R000277 rationality

Economic theory takes the individual consumer and firm as a primitive unit
of analysis, and so a theory of individual agency is required to derive hy-
potheses about the behaviour of markets and other systems of economic
interest. One such theory is the principle of rationality, whereby agents act in
their perceived best interest. This article surveys the implementation of this
principle in economic models, and discusses the critiques of the rationality
principle and some proposed alternatives from the perspective of the eco-
nomic modeller.

I shall not today attempt further to define the kinds of material I

understand to be embraced within that shorthand description; and
perhaps I could never succeed in intelligibly doing so. But I know it
when I see it, y
(Justice Potter Stewart, 378 U.S. 184, 197)

Rationality is for economists as pornography was to the US Supreme

Court, undefinable but nonetheless easily identified; and yet, like the Justices
of the Court, no two economists share a common definition. This article
details some of the common meanings of individual (as opposed to social)
rationality and discusses their uses. Our point of view is that of working
economists rather than that of psychologists. Economics is committed to
METHODOLOGICAL INDIVIDUALISM, the claim that social phenomena must be
explained in terms of individual actions which in turn must be explained
through individuals’ motivations. This commitment requires a theory of hu-
man action. The rationality principle, that individuals act in their best interest
as they perceive it, provides such a theory. In this article we evaluate the
rationality hypothesis and its alternatives from the perspective of how they
explain social phenomena such as the behaviour of a market. Our interest is
in social life rather than in the psychology of an individual.

History and description

The use of the rationality principle in economics certainly predates the util-
itarianism with which it is so often conflated. Adam Smith (1789, p. 19)
describes, in his discussion of the division of labour, a tribe of hunters in
which one person is particularly deft at making bows and arrows. ‘He fre-
quently exchanges them for cattle or for venison with his companions; and he
finds at last that he can in this manner get more cattle and venison, than if he
himself went to the field to catch them. From a regard to his own interest,
therefore, the making of bows and arrows grows to be his chief business, y’
Moving from intuition to analysis, however, requires a sharp understanding
of what it means to regard one’s own interest, and this has become a source
of endless debate for rational-actor social scientists.
The utility-maximization version of rationality springs from the utilitar-
ianism of Bentham and Mill. According to Bentham (1789, p. i),

Nature has placed mankind under the governance of two sovereign

masters, pain and pleasure. It is for them alone to point out what we
ought to do, as well as to determine what we shall do. On the one hand
2 rationality

the standard of right and wrong, on the other the chain of causes and
effects, are fastened to their throne.

Although many thinkers toyed with utilitarian approaches to economic

analysis, it was not until the 1870s, through the work of Jevons, Menger, and
Walras, that utility maximization began to assume the important role in
economic analysis it has since held. For this trio, utility was a short cut to a
theory of value. Perhaps this is why they were not overly concerned with the
issues of measurable utility and the possibility of interpersonal utility com-
parisons which so exercised their successors. Utility for Bentham, on the
other hand, was a physical measure of pain and pleasure which could be
computed according to his ‘felicific calculus’. Although utility as a merely
hedonic measure was rejected even by Mill, only in the 1930s, and after a half
century’s work beginning with Fisher (1892) and Pareto (1895) was it gen-
erally recognized that properties of demand derived from the shape of in-
difference curves, and so utility could admit a purely ordinal interpretation.
This ‘shift in emphasis away from the physiological and psychological he-
donistic, introspective aspects of utility’, as Samuelson (1947, p. 90–1) put it,
led to the ‘purging out of objectionable, and sometimes unnecessary, con-
notations y of the Bentham y variety’. The ultimate expression of this
non-psychological view is the theory of revealed preference, whose purpose is
‘yto develop the theory of consumer’s behavior freed from any vestigial
traces of the utility concept’ (Samuelson, 1938a, p. 71). The result is a
mathematical structure that Edgeworth would have understood, interpreted
in a manner completely foreign to his way of thinking.
Expected utility in the theory of choice under uncertainty is older than
Benthamite utilitarianism. Both an expectation argument and a dominance
(admissibility) argument for the existence of God were carefully laid out by
Pascal (1672, p. 233). These remarkable few paragraphs touch on many
important issues in contemporary decision theory, including the principle of
insufficient reason, the problem of infinite utility payoffs, and incomplete
preferences: ‘Yes; but you must wager. It is not optional. You are embarked.
Which will you choose then?’ Even the concept of marginal utility predates
Bentham, in Gabriel Cramer’s and Daniel Bernoulli’s famous near-resolu-
tions of the St. Petersburg paradox. But despite this early progress, the
formalization of the modern theory of choice under uncertainty begins only
with Wald (1939), who at one go describes the key structures of statistical
decision theory: loss functions, a priori distributions, and Bayes, admissible,
and minimax decision rules. Interest quickly coalesced, however, around the
expected utility models described in the two great testaments of decision
science, von Neumann and Morgenstern (1947) and Savage (1956). Expected
utility (EU) quickly became such a dominant paradigm for choice under
uncertainty that research into alternatives was a backwater for 20 years. But
criticisms of the expected utility models emerged almost before the ink was
dry on the two manuscripts, in Allais’ (1953) experiments and Cyert, Simon
and Trow’s (1956) empirical studies of firm behaviour, and by the late 1970s
behavioural economics and non-EU decision theory were active areas of
research.
Psychological utilitarianism and decision theory are the two traditions
which most inform the modern economist’s thinking about ‘rationality’, and
yet, despite the long intellectual history of these ideas, no single vision of
what it means to be a ‘rational actor’ has emerged. In the remainder of this
article we single out several sources of confusion and disagreement. We
discuss five models of rationality.
rationality 3

General choice theory (GCT)

