Loss aversion: why losses hurt more than gains, with real examples

In a 1963 paper, the economist Paul Samuelson described a bet he had offered a colleague: a 50% chance to win 200 dollars and a 50% chance to lose 100 dollars. The colleague said no. His reason, as Samuelson told it, was that he would feel the 100 dollar loss more than the 200 dollar gain. He then added that he would happily accept 100 such bets.

The bet was worth 50 dollars a toss on average. Turning it down makes sense only if losing 100 dollars weighs at least twice as much as winning 100 dollars, and his stated reason, a loss that would feel bigger than the gain, is what loss aversion describes.

Loss aversion is the tendency to weigh a loss more heavily than a gain of the same size, with both measured from a reference point such as the price paid, the current balance or the amount a person expected to have.

A portfolio that falls from 120,000 to 110,000 dollars after being bought for 100,000 is still up 10,000 dollars, and it feels like a small setback. The same 10,000 dollar fall from 105,000 to 95,000 crosses the purchase price and feels like a loss.

Prospect theory, the 1979 paper that introduced loss aversion

Daniel Kahneman and Amos Tversky set out the idea in "Prospect theory: an analysis of decision under risk", published in Econometrica in 1979. The paper was a direct challenge to expected utility theory, which assumes people judge a gamble by the final wealth it leaves them with. Kahneman and Tversky showed that people judge outcomes as changes from where they stand, and that they treat gains and losses differently.

Their evidence came from hypothetical choices put to students and faculty in Israel, with amounts in Israeli pounds at a time when the median family earned about 3,000 pounds a month after tax. The same problems given to groups in Stockholm and at the University of Michigan produced essentially the same pattern. Two of the problems show the asymmetry in a single pair of questions:

The gamble in the first problem has an expected value of 3,200, more than the sure 3,000, and most people still preferred certainty. Flip the signs and 92% of respondents gambled to avoid a certain loss. Kahneman and Tversky called this the reflection effect. For an investor it describes a familiar sequence: take the profit on a winner early, and hold a loser in the hope it comes back.

From choices like these, the paper drew a value function with three properties. It is defined on gains and losses from a reference point. It is concave for gains and usually convex for losses, so each extra dollar matters a little less the further one moves from the starting point. And it is steeper for losses than for gains. In the paper's own words, losses loom larger than gains, and "most people find symmetric bets" of the 50-50 kind "distinctly unattractive", more so as the stakes rise.

The reflection effect comes from the curvature of the function. Loss aversion comes from the kink at the reference point, where the loss side falls faster than the gain side rises.

Tversky and Kahneman extended the idea to choices without risk in a 1991 paper in the Quarterly Journal of Economics, which used it to explain why people are reluctant to trade away things they already own. In 1992 they published cumulative prospect theory, a revised version that puts a number on the asymmetry. Kahneman received the 2002 Nobel memorial prize in economic sciences for this line of work. For the wider field these papers opened, see our introduction to behavioral finance.

The loss aversion coefficient and the figure of 2

In cumulative prospect theory, loss aversion is a single parameter, written as the Greek letter lambda. The value of a gain of x is x raised to a power alpha. The value of a loss of x is minus lambda times x raised to a power beta. Tversky and Kahneman estimated alpha and beta at 0.88 and lambda at 2.25. A lambda of 1 would mean losses and gains count equally. A lambda of 2.25 means a loss weighs 2.25 times as much as a gain of the same size.

Here is what that implies for a coin toss with a 100 dollar loss on one side, using the value function with the 1992 curvature of 0.88 on both sides and leaving aside how people weight the 50% probabilities:

• With lambda at 1, a gain of 100 dollars on the other side makes the bet feel fair.
• With lambda at 2, the gain has to reach about 220 dollars.
• With lambda at 2.25, the 1992 estimate, it has to reach about 251 dollars.
• With lambda at 1.31, the lower meta-analytic estimate discussed below, about 136 dollars is enough.

Samuelson's colleague, offered 200 dollars against 100, sat below the 251 dollar break-even of the 1992 parameters, so his refusal fits them. The popular claim that losses hurt "twice as much" as gains is a rounded version of that 2.25 estimate.

How big is loss aversion? What the meta-analyses found

Three large reviews published since 2024 have tried to settle the size of lambda. They reach different answers because they pool different kinds of study.

The first review takes each paper's published number at face value, and the average lands close to the folk figure of 2. The second goes back to the raw choices where the data could be obtained and fits the same model to all of them. Its average is much lower, and in 12 of the 19 data sets the confidence interval for lambda included 1, the point of no loss aversion at all. Only 6 of the 19 intervals included 2.25.

The third review explains part of the gap. When an experiment offers gambles whose losses are smaller than their gains, or lists the options in order of size, measured loss aversion is strong, with an average lambda of 2.295 for ordered designs in which losses were smaller than gains. When gains and losses cover the same range and appear in random order, lambda falls to about 1.07, statistically indistinguishable from 1.

Gal and Rucker's challenge and the 2020 replication studies

The main criticism comes from the psychologists David Gal and Derek Rucker. In 2006 Gal had argued that many effects credited to loss aversion are better explained by inertia, a plain preference for leaving things as they are. In 2018 Gal and Rucker went further in the Journal of Consumer Psychology, concluding that current evidence does not show losses to be more impactful than gains on balance. Their paper opens with a show of hands among about 80 people at a 2016 conference session on consumer decision making: all but three said losses loom larger than gains, although the session had been announced as a challenge to the idea.

Kellen Mrkva, Eric Johnson, Simon Gächter and Andreas Herrmann tested loss aversion across five samples totalling 17,720 people in 2020. People at every level of knowledge and experience were loss averse. Those who knew more about a product attribute were less loss averse about that attribute, and older consumers were more loss averse across measures. The same year, a team led by Kai Ruggeri repeated the original 1979 questions with 4,098 participants in 19 countries and 13 languages. The results replicated for 94% of the items, with somewhat weaker effects than in 1