[Paleopsych] Vernon L. Smith: Constructivist and Ecological Rationality in Economics
checker at panix.com
Sat Jan 21 01:06:11 UTC 2006
Vernon L. Smith: Constructivist and Ecological Rationality in Economics
The American Economic Review, 2003 June, pp. 465-508
[This article is a fabulous one. It brings together the concepts of
Friedrich Hayek and his notion of dispersed knowledge, the separate
concepts of rationality in the classical SSSM (Standard Socioeconomic
Science Model, which is NOT the SSSM (Standard Social Science Model, with
its avoidance of Darwin, whose collapse last year I reported) and the
rationality of brains with shaply limited abilities to know and to
calculate, experimental economics (the work in which garnered Smith the
Nobel Prize in economics), game theory, and neurobiology.
[Read it as closely as you wish, though it assumes a fairly high level of
understanding of work in game theory and economics. The talk itself is
given at http://nobelprize.org/economics/laureates/2002/smith-lecture.pdf
and appears to be shorter that what you see below. I tried mightily to
modify the output of Adobe's PDF --> TXT converter of the American
Economic Review article to something more readable. If you'd like the AER
PDF, send me an e-mail.]
This article is a revised version of the lecture Vernon L. Smith delivered
in Stockholm, Sweden, on December 8, 2002, when he received the Bankof
Sweden Prize in Economic Sciences in Memory of Alfred Nobel.
The title was suggestedto the author in the paper by Joel Norman, "Two
Visual Systems and Two Theories of Perception: An Attempt to Reconcile
the Constructivist and Ecological Approaches" (Behavioral and Brain
Sciences, 2002). After .nishing this paper I found that my use of the term
below had been used byGerd Gigerenzer et al. (1999), for application to
"fast and frugal decision making," by individuals: "A heuristic is
ecologically rational to the degree thatitis adapted to the structure
ofan environment" (p. 13). Hayek, in the citations below, identi. es both
kinds of rationality.
* Department of Economics and Law, Interdisciplinary Center for Economic
Science, George Mason University, 4400 UniversityDrive, MSN 1B2, Fairfax,
VA 22030. Doing experimental economics has changed the way I thinkabout
economics. There are many reasons for this, but one of the most prominent
is that designing and conducting experiments forces you to think through
the process rules and procedures of an institution. Few, like Einstein,
can perform detailed and imaginative mental experiments. Most of us need
the challenge of real experiments to discipline our thinking. In this
paper I hope to indicate how my thinking has been changed in some detail.
Iam indebted to Sid Siegel for technical and conceptual inspiration; to
John Hughes, Stan Reiter, andthe Purdue Universityfacultyfrom 1955 to 1967
for warm, tolerant support beginning with my .rst experiment; to Charles
Plott, Charles Holt, and Martin Shubik for many valuable encounters over
the years on institutional and experimental issues; to students,
visitors, the current ICES team, and especially to my growing but
tolerating family who have made all these years the best years of my
When we leave our closet, and engage in the common affairs oflife,
(reason's) conclusions seem to vanish, like the phantoms ofthe nighton
the appearance of the morning;and 'tis dif. cult for us to retain even
that conviction, which we had attained with dif.culty... (David Hume,
1739, 1985, p. 507).
... we must constantly adjust our lives, our thoughts and our emotions, in
order to live simultaneouslywithin different kinds of orders according to
different rules. If we were to apply the unmodi.ed, uncurbed rules
(ofcaring intervention to do visible "good") of the ... small band or
troop, or ... our families ... to the (extendedorder of cooperation
throughmarkets), as our instincts and sentimental yearnings often make us
wish to do, we would destroy it. Yet ifwe were to always apply the
(competitive) rules of the extendedorder to our more intimate groupings,
we would crush them (Friedrich A. Hayek, 1988, p. 18; italics are his,
parenthetical reductions are mine).
We have become accustomed to the idea that a naturalsystem like the human
body or an ecosystem regulates itself. To explain the regulation, we
lookfor feedback loops rather than a centralplanningand directing body.
But somehow our intuitions about selfregulation do not carry over to the
arti. cial systems of human society. (Thus) ... the ... disbelief always
expressed by (my) architecture students (about)... medieval cities as
marvelously patterned systems that had mostly just "grown" in response to
myriads of individualdecisions. To my studentsa pattern implieda planner
... . The idea that a city could acquire its pattern as "naturally" as a
snow. ake was foreign to them (Herbert Alexander Simon, 1981, 1996, p.
Historically, a recurrent theme in economics is that the values to which
people respond are notcon.ned to thoseone wouldexpectbasedon the narrowly
de.ned canons of rationality. These roots go back to Adam Smith (1759,
1776), who examined the moralsympathiesthat characterize natural human
sociality.1Contrary to vulgar impressions, in Smiths view, each
individual de. ned and pursued his own interest in his own way, and
individuals were mischaracterized
1 Economistsare largelyuntouched by Smiths .rst great work, which was
eclipsed by the Wealthof Nations. Thus, one of the professions bestknown
historian of economic thought, "found these two works in some measure
basically inconsistent" (Jacob Viner, 1991, p. 250). For a contrary
interpretation seeSmith (1998). Many of the references herein to my own
andcoauthored work have been reprinted in Smith (1991, 2000).
by the metaphor, "economic man." (cf., Hayek, 1991, p. 120). This careless
scholarship fails to recognize the key proposition articulated by the
Scottish philosophers: to do good for others, does not require deliberate
action to further the perceived interest ofothers. As Mandeville so
succinctlyput it, "The worst of all the multitude did something for the
common good." (See Mandevilles poem, "The Grumbling Hive" or "Knaves
Turned Honest," 1705; quoted in Hayek, 1991, p. 82.) Manycontemporary
scholars, and not only popular writers, have reversed Mandevilles
proposition, and argued that the standard socioeconomic science model
(SSSM) requires,justi.es, and promotes sel.sh behavior.2On the contrary,
because enforceable rights can never cover every margin of decision,
opportunism in all relational contracting and exchange across time are
costs, not bene.ts, in achievinglongterm value from trade;an ideology of
honesty3 means that people play the game of "trade," rather than "steal,"
although crime may often pay the rational lawbreaker who always chooses
dominant strategies. Nor does nonsel. sh behavior in ordinary market
transactions prevent those transactions from promoting specialization and
Cultures that have evolved markets have enormously expandedresource
specialization, created commensurate gains from exchange, and are
wealthier than those that have not. This proposition says nothing about
the necessity of human sel. shness--the increased wealth of particular
individuals can be used for consumption, investment, to pay taxes, for
Macarthur Fellows, gifted to the symphony, the Smithsonian, or the poor.4
2 That A implies B in no sense allows the reverse statement. But why
would we economists confuse necessary with suf. cient conditions?The
textfrom Hume provides the answer. No one can consistentlyapplyrational
logical principles to everything he or she does; if there are cognitive
costs in every application then the effort cost will often exceed the
bene. ts (Smith and Ferenc Szidarovszky, 2003). Theorists live by
provingtheorems, and when in this mode we rarely make such errors. A
missingchapter in the study of bounded rationality is its application to
understanding, and accepting with a little humility, the severe
limitations it imposes on our developmentof economic theory.
3 Douglass Cecil North (1981) has emphasized the importance of ideology
in promoting economic growth. 4In the Potlatch, some wealth-- created in
part by private property rights in .shing grounds--was publicly
destroyed. Markets economize on the need for virtue, but do not eliminate
Researchin economic psychology5has prominently reported examples where
"fairness" considerations are said to contradict the rationality
assumptions of the standardsocioeconomic science model. But
experimentaleconomists have reported mixed results on rationality: people
are often better (e.g., in twoperson anonymous interactions), in
agreement with (e.g., in .ow supply and demand markets), or worse (e.g.,
in asset trading), in achieving gains for themselves and others than is
predicted by rational analysis. Patterns in these contradictions and con.
rmations provide important clues to the implicit rules or norms that
people may follow, and can motivate new theoretical hypotheses for
examination in both the .eld and the laboratory. The pattern of results
greatly modi.es the prevailing, and I believe misguided, rational SSSM,
and richly modernizes the unadulterated message of the Scottish
I. On Two Forms of Rationality
The organizing principle throughout this paper is the simultaneous
existence of two rational orders. I shall try to make the case that both
orders are distinguishingcharacteristics of what we are as
socialcreatures; that both are essential to understanding and unifying a
large body of experience from socioeconomic life and the experimental
laboratory, and in charting relevant new directions for economic theory as
well as experimentalempirical programs.
A. Constructivist Rationality
The .rst concept of a rational order derives from the SSSM going back to
the seventeenth century. The SSSM is an example of what Hayek has called
constructivist rationality (or "constructivism"), which stems particularly
5I will use the term "economic psychology" generally to refer to cognitive
psychology as it has been applied to economic questions, and to a third
sub.eld of experimental methods in economics recently
productdifferentiated as "behavioral economics" (Sendhil Mullainathan and
Richard H. Thaler, 2001), and further differentiated into "behavioral game
theory" (Colin F. Camerer, 2002); the original foundations were laid byW.
Edwards, DannyKahneman, Anatol Rapoport, Paul Slovic, and Amos Tversky to
name some of the most prominent.
from Descartes (also Bacon and Hobbes),6 who believed and argued that all
worthwhile social institutions were and should be created byconscious
deductiveprocesses of human reason.7In economics the SSSM leads to
rational predictive models of decision that motivate research hypotheses
thatexperimentalists have been testing in the laboratory since the
midtwentieth century. Although the test results tend to be con.rming in
impersonal market exchange, the results are famously and recalcitrantly
mixedin "personal exchange," notably in a great variety of twoperson
extensiveform games where some half of the people attempt and frequently
succeed when risking cooperation, even when anonymously paired.8These
results have motivated constructivist extensions of game theory based on
otherregarding, in addition to ownregarding, preferences (e.g., Gary E
Bolton, 1991; Matthew Rabin, 1993), and on "learning"--the idea that the
predictions of the SSSM might be approached over time by trialanderror
adaptation processes (Ido Erev and
6 In the nineteenth century, Bentham and John Stuart Mill were among the
leading constructivists. Bentham (and the utilitarians) sought to "...
remake the whole of... (British) ... law and institutions on rational
principles" (Hayek, 1960, p. 174). Mill introduced the muchabused
constructivist concept of(but not the name)"natural monopoly." To Mill it
was transparently wasteful and duplicative to have two or more mail
carriers operating on the same route. He is the intellectual father of the
U.S. and other postal monopolies around the world, their resistance to
innovation, and their demise in the face of the privatization movement in
some countries and the growth of superior substitutes in others. Mill
could not imagine that it would be ef.cient for two cities to be connected
by two parallel railroad tracks (Mill, 1899, Vol. 1, pp. 131, 141- 42;
Vol. 2, p. 463). Mill died in 1873. I would conjecture that by that date,
or soon thereafter, men with gradeschool educations had become rich
constructing the .rst parallelroute railroads. These emergent
contradictions to constructivist natural monopoly are examples ofwhat we
shall call ecological rationality, as detailed below.
7 "... Descartes contended that all the useful human institutions were
and ought to be deliberate creation(s) of conscious reason ... a capacity
of the mind to arrive at the truth by a deductive process from a few
obvious and undoubtable premises" (Hayek, 1967, p. 85).
8 Behavioral economists have made a cottage industry of showing that the
SSSM assumptions seem to apply almost nowhere to real decisions. This is
because their research program has been a deliberate search in the tails
of distributions for "Identifyingthe ways in which behavior differs from
the standard model ..." (Mullainathan and Thaler, 2001, Vol. II, p. 1094),
a search that can only succeed. Alvin E. Roth, 1998; Camerer and TeckHua
An alternative and perhaps complimentary explanation of some of these
contradictions to theory is that people may use socialgrown norms of
trust and reciprocity9 (including equity, meaning to each according to
his justly earned dessert; i.e., equality ofopportunity, not outcome)to
achieve cooperative states superior to individuallyrational defection
outcomes. We will report some experimental tests designedto separate
competing preference and reciprocity theories of behavior in personal
exchange. Although reciprocity seems to be a leader in the comparisons we
summarize, its strength is not uniform across alltests, andmuch remains to
be learned aboutthe hidden recesses ofmeaningin human behaviorandthe
circumstances in which cooperative or noncooperativebehavior is
manifest.10Technically, the issue is how most productively to model
agent "types" byextending game theory so that types are an integral part
of its predictive content, rather than merely imported as an ex post
technical explanation of experimental results. For example, moves can
signal types, andeffect decision,which explains why game form matters,
andwhy payoffs available, but forgone, can effect outcomes. These
elements must be part ofthe internal structure of the theory such that
outcomes become predictions conditional on the elementarycharacteristics
of players who read each others intentions. If successful, many of the
basic results in game
9 Dissatis.ed with the utilitarian approach because its predictions fail
to account for the observed importance of instructions/procedures, we
began investigating the reciprocity hypothesis in Elizabeth Hoffman et
al. (1994). Mechanically, utilities can serve as intermediate
placeholders for reciprocal trust, but, as surface indicators, serve
poorly to generate new hypotheses designed to understandinteractive
processes. Good theory must be an engine for generating testable
hypotheses, and utility theory runs outof fuel quickly. Utility values are
seen as providing the ultimate "given" data, and the conversation stops.
10I am reminded of a department head from HewlettPackard visitingour lab.
Inaively assumed that he would be most interested in demonstrations of
some of our market experiments. Not so. He was more interested in the
"trust" experiments. Why?He saw the HP management problem as one
ofgettingteams to cooperate internallyby buildingtrust and
trustworthiness, while being vigorous competitors externally. Could the
trust games serve as a measurement and teaching tool for helpingto solve
this problem? This nicely illustrates the tension in Hayeks twoworlds
quote in the text. theory would become special cases of the extended
In market experiments--where cooperation can occur through the
coordination function of prices produced by, but simultaneously resulting
from, interaction with individual choice behavior--the results are more
commonly in accord with standard competitive models that maximize group
welfare. This professionalvictory is hollowed by the failure of standard
theory to predict the "surprisingly"11 weak conditions under which the
results obtain.12 Thus, for tractability, Cartesian rationalism
provisionally assumes or "requires" agents to possess complete payoff and
other information-- far more than could ever be given to one mind. In
economics the resulting exercises are believed to sharpen economic
thinking, as if-then parables. Yet, these assumptions are unlikelyto
approximate the levelofignorance thathas conditioned either individual
behavior, or our evolved institutions, as abstract norms or rules
independent of particular parameters, which have survived as part of the
world of experience.13 The temptation is to ignore this reality because
it is poorly understood, and does not yieldto our familiar but inadequate
modeling tools, and to proceed in the implicit belief that our parables
capture what is most essential
11Robert B. Wilson (1992, p. 256)discusses an ef. ciency theorem, and
suggests that the phenomenon is "perhaps unsurprising." It is, nowadays,
but few believe it; also, theory has lagged well behind the evidence, and
yields inadequate testable insight into the process dynamics operating in
12I want to acknowledge correspondence with Charles Plott andaddthe
following:Althoughthis is a giant victory for the economic theory of
markets it simultaneously demonstrates that the theory is incomplete. The
unexpectedly weak conditions under which the results obtain are good news
for market performance, but not such good news for the scienti. c
community because itdemonstrates that we do not understand why markets
work as they do. You do not have to have large numbers of agents, each an
insigni. cant part ofthe whole--three or four buyers andas manysellers are
entirely adequate in a wide range of economic environments; they do not
have to have complete or perfect or common information--each can have only
private information; nor is it required that individuals make decisions
systematically or be economically sophisticated.
13Throughout the paper I will use "environment" to mean the collection of
agentvalues (preferences) that de.ne the gains from trade; "institution"
to refer to the language (messages), rules of message exchange, and
contract in a market; and "behavior" for agent message choices
conditional on the environment and institution (Smith, 1982).
about what we observe. Having sharpened our understanding on Cartesian
complete information parables we carry these tools into the world for
applicationwithoutall the necessary caveats that re. ect the tractability
constraints imposed by our bounded professional cognitive capacities as
In summary, constructivism uses reason to deliberately create rules of
action, and create human socioeconomic institutions that yield outcomes
deemed preferable, given particular circumstances, to those producedby
alternative arrangements. Although constructivism is one of the crowning
achievements of the human intellect, it is important to remain sensitive
to the fact that human institutions and most decision makingis not guided
primarily, ifat all, by constructivism. Emergent arrangements, even if
initially constructivist in form, must have survival properties that take
accountofopportunity costs and environmentalchallenges invisible to our
B. Limitations and Distractions of Constructivist Rationality
Since our theories and thought processes about social systems involve the
conscious and deliberate use of reason, it is necessary to constantly
remindourselves that human activity is diffused and dominated by
unconscious, autonomic, neuropsychologicalsystems that enable people to
function effectively without always calling upon the brains scarcest
resource-- attentional and reasoning circuitry. This is an important
economizing property of how the brain works. If it were otherwise, no one
could get through the day under the burden of the selfconscious
monitoring and planning of every trivial action in detail.14 Also, no one
can express in thoughts, let alone words, all that he or she knows,
anddoes not know but might call upon, or need to discover, for some
purposive action. Imagine the strain on the brains resources if at the
supermarket a shopper were required to explicitly evaluate his preferences
for every combination of the tens of thousands of grocery items that are
feasible for a given
14 "If we stopped doing everythingfor which we do not know the reason, or
for which we cannot provide ajusti. cation ... we would probably soon be
dead" (Hayek, 1988, p. 68).
budget. Such mental processes are enormously opportunitycostly and
implicitly our brain knows, if our conscious mind does not know, that we
must avoid incurring opportunity costs that are not worth the bene. t.15
The challenge ofanyunfamiliar action or problemappears .rst to trigger a
search bythe brain to bring to the conscious mind what one knows that is
related to the decision context. Context triggers autobiographic
experiential memory, which explains why context surfaces as a nontrivial
treatment, particularly in small group experiments. The brain (including
the entire neurophysiological system) takes over directly in the case of
familiar, mastered tasks, and plays the equivalent of lightening chess
when the "expert" trades, plays Beethovens Fifthpiano concerto, or
connects with a 95mile/hour fastball--all without selfaware "thinking"
by the mind.
We fail utterly to possess natural mechanisms for reminding ourselves of
the brains offline activities and accomplishments. This important
proposition has led Michael S. Gazzaniga (1998)to askwhy the brain fools
the mind into believing it is in control.16 Andto Hayek, who
15Expected utility theory is for teaching (as Wassily Leontief once
suggested), but also for the constructivist modeling of consistent choice.
It seems inadequate for the prediction, or the ecological understanding,
ofbehavior. Its inadequacy for prediction has been plainly emphasized in
the many contributions of Amos Tversky and Daniel Kahneman (see their
1987 paper for an excellent summary statement), some of which have been
quali. ed and reinterpreted in the work of Gigerenzer et al. (1999). The
results are far more encouraging in the context of markets, where subjects
are not consciously maximizing. See, e.g., Smith (1991, 2000)and Plott
(2001). Joyce Berg et al. (1994), in a signi. cant paper, .nd that the
measurement of risk aversion varies with the type of market institution
or procedure used in extracting risk measures from decisions. See Smith
and Szidarovszky (2003) for a constructivist utilitarian treatmentof
decision whose reward outcome requires cognitive costs to be incurred:
objective rationality is not subjectively rational, and therefore it is
not optimal for the individualto apply the objectively "optimal"
16 "By the time we thinkwe know something--it is part ofour conscious
experience--the brain has already done its work. It is old news to the
brain, but fresh to us. Systems built into the brain do their work
automatically and largely outside ofour conscious awareness. The brain
.nishes the work half a second before the information it processes reaches
our consciousness. ... We are clueless about how all this works and gets
effected. We dont plan or articulate these actions. We simply observe the
output. ... The brain begins to cover for this done deal aspect ofits
functioning thoroughly understood this proposition, what was the "fatal
conceit"? "The idea that the ability to acquire skills stems from
reason." The constructivist mindmakes a fatal "error," blinding itself to
understanding, as we are warned, "one should never suppose that our reason
is in the higher critical position andthatonly those moral rules are valid
that reason endorses" (Hayek, 1988, p. 21). But the anthropocentric
(morphic) mind routinely makes this signi. cant error.
Most ofour operating knowledge17 we do not remember learning. Natural
language is the most prominent example, but also music and virtually
everything that constitutes our developmentalsocialization. We learn the
rules ofa language and of ef. cient social intercourse,
withoutexplicitinstruction, simply byexposure to family and extended
family social networks (Jerome Kagan and Sharon Lamb, 1987; Alan Page
Fiske, 1991;Kagan, 1994;Steven Pinker, 1994). That the brain is capable of
offline subconscious learning is shown by experiments withamnesiacs who
are taught a new task. They learn to perform well, but memory of having
learned the task escapes them (Barbara J. Knowlton et al., 1996).
C. Ecological Rationality
These considerations lead to the secondconcept of a rational order, as an
undesigned ecological system that emerges out of cultural and biological
evolutionary18 processes: homegrown
by creating in us the illusion that the events we are experiencingare
happeningin real time--notbefore our conscious experience ofdeciding to do
something" (Gazzaniga, 1998, pp. 63-64).