A set A of alternatives is given, along with a collection B of non-empty
subsets of A. The set A is the set of possible alternatives and any member B
of B is a set of feasible alternatives, a set from which the decision-maker
must choose. A choice function C assigns to each B 2 B a nonempty subset of
B, the objects chosen by the decision-maker from the feasible set. In the
theory of demand, for instance, A is the consumption set, B is the collection
of possible budget sets and the choice function is the demand function. A
textbook treatment of the rational decision-maker requires that she have a
preference relation k on A, and we understand akb to mean that she finds a
to be ‘at least as good as’ b. By ‘preference relation’ we mean a binary
relation which is complete, all alternatives can be compared, and transitive.
Transitivity means that if a is at least as good as b and b is at least as good as
c, then a is at least as good as c. Chosen objects in a set B of feasible
alternatives are those maximal with respect to the preference relation; b is
chosen from B, that is, b 2 CðBÞ, if and only if bka for all a 2 B. Preference
is the primitive expression of rationality. The role of utility is to provide a
convenient representation of preference. A utility function u is a real-valued
function on A, and to say that u represents k means that uðaÞkuðbÞ if and
only if akb. While the decision theory toolkit of the working economist
mostly specializes this model, much contemporary economic theory does not
require this much of rationality. In particular, the completeness and tran-
sitivity assumptions can be done away with in general equilibrium theory,
and numerical representations for incomplete and non-transitive preferences
are available. See, for instance, Aumann (1962), Chipman et al. (1971), and
Gale and Mas-Colell (1975).

Expected utility theory (EU)

Expected utility is a specialization of GCT in which the set A and the pref-
erence relation have a specific structure. In EU theory, X is a finite set of
prizes or outcomes, and the alternative set A is the set of all probability
distributions on X. Preferences have the following representation: A payoff
function v is a real-valued function on X, and any two probability distribu-
tions p and q in A are compared according to their expected values of v; that
is pkq iff E p vkE q v. The content of this theory is that, geometrically speak-
ing, indifference curves are parallel straight lines (hyperplanes). The first
characterization of EU preferences was provided by von Neumann and
Morgenstern (1947); today’s standard axiomatic characterization of EU
preference orders is due to Herstein and Milnor (1953).

Subjective expected utility theory (SEU)

When we choose whether to play roulette or a slot machine, we are choosing
among probability distributions. When we bet on the outcome of a horse or
political race, we are betting on the realization of uncertain outcomes, but
not objects to which probabilities are necessarily attached. Savage’s (1956)
contribution was to provide a theory of what he called ‘personal probability’,
a specialization of GCT, here interpreted as a decision-maker’s degree of
belief in the occurrence of some event. He characterized those preference
relations which could be represented by the expectation of some payoff
function with respect to a personal probability. In Savage’s subjective ex-
4 rationality

pected utility (SEU) theory, S is a set of states, such as the possible outcomes
of the election. There is also a set X of outcomes. A bet on the election is a
function which assigns an outcome to every state. Savage called such func-
tions f : S ! X acts, and the set of acts is the alternative set A. A preference
relation k on the set A has an SEU representation if there is a payoff
function v on outcomes X and a probability distribution p on states S such
that f kg if and only if E p fvðf ðsÞÞgkE p fvðgðsÞÞg.
Methodological individualism requires the analysis of social phenomena to
be ‘bottom-up’, that is, to begin with individuals. It is a stronger statement to
claim, however, that the description of the individual is entirely pre-social;
that in economic models, for instance, that individuals come to the market
with preferences and beliefs already formed. Most modern economists do not
make this claim, and instead work with models in which the description of
the individual is an equilibrium outcome. The two most prominent examples
of this method are rational expectations equilibrium and non-cooperative
game theory.

Rational expectations equilibrium (REE)

The rational expectations hypothesis supposes a population of individuals
solving decision problems which have a common state space, and further-
more that the state will be chosen according to the ‘true distribution’ m,
which is determined by the individuals’ choices. The payoff v(c,s) to a choice
c depends on the state realization s, and preferences over choices are EU:
U i ðcÞ ¼ E m vfc; sg. The hypothesis asserts that all beliefs will be correct; that
is, that all SEU decision-makers have preference representations in which the
beliefs are in fact the probability distribution m, and m in turn is the dis-
tribution of states which is determined by their actions. Rational expecta-
tions is a misuse of the adjective. Unfortunately it is probably too late to
abandon the term. There is no connection between the rationality principle,
which claims that individuals act in their perceived best interest, and the
rational expectations hypothesis, which claims that those perceptions meet
some ex ante standard of correctness. But so labelling a theory is certainly a
nice rhetorical move for how it structures subsequent debate.

Non-cooperative game theory (NGT)

A population of individuals chooses actions. Individual i’s payoff to action
ci, vðci ; c
i Þ depends upon the choices c
i of the others. He holds probabilistic
beliefs about the actions of others, and evaluates a choice according to EU.
The social construction of the individual is seen in the determination of
beliefs. Undominated strategies are those which can be rationalized by some
choices of beliefs. Rationalizable strategies are those which can be justified by
some beliefs satisfying a belief restriction, that it be common knowledge that
all members of the population are EU-rational with some beliefs, and that
payoffs be common knowledge (see EPISTEMIC GAME THEORY: AN OVERVIEW).
Nash equilibrium requires, like REE, that everyone’s beliefs are correct.
Various Nash equilibrium refinements also have belief interpretations (see
NASH EQUILIBRIUM, REFINEMENTS OF).
rationality 5

