Preferences and Positivist Methodology in
Economics

Christopher Clarke∗

Forthcoming in Philosophy of Science

Abstract

I distinguish several doctrines that economic methodologists have found at-
tractive, all of which have a positivist flavour. One of these is the doctrine
that preference assignments in economics are just shorthand descriptions of
agents’ choices. Although most of these doctrines are problematic, the lat-
ter doctrine about preference assignments is a respectable one, I argue. It
doesn’t entail any of the problematic doctrines, and indeed it is warranted
independently of them.

1 The Legacy of Positivism in Economics

In the middle decades of the twentieth century, economic methodologists began
to advance several doctrines that had a strikingly positivist flavour. This paper will
examine an organized package of seven such doctrines:

(1) The ultimate, defining aim of the sciences is merely: to describe the ob-
servable features of the world, and to explain these features by reference
to other observable features of the world. (2) In fact, the defining aim of
economics is merely to explain agents’ choices by reference to external con-
straints such as prices, taxes, technology shocks and the availability of natu-
ral resources. (And by this I mean both the choices made by a single agent,

∗I am indebted to Anna Alexandrova, Christopher Cowie, and Zina Ward for their invaluable
comments on an ancestor of this manuscript; to Ken Binmore and Samir Okasha for many fruitful
discussions on the topic; and to several anonymous referees for their helpful comments. This work
has received funding from the European Research Council under the European Union’s Seventh
Framework Programme (FP7/2007-2013)/ERC Grant agreement no 284123.

1



and the aggregated choices of multiple agents.)1 It follows that (3) economic
models are just hypotheses about agents’ choices given the external circum-
stances. Admittedly, many economic models appear to describe agents’ psy-
chology—including their preferences, expectations, knowledge and the like.
But such models are not to be taken at face value. (4) What’s more, hy-
potheses are only to be tested by their observable implications. So (5) the
only evidence that bears on economic models is standard choice data: the
choices that agents make, and the external circumstances under which they
make them. Cognitive and neurobiological data, in contrast, are irrelevant
to testing economic models—even if these data show the psychological as-
sumptions behind economic models to be false. Fortunately (6) there is a
practical, theory-independent procedure for observing agents’ choices. This
is what allows economists to test their models, and to do so independently
of psychological theory. But it turns out that basic economic theory can ac-
commodate the currently available data about agents’ choices. So (7) basic
economic theory is well-confirmed.2

This is somewhat of a caricature, and indeed these doctrines are less popular
in some quarters of contemporary economics than they once were in the mid
twentieth century (Kahneman, Wakker, and Sarin 1997; Camerer, Loewenstein,
and Prelec 2005; Rubinstein and Salant 2008). Nevertheless, some economists
today are happy to champion the above package of ideas almost in its entirety
(Gul and Pesendorfer 2008).

I follow the majority of philosophers however in thinking that many of these
positivist ideas are deeply mistaken. I’d deny (1), that the defining aim of the
sciences is merely to explain the observable in terms of the observable. I think (2)
is overstated. I’m happy to think of economics as encompassing many defining
aims other than to explain choices by reference to external circumstances. I’d also
deny (4), that hypotheses are only to be tested by their observable implications;
and accordingly I’d deny (5) that only standard choice data bear on economic
models. I’d also deny (6) that there is a practical, theory-independent procedure
for observing agents’ choices. Finally, I have doubts about what the data show, and
so I’m reluctant to take a position on (7), the claim that basic economic theory is
well-confirmed.

Nevertheless, I think that doctrine (3) is defensible: economic models are just
hypotheses about agents’ choices given the external circumstances. Although
many economic models appear to describe agents’ preferences, expectations and
the like, this is just a convenient shorthand for these choice hypotheses. (This

1For example, the level of unemployment is an aggregate derived from the choices of workers
and firms.

2See the citations throughout this paper at the point that each individual doctrine is discussed.

2



‘shorthand story’ about economic models remains an official doctrine of main-
stream economics, as codified in some of the leading textbooks in microeconomics
and game theory.)3

Now, many philosophers reject this shorthand story as a piece of outmoded
positivism. I will argue however that this critique is unfounded. Firstly, the short-
hand story does not issue in any of the problematic positivist doctrines (Sections
3 and 4). And, secondly, the shorthand story can be warranted entirely indepen-
dently of them (Section 5). To put it briefly: the warrant for the shorthand story
is to shield economic theory from misguided criticism from psychology. (But this
does not entail that it shields economic theory from all criticism from psychol-
ogy—far from it; see Section 4).

Thus the central aim of this paper is to present the shorthand story in such
a way that it’s sharply separated from the other positivist doctrines (Section 2).
This clears the path for an economic methodology that dispenses with these prob-
lematic doctrines, but that retains the shorthand story. Thus far, philosophers
have overlooked the possibility of doing this. I recommend this methodology to
economists as the best way of doing justice to their desire to shield economic the-
ory from misguided criticism.

I should note some limitations in the ambitions of this paper. Firstly, this
paper will defend the shorthand story for the paradigm case of assignments of
preferences within economic models. But it will leave aside the wider case of as-
signments of expectations and knowledge. Secondly, it will focus on economics as
a descriptive science rather than as a normative enterprise, a science of policy or
of rationality. Thirdly, limitations of space will also prevent me from discussing
each and every objection to the shorthand story. Most notably I will not address
Sen’s (1973) worry with respect to other-regarding agents, nor Hausman’s (2012)
argument that appeals to the fact that preferences typically need to combine with
beliefs in order to generate action.

Fourthly, the main aim of this paper is not to critique any of the more problem-
atic positivist doctrines. Fifthly, this paper will not aim to resolve the controversy
over ‘idealized assumptions’ in economic models: Take the case in which some
of a model’s content is false—at least at face value. Does this always hinder the
aims for which the model is used, or indicate that the model is not fit for pur-
pose; or can it in fact help to further these aims (Friedman 1953; Musgrave 1981;
Mäki 2002)? Instead of providing a general answer to such questions, this paper
will speak to the issue of ‘false assumptions’ only for the case of preference assign-
ments in economic models. So, of course, my discussion will not apply to those
macro-economic models, for example, that do not assign preferences to agents.

3See Mas-Colell, Winston, and Green (1995, 11), Varian (2005, 120), Varian (1992, 131–133),
Binmore (2007b, chaps. 1.4.2, 14.1–14.2), and also the citations in Section 2.

3



Nor will it apply to parts of economic models other than preference assignments.