17 Hayek (1967, p. 44) notes that "... modern English usage does not
permit generallyto employthe verb can (in the sense of the German
konnen) to describe all those instances in which an individualmerely
knows how to do a thing... (including)... the capacity to act according
to rules which we may be able to discover but which we need not be able to
state in order to obey them."
18Manyrecognize that evolutionary processes are necessarily
coevolutionary,butthere is also deep denial of this, and bias, that all is
due to "culture" (which is even more poorly understood than biology),
leading Pinker (2002)to investigate why. Heritable abstract function can
become dormant, atrophied, or malfunctional in the absence of
initializing input on a developmental time schedule for the brains
vision, language, and socialization circuitry. That these processes are
coevolutionary is evidentin the study of twins (Nancy L. Segal, 1999).
Deconstructivist reports argue that these studies exhibit many of the
usual data and statistical identi.cation problems (Arthur S. Goldberger,
1979), but the need is for positive revisionist analysis.
principlesofaction, norms, traditions, and"morality."19 Ecological
rationality uses reason-- rationalreconstruction--to examine the behavior
of individuals based on their experience and folkknowledge, who are
"naive" in their ability to apply constructivist tools to the decisions
they make;to understandthe emergentorder in human cultures; to discover
the possible intelligence embodiedin the rules, norms, and institutions
of our cultural and biological heritage that are createdfrom human
interactions but not bydeliberatehuman design. People follow rules without
being able to articulate them, but they can be discovered. This is the
intellectualheritage ofthe Scottish philosophers, who described and
interpreted the social and economic order they observed.
An eighteenthcentury precursor of Herbert Simon, David Hume was concerned
with the limits of reason, the bounds on human understanding, and with
scalingbackthe exaggerated claims of Cartesian constructivism. To Hume,
rationality was phenomena that reason discovers in emergent institutions.
Thus, "the rules of morality... are not conclusions of(our) reason."
(Hume, 1985, p. 235). Adam Smith developed the idea ofemergent order for
economics. Truth is discovered in the form of the intelligence embodied in
rules and traditions that have formed, inscrutably, outofthe
ancienthistory of human social interactions. This is the antithesis of the
anthropocentric beliefthat ifan observed social mechanism is functional,
somebody in the unrecordedpast must have used reason con
19 "Morality" refers to any maxim of cohesive social behavior that
survives the test of time, and is prominently represented by the great
"shalt not" prohibitions of the leading world religions: thou shalt not:
(1) steal, (2) covet the possessions of others, commit (3)murder, (4)
adultery, or (5) bear false witness. The .rst two de.ne and defend
property rights in the product of ones labor, and all resources
accumulated bysuch labor, enabling the emergence of the extended order of
mind through markets. The last three commandments protect the sanctity of
social exchange--the external order of the mind. These modest exclusionary
constraints leave an immense scope for freedom within their bounds.
Corollaries, like the Buddhist liveandletlive version of the golden
rule, are explicit in this respect:"do not unto others as you would have
them not do unto you." sciously to create it to serve its perceived
In experimental economics the eighteenthcentury Scottish tradition is
revealed in the observation ofemergent order in numerous studies of
existing market institutions such as the continuous double auction
(CDA).21To paraphrase Adam Smith, people in these experiments are led to
promote group welfareenhancing social ends that are not part of their
intention. This principle is supported by hundreds of experiments whose
environments and institutions (sealed bid, posted offer, and others
besides CDA) may exceed the capacity offormalgametheoretic analysis to
articulate predictive models. But they do not exceed the functional
capacity of collectives of incompletely informed human decision makers,
whose autonomic mental algorithms coordinate behavior through the rules
of the institution--social algorithms--to generate high levels of
Acknowledging and investigating the workings of unseen processes are
essential to the 20In cultural and biological evolution, order arises from
mechanisms for generating variation to which is applied mechanisms for
selection. I am indebted to Todd Zywicki, who, at a recent Liberty Fund
conference on "Hayek, Experiment and Freedom," observed that reason is
good at providing variation, but not selection. Constructivism is indeed
an engine for generating variation, but is far too limited in its ability
to comprehend and apply all the relevant facts to serve the process of
selection, which is better left to ecological processes. 21What
experimentalists have unintentionally brought to the table is a
methodology for objectively testing the ScottishHayekian hypotheses under
scienti. c controls. This answers the question Milton Friedman is said to
have raisedconcerningthe validity of Hayeks theory/reasoning: "How would
you know?" (I am unable to .nd/provide the reference). Remarkably, Hayek
stated on the very edge of recognizing what experiments could do for
testing his theory, then dismissed it. Thus,"We can test it (competition
as a discovery procedure) on conceptual models, and we might conceivably
test it in arti. cially created real situations, where the facts which
competition is intended to discover are already known to the observer. But
in such cases it is of no practical value, so that to carry out the
experiment would hardly be worth the expense" (Hayek, 1984, p. 255;also
see Smith, 2002, pp. 95-96). Economic historians, e.g.,
North(1981),andpolitical economists, e.g., Elinor Ostrom (1990), have long
explored the intelligence and ef. cacy embodied in emergent socioeconomic
institutions that solve, or fail to solve, problems of growth and
resource management. They study "natural" ecological experiments from
which we have learned immeasurably.
growthofour understandingof social phenomena, and enable us to probe
beyond the anthropocentric limitations ofconstructivism.
Both kinds of rationalityhave in. uenced the design and interpretation of
experiments in economics. Thus, if people in certain contexts make choices
that contradict our formaltheory of rationality, rather than conclude that
they are irrational, some ask why, reexamine maintained hypotheses
including all aspects of the experiments--procedures, payoffs, context,
instructions, etc.--and inquire as to what new concepts and experimental
designs can help us to better understandthe behavior. What is the
subjects perceptionofthe problem that they are trying to solve?
Finally, understanding decision requires knowledge beyond the
traditionalbounds of economics,22 a challenge to which Hume and Smith
were not strangers.23 This is manifest in the recentstudies of the
neuralcorrelates ofstrategic interaction (McCabe calls it
neuroeconomics) using fMRI and other brainimaging technologies. That
researchexplores the neurocorrelates of intentions or "mind reading," and
other hypotheses aboutinformation,choice, and own versus other payoffs in
determining interactive behavior.
The above themes will be illustrated and discussed in a wide variety of
examples drawn from economics, law, experimental economics, and
psychology. I will begin with impersonal exchange through markets, drawing
on the learning from experiments and .eld observations to illustrate how
the contrast between constructive and ecological rationality informs
learning from observation. Then I willexamine
personalexchange,particularlyin the context of twoperson extensiveform
games, asking why constructivist models are of limited success in
predicting behavior in singleplay games, even when subjects are
22I importune students to read narrowly within economics, but widelyin
science. Within economics there is essentially only one model to be
adapted to every application: optimization subject to constraints due to
resource limitations, institutional rules, and/or the behavior ofothers,
as in CournotNash equilibria. The economic literature is not the best
place to .ndnew inspiration beyond these traditional technical methods of
23Thus, for Hayek, "an economist who is onlyan economist cannot be a good
II. Impersonal Exchange: The Extended Order of the Market
A. How Are the Two Concepts of a Rational Order Related?
Constructivism takes as given the social structures generated by emergent
institutions that we observe in the world, and proceeds to model it
formally. An example would be the Dutch auction or its alleged isomorphic
equivalent, the sealedbidauction (William Vickrey, 1961; Paul Milgrom
and Robert J. Weber, 1982). Constructivist models need notask why or how
an auction institutionarose or what were the ecologicalconditions that
createdit, or why there are so many distinctauction institutions. In some
cases it is the other way around. Thus, revenue equivalence theorems show
that the standard auctions generate identical expected outcomes, which, if
taken literally, leave no modeled economicreason for choosingbetween them.
More generally, using rational theory, one represents an observed
socioeconomic situation withan abstract interactive game tree.
Contrarily, the ecological concept of rationality asks from whence came
the structure capturedby the tree? Why this social practice, from which we
can abstract a particular game, andnot another? Were there other practices
and associated game trees that lacked survival properties and were
successfully invaded by what we observe? There is a sense in which
ecological systems, whether cultural or biological, must necessarily be,
or are in the process of becoming, rational: they serve the .tness needs
of those who unintentionally created them through their interactions.
Constructivist mental models are based on assumptions about behavior,
structure, and the valueknowledge environment. These assumptions might
be correct, incorrect, or irrelevant, and the models may or may not lead
to rational action in the sense of serving well the needs ofthose to whom
the models apply. As theorists the professional charge for which we are
paid is to formulate and prove theorems. A theorem is a mapping from
assumptions into testable or observable implications. The demands
oftractabilityloom large in this exercise, and to get anything much in the
way of results it is necessary to consider both the assumptions andtheir
implications as variables. Few game theorists, buildingon the assumption
thatagents always choose dominant strategies, believed this to
characterize the behavior ofall agents in all situations. Hence, the
nearuniversal justi. cation oftheory as an exercise in "understanding."
But the temptation is to believe that our "castles in the sky" (as W.
Brock would say) have directmeaningin our worlds ofexperience andproceedto
impose them where itmaynotbe ecologically rational to do so.
To understandwhat is--the tip of the knowledge iceberg--requires
understandingofa great deal that is not. In the laboratory we can not
onlyrationally reconstruct counterfactuals, as in economic history, but
also use experiments to test and examine their properties. Let us lookat
two contemporary examples.
1. Deregulating Airline Routes.--Airline route deregulation brought an
unanticipated reorganizationofthe network, calledthe hubandspoke
system. (See, e.g., George Donahue, 2002). This is an
ecologicallyrationalresponse, apparently anticipated bynone of the
constructivist arguments for deregulation, and predicted byno one. Nor
could it have been uncovered, I submit, in 1978by surveys of airline
managers, or by marketing surveys of airline customers. Unknown to both
managers and customers was the subsequently revealed decision preferences
of customers who unknowingly favored frequencyof daily departure
andarrival times--a preference that had to be discovered through market
experimentation. Nonstop service between secondary cities was simply not
sustainable in a deregulatedworldof free choice. The onlyway to achieve
ef. ciency, both the demand for frequency of service and pro.table load
factors, among secondary cities was for the .ights to connect through
hubs. Hence, the hypothesis that a rational ecologicalequilibrium emerged
to dominate repeated constructivist attempts, by business entrants and
startups, to satisfy an incompatible set of constraints provided by the
microstructure of demand, pro.tability, and technology.
Might it have been otherwise if airport runway rights, or "slots," had
been an integral part of the deregulation of airline routes, and the
timeofday spotpricing ofslots had emerged to re.ect hub congestion
costs? (Stephen Rassenti et al., 1982). We do not know, but the effectof
this hypotheticalcounterfactualon the viability of hub bypass could be
assessed in laboratory experiments. As in allstudies of what is not, the
challenge is to estimate the parameters that would implement the
appropriate economic environment.
2. The California Energy Crisis.--Asecond, and very troubling, example is
the circumstances leadingto the California energycrisis. As in other
regions of the country and the world, deregulation was effected as a
planned transition with numerous political compromises. In California it
took the form of deregulating wholesale markets and prices while
continuing to regulate retail prices at .xed hourly rates over the daily
and seasonalcycles in consumption. The utilities negotiated an increase
in these average retail rates to meet the revenue requirements of capital
investments that were "stranded" (i.e., werebelieved to be unable to
recover their costs under competition). This preoccupationwith the past,
and with average revenue/cost thinkingbyregulators and regulated alike,
illprepared the state for the consequences of having no dynamic
mechanisms for prioritizing the end use consumption of power.
As expected, traditionalvolatilityin the marginal cost of
generatedelectricity was immediatelytranslatedinto volatileintraday
wholesale prices. What was not expected was that a combinationoflow
rainfall(reducingPaci. c Northwest hydroelectric output), growth in
demand, unseasonably hot weather, generators down on normal maintenance
schedules, etc., caused the temporary normal daily peakingof prices to be
greatly accentuated, and to be muchmore lastingthan had occurred earlier
in the Midwestand South. Events of small probability happen at about the
expected frequency, and since there are many such events the unexpectedis
not that unlikely. Constructivist planning failed to provide for retail
competition to experiment with programs allowingconsumers to save money by
enablingtheir lower priorityuses of power to be interrupted in times of
supply stress. Interruptible deliveries are a direct substitute for both
energy supply and energy reserves, and are an essential means of assuring
adequate capacity and reserves that cover all the various supply
contingencies faced bythe industry.
Because of the regulatory mandate that all demand must be served at a .xed
price, the planning did notallow for the earlyintroduction of demand
responsive retail prices and technologies to enable peak consumption to
be reduced. Instead of mechanism design we had .xed retail price "design"
to generate average revenue that was supposed to cover average cost, and
it failed. The regulatory thought process is as follows: the function of
price is to provide revenue, and the function of revenue is to cover cost.
But this is the antithesis of the market function of price. For neither
management nor the regulators was itnaturalto thinkin terms of pro.ting
from selling less power. Yet that was precisely the route by which the
California distributors could have avoided the loss of an estimated $15
billion: every peak kilowatthour not sold at the average retail rate
wouldhave saved upto ten times that amountof energy cost. Static
technology, and the utter fantasy that all load can always be served, was
protected from innovation bythe legally franchised local wires monopoly.
An entrant could not seekto win customers by offering discounts for
switching from peak to offpeak consumption, and, at the entrants
investment risk, installing the required control devices on enduse
appliances. This legacy--long entrenched, and jealously sheltered by local
franchised monopolies after deregulation--gave California dispatchers no
alternative but to trap people in elevators and shut down highend
computer programming facilities at critical times of peak power shortage.
All power delivery systems are vulnerable to a combination of unfavorable
events that will produce short supplies at peak demand. Constructivism
alone, without competitive trialanderror ecologicalexperimentationwith
retail delivery technologies and consumer preferences, cannot design
mechanisms that process all the distributed knowledge that individuals
either possess or will discover, and that is relevant to .nding an ef.
cient mix of both demand and supply responsiveness.24
24This is illustrated bya November 6, 2002 press release by Puget
SoundEnergy. "PSE Proposes to End Pilot TimeofUse Program Ahead of
Schedule: PSEs timeofuse (TOU) program was created in 2000 duringthe
energycrisis and was intended to provide .nancial incentives for
customers to shift some of their electricity consumption to less
expensive, offpeak times of the day. The program was restructured in July
2002to re. ect a calmer energy market.
Economic Systems Design. 25--What can we learn from experiments about how
demand responsiveness could impact energy shortages as in the California
crisis? Rassenti et al. (2003) measure this impact by creating a market in
which a modest and achievable 16 percent of peak retail demand can be
interrupted voluntarily at discount prices by wholesale energy
providers. In the experiments, demand cycles throughfour levels each
"day" and is expressed in the wholesale market with two contrasting
experimental treatments: (1) robot buyers who reveal all demand at the
spot market clearing price; (2) four pro.tmotivated human buyers who are
free to bid strategically in the market to obtain the lowest available
prices. In each case bids to supply power are entered by .ve
pro.tmotivated human suppliers. In the passivedemand treatment prices
average much above the benchmark competitiveequilibrium, and are very
volatile. In the treatment with human buyers, prices approach the
competitive equilibrium, and price volatility becomes miniscule. By
empowering wholesale buyers, in addition to sellers, to bidstrategicallyin
their own interest, even though 84 percent of peak demand is "must
serve," buyers are able to effectivelydiscipline sellers and hold prices
at competitive levels.26
Since that time it has resulted in most participants bills being slightly
higher than on .at rates. ... Reynolds (a PSE spokesman, said) However,
when exploring new territory, you need to be able to recognize when the
program is not working as you had hoped... and begin a rigorous analysis
ofthe program andhow it could be successfully restructured for the future
energy marketplace. " It is because no one knows what will work best that
you have to open retailing up to the .eld experiment called "free entry
and exit." One experimental possibility is a contract that would share the
inherently unknown and unpredictable savings with the customer. When a
total cannot be known in advance, use a proportionalityrule. New Zealands
tradable .sh catch quotas were originally speci.ed in quantities, and
were redesigned as proportions of the changing quantity available (oral
communication with Maurice McTigue, 2002). An advantage of laboratory
experiments is that these kinds of errors are exposed, and corrective
alternatives tested, at very low cost.
25I have never been comfortable with this label because it is reminiscent
of the idea that we can engineer best social arrangements, which the
reader will see is not my interpretation. See footnote 20.
26The most widely agreedupon design failure in the California crisis was
the rule preventing the distribution utilities from engaging in longterm
contracts to supply power (Wilson, 2002, p. 1332). Beware this simplistic
popular explanation:it is a twowrongsmakearightargument: yes, of
course, given that you were going to protect the monopoly power ofthe
distribution utilities to tie the rental of the wires to the metering and
sale of energy at a .xed regulated price, then one way to protect them
temporarily from the consequent wholesale price volatility might be to
encourage longterm contracts at a .xed average delivery cost. But
suppliers willwant higher prices and/or shortterm contracts if they
anticipate shortages--you cannot get blood outofa turnip; long contracts
work to lower cost only to the extent that suppliers are surprised by high
spot prices, but when it comes time to renegotiate expiring contracts they
will not replicate the error. California discovered this when they
intervened to sign longterm contracts, and encountered high prices. This
whole argument turns the design problem on its head. You must (1) remove
the legal power of the local wires monopoly to prevent competing energy
suppliers from contracting with customers to discount offpeak energy,
charge premiums for peak energy, and install the supporting control
devices; (2)let this competition determine the dynamic price structure,
and investment required to implement it; (3) simultaneously, let .nancial
instruments evolve to hedge whatever risk is left over as prices become
less volatile. Financial instruments can hedge price volatility,notload
volatility. Only demandresponsive interruptible loads can relieve supply
stress and provide the demandside reserves that reduce the risk of lost
load. No one mind or collective can anticipate and plan the needed mix of
technologies to enable the market to manage demand. Therefore it is
essential to remove all entry barriers, and allow .rms to experiment
through competition to discover and innovate ef. cient ways of organizing
retail delivery systems. Claims that shortrun retail demand is
"notoriously" inelastic miss the point: how would you know if
loansheddingtechnologyis in.exible? Competition and incentives to
innovate have never been part ofthe structure.
This example illustrates the use ofthe laboratory in economic systems
design. In these exercises we can testbed alternative market auction
rules andthe effect oftransmission constraintson generator supply
behavior(Steven R. Backerman et al., 2001), vary the degree of
marketconcentration, or "power" in a nonconvex environment (Michael J.
Denton et al., 2001), compare the effectof more or less strategic demand
responsiveness (Rassenti et al., 2003), study network and multiple market
effects also in a nonconvex environment (Mark Olson et al., 2003), and
testbed markets to inform, but not .nalize, market liberalization policy
(Rassenti et al., 2002). For a survey of many examples, see McCabe et al.
The two types of rational order are bothexpressedin the
experimentalmethodology developedfor economic systems design. This branch
of experimentaleconomics uses the labas a test bed to examine the
performance of proposed new institutions, and modi. es their rules and
implementation features in the lightofthe test results. The proposed
designs are initially constructivist, although most applications, such as
the design ofelectricitymarkets or auctions for spectrum licenses, are far
too complicated for formalanalysis (Jeffrey Banks et al., 2003;Rassenti
et al., 2003).
But when a design is modi.ed in the lightof test results, the modi.cations
tested, modi.ed again, retested, and so on, one is using the laboratory to
effect an evolutionary adaptation as in the ecological concept ofa
rationalorder. If the .nalresult is implemented in the .eld, it certainly
undergoes further evolutionary change in the light ofpractice,
andofoperational forces not tested in the experiments because they were
unknown, or beyondcurrentlaboratory technology.27 In fact this
evolutionary process is essential if institutions, as dynamic social
tools, are to be adaptive and responsive to changing conditions. How can
such .exibility be made part of their design? We do not know because no
one can foresee what changes will be needed.
Market Institutionsand Performance.--Noncooperative or CournotNash
competitive equilibrium (CE)theory has conventionallyoffered two speci.
cations concerningthe preconditions for achievinga CE:(1)agents require
complete, or "perfect," information on the equations de.ning the CE; also
common knowledge--all must know that all know that all know that they have
this information. In this way all agents have common expectations of a CE
and their behavior must necessarily produce it; (2) another tradition,
popularly articulated in textbooks, and showing, perhaps, more
sensitivity for plausibility,has argued for a weaker requirementthat
agents needonly be pricetakers in the market.
27People often ask, What are the limits of laboratory investigation? I
think any attempt to de. ne such limits is very likely to be bridged by
the subsequent ingenuity and creativity (the primary barriers at any one
time) of some experimentalist. Twenty. ve years ago I could not have
imagined being able to do the kinds of experiments that today have become
routine in our laboratories. Experimentalists also include many of us who
see no clear border separating the lab and the .eld.
The alleged"requirement" ofcomplete, common, or perfect information is
vacuous: I know of no predictive theorem stating that when agents have
such information their behavior produces a CE, andin its absence their
behavior fails to produce a CE. If such a theorem existed, it could help
us to design the experiments that could test these dichotomous
predictions. I suggest that the idea that agents need complete
information is derived from introspective error: as theorists we need
complete information to calculate the CE. But this is not a theory of how
information or its absence causes agent behavior to yield or not a CE. It
is simplyan unmotivated statement declaring, without evidence, that every
agentis a constructivist in exactlythe same sense as are we as theorists.
