Chapter Summary

Numerous examples of behaviour contradict the predictions of the standard rational choice model. People apparently often fail to ignore sunk costs. They play tennis indoors when, by their own account, they would prefer to play outside. They behave differently when they lose a ticket than when they lose an equivalent amount of cash. Psychologists argue that such behaviour is the result of limitations in human cognitive capacity. People use mental accounting systems that reduce the complexity of their decisions, sometimes at the expense of consistency with the axioms of rational choice.

An important class of departures from rational choice appears to result from the asymmetric value function described by Kahneman and Tversky. In contrast to the rational choice model, which uses a utility function defined on total wealth, Kahneman's and Tversky's descriptive theory uses a value function defined over changes in wealth. Unlike the traditional model, it gives losses much heavier decision weight than gains. This feature makes decisions extremely sensitive to how the alternatives are framed. A pair of gains, for example, is more attractive if presented separately than if lumped together. Losses, in contrast, have less impact if amalgamated than if taken separately. Also, a loss combined with a slightly larger gain produces a positive effect, whereas taken separately their net effect is negative; and finally, a small gain segregated from a large loss produces less of a negative effect than the two lumped together. The rational choice model, in contrast, says that none of these framing effects should matter.

Decisions under uncertainty also often violate the prescriptions of the expected utility model. And again, the asymmetric value function provides a consistent description of several important patterns. People tend to be risk averse in the domain of gains but risk seeking in the domain of losses. The result is that subtle differences in the framing of the problem can shift the mental reference point used for reckoning gains and losses, which, in turn, can produce radically different patterns of choice.

Another important departure from rational choice occurs in the heuristics, or rules of thumb, people use to make estimates of important decision factors. The availability heuristic says that one way people estimate the frequency of a given class of events is by the ease with which they can recall relevant examples. This leads to predictable biases because actual frequency is not the only factor that governs how easy it is to recall examples. People tend to overestimate the frequency of vivid or salient events, and of other events that are especially easy to retrieve from memory.

Another important heuristic is representativeness. People estimate the likelihood that an item belongs to a given class by how representative it is of that class. We saw that this often leads to substantial bias because representativeness is only one of many factors that govern this likelihood. Shyness may indeed be a trait representative of librarians, but if there are many more salespeople than librarians, it is more likely that a randomly chosen shy person is a salesperson rather than a librarian.

Anchoring and adjustment is a third heuristic that often leads to biased estimates of important decision factors. This heuristic says that people often make numerical estimates by first picking a convenient, but sometimes irrelevant, anchor and then adjusting from it (usually insufficiently) on the basis of other potentially relevant information. This procedure often causes people to underestimate the failure rate of projects with many steps. If such a project fails when any one of its essential elements fails, then even if the failure rate of each element is extremely low, a project with many elements is nonetheless very likely to fail. Because people tend to anchor on the failure rate for the typical step, and adjust insufficiently from it, they often grossly overestimate the likelihood of success. This may help explain the naive optimism of people who start new businesses.

Another departure from rational choice is analogous to a phenomenon in the psychophysics of perception. Psychologists have discovered that the just perceptible change in any stimulus tends to be proportional to its initial level. A related pattern seems to apply when the stimulus in question is the price of a good or service. People think nothing of driving across town to save $5 on a $25 radio, but would never dream of doing so to save $5 on a $500 TV set.

Finally, departures from rational choice may occur because people simply have difficulty choosing between alternatives that are hard to compare. The rational choice model assumes that we have complete preference orderings, but in practice, it often seems to require a great deal of effort for us to decide how we feel about even very simple alternatives.

Behavioural models of choice often do a much better job of predicting actual decisions than the rational choice model. It is important to remember, however, that the behavioural models claim no normative significance. That is, the mere fact that people are often influenced in their decisions by sunk costs should not be taken to mean that people should be influenced by them. The rational choice model says we can make better decisions by ignoring sunk costs, and most people, on reflection, strongly agree. In this respect, behavioural models of choice are an important tool for helping us avoid common pitfalls in decision making.