Rationality and mind

The merits of the rational choice foundation of economics have been much
discussed, both by its practitioners and by its critics. This discussion is often
confused, in part because economists are not consistent in how they under-
stand the contents of the rationality hypothesis. Economic theory holds two
views of rationality. One is that rationality is consistency of choice, that the
tools of choice theory are just an alternative encoding of certain choice
functions; the other is that rationality is a theory of intentional behaviour, in
which beliefs and desires are meaningful constructs.
Revealed preference theory is the sharpest formulation of the consistency
view. It takes demand as primitive and asks if it is consistent with the max-
imization of a preference order. It recovers desires from choice, and only to
the extent that choices are different can two desires, preference orders, be
distinguished. This view permeates the foundations of decision theory. For
Savage (1956, p. 17), ‘It is possible that the person prefers f to g. Loosely
speaking, this means that, if he were required to decide between f and g, no
other acts being available, he would decide on f’. In this account, preference
is defined by choice. This means specifically that if the choice function C on a
collection B of choice sets satisfies certain conditions, then there is a com-
plete and transitive binary relation such that for every B 2 B, C(B) contains
exactly the elements of B which are maximal in B with respect to the relation.
The binary relation is nothing more than an alternative description of C on
B. Suppose a new choice set B0 eB is considered. What can we guess about
the contents of C(B)? Knowing that the decision-maker is consistent on B
allows the observer to infer nothing at all about C(B0 ).
If revealed preference represents at all a psychology of choice, that psy-
chology is a form of radical behaviourism. Radical behaviourism asserts that
two mental states are distinguishable only to the extent that some observable
behaviour distinguishes them. Behaviours are all that one can theorize about.
Samuelson (1938b, p. 344) writes ‘of a steady tendency toward the removal of
moral, utilitarian, welfare connotations y’ and of ‘the rejection of hedon-
istic, introspective, psychological elements’. Although the behaviourist po-
sition seems extreme, the leading graduate microeconomics textbook writes
approvingly of reveal preference: ‘Perhaps most importantly, it makes clear
that the theory of individual decision making need not be based on a process
of introspection but can be given an entirely behavioral foundation’ (Mas-
Colell, Whinston and Green, 1995, p. 5). Consistency is often justified as
discipline. It requires a minimum of assumptions about the beliefs and de-
sires of individuals, and minimizes the possibility of researchers’ values and
beliefs slipping unbidden into their analyses. It allows the data maximal
scope to speak for itself.
Although received economics talks approvingly of rationality as mere
consistency, this is not in fact what most economists do. Much of economics
involves invisible-hand explanations; aggregate market behaviour emerges
from the decisions of many agents. Whether the invisible hand lifts the cup
aloft or knocks it over, economic explanation entails explaining how it co-
ordinates for good or ill the motives and interests of diverse individual ac-
tors. These kinds of question call for explanations based on the motivations
of economic actors, which purely behaviouralist explanations cannot pro-
vide. So economists in practice take an intentional view.
The intentional view holds that rational choice theory is a common-sense
or ‘folk’ (as opposed to ‘scientific’) psychology. Just as in our everyday
transactions we use the language of beliefs and desires to interpret and fore-
cast the behaviour of others, so do economists interpret choice behaviour.
6 rationality

The investor believes that the asset price will be higher tomorrow. She wants
greater wealth tomorrow. So she acts by purchasing the asset. In this view
belief and desire are in fact mental states that are connected to action. The
folk psychology is a theory of mind which is presumed by economists to be
both adequate for a descriptive psychology of decision and accurate enough
in its predictions of individual behaviours for the uses to which it is put.
Although utility does not exist as a psychophysical quantity, rational choice
models provide a representation of the mental states involved in judgment
and decision. (The stronger claim that mental process is a more or less
efficient utility maximization algorithm is a view held only by the straw man
regularly beaten up by rationality’s critics.)
The economist’s folk psychology goes further than everyday folk psy-
chology by specifying analytic representations of beliefs, desires, and how
they interact. No matter what representation is ultimately chosen by the
textbook economist, his folk psychology rests on two points. (1) Rationality
is instrumental. Its concern is the efficient pursuing of ends by available
means, not the sensibility of the ends. (2) Desire is not anchored by any other
aspect of the decision problem, whether the feasible set or the context of
choice. Formally, desires are captured by a preference ordering on possible
objects of choice whose existence is independent of the feasible set or the
context of choice. This is the content of GCT.
The tension between the demands for a parsimonious behavioural theory
and the need for an intentional theory of choice is often resolved by holding
that, of course beliefs and desires exist, but we economists have access to
them only as they are revealed in observed choice behaviour. In a recent
critique of neuroeconomics, two well-known theorists write, ‘In standard
economics, the testable implications of a theory are its content; once they are
identified, the non-choice evidence that motivated a novel theory becomes
irrelevant’ (Gul and Pesendorfer, 2005, p. 6). This view has a long history,
perhaps with origins in the defence of marginal analysis against its early
critics. Machlup (1946, p. 537) writes, ‘Psychologists will readily confirm that
statements by interviewed individuals about the motives and reasons for their
actions are unreliable or at least incomplete’, and also raises the oft-heard
incentive problem of eliciting survey data, namely, that survey respondents
may choose answers to meet their own goals, which may not include truth.
One source of confusion in evaluating claims for and against the econ-
omist’s psychology is that the theory has both positive and normative com-
ponents. According to Marshak (1950, p. 111), ‘The theory of rational
behavior is a set of propositions that can be regarded either as idealized
approximations to the actual behavior of men or as recommendations to be
followed’. Savage’s early work with Milton Friedman (1948; 1952) was ex-
plicitly descriptive, but Savage (1956) is just as explicitly normative. It is not
surprising that a description of decision in terms of beliefs and desires should
have a normative component which evaluates how well goals are achieved.
Confusion arises, however, when the descriptive and prescriptive positions
are inappropriately conflated to justify the rationality assumptions as a
statement of fact. Many undergraduate microeconomics texts justify tran-
sitivity assumptions by a money pump argument as a prelude to demand
theory. The Dutch book is used to defend probabilistic descriptions of belief.
But both of these arguments are, at their source, explicitly normative (see
Davidson, McKinsey and Suppes, 1955, p. 146; Ramsey, 1931).
A descriptive theory of choice which is grounded not in empirical reality
but in logical deductions from normative principles, like Dutch books and
money pumps, is not science, but metaphysics. Furthermore, normative jus-
tifications are implicitly introspective. A money pump argument really says,
rationality 7