Andthe claim that it is "as if" agents had complete information, helps
not a wit to understand the wellsprings ofbehavior. What is missing are
models of the process whereby agents go from their initial circumstances,
and dispersed information, using the algorithms ofthe institution to
update their status, andconverge (or not) to the predictedequilibrium.28
28The inherent dif. culty in equilibrium modeling of the CDA is shown by
the fact that so few have even attempted. Wilson (1987),
characteristically, has had the courage and competence to log progress.
Daniel Friedman (1984) uses an unconventional nocongestion assumption to
.nesse NashCournot analysis, concluding ef. ciency and a .nal competitive
clearing price. Wilson (1987) uses standard assumptions of what is common
knowledge--number of buyers (sellers), each with inelastic
demand(supply)for one unit, preferences linear in payoffs, no risk
aversion or wealth effects, valuations jointly distributed, and agent
capacity to "compute equilibrium strategies and select one equilibrium in
a waythat is common knowledge" (p. 411). This is an abstract
asifallagentsweregametheorists constructivist modelofa thought
process that no game theorist would or does use when participatingin a
CDA. The model itself generates its own problems, such as degeneracy in
the endgame when there is only one buyer and seller left who can feasibly
trade--a problem that is not a problem for the subjects, who do not know
this, and see imperfectly informed buyers and sellers still attempting to
trade and thereby disciplining price. Extra marginal traders provide
opportunity cost endgame constraints on price. Agents need have no
understandingof opportunity cost in order for their behavior to be shaped
by it. Wilson recognizes these considerations: "The crucial de. ciencies,
however, are inescapable consequences of the gametheoretic formulation"
(Wilson, 1987, p. 411). We are squarely up against the
limitations--perhaps the deadend ultimate consequences-- of Cartesian
constructivism. We have not a clue, any more than the socalled "naive"
subjects in experiments, how it is that our brains so effortlessly solve
the equilibration prob
As a theory the pricetakingparable is also a nonstarter: who makes price
if all agents take price as given?If it is the Walrasian auctioneer, why
have such processes been found to be so inef. cient? (Corrine Bronfman et
Hundreds ofexperiments in the past 40 years (Smith, 1962, 1982; Douglas D.
Davis and Charles A. Holt, 1993, 1995;John H. Kageland Roth, 1995;Plott,
2001)demonstrate that complete information is not necessary for a CE to
form out of a selfordering interaction between agent behavior and the
rules of information exchange and contract in a variety of different
institutions,but most prominentlyin the continuous bid/ask double auction
(CDA).29 That complete information also may not be suf. cient for a CEis
suggested(the samples are small)by comparisons showing that convergence is
slowed or fails under complete information in certain environments (Smith,
An interesting contribution by Dhananjay K. Gode and Shyam Sunder
(hereafter GS; see Shyam Sunder, 2003, and the references itcontains) is
to demonstrate that an important component of the emergentorder observed
in these marketexperiments derives from the institution, not merely the
presumedrationalityof the individuals. Ef. ciency is
necessarilyajointproduct ofthe rules ofthe institutionandthe behaviorof
agents. What Sunder and his coauthors have shown is that in the
doubleauctionmarket for a single commodity (we know not yet how far it
can be generalized),ef. ciency is higheven with "zero" intelligence robot
agents, each of whom chooses bids (asks)completely at random from all
those thatwillnotimpose a loss on the agent. Thus, agents who are
notrational constructivist pro.t maximizers, and use no learning or
updating algorithms, achieve most of the possible social gains from trade
using this institution. Does this example illustrate in a small way those
"superindividualstructures within which individuals found great
opportunities ... (and that)... could take account of more factual
circumstances than individualscould perceive,and in consequence... is in
some respects superior to, or 'wiser' than, human reason ... "? (Hayek,
1988, pp. 77, 75). We do not know if the GS results generalize to multiple
market settings as discussed in the next paragraph. Ross M. Miller (2002),
however, has shown that in a very elementarytwomarket
environment--intertemporally separated markets for the same
commodity--the GS results are quali. ed. Complex price dynamics,
including "bubbles," appear, and there is loss of ef.ciency, although
the loss is not substantial. On average the decline is apparently from
around 94 percent to 88 percent.
lem in interacting with other brains though the CDA (and other)
institutions. We model not the rightworld to capture this important
experimental .nding. 29See Jon Ketcham et al. (1984) for a comparison of
CDA with the posted offer (PO)retail pricing mechanism. CDA converges more
rapidly andis more ef. cient than PO. So why does not CDA invade and
displace PO? It is the high cost of training every retail clerk to be an
effective negotiator for the .rm. Institutions re.ect the .ne structure of
opportunitycost, and the loss of exchange ef.ciency in PO is more than
offset by the distributional productive ef. ciency of the mass
retailinginnovation of the 1880s that led price policy to be centralized.
As I write, those policies are being modi. ed on the Internet where prices
can be adjusted to the opportunity cost characteristics of buyers, such as
how many other Internet sites they have visited (Cary Deck andBart Wilson,
2002).Institutional changes in response to innovations like mass retailing
are part of the emergence of an ecologically rational equilibrium.
In multiple market trading in nonlinear interdependent
demandenvironments, each individuals maximum willingnesstopay for a
unitof commodity A depends on the price of B, and vice versa, and in this
more complex economy double auction, markets also converge to the vector
of CE prices and trading volumes. A twocommodity example is reported in
Smith (1986), based on nonlinear demand (CES payoff function)and linear
supply functions found in Arlington Williams and Smith (1986); also see
Williams et al. (2000). In these experiments, numerical tables based on
the preference and cost information de.ning the generalequilibrium
solution of four nonlinear equations in two prices andtwo quantities are
dispersedamong the undergraduate subjects. They buy and sell units
ofeachof the two commodities in a series of trading periods. Prices and
trading volume converge, after several trading periods, to the CE de. ned
by the nonlinear equations. The subjects would not have a clue as to how
to solve the equations mathematically. The experimenter applies the tools
of constructivist reason to solve for the benchmark CE, butin repeat play
this "solution" emerges from the spontaneous order created by the
subjects trading under the rules ofthe doubleauction market institution.
Numerous other experiments with many simultaneousinterdependentmarkets
show similar patterns ofconvergence (Plott, 1988, 2001).
The Iowa Electronic Market.--What evidence do we have that the laboratory
ef.ciency properties ofcontinuous double auction trading apply also in the
.eld? One of the best sources of evidence, I believe, is found in the Iowa
Electronic Market (IEM) used widely around the world (Robert Forsythe et
al., 1992, 1999). These markets are used to study the ef. cacy of futures
markets in aggregatingwidelydispersed information on the outcomes of
political elections, or any wellde. ned extralaboratory event, such as
a change in the discount rate by the Fed. The "laboratory" is the
Internet. The "subjects" are all who log on and buy an initial portfolio
of claims on the .nalevent outcomes; they consist of whomever logs in, and
are not anykind ofrepresentativeor"scienti.c" sample as in the polls with
which they are paired. The institution is the openbook double auction.
In the IEM, traders make a market in shares representing pairmutuel
claims on the popular vote (or winnertakeall) outcome of an election,
referendum, etc. For example, the .rst IEM was on the 1988 presidential
election. Each person wanting to trade shares deposits a minimum sum, $35,
with the IEM and receives a trading account containing $10 cash for
buyingadditionalshares, and ten elemental portfolios at $2.50 each,
consisting of one share of each of the candidates--Bush, Dukakis, Jackson,
and"restof. eld." Tradingoccurs continuously in an openbook bidask
market for several months, and everyone knows that the market will be
called (tradingsuspended) in November on election day, when the
dividendpaidon each share is equal to the candidates fraction of the
popular vote times $2.50. Hence if the .nal two candidates and all others
receive popular vote shares (53.2 percent, 45.4 percent, 1.4 percent),
these proportions (times $2.50) represent the payoff to a trader for each
share held. Consequently, at any time t, normalizing on $1, the price of
a share(4$2.50) re. ects the market expectationofthat candidates share
ofthe total vote. A price, $0.43, means the market predicts that the
candidate will poll 43 percent of the vote. Other forms of contract that
can be traded in some IEMs include winnertakeall, or number of seats in
the House, and so on.
The IEM data set includes 49 markets, 41 worldwide elections, and 13
countries. Several results standout:the closingmarket prices, produced
bya nonrepresentativesample of traders, show lower average absolute
forecasting error (1.5 percent) than the representative exit poll samples
(1.9 percent); in the subset of 16 national elections, the market
outperforms the polls in 9of 15 cases; in the course ofseveral months
preceding the election outcome, the market predictions are consistently
much less volatile than the polls; generally, larger and more active
markets predict better than smaller, thinner markets; surveys ofthe market
traders show that their share holdings are biased in favor of the
candidates they themselves prefer.
In view of this last result why do markets outperform the polls? Forsythe
et al. (1992) argue that its their marginal trader hypothesis. Those who
are active in price "setting," that is, in entering limit bids or asks,
are found to be less subject to this bias, than those traders accepting
(selling and buying "at market") the limit bids and asks. Polls record
unmotivated, representative, average opinion. Markets record motivated
marginal opinion that cannot be described as "representative." This
analysis helps to provide a good mechanical, if notultimate, understanding
of how human interaction with the rules of a bid/ask CDA yield ef. cient
30 Other markets besides the IEM are known to have ef. cient
informationaggregating properties. Parimutuel racetrack markets are an
example where, interestingly, the environmentis much like the IEM: the
settlements occur at a wellde. ned endstate known to all agents, unlike
stock market trading where expectations .oat continuously with no clear
value revelation endpoint. "The racetrack betting market is surprisingly
ef. cient. Market odds are remarkably good estimates ofwinning
probabilities. This implies (sic) that racetrack bettors have considerable
expertise, and that the markets shouldbe taken seriously" (Thaler
andWilliam T. Ziemba, 1988, p. 169). It is surprising to behavioral
economists because their methodologyis restricted to looking for
deviations from the standard model. What is unusual here is that in
racetracks they have found reportable evidence for ef. cient outcomes.For
those who follow the experimental economics, IEM and similar
controlledenvironmentmarket studies, ef. ciency is notonly commonplace
(if notuniversal), it cannotbe attributed to agents with "considerable
expertise." The agents are mostly naive, although they get repeat
interaction experience, which, from the evidence, clearly gives them
expertise enough. But, as in the IEM and experimental markets, racetrack
markets are not perfect: there are inef. ciencies in the "place" and
"show" options and the favoritelongshot bias, with the
We have seen that markets economize on information, understanding, the
number of agents, and individualrationality. Can they also economize on
the need for externalintervention to protect particular interests, ifall
are empowered by the trading institution to act in their
B. Strategyproofness: Theory and Behavior
Preferences are private and unobservable, and institutions have to rely on
the messages reported by agents, not their true preferences. This follows
from the fact that no one mindhas all the information known together by
all those in the market. It is therefore possible for an agent to affect
prices andoutcomes in a market by strategically misreporting preferences.
This prospect has motivated the literature seeking strategyproof
mechanisms: "An allocation mechanism is strategyproof if every agents
utilitymaximizing choice of what preferences to report depends only on
his own preferences andnot on his expectationsconcerning the preferences
that other agents will report" (Mark Satterthwaite, 1987, p. 519). This
requires each agent to have a dominantstrategyto report true preferences,
and has led to impossibility theorems establishing the nonexistence of
such a mechanism under certain conditions.
In view of such negative theoretical results and the narrow conditions
under which solutions have been investigated, it is important to ask what
people actually do in experimental environments in which the experimenter
induces preferences privately on individual subjects. We know what is
impossible, but what is possible in more openended systems than are
modeled by theory? Is it possible that when all are free to choose from a
large space of strategies, ecologically rational strategies will emerge
that immunize against strategic manipulation? Given that information is
inherently latter more pronounced in the last two races of the day.
Various hypotheses have been offered to explain these inef. ciencies,but
more signi.cantis that computer programs have been written to arbitrage
(yielding returns of some 11 percent per bet), the place, show, and
longshot inef. ciencies. (It is my understandingthat goodpro.ts have been
accumulated on these programs, so far without neutralizing the arbitrage
opportunities--let the good times roll!) dispersed, has society evolved
institutions in which forms of behavior arise that result in practical
ifnotuniversal solutions to the problem ofstrategyproofness?
The double auction is a wellknown example yielding CE in a wide range of
economic environments including small numbers. Are there other examples,
and if there are, what are the strategic behavioral mechanisms that people
adopt to achieve strategyproofness?
One example is the sealedbidoffer auction: in each contract period the
submitted bids are orderedfrom highest to lowest, the offers (asks) from
lowest to highest, with the intersection (cross) determining the uniform
clearing price andvolume exchanged (see TimothyCason and Friedman, 1993;
Friedman, 1993; and Wilson, 1993). Also see Smith etal. (1982)for
comparative studies of different versions of the sealedbidoffer
mechanism andthe continuousdouble auction.
In experiments with stationary supply and demand, initiallybothbuyers
andsellers greatly underreveal their true individual willingness to buy or
sell. Volume is very low (10-15 percent ofoptimal), the market is inef.
cient andeach agent can see that at the initial clearing price they are
leaving money on the table. In repeat play they increase revelation, but
mostly of units near the last periods clearing price. As volume increases
and the clearing price closes in on the CE, the realized inverse demandand
supply become very .at near the true clearing price with many tied or
nearly tied bids and asks that exceed the capacity of any single buyer or
seller. At this steady state, and given this behavior, if anyone withholds
purchases or sales she is denied an allocation as other more competitively
traded units substitute for hers. This results in a "behavioral
strategyproof equilibrium." Such is the power of motivated, privately
informed agents in trialanderror repeat interaction.
These experimentalresults make it plain that the theoretical condition for
a strategyproof equilibrium--that each agent have a dominate strategy to
reveal true willingnesstopay or willingnesstoaccept for allunits,
andnotjust units near the margin--is much too strong. The above
description from blind twosided auctions, however, also shows that there
is a social cost to the achievement ofa strategyproofequilibrium: blind
twosided auctions converge more slowlyto the competitiveequilibriumthan
continuous double auctions, andupon converging, may not be quite as ef.
cient if agents occasionally attempt manipulation, are disciplined, and
return to the full exchange volume.
A second example is the uniform price double auction (UPDA), a realtime
continuous feedback mechanism clearing all trades at a single price in
each trading period. This is a "designer market" inventedby
experimentalists who asked, "Can we combine the continuous information
feedback advantages of the double auction with the uniform price (zero
withinperiod volatility)advantages of the sealedbidoffer auction?" As
we have seen above, with blind bidding several repeat interactions are
required to reach optimality, with many lost trades in the process. Can
we accelerate the price discoveryprocess bycontinuouslyfeeding back
information on the tentative state of the market, and allowing bids (asks)
to be adjusted within each period?
This institution is made possible by highspeed computer and communication
technology. It comes in several .avors, or variationson the rules. In
allversions ateach time, t # T 5 time market is "called" (closed), the
tentative clearing price, pt, is displayed and each agent knows the
acceptance state of all her bids (asks). This allows bids andasks to be
adjusted in realtime. See the chapter by McCabe et al. (1993, pp. 311-16)
for a report of 49 UPDA experiments comparing these different versions
with the continuous double auction. UPDA exhibits even more
underrevelation of demand and supply than the blind twosided auction
discussed above, but ef. ciency tends to be much higher, especially in the
.rst periods, and, in one form (endogenous close, open book, the "other
side" rule with conditional time priority), exceeds that of the continuous
Experiments using UPDA in a randomly .uctuating supply and demand
environment routinelyexhibit ef. ciencies of95-100percent, sometimes
withas little as 5-10 percentof the available surplus revealed. This is
shown in Table 1 for summary data from UPDA experimentup43. Most agents
enter bids (asks) equal to or near the clearing price as it is
continuously displayed in real time. It is ofcourse true, hypothetically,
that if all agents reveal their true demand or supply with the exception
of one intramarginal buyer or seller, then that agent
TABLE1--SUMMARY OFRESULTS: UPDA EXPERIMENT UP 43; 5, 5
Notes: (Pe, Qe) 5 equilibrium price and quantity. (Pr, Qr) 5 realized
price and quantity. Eff% 5 ef. ciency, % max surplus. Rev% 5 %of surplus
revealed. can manipulate the price to his or her advantage. But this
parable is irrelevant. The relevantquestion is what behavior is manifest
when every agent has the potential for manipulating the price. Without
knowledge or understandingof the whole, and without design or intention,
the participants use the rules at their disposal to achieve three
properties observed by the experimenter: (1)highef. ciency, (2) maximum
individual pro.t given the behavior of all other agents, and(3)protection
from manipulation by their protagonists.31This ecologicalresult
illustrates the perceptive insight ofHayek (1988, pp. 19-20). "Rules
alone can unite an extended
31Space prevents me from dealing fully with the many importantissues
raised when a subset ofagents have asymmetric advance information on
product quality or value characteristics. The analysis shows that such
conditions generate market failure or inef. ciency. Some of these
problems, however, arise because the analysis is inadequate in examining
both sides of the market, and the implications of the information content
of prices. Experiments have establishedthat constructivist inef. ciency
is often alleviated by one of several ecologically rational response
mechanisms: competition among sellers for reputations, quality (brand)
signaling,product warranties, andthe aggregationofprivate asymmetric
information into public price patterns that selfcorrect the alleged
problems. See, e.g., Plott and Louis Wilde (1982); Plott and Sunder (1982,
1988); Miller and Plott (1985); Camerer and Keith Weigelt (1988).
order ... . Neither all ends pursued, nor all means used, are known or
need be known to anybody, in order for them to be taken account of within
a spontaneous order. Such an order forms of itself... ."
C. Greshams Law: If It Isnt CournotNash, Why Is It a Law?
In this section Ihave given manyexamples of institutions in whichthe CE
theoryof markets predicts their observed behavior. Do we have contrary
examples? Yes. Greshams Law: bad money drives out good. This "law," while
sometimes claimed to be an observed phenomena in countries all over the
world, is not a CournotNash equilibrium.32 If currencies A and B are both
available, A having an intrinsic worth while B is worthless .at money,
then the theory predicts that A will drive out B. This is because each
agent believes other agents are rational, and will accept only A in
exchange. Each agent will therefore avoid getting stuck with the inferior
B by accepting only A, which becomes the dominant circulating medium of
exchange, while B is "horded." Experiments have con.rmed that if both
types of money are initially available, subjects use onlythe superior
currency (an interestbearing consol) as a medium of exchange. But in
treatments in which subjects .rst experience a history ofusing .at money,
it being the only medium of exchange available, and then the consol is
introduced, subjects continue trading withthe .at money, hording the
interestbearing consol (Gabriele Camera etal., 2003). This is entirely
rational if each agent believes others will accept the .at money in
exchange and this belief is supported by experience. Think of Greshams
Law as a belief equilibrium in which theory alone is unable to predict
when it might occur (Ledyard, 1986).
Complementingthese results, another experimentalstudyshows that when .at
money is the only currency, it will be used even under the 32 Hayek (1967,
p. 318) notes that Greshams Law is not due to Gresham nor is it a "law"
in the theoretical sense, and "... as a mere empirical rule is practically
worthless." In the 1920s when people started using dollars and other hard
currencies in substitution for the depreciating mark, the claim emerged
that Greshams Law was wrong--that it was the other way around.
conditionthat it is abandonedand replaced with a new .at money issue at
the end of a .nite horizon. In this studythe realeconomy is found to
suffer some loss in ef. ciency relative to the use of "backed" (commodity)
money, but the economy does not collapse even in short horizon
treatments. Collapse in real sector ef.ciency is observed only when a
"government" sector prints .at money to purchase real goods from the
private sector. Moreover, additional experimental tests show thatthe
collapse cannot be due to the resulting in.ation, but to interference
with the real price discovery of markets when some agents are able to
crowd out private real purchases with printing press money33 (Deck et al.,
D. Psychology and Markets
Psychologists and "behavioral economists" who study decision behavior
almost uniformly report results contrary to rational theory (Robin Hogarth
and Melvin Reder, 1987). It was not always so,34but the focus on
"anomalies," beginning in the 1970s, converted the emerging discovery
enterprise into a search for contradictions between reports of behavior
andthe caricatures35 of mainstream theory that constitute much of its
core. Psychologists, to their credit,
33This is demonstrated by comparison experiments in which there are no
government agents, but .at money is in. ated each period by the average
rate that is observed in those experiments with government agents present.
34 "Prior to 1970 or so, most researchers in judgment and decisionmaking
believedthatpeople are pretty good decisionmakers. ... Since then,
however, opinionhas taken a decided turn for the worse, though the decline
was not in any sense demanded by experimental results. Subjects did not
suddenly become any less adept at experimental tasks nor did
experimentalists begin to grade their performance against a tougher
standard. Instead, researchers began selectively to emphasize some results
at the expense of others." "The view that people are irrational is real in
the sense that people hold it to be true. But the reality is mostly in the
rhetoric" (Lola Lopes, 1991, pp. 66, 80).
