Smith & Ross Curnow,"Arousal Hypothesis" and the Effects of Music
on Purchasing Behavior, 50 J. Applied Psychol. 255, 255-56 (1966) (showing level of music
affects length of time consumers spend in stores); see also Ronald E. Milliman, Using
Background Music to Affect the Behavior of Supermarket Shoppers, J. Marketing, Sum-
mer 1982, at 86, 90 (finding tempo of music in store can affect pace of customer flow and
sales volume).
547 See Russel L. Ackoff & James R. Emshoff, Advertising Research at Anheuser-
Busch, Inc. (1968-74), Sloan Mgmt. Rev., Spring 1975, at 1 (reviewing market research
analyzing link between drinking patterns and advertising). A similar tactic was apparently
investigated by the Federal Trade Commission in 1976: "According to [FTCJ staff, studies
[that identified problem drinkers] were used in an advertisement for Johnnie Walker Black
Label which appeared to offer the product to relieve stress and tension associated with
striving for success-'The road to success is paved with rocks. Let us smooth them for
you."' Eric Clark, The Want Makers 276 (1988) (one set of internal quotation marks omit-
ted) (reviewing tactics used by alcohol marketers).
548 See Hanson & Kysar, TBS II, supra note 25.
549 See Hanson & Kysar, TBS III, supra note 26.

Imaged with the Permission of N.Y.U. Law Review

June 1999] TAKING BEHAVIORALISM SERIOUSLY

to the problem of market manipulation. Implications of the problem

of market manipulation for areas of law outside of products liability
we will leave for still later articles.

Imaged with the Permission of N.Y.U. Law Review