‘you wouldn’t fall into this trap, would you?’ Significant empirical work in
psychology (Nisbett and Wilson, 1977), however, indicates that introspective
evidence is simply unreliable. When individuals turn to review and justify
their decisions, they may have no access to the mental states which guided
their choice. On the other hand, it seems to us quite reasonable to build
models of financial asset pricing which assume that traders are probabilis-
tically sophisticated, on the supposition that traders who are not will either
not long survive in the market or not, as a group, be large enough to have a
significant effect on prices. Financial markets, unlike Dutch books, actually
exist, and the claim that individuals with probabilistically incoherent beliefs
do not fare well is a claim of fact, to be tested against market data.
The conflation of positive and normative concerns in decision theory is
more fundamental than simple carelessness in an argument. In his criticism
of the fact/value dichotomy, Putnam (2002) asks us to consider the word
‘cruel’. He observes that the word often has both descriptive and normative
content, and in most uses they cannot be separated. The same could be said
of the adjective ‘rational’ in economists’ usage. Marshak (1950, p. 111) il-
lustrates this perfectly when he writes that the purpose of EU is ‘y to
describe the behavior of men who, it is believed, cannot be ‘‘all fools all the
time’’y.’ When the word ‘rational’ is used to describe a system in which all
agents hold accurate probabilistic beliefs, the implication is that someone
holding inaccurate beliefs gets it wrong. REE is often informally defended by
the assertion that, if an economic actor’s beliefs were incorrect, he would
observe this and form new ones. The assertion is either a positive assertion,
that actors do indeed have such beliefs, or a normative assertion, that they
should hold such beliefs. The normative assertion is a metaphysical defence
of the validity of the rational expectations hypothesis. The positive assertion
is a claim of fact whose validity could in principle be put to test, but testing
the claim would in fact require so rich a set of ancillary maintained hypoth-
eses that practically it is infeasible.
Given all the problems of the two views of rationality, one might wonder
why economics needs a rational actor. Dennett (1971, p. 92) provides per-
haps the best defence of belief/desire explanations. He contrasts what he calls
the design stance, predicting behaviour from an understanding of how an
agent is designed, or built, with the intentional stance, attributing to the agent
beliefs and desires, and predicting from them. The intentional stance is use-
ful, he writes, ‘Whenever we have reason to suppose the assumption of op-
timal design is warranted, and doubt the practicality of prediction from the
design ystance’. Warranting the optimal design assumption means for
Dennett not that the system actually be designed to achieve a fixed set of
goals, but that this assumption is a useful first approximation. ‘Not surpris-
ingly’, he observes,
as we discover more and more imperfections y, our efforts at inten-
tional prediction become more and more cumbersome and undecid-
able, for we can no longer count on the beliefs, desires, and actions
going together that ought to go together. Eventually we end up, fol-
lowing this process, by predicting from the design stance; we end up,
that is, dropping the assumption of rationality. (p. 95)

This movement, from rationality to realism, is the motivation for taking

behaviour more seriously.
8 rationality

Rationality and behaviours

Game theory and general equilibrium theory are ‘system frameworks’. They
imagine a collection of individual agents interacting in some systematic way,
strategically in game theory, as described by the normal or extensive form of
the game, and through markets in general equilibrium theory. In each case,
the model produces an ‘equilibrium’ of the system. The first stage in the
development of a system framework involves determining its consistency and
internal coherence, that is, conditions which guarantee the existence of equi-
librium. This analysis will be as abstract and general as possible, to encom-
pass as large a repertory of behaviours as possible. The second stage is the
application of the framework to derive useful statements about the world.
This requires explicit behavioural assumptions about agent behaviour and
describing the resulting equilibrium. These statements – predictions about
market or game behaviour – can be examined empirically.
There are two difficulties with the received models of decision theory such
as expected utility and dynamic programming in this kind of research pro-
gram. First, as these models are formulated, behaviours are not accessible.
For example, using expected utility to derive home bias in financial asset
markets – that is, investors tend not to take positions in foreign assets –
requires complicated assumptions about traders’ beliefs. Second, these mod-
els are insufficiently rich to capture all the behaviours one might want to
examine. For instance, the additively separable intertemporal expected utility
model conflates time preference and risk aversion because the model is too
thinly parametrized.
Behavioural economics is a research program which will, its proponents
argue, replace rational actor models with a more psychologically informed
view of human decision making. Much of behavioural economics, however,
is less ambitious (and thus, perhaps, more useful). This work can be de-
scribed as reformulating or extending rational actor models so as to make
those observable behaviours whose implications we wish to examine more
accessible. While much of this work is at the core of behavioural economics,
many who do this work eschew the label; not only behavioural economists
are interested in behaviour. Here we discuss four categories of research which
cover much work on behaviours, both by behavioural economics and by its
critics.

Recontextualizing decision
GCT is a very parsimonious simplification of a decision problem. In mod-
elling there is a trade-off between behavioural accuracy and parsimony in the
description of decision problems. In general equilibrium models, for exam-
ple, behavioural accuracy may improve descriptive and explanatory power,
but parsimony is required because individual decisions are only one piece of
the analysis, and complicated models of individual behaviour may generate
only intractable market models.
One implication of GCT is that preferences are not choice-set dependent.
Even in the early days of decision theory, important models such as minimax
regret (Savage, 1951) violated the requirement of a single preference order on
a universal space of potential choices. Furthermore, many choice-set effects
appear to be perfectly rational. Consider the behaviour of a well-mannered
but very hungry person at a dinner party. A plate is passed to him with three
pieces of the main course, ordered in size such that a  b  c. Being both
well-mannered and hungry, he chooses the second largest piece, b. Suppose
rationality 9