35I say "caricatures" because economics has long offered much in the way
of theoretical exceptions to the core neoclassical model of
selfinterested market competition: externalities in choice, public
goodeffects, and "anomalies" in choice under uncertainty
requiringexplanation (Friedman and L. J. Savage, 1948; Harry Markowitz,
1952). But it is the neoclassical assumption of selfinterested agents
that hasbeen the most productive of theoretical results and therefore is a
prominent and easy target of criticism.
have maintained an intensive program examiningthe behavioralnature
ofthese contradictions to the classical model. For example, Sidney Siegel
(1959) and Lawrence E. Fouraker and Siegel (1963) reported both
con.rmations and contradictions, and used the pattern to propose improved
models. Similarly, in prospect theory Kahneman and Tversky (1979) have
proposed modi.cations in both the utility and probability
weightingfunctions of standard expected utility theory.36 Research
strategies that focus on the study of errors, however, can distort
professional beliefs, to say nothing of popular representations, if the
primary emphasis is on the failures, to the exclusion ofthe predictive
successes, ofthe theory.37
E. Psychology, Economics, and the Two Kinds of Rationality
Curiously, the image ofeconomists and psychologists as protagonists
obscures their underlyingagreement on foundations. Bothrely upon
constructivism:(1)to the extentthatmarkets are
36Their most important contributions in prospect theory were in empirical
tests demonstrating the relevance of two ideas suggested originally by
Markowitz (1952): the idea that the theory applies to changes in wealth
(income) relative to the individuals current asset state, andthat people
are risk preferring in losses and risk averse in gains. This much is
consistent with standard expected utility theory, which requires onlythat
the prizes ofchoice can be ordered, and therefore applies either to wealth
or income. Which prizes the theory is best applied to has always seemed to
me to be inherently a subject for empirical determination. If applied to
wealth, the theory starts to infringe on preference theory over time, long
recognized as especially dif. cult modeling terrain.
37As Isee it experimental market economics and behavioral economics are
in principle complementary. Experimental economists study market
performance (rationality) given individual valuations, while cognitive
psychologists study the valuations (rationality)of individuals. If the
objects traded are prospects the appropriate valuations are their "cash
values," whether based on expected utility, prospect theory (Kahneman and
Tversky, 1979), or some other representation. Thus Plottand Jonathan T.
Uhl (1981)study experimental markets in which the items traded are
gambles, and reportconvergence to a CEde. ned by demand and supplybased
on the expected values ofthe gambles. But the connective interface between
rationality at the individual and at the market level and how institutions
modulate the interface has not been well explored. Markets do their thing
with whatever are the values--rational, irrational, or
nonrational--thatare provided by individuals. rational38 or irrational,39
this derives directly andonly from the rationality or irrationality of
agents;40 (2) individual rationality is a selfaware, calculating process
(3) predominantly both are reluctant to allow that naive, unsophisticated
agents can achieve socially optimalends withoutcomprehensionof the whole,
as well as their individual parts, implemented by deliberate action (there
is no "magic," and no room for the GS zero intelligence traders); (4)
consequently, psychologists test the rationality of individual decisions
largely byaskingfor subject responsesto choice problems to discover how
they "reason." Rather than challenge this constructivist view,
economists, subject to the identical vision (how do agents consciously
think?), are critical of the questionresponse survey methods used in
cognitive psychology: the stakes are zero or too low,42and the subjects
are too unsophisticated, inexperienced, or untrained to allow a serious
researcher to .nd out how "real agents really think." Many psychologists
appear to .nd irrationality everywhere, and many economists appear to see
the .ndings as everywhere irrelevant. To these economists, how agents
think indeed exhausts the core of empirical econom
38For example, the doubleauction markets discussed above.
39Experimental asset markets bubble and crash on the long path
ofexperience to equilibrium (Smith et al., 1988; Porter and Smith, 1994).
For a new study of subject experience and asset bubbles see Martin
Dufwenberg et al. (2003).
40 Thus, even a "... monopolist ... has to havea fullgeneralequilibrium
model of the economy" (Kenneth Arrow, 1987, p. 207). Also see footnote 30
above on racetrack market ef. ciency, and the inference that the bettors
must therefore have considerable expertise. Thus, market rationality is
automatically assumed to derive entirely from individual rationality.
41 Here is a particularly clear statement of decision as rational
constructivist action: "Incentives do not operate by magic: they work by
focusing attention and by prolonged deliberation" (Tversky and Kahneman,
1987, p. 90).
42The use of cash or other reward medium in decision behavior experiments
is listed by Ralph Hertwig and Andreas Ortmann (2001)as one ofthe key
differences between psychology and economics experiments. The controversy
over paying subjects, however, is rapidly being eroded as cognitive
psychologists and experimental economists join with
neurobiologists--including those who are informed on animal behavior
models--and subjects are paid salient rewards (Gregor Thut et al., 1997;
Hans C. Breiter et al., 2001;McCabe et al., 2001). ics; psychologists
merely "fail" to properly implement their investigation of this core.43
In point of fact, opinion surveys can provide important insights:
sometimes survey .ndings can be tested more rigorously with
rewardmotivated choices in the laboratory or the .eld and are found to
have predictive content (e.g., the asymmetry between losses and gains in
wealth). Sometimes what people actually do completely contradicts what
they say, and sometimesyou cannot .nd out by asking because the agents
themselves do not know what they will do or are doing. For example:
Comparisons of riskpreferences under low and high monetary stakes have
shown that actual reward levels have a statistically signi.cant effect on
decision, but that the qualitative conclusions from hypothetical choice
response surveys are not refuted by studies using very high stakes--the
accumulated payoffs average three times subjects normal monthly living
expenses (Steven J. Kachelmeier and Mohamed Shehata, 1992; also see Hans
P. Binswanger, 1980, for similar .ndings).
Consider the double auction in classroom demonstration experiments:in
debrie. ngs afterwards students deny that there is any kind of
quantitative model that could predict their market price and exchange
volume, or that they were able to maximize their pro.ts; but a participant
with an envelope containing the predictions provided in advance, opens it
showing that this consensus is false. The dispersed private value/cost
information is aggregated into prices thatare at the equilibrium andeach
agentis maximizinghis or her pro.t given the behavior ofall others. Here
there is indeed a kindof"magic," but only, I think, in not being well
understood or modeled at the gametheoretic level of individual choice.44
43 Kahneman clearly does not see people as irrational except in the narrow
context used in economic modeling based on dominant choice. In fact he
describes his empirical .ndings contradicting the SSSM as having been
easy, thanks to the implausibility to any psychologist of the SSSM. See
the Nobel Foundation interview of the 2002 Nobel Laureates in economics at
44 At the macromarket level, convergence, and cases of stable and
unstable equilibrium, are well predicted by the classical Walrasian
adjustment model, but the paths taken, including jumps across alternative
unstable equilibria, are
Our bounded rationality as economic theorists is far more constraining on
economic science, than the bounded rationality of privately informed
agents is constraining on their ability to maximize the gains from
exchange. In asset trading, participant survey responses re. ect the
disconnect between their information on fundamental value and their
puzzling experience of a price bubble and crash generatedon the long path
to the rational expectations equilibrium (T. Schwartz and J. S. Ang,
1989). Opinion polls administered to the IEM traders show the same
judgment biases that psychologists and political scientists .nd in public
opinion polls, but these biases did not interfere with the markets
ability to predict the popular vote outcomes (Forsythe et al., 1992). In
preference reversal survey experiments subjects report many inconsistent
choices: gamble A is preferredto B buta subject will sell A for less than
B. Arbitraging the subjects cashmotivated choices quicklyreduces these
inconsistencies (Yun Peng Chu and RueyLingChu, 1990, p. 906), andit has
been shown that the inconsistencies are unbiased random errors under some,
but not all, conditions (James C. Cox and DavidM. Grether, 1996); also
see Barry Soper and Gary Gigiolotti, 1993, where choice intransitivity is
studied directly and the errors are found to be random. Kahneman et al.
(1986; hereafter KKT) provide many examples in which respondents are
asked to rate the fairness,45 on a four
not well predicted by the model. See the outstanding summary by Plott
(2001). The disconnect with choice behavior is evident in the following:
Walrasian dynamics makes ad hoc assumptions about price adjustments in
response to excess demandsayingnothing about the correspondingpayoff
motivation of the agents who drive the price changes. Walrasian dynamics
is a story about the tatonnement mechanism in which there are no
disequilibrium trades, whereas Plotts (2001)summary is about continuous
doubleauction trading witha great many disequilibrium trades.
45The descriptor "fairness" has so many meanings in different contexts
that I believe it is best to avoid the term entirelyin experimental
science except where it is explicitly modeled and the model tested in
environments where subjects make decisions on the basis of the de.
ningparameters of the model; then the descriptor "fair" and its ambiguity
can be avoided altogether. This is the way it is used in the utilitarian
de. nitions by Robert Franciosi et al. (1995), in Ernst Fehr and Klaus M.
Schmidt (1999), and Bolton and Axel Ockenfels (2000). Of course it is
appropriate to use the point scale, of elementary business actions in
competitive environments. In one case a hardware store raises the price
ofsnow shovels from $15 to $20 after a snowstorm. Eightytwo percent of
the respondents consider this action either unfair or very unfair.
Franciosi et al. (1995, pp. 939- 40)substitute the words "acceptable" for
"fair" and "unacceptable" for "unfair"46and add one additional sentence
to this KKT example: "The store does this to prevent a stock out for its
regular customers since another store has raised its price to $20." Now
only 32percent rate the action unfavorably. This exercise suggests the
possible sensitivity of survey results to emotive words and/or perceived
"justi. cation" in terms of impersonal market forces.
Note that it is in private information environments, where the market is
aggregating information far beyond the reach of what each individual
knows, and is able to comprehend, that the solicited opinions are so far
off the mark. The surveys yield no useful understanding because the
subjects have none to relate. In the complete information asset market,
subjects are aware of its fundamental value structure, and come to have
common expectations through an experiential process of repetition; i.e.,
initialcommon information is not suf.cient to induce common
expectations.47 They play myopically and their expressed baf.ement
("prices rise without cause") re. ects this myo
descriptor if the purpose is to see how its instructional use might have
an emotive affect on behavior. The emotive content of "fairness" is clear
in the important work of Edward E. Zajac (1995), who has also examined the
rhetoric of fairness arguments as selfinterest serving in the Florida,
2000, election controversy (Zajac, 2002).
46 KKT state that "... the phrase it is fair, is simply an abbreviation
for a substantial majority of the population studied thinks it is fair"
(KKT, 1986,p. 201). But their main interest is in whether .rm behavior is
affected by communitynorms. Whether or not an action is "acceptable"
would seem to be just as important in determining .rm behavior as whether
or not it is "fair." If the two terms map into different attitudes, then
there is inherent ambiguity in specifying the effect on .rm behavior.
47This interpretation is consistent with assettrading experiments using
undergraduates, small businesspersons, corporation managers, and
overthecounter traders (Smith et al., 1988, Porter and Smith, 1994).
Exceptions using inexperienced subjects, to my knowledge, have only been
observed with advanced graduate students (McCabe and Smith, 2000).
pia. These comments suggest that much insight might be obtained from the
systematic study of the conditions under which surveyresults are robustly
informative and the conditions where they are not.
F. Fairness: An Experimental Market Test
In developing a descriptive theory48 of the "reference transaction," KKT
state that what is considered "fair" may change: "Terms of exchange that
are initially seen as unfair may in time acquire the status of a reference
transaction" (KKT, 1991, p. 203). This paves the way for the adaptation
of "fairness beliefs" to changes in the competitive equilibrium.
Althoughthe competitivemodelis the one thathas static predictive content,
its prediction is silent as to how long it will take to respond to a
change in parameters. KKTs arguments are not predictive, but they tella
story about why markets might be sluggish in respondingto change. How
good is their story?
Franciosi et al. (1995) state a preference modelofoptimal choice that
allows for a utilitarian tradeoff between own consumption and
"fairness." For example, the utility of two commodities (x, z)is given
by: u(x, z)5 z 1 ax 2 (b/2)x2 2 ax[(p/p0) 2 1], in which the sellers
pro.t, p, relative to a reference pro. t, p0, appears as an "externality"
in the buyers utility function. The usual maximization subject to an
income constraint yields the inverse demand equation:p 5 a 2 bx 2
a[(p/p0)2 1]. Thus, for a . 0 any change in the environment that increases
a .rms pro.t relative to the reference pro.t has an external effect that
lowers the buyers inverse demand for units x. If a 5 0, then we have the
standard ownmaximizingtheory. Consequently, Franciosi et al. (1995) can
test the hypothesis, never using the word "fairness," that if subject
buyers have a utilitarian concern for pro.ts not being increased relative
to a baseline then after a change from the baseline this should alter the
observed equilibrium relative to the standard predicted equilibrium withno
externaleffect, a 5 0. In a posted offer market giving KKT their best shot
(sellers cannot see each other's posted prices, andtherefore cannot
knowingly emulate or undercut each other's prices), Franciosi etal.
(1995).nd that when a 5 0 (implemented by either no disclosure, or by
marginal costjustifying disclosure) the market converges quickly to the
new competitive equilibrium. When a . 0 (implemented by pro.t p and
p0disclosure)prices converge more slowly, but precisely, to the new
equilibrium. Hence, under conditions most favorable to a "fairness"
effect, the response dynamics is changed,butnotthe equilibriumas predicted
by the standard competitive model. The discipline of the market swamps all
but a transient "fairness" effect. If, realistically,sellers can see each
others prices, I would predict a much smaller "fairness" effect, if any.
48This methodology is driven by the untenable belief that general theories
can be derived directly from observations if youjust have enoughdata (see
Smith, 2002, and the references therein). "Perhaps the most important
lesson learned from these studies is thatthe rules of fairness cannot be
inferred either from conventional economic principles or from intuition
and introspection. In the words of Sherlock Holmes in The Adventure of
the Copper Beaches: Data! Data! Data! I cannot make bricks without clay
" (KKT, 1986, pp. 115-16). Neither can a predictive theory of"fairness"
be inferred from any amount of the KKT data. If N "fairness" rules are
discovered by trial and error modi. cations in the survey questionnaires,
you cannot reject the hypothesis that there is an N 1 1 variation that
will identify a new one. More data will not help, as the fairness concept
is used here as a word that provides no effective means of
modifyingstandard theory to correct for its predictive .aws.
III. Personal Social Exchange
One of the most intriguing discoveries of experimental economics is that
(1) as we have seen, people invariably behave noncooperatively in small
and large group "impersonal" market exchange institutions; (2) many (up to
half in single play; over 90 percent in repeat play) cooperate in
"personal" exchange (twoperson extensiveform games); (3)yet in both
economic environments all interactions are between anonymous players. In
this section Ishall attempt to summarize some of the most compelling
evidence of cooperation in personal exchange--in the .eld as well as the
laboratory-- and review some of the test results designed to discriminate
among the more prominentpredictive hypotheses for modeling cooperative
behavior. Whatever might be the most useful way to model and explain
cooperation, unaided by market incentives, my working hypothesis
throughoutis that it is a product of an unknown mix of cultural and
biological evolution, with the biologyprovidingabstract function de. ning
potential, and culture shaping the emergent forms that we observe. But to
motivate the whole exercise in thought, I will begin with a discussion
ofsome persistent crosscultural social practices from business, law,
anthropology, and American economic history.
How might a social rule (practice, norm) emerge, become a cultural .xture,
and be widely emulated? I will use a parable to illustrate how a rule for
"bargaining in good faith" might become established.
In bargaining over the exchange price between a buyer and seller, suppose
the seller begins byannouncinga selling price, the buyer responds with a
lower buying price, the seller reduces his asking price, and so on. In
this concessionary process it is considered bad form for the buyer (or
seller), once having made a concession, to return to a lower (or higher)
price. This violates a principle of"bargainingin good faith" (see Siegel
and Fouraker, 1960, p. 20). How might this come about? One can suppose
that those who failto bargain in goodfaith would be less likely to be
sought out by others for repeat transactions. Such behavior raises
transactions cost by increasing the time it takes to complete a sale.
Trading pairs would be expected to selfselect, tending to isolate the
more timeconsuming bargainers, and it would take them longer to .ndthose
willingto tolerate the time cost of bargaining. Such practices--
inherently economizing in this parable--might then become part of a
cultural norm, powerful enoughto be codi.edultimately in contractlaw andin
stock exchange rules. Proposition:in this manner collectives discover law
in those rules that persist long enough to become entrenched practices. In
this example the emergent rule reduces transaction cost, leaving open the
classical question of how equilibrium can be characterized in bilateral
A. Spontaneous Order Without the Law49
The early "lawgivers" didnotmake the law they "gave"; they studied social
49Experimental studies have inquired as to whether emergent norms
ofcooperation and constructivist incentive schemes are substitutes, the
latter crowding out the former. See Iris Bohnet et al. (2001) and Fehr and
Simon Ga¨chter (2002) for studies suggesting that they are substitutes
(formal rules undermine informal cooperative norms), and Ser gio G.
Lazzarini et al. (2002) for new results suggesting that they are
complements-- contracts facilitate the selfenforcement of relational
elements beyond contractibility. I would hypothesize that both must be
true: constructivist rules ultimately must pass .tness tests of ecological
rationality. Formal rules that are incompatible with informal rules will
be modi.ed or eliminated; those that are compatible willpersist. Hence,
at any time slice in history, both must necessarilybe observedacross
informal rules andgavevoice to them, as Gods, or natural, law.50 The
common lawyer, Sir Edward Coke, championed seventeenthcentury social
norms as law commandinghigher authority than the king. Remarkably, these
forces prevailed, paving the way for the rule of law in England.51
Similarly, the cattlemens associations, land clubs, and mining districts
in the American West allfashionedtheir own rules for
establishingpropertyrights andenforcingthem: the brandon the hindquarters
ofhis calf was the cattlemens indelible ownership signature on his
property, enforcedbygunmen hiredthrough his cattle club;52squatters
rightswere defended ably (possession is nine points of the law?) by the
land clubs composedof those brave enough to settle wilderness lands in
advance ofveterans exercising their land script claims, and of set
50 "... ([A]ll early (my insertion: as in Sumar with the beginning of
writing) lawgiving consisted in efforts to record and make known a law
that was conceived as unalterably given. A legislator might endeavor to
purge the law of supposed corruptions, or to restore it to its pristine
purity, but it was not thought that he could make new law. ... But if
nobody had the power or intention to change the law ... this does not
meanthat law did notcontinue to develop" (Hayek, 1973, p. 81).
51What allowed the rule of "natural" or found law to prevail in England
"... was the deeply entrenched tradition of a common law that was not
conceived as the product of anyones will but rather as a barrier to all
power, including that of the king--a tradition which Edward Coke was to
defend against King James I and Francis Bacon, and which Matthew Hale at
the end of the seventeenth century masterly restated in opposition to
Thomas Hobbes" (Hayek, 1973, p. 85; also see pp. 167, 173-74).
52These voluntary private associations for sharing the cost ofa common
good--policing--were subsequently undermined by statehood, and the
publicly .nanced local sheriff as the recognized monopolylaw enforcement
of. cer. This observation contradicts the myththat a central function of
government is to "solve" the freerider problem in the private
provisionofpublic goods. Here we have the reverse: the incentive of the
cattlemens clubs was to free ride on the general taxpayer, assigningthe
sheriff the task of enforcing property rights in cattle. The same
freeridingoccurs with school busingprograms, andin
publicly provided education itself in which government .nancing need not
require government provision. tlers under the Homestead Act; mining
claims were de.ned, established, and defended by the guns of the mining
clubbers, whose rules were later to become part of public mining law
(Terry L. Anderson andPeter J. Hill, 1975;John Umbeck, 1977). For over a
century, the Maine lobstermen have establishedrights, usedthreats, then
force, to defend exclusive individual lobster.shing territories in the
ocean (James Acheson, 1975). Eskimo polar bear hunting teams awardedthe
upperhalfofthe bears skin (prized for its long mane hairs used to line
womens boots)to that person who .rst .xed his spear in the prey (Peter
Freuchen, 1961). Extant huntergatherers have evolved sharingcustoms for
the products of communal hunting and gathering. For example, the Ache of
Eastern Paraguay share the volatile products of the hunt widely within the
tribe, while the low variance products of gatheringare sharedonly within
the nuclear family (Hillard Kaplan and Kim Hill, 1985; Kristen Hawkes,
B. Ellickson OutCoases Coase
Using the rancher/farmer parable, Ronald H. Coase (1960)argued that if
there were no costs of transacting, then theoretically ef. ciency could
not depend on who was liable for damages to crops caused by stray
animals. Legal liability gives the rancher an incentive to employcostef.
cient measures to control straying cattle. But if she were not liable,
then in a world ofzero transactions cost, victims would be led in their
own interest to negotiate a settlement paying the rancher to undertake the
same ef.cient control measures induced by legal liability. In so doing,
trespass victims save the costof crop damage, assumed to be more than the
cost of cattle control--otherwise it is inef. cient to control them. The
externality is internalizedby market negotiation incentives. Curiously,
the Coase Theorem--that in the absence of transactions cost ef.
ciencydoes notdepend upon the locus of liability--was controversial. It
was clearly intended as a kindly spoof of oversimpli. ed theories that,
in particular, ignored transactions cost.53 The real problem, addressed
53Later gametheoretic formulations have allowed that with two or more
alternatives there may exist "standoff equilibria" that stall agreement in
Coase bargaining (see
brilliantlyby Coase, was to deal with the question of ef. cient liability
rules in a world of signi. cant transactions cost. He then proceeded to
use the transactions cost framework to examine the problem of social cost
in a variety of legal precedents andcases.54
In the beginning Shasta County California was governed by "open range"
law, meaning that in principle ranchers are not legally liable for damages
resulting from their cattle accidentally trespassing on unfenced land.
