Loss Aversion and the Bet You Turn Down
In 2018 two consumer researchers argued that loss aversion had been accepted as a general rule without the evidence to back it. The most persuasive reply came from a five-cent charge on shopping bags in Maryland, not from a laboratory.
Answer to the question
Do people really hate losing more than they like winning?
On average, yes, by about two to one. A 2024 meta-analysis led by Alexander Brown pooled 607 estimates and put the mean loss aversion coefficient between 1.8 and 2.1. The size varies widely between people and settings, and Kellen Mrkva and colleagues found in 2020 that it shrinks with knowledge and experience without disappearing.
Key takeaways
- Tversky and Kahneman's famous coefficient of 2.25 was the median from 25 graduate students in their 1992 paper.
- A 2024 meta-analysis of 607 estimates from 150 articles put the mean loss aversion coefficient between 1.8 and 2.1, with individual estimates ranging from close to zero to above twenty.
- In 2020 Kellen Mrkva and colleagues found, across 17,720 people, that loss aversion is smaller with knowledge and experience and larger in older consumers, but present at every level.
- A five-cent charge on disposable bags cut their use by more than forty percentage points, while a five-cent bonus for reusable bags had virtually no effect.
Two consumer researchers set out to take loss aversion apart. Writing in the Journal of Consumer Psychology, David Gal and Derek Rucker reviewed the evidence offered for it and concluded that losses do not, on balance, weigh more than gains. They asked why social scientists kept treating the idea as a general principle despite evidence to the contrary, and linked its persistence to the way beliefs take hold in science. Their target was one of the most quoted claims in behavioural economics, that losing something hurts about twice as much as gaining it pleases. Testing their case means going back to a coin, and to the book that made the bet famous.
What does refusing a fair coin bet reveal about you?
Daniel Kahneman's Thinking, Fast and Slow puts the bet to the reader in its chapter on prospect theory. Tails, you lose a hundred dollars. Heads, you win a hundred and fifty. The odds favour you, and the book predicts you will dislike the offer, as most people do. It follows with a sharper question: what is the smallest gain that would make an even chance of losing a hundred dollars acceptable? For many people, it says, about two hundred dollars, twice the loss.
On page 284 the book gives a range rather than a single figure. The loss aversion ratio, estimated in several experiments, is usually 1.5 to 2.5. It tends to rise with the stakes, though not dramatically, and becomes very large when a loss would be ruinous.
The asymmetry first appeared inside prospect theory, which Kahneman and Amos Tversky published in Econometrica in 1979. Its central move was to say that people judge outcomes as gains and losses from a reference point, usually where they stand now, rather than as changes to their total wealth. The curve that turns money into felt value is steeper on the loss side than on the gain side, and the 1979 paper gives that steepness no number.
Kahneman was modest about why the theory caught on. A few pages after the bet he writes that it was accepted by many scholars because the reference point and loss aversion yielded new predictions that turned out to be true, and adds, "We were lucky." The coin bet is a demonstration. It shows that you, probably, would refuse, and it says nothing about how far the refusal would shift if the stake were a mug, a holiday or a pension.
Where did the famous figure of two come from?
The number arrived in two later papers, both with Tversky as first author. In 1991, in the Quarterly Journal of Economics, he and Kahneman showed the asymmetry in riskless choice, in simple trades of everyday goods, and put the coefficient at about two. In 1992, in the Journal of Risk and Uncertainty, they fitted their extended model to the choices of 25 graduate students from Berkeley and Stanford. The median coefficient was 2.25, so the most precise version of the constant comes from one study of 25 students.
In 2024 came a test on a different scale. A meta-analysis led by Alexander Brown gathered 607 estimates of the coefficient from 150 articles in economics, psychology and neuroscience, published between 1992 and 2017, and reported them in the Journal of Economic Literature. The mean came out between 1.8 and 2.1. In the working version of the paper more than nine in ten estimates were above one, the point at which losses and gains weigh the same, but individual estimates ran from close to zero to above twenty. The authors also coded each study's design and found that few of its features shifted the average by much.