2 Preference Assignments as a Shorthand

I want to sharply distinguish the shorthand story from the problematic doctrines
with which it is often conflated. To this end, this section will articulate the short-
hand story about preferences in economics in a non-standard way—in two re-
spects. Firstly, I will articulate the shorthand story in terms of the distinction
between belief and acceptance, a distinction familiar from general philosophy of
science. Secondly, I will exclude any extraneous elements or misleading termi-
nology—especially talk of ‘revealed preferences’.

Commonsense Preferences. Standard economic models, when taken at face value,
say that agents are maximizers of their preferences—their preferences as under-
stood by commonsense psychology. One can show that this entails the following
principle that connects these commonsense preferences and an agent’s choices:4

() Take any given agent, Eve for example. If Eve has the following
commonsense preferences then she would make the following choices; and
if she would make the following choices then she has the following common-
sense preferences.

Preferences: Eve weakly prefers outcome x to outcome y. (I use ‘weakly
prefers’ here in the standard sense to contrast with ‘strictly prefers’.)5

Choices: In some circumstance, Eve would choose to receive outcome x
guaranteed, despite the option of receiving outcome y guaranteed also being
on the menu.

As an illustration, imagine four circumstances in which Eve might find herself. In
the first circumstance she has the choice between lobster and steak, in the second
she has the choice between steak and salad, in the third between lobster and salad,
and in the fourth between all three outcomes. Suppose that, in all four circum-
stances, Eve would always choose the lobster if it’s on the menu; otherwise she
would choose the steak if it’s on the menu. If one knew of this pattern of choices
it would be natural to infer that Eve’s preferences over these three outcomes are:
Eve strictly prefers lobster to steak, and steak in turn to salad. Conversely, if one
knew these preferences, it would be natural to infer that Eve’s choices adhere to
the pattern above. And the  principle licenses these inferences: it says that

4Modulo the assumption that all permutations of outcomes constitute possible menus.
5To say that Eve weakly prefers A to B is to leave open the possibility that she also weakly

prefers B to A. Intuitively, it leaves open the possibility that she is indifferent between A and B. In
contrast, stipulate that Eve strictly prefers A to B if and only if she weakly prefers A to B, but not
vice versa (Mas-Colell, Winston, and Green 1995, chap. 1B).

4



Eve has the above commonsense preferences if and only if she would make the
above choices in the above circumstances.6

To be clear, this talk of circumstances is meant to include circumstances that
will never actually materialize. In this sense the above circumstances are ‘coun-
terfactual’, as are the above choices. (Qualification: these counterfactual circum-
stances are not meant to include those that are too far-fetched.)7

Technical Preferences. The next step is to use  as the template for a defini-
tion of an agent’s ‘T-preferences’. T-preference is my label for an agent’s prefer-
ences in a stipulative technical sense. These are to be contrasted with an agent’s
preferences as understood by commonsense psychology. Here’s the definition:

() Any given agent Eve weakly T-prefers outcome x to outcome y if
and only if, in some circumstance, Eve would choose to receive x guaranteed,
despite the option of receiving y guaranteed also being on the menu.

T-preferences are therefore defined as patterns in an agent’s choices across coun-
terfactual circumstances. Take for example the fact that Eve weakly T-prefers
lobster to steak. This fact is, by , the very same thing as the fact that there
is some circumstance in which Eve would choose lobster when steak is on the
menu.

The Shorthand Story. We are now in a position to precisely state the shorthand
story, a story about economic models that assign commonsense preferences to
agents. The shorthand story says that the content of such preference assignments
is just the corresponding assignment of T-preferences. As I will explain in a mo-
ment, another way of putting this is:

() Take any economist who accepts that Eve weakly prefers outcome x
to outcome y. This acceptance is just the economist’s belief that Eve weakly
T-prefers x to y.

In other words, applying : to accept this model is just for the economist
to believe that in some circumstances Eve would choose to receive x guaranteed,
despite the option of receiving y guaranteed being on the menu. So, more gener-
ally, for an economist to accept an assignment of commonsense preferences to an
agent is merely for her to believe that the agent’s counterfactual choices conform
to a particular pattern.8

6See footnote above to see how to translate between talk of weak preferences and strict prefer-
ences.

7A circumstance is not too far-fetched, I stipulate, if the agent’s commonsense preferences are
the same, in that circumstance, as they actually are. (Obviously, this precludes  from offering
a reductive definition of T-preferences. See extensive discussion of reductive doctrines in Section
3.)

8In addition to the citations in Section 1, refer to Friedman and Savage (1952), Binmore (2007a,
321), Binmore (2009, 14), Gul and Pesendorfer (2008), Bernheim and Rangel (2008), Wakker
(2010, 366), and Gilboa (2010, 20).

5



Take any economist, for example, who models Eve as a maximizer of her com-
monsense preferences. One can show that, according to , the content of this
model is roughly the following hypothesis called the ‘weak axiom’:9

Weak Axiom: Suppose that in some circumstance Eve would choose to re-
ceive outcome x (guaranteed) despite the option of receiving outcome y (guar-
anteed) also being on the menu. And suppose that in some circumstance Eve
would choose y from a menu also containing x. Then there is also some
circumstance in which Eve would also choose x from that same menu.

The Face-Value Story. This shorthand story is to be contrasted with what I will
call the ‘face value’ story. According to the face value story, the content of the
preference assignment on the last page, for example, is just as it appears: Eve has
a weak commonsense preference for lobster over steak. According to the short-
hand story, in contrast, the content is: there is some circumstance in which Eve
would choose lobster when steak is on the menu. But I will follow most philoso-
phers in thinking that commonsense preferences are more than such facts about
an agent’s counterfactual choices. (For one thing, commonsense preferences are
often supposed to be the causes of such facts, not the facts themselves.)10 So the
face-value story and the shorthand story make contradictory claims about the
content of preference assignments in economic models.

Acceptance versus Belief. An important feature of many scientific models is the gap
between their face-value content and their real content. Take for example the
equation F = g.cos(θ). This abstract equation can be used as a model for the
motion of a ball on an inclined plane. F can be construed as the net force acting
upon the ball down the plane, θ as the angle of the plane, and g as Newton’s
gravitational constant. This assignment of the parts of the model to some parts
of the world treats the model as saying, at face value, that the plane is perfectly
frictionless and so gravity is the only force acting. So construed, I will call this the
face-value content of the model.