Then, in 1945 a California law authorized the Shasta County Board of
Supervisors to substitute a "closed range" ordinance in subregions of the
county. Dozens of conversions have occurred since this enabling law. Under
a closed range law the rancher is strictly liable--even if not
negligent--for damage caused by his livestock. Robert C. Ellickson (1991),
outCoased Coase by, in effect, asking, "Given that this county applies
the polar legal rules used in Coases illustration, how do neighbors in
Shasta County actually handle the problem of stray cattle?" The answer:
"Neighbors in fact are strongly inclined to cooperate, but theyachieve
cooperative outcomes not by bargaining from legally established
entitlements,55 as the parable supposes, but rather developing and
enforcing adaptive norms of neighborliness that trump formal legal
arrangements. Although the route chosen is not the one that the parable
anticipates, the end reached is exactly the one that Coase predicted:
coordination to mutual advantage without supervision by the state"56
(Ellickson, Roger B. Myerson, 1991, p. 506). These cases may limit
extensions of the Coase Theorem, butdo not, Ithink,detract from its
essential message that the locus of liability was irrelevant.
54Coase (1974) also noticed that the lighthouse was frequently cited by
theorists as an example of a "pure" public good. As was his style (to
confront the casual parables oftheory that .nessed certain costs by .at),
his response in effect was, "Well, lets see what people have done who
actually operate lighthouses, or who use the services of lighthouses." It
turned out that early lighthouses were private enterprise, not
government, solutions to a public good problem, and the alleged
inevitability of freeriding was solved by the owner who contracted with
port authorities to collect lighthouse fees when ships arrived portside to
load or unload cargo.
55These are the outside options or threat points in game theory.
56The same results emerge in laboratory experiments reported by Hoffman
and Matthew L. Spitzer (1985). 1991, p. 4). Thus, Shasta County citizens,
including judges, attorneys, and insurance adjustors, do not have full
working knowledge of formal local trespass law.57Citizens notifyowners and
help catchthe trespassinganimal; use mental accounting(reciprocity)to
settle debts, e.g., a rancher, whose cattle have strayedmaytellthe victim
to come down and take some hay, or if your goat eats my tomato plants, you
offer to help me replant them; use negativegossip, complain to of. cials,
submit informal claims for money (but not through a lawyer) to punish
deviant neighbors; rarely use lawyers to seek monetary compensation; share
the building of fences, most often bya rule ofproportionality-- you pay
more if you have more animals than your neighbor; ignore fence law as
irrelevant; and do not change fence obligations with the plantingofcrops.
Finally, contrary to the Coasian parable, the main cost of trespass is
not from crop damage, but from highwaycollisions that killanimals and
C. Extensive Form Interactions Between Anonymously Paired Individuals
Cooperation has also emerged in anonymous twoperson extensiveform games
in laboratory experiments. Although such behavior is contrary to rational
prescriptions, it is not inconsistent with our examples of spontaneous
order without externally imposed law. Why do we study anonymous
interactions in the laboratory?The model ofnonrepeated game theory is
about strangers without a historyor a future (Robert W. Rosenthal, 1981),
but anonymity has long been used in small group experiments to control
for the unknown complexities of natural social intercourse (Siegeland
Fouraker, 1960). It is well documented that facetoface interaction
swamps subtler procedural effects in yielding cooperative outcomes
(Hoffman and Spitzer, 1985; Kagel and Roth, 1995). But more important, I
believe it is this condition that provides the greatest scope for
exploring the human instinct for social exchange, and how it is affected
by context, re
57 Under open range the animal owner is liable for intentional trespass,
trespass ofa lawfulfence, andtrespass by goats, whatever the
circumstances, suggestingthe hypothesis that goat behavior had long been
recognized in pastoral norms and now captured in codi. ed law. ward, and
procedural conditions that vary elements of social distance. Again,
studying what is not helps us to understand what is.
D. Perception and the Internal Order of the Mind: Why Context Matters
Two decision tasks, represented by the same abstract game tree, may lead
to different responses because they occur in different contexts. Why?
The answer may be found in the process by which we perceive the external
world. Hayek (1952)58 was a pioneer in developing a theory of perception,
which anticipated recent contributionsto the neuroscience of perception.
It is natural for our minds to suppose that experience is formed from the
receipt of sensory impulses that re. ect unchanging attributes
ofexternalobjects in the environment. Instead, Hayek proposed that our
current perception results from a relationship between external impulses
and our past experience of similar conditions. Categories formed in the
mind are based on the relative frequency with which current and past
perceptions coincide. Memory consists of external stimuli that have been
modi.ed by processing systems whose organizationis conditionedbypast
experience59 (Hayek, 1952, pp. 64, 165). There is a "constant dynamic
interaction between perception and memory, which explains the ... identity
of processing and representational networks of the cortex that modern
evidence indicates." "Although devoid of mathematical elaboration,
Hayeks model clearly contains most of the elements of those later network
models of associative memory ..." (Joaquin M. Fuster, 1999, pp. 88- 89).
Hayeks model captures the idea that, in the internal order of the mind,
perception is selforganized:abstract function combines withexperience to
determine network connectivityand expansion.60Loss can occur either from
58The Sensory Order was notpublished until 1952when Hayek revised a
manuscript, originally written in the 1920s, entitled in English
translation, "Contributions to a Theory of How Consciousness Develops"
(noted in correspondence to me by Bruce Caldwell).
59The interdependence between perception andmemory is revealed by the
different descriptions of the same event by two eyewitnesses (Gazzaniga et
al., 1998, pp. 484- 86).
60Built into your brain is the maintained hypothesis that the world around
you is stationary. Look at the wall and function or the stimulus of
developmental experience. Block or distort sensory input, and function is
impaired; impair function by brain lesions or inherited de.ciency, and
development is compromised.
This model is consistent with the hypothesis that the mind is organized by
interactive modules (circuits) that are specialized for vision, for
language learning, for socialization, and a host ofother functions(see
Leda Cosmides andJohn Tooby, 1992; Pinker, 1994). In this view, mind is
the unconscious product of coevolution between the biological and
cultural development of our brains that distinguished us from other
primates. It was what made reason possible. Our folk predilection for
believing in the "blank slate" concept of mind (Pinker, 2002) makes plain
that this interpretation of mind is just as consonant with our direct
experience as was once the idea that the earth is .at, or that witches had
to be destroyed. In each case to escape from the folk perception requires
the falsifyingindirect evidence, based on reason, to become partofour
"felt" experience. Constructivist rationality then becomes ecologically
E. Experimental Procedures
The experiments I will report will show how social contextcan be
importantin the interactive decision behavior we observe. This
possibilityfollows from the autobiographical character of memoryand the
manner in which past encoded experience interacts with current sensory
input in creating memory. I will be reporting the results of
decisionmaking in single play, twoperson, sequentialmove game trees.
Subject instructions do not use technical and role
move your eyes back and forth with head still. The wall does not move. Now
press your eyeball with your .nger through the eyelid from the side. The
wall moves as you jiggle your eyeball. Why the difference? When you .ex
the eye muscles and move your eyes back and forth, a copy of the signal
goes to the occipital cortex to offset apparent movementofthe wall so that
the net perception is that ofa stationary wall. This stabilizing
selfordered system for seeing also makes you vulnerable to optical
illusions of motion. Moving your eyes back and forth between the tunnel
gate and your airplane as it docks, you ambiguously "sense" that either
the gate or the plane, or both, is in motion. The ambiguity is resolved
only when the gate, or plane, stops.
suggestive words like "game," "play," "players," "opponent," and "partner"
(except where variations on the instructions are used as systematic
treatments to identify their effect);61 rather, reference is made to the
"decision tree," "decision maker 1" (DM1) or 2 (DM2), and "your
counterpart," etc. The purpose is to provide a baseline context, which
avoids emotive words that might trigger unintended meanings by the
experimenter.62 I do not mean that the baseline is "neutral," a concept
that is not clearly de.nable, given thatcontext effects can dependon
autobiographical experience. The effect of instructionalvariation on
decision is an empirical matter and any particular set of instructions
must always be considered a treatment unless the observations are shown
to be robust to changes in the instructions. All observations must be
seen as ajoint productof experimental procedures and the theoretical
hypotheses, implemented by particular parameters that it is our intention
to test. This is not unique to laboratory observations, but a
characteristic also of .eld observations, and the whole of science (see
Smith, 2002, for examples from physics, astronomy, andexperimental
economics). It is therefore important to understandhow procedures as well
as different parameterizations (games, payoffs) affect behavior.
Subjects are recruited in advance for an economics experiment. Upon
arrival at the appointed time they register, receive a showup fee, and
are assigned to a private computer terminalin a large room with
40stations. Commonly there are 11 other people, well spaced throughout
the room, in the experiments reported below. After everyone has arrived,
each person logs into the experiment, reads through the instructions for
the experiment, responds to instructional questions, andlearns that he or
she is matched anonymously with another person in the room, whose identity
willnever be known,
61See Terence Burnham et al. (2000), discussed below, for a study
comparing the effect of using "partner" and "opponent" in a trust game.
62It is not meaningful or helpful to talk about "experimenter effects."
There are instructional and procedural effects, includingthe presence or
absence ofan experimenter, what he/she knows or does not know (as in
double blind behavioral experiments), and what he/she does or does not do.
All of the elementary operations used to implement an experiment are
treatments that may or may not have a signi. cant effect on observed
outcomes. and vice versa. This does not mean that a subject knows nothing
about their matched counterpart. For example, it mayappear evident that
he or she is another "like" person, such as an undergraduate, or an
industry executive with whom one may feel moreorless an ingroup
identity. Obviously, each person imports into the experiment a host of
different past experiences and impressions that are likely to be
associated with the currentexperiment.
F. The Contextof Decision: The Ultimatum Game Example
Consider the ultimatum game, a twostage, twoperson game with the
following abstract form: for each pair the experimenter makes a .xed sum
of money, m, available (e.g., m will be 10 onedollar bills, or 10
tendollar bills); Player 1 moves .rst offering 0 # x # m units of the
money to Player 2, Player 1 retaining m 2 x;Player 2 then responds by
either accepting the offer, in which case Player 1 is paid m 2x, and
Player 2 is paid x, or rejecting the offer, in whichcase each player
Below I report ultimatum results from four different
instructional/procedural treatments (contexts) that have the same
underlying abstract game structure. In each case imagine that you are
Player 1. See Hoffman et al. (1994; hereafter HMSS) for instructional
details, and for references to the literature and the origins of the
Divide $10.--You and your counterpart have been provisionallyallocated$10,
and randomly assigned to positions. Your task as Player 1is to divide the
$10 by .lling out a form that will then go to your counterpart who will
acceptor reject it.
Contest Entitlement.--The 12 people in the room each answer the same 10
questions on a general knowledgequiz. Your score is the number of
questions answered correctly; ties are broken in favor of the person who
.rst .nished the quiz. The scores are ranked from 1(highest) through 12
(lowest). Those ranked 1-6 are informed that they have earned the right
to be Player 1, the other six will be Players 2.
Exchange.--Player 1 is a seller, and Player 2 is a buyer. A table lists
the buyer and seller pro.t for each price $0, $1, $2, ... , $10charged by
the seller, and the buyer chooses to buy or not buy. The pro.t of the
seller is the price chosen;the pro. tof the buyer is ($10 2 price). Each
receives nothing if the buyer refuses to buy.
Contest/Exchange.--This treatment combines Contest withExchange; i.e., the
sellers andbuyers in Exchange are selected by the contest scoring
procedure. In one version the total amount is 10onedollar bills, and in
the second it is 10 tendollar bills.
Whatever the context there is a gametheoretic concept ofequilibrium
(subgame perfect) that yields the same prediction in all four treatments
(Reinhard Selten, 1975): Player 1 offers the minimum unit ofaccount, $1
($10)if m 5 $10 ($100), andPlayer 2 accepts the offer. This follows from
the assumption that each player is selfinterested in the narrow sense of
always choosing the largest of two immediate payoffs for herself; that
this condition is common knowledge for the two players; and that Player 1
applies backward induction to the decision problem faced by Player 2,
conditional on Player 1s offer. Thus Player 1 reasons that
anypositivepayoff is better than zero for Player 2andtherefore, Player
1need onlyoffer x 5 $1 ($10).
One dif. culty with this analysis is that, depending on context, the
interaction may be interpreted as a social exchange between the two
anonymously matched players who in daytoday experience read intentions
into the actions ofothers (S. BaronCohen, 1995). Suppose the situation is
perceived as a social contract as follows: if Player 2 has an entitlement
to more than the minimum unitofaccount, then an offer of less than the
perceived entitlement(say, only $1, or even $2-$3) may be rejected by some
Players 2. Player 1, introspectively anticipating this possible mental
state of Player 2, might then offer $4 or $5 to insure acceptance of his
offer. Alternatively, Player 1 might enjoy (get utility from) giving money
to his counterpart. The point is simply that there are alternative models
to that of subgame perfection that predict choices in the ultimatum game,
and these alternatives leave wide latitude for the possibility of context
affecting the behavior of both players. Abstract game theory can embrace
TABLE 2--MEAN PERCENTAGE OFFERED BY TREATMENT IN ULTIMATUM GAMES
these alternatives through the arti.ce of "types"-- utilities, or beliefs
states suchas trust, trustworthiness, reciprocation, etc. Ultimately the
predictive success of such models depends on relating task descriptions
de. ning context to autobiographical characteristics of individuals that
are then identi.ed by types that in turn determine behavior. The point
that needs emphasis is that it is easy to go from types (traditionally
utility or beliefs about states) to gametheoretic choice;the hard part is
to relate types to characteristics of the individuals memorysensory
system. Given the directions ofneuroscience andthe learning from brain
imaging, I do not think this is an impossible order.
Observe thatin "Divide $10" the original$10 is allocated imprecisely to
both players. Moreover, a common de. nition of the word "divide"
(Webster) includes the separation of some divisible quantity into equal
parts. Finally, random devices are recognized as a standard mechanism for
"fair" (equal)treatment. Consequently, the instructions might be
interpretedas suggesting that the experimenter is engaged in the "fair"
treatment of the subjects cueing them to be "fair" to each other.
As an alternative, Contest deliberately introduces a pregame procedure
that requires Player 1 to "earn" the right to be the .rst mover. This may
cue some insipient norm of just desserts based on the pregame quiz. In
Exchange the ultimatum game is imbedded in the gains from exchange from a
transaction between a buyer and a seller. In an exchange, both the buyer
and the seller are made better off, and buyers in our culture may accept
the right of a seller to move.rstby quotinga price. Contest/Exchange
combines the implicit property right norm of a seller with a mechanism
for earning the property right.
Table 2 summarizes the results from two different studies ofultimatum game
bargaining with stakes of either 10 onedollar or 10 tendollar bills for
eachofNpairs of players, where N varies from 23 to 33 subject pairs.
1. Comparing Divide $10 with Divide $100 under the random entitlement we
observe a trivial difference in the amount offered between the low stakes
(43.7 percent) and the tenfold increase in the stakes (44.4 percent).
Also, there is no signi. cant difference in the percentage rate at
whichoffers are rejected, 8.3 percent, and 3.7 percent, respectively.
2. When Exchange is combined with an earned entitlement the increase in
stakes lowers the offer percentage from 30.8 percent for $10 stakes to
27.8 percent for $100 stakes, but this difference is within the normal
range of sampling error usingdifferentgroups ofsubjects and is not signi.
cant. Surprisingly, however, this miniscule decline in the mean offer
causes the rejection rate to go up from 12.5 percent to 21.7percent. Three
of four subject Players 1 offering $10 are rejected, and one offer of $30
is rejected in the game with $100 stakes. As has been shown in
trust/punishment games, this behavior is associated with a strong human
propensity to incur personal cost to punish those who are perceived as
cheaters, even under strict anonymity.
3. We note that comparing the Divide $10/ Random entitlement condition
withthe Exchange entitlement, the offer percentage declines from 43.7
percent to 37.1 percent, and comparing the former to the Earned
entitlement the decline is from 43.7 percent to 36.2 percent, both
reductions being statisticallysigni. cant. Even more signi. cant is the
reduction from 43.7 percent to 30.8 percent when the Earned and Exchange
entitlements are combined. Moreover, in all four of these comparisons the
rejection rate is null or modest (0 to 12.5 percent).
4. The small proportion of the offers that were rejected, except when the
stakes were $100 in the Earned/Exchange context and the mean offers
declined to a low of 27.8 percent, indicates thatPlayers 1read their
counterparts well, and as the context is altered, normally offer a suf.
cient amount to avoid being rejected. The one exception shows clearly that
pushing the edge, even if it seems justi.ed by the higher stakes, may
invite an escalation ofrejections.
These data indicate that context is important in the ultimatum game: the
percentage offered varies by over a third as we move from the highest
(44percent)to the lowest (28 percent) measured effect. Also see Hoffman et
al. (2000). Like variation is reported in crosscultural experiments: a
comparison of two huntergather and .ve modern cultures reveals variation
from a high of 48 percent (Los Angeles subjects) to a low of 26 percent
(Machiguenga subjects from Peru) (Joseph Henrich, 2000). These
comparisons used care in attempting to control for instructional
differences across different languages, but this is inherently
problematic, given the nature ofperception, in that one cannot be sure
thatthe instructions, translations, payoffs, or the procedures for
handling the subjects,control adequately for context across cultures. In
each culture one needs to vary the instructions/procedures and observe
the sampling distribution of outcomes, then compare the sample
distributions across cultures. These instructionaltreatment effects call
into question the extent to which one can de.ne what is meant by
"unbiased" instructions. If results are robust with respect to
instructional changes, this can only be established empirically. Without
such studies no claims can be made concerning the relative "neutrality" of
instructions. The main lesson is that, because of the nature of perception
and memory, context should matter, and in the ultimatum game the variation
of observed results with systematic instructional changes designed to
alter context shows clearly that context can and does matter.
Experimenters, subject to the same perception/ memory variations, are
likely to disagree as to what is "neutral."
G. Dictator Games With and Without Gains from Exchange
The ultimatum game is converted into a dictator game byremoving the
rightof the second mover to veto the offer of the .rst. Forsythe et al.
(1994; hereafter FHSS) note that if the observed tendency toward equal
split of the prize is due primarily to "fairness"--a socialnorm of just
division--then it should be of little consequence if this right is
eliminated. But if it is the prospectofrejection-- however
irrational--that tempers the amount offered by Player 1, then the outcomes
should be materially affected by removing the right of rejection, which
converts the ultimatum game into what is called the dictator game. Thus a
signi. cant reduction in the mean percent offered in the dictator game
wouldbe consistentwith the secondhypothesis, while no signi. cant
reduction would be consistent with the .rst. Comparing the results in
Table 3, column 1 with those for Divide $10, Random entitlement in Table
2, we see that the mean dictator offer is only 23.3 percent compared with
the mean ultimatum offer of 43.7 percent. FHSS conclude that fairness
alone cannotaccountfor behaviorin the ultimatumgame. This is correct,
but, equallyof interest, why are dictators giving away nearly a quarter of
their endowment? This research puzzle was picked up by HMSS who
conjectured thatsuch gener
TABLE3--DICTATOR GIVING: WITH AND WITHOUT GAINS FROM EXCHANGE AND SOCIAL
ositymight be, at least in part, a consequence of the incompleteness of
anonymity. In all the games prior to the HMSS studythe members of each
player pair were anonymous with respect to each other butnotwithrespect to
the experimenter who knew every persons decision. Hence, they introduced
a "double blind" treatment category (two versions) in which the protocol
made it transparent that no one, including the experimenter, could learn
the decisions of any player. Data from the secondversion, Double Blind 2,
are reported in Table 3. In this treatment mean dictator offers decline to
only 10.5 percent. Consequently, context--in this case social
connectedness or distance--has an important affect on dictator
transfers.63 These issues are explored more fully in Hoffman et al.
(1996b), who vary social distance by varying the instructional and
protocol parameters that de.ne various versions of singleand
doubleblinddictator games. Also reported in Table 3is the percent given
by the top 50 percent ofthe most generous dictators: 38.3 percent for
Single Blind and 21 percent for Double Blind. Berg et al. (1995, hereafter
BDM)modify the dictator game to introduce gains from "ex
63These double blind procedures and treatment effects have been replicated
by two other investigations (Catherine C. Eckel and Philip J. Grossman,
1996, and Burnham, 1998). Bolton et al. (1998), usinga different double
blind procedure failed to replicate the results, suggesting that
procedures matter and interact with the double blind condition.
change."64Their investmenttrust twostage dictator game also uses the
Double Blind 2 protocol: dictators in room Asend any portion of their $10
(0 to $10)to their random counterpart in room B. People in bothrooms know
that if $x is sent byanyone, it is tripled, so that the counterpart
receives $3x. Thus, the most generous offer, $10, yields a gain of $30.