The meta-analysis deliberately left out studies in which people buy and sell objects such as mugs, the experiments behind the endowment effect, so it says nothing about the riskless trades of the 1991 paper. With a spread that wide, the average describes a crowd. Any single person, or any single setting, can sit a long way from it.
Is loss aversion a law of nature or of context?
Gal and Rucker's 2018 case was about generality. Their review did not deny that people sometimes weigh losses more heavily. It argued that the evidence does not establish the asymmetry as a general rule, and that consumer research would gain by asking when losses loom larger, when they weigh the same as gains, and when they weigh less.
Kahneman answered in conversation rather than in a journal. Asked about the debate in a 2021 interview with the science magazine Undark, he gave ground on generality and kept the principle. "It's not a law of human nature that you have to find it in every context," he said, and later, "There are experiments where people don't find loss aversion." He defended the idea as a tool: "That doesn't mean that loss aversion's true. It means that it's useful."
Kellen Mrkva and colleagues replied with data in 2020, in the same journal. Across five samples totalling 17,720 people, they found that loss aversion was smaller in people with more knowledge and experience of the thing being traded, and larger in older consumers. Among car buyers, knowing more about one feature of a car reduced loss aversion for that feature only. Yet people at every level of knowledge were still loss averse. They read the results as support for a view in which preferences are partly built at the moment of choice, and as evidence against claims that loss aversion is a fallacy or disguised status quo bias. Their title said that reports of its death were greatly exaggerated.
The asymmetry is common and it is not constant. Across hundreds of gambles, losses weigh more than gains and the average sits near two, while the size shifts with experience, age and what is at stake, so "twice as much" works as a summary of studies and badly as a prediction for any particular shopper.
The asymmetry is common and it is not constant.
What does it explain about free trials and shop discounts?
In 2012 Montgomery County in Maryland began charging five cents for each disposable shopping bag. The economist Tatiana Homonoff had observers record the bags used by 16,251 customers in 16 stores in the Washington area. The charge cut disposable bag use by more than forty percentage points. A five-cent bonus for bringing your own bag, which half of the largest grocery chains in the area already offered, had virtually no effect. Identical sums were framed as a fee in one case and a reward in the other, and only the fee changed behaviour. Homonoff reads the result as consistent with loss aversion.
Identical sums were framed as a fee in one case and a reward in the other, and only the fee changed behaviour.
Shops borrow that result by framing a discount as a loss avoided, such as a delivery charge waived.
Free trials are the case most often credited to loss aversion, and the evidence there runs another way. In a field experiment published in 2023, Hema Yoganarasimhan and colleagues randomly gave 337,724 new users of a software company a 7, 14 or 30 day trial. The 7 day trial produced the most subscriptions. The authors trace the result to how people learn a product: long stretches of inactivity near the end of a trial went with fewer people paying, which points to habit and inertia more than to the sting of giving something up.
Experienced users gained more from longer trials in that experiment, and beginners from shorter ones.
Thinking, Fast and Slow draws its own boundary on page 294: in routine trade, where the merchant's shoes and the buyer's money are both held for exchange, losses do not loom larger on either side. The bag study sits awkwardly with that boundary, since a five-cent charge at a till is as routine as commerce gets. The book's own examples of goods held for use are a collector's wine, a ticket to a big game, leisure time and the standard of living a family has grown used to, and on that list a carrier bag looks badly out of place.
Common questions
Do people really hate losing more than they like winning?
On average, yes, by about two to one. A 2024 meta-analysis led by Alexander Brown pooled 607 estimates and put the mean loss aversion coefficient between 1.8 and 2.1. Individual estimates varied widely, and Kellen Mrkva and colleagues found in 2020 that the asymmetry shrinks with knowledge and experience.
Where does the 2.25 figure come from?
It is the median loss aversion coefficient in Tversky and Kahneman's 1992 paper in the Journal of Risk and Uncertainty, estimated from the choices of 25 graduate students at Berkeley and Stanford. The 1979 prospect theory paper gives no number, and Thinking, Fast and Slow gives a range of 1.5 to 2.5.