now that the plate had been passed around the table in the other direction, so
that when it comes to him there remains only a and b. Now according to his
rule he chooses a. Is he called irrational by the GCT theorists at the table?
Kahneman and Tversky’s (1979) prospect theory illustrates another way in
which decision problems can be recontextualized. Here additional context, a
status quo, is added to the description of the decision problems. Gambles are
viewed as probability distributions over gains and losses relative to the status
quo. Given a status quo, a preference order over all possible final outcomes
exists, but that preference order varies with the status quo. There is, however,
a stable preference order over the universe of all possible gains and losses;
more context is added by redefining the objects of choice. A similar trans-
formation is accomplished in Gul and Pesendorfer’s (2004) model of choice
with self-control problems. In the conventional infinite-horizon optimal
consumption problem, the objects of choice are consumption paths. Gul and
Pesendorfer, on the other hand, take the objects of choice to be pair con-
sisting of a current period consumption and a decision problem to be solved
tomorrow. Gul and Pesendorfer’s model is an example of a menu choice
model. Although used somewhat earlier, the first formal development of such
models was by Kreps (1979) to describe preferences for flexibility.

Constructing rationality
The economist’s conventional view of market interaction posits a collection
of individuals with well-formed preferences meeting in a marketplace. The
preferences, along with endowments and technologies, are exogenous to the
system. On the other hand, some attendees at a large outdoor concert are
there because they like the music, while others are there because of the
crowd. Teenagers’ evaluation of clothing style has perhaps as much to do
with who wears such clothes as with their cut and pattern. These are all
examples of socially constructed preferences.
Socially constructed preferences are a part of conventional economic the-
ory. Both NGT and REE are models of socially constructed preferences. In
each case desires are fixed, but beliefs adjust. However, neither of these
equilibrium concepts is particularly well-supported by belief adjustment
(learning) processes. The literature on learning Nash equilibrium is huge, and
the state of the art is that, while one can construct learning dynamics that will
find a Nash equilibrium, many intuitive learning processes will often fail.
Blume and Easley (1982) show that rational equilibrium can easily fail to be
reached by any reasonable learning process.
Restricting the socially determined component of preferences only to be-
liefs is an artificial constraint, and to limit social influence on preference
formation to learning is to miss most of the interplay between the individual
and the group. Manski (2000) observes that the implications of social in-
teractions through learning and through tastes are distinct, and the difference
is significant for policy analysis. Any theory of the interaction of desires
requires a new set of primitives which describe the preference formation
mechanism. One popular approach has been to model the evolution and
workings of pro-social norms of cooperation and trust. Bowles (1998) is an
engaging survey of this work. Much less has been done on the evolution and
workings of anti-social norms, such as discrimination and stigmatization.
Others have turned to biological metaphors. Here one might look at the
population dynamics of rules or preferences on a game form or market where
game or market outcomes (not utilities) determine the composition in the
next round of the population’s decision rules or preference orders (Güth and
10 rationality

Kliemt, 1998; Blume and Easley, 1992; 2006). Pro-social behaviour such as
reciprocity and altruism has also been investigated from the biological
standpoint (Bergstrom, 2002; Sethi and Somanathan, 2003). One lesson of
this literature is that the nature of the interaction between agents is at least as
important as the model of choice in determining system outcomes. About the
embeddedness of economic action in social life, Granovetter (1985, p. 506)
writes,

The notion that rational choice is derailed by social influences has long
discouraged detailed sociological analysis of economic life and led re-
visionist economists to reform economic theory by focusing on its naive
psychology. My claim here is that however naive that psychology may
be, this is not where the main difficulty lies – it is rather in the neglect of
social structure.

The content of preferences and beliefs

It has been conventional in economic analysis to construe self-interest very
narrowly. No ‘other-regarding’ values are expressed in preferences, and
conventionally to do otherwise is frowned upon. For instance, it is hard to
explain why an individual votes in an election by her effect on the outcome,
without referring to the psychic rewards of the act of voting. Yet the claim
that people vote because of norms of citizenship and the like is often re-
garded as ‘nearly tautological’ (Ordeshook, 1986, p. 50.). On the other hand,
critics of economic man often incorrectly assert that rational actors are ex-
cessively self-interested; incorrectly, because the existence of preferences and
the content of preferences are distinct issues. The rationality hypothesis does
not preclude other-regarding desires. Interest in those externalities that arise
from ethical concerns, social norms, and other social constructions has in-
creased enormously since the mid-1990s. Much of the literature on social
interactions is a study of the consequences of other-regarding preferences.
Not surprisingly, other-regarding preferences usefully model both altruism
and racism. This is not a fix for those critics who see the selfishness of
traditional neoclassical models as a moral failing rather than a behavioural
one.
A distinct problem which, unfortunately, has not been much addressed by
behavioural economics is the use of individual preferences in the economist’s
version of moral philosophy. The same preferences which are revealed
through shopping behaviour at the grocery store are supposed to be in-
formative for the ethical questions posed by welfare economics. One could,
in fact, distinguish ‘ethical preferences’ from ‘subjective preferences’ as Ha-
rsanyi (1955) has done, and it would be interesting to know if social psy-
chology has anything to say about the relationship between the two types of
decision problems, individual and social, which economists address.

Different psychologies
Some economists look to replace the folk psychology of beliefs and wants
with something altogether different. Neuroeconomics is one such attempt,
although the neuroeconomics literature seems to eschew drawing economic
conclusions from imaging data. Unfortunately, the link between brain and
rationality 11

mind is elusive. Eliminative materialism is a position taken by some cognitive

scientists, which claims that beliefs and desires do not exist as mental states,
and will have no place in an accurate account of the mind. Theoretical and
methodological arguments in its support can be found, for instance, in
Churchland (1981). An economics which takes its microfoundations entirely
from cognitive science could look extremely different than the economics of
today. But even if one is more hopeful than Churchland for the utility of the
economist’s folk psychology, the goal is far off. As one leading neuroimaging
specialist puts it, ‘Despite fantastic technical developments, lingering meth-
odological and conceptual limitations hinder progress in understanding how
mental processes (wrapped up in folk psychology) reduce to or emerge from
neural processes’ (Schall, 2004, p. 44). Savoy (2001, p. 36) has a bleaker view:
Do the new discoveries about human brain function based on ne-
uroimaging experiments really teach us things that are relevant for the
study and understanding of behaviour? That is a question which you
must answer. My own impression is that, at present, the overwhelming
thrust of these data are toward understanding brain organisation,
rather than human behaviour. Of course, we assume that when brain
organisation is sufficiently well understood, it will lead to increases in
our understanding of behaviour. But I do not think, as yet, there is a
great deal of progress in that direction.