The counterpart can then respond by sending any part (0 to $3x)of the
amountreceived backto his or her matchedsender. Now an exchangewithgains
to both parties is possible, and BDM ask if this context is a signi. cant
treatment. Note that the analysis of the game is no different than the
onestage dictator game: by the principle of backward induction Player 1
can see that Player 2s interest is to keep all the money received, and
therefore nothing should be sent. The fact thatthe senders transfer
willbe tripled is irrelevant. But it is not irrelevant if both players see
the interactionas an exchange based on trust by Player 1 and
trustworthiness by Player 2.
In Table 3 sender Players 1 now give 51.6 percent when the transfer is
tripled, compared with 23.3 percent when it is not. Furthermore, the top
50 percent of the givers send 74.4 percent of the money, up from 38.3
percent. This shows how the tripledpie shifts the distribution 64For a
recent extension and replication of the BDM .ndings see Madan Pillutla et
al. (2003). Also see Ortmann et al. (2000). toward larger transfers by
Players 1. But on average the senders do not quite break even: an average
of 27.2 percent of the amount received byPlayers 2is returned to Players
1(break even would be 33.3 percent since x is tripled). In the social
history treatment the instructions and protocolare the same as described
above except that the second treatment group is shown the
distributionofamounts transferredand returned for the .rst group.
Comparingthe social history with the baseline mean percent given and
returned reveals the effect ofbeingexposedto the decision data ofthe .rst
group. Social history does not cause a reduction in transfers, which
actually increases marginally from 51.6 percent to 53.6 percent. The
average percent returned increases from 27.2percent to 35.5percent,just
above the breakeven level.
These results are not explicable bythe canons of traditional game theory
that assumes selfinterested (in the sense of always choosing larger
payoffs) types. Byintroducinggainsfrom the investment by Player 1, who can
only bene.t if Player 2 perceives the process as an exchange calling for
payment for services rendered, dictator giving more than doubles. And the
effect ofsocial history does not precipitate a decline in investment nor
in the return to Players 2--in fact both increase slightly. The same
behaviors have been observed in chimpanzee and capuchin monkey communities
(Frans B. M. deWaal, 1989, 1997). Should such trustingand trustworthy
behaviors be diminishedin human communities characterized by the maxim
that "the rules ofmoralityare not ... conclusions of (our) reason?"
H. Trust Games
Ultimatum and dictator games have been studiedextensively, butare much too
simple to allow an adequate understandingofsome ofthe underlying behavior
manifest in them. It is tempting to overinterpret them in terms of a mixed
utility for own and other reward. The potential for greatly expanding what
can be learned is illustrated in Table 3 where BDM extendedthe dictator
game to a twostage game with gains from voluntary exchange. We turn
therefore to a somewhat richer class of twoperson extensiveform trust
games in which equilibrium play, cooperation, and the prospect ofdefection
can be studiedin a richer parameter space than the ultimatum game.65
Figure 1 is a typical"trust" game tree.66 Play starts at the topofthe
tree, node x1, withPlayer 1 who can move right, which stops the game
yielding the upper payoff to Player 1, $10, and the lower payoffto Player
2, $10, or move down in which case Player 2chooses a move at node x2. If
Player 2s move is right, Player 1receives $15, and Player 2, $25. This is
the cooperative (C) outcome. If, however, Player 2 moves down, the payoffs
to 1 and 2are, respectively, $0and $40. This is the defection (D) outcome,
in which Player 2 defects on Player 1s offer to cooperate. The
subgameperfect equilibrium (SPE) is $10 for each player. This follows
because at node 1 Player 1 can apply backward inductionbyobservingthat if
play reaches node x2, Player 2s dominant choice is to defect. Seeing that
this is the case, Player 1s dominant choice is to move right at the top
of the tree, yielding the SPE outcome.
These gametheoretic assumptions are very strong. As we see from the above
discussion, however, they have the dubious meritof allowing "unambiguous"
predictions to be made about behavior.67 We particularly want to note that
if every player is exactly like every other player, and is strictly
selfinterested, there is no room for "mind reading," or
inferringintentions from actions, and no room for more sophisticated and
subtle action in the selfinterest.
To illustrate, suppose that you have been throughthe standard economics
course in game
65Space does not permitexamining also the effect of beingable to punish
defection. See McCabe et al. (1996)for a more complete report of trust
games with and without punishment of defection and witha wide variety of
66 As indicated above, the word "trust" never appears in the instructions.
Of interest, however, is that the subjects use this word when you ask them
openended questions about their analysis and perceptions of the game.
"Its all a question of whether you can trust your partner." Neither do we
use the word "partner."
67Except see Smith (2002), wherein it is shown that if, in addition to the
research hypothesis derived from game theory (e.g., CournotNash or SPE),
there is one auxiliary hypothesis (e.g., payoffs are adequate, types have
been accuratelyde. ned, or subjects are sophisticated), then either the
theoretical hypothesis is not falsi.able, or it has no predictive content.
The belief, however, persists that the predictions of game theory are
sharp andunambiguous (see, e.g., Camerer et al., 2001).
theory and that you are in the position of Player 2 in Figure 1.
Consequently, you expect Player 1 to move right at the top of the tree. He
does not. He moves down, and it is your turn. Surely he did not move down
because he prefers $0 to $10, or expects you to defect. He must thinkthat
you thinkthat he wants you to choose C. Whatever else can he have in
mind? Maybehe cannot do backward induction, or thinks you are not
selfinterested. So how are you going to respond? He is making it
possible for you to increase your payoff by 150percent, and his by 50
percent, compared with the SPE. He is not even asking for the larger share
ofthe pie that his action has created! According to reciprocity theory,
ifyou choose C, you will reciprocate his inferred intentions, and complete
an exchange--exactlyin the same way that you trade favors, services, and
goods across time with your closer friends andassociates in life (except
for those who are victims of antisocial personality disorder, or
sociopaths, and are unable to maintain social relations based on
reciprocity),68 also in the same way that you do not hesitate to leave a
"tip" ("to insure promptness"?) for good service at a restaurant even in
a foreign city. Without a conscious thoughtyou often say, "I owe you one,"
in response to an acquaintances favor. So you might choose C with hardly
a thought, or, since he will never know your identity, upon closer re.
ection, you may thinkthat it just makes no sense not to take the $40.
Although you are nota clinical sociopath, here is an opportunityto cuta
corner and no one can know. As Player 1 in Figure 1 are you certain that
you wouldwant to play SPE?69
68 "Sociopaths, who comprise only 3-4 percent of the male population and
less than 1 percent of the female population, are thought to account for
20 percent of the UnitedStates prison population andbetween 33 percent and
80 percentof the population of chronic criminaloffenders. Furthermore,
whereas the "typical" U.S. burglar is estimated to have committed a
median .ve crimes per year before being apprehended, chronic
offenders--those most likely to be sociopaths--report committing upwards
of 50 crimes per year and sometimes as many as two or three hundred.
Collectively, these individuals are thought to account for over 50
percent of all crimes in the U.S." (see Linda Mealy, 1995, p. 523, and pp.
587-99 for references and caveats; also David T. Lykken, 1995).
69This thought process may explain why, in data reported by Giorgio
Coricelli et al. (2000)comparing faculty with undergraduate subjects, the
faculty take much longer (and earn less money) than the undergraduates to
FIGURE1. INVEST$10 TRUSTGAME: FREQUENCY OF PLAY
In regard to this reciprocity analysis of the game, we should note that
the game in Figure 1 is a much reducedform version of the BDM game:
thinkof Player 1 as sending $10, which becomes $30;Players 2can either
split the $30 equally with Player 1, givingthe imputation C, or Player 2
can keep it all, yielding the D outcome. But there is another difference,
one of context. In the experiment using Figure 1 the subjects play an
abstract game, one that is not embedded in a BDMtype story about sending
$10 upstairs, which becomes $30, and the receiver can either keep it all
or split the gain made possible by the sender. But given the BDM outcomes
reported above we shouldnot be too surprised that some subject pairs might
The outcomes are shown in Figure 1 for 24 undergraduate subjects: 50
percent move down, and of these 75 percent "reciprocate." whether to offer
cooperation, and whether to defect. Yet given knowledge of game theory,
and knowing that ones counterpart has the equivalent knowledge, what is
there to think about?
Why So Much Cooperation?--Mycoauthors and I have interpreted the outcome
Cas due to reciprocity. But there are other interpretations; e.g., that
the subjects are gametheoreticallyunsophisticated or have nonsel. sh
preferences. The effectof subjects on outcomes is an empiricalmatter,
andis most important, butcannotbe pursued here in depth. It is essential
to programs concerned with extending game theory to "player types."
Subject background diversity and resulting choice behavior can help inform
the identi.cation andclassi.cation of "types," whether reciprocators,
sophisticates, or utilitarian (see McCabe and Smith, 2001; McCabe et al.,
Is It the Subjects? Undergraduates Versus Graduates.--In the above trust
game nearly half the Players 1forgo the sure thing, SPE, and
threequarters of the responses are cooperative. We have often heard
suchresults dismissedas a consequence of usingnaive subjects. (This
dismissal has the logicalimplication that the original
theoreticalhypothesis is either not falsi. able or has no predictive
content. See footnote 67.) McCabe andSmith(2000) examinedthis explanation
using advanced graduate students from the population, a sample of whom
participated in the $100 Exchange/Entitlement version of the ultimatum
game reported in Table 2, showing almost identical results for graduate
and undergraduate students. They used the trust game shown in Figure 1.
For comparison, McCabe and Smith (2000)used the undergraduate data shown
in Figure 2. In both groups 50 percent of Players 1offer cooperation,
while 75 percent of the undergraduate and 64 percent of the graduate
student Players 2 reciprocate. However naive undergraduates are alleged to
be, these tests suggest that graduate students with training in economic
theory are capable of showingvery similar behaviorin this extensiveform
game, andin the ultimatum game reported in Table 2.
Is It Utility for Other?--Bolton (1991), Rabin (1993), Fehr and Schmidt
(1999), and Bolton and Ockenfels (2000) have proposed usefulpreference
models ofdecision that aim at accounting for behavior in a variety of
experiments, most particularlyultimatum and dictator games. The idea
behindthese models is that we can explain cooperation in bargaininggames
FIGURE2. INVEST$10 TRUST GAME COMPARING UNDERGRADUATES (U) AND GRADUATES
saying that people have a taste for altruism or for "fair" outcomes (or a
distaste for "unfair" outcomes), where fairness is understoodas payoff
equity, as in Franciosi et al. (1995). The hypothesis is that subjects
seek to maximize these adjusted utilities. It is only the intrinsic
propertiesofoutcomesthatare assumed to drive behavior; what alternatives
the players facedat a previous decision node are irrelevant. This means
that intentions, as re.ected in the move choices, are assumedto be super.
uous in the interactions between the parties. The former approach identi.
es utilitytypes. The latter identi.es types who signal intentions, who
are into reading move signals, and risk misidentifying reciprocity versus
defection types. The important testable distinctions are that the former
are immune to instructional procedures and to path dependency--the
opportunity costs of forgone options; this constitutes the core of the
research program of my coauthors andme. This program leads naturally to an
understanding of its own limitations as well as that of the utilitarian
program, i.e., both types may be needed.
Otherregarding preference models are unable to account for our previous
data demonstrating that procedures and context variables matter. In the
ultimatum game data reported above where the context is variedfrom "Divide
$10" to "Context/Exchange," ostensibly the utilities to the participants
are the same under each ultimatum condition. However, behavior varies
dramatically. These models also cannot explain the results reported above
in the single versus double blind protocol in dictator games, and the
dramatic change in dictator behavior in the BDM investment trust game.
Clearly, the behavioris muchmore variable than is expected from
outcomebased utility models.
An altruistic utility interpretation of cooperation can be invokedin
trust games like that in Figure 1: Player 2 may move down because her
utility for reward is increasing in bothown and other payoff. Figure 3 is
a trust game that enables us to distinguish subjects who cooperate from
motivations of altruism, and those whose cooperation derives from
reciprocity in an exchange. The game starts at the top, node x1, with
Player 1 who can move right, which stops the game yielding the upper
payoff to Player 1, $7, andthe lower payoff to Player 2, $14, or move down
in which case Player 2 chooses a move at node x2. If the move is right,
each player gets $8. If Player 2 moves down, Player 1can then move rightat
node x3yielding $10 for each, or down, yielding $12 for Player 1 and $6
for Player 2. The subgameperfect equilibrium (SPE) is $8 for each player.
This follows because at node x1, Player 1 can apply backward induction by
observing that if play reaches node x3, Player 1 will want to move down.
But Player 2, also usingbackward induction will see that at node x2 he
should move right. Since right at node x2 yields a higher payoff to Player
1, at node x1 Player 1 will conclude that he should move down. Hence, the
SPE outcome would prevail by the logic of selfinterestedplayers who
always choose dominantstrategies, andapply the principleofbackward
If Player 1 has otherregarding preferences (altruism) and is willing to
incur some cost to greatly increase the payoff to Player 2, Player 1 may
move right at x1. His payoff of $7 is only oneeighth smaller than his
payoff at the SPE, and yields $14 for Player 2. Hence, at a cost of $1 to
himself, Player 1 can increase his coun
FIGURE3. TRUST GAME WITH ALTRUISM: FREQUENCY OFPLAY
terpart Player 2s payoff by $6. Player 1 need have onlya modest
preference for an increase in Player 2s welfare to induce him to move
right because of the 6 to 1 return for the other player over the cost to
At x2, Player 2 maymove down signalingto Player 1 that such a move enables
both to pro.t (gains from exchange), provided that at x3 Player 1
cooperates by reciprocatingPlayer 2s favor. Alternatively at x3 Player 1
can defect (D) on the offer to cooperate by choosing his dominate
strategy, and move down.
The outcome frequencies for the trust game (N 5 26pairs) are
entereddirectlyon the tree in Figure3.The .rst result-- overwhelmingly
decisively--is that no Player 1 chooses the A
FIGURE4A. VOLUNTARYTRUST GAME
(altruistic) outcome;all choose to pass to Player 2 seeking a higher
payoff for themselves, and being content to give Player 2 a smaller payoff
than is achieved at A, dependingupon the .nal outcome of the move
sequence. Secondly, 46 percent offer to cooperate (down), and 50 percent
I. Utility Versus Exchange: Does Opportunity Cost Matter?
If reciprocity is perceived as an exchange in which each player gains
relative to the default outcome (SPE), then the cooperative outcome must
yieldan increase in the size of the prize to be splitbetween the two
players (see McCabe et al., 2003). Also, Player 2 must believe that (i)
Player 1 made a deliberate choice to make this outcome possible, and (ii)
incurred an opportunity cost in doing so, i.e., gave up a smaller certain
payoff riskinga still smaller payoff if C is not attained. It then becomes
credible to Player 2that Player 1 did a favor for Player 2, and reasonably
can expect reciprocal action in return. Notice that our argument is in the
FIGURE4B. INVOLUNTARY TRUST GAME
of a constructivist theory that need not characterize the subjects
reasoning, even if it has predictiveaccuracy; i.e.,
constructiverationality may predict emergent ecologically
rationaloutcomes, just as CE theory predicts market outcomes notpart
ofthe consciousintentionsofthe agents. McCabe et al. (2001), however,
report fMRI brainimaging data supporting the hypothesis that subjects
who cooperate use the "mind reading" circuit modules in their brains (see
BaronCohen, 1995). This circuitry is not activatedin subjects who choose
not to cooperate (SPE). In responding to postexperiment
questionsaskingthem to write their impressions oftheir decisions, subjects
frequentlyreport that the experiment is all about whether you can trust
your counterpart. They do not refer to returning a favor, to reciprocity,
an exchange, fairness, etc., suggestingthat if their actions are driven by
reciprocity motives, such are not part of a selfaware reasoning process.
Reciprocity reasoning motivated the alternative game trees shown in
Figure 4 designed to test reciprocity against utility interpretations of
choice. In Figure 4A, ifPlayer 1moves down at the topthe potential prize
increases from $40to $50. Player 2 can defect at a cost to Player 1, and
can clearly infer that Player 1 deliberately enabled the outcome to
increase from ($20, $20)to ($25, $25). But in Figure 4B Player 2 can see
that Player 1 was presented with no voluntary choice to move down.
Consequently, Player 1 incurred no opportunitycost to enable Player 2 to
achieve C. Player 1 did nothing intentionally for Player 2, and according
to reciprocity reasoning, Player 2 incurredno implied debt that neededto
be repaid. Player 2can thus move down with impunity. Consequently,
reciprocity theory predicts a greater frequency of right moves by Players
2 in Figure 4A than in Figure 4B. Since only outcomes matter, both
ownandotherregardingutilitytheories predict no difference in Player 2s
choices between Figures 4A and 4B. In fact, as shown by the frequency data
on the trees, a third of the Players 2on the right play C, whereas nearly
twice that manychoose C in the left game.
Intentionshave also been foundto matter in a study of ultimatum
bargaining: "... we show that identical offers in an ultimatum game
trigger vastlydifferent rejection rates dependingon the other offers
available to the proposer" (Armin Falk et al., 1999, p. 2).
J. ExtensiveVersus NormalForm Games
A fundamental principle of game theory is that rational behavior is
invariant to the form-- extensive or normal--of the game. Behavior in the
extensive andnormal forms has been compared byAndrew Schotter et al.
(1994), Amnon Rapoport (1997), and McCabe et al. (2000). All three reject
the invariance principle, but in the .rst study the rationality principles
they proposedto explain the invariance either failedto predict the
differences, "or they were notwhat we expected" (Schotter et al., 1994,
pp. 446- 47). Rapoport provides two versions of the "BattleoftheSexes"
game to show how orderofplay information in the extensive form allows
players to better coordinate their actions. McCabe et al. (2000) argue
that the important principle that allows better coordination "derives
from the human capacity to read another persons thoughts or intentions by
placing themselves in the position and information state ofthe other
person" (p. 4404). This"mindreading" to detect intentions underlies
($7, $14) 0 26 5 0.0 0 245 0.0 12 7 26 5 0.46 245 0.29 ($8, $8) 6 12 5
0.50 1 75 0.14 ($12, $6) 6 12 5 0.50 6 75 0.86
We summarize here the .ndings of McCabe et al. (2000)for the trimmed
version of the game they study, which is the game we have depicted in
Figure 3. In the extensive form of the game in Figure 3, Player 2 sees the
move of Player 1 before Player 2 chooses to move. In this form of the
game, intentions can be clearly communicated along the lines storiedabove.
In the normal (or strategic) form of the same game each player chooses a
move at each node without knowing whether thatnode willactuallybe
reachedin the move sequence. Decisions are thus contingent on the node
beingreachedandmay be irrelevant in determining the payoffs. In the normal
form, therefore, we can present the game as an n 3 m matrix ofthe n 5
3strategies ofPlayer 1--right at node x1, right or down at node x3, and
the m 5 2 strategies of Player 2, right or down at node x2. Players 1 and
2 each simultaneously choose among these alternatives not knowing the
choice of the other.
McCabe et al. (2000) predict that Players 2 willmove down at x2, with
higher frequency in the extensive than the normal form. They also predict
higher rates of cooperation by Players 1 (andlower defection rates) in the
extensive than normal forms. Neither ownor otherregarding utility
theorycan support these predictions. The data are shown in Table 4: 46
percent of the Players 1 offer to cooperate in the extensive form, only 29
percent in the normal form. Similarly, they observe a 50 percent
cooperative rate byPlayers 2in the extensive form, but only 14 percent in
the normal form.
These results and that of others cited above imply that the extensive and
normal forms are not played as if they were the same games. Players moves
signalintentionsthat are not the same when actually experienced in
extensive form as when imaginedin a mental experiment corresponding to the
normal form. I would argue that experience and its memory in life is an
extensive process that encodes context along with outcome. The brain is
not naturally adapted to solve all sequential move problems byreducing
them to a single strategyvector as in a highly structured game.
Apparently, we have a builtin tendency to wait, observe, then decide--a
process that conserves cognitive resources by applyingthem only to
contingencies that are realized, and avoids the need for revision, given
the inevitable surprises in the less structured games of
life.70Constructivist modeling glosses over distinctions of which we are
unaware that govern the ecologyofchoice. Experimental designs conditioned
only by constructivist thinking, illprepare us to collect the data that
can inform needed revisions in our thinking. It is both cost effective and
faithfulto gametheoretic assumptions in experiments to collect move data
from each subject under all contingencies, but it distorts
interpretability if game forms are not equivalent. The assumptions of
game theory, such as those leadingto the logical equivalence of the two
game forms, shouldnot be imposedon experimental designs, thereby
constraining our understanding of behavior beyond those assumptions.