Did Gal and Rucker disprove loss aversion?
They challenged its generality. In 2018 David Gal and Derek Rucker argued that the evidence does not show losses to be generally more impactful than gains. Mrkva and colleagues replied in 2020 with 17,720 people and found loss aversion at every level of knowledge, though smaller among experienced consumers.
Sources
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux (paperback 2013). Read in full by this publication, with a reliability audit. Ch. 26, pp. 283-284: the coin-toss bet (lose $100, win $150), the $200 answer and the 1.5 to 2.5 range, no researcher named; p. 288: 'We were lucky'; ch. 27, p. 294: no loss aversion on either side of routine commercial exchanges.
Kahneman, D., & Tversky, A. (1979). Prospect theory: an analysis of decision under risk. Econometrica, 47(2), 263-291. doi:10.2307/1914185. Abstract read: value assigned to gains and losses rather than final assets; value function concave for gains, convex for losses and steeper for losses.
Tversky, A., & Kahneman, D. (1991). Loss aversion in riskless choice: a reference-dependent model. Quarterly Journal of Economics, 106(4), 1039-1061. doi:10.2307/2937956. Full text read (technical report version, DTIC ADA231109): the data 'yield a loss aversion coefficient of about 2'.
Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297-323. doi:10.1007/BF00122574. Full text read: 25 graduate students from Berkeley and Stanford; 'The median lambda was 2.25, indicating pronounced loss aversion'.
Brown, A. L., Imai, T., Vieider, F. M., & Camerer, C. F. (2024). Meta-analysis of empirical estimates of loss aversion. Journal of Economic Literature, 62(2), 485-516. doi:10.1257/jel.20221698. Abstract read (published version): 607 estimates from 150 articles, 1992 to 2017; mean 1.955, 95 per cent probability between 1.820 and 2.102. Full text read (working paper, LMU discussion paper 275): riskless WTP-WTA studies excluded; range 0.040 to 23.460; 93.9 per cent of estimates above 1.
Gal, D., & Rucker, D. D. (2018). The loss of loss aversion: will it loom larger than its gain? Journal of Consumer Psychology, 28(3), 497-516. doi:10.1002/jcpy.1047. Abstract read: 'current evidence does not support that losses, on balance, tend to be any more impactful than gains'; asks why acceptance persists among social scientists.
Mrkva, K., Johnson, E. J., Gächter, S., & Herrmann, A. (2020). Moderating loss aversion: loss aversion has moderators, but reports of its death are greatly exaggerated. Journal of Consumer Psychology, 30(3), 407-428. doi:10.1002/jcpy.1156. Abstract read: five samples, 17,720 participants; less loss aversion with domain knowledge and experience, more in older consumers, attribute-specific among car buyers, present at every knowledge level.
Schulson, M. (2021). Interview with Daniel Kahneman in 'Among social scientists, a vigorous debate over loss aversion', Undark, republished by Salon, 27 November 2021. Full text read: Kahneman quotations as given; 'Kahneman has not responded publicly to Gal, Rucker, or other recent challengers'.
Homonoff, T. A. (2018). Can small incentives have large effects? The impact of taxes versus bonuses on disposable bag use. American Economic Journal: Economic Policy, 10(4), 177-210. doi:10.1257/pol.20150261. Full text read: Montgomery County, Maryland, tax from 1 January 2012; 16,251 customers in 16 stores; tax cut disposable bag use by over forty percentage points; bonus had virtually no effect; consistent with loss aversion.
Yoganarasimhan, H., Barzegary, E., & Pani, A. (2023). Design and evaluation of optimal free trials. Management Science, 69(6), 3220-3240. doi:10.1287/mnsc.2022.4507. Full text read: 337,724 new users randomly assigned to 7, 14 or 30 day trials of SaaS products; the 7 day trial gave the most subscriptions; consumer learning, not cannibalisation, supported.