Journal of Economic Psychology 19 (1998) 397±401

Experimental economics and deception: A comment

*
John D. Hey
Department of Economics, University of York, Heslington, York Y01 5DD, UK

Let me begin at the end: with what I have called elsewhere ``Ill-De®ned Ex-
periments'' and which are referred to in Bonetti's ``Experimental Economics
and Deception'' as my suggestion of part of ``a vision splendid of the future
of experimental economics''. I argue there (Hey, 1991) that ill-de®ned exper-
iments are an important new tool in the economist's armoury ± a tool that
will enable the economist to explore brave new worlds, worlds into which
the neoclassicist economist has yet to venture. Ill-de®ned experiments are ex-
periments in which the subject is not told all the features of the experiment at
the beginning of the experiment: experiments in which, for example, there are
possible decisions that the subject may take which are not revealed to the
subject right from the beginning of the experiment; experiments in which,
for example, there is information available to the subject which is not re-
vealed right from the beginning of the experiment. I am not suggesting, or
indeed advocating, that the subjects are deliberately deceived ± on the con-
trary, I will still maintain that in this kind of experiment, as well as in
well-de®ned experiments, subjects should not be deceived. I will argue this
shortly, but let me ®rst emphasise the point that Bonetti has missed in his al-
leged coup de grace: there is a world of di€erence between not telling subjects
things and telling them the wrong things. The latter is deception, the former is
not. I am advocating the former, not the latter. In ill-de®ned experiments
subjects have to discover what extra information is available, what extra

*
Present address: Tel.: 0039 80 5049195; fax: 0039 80 5049147; e-mail: [email protected].

0167-4870/98/$19.00 Ó 1998 Elsevier Science B.V. All rights reserved.

PII: S 0 1 6 7 - 4 8 7 0 ( 9 8 ) 0 0 0 1 3 - 0
398 J.D. Hey / Journal of Economic Psychology 19 (1998) 397±401

decisions possible; but if they ask we do not tell lies ± we do not deceive them.
True, we may have to tell them what questions they are allowed to ask and
how we will answer them, but, once again, this must be done without decep-
tion. So my advocating ill-de®ned experiments is not, as Bonetti appears to
conclude, an advocation by me of deception.
In one crucial thing, Bonetti is right: economics experiments are not psy-
chological experiments. Or perhaps I should qualify that: psychological ex-
periments are not a homogeneous entity and economics experiments are
not like some psychological experiments (though there are some psychologi-
cal experiments which should be conducted like economics experiments 1). In
general, we economists are doing di€erent things from psychologists: eco-
nomics is a di€erent subject from psychology. Although both disciplines
are interested in human behaviour, the two disciplines are interested in it
from di€erent perspectives.
The majority of economics experiments conducted so far have been either
theory-testing experiments or theory-suggesting experiments, with a common
unifying feature that the majority of experiments have been conducted
against a backdrop of neoclassical economics. What is the crucial feature
of this form of economics is that everything in it is well-speci®ed, and partic-
ularly the `rules of the game' under which the economic agents in the theory
are operating. Let me consider theory-testing experiments ± these must be
testing the behaviour of the agents in the theory. 2 It is therefore crucial that
the subjects are told exactly the `rules of the game' under which they are op-
erating, for otherwise the experiment is no longer a test of the theory as spec-
i®ed: it is more a test of some other theory in which the `rules of the game' are
those constructed by the subject in some manner which cannot possibly be un-
ravelled by the experimenter. And this is the key problem with deception: once
subjects entertain a suspicion that what the experimenter tells them about the
`rules of the game' (or more seriously, as I will argue later, about the behav-
iour of the other subjects in the experiment) then the experimenter no longer
knows the nature of the theory now being tested by his experiment. The ex-
perimenter has lost control. Experimental economics loses one of the key rea-
sons for its existence.

1
Just because (some) psychologists run experiments badly is not a good reason for economists to run
experiments badly.
2
I ignore the argument that, if everything in the theory is reproduced exactly in the laboratory then the
theory must be tautologically true, as correct but uninteresting.
 J.D. Hey / Journal of Economic Psychology 19 (1998) 397±401 399

Bonetti also highlights another crucial di€erence between economics exper-

iments and (some) psychological experiments: economists regard as crucial a
payment mechanism linked to the behaviour of the subject; psychological ex-
periments, in contrast, typically do not. This is not the result of chance but
rather the consequence of the fact that economics experiments are testing
(or investigating) theories of economic behaviour in which a particular vari-
able (or set of variables) is posited to be the objective of the economic agent
of that theory: it is therefore crucial that that variable (or that set of vari-
ables) is also the objective of the subject in the experiment; for otherwise,
once again, the experiment is no longer an appropriate test of the theory
in question. (As a side issue, I think that Bonetti is wrong in stating that
the existence of the magnitude of such payments ± linked to the behaviour
of the subjects ± or the magnitude of such payments does not a€ect the be-
haviour of the subjects. See Smith and Walker (1993) in particular for evi-
dence to the contrary.)
Given such a payment mechanism, one form of deception is clearly ruled
out ± deception about the relationship between the behaviour of the subject
and the payment to the subject. For such a deception could be detected ±
which could expose the experimenter to litigation. But there is a more subtle
form of deception which should also be ruled out: deception about the behav-
iour of the other subjects in the experiment. For if such deception were to be
practised then there is a chance that it might be detected (by subjects talking
to each other after the experiment) ± and once again the experimenter ex-
posed to potential litigation. The experimenter could try and avoid this pos-
sibility by endeavouring to ensure that subjects did not make contact after
the experiment, or by performing the experiment with the subjects visually
separated from each other. The trouble with this is that such procedures
would kindle the embers of any suspicions of deception present amongst
the subjects. Of course, there are good reasons why subjects might suspect de-
ception ± they can easily reason that deception can work to the experiment-
er's interest: it can save the experimenter money. 3 Bonetti cites approvingly
experiments in which such motives can justify deception: in particular, he
quotes Public Goods experiments in which the actual numbers of subjects dif-
fer from the numbers told to the subjects. Interestingly, he does not cite the
classic reference in which such deception is carried out ± that by Kim
and Walker (1984). They told subjects that there were 100 subjects in the