Neuroeconomics hopes to replace the belief/desire folk psychology that

informs most of modern analytical economics with a more accurate scientific
psychology. Alternatively, one could construct a different folk psychology
which, like utility maximization, has no scientific pretensions, but is more
descriptively accurate. Models of intrapersonal conflict are the most familiar
example of this kind of framework. Strotz (1955) demonstrated the possi-
bility of time-inconsistent planning in intertemporal utility maximization
problems, and Pollak (1968) subsequently displayed the essentially strategic
nature of the planning problem as a problem of competition between the
selves choosing at different dates. Schelling (1984) described a variety of
decision problems with aspects of intrapersonal conflicts, and discussed them
from a game-theoretic perspective. He, for instance, wrote about the com-
petition between that part of him which desires nicotine and that part which
wants to give it up. This is a contest for self-control. The literature today
contains a number of intertemporal models which, following Pollak (1968)
distinguish two kinds of behaviour. Sophisticated behaviour chooses today
with full knowledge that her future selves may try to undo her decision. A
choice is a subgame perfect equilibrium of a game played by all her selves.
Naive behaviour chooses today assuming, perhaps incorrectly, that her fu-
ture selves will stick with her decisions. These two models are intrinsically no
more realistic than GCT, just different.
For the working economist, the ultimate test of a psychologically more
accurate theory of individual choice is how it performs in explaining market
and other social outcomes rather than how well it predicts the behaviour of
an individual. Could theory A, more informed with insights from psychology
and cognitive science, possibly be less useful for economists? Here are three
possibilities: (1) Theory A might be extremely complex. Its application to a
heterogeneous-agent financial market model, for instance, is simply impos-
sible to work with, and no conclusions can be derived. (2) Theory A might
require for its application data that we can observe in a controlled and
heavily instrumented setting but simply cannot collect in the field. (3) Theory
A may not be posed with concepts which are useful for the economist’s
12 rationality

interpretation of social outcomes. For instance, theory A might be couched

in terms of chemical states of the brain, and not speak at all about agents’
intentions, beliefs or desires. While it may be possible to construct a bio-
chemical model of the invisible hand, it would not be useful for welfare
economics.
Evidence on the question of whether these models lead to better market
analyses is sparse, and mixed, and there is no evidence on how these models
perform relative to menu choice models, which address the same questions
from a rational choice perspective. More generally, more work needs to be
done in evaluating behavioural models with respect to their economic per-
formance. How useful are they for deriving implications about the perform-
ance of aggregate economic variables such as prices? This kind of research is
already under way. Two examples are Kocherlakota (2001) and Laibson
(1997).
An instance of point (3) can be seen in the time-inconsistency literature.
Pollak (1968) and his followers (O’Donoghue and Rabin, 1999) see choice
not as the expression of a single desire, but as the outcome of conflict,
perhaps inefficient and destructive, between competing desires. Now the
Pareto ranking of alternatives in a social interaction either becomes depend-
ent on which of the many competing preference orders we modellers choose
for each individual or it becomes empty if we try to respect them all. The
advantage of menu choice models, the rational-choice approach to modelling
problems of self-control, is that there is a well-defined notion of preference
for each agent, from which a Pareto ranking can be constructed. To be fair,
rational choice modelling also poses problems for welfare economics. If in-
dividuals make consistent errors in a class of choice problems, what can
revealed preference say about intentions? In the presence of systematic error,
a welfare economics built from revealed choice is at best misleading.

Conclusion
The purpose of decision models in economics is to explain the behaviour not
of a single individual but of aggregates of individuals. Sometimes economists
explain by appeal to ‘Laws’, such as ‘the Law of Supply and Demand’. But
this mode of explanation is mostly an intermediate product; useful, perhaps,
for generating back-of-the-envelope predictions about the effects of a tax on
market price, but not a source of understanding. There are few natural laws
in the social sciences, and the domains of the few we can identify are very
limited.
More often, economists appeal to ‘mechanism’. We try to understand
economic phenomena, such as the determination of prices in different kinds
of markets, in terms of the mechanisms which generate them. Given our
commitment to methodological individualism, this requires an explanation
of how individual economic actors interact with one another. This is where
rational actor theories are employed, and it is with respect to how these
models do in this discussion rather than how they do in other domains, such
as explaining individual behaviour, that the rationality principle should be
evaluated.
Unfortunately, perhaps, at this point there are no serious alternatives to
the rationality principle. For all of its buzz, proponents of bounded ration-
ality, by which we mean models of behaviour that consider beliefs and desires
but that do not optimize, have so far failed to deliver decision models which
are robust and not tightly tied to a small class of decision problems.
rationality 13