Note however that a physicist who accepts this model isn’t thereby commit-
ted to believing its face-value content. She needn’t believe the obvious falsehood
that the plane is perfectly frictionless. This highlights the crucial and intuitive
distinction between belief and acceptance: to accept a model is to be willing to
use that model for various purposes, but it need not involve believing the model’s
face-value content.11 Instead, in virtue of accepting the model, what the physi-

9More generally, it’s the hypothesis that an agent’s choices conform to the so-called strong
axiom (Mas-Colell, Winston, and Green 1995, chap. 1), which is a slightly bolder hypothesis than
the weak axiom.

10See Armstrong (1968, 84–85) and Fodor (1968, 68–71) for discussion.
11See Harman (1986, 46), Maher (1993, chap. 6), van Fraassen (1994), Kaplan (2002, 451–453),

and Frankish (2004, chap. 4) for elaboration of the distinction between belief and acceptance.

6



cist believes is that the frictional force is negligible (Musgrave 1981). Accordingly
let’s stipulate that the real content of a model, in a given context, is the content
of the beliefs that a scientist has in virtue of accepting the model in that context.
This allows the face-value content of a model to come apart from the model’s real
content. How an economic model appears to describe the world can differ con-
siderably from how the model itself is used to describe the world by a particular
group of economists in a particular context (Morgan 2012).

How then should one characterize the role in economic models of assumptions
about commonsense preferences? As I render it, the shorthand story says that the
role of these assumptions is one of concision: they provide a concise shorthand in
which one can formulate hypotheses about complex patterns in an agent’s choices
(such as the Weak Axiom, for instance). This role differs from the three roles of
‘idealized assumptions’ that have already been discussed in the literature (Mus-
grave 1981; Mäki 2000; Hindriks 2006). Firstly, the real content of assignments
of commonsense preferences isn’t that a factor in the model is negligible; contrast
the factor of friction in the physicist’s model. Secondly, the real content of as-
signments of commonsense preferences isn’t that some relationship holds, but in
a limited domain. Contrast for example the hypothesis that F = g.cos(θ) holds,
but only for frictionless planes. Thirdly, assumptions about commonsense prefer-
ences aren’t required in order to make economic models mathematically tractable.
That is to say, to allow economists to calculate the deductive consequences of the
model’s assumptions. After all, economists have developed the mathematical tools
to express the theory of consumer demand, for example, solely in terms of coun-
terfactual choices, without the need to refer to preferences of any sort (Varian
1992, chaps. 8–9). Contrast for example the assumption in macro-economics
that there is only one consumer in the economy, an assumption which is needed
for mathematical tractability.

3 Theory-Independent Observation and Behaviourism

This section fends off some misplaced objections. It does so by pointing out that
the shorthand story does not entail positivist doctrine (6), nor does it entail be-
haviourism about commonsense psychology.

Observing Choices Independent of Theory. Positivist doctrine (6) says that there is a
practical procedure for conclusively verifying or falsifying any hypothesis about
any agent’s choices. And this procedure doesn’t presuppose any background the-
ory. In particular, one can observe an agent’s choices ‘directly’, independent of
any background psychological theory, it is claimed. On this basis, one might
reason as follows: ‘T-preferences are just patterns in an agent’s choices. So hy-

7



potheses about an agent’s T-preferences can be independently and conclusively
tested. But the shorthand story takes the content of preference assignments in
economic models to be hypotheses about T-preferences. So the shorthand story
helps such economic models to be independently and conclusively tested. And
this is a virtue of the shorthand story. Take for example a simple model in which
Eve weakly prefers lobster to steak. This model is conclusively established when
one comes to know that Eve chose lobster out of a menu that also contained steak.’

Whatever its shortcomings, this line of thought may have been one of the
original motivations behind the shorthand story in the 1930s and 1940s. In his
seminal paper advocating the shorthand story, Samuelson (1948, 251) contends:
‘The whole theory of consumer’s behaviour can thus be based upon operationally
meaningful foundations in terms of revealed preference’. And, when he talks of
‘operationally meaningful foundations’ here, Samuelson distinctly echoes Bridg-
man’s (1927) infamous insistence on conclusive, theory-independent observation
(Cohen 1995; Bruni and Guala 2001).

Now, in Section 5, I will argue that the shorthand story can be warranted
entirely independently of this line of thought. For the moment, however, I merely
want to register the fact that the shorthand story does not entail positivist doctrine
(6) above. To see this, note that the shorthand story treats the content of preference
assignments in economic models as hypotheses about choices. But it says nothing
about the manner in which (and the extent to which) economists can come to
know these choices. The shorthand story does not say that there is a practical
procedure to conclusively identify patterns in an agent’s choices, and indeed to do
so without presupposing any psychological theory.

This simple point is worth making because the shorthand story and positivist
doctrine (6) are often treated together under the label ‘the theory of revealed pref-
erence’. On the one hand, for example, Hausman, Rosenberg and other philoso-
phers take revealed preference theory to mean the shorthand story: ‘Taken se-
riously, [revealed preference theory] means that talk about preferences revealed
in behavior is just a useful fiction, a handy instrument. It is the most convenient
description of the behavior from which all the interesting results about markets
and economies follow’ (Rosenberg 1988, 87). ‘. . . revealed-preference theory de-
fines an agent’s preferences in terms of what the agent would choose if the agent were
able to choose’ (Hausman 2012, 31). On the other hand, when these philosophers
attack revealed preference theory, what they attack is something roughly similar
to positivist doctrine (6): ‘. . . the whole pretense of the behaviorists of revealed-
preference theory must be surrendered. For there is no way to read my beliefs from
my behavior except against some background assumptions about my preferences’
(Rosenberg 1988, 88). ‘Economists can infer preferences from choices or choices
from preferences only given premises concerning the agent’s beliefs’ (Hausman

8



2012, 30). (This is not quite an attack on doctrine (6) itself; for which see the cri-
tiques of revealed preference theory in Wong (1978), Sen (1993, 499–501, 515),
Rosenberg (1992, 123), and Cohen (1995).)

Thus these philosophers conflate the shorthand story with something akin to
positivist doctrine (6), a doctrine which Wong and others show to be false. And
so—whether or not they intend to do so—they give the strong impression that
the falsity of doctrine (6) is a decisive blow to the shorthand story itself. But this
conclusion is premature: although the falsity of (6) removes one potential reason
to favour the shorthand story, it does not decisively show that the shorthand story
itself is false.

Philosophical Behaviourism. Philosophical behaviourism states that commonsense
psychological facts can be characterized without using any psychological vocab-
ulary. For example, Adrian’s belief that it is raining might be characterized as
the tendency of Adrian’s body to move in particular ways, under various circum-
stances. For instance, his hand moves towards an umbrella before his body moves
outdoors. Thus the surprising behaviourist claim is that the facts of common-
sense psychology are just facts about motor movements given various external
circumstances—including counterfactual motor movements. In particular, then,
commonsense preferences are just patterns amongst these counterfactual motor
movements. But, behaviourists would add, motor movements are themselves
choices. So commonsense preferences are just patterns in an agent’s counter-
factual choices.

