An evaluation of the consistency of individuals' preferences in
decisions involving uncertainty and time and consideration of
the implications for economic modelling.
For more details contact:
Professor Robert Sugden, School of Economics and Social Studies,
University of East Anglia, Norwich, NR4 7TJ; Tel. +1603 593423;
Fax. +1603 250434.
Conventional economic analysis relies on the assumption that individuals
act as if they have highly articulated preferences which obey
certain fundamental principles of consistency. This assumption
underpins many of the predictions derived from economic models,
and provides the basis for welfare economics and for associated
techniques of public policy evaluation, such as cost-benefit analysis.
However, a substantial body of evidence, generated largely by
experimental and survey research methods, has raised serious doubts
about the internal consistency of people's preferences. In particular,
decisions involving uncertainty and time have been found to contravene
the conventional axioms of consistency in systematic and predictable
ways. These unpredicted patterns of behaviour are often referred
to as anomalies.
Economists and psychologists have tried to explain this lack of
fit between theory and evidence in at least two broad ways. One
is to discount the evidence on the grounds that in experimental
and survey research, data are collected from inexperienced respondents
who have little or no incentive to reflect about the tasks they
are set. Thus, it is claimed, these data are subject to errors
and biases that are not present when people make "real"
choices in familiar market settings. The other strategy of explanation
is to challenge the idea that people come to decision problems
with complete and consistent preferences, and to suggest instead
that preferences are constructed only in response to specific
problems. Thus, consistency of preferences across different problems
is not to be expected.
The main objective of the project was to evaluate the significance
of this body of evidence and to consider its implications for
economic modelling. In particular, we sought to discriminate between
the two main alternative strategies for explaining the mismatch
between theory and evidence. An important aspect of the project
was the integration of the theoretical insights and research methods
of economics and psychology. All of our work has been informed
by the two disciplines, but the tragic death of Jane Beattie has
prevented us from achieving one of our main aims: the establishment
of a cross-disciplinary research team which would continue to
work at the interface of economics and psychology.
Our work focused on the following issues:
Diagnosing anomalies. We investigated a number of previously-known
and predictable patterns of choice behaviour which contravene
the standard theory of consistent preferences. These included:
the divergence between people's willingness to pay for goods that
they do not possess and their willingness to accept compensation
for giving up goods that they do possess; the differences
between the preferences people reveal in straight choices and
those that they reveal when assigning money values to options;
the tendency for people to give greater implicit weight to an
event with a given probability if it is redescribed as two events
with the same total probability; and systematic cycles of pairwise
choice (e.g. A chosen over B, B chosen over C, and C chosen over
A). We used a number of different experimental designs to try
to uncover the underlying causes of these anomalies. Our findings
are broadly consistent with the hypothesis that individuals do
not have precise and well-defined preferences prior to confronting
specific choice problems. Rather, they use a mix of rules of thumb;
anomalies arise because different rules of thumb are used in different
contexts. A particularly interesting finding was that people can
often give what they regard as coherent reasons for sets of decisions
which, from an external point of view, seem totally inconsistent.
Incentives in experiments. Some commentators have argued
that much of the existing evidence about anomalies is unreliable,
because it comes from experiments in which the incentives for
"correct" reasoning are very small. We ran a series
of experiments to test for differences between conventional experimental
designs (in which incentives are small or non-existent) and a
"single choice" design in which each individual faces
just one decision problem, with significant amounts of money at
stake. Generally we found no significant differences between the
two types of design and therefore no support for the hypothesis
that evidence from earlier experiments should be discounted.
Dynamic choice. Principles of "dynamic choice"
specify how a person acts when he faces a multi-stage decision
problem - for example, one which requires a sequence of decisions,
or one in which decisions are preceded by chance events. The standard
theory of choice under uncertainty, expected utility theory (EUT),
can be derived from a set of apparently reasonable principles
of dynamic choice. Prior to our work, little was known about which
of these principles is contravened when (as in several well-known
classes of anomaly) people behave contrary to EUT. We ran an experiment
to investigate this issue. Surprisingly, the one principle that
was clearly violated was a principle of "timing independence"
which, to our knowledge, has not previously been questioned by
decision theorists.
Stochastic choice. By running an experiment in which each
participant made a very large number of decisions involving risk,
we were able to investigate the stochastic element in choice behaviour.
We tested two models in which individuals have "true"
underlying preferences but make random errors, and a third "random
preference" model in which preferences are imprecise. Our
data allowed us to reject the hypothesis that derivations from
EUT are merely the result of random error. The model which best
fitted the data was based on a recently-developed alternative
to EUT (rank-dependent expected utility theory), but with imprecise
preferences.
Money pumps and imitation. Economic theorists sometimes
argue that in the long run, market environments tend to select
preferences which satisfy the consistency principles of conventional
theory. As a corollary, it is argued that the usual kinds of experimental
tests of those principles are inappropriate, as the laboratory
environment does not include the relevant selection mechanisms.
Two versions of this strategy of argument are particularly well
known. One is the money pump argument, which seeks to show
that a person whose preferences violate the standard consistency
conditions can be exploited by a money-seeking arbitrageur. The
other is to appeal to some evolutionary process analogous with
natural selection in biology. We subjected each of these arguments
to rigorous theoretical analysis. In each case, our strategy was
to build a model which does not presuppose any consistency of
preferences, but which includes the relevant selection mechanisms.
In each case, we were able to show that these mechanisms did not
select preferences with standard consistency properties.