It is perhaps too early in its intellectual history to ask for as much from
cognitive models. We are sceptical about the value for social and economic
systems analysis of unpacking the black box of consumer behaviour by de-
ploying a rich and sophisticated model of cognitive process within a general
equilibrium or game theoretic model. There is a point to reductionism. On
the other hand, we are enthusiastic about the possibility that cognitive sci-
ence will contribute to sharpening the rationality principle. The focus of
much modern decision theory, such as Kahneman and Tversky (1979), Gil-
boa and Schmeidler (1989) and Gul and Pesendorfer (2004), has been to
make the black-box model better by looking for formulations of rational
choice models that better conform to the data. A better understanding of
decision mechanism will doubtless suggest constraints on black-box behav-
iour which can be captured in reduced-form decision models, and perhaps it
will uncover constraints that cannot be observed from behaviour alone.
Evolutionary models have also been proposed as an alternative framework
to rational choice decision-making. Market forces, or a combination of
markets and biology, favour some decision rules over others. In the long run,
the market will be populated mostly by those decision rules that are ‘most
fit’, rational or not. One can indeed ask if the forces of market selection
favour rational decision rules (Blume and Easley, 1992; Sandroni, 2000), but
the study of market population dynamics is complementary to rather than a
substitute for rational choice models. Blume and Easley (2006), for instance,
demonstrate how market forces select within the class of rational decision
rules, favouring some kinds of preferences and beliefs over others.
Although there appear no be no serious alternatives to the rational choice
paradigm on the near horizon, there is much to regret in how the rationality
principle is discussed. The following statements should be self-evident, but
clearly are not, judging by our reading of the literature: (1) Rationality does
not mean complete or symmetric information. In fact, much of rational actor
social science attempts to understand social outcomes when these conditions
do not obtain. (2) Rationality does not require individuals to be entirely
selfish. While much effort has been made to understand social norms from
the point of view of entirely individualistic preferences, the insistence on
relying on self-regarding rather than pro-social preferences is a matter of the
content of preferences, rather than an axiom of rationality per se. (3) Ra-
tionality does not mean expected utility. Expected utility is one small class of
decision models for choice under uncertainty. Its dominance in application
was understandable in the 1970s, when few alternatives were on the table.
Since then decision theorists have been creative in developing better-behaved
alternatives, and equilibrium and game theorists have been clever in applying
them. (4) Rationality does not mean ‘rational expectations’. For a belief
restriction to be a requirement of rationality, it must be clear that all those
who are not ‘all fools all the time’ must have correct beliefs. No research into
learning in economics suggests this is the case in any kind of complex en-
vironment.
There is also much to regret in how the rationality principle has been
deployed in economic analysis. Given the explosion of decision-theoretic
research since the 1970s, it is surprising how little this research has affected
market and game theoretic analysis. The norm still seems to be self-interested
preferences, expected utility and rational expectations (or Nash equilibrium).
At this point the question of whether contemporary decision models such as
Choquet expected utility and cumulative prospect theory have anything new
to say about, say, asset pricing, is open. The value to economists of new
decision theories, rational choice or not, is not in how they perform in a
14 rationality

laboratory but how they perform in the analysis of markets and other social
systems. Too rarely have modern decision theories been exposed to this test.
Rational actor social science is a broader tent than both its supporters and
its critics make it out to be. We expect the rational choice framework to be as
dominant when the next edition of the New Palgrave goes to press as it is
today. But we also expect the set of decision-theoretic models deployed in the
analysis of social systems will be quite different, and probably more diverse,
than it is now.
Lawrence E. Blume and David Easley

See also

<xref=E000178> expected utility hypothesis;

<xref=M000352> methodological individualism;
<xref=R000256> rationality, history of the concept;
<xref=xyyyyyy> savage’s subjective expected utility model;
<xref=U000005> uncertainty;
<xref=U000073> utilitarianism and economic theory;
<xref=xyyyyyy> utility.

Bibliography
Allais, M. 1953. Le comportement de l’homme rationel devant le risque: Critique des
postulats et axiomes de l’Ecole Américaine. Econometrica 21, 503–46.
Aumann, R.J. 1962. Utility theory without the completeness axiom. Econometrica 32,
445–62.
Bentham, J. 1789. An Introduction to the Principle of Morals and Legislations. Based
on information from English Short Title Catalogue. Eighteenth Century Collec-
tions Online. Farmington Hills, MI: Gale Group.
Bergstrom, T.C. 2002. Evolution of social behavior: individual and group selection.
Journal of Economic Perspectives 16(2), 67–88.
Blume, L. and Easley, D. 1992. Evolution and market behavior. Journal of Economic
Theory 58, 9–40.
Blume, L.E. and Easley, D. 1982. Learning to be rational. Journal of Economic The-
ory 26, 340–51.
Blume, L.E. and Easley, D. 2006. If you’re so smart, why aren’t you rich? Belief
selection in complete and incomplete markets. Econometrica 74, 929–66.
Bowles, S. 1998. Endogenous preferences: the cultural consequences of markets and
other economic institutions. Journal of Economic Literature 36, 75–111.
Chipman, J.S., Hurwicz, L., Richter, M.K. and Sonnenschein, H.F. 1971. Prefer-
ences, Utility, and Demand. New York: Harcourt Brace Jovanovich.
Churchland, P.M. 1981. Eliminative materialism and the propositional attitudes.
Journal of Philosophy 78(2), 67–90.
Cyert, R.M., Simon, H.A. and Trow, D.B. 1956. Observation of a business decision.
Journal of Business 29, 237–48.
Davidson, D., McKinsey, J.C.C. and Suppes, P. 1955. Outlines of a formal theory of
value, I. Philosophy of Science 22, 140–60.
Dennett, D.C. 1971. Intentional systems. Journal of Philosophy 68(4), 87–106.
Fishburn, P.C. 1991. Nontransitive preferences in decision theory. Journal of Risk and
Uncertainty 4, 113–34.
Fisher, I. 1892. Mathematical Investigations in the Theory of Value and Prices. New
Haven: Connecticut Academy of Arts and Sciences.
rationality 15