Neuroeconomics is concerned with studying the connections between how the
mind/brain works--the internal order of the mind--and behavior in (1)
individual decision making, (2) social exchange, and (3) institutions such
as markets. The working hypothesis is that the brain has evolved
different, but interdependent, adaptive mechanisms for each of these tasks
involving experience, memory, and perception. The tools include
brainimagingtechnology and the existence of patients with localized brain
lesions associated with speci.c loss of certain mental functions. Decision
making has drawn the attention of
70 Any such natural process must be deliberately overcome,
constructively, in situations where nature serves us poorly.
neuroscientists who study the deviant behavior of neurological patients
with speci.c brain lesions, suchas front lobe (ventromedialprefrontal)
Such patients have long been known to be challenged by tasks involving
planning and coordination over time, although they score normallyon
batteries of psychological tests (Antonio R. Damasio, 1994). A landmark
experimental study of such patients (compared with controls) in individual
decisionmaking under uncertainty is that of Antoine Bechara et al.
(1997). Starting with (. ctional) endowments of $2,000, each subject on
each trialdraws a card from one offour freely chosen decks (A, B, C, D).
In decks (A, B) each card has a payoff value of $50, whereas in decks (C,
D), each is worth $100.
Also, the $100 decks contain occasional large negative payoff cards, while
the $50deckshave much smaller negative payoff cards. All this must be
learned from single card draws in a sequence of trials, with a running
tally of cumulative payoff value. A subject performs much better by
learning to avoid the $100decks in favor of the $50decks. By period 60,
normal control subjects have learned to draw onlyfrom the $50(C, D) decks,
while the braindamaged subjects continue to sample disadvantageously the
$100 (A, B) decks. Furthermore, the controlsubjects shift to the (C, D)
decks before they are able to articulate why, in periodic questioning.
Also, they preregister emotional reactions to the (A, B) decks as
measured by realtime skin conductivity test (SCT)readings. But the
braindamaged patients tend to verbally rationalize continued sampling of
the (A, B) decks, and some types (with damaged amygdala) register no SCT
response. Results consistent with those of Bechara et al. (1997)have
been reported byVinod Goel et al. (1997) who study patient performance in
a complex .nancial planning task.
Over 50 years ago experiments with animal behavior demonstrated that
motivation was based on relative or forgone reward-- opportunity cost--and
not on an absolute scale of values generated by the brain. Thus David J.
Zeaman (1949)reported experiments in which rats were trained to run for a
large rewardmotivatedgoal. When shifted to a small reward, the rats
responded by running more slowly than they wouldto the small rewardonly.
Acontrolgroup began witha smallreward andshiftedto a large one, andthese
rats immediately ran faster than if the large reward alone was applied.
Monkeys similarly respondto comparisons of differential rewards. It is now
establishedthat orbitofrontal cortex (just above the eyes) neuron activity
in monkeys enable them to discriminate between rewards that are directly
related to the animals relative, as distinct from absolute, preference
amongrewards suchas cereal, apple,andraisins (in order of increasing
preference in monkeys) (Leon Tremblay and Schultz, 1999). Thus suppose A
is preferred to B is preferred to C based on choice response. Then
neuronal activity is greater for A than for B when the subject is viewing
A and B, and similarlyfor B and C when comparing B and C. But the
activity associated with B is much greater when comparedwith C than when
it is compared with A. This is contrary to what one would expect to
observeif A, B, and C arecodedon a .xed properties scale rather than a
relative scale (Tremblay and Schultz, 1999, p. 706).
These studies have a parallel signi.cance for humans. Prospect theory
proposes that the evaluationofa gamble depends noton the totalasset
position but focuses myopically on the opportunitycost, gain or loss,
relative to ones current asset position. There is also asymmetry--the
effectofa loss looms larger than the effect ofa gain of the same magnitude
(Kahneman and Tversky, 1979).Barbara A. Mellers et al. (1999) have shown
that the emotional response to the outcome ofa gamble depends on the
perceived value and likelihoodof the outcome andon the forgone outcome. It
feels better (less bad) to receive $0from a gamble when you forgo 1$10
than when you forgo 1$90. (Theyuse the term "counterfactual" rather than
"opportunity cost" to refer to the alternative that might have
prevailed.) Thus, our ability to form opportunity cost comparisons
receive importantneurophysiological support from our emotional circuitry.
Breiter et al. (2001)use these same principlesin the design of a
functional magnetic imaging (fMRI) study of human hemodynamic responses
to both the expectation andexperience ofmonetary gains and losses under
uncertainty. They observedsigni.cantactivation responses in the amygdala
and orbital gyrus, with both activations increasing with the expected
value of the gamble. There was also some evidence that the right
hemisphere is predominantly active for gains, and the left for losses--a
particularly interesting possibility inviting deeper examination, perhaps
by imaging splitbrain subjects in the same task. Also see Kip Smithet al.
The effect of paying subjects is informed by Thut et al. (1997)who compare
brain activation under monetary rewards with the feedbackofan "OK"
reinforcement in a dichotomous choice task. The monetary rewards yielded a
signi. cantly higher activationofthe orbitofrontal cortex and other
related brain areas (see also Schultz, 2000, 2002).
The neuralcorrelates of individual decisionmaking were extended, by
McCabe et al. (2001), to an fMRI study of behavior in twoperson strategic
interactions in extensiveform trust games like those in Figures 1 to 4.
The prior hypothesis, derived from reciprocity theory, the theoryofmind
literature, and supported by imaging results from individual studies of
cued thought processes (P. C. Fletcher et al., 1995), was that cooperators
would show greater activation in the prefrontal cortex (speci.cally BA8)
and supporting circuitry, than noncooperators. The control, for
comparison with the mental processes used when a subject is playing a
human, is for the subject to play a computer knowing the programmed
response probabilities and therefore havingno needto interpretmoves as
intentions. The predicted activations were signi. cantly greater, relative
to controls, for cooperators than noncooperators, and are consistent with
the reciprocity interpretation of behavior discussed above.
Cartesian constructivism applies reason to the design ofrules for
individualaction, to the design of institutions that yield socially
optimal outcomes, and constitutes the standard socioeconomic science
model. But most ofour operating knowledge, and ability to decide
andperform is nondeliberative. Our brains conserve attentional,
conceptual, and symbolic thoughtresources because they are scarce, and
proceeds to delegate most decisionmaking to autonomic processes
(including the emotions) that do not require conscious attention. Emergent
arrangements, even if initially constructivist, must have survival
properties that incorporate opportunity costs andenvironmental challenges
invisible to constructivist modeling. This leads to an alternative,
ecological concept of rationality: an emergent order based on
trialanderror cultural and biological evolutionary processes. It yields
homeand socially grown rules of action, traditions and moral principles
that underlie property rights in impersonalexchange, and socialcohesion
in personalexchange. To studyecological rationality we use rational
reconstruction--for example, reciprocity or otherregarding
preferences--to examine individual behavior, emergent order in human
culture and institutions, and their persistence, diversity, and
development over time. Experiments enable us to test propositions derived
from these rational reconstructions.
The study of both kinds of rationality has been prominent in the work of
experimental economists. This is made plain in the many direct tests of
the observable implications of propositions derived from economic and game
theory. It is also evident in the great variety of experiments that have
reached far beyond the theory to ask why the tests have succeeded, failed,
or performed better (under weaker conditions) than was expected. What
have we learned, not as.nal truth, but as compelling working hypotheses
for continuing examination?
1. Markets constitute an engine of productivity by supporting resource
specialization throughtrade andcreatinga diverse wealth of goods
2. Markets are rulegoverned institutions providing algorithms that
select, process, and order the exploratory messages of agents who are
better informedas to their personal circumstances than that of others. As
precautionary probes by agents yield to contracts, each becomes more
certain ofwhat must be given in order to receive. Out of this interaction
between minds through the intermediary of rules the process aggregates
the dispersed asymmetric information, converging moreorless rapidly to
competitive equilibria if they exist. Each experimental market carries its
own unique mark with a different dynamic path.
3. Allthis information is captured in the static or timevariable supply
and demand environment and must be aggregated to yield ef. cient
clearingprices. We can never fully understand how this process works in
the world because the required information is not given, or available, to
any one mind. Thus, for many, the arguments of the Scottish philosophers
and of Hayekare obscure and mystical. But we can design experiments in
whichthe information is not given to any participant, then compare market
outcomes with ef. cient competitive outcomes and gauge a market
4. The resulting order is invisible to the participants, unlike the
visible gains they reap. Agents discover what they need to know to achieve
outcomes optimal against the constraining limits imposed byothers.
5. Rules emerge as a spontaneousorder--they are found--not deliberately
designed by one calculating mind. Initially constructivist
institutionsundergo evolutionary change adapting beyond the circumstances
that gave them birth. What emerges is a form of "social mind" that solves
complex organization problems without conscious cognition. This "social
mind" is born of the interaction among all individuals through the rules
of institutions that have to date survived cultural selection processes.
6. This process accommodates tradeoffs between the cost
oftransacting,attending,and monitoring, andthe ef.ciencyofthe allocations
so that the institution itself generates an order of economy that .ts the
problem it evolved to solve. Hence, the hundreds of variations on the .ne
structure of institutions, each designedwithout a designer to accommodate
disparate conditions, but all of them subservient to the reality of
dispersed agent information.
7. We understand little about how rule systems for social interaction and
markets emerge, but it is possible in the laboratory to do variations on
the rules, and thus to study that which is not.
8. Marketsrequire enforcement--voluntaryor involuntary--of the rules of
exchange. These are: the right possession, its transference by consent,
and the performance of promises (Hume). Voluntary enforcement occurs when
people in the market reward good services with gratuities or "tips," an
example, perhaps, of an emergent cultural norm in which people recognize
that tips are part of an informal exchange. If selfor
communityenforcement conditions are not present, the result is unintended
consequences for the bad, as markets are compromisedor may fail. The
game of "trade" must not yield to the game of "steal."
9. Reciprocity, trust, and trustworthiness are important in personal
exchange where formal markets are not worth their cost, yet there are
gains from exchange to be captured. They are also important in
contracting as not every margin of gain at the expense of other can be
anticipated and formalized in written contracts.
10. People are not required to be sel.sh;rather the pointofthe Scottish
philosophers was that people did not have to be good to produce good.
Markets economize on information, understanding, rationality, numbers of
11. Markets in no way need destroythe foundation upon which they probably
emerged--social exchange between family, friends, andassociates. This is
supported in the studies reported by Henrich (2000). Thus, individuals can
be habitual socialexchangers and vigorous traders as well, but as in
Hayeks "two worlds" text, the ecologically rational coexistence of
personal and impersonal exchange is not a selfaware Cartesian construct.
Consequently, there is the everpresent danger that the rules of "personal
exchange" will be applied inappropriately to govern or modify the
extended order ofmarkets. Equallydangerous, the rules of
impersonalmarketexchange may be applied inappropriately to our cohesive
12. New brainimaging technologies have motivatedneuroeconomic studies of
the internal order of the mindand its links with the spectrum of human
decision from choice among.xedgambles to choice mediatedby market and
other institutionalrules. We are onlyat the beginningof this enterprise,
but its promise suggests a fundamental change in how we think, observe,
and model decision in all its contexts.
Acheson, James. "The Lobster Fiefs: Economic and Ecological Effects of
Territoriality in the Maine Lobster Industry." Human Ecology, September
1975, 2(3), pp. 183-207.
Anderson, Terry L. and Hill, Peter J. "The Evolution of Property Rights:
A Study of the American West." Journal of Law and Economics, April 1975,
18(1), pp. 163-79.
Arrow, Kenneth J. "Rationality of Self and Others in an Economic System,"
in Robin M. Hogarthand Melvin Warren Reder, eds., Rationalchoice: The
contrast between economics and psychology. Chicago: University of Chicago
Press, 1987, pp. 201-16.
Backerman, Steven R.; Denton, Michael; Rassenti, Stephen and Smith,
Vernon L. "Market Power in a Deregulated Electrical Industry." Journal of
Decision Support Systems, 2001, (30), pp. 357- 81.
Banks, Jeffrey; Olson, Mark; Rassenti, Stephen and Smith, Vernon L.
"Theory, Experiment and the Federal Communications Commission Spectrum
Auctions." Journal of Economic Behavior and Organization, 2003
BaronCohen, S. Mindblindness: An essay on autism and theoryofmind.
Cambridge, MA: MIT Press, 1995.
Bechara, Antoine; Damasio, Hanna; Tranel, Daniel and Damasio, Antonio R.
"Deciding AdvantageouslyBefore Knowing the Advantageous Strategy."
Science, February 28, 1997, 275(5304), pp. 1293-95.
Berg, Joyce; Dickhaut, John and McCabe, Kevin. "Risk Preference
Instability Across Institutions: A Dilemma." Working paper, Carlson
School of Management, University of Minnesota, 1994. . "Trust,
Reciprocity, and Social History." Games and Economic Behavior, July 1995,
10(1), pp. 122-42.
Binswanger, Hans P. "Attitudes Toward Risk: Experimental Measurement in
Rural India." American Journal of Agricultural Economics, August 1980,
62(3), pp. 395-407.
Bohnet, Iris; Frey, Bruno S. and Huck, Steffen. "More Order with Less Law:
On Contract Enforcement, Trust, and Crowding." American Political Science
Review, March 2001, 95(1), pp. 131- 44.
Bolton, Gary E. "A Comparative Model of Bargaining: Theory and Evidence."
American Economic Review, December 1991, 81(5), pp. 1096-136.
Bolton, Gary E; Katok, Elena and Zwick, Rami. "Dictator Game Giving: Rules
of Fairness versus Acts ofKindness." International Journal of Game Theory,
July 1998, 27(2), pp. 269- 99.
Bolton, Gary E and Ockenfels, Axel. "ERC: A Theory of Equity, Reciprocity,
and Competition." American Economic Review, March 2000, 90(1), pp.
Breiter, Hans C.; Aharon, Itzhak; Kahneman, Daniel; Dale, Anders and
Shizgal, Peter. "Functional Imaging of Neural Responses to Expectancy and
Experience of Monetary Gains and Losses." Neuron, May 2001, 30(2), pp.
Bronfman, Corrine; McCabe, Kevin; Porter, David; Rassenti, Stephen and
Smith, Vernon L. "An Experimental Examination of the Walrasian
Tatonnement Mechanism." RAND Journal of Economics, Winter 1996, 27(4),
Burnham, Terence C. "Engineering Altruism: An Experimental
InvestigationofAnonymity and Gift Giving." Unpublished manuscript, Kennedy
School of Government, Harvard University, 1998.
Burnham, Terence; McCabe, Kevin and Smith, Vernon L. "FriendorFoe
Intentionality Priming in an Extensive Form Trust Game." Journal of
Economic Behavior and Organization, September 2000, 43(1), pp. 57-73.
Camera, Gabriele;Noussair, Charles and Tucker, Steven. "RateofReturn
Dominance and Ef.ciency in an Experimental Economy." Economic Theory,
Camerer, Colin F. Behavioralgame theory: Experiments on strategic
interaction. Princeton, NJ: Princeton University Press, 2002.
Camerer, Colin and Ho, TeckHua. "ExperienceWeighted Attraction Learning
in Normal Form Games." Econometrica, July 1999, 67(4), pp. 827-74.
Camerer, Colin; Ho, TeckHua and Chong, JuinKuan. "Behavioral Game
Theory: Thinking, Learning, and Teaching." Nobel Symposium on Behavioral
and Experimental Economics, Stockholm, December 2001.
Camerer, Colin and Weigelt, Keith. "Experimental Tests ofa
SequentialEquilibrium Model." Econometrica, January 1988, 56(1), pp. 1-36.
Cason, Timothy and Friedman, Daniel. "An Empirical Analysis of Price
Formation in DoubleAuction Markets," in D. Friedman and J. Rust, eds.,
The double auction. Redwood City, CA: AddisonWesley, 1993, pp. 253-83.
Chu, Yun Peng and Chu, Ruey Ling. "The Subsistance of Preference
Reversals in Simpli.edand Marketlike ExperimentalSettings: A Note."
American Economic Review, September 1990, 80(4), pp. 902-11.
Coase, RonaldH. "The Problem ofSocial Cost." Journal of Law and Economics,
October 1960, 3(1), pp. 1- 44.
. "The Lighthouse in Economics." Journal of Law and Economics, October
1974, 17(2), pp. 357-76.
Coricelli, Giorgio; McCabe, Kevin and Smith, Vernon L. "TheoryofMind
Mechanism in PersonalExchange," in G. Hatano, N. Okada, and H. Tanabe,
eds., Affectiveminds. Amsterdam: Elsevier, 2000.
Cosmides, Leda and Tooby, John. "Cognitive Adaptations for
SocialExchange," in Jerome H. Barkow, Leda Cosmides, andJohn Tooby, eds.,
The adapted mind. New York: Oxford University Press, 1992, pp. 163-228.
Cox, James C. and Grether, DavidM. "The Preference Reversal Phenomenon:
Response Mode, Markets and Incentives." Economic Theory, April 1996, 7(3),
pp. 381- 405.
Damasio, Antonio R. Descartes error: Emotion, reason, and the human
brain. New York: G. P. Putnam, 1994.
Davis, Douglas D. and Holt, Charles A. Experimental economics. Princeton,
NJ: Princeton University Press, 1993.
Deck, Cary; McCabe, Kevin and Porter, David. "Why Stable Flat Money
Hyperin.ates: Results from an Experimental Economy." Unpublished
manuscript, Department of Economics, Universityof Arkansas, 2001.
Deck, Cary and Wilson, Bart. "Tracking Customer Search to Price
Discriminate." Working paper, InterdisciplinaryCenter for Economic
Science, George Mason University, 2002.
Denton, Michael J.; Rassenti, Stephen J. and Smith, Vernon L. "Spot Market
Mechanism Design and Competitivity Issues in Electric Power."
JournalofEconomic Behavior and Organization, April 2001, 44(4), pp.
de Waal, Frans B. M. "FoodSharing and Reciprocal Obligations Among
Chimpanzees." Journal of Human Evolution, May 1989, 18(5), pp. 433-59. .
"Food Transfers Through Mesh in Brown Capuchins." Journal of Comparative
Psychology, December 1997, 111(4), pp. 370-78.
Donahue, George. "The U.S. Air Transportation System: A Bold Vision for
Change." Paper preparedfor the Commissionon the Future of the U.S.
AirspaceIndustry, September 19, 2002.
Dufwenberg, Martin; Lindqvist, Tobias and Moore, Evan. "Bubbles and
Experience: An Experiment on Speculation." Working paper, Department of
Economics, Stockholm University, January 2003.
Eckel, Catherine C. and Grossman, Philip J. "Altruism in Anonymous
Dictator Games." Games and Economic Behavior, October 1996, 16(2), pp.
Ellickson, Robert C. Order without law: How neighbors settle disputes.
Cambridge, MA: Harvard University Press, 1991.
Erev, Ido and Roth, Alvin E. "Predicting How People Play Games:
Reinforcement Learning in Experimental Games with Unique, Mixed
StrategyEquilibria." American Economic Review, September 1998, 88(4), pp.
Falk, Armin; Fehr, Ernst and Fischbacher, Urs. "On the Nature of Fair
Behavior." Working Paper No. 17, University of Zurich, August 1999.
Fehr, Ernst and Ga¨chter, Simon. "Do Incentive Contracts Crowd Out
Voluntary Cooperation?" Working Paper No. 34, Universityof Zurich, April
Fehr, Ernst and Schmidt, Klaus M. "ATheory of Fairness, Competition, and
Cooperation." Quarterly Journal of Economics, August 1999, 114(3), pp.
Fiske, Alan Page. Structures of social life: The four elementary forms of
human relations. New York: Free Press, 1991.
Fletcher, P. C.; Happe´, F.; Frith, U.; Baker, S. C.; Dolan, R. J.;
Frakowiak, R. S. J. and Frith, C. D. "Other Minds in the Brain: A
Functional Imaging Studyof Theory of Mind in Story Comprehension."
Cognition, November 1995, 57(2), pp. 109-28.
Forsythe, Robert; Horowitz, Joel L.; Savin, N. E. and Sefton, Martin.
"Fairness in Simple BargainingExperiments." Games and Economic Behavior,
May 1994, 6(3), pp. 347- 69.
Forsythe, Robert; Nelson, Forrest; Neumann, George R. and Wright, Jack.
"Anatomyofan ExperimentalPoliticalStockMarket." American Economic Review,
December 1992, 82(5), pp. 1142- 61.
Forsythe, Robert; Rietz, Thomas A. and Ross, Thomas W. "Wishes,
Expectations and Ac tions: A Surveyon Price Formation in Election Stock
Markets." Journal of Economic Behavior and Organization, May 1999, 39(1),
Fouraker, Lawrence E. and Siegel, Sidney. Bargaining behavior. New York:
Franciosi, Robert; Kujal, Praveen; Michelitsch, Roland; Smith, Vernon and
Deng, Gang. "Fairness: Effect on Temporary and Equilibrium Prices in
PostedOffer Markets." Economic Journal, July 1995, 105(431), pp. 938-50.
Freuchen, Peter. Peter Freuchens book of the eskimos. Cleveland, OH:
World Publishing Co., July 1961.
