Economics•Chapter 3•6 min read•Updated September 24, 2026

Behavioral Economics — Prospect Theory II: Probability Weighting and the Endowment Effect

O
OIYO EditorialContributor
3/10

Behavioral Economics — Probabilities Are Not Used as They Are, and What We Own Becomes Dearer

The value function of Chapter 2 dealt with how outcomes are evaluated. The second element of prospect theory is how probabilities are used. In expected utility theory, outcomes are weighted in proportion to their probabilities. In prospect theory, probabilities are converted into decision weights before they are used, and the conversion is not linear.

1. The Allais paradox and the certainty effect

The structure of the Allais paradox
ProblemOption AOption BCommon choice
1100 million won for sure10%: 500 million; 89%: 100 million; 1%: 0A
211%: 100 million won; 89%: 010%: 500 million won; 90%: 0B

Apart from the common outcome “89% chance of 100 million won,” the two problems are the same choice. Expected utility theory ignores common outcomes, so someone who picks A in Problem 1 should also pick the corresponding 11% option in Problem 2 (the independence axiom). In practice many people pick A in 1 and B in 2, because a fall in probability from 100% to 99% feels much larger than a fall from 11% to 10%. This is the certainty effect.

2. The probability weighting function

Probability weighting function (Tversky and Kahneman 1992)
w(p)=pγ(pγ+(1−p)γ)1/γw(p) = \frac{p^{\gamma}}{\left(p^{\gamma}+(1-p)^{\gamma}\right)^{1/\gamma}}
The estimate for gains is γ ≈ 0.61. With γ = 1, w(p) = p, as in expected utility.

With γ=0.61\gamma=0.61, a 1% probability is weighted as about 5.5% and 99% as about 91%; 50% becomes about 42%. Small probabilities are overweighted, and medium to large probabilities underweighted. This inverse-S shape explains both the certainty effect and lottery purchases.

The probability weighting function (γ = 0.61)
p w(p) 50 42 w(p) w = p 1% → 5.5%

Above the diagonal (small probabilities) is overweighting; below it (medium to large probabilities) is underweighting. The two curves cross at around 30–40%.

3. The fourfold pattern of risk attitudes

Combining the shape of the value function (risk aversion in gains, risk seeking in losses) with probability weighting yields four patterns.

The fourfold pattern
GainsLosses
High probabilityRisk averse — accepting a settlement instead of going to trialRisk seeking — holding on instead of locking in a loss
Low probabilityRisk seeking — buying lottery ticketsRisk averse — buying insurance

Risk seeking appears for low-probability gains because the overweighting of small probabilities overwhelms the concavity of the value function. A lottery paying 1 million won with 1% probability has an expected value of 10,000 won, but with a decision weight of 5.5% its subjective value can exceed that of a sure 10,000 won. By the same logic, people pay more than the expected loss of 10,000 won for insurance against a 1% chance of losing 1 million won.

4. The endowment effect: what we own is dearer

Kahneman, Knetsch and Thaler (1990) gave half the participants in an experiment a mug and let them trade. Since the mugs were assigned at random, about half should have traded if preferences were the same. Actual trading volume was far lower, and the median price demanded by sellers was more than about twice what buyers were willing to pay.

In terms of the loss aversion of Chapter 2, selling the mug is a loss for its owner and buying it is a gain for a non-owner. Because losses loom larger, the willingness to accept (WTA) exceeds the willingness to pay (WTP).

Loss aversion interpretation of the endowment effect
WTA≈λ⋅WTP\text{WTA} \approx \lambda \cdot \text{WTP}
If the value function is linear near the origin, the WTA/WTP ratio equals the loss aversion coefficient. The experimental ratio of around 2 broadly matches estimates of λ.

The endowment effect weakens the prediction of the Coase theorem (Public Finance Chapter 1). Depending on who is given the right, either WTA or WTP becomes the benchmark, so the initial allocation can affect the final allocation even without transaction costs.

5. Status quo bias

When there are several alternatives, the option of keeping the current state is chosen too often. Samuelson and Zeckhauser (1988) showed that presenting the same options but labelling one as the “status quo” raises the share choosing it. Every change away from the status quo involves both gains and losses, and if losses feel larger, not changing looks better. Switching costs and decision avoidance work in the same direction. The default design of Chapter 9 exploits this bias.

Check your understanding

With w(0.01) = 0.055 and a value function for gains of v(x)=x0.88v(x)=x^{0.88}, which would someone choose: “1 million won with 1% probability” or “20,000 won for sure”? In units of 10,000 won, the lottery’s value is 0.055×1000.88≈0.055×57.5≈3.20.055\times 100^{0.88}\approx 0.055\times 57.5\approx 3.2, and the value of a sure 20,000 won is 20.88≈1.842^{0.88}\approx 1.84. The person prefers a lottery with an expected value of 10,000 won to a sure 20,000 won. Using the probability as it is (w=0.01w=0.01), the lottery’s value would be 0.58, reversing the choice.

References

  • Maurice Allais, “Le comportement de l’homme rationnel devant le risque,” Econometrica (1953)
  • Daniel Kahneman, Jack Knetsch and Richard Thaler, “Experimental Tests of the Endowment Effect and the Coase Theorem,” Journal of Political Economy (1990)
  • William Samuelson and Richard Zeckhauser, “Status Quo Bias in Decision Making,” Journal of Risk and Uncertainty (1988)
  • Charles Plott and Kathryn Zeiler, “The Willingness to Pay–Willingness to Accept Gap,” American Economic Review (2005)
O

OIYO Editorial

Editorial Desk

The OIYO editorial desk researches money, law, lifestyle, and self-understanding topics against primary sources and public statistics. Every piece carries source notes and is reviewed on a regular cycle for accuracy and usefulness.