3
Actually one would hope that subjects could reason this way.
400 J.D. Hey / Journal of Economic Psychology 19 (1998) 397±401

experiment when in fact there were just ®ve. Elaborate procedures were need-
ed to keep up this deception ± even so far as adding a random term to the
aggregate public good contributions of the `100' subjects. Even so, as the au-
thors themselves admitted, the experiment collapsed remarkably quickly ±
with complete free-riding setting in after very few repetitions. 4 I must admit
that this does not surprise me ± if I had been a subject in that experiment,
with all the devious procedures in place, I would have become very suspi-
cious. I am sure my response would have been to give up any attempt to con-
tribute to the public good.
Indeed this experiment, and others I have seen being run, have strength-
ened my belief that there should be some way, after the experiment, of sub-
jects verifying that the information they have been given during the
experiment does indeed agree with what the other subjects have done and
have been told. A printout of the results, signed by the participants, could
be publicly posted, for example. I suggest this simply for the reason I have
given above: once subjects entertain a suspicion that what they have been
told by the experimenter is possibly not true, then it is no longer possible
for the experimenter to know what alternative story the subjects have in
mind.
A second scenario in which Bonetti approvingly cites the use of deception
is when the experimenter wishes to manipulate the behaviour of one or more
of the subjects in an interactive game. I appreciate the reasons why he does
suggest this: if one is interested in seeing how subjects respond to a particular
pattern of behaviour adopted by other subjects, then if those other subjects
do not display that pattern of behaviour, one has a singularly uninformative
experiment. He suggests therefore that one should simulate those `other sub-
jects' through a computer subject (rather than a human subject) and that one
should not tell the (human) subjects that you are doing this. Well, I agree with
the ®rst of these but not the second. I do not agree that deception is neces-
sary. 5
Given my earlier arguments against deception, why does Bonetti advocate
it in this instance? I suspect because, if humans knew they were playing
against a computer, they would play di€erently. Although Bonetti does not
give any evidence as to whether this might be true or not, I suspect the reason

4
So much so that they had to restart the experiment.
5
Though I should admit that I have been guilty of this practice myself (Hey and Reynolds, 1992) for
reasons which I cannot recall and which I still feel guilty about. I do not think I gained any bene®t from
using that deception.
 J.D. Hey / Journal of Economic Psychology 19 (1998) 397±401 401

that he makes this claim is that he believes that humans feel di€erently about
money being `paid' to a computer subject rather than to a human subject.
This may well be true, but one can simply respond by paying the money that
the computer subject earns to a human subject who is recruited for the exper-
iment but who is not selected (perhaps by a random device) to participate.
(Or indeed to a charity, though I suspect this will alter things again.) I do
not see how the deception, per se, helps the experiment ± though it could,
as I have argued above, harm it.
Bonetti cites one further practice which he terms `deception' ± using di€er-
ent terms to describe features of the experiment. This not deception: as long
as the subjects know the implications of the actions of themselves and of the
other subjects in the experiment (in terms of the outcomes for the various
subjects) then it does not matter what one calls the various components of
the experiment. So calling something `an investment in account B' rather
than `a contribution to the public good' is not deception ± if the implications
of each are the same and known as the same to all the subjects. While I ap-
preciate that it is interesting to test the e€ect of di€erent forms of wording on
the behaviour of subjects doing so is not deception as I understand it.
The central point then is that if subjects entertain any suspicion that the
experimenter might be practising undetectable 6 deception then the experi-
menter has lost control. Since control is the main justi®cation for the exper-
imental approach, I therefore conclude that deception is unacceptable.
Furthermore, Bonetti's examples do not justify deception but rather the
adoption of novel experimental practices. I would hope that he will concen-
trate his energies on those rather than on the practice of deception!

References

Bonetti, S., 1998. Experimental economics and deception. Journal of Economic Psychology (this issue).
Hey, J.D., 1991. Experiments in Economics, Blackwell, Oxford.
Hey, J.D., Reynolds, M.R., 1992. Experiments, games and economics. In: Moss, S., Rae, J. (Eds.),
Arti®cial Intelligence and Economic Analysis: Prospects and Problems, Edward Elgar, Chelterhan, pp.
81±115.
Kim, O., Walker, M., 1984. The free-rider problem: experimental evidence. Public Choice 43, 3±24.
Smith, V.E., Walker, J.M., 1993. Monetary rewards and decision cost in experimental economics.
Economic Inquiry 31, 245±261.

6
I rule out as simply stupid the exercise of detectable deception ± for reasons already given.