In the past, the popularity of the shorthand story may have been underpinned
by the popularity of philosophical behaviourism, or related behaviourist doctrines:
‘Samuelson and others employed [the shorthand story] to get purpose and inten-
tionality out of science. . . . For Samuelson and the behaviourists . . . since purpose
and intention are not observable in this sense, they have no place within science’
(Hands 2004, 961).

Now, in Section 5, I will argue that the shorthand story can be warranted en-
tirely independently from behaviourism. For the moment, however, I just want
to point out that the shorthand story itself does not entail behaviourism, a con-
troversial doctrine to say the least. Thus I take issue with Rosenberg (1988, 86),
for example, who gives the strong impression to the contrary: ‘The implicit ex-
cuse [for keeping psychology out of economics] is founded on a self-consciously
behaviorist interpretation of rational choice, the “theory of revealed preference”.’
And so, since behaviourism is probably false, Rosenberg seems to suggest that the
short-hand story is probably false too (88).

Does the shorthand story itself entail behaviourism, as Rosenberg seems to
suggest? Firstly, note that the shorthand story itself does not say anything about
the choices to which economic models refer. On the one hand, choices may be

9



construed in a commonsense way as intentional choice: for Eve to choose lobster
(guaranteed) requires that Eve know that her bodily movements will result in a
lobster arriving; or that she intend this to be the case. Alternatively, choices may
be construed in a behaviourist way: for Eve to choose lobster (guaranteed) merely
requires that her bodily movements will result in a lobster arriving, as a matter of
actual fact. Thus the shorthand story itself leaves open the nature of choice.

Secondly, consider the following analogy. Imagine a map which models the
British transport network. On this map London is connected to Bristol via a blue
line. And consider a story that says that the real content of such a blue line is the
hypothesis that London and Bristol are linked by a motorway. Assignments of blue
lines are just a shorthand, as it were, for hypotheses about motorways. Note the
following obvious fact: this story does not entail that motorways are blue lines.
To deny this would be to ignore the distinction between the model’s face-value
content and its real content. Now consider the analogous case in economics: the
shorthand story says that the real content of assignments of commonsense pref-
erences in economic models are hypotheses about an agent’s T-preferences. The
former is just a shorthand for the latter. But, by analogy, this story does not entail
that commonsense preferences are T-preferences. In other words, the shorthand
story does not entail the behaviourist doctrine that commonsense preferences are
patterns of counterfactual choices. Again, this becomes clear once one distin-
guishes between the model’s face-value content and it’s real content.

Conclusion. This section has fended off some misplaced objections. It has done
so by pointing out that the shorthand story does not itself entail positivist doc-
trine (6), which says that there is a practical, theory-independent procedure for
observing agents’ choices. Nor does it entail behaviourism about commonsense
psychology. Given the ambiguities of their talk of ‘the theory of revealed pref-
erence’, Hausman and Rosenberg’s treatment obscures the possibility that the
shorthand story can be separated from these problematic doctrines.

4 Shielding the Status Quo from Criticism?

This section will argue that the shorthand story doesn’t issue in problematic pos-
itivist doctrines (4) (5) or (7).

To keep things simple I will start with a toy example. Imagine that some
researchers offer an individual—call him Conor—the choice of x cans of Coke
and y cans of Pepsi, versus x∗ cans of Coke and y∗ cans of Pepsi. And they re-
peat this experiment many times over the course of a year, each time varying
the quantity of Coke and Pepsi on offer, x x∗ and y y∗. Imagine that these re-
searchers discover that Conor chooses the first option over the second if and only

10



if 2
3
ln(x)+ 1

3
ln(y) > 2

3
ln(x∗)+ 1

3
ln(y∗). (Intuitively 2

3
ln(x)+ 1

3
ln(y) measures Conor’s

tendency to choose the bundle of x cans of Coke and y cans of Pepsi on offer.)
Suppose that these researchers find a similar pattern amongst the other nine-

teen subjects in their experiment. That is to say, they find that a subject’s pattern
of choices can be accurately described by the function a.ln(x)+(1−a)ln(y), where
a = 2

3
for Conor, a = 1

4
for Rose, and a = 1

2
for Helen, and so on, for example.

The details aren’t important. The point is just that it looks like individuals’ choices
amongst Pepsi and Coke can be summarized by a single number a. The higher
any individual’s a score, the more inclined that individual is to choose options with
more Coke than Pepsi.

This is exciting news for Dr Terzian, the chief economist at Coke: she now
knows the distribution of a scores amongst twenty experimental subjects. Ap-
plying standard statistical techniques, she can then estimate the distribution of a
scores amongst all soft drink consumers in the US. Call this estimate ‘the Terzian
model’.

Of course, twenty data points is not an ideal basis on which to estimate the
distribution of a values amongst millions of soft drinks consumers. Ideally one
would like a thousand data points, say. Unfortunately it is prohibitively expensive
to calculate the a scores of a thousand individuals using the above method, because
the method requires a year long experiment for each individual.

Imagine however that, during the initial experiment with twenty subjects, the
experimenters discover that an individual’s galvanic skin response, when drinking
a can of Coke, is an excellent predictor of their a score. Or alternatively one
might imagine that their EEG reading when drinking a can of Coke is an excellent
predictor of their a score. So the researchers now have a cheaper way of improving
the Terzian model: measure the galvanic skin responses or EEG responses of, say,
a thousand subjects.

In summary, the content of the Terzian model is just a description of indi-
viduals’ counterfactual choices. It’s content includes nothing specifically about
these individuals’ psychological states. Nevertheless this model can be tested us-
ing biological and neurobiological data, such as galvanic skin responses and EEG
responses. (And one can easily modify the example to illustrate how the Terzian
model could be tested by cognitive data and introspective data too.) This illus-
trates a general point: to treat the content of an economic model as purely about
choices given external circumstances does not shield the model from being under-
mined by cognitive, neurobiological and introspective data, and so on—at least
not completely shielded (Clarke 2014). In other words, the shorthand story does
not issue in positivist doctrine (5), that the only evidence that bears on economic
models is observations of agents’ choices, and the external circumstances under
which they make them.

