Prospect Theory and Loss Aversion
The unit takes the standard theory of choice under risk apart and puts a better one in its place. It starts with Bernoulli's game and expected utility, shows with the Allais paradox, the reflection effect, and framing where the theory fails, and uses Rabin's calibration theorem to prove that the curvature of the utility of wealth cannot explain why people refuse small bets. Prospect theory follows, with the value function, loss aversion, the weighting of probabilities, and the fourfold pattern that explains insurance and lottery tickets at once, and the unit ends with the endowment effect and with reference points in the field, from taxi drivers to golfers and home sellers.
Offered a coin toss that pays 110 euros for heads and costs 100 euros for tails, most people decline, although the toss is worth 5 euros on average to whoever plays it. Many of the same people buy a lottery ticket now and then, insure a telephone against theft at a premium far above the expected loss, and would not sell a mug they were given a minute ago for the price they would have paid for it. The standard model of Units 2 to 12 handles all of these choices with one tool, the expected utility of wealth. This unit shows where that tool works, where it breaks, and what Daniel Kahneman (1934-2024) and Amos Tversky (1937-1996) put in its place in 1979. Unit 13 tested the beliefs the standard model assumes; this unit tests its theory of choice when the outcome is uncertain.
1 Expected utility and the price of risk
A prospect is a list of outcomes with their probabilities, such as “110 euros with probability one half and \(-100\) euros with probability one half”. Unit 1 defined the expected value of such a list as the probability-weighted average of its outcomes, which is 5 euros here, and it is tempting to take the expected value as the worth of the prospect. A famous game, known as the St Petersburg game after the city where its solution was published, shows that nobody does, and it works as follows. A fair coin is tossed until it first shows heads. If heads comes on the first toss, the player receives 1 ducat, which was a gold coin in use in the eighteenth century; if it first comes on the second toss, 2 ducats; on the third, 4 ducats; and so on, the prize doubling with every toss. The chance of the first heads on toss \(k\) is \(1/2^{k}\) and the prize is \(2^{k-1}\) ducats, so every toss contributes \(1/2^{k} \times 2^{k-1} = 1/2\) ducat to the expected value, and the sum over infinitely many tosses is infinite. Nobody would pay more than a few ducats to play, and Daniel Bernoulli (1700-1782), who published the game in 1738, wrote that any reasonable man would sell his chance with pleasure for 20 ducats.
Bernoulli’s answer was that money has a value to a person which is not proportional to its amount. One extra ducat means more to a poor man than to a rich one, so the gain in well-being from a sum of money grows more slowly than the sum itself. Economists call that value the utility of wealth, written \(u(W)\) for a wealth \(W\), and the claim that each further euro adds less is the diminishing marginal utility of Unit 3, now applied to wealth as a whole rather than to one good. A person who follows Bernoulli values a prospect by its expected utility, the probability-weighted average of the utilities of its outcomes, and not by its expected value. John von Neumann (1903-1957) and Oskar Morgenstern (1902-1977) showed in their book Theory of Games and Economic Behavior of 1944 that a person whose choices obey a few axioms of consistency must choose as if maximizing expected utility for some function \(u\). The axioms are simple rules such as “if A is preferred to B and B to C, then A is preferred to C”; they are stated in the 1944 edition, and the proof was added to the second edition of 1947. Since then expected utility theory has been the standard model of choice under risk.
Two terms in the box, the certainty equivalent and the risk premium, are easiest to see in a picture, and Figure 1 draws one for a concave utility of wealth and a prospect that gives 100,000 or 900,000 euros with equal probability. The expected wealth is 500,000 euros. The expected utility is the average of \(u(100{,}000)\) and \(u(900{,}000)\), and the average of two heights on a curve is the height of the midpoint of the straight line that joins them, which is called a chord. In the figure that midpoint lies at the height of \(u(400{,}000)\), so the expected utility equals the utility of a sure 400,000 euros. The certainty equivalent of the prospect is therefore 400,000 euros, and the risk premium, the gap between expected wealth and certainty equivalent, is 100,000 euros. The chord lies below the curve only because the curve bends downward, so in expected utility theory risk aversion and the concavity of the utility of wealth are one and the same thing.