Friedman, M. and Savage, L.J. 1948. The utility analysis of choices involving risk.
Journal of Political Economy 56, 279–304.
Friedman, M. and Savage, L.J. 1952. The expected-utility hypothesis and the meas-
urability of utility. Journal of Political Economy 60, 463–74.
Gale, D. and Mas-Colell, A. 1975. An equilibrium existence theorem without ordered
preference. Journal of Mathematical Economics 2, 9–15.
Gilboa, I. and Schmeidler, D. 1989. Maximin expected utility with a non-unique
prior. Journal of Mathematical Economics 18, 141–53.
Granovetter, M. 1985. Economic action and social structure: the problem of embed-
dedness. American Journal of Sociology 91, 481–510.
Gul, F. and Pesendorfer, W. 2004. Self-control and the theory of consumption.
Econometrica 72, 119–58.
Gul, F. and Pesendorfer, W. 2005. The case for mindless economics. Working paper,
Princeton University.
Güth, W. and Kliemt, H. 1998. The indirect evolutionary approach: bridging the gap
between rationality and adaptation. Rationality and Society 10, 377–99.
Harsanyi, J.C. 1955. Cardinal welfare, individualistic ethics, and interpersonal com-
parisons of utility. Journal of Political Economy 63, 309–21.
Herstein, I. and Milnor, J. 1953. An axiomatic approach to measurable utility.
Econometrica 21, 291–7.
Kahneman, D. and Tversky, A. 1979. Prospect theory: an analysis of decision under
risk. Econometrica 47, 263–91.
Kocherlakota, N. 2001. Looking for evidence of time-inconsistent preferences in asset
market data. Federal Reserve Bank of Minneapolis Quarterly Review 25(3), 13–34.
Kreps, D.M. 1979. A representation theorem for ‘preference for flexibility’. Econo-
metrica 47, 565–78.
Laibson, D. 1997. Golden eggs and hyperbolic discounting. Quarterly Journal of
Economics 112, 443–77.
Machlup, F. 1946. Marginal analysis and empirical research. American Economic
Review 36, 519–54.
Manski, C.F. 2000. Economic analysis of social interactions. Journal of Economic
Perspectives 14(3), 115–36.
Marshak, J. 1950. Rational behavior, uncertain prospects, and measurable utility.
Econometrica 18, 111–41.
Mas-Colell, A., Whinston, M.D. and Green, J.R. 1995. Microeconomic Theory. Ox-
ford: Oxford University Press.
Nisbett, R.E. and Wilson, T.D. 1977. Telling more than we can know: verbal reports
on mental processes. Psychological Review 83, 231–59.
O’Donoghue, T. and Rabin, M. 1999. Doing it now or later. American Economic
Review 89, 103–24.
Ordeshook, P.C. 1986. Game Theory and Political Theory: An Introduction. Cam-
bridge: Cambridge University Press.
Pareto, V. 1895. Considerazioni sui Principi Fondamentali dell’Economia Politica
Pura, part V. Giornale degli Economisti, ser. 2(5), 119–57.
Pascal, B. 1672. Pascal’s Pense´es. New York: E.P. Dutton and Co, 1958.
Pollak, R.A. 1968. Consistent planning. Review of Economic Studies 35, 201–8.
Putnam, H. 2002. The entanglement of fact and value. In The Collapse of the Fact/
Value Dichotomy and Other Essays. Cambridge, MA: Harvard University Press.
Ramsey, F.P. 1931. Truth and probability. In The Foundations of Mathematics and
other Logical Essays, ed. R.B. Braithwaite. London: K. Paul, Trench, Trubner and
Co.
Samuelson, P.A. 1938a. A note on the pure theory of consumer’s behavior. Econo-
mica 5(17), 61–71.
Samuelson, P.A. 1938b. The empirical implications of utility analysis. Econometrica 6,
344–56.
Samuelson, P.A. 1947. Foundations of Economic Analysis. Cambridge, MA: Harvard
University Press.
Sandroni, A. 2000. Do markets favor agents able to make accurate predictions?
Econometrica 68, 1303–42.
Savage, L.J. 1951. The theory of statistical decision. Journal of the American Sta-
tistical Association 46(253), 55–67.
Savage, L.J. 1956. The Foundations of Statistics, 2nd edn. New York: Dover, 1972.
Savoy, R.L. 2001. History and future directions of human brain mapping and func-
tional neuroimaging. Acta Psychologica 107(1–3), 9–42.
Schall, J.D. 2004. On building a bridge between brain and behavior. Annual Review of
Psychology 55, 23–50.
16 rationality

Schelling, T. 1984. The intimate contest for self-command. In Choice and Conse-
quence: Perspectives of an Errant Economist. Cambridge, MA: Harvard University
Press.
Sethi, R. and Somanathan, E. 2003. Understanding reciprocity. Journal of Economic
Behavior and Organization 50, 1–27.
Smith, A. 1789. An Inquiry into the Nature and Causes of The Wealth of Nations, 5th
edn. Chicago: University of Chicago Press, 1976.
Strotz, R.H. 1955. Myopia and inconsistency in dynamic utility maximization. Review
of Economic Studies 23(3), 165–80.
von Neumann, J. and Morgenstern, O. 1947. Theory of Games and Economic Be-
havior, 2nd edn. Princeton: Princeton University Press.
Wald, A. 1939. Contributions to the theory of statistical estimation and testing hy-
potheses. Annals of Mathematical Statistics 10, 299–326.

Index terms

altruism
behavioural economics
Bentham, J.
bounded rationality
choice under uncertainty
cognitive models
consistency
cooperation
economic laws
economic man
evolutionary models
expected utility
general choice theory
hyperplanes
interpersonal utility comparisons
marginal utility
mechanisms
menu choice models
methodological individualism
neuroeconomics
non-cooperative game theory
non-expected utility decision theory
preference formation
preference relation
prospect theory
rational choice
rational expectations
rational expectations equilibrium
rationality
rationality principle
reciprocity
revealed preference theory
Savage’s subjective expected utility model
self-interest
social norms
social preferences
statistical decision theory
transitivity
trust
utilitarianism
rationality 17

utility

Index terms not found:

economic laws
non-expected utility decision theory