11



Once this point is clearly illustrated, it becomes obvious, one might think.
Nevertheless Craver and Alexandrova (2008, 393) argue against the shorthand
story because they think that it ‘isolates economics from contributions from other
disciplines.’ Similarly Alexandrova and Haybron (2011, sec. 2) argue against the
shorthand story because ‘[positivist] epistemology, which eschews everything but
observable behavior—and thus throws out the baby, much of our best physics, and
all good sense with the bathwater—has long since been consigned to the flames
elsewhere in the academy.’ If I am right, then such philosophers are mistaken.
Why then was this mistaken reasoning so compelling?

The first reason is this: the shorthand story does indeed shield economic theory
from being undermined by some psychological data—on some occasions. Imag-
ine a psychologist in the Coke research group who complains: ‘Our laboratory
work shows that a.ln(x)+(1−a)ln(y) is an excellent model of individuals’ choices
between Coke and Pepsi, I agree. But I suggest that we start interpreting this
model as a description of an individual’s strength of desire for Coke and Pepsi,
desire being a posit of commonsense psychology. Now, I’ve examined the psy-
chological literature, and it turns out that people very rarely choose the option
that they desire mostly strongly. In fact, an individual’s choices are a poor guide
to their desires. Thus this psychological data from the literature casts considerable
doubt on our model, as a model of desire for Coke.’ One might imagine that the
other members of the research group respond: ‘No, we are going to continue to
interpret the a.ln(x) + (1 − a)ln(y) model as a model of choices, not as a model of
desires as you insist. So your psychological data are irrelevant in this instance.’ On
this occasion, the shorthand story does shield the model from being undermined
by psychological data.

Thus it is crucial, I suggest, to distinguish two ways in which one might try
to shield a model from being undermined. The first way is to insist that there
are limits on what sort of evidence can support or undermine a particular sort of
hypothesis. Call this an ‘epistemological restriction’. The second way is to limit
the hypotheses that form part of the real content of a model. Call this a ‘content
restriction’.

The Coke example highlights that the shorthand story itself is a content restric-
tion, rather than an epistemological restriction. The shorthand story is consistent
with the idea that all sorts of psychological data can support or undermine stan-
dard choice hypotheses. But the shorthand story insists that content of preference
assignments in economic models are just standard choice hypotheses, not distinc-
tively psychological hypotheses. (I will wait until Section 5 to motivate this content
restriction.) It follows that psychological data that bear on the latter hypothesis,
without bearing on the former, are irrelevant to testing economic models.

Now for a second reason why many philosophers have mistakenly thought that

12



the shorthand story issues in positivist doctrine (5). This reason is rather ironic:
many such philosophers seem to implicitly accept positivist doctrine (4) that hy-
potheses are only to be tested by their observable implications. That is to say, by
noting what observations an hypothesis deductively entails, and then discovering
whether or not these predicted observations are indeed observed (Friedman 1953;
Gul and Pesendorfer 2008). Now, the shorthand story says that a model’s content
is just an hypothesis about agents’ choices given external circumstances. It follows,
given (4), that (5): only data about an agents’ choices, and the external circum-
stances under which they make them, can support or undermine economic mod-
els. (This line of reasoning fails, however, because (4) is false. The Coke example
illustrates this nicely: a wide range of observations can support or undermine an
hypothesis, without the hypothesis deductively entailing those observations. See
also Clarke (2014, sec. 4).)

Finally positivist doctrine (7), the doctrine that basic economic theory is well-
confirmed. To examine this doctrine, it’s important to distinguish two things: it’s
one thing to think that a model has a particular content, and it’s another thing to
think that this content is true (or well-confirmed). Take, for example, a philoso-
pher who thinks that the real content of all scientific models is some hypothesis
about observables (van Fraassen 1980). For example, the real content of string the-
oretic models is that various measuring devices in a physics laboratory will give
particular readings. But let’s imagine that this philosopher doesn’t accept string
theoretic models. Obviously, therefore, this philosopher isn’t committed to the
truth of the corresponding hypothesis about these measuring devices.

Analogously, the shorthand story is committed to treating economic models
as merely hypotheses about patterns in agents’ choices. The real content of the
model of Eve as a preference maximizer, for example, is that Eve’s choices conform
to the Weak Axiom (Section 2). But, the shorthand story itself isn’t committed to
accepting this economic model, and so it isn’t committed to the truth of the Weak
Axiom. This illustrates, more generally, that the shorthand story isn’t committed
to the claim that basic economic theory is true (or well-confirmed). In other words,
the shorthand story doesn’t entail positivist doctrine (7). (Of course, one might
again wonder what the point of the shorthand story could conceivably be, if not
as part of a dogmatic defense of basic economic theory. I will wait until Section 5
to address this.)

This simple but important point goes unrecognized, however, because talk of
‘the theory of revealed preference’ in the literature conflates the shorthand story
with the Weak Axiom of revealed preference. Take for example Reiss (2013, 35),
who points out that agents ‘make all sorts of mistakes when they choose due to
inattentiveness, weakness of the will, or false beliefs.’ And so agents’ choices don’t
always adhere to the Weak Axiom. But the Weak Axiom is part of the content of

13



standard economic models, when interpreted strictly according to the shorthand
story (Section 2). Therefore, Reiss points out, the content of standard economic
models is false, when so interpreted. Reiss then concludes from this that ‘the
scientific value of the technical [shorthand] concept of preference as choice is
dubious’.

Reiss’ last move is too quick. Note crucially that people are not always maxi-
mizers of their commonsense preferences. This is for exactly the reason that Reiss
gives above: people ‘make all sorts of mistakes when they choose’. It follows that
the content of standard economic models is also false, when strictly treated at face
value as an hypothesis about commonsense preferences.12 So Reiss’ point above
does not tell in favour of the face-value story over the shorthand story. It only
appears to do so if one conflates the shorthand story with the Weak Axiom of
revealed preference.

In sum, this section has shown that the shorthand story doesn’t itself issue in
problematic positivist doctrines (4) (5) or (7).

5 The Rationale behind the Shorthand Story

What then is the rationale behind the shorthand story, if not to provide a suppos-
edly theory-independent and conclusive way to evaluate economic models (Sec-
tion 3) or to shield basic economic theory from all psychological criticism (Section
4)?

To set the scene, consider the economic model of the consumer, whose face-
value content is the following: (i) There are a range of commodities. And an
agent—call him Adam—chooses how much of each of these commodities to con-
sume. (ii) Adam has commonsense preferences over such ‘commodity bundles’.
In the version of the model that I want to consider, the model attributes to Adam
a maximally-specific assignment of preferences, HP . These preferences are stable
over time. (iii) Adam is an optimizer: he always chooses the most preferred bundle
on the menu of bundles available to him. So, on this basis, the model describes
which commodity bundle Adam would choose from any given menu of bundles.
It describes his counterfactual choices, HC. (iv) The menu of commodity bundles
from which Adam can choose is constrained by the prices of each commodity, by
Adam’s income and by the tax regime. Therefore this description HC of Adam’s
choices specifies how price, income and tax make a difference to Adam’s choices
as a consumer.

12Note from Section 2 that the face-value story applied to standard economic models is com-
mitted to . Note also from Section 5 that the face-value story entails that the content of
standard economic models is logically stronger than the content according to the shorthand story.

14



Let’s imagine however that much of what this model says at face value about
Adam’s psychology is false. (a) The model describes some of Adam’s common-
sense preferences incorrectly. In particular the model does not accurately describe
Adam’s preferences between enormous bundles, such as one thousand loaves of
bread versus nine-hundred apples. And it does not accurately describe his pref-
erences between minuscule bundles, such as one loaf of bread in a given month
versus one apple that month. (b) Adam’s commonsense preferences are not stable,
as assumed, but vary somewhat depending upon his mood and upon environmen-
tal cues. (c) Sometimes ‘interfering factors’ result in Adam choosing one bundle
over a second bundle, even though he has a commonsense preference for the sec-
ond bundle over the first. For example when Adam is tired or stressed he tends
to make mistakes. And, suffering from a clinical pathology, on some occasions
Adam compulsively buys kitchenware, rather than renting movies for instance.
This is despite the fact that Adam has a strong and stable commonsense prefer-
ence for movies over kitchenware. Indeed he has no use for kitchenware, nor does
it afford him any pleasure either during or after the purchase.

In spite of all this, it is perfectly possible that the model under consideration
accurately describes a large proportion of Adam’s counterfactual choices. Here’s
a few ways in which this is possible. In relation to c: imagine that the model de-
scribes Adam as possessing a stronger commonsense preference for kitchenware
than he in fact does possess. As a result the model describes more or less accu-
rately Adam’s consumption of kitchenware. This slight mis-description of Adam’s
commonsense preferences offsets, if you like, the mis-description of Adam as a
perfect optimizer of these preferences. In relation to b: imagine that the model
does not accurately describe Adam’s commonsense preferences at any moment
in time, preferences which are rather unstable. But instead the stable preferences
that the model attributes to him are a good ‘average’ over this range of unsta-
ble preferences. As a result, the model describes more or less accurately Adam’s
total monthly consumption of each commodity. In relation to a: imagine that
Adam’s income would never be low enough that only minuscule commodity bun-
dles would be on his menu of available bundles. Similarly it would never be high
enough that any enormous commodity bundles would be on the menu. As a
result, the model accurately describes his choices with respect to the menus of
commodity bundles that are in some sense ‘practically feasible’. In this respect it
does not matter that the model mis-describes Adam’s commonsense preferences
concerning minuscule and enormous bundles.

In summary, at face value, the above model of the consumer mis-describes
much of Adam’s psychology; for one thing, HP mis-describes Adam’s common-
sense preferences. Nevertheless the model’s description of Adam’s choices HC
is more or less accurate—at least when considered month by month, and with

15



respect to menus that are practically feasible. But recall that HC is in effect a de-
scription of how price, income and tax make a difference to Adam’s choices. So
the model describes these influences more or less accurately too.

Adam’s case can be used to highlight the virtues of the shorthand story. Con-
sider an economist who has two ultimate aims as far as Adam is concerned. Her
first aim is to describe Adam’s choices more or less accurately. Her second aim
is to describe Adam’s psychology more or less accurately. So the above model
is very good at meeting her first aim, but fails to meet her second aim. What
then should this economist’s attitude toward this model be? I will suggest that she
take an ambivalent, ‘aim-relative attitude’. In the context of pursuing her aim of
describing Adam’s choices, she should accept the model. In the context of pursu-
ing her aim of describing Adam’s psychology, she should reject the model. This
is just an extension of the idea, familiar from the modeling literature, that one’s
attitude towards a model will depend upon one’s aims (Section 2). Incidentally,
Daniel Kahnemann I think is an great example of an economist who switches in
this manner between treating economic models of decision-making as models of
choices, in one context, and then as models of an agent’s psychology in another
(Kahneman and Tversky 1979; Kahneman, Wakker, and Sarin 1997).

One can now proceed to argue as follows: (A) One of the defining aims of eco-
nomics is to describe how external circumstances such as price, income and taxes
make a difference to an agent’s or agents’ choices. But (B): it is methodologically
felicitous for a discipline to take an aim-relative attitude to its models. So there is a
context in which it is methodologically felicitous for economics to treat preference
assignments as just describing agents’ choices given the external circumstances.
But (C) this treatment is just what the shorthand story about preference assign-
ments () offers. Therefore  is methodologically felicitous, where one is
able to implement it coherently. But one is able to implement  coherently for
the overwhelming majority of preference assignments in current economic mod-
els.13 So this shorthand story provides a methodologically felicitous treatment of
such preference assignments.

I will now bolster this argument by lending further support to premise A and
B. (See Section 2 for premise C.)

Premise A claims that one of the defining aims of economics is to describe how
external circumstances make a difference to an agent’s or agents’ choices; for ex-
ample the consumption of fruit in Belgium aggregated over a month. Here are
some examples of the external circumstances I have in mind: the prices of each

13When is one able to apply  to a model without falling into logical contradiction? There
are a few pre-requisites. One of these is that the model takes agents to be optimizers, at face value
that is. And a second is that it all permutations of outcomes constitute possible menus. Contrary to
Hausman (2012), this is true of the over-whelming majority of current economic models, I claim.
This paper makes no claims about future economic models.

16



commodity in a market; the technological ability of firms to produce commodi-
ties by transforming other commodities; the initial endowment of commodities
amongst agents and firms; and the tax regime. For example, one of the defining
aims of economics is to describe how technology shocks and commodity shocks
make a difference to prices, and thereby to an agent’s or agents’ supply and de-
mand behaviour.14 Similarly one of the defining aims of economics is to de-
scribe how government intervention via taxation makes a difference to prices,
and thereby to an agent’s or agents’ supply and demand behaviour.

More controversially, to this list of external circumstances, one might also add
the availability and distribution of information—for example information about
the quality of a commodity (second-hand cars being the classic example) or about
other agents’ choices (whether one’s partner co-operated or defected in the last
round of an iterated prisoners dilemma for example). Thus one of the defining
aims of so-called informational economics is to describe the way in which agents’
choices are influenced by external circumstances such as the quality of commodi-
ties, and the choices of other agents. (Of course, a complete discussion of the
content of models in information economics would also have to say something
about the content of its assignments of knowledge and expectation—a task I have
had to set aside for another day.)

There may of course be many other defining aims of economics. For example,
one might think that it is a defining aim of economics to describe the cognitive and
neurobiological mechanisms that underlie decision-making (Camerer, Loewen-
stein, and Prelec 2005). Nevertheless it’s clear that one of the central defining
aims of economics is (also) to describe how external circumstances make a differ-
ence to an agent’s or agents’ choices.

Finally and most importantly, premise B, which I put forward as a prosaic
methodological guideline for model-based science. This premise says that it is
methodologically felicitous for a discipline to take an aim-relative attitude to its
models. To see this guideline in action, consider a climate model that, at face
value, divides the globe into regions, and that describes the rainfall in each re-
gion for each month in the next year. Let’s say that one of the defining aims of
climatology relates to tropical seasons. The aim is namely to describe how rain-
fall in the Southern tropics changes from month to month. And a second aim is
a cross-regional comparison. The aim is namely to describe how yearly rainfall
varies across the various regions of the globe.

A climate model might (seem to) be very good at achieving the aim of pre-
dicting tropical seasons, for example, and (seem to) be bad at achieving the aim
of cross-regional comparison. On these occasions it is methodologically felicitous

14See Varian (2005, chap. 22) and Mas-Colell, Winston, and Green (1995, chaps. 5C, 322) for
some textbook examples.

17



for climatology to take an ambivalent, aim-relative attitude to this model. This
involves treating the climate model, at one point, as a description just of seasonal
changes in tropical rainfall. In this context one accepts the model. And then,
at another point, one uses the model to describe just regional variance in yearly
rainfall. In this second context one rejects the model. (It is important to recognize
that this is not to deny that a model’s ability to describe seasonal change in the
tropics accurately isn’t sometimes strong evidence as to whether it describes cross-
regional variance accurately. As I discussed in Section 4, the sort of evidence that
bears on an hypothesis may be very broad.)

Note that premise B does not say that this way of proceeding is mandatory;
nor that it is always the best way on balance. It just says that it’s one desirable
way to proceed. As Section 2 illustrated, it’s a well-attested feature of the model-
based approach to science that the same model can be used in a number of dif-
ferent ways (Morgan 2012). Importantly, premise B is consistent with the idea
that some disciplines aim to describe unobservable features of the world. This is
in contrast to positivist doctrine (1). Similarly, premise B is consistent with the
idea that one of the defining aims of economics is to model agents’ psychological
states, and the cognition and neurobiology of decision-making. This is in contrast
to positivist doctrine (2). All that premise B requires is that one of the ultimate
aims of economics is—in addition to any aims pertaining to cognition and neuro-
biology—to describe how external circumstances make a difference an agent’s or
agents’ choices.

In summary, this section has made the warrant for the shorthand story entirely
explicit. The shorthand story allows economists to focus on pursuing one of the
central aims of economics, namely describing how external circumstances make
a difference to choice behaviour.

6 Conclusion

This paper has examined the shorthand story about assignments of preferences in
economic models (and it has put aside the wider issue of knowledge and expecta-
tion in economic models). According to the shorthand story, when an economist
assigns preferences to an agent, this is for the economist to believe that the agent’s
choices conform to a particular pattern (Section 2).

Section 3 pre-empted some misguided criticisms of the shorthand story, criti-
cisms that Hausman, Rosenberg and several others give the impression of endors-
ing. It pre-empted these criticisms by pointing out that the shorthand story should
be distinguished from positivist doctrine (6), which says that there is a practical,
theory-independent procedure for observing an agent’s choices. The shorthand

18



story does not entail doctrine (6); nor indeed does it entail behaviourism about
commonsense psychology.

Section 4 refuted the positivist doctrine (4) that models are only to be tested by
testing their observable implications. It follows—contrary to Alexandrova, Craver
and Haybron—that the shorthand story doesn’t issue in positivist doctrine (5), the
doctrine that the only evidence that bears on economic models is observations
of agents’ choices, and the external circumstances under which they are made.
What’s more—contrary to Reiss—the shorthand story itself is not committed to
endorsing basic economic theory, unlike positivist doctrine (7).

Thus the first take-home message is this: by carefully spelling out a judicious
version of the shorthand story, it becomes evident that the shorthand story doesn’t
entail positivist doctrines (4) (5) (6) or (7)—despite what its critics suggest.

The reason that is so important to distinguish these problematic doctrines from
the shorthand story, I contend, is that the failure to do so has prevented philoso-
phers of economics from credibly engaging with economists. When economists
see philosophers striving to criticize the shorthand story—an entirely respectable
part of economic methodology—they are less likely to take seriously our criticisms
of the genuinely problematic positivist doctrines.

But what is the warrant for the shorthand story? To answer this, I applied some
familiar ideas from the literature on modeling. My answer was that the shorthand
story allows economics to focus on one of its defining aims: to describe how ex-
ternal factors make a difference to choices. Thus the shorthand story shields eco-
nomic theory from being undermined by any cognitive and neurobiological data
and the like that is not relevant to this aim. It shields via ‘content restriction’ but
not by ‘epistemological restriction’ (Section 4). And this restriction is warranted
by the principle that it is methodologically felicitous for a discipline to take an
aim-relative attitude to its models (Section 5).

So the second take-home message is this: the shorthand story doesn’t need to
be warranted by appealing to any problematic positivist doctrines. In particular,
one does not need to claim that (1) the defining aim of the sciences is merely to
explain the observable in terms of the observable; or (2) the sole defining aim
of economics is to explain choices in terms of external circumstances. Similarly,
in the past, the motivation for the shorthand story may have been to provide a
theory-independent and conclusive way to test economic models (Section 3) or to
shield basic economic theory from all psychological criticism (Section 4). But this
needn’t be its rationale today.

This addresses the concerns of critics who charge that ‘it is unclear what [the
shorthand] reduction of preference to hypothetical choice would accomplish’ (Haus-
man 2008, 137). It rises to the challenge posed by Alexandrova and Haybron
(2011, sec. 3) who note that ‘what is absent is a good argument for why this is the

19



right view of economics.’ And it corrects Hausman’s (2000, 134) contention that
there was ‘never any reason to develop [the shorthand story] in the first place.
. . . There is nothing to be ashamed of in the notion of a subjective preference
ranking, and it is no good pretending that economics can make do without it.’