Friedman, Daniel. "On the Ef. ciencyof Double Auction Markets." American
Economic Review, March 1984, 74(1), pp. 60-72.
. "The Double Auction Market Institution: A Survey," in D. Friedman and
J. Rust, eds., The double auction market: Institutions, theory, and
evidence. Reading, MA: AddisonWesley Publishing Co., 1993, pp. 3-25.
Friedman, Milton and Savage, L. J. "The Utility AnalysisofChoices
InvolvingRisk." Journal of Political Economy, August 1948, 56(4), pp.
Fuster, Joaquin M. Memory in the cerebralcortex: An empirical approach to
neural networks in the human and nonhuman primate. Cambridge, MA: MIT
Gazzaniga, Michael S. The minds past. Berkeley, CA:
UniversityofCalifornia Press, 1998.
Gazzaniga, MichaelS.;Ivry, Richard B. and Mangun, G. Cognitive
neuroscience: The biology of the mind. New York: Norton, 1998.
Gigerenzer, Gerd; Todd, Peter M. and the ABC Research Group. Simple
heuristics that make us smart. New York: Oxford, 1999.
Goel, Vinod; Grafman, Jordan; Tajik, Jinous; Gana, Sheldon and Danto,
David. "A Study of the Performance of Patients with Frontal Lobe Lesions
in a Financial Planning Task." Brain, October 1997, 120(10), pp. 1805-22.
Goldberger, Arthur S. "Heritability." Economica, November 1979, 46(184),
pp. 327- 47.
Hawkes, Kristen. "Showing Off: Tests of an Hypothesis about Mens Foraging
Goals." Ethology and Sociobiology, January 1991, 12(1), pp. 29-54.
Hayek, Friedrich A. The sensory order: An inquiry into the foundations of
theoretical psychology. Chicago: University of Chicago Press, 1952. . The
constitution of liberty. Chicago: University of Chicago Press, 1960.
. Studies in philosophy, politics and economics. Chicago: University of
Chicago Press, 1967.
. Law, legislation,andliberty, volume I, rules and order. Chicago:
University of Chicago Press, 1973.
. "Competition as a Discovery Procedure," in The essence of Hayek.
Stanford, CA: Hoover InstitutionPress, 1984, pp. 254- 65.
. The fatalconceit:The errors of socialism. Chicago: University of
Chicago Press, 1988.
. "Adam Smith (1723-1790): His Message in Todays Language," in The trend
of economic thinking:Essays on politicaleconomists and economic history.
Chicago: Universityof Chicago Press, 1991, pp. 119-24.
Henrich, Joseph. "Does Culture Matter in Economic Behavior? Ultimatum
Game Bargaining Among the Machiguenga of the Peruvian Amazon." American
Economic Review, September 2000, 90(4), pp. 973-79.
Hertwig, Ralph and Ortmann, Andreas. "Experimental Practices in
Economics: A Methodological Challenge for Psychologists?" Behavioral and
Brain Sciences, June 2001, 24(3), pp. 383- 451.
Hoffman, Elizabeth; McCabe, Kevin A.; Shachat, Keith and Smith, Vernon L.
"Preferences, Property Rights, and Anonymity in Bargaining Games." Games
andEconomic Behavior, November 1994, 7(3), pp. 346- 80.
Hoffman, Elizabeth; McCabe, Kevin A. and Smith, Vernon L. "On Expectations
and the Monetary Stakes in Ultimatum Games." InternationalJournalof Game
Theory, 1996a, 25(3), pp. 289-301.
. "SocialDistance and OtherRegarding Behavior in Dictator Games."
American Economic Review, June 1996b, 86(3), pp. 653- 60.
. "The Impact of Exchange Context on the Activation of Equity in Ultimatum
Games." Experimental Economics, June 2000, 3(1), pp. 5-9.
Hoffman, Elizabethand Spitzer, Matthew L. "Entitlements, Rights, and
Fairness: An ExperimentalExaminationof Subjects Concepts of Distributive
Justice." Journal of Legal Studies, June 1985, 14(2), pp. 259-97.
Hogarth, Robin M. and Reder, Melvin Warren. Rational choice: The contrast
between economics and psychology. Chicago: University of Chicago Press,
Holt, Charles. "Industrial Organization: A Survey of Laboratory
Research," in J. Kageland A. Roth, eds., The handbookof experimental
economics. Princeton, NJ: Princeton University Press, 1995, pp. 349- 443.
Hume, David. A treatise of human nature. London: Penguin Books, 1985
(.rst publishedin 1739).
Kachelmeier, Steven J. and Shehata, Mohamed. "Examining Risk Preferences
under High Monetary Incentives: Evidencefrom the Peoples Republic of
China." American Economic Review, December 1992, 82(5), pp. 1120- 41.
Kagan, Jerome. Galens prophecy: Temperament in human nature. New York:
Basic Books, 1994.
Kagan, Jerome and Lamb, Sharon. The emergence of morality in young
children.Chicago: University of Chicago Press, 1987.
Kagel, John H. and Roth, Alvin E., eds. The handbook of experimental
economics. Princeton, NJ: Princeton University Press, 1995.
Kahneman, Daniel; Knetsch, Jack L. and Thaler, Richard. "Fairness as a
Constrainton Pro.t Seeking: Entitlements in the Market." American
Economic Review, September 1986, 76(4), pp. 728- 41.
. "Fairness andthe Assumptions of Economics,"in R. Thaler, ed., Quasi
rational economics. New York: RussellSage Foundation, 1991, pp. 220-35.
Kahneman, Danieland Tversky, Amos. "Prospect Theory: An Analysis of
Decision Under Risk." Econometrica, March1979, 47(2), pp. 263-91.
Kaplan, Hillard and Hill, Kim. "Food Sharing among Ache Foragers: Tests of
Explanatory Hypotheses." Current Anthropology, April 1985, 26(2), pp. 223-
Ketcham, Jon; Smith, Vernon L. and Williams, Arlington. "A Comparison of
PostedOffer and DoubleAuction Pricing Institutions." Review of Economic
Studies, October 1984, 51(4), pp. 595- 614.
Knowlton, Barbara J.; Mangels, Jennifer A. and Squire, Larry R. "A
Neostriatal Habit Learn ing System in Humans." Science, September 6,
1996, 273(5280), pp. 1399- 402.
Lazzarini, Sergio G.; Miller, Gary and Zenger, Todd. "Order with Some Law:
Complementarity vs. SubstitutionofFormalandInformal Arrangements."
Interdisciplinary Conference on Trust and Reciprocity in Experimental
Economics, Washington University, St. Louis, October 11-12, 2002.
Ledyard, John D. "The Scope of the Hypothesis of Bayesian Equilibrium."
Journal of Economic Theory, June 1986, 39(1), pp. 59- 82.
Lopes, Lola. "The Rhetoric of Irrationality." Theory and Psychology, 1991,
1(1), pp. 65- 82.
Lykken, David T. The antisocial personalities. Hillsdale, NJ: Lawrence
Erlbaum Associates, 1995.
Markowitz, Harry. "The Utility of Wealth." Journal of Political Economy,
April 1952, 60(2), pp. 151-58.
McCabe, Kevin A.; Houser, Daniel; Ryan, Lee; Smith, Vernon L. and Trouard,
Theodore. "A FunctionalImaging Study of Cooperation in Twoperson
ReciprocalExchange." Proceedings of the National Academy of Sciences,
September 25, 2001, 98(20), pp. 11832-35.
McCabe, Kevin A.; Rassenti, Stephen J. and Smith, Vernon L. "Smart
ComputerAssisted Markets," Science, October 1991, 254(5031), pp. 534-38.
. "Designing a UniformPrice Double Auction: An ExperimentalEvaluation,"
in D. Friedman and J. Rust, eds., The double auction market:
Institutions, theory, and evidence. Reading, MA: AddisonWesley
Publishing Co., 1993, pp. 307-32.
. "Game Theory and Reciprocity in Some ExtensiveForm Experimental Games."
Proceedings of the National Academy of Sciences, September 16, 1996,
93(23), pp. 13421-28.
McCabe, Kevin A.; Rigdon, Mary L. and Smith, Vernon L. "Positive
Reciprocity and Intentions in Trust Games." Journalof Economic Behavior
and Organization, 2003 (forthcoming).
McCabe, Kevin A. and Smith, Vernon L. "A Comparison of Nave and
SophisticatedSubject Behavior with Game Theoretic Predictions."
Proceedings of the NationalAcademy of Sciences, March 28, 2000, 97(7), pp.
. "GoodwillAccountingand the Process of Exchange," in G. Gigerenzer and R.
Selten, eds., Bounded rationality: The adaptive toolbox. Cambridge, MA:
MIT Press, 2001, pp. 319- 40.
McCabe, Kevin A.;Smith, Vernon L. and LePore, Michael.
"IntentionalityDetectionand Mindreading: Why Does Game Form Matter?"
Proceedings of the National Academy of Sciences, April 11, 2000, 97(8),
pp. 4404- 09.
Mealy, Linda. "The SociobiologyofSociopathy: An Integrated Evolutionary
Model." Behavioral and Brain Sciences, September 1995, 18(3), pp. 523-99.
Mellers, Barbara A.; Schwartz, A. and Ritov, I. "EmotionBased Choice."
Journalof Experimental Psychology: General, March 1999, 128(1), pp. 1-14.
Milgrom, Paul R. and Weber, Robert J. "A Theory of Auctions and
Competitive Bidding." Econometrica, September 1982, 50(5), pp. 1089-122.
Mill, John Stuart. Principles of politicaleconomy, with some oftheir
applicationsto social philosophy, volumes 1 and 2. London: Colonial
Miller, Ross M. "Dont Let Your Robots Grow Up to Be Traders." Working
paper, Miller Risk Advisors, 2002.
Miller, Ross M. and Plott, Charles R. "Product Quality Signaling in
Experimental Markets." Econometrica, July 1985, 53(4), pp. 837-72.
Mullainathan, Sendhil and Thaler, Richard H. "BehavioralEconomics."
Internationalencyclopedia of the social and behavioral sciences, volume
II. Oxford: Pergamon, 2001, pp. 1094-99.
Myerson, Roger B. Game theory: Analysis of con.ict. Cambridge, MA: Harvard
University Press, 1991.
North, Douglass Cecil. Structure and change in economic history. New York:
Olson, Mark;Rassenti, Stephen; Smith, Vernon L. and Rigdon, Mary. "Market
Design and Motivated Human Trading Behavior in Electricity Markets."
Institute of Industrial EngineeringTransactions, 2003(forthcoming).
Ortmann, Andreas; Fitzgerald, John and Boeing, Carl. "Trust, Reciprocity,
andSocialHistory: A Reexamination." Experimental Economics, June 2000,
3(3), pp. 81-100.
Ostrom, Elinor. Governing the commons: The evolution of institutions for
collective action. Cambridge: Cambridge University Press, 1990.
Pillutla, Madan; Malhotra, Deepak and Murnighan, J. Keith. "Attributions
of Trust and the Calculus of Reciprocity." Journalof Experimental Social
Psychology, 2003 (forthcoming).
Pinker, Steven. The language instinct. New York: W. Morrow and Co., 1994.
. The blankslate: The modern denialof human nature. New York: Viking,
Plott, Charles R. "Research on Pricing in a Gas Transportation Network."
Of.ce of Economic Policy Technical Report No. 88-2. Federal Energy
Regulatory Commission, Washington, DC, July 1988.
. "Equilibrium, Equilibration, Information and MultipleMarkets: From Basic
Science to InstitutionalDesign." NobelSymposium, Behavioral
andExperimental Economics,Grand Hotel Saltsjobaden, December 4- 6, 2001.
Plott, Charles R. and Sunder, Shyam. "Ef. ciency of Experimental Security
Markets with Insider Information: An Application of RationalExpectations
Models." Journal of Political Economy, August 1982, 90(4), pp. 663-98.
. "RationalExpectationsandthe Aggregation of Diverse Information in
Laboratory Securities Markets." Econometrica, September 1988, 56(5), pp.
Plott, Charles R. and Uhl, Jonathan T. "Competitive Equilibrium with
Middlemen: An Empirical Study." Southern Economic Journal, April 1981,
47(4), pp. 1063-71.
Plott, Charles R. and Wilde, Louis. "Professional Diagnosisvs.
SelfDiagnosis:An Experimental Examination of Some Special Features of
Markets with Uncertainty," in V. L. Smith, ed., Research in experimental
economics, volume 2. Greenwich, CT: JAI Press, 1982, pp. 63-112.
Porter, David and Smith, Vernon L. "StockMarket Bubbles in the
Laboratory." Applied Mathematical Finance, December 1994, 1, pp. 111-27.
Rabin, Matthew. "Incorporating Fairness into Game Theory and Economics."
American Economic Review, December 1993, 83(5), pp. 1281-302.
Rapoport, Amnon. "Order of Play in Strategically Equivalent Games in
Extensive Form." International Journalof Game Theory, 1997, 26(1), pp.
Rassenti, Stephen; Smith, Vernon L. and Bul. n, R. "A Combinatorial
Auction Mechanism for Airport Time Slot Allocation." Bell Journal of
Economics, Autumn 1982, 13(2), pp. 402- 17.
Rassenti, Stephen; Smith, Vernon L. and Wilson, Bart J. "Using Experiments
to Inform the Privatization/Deregulation Movement in Electricity." Cato
Journal, Winter 2002, 21(3), pp. 515- 44.
. "Controlling Market Power and Price Spikes in Electricity Networks:
Demandside Bidding." Proceedings of the NationalAcademy of Sciences,
March 2003, 100(5), pp. 2998-3003.
Rosenthal, Robert W. "Games of Perfect Information, Predatory Pricing and
the ChainStore Paradox." JournalofEconomic Theory, August 1981, 25(1),
Satterthwaite, Mark. "StrategyProofAllocation Mechanisms," in J. Eatwell,
M. Milgate, and P. Newman, eds., The new Palgrave, volume 4. London:
Macmillan, 1987, pp. 518-20.
Schotter, Andrew; Weigelt, Keith and Wilson, Charles. "A Laboratory
Investigation of Multiperson Rationality and Presentation Effects."
Games and Economic Behavior, May 1994, 6(3), pp. 445- 68.
Schultz, Wolfram. "Multiple Reward Signals in the Brain." Nature Reviews:
Neuroscience, December 2000, 1(3), pp. 199-207.
. "Getting Formal with Dopamine and Reward." Neuron, October 2002, 36(2),
Schwartz, T. and Ang, J. S. "SpeculativeBubbles in the Asset Market: An
Experimental Study." Paper presented at the American Finance Association
meeting, Atlanta, December, 1989.
Segal, Nancy L. Entwinedlives: Twins and what they tell us about human
behavior. New York: Plume, 1999.
Selten, Reinhard. "Reexamination of the Perfectness Concept for
Equilibrium Points in Extensive Games." International Journal of Game
Theory, 1975, 4, pp. 25-55.
Siegel, Sidney. "Theoretical Models of Choice and Strategy Behavior."
Psychometrika, 1959, 24, pp. 303-16.
Siegel, Sidney and Fouraker, Lawrence E. Bargainingand group decision
making: Experiments in bilateral monopoly. New York: McGrawHill, 1960.
Simon, Herbert Alexander. The sciences of the arti.cial, 3rd ed.
Cambridge, MA: MIT Press, 1981, 1996.
Smith, Adam. The theory of moral sentiments. Indianapolis, IN: Liberty
Classics, 1759 (republished in 1982).
. An enquiry into the nature and causes of the wealth of nations.
Indianapolis, IN: Liberty Classics, 1776(republishedin 1981).
Smith, Kip; Dickhaut, John; McCabe, Kevin and Pardo, Jose´ V. "Neuronal
Substrates for Choice Under Ambiguity, Risk, Gains, and Losses."
Management Science, June 2002, 48(6), pp. 711-18.
Smith, Vernon L. "An Experimental Study of Competitive Market Behavior."
Journal of Political Economy, April 1962, 70(12), pp. 111-37.
. "Experimental Economics: Induced Value Theory." American Economic
Review, May 1976 (Papers and Proceedings), 66(2), pp. 274-79.
. "Relevance of Laboratory Experiments to Testing Resource Allocation
Theory," in Jan Kmenta and James Bernard Ramsey, eds., Evaluation
ofeconometric models. New York: Academic Press, 1980, pp. 345-77.
. "MicroeconomicSystems as an Experimental Science." American Economic
Review, December 1982, 72(5), pp. 923-55.
. "ExperimentalMethodsin the Political Economy ofExchange." Science,
October 10, 1986, 234(4773), pp. 167-73.
. Papers in experimental economics. Cambridge: Cambridge UniversityPress,
. "The Two Faces of Adam Smith." Southern Economic Journal, July 1998,
65(1), pp. 1-19.
. Bargaining and market behavior: Essays in experimental economics.
Cambridge: Cambridge University Press, 2000.
. "Method in Experiment: Rhetoric and Reality." Journal of Experimental
Economics, 2002. Smith, Vernon L.; Suchanek, Gerry L. and Williams,
Arlington W. "Bubbles, Crashes and Endogenous Expectations in Experimental
Spot Asset Markets." Econometrica, September 1988, 56(5), pp. 1119-51.
Smith, Vernon L. and Szidarovszky, Ferenc. "Monetary Rewards and Decision
Cost in Strategic Interactions," in M. Augier and J. March, eds., Models
of a man: Essays in memory of Herbert A. Simon. Cambridge, MA: MIT Press,
Smith, Vernon L.; Williams, Arlington W.; Bratton, Kenneth and Vannoni,
Michael G. "CompetitiveMarket Institutions: Double Auctions vs. Sealed
BidOffer Auctions." American Economic Review, March 1982, 72(1), pp.
Soper, Barry and Gigiolotti, Gary. "Intransitive Cycles: Rational Choice
or Random Error?" Theoryand Decision, 1993, 35, pp. 311-36.
Sunder, Shyam. "Market as Artifact: Aggregate Ef.ciency from Zero
Intelligence Traders," in M. Augier and J. March, eds., Models of a man:
Essays in memory of Herbert A. Simon. Cambridge, MA: MIT Press, 2003
Thaler, Richard H. and Ziemba, William T. "Parimutuel Betting Markets:
Racetracks and Lotteries." Journal of Economics Perspectives, Spring
1988, 2(2), pp. 161-74.
Thut, Gregor;Schultz, Wolfram; Roelcke, Ulrich; Nieuhusmeier, Matthias;
Missimer, John; Maguire, R. Paul and Leenders, Klaus L. "Activation of
the Human Brain by Monetary Reward." NeuroReport, January 1997, 8(5), pp.
Tremblay, Leon and Schultz, Wolfram. "Relative Reward Preference in
Primate Orbitofrontal Cortex." Nature, April 22, 1999, 39, pp. 466- 83.
Tversky, Amos and Kahneman, Daniel. "Rational Choice and the Framing of
Decisions," in Robin M. Hogarth andMelvinWarren Reder, eds., Rational
choice: The contrast between economics andpsychology. Chicago:University
of Chicago Press, 1987, pp. 67-94.
Umbeck, John. "The California Gold Rush: A Study of Emerging Property
Rights." Explorations in Economic History, July 1977, 14(3), pp. 197-226.
Vickrey, William. "Counterspeculation, Auctions, and Competitive Sealed
Tenders." Journal of Finance, March 1961, 16(1), pp. 8-37.
Viner, Jacob. "Adam Smith," in D. Irvin, ed., Essays on the
intellectualhistoryofeconomics. Princeton, NJ: Princeton University
Press, 1991, pp. 248- 61.
Williams, Arlington and Smith, Vernon L. "Simultaneous Trading in Two
Competitive Markets." Unpublished manuscript, Indiana University
Department of Economics, Bloomington, Indiana, 1986.
Williams, Arlington W.; Smith, Vernon L.; Ledyard, John O. and Gjerstad,
Steven. "Concurrent Trading in Two Experimental Markets with Demand
Interdependence." Economic Theory, November 2000, 16(3), pp. 511-28.
Wilson, Robert B. "On Equilibria of BidAsk Markets," in George R. Feiwal,
ed., Arrow and the ascent of modern economic theory. New York: New York
University Press, 1987, pp. 375- 414.
. "Strategic Analysis of Auctions," in Robert J. Aumann and Sergi Hart,
eds., Handbook of game theory. Amsterdam: NorthHolland, 1992, pp. 227-79.
. "Design of Ef. cient Auction Procedures," in D. Friedman and J. Rust,
eds., The double auction market: Institutions, theory, and evidence.
Reading, MA: AddisonWesley Publishing Co., 1993, pp. 125-52.
. "Architecture of Power Markets." Econometrica, July 2002, 70(4), pp.
Zajac, Edward E. Politicaleconomy of fairness. Cambridge, MA: MIT Press,
. "What FairnessandDenial Research Could Have Told the Florida Supreme
Court (and Can Tell the Rest of Us)." Independent Review, Winter 2002,
6(3), pp. 377-97.
Zeaman, David J. "Response Latency as a Func tion of the Amount of
Reinforcement." Ex perimental Psychology, 1949, 39, pp. 466- 83.
More information about the paleopsych