Economic View: When proven economic theory falls flat on its face

Does the fact that taking Prozac can make somebody more inclined to impulse buy when they are out shopping invalidate economic theory?

It is a serious question. The foundation stone of economics is the assumption that individuals maximise their expected utility - that is, the benefit they expect to gain from their choices on work, saving and spending. Aggregating the choices of all the individuals in the economy delivers the familiar results of economics, such as lower prices, causing demand to increase.

However, the results depend on people's preferences having certain key characteristics, which include consistency. For instance, if I like tuna sandwiches better than chicken, and chicken better than cheese, then I must prefer tuna to cheese. In addition, people are assumed to think more is better, but with diminishing marginal returns: the tenth tuna sandwich is less appealing than the first. Given a consistent set of preferences, economic choices are predictable.

Experiments dating back to a study by the French Nobel laureate Maurice Allais in 1953 suggest that, alas, actual people do not behave like rational economic beings. The basic theory of consumer choice is systematically violated by how people behave in experimental situations. In fact, experiments over the years have shown that economics students are the only group of people ever to behave as they are supposed to in economic theory.

Some recent experiments directed by Professor Robert Sugden at the University of East Anglia shed new light on the underlying patterns of consumer behaviour. Volunteers were given either cash or cash and a voucher for a main course, or a dessert, or both, at the local pizza restaurant. They were then asked questions about how much they would be willing to pay for a main course or a dessert voucher, and how much they would be willing to sell their vouchers for.

According to economic theory, the prices subjects were willing to pay for the vouchers separately ought to add up to the price they would pay for both. Equally, the price which they would accept for both should equal the sum of the amounts accepted for the two types of voucher separately.

However, testing this by comparing the implicit value placed on a dessert voucher - calculated as willingness to pay for both vouchers less willingness to pay for a main course voucher - showed it to be consistently lower than the actual willingness to pay for the main course voucher in the experiment.

The punch-line is obvious. The sum of the parts added up to more than the whole. This is parallel to survey results which show that the sum of the amounts people claim they would like to see spent on separately cleaning up several beaches exceeds the amount they would be willing to see spent on cleaning all the beaches at once. They might think it was worth spending several millions of pounds on cleaning the oil from the Pembrokeshire coast after the grounding of the tanker Sea Empress earlier this year but would not name a figure 10 times higher for cleaning up 10 beaches simultaneously.

Similarly, the amount people say would compensate them for the loss of clean air or clean beaches is far greater than the amount they would be willing to contribute to cleaning the atmosphere through higher taxes. This result also had its parallel in the pizza experiment, where compensation required for giving up vouchers exceeded willingness to pay for them.

This phenomenon means there is a problem in various areas of public policy, not just the environment. Take the question of compensation for injuries, for example. Two researchers at Newcastle University, Michael Jones-Lee and Graham Loomes, recently reported that surveys showed the amounts of money people said they were willing to pay to reduce the risk of road accidents was extremely sensitive to the size of the initial sum suggested to them. Yet such "contingent value" surveys are widely used for the cost- benefit analysis of spending on safety or environmental improvements.

Further results indicate that these inconsistencies in people's responses are not mistakes. Subjects remain happy with their original answers even when the inconsistencies are pointed out to them. What's more, in experiments where there definitely is a right answer - when the same prize is available with a greater probability in one case than the other - the subjects very rarely get the answer wrong. They are remarkably astute at working out the odds and choosing the option that will deliver the highest expected reward.

According to Professor Sugden, experimental evidence clearly suggests that consumer preferences are conditional on a reference point - the initial endowment of cash and vouchers in the pizza experiment. Given that starting point, subjects' preferences were consistent. Basic consumer choice theory needs to be amended to take account of reference points, he suggests.

This is an argument that appeals to common sense. However, the notion of the rational economic person has its defenders. One is Alvin Roth, Professor of economics at the University of Pittsburgh, whose work in experimental economics is reported on his Internet page.

Professor Roth argues that the economic theory of consumer choice is a good approximation to the world most of the time. He defends this by pointing out that critics oppose the notion of the more-or-less rational economic man with psychological man. The experimental model sees people's choices as the result of mental processes which differ in different contexts, rather than a set of consistent preferences.

However, the idea of psychological man is only an approximation to the world, too. For neuro-biologists see behaviour as the result of bio-chemical changes in the brain. Different levels of chemicals - whether caused by a deficiency of chocolate, a sunny start to the day, or even taking Prozac - will lead to different decisions (more shopping in each case). The psychological metaphor will sometimes break down too.

He goes on to argue that experiments can be useful in pointing out where the economic model breaks down. He calls for "experiments which show not merely that individuals are systematically non-rational but which would start to give us some feeling for how important this phenomenon might be". He suggests that "psychological" models, rather than "economic" models will find their niche in the study of areas such as marketing and advertising.

This does not convince Professor Sugden. Many people will continue to buy Coca-Cola, even when they have preferred Pepsi in blind tastings, he says. Perhaps an economist will think this is not really economics, but the need to build brand loyalty involves serious strategy questions and the spending of millions and millions of pounds.

"What economists think of as the realm to which their theory applies might turn out to be very small if you define the area of the subject to fit the theory," he said.

Alvin Roth's experimental economics Web page can be found at http://info.pitt.edu/alroth/rational.html.