20



6 References

Alexandrova, Anna, and Daniel M. Haybron. 2011. “High Fidelity Economics.” In Elgar
Companion to Recent Economic Methodology, eds. John Davis and Wade Hands, 94–120.
Cheltenham: Edward Elgar.

Armstrong, David M. 1968. A Materialist Theory of Mind. London: Routledge.

Bernheim, B. Douglas, and Antonio Rangel. 2008. “Choice-Theoretic Foundations for
Behavioral Welfare Economics.” In Caplin and Schotter 2008, 155–92.

Binmore, Ken. 2007a. Does Game Theory Work? The Bargaining Challenge. Cambridge MA:
MIT Press.

. 2007b. Playing for Real: A Text on Game Theory. Oxford: Oxford University Press.

. 2009. Rational Decisions. Princeton NJ: Princeton University Press.

Bridgman, Percy Williams. 1927. The Logic of Modern Physics. New York: Macmillan.

Bruni, Luigino, and Francesco Guala. 2001. “Vilfredo Pareto and the Epistemological
Foundations of Choice Theory.” History of Political Economy 33:21–49.

Camerer, Colin F., George Loewenstein, and Drazen Prelec. 2004. “Neuroeconomics:
Why Economics Needs Brains.” Scandinavian Journal of Economics 106:555–79.

. 2005. “Neuroeconomics: How Neuroscience Can Inform Economics.” Journal
of Economic Literature 43:9–64. Revision of Camerer, Loewenstein, and Prelec (2004).

Caplin, Andrew, and Andrew Schotter, eds. 2008. The Foundations of Positive and Normative
Economics: A Handbook. Oxford: Oxford University Press.

Clarke, Christopher. 2014. “Neuroeconomics and Confirmation Theory.” Philosophy of
Science 81:195–215.

Cohen, Joshua. 1995. “Samuelson’s Operationalist–Descriptivist Thesis.” Journal of
Economic Methodology 2:53–78.

Craver, Carl, and Anna Alexandrova. 2008. “No Revolution Necessary: Neural Mech-
anisms for Economics.” Economics and Philosophy 24:381–406.

Fodor, Jerry. 1968. Psychological Explanation: An Introduction to the Philosophy of Psychology. New
York: Random House.

Frankish, Keith. 2004. Mind and Supermind. Cambridge: Cambridge University Press.

Friedman, Milton. 1953. “The Methodology of Positive Economics.” In Essays in Positive
Economics, 3–46. Chicago IL: University of Chicago Press.

Friedman, Milton, and Leonard J. Savage. 1952. “The Expected-Utility Hypothesis
and the Measurability of Utility.” Journal of Political Economy 60:463–74.

Gilboa, Itzhak. 2010. Rational Choice. Cambridge MA: MIT Press.

Gul, Faruk, and Wolfgang Pesendorfer. 2008. “The Case for Mindless Economics.” In
Caplin and Schotter 2008, 3–39.

21



Hands, D. Wade. 2004. “On Operationalisms and Economics.” Journal of Economic Issues
38:953–68.

Harman, Gilbert. 1986. Change in View: Principles of Reasoning. Cambridge MA: Brad-
ford–MIT.

Hausman, Daniel M. 2000. “Revealed Preference, Belief and Game Theory.” Economics
and Philosophy 16:99–115.

. 2008. “Mindful or Mindless Economics: A Methodological Evaluation.” In
Caplin and Schotter 2008, 125–53.

. 2012. Preference, Value, Choice and Welfare. Cambridge: Cambridge University
Press.

Hindriks, Frank. 2006. “Tractability Assumptions and the Musgrave–Mäki Typology.”
Journal of Economic Methodology 13:401–23.

Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Deci-
sion Under Risk.” Econometrica 47:263–92.

Kahneman, Daniel, Peter Wakker, and Rakesh Sarin. 1997. “Back to Bentham? Ex-
plorations of Experienced Utility.” Quarterly Journal of Economics 112:375–405.

Kaplan, Mark. 2002. “Decision Theory and Epistemology.” In The Oxford Handbook of
Epistemology, ed. Paul K. Moser, 434–62. Oxford: Oxford University Press.

Maher, Patrick. 1993. Betting on Theories. Cambridge: Cambridge University Press.

Mäki, Uskali. 2000. “Kinds of Assumptions and their Truth: Shaking an Untwisted
F-Twist.” Kyklos 53:317–35.

, ed. 2002. Fact and Fiction in Economics. Cambridge: Cambridge University Press.

Mas-Colell, Andreu, Michael D. Winston, and Jerry R. Green. 1995. Microeconomic
Theory. Oxford: Oxford University Press.

Morgan, Mary. 2012. The World in the Model: How Economists Work and Think. Cambridge:
Cambridge University Press.

Musgrave, Alan. 1981. “Unreal Assumptions in Economic Theory: The F-Twist Un-
twisted.” Kyklos 34:377–87.

Reiss, Julian. 2013. Philosophy of Economics: A Contemporary Introduction. Abingdon: Rout-
ledge–Taylor.

Rosenberg, Alexander. 1988. Philosophy of Social Science. Boulder CO: Westview. Citations
refer to 3rd Edition (2008).

. 1992. Economics: Mathematical Politics or Science of Diminishing Returns? Chicago IL:
University of Chicago Press.

Rubinstein, Ariel, and Yuval Salant. 2008. “Some Thoughts on the Principle of Revealed
Preference.” In Caplin and Schotter 2008, 116–24.

Samuelson, Paul A. 1948. “Consumption Theory in Terms of Revealed Preference.”
Economica 15:243–53.

22



Sen, Amartya K. 1973. “Behaviour and the Concept of Preference.” Economica
40:241–59.

. 1993. “Internal Consistency of Choice.” Econometrica 61:495–521.

van Fraassen, Bas C. 1980. The Scientific Image. Oxford: OUP–Clarendon.

. 1994. “Gideon Rosen on Constructive Empiricism.” Philosophical Studies
74:179–92.

Varian, Hal R. 1992. Microeconomic Analysis. 3rd edition. London: Norton.

. 2005. Intermediate Microeconomics: A Modern Approach. 7th edition. London: Norton.

Wakker, Peter. 2010. Prospect Theory: For Risk and Ambiguity. Cambridge: Cambridge
University Press.

Wong, Stanley. 1978. The Foundations of Paul Samuelson’s Revealed Preference Theory. London:
Routledge. Citations refer to 2nd Revised Edition (2006) from Routledge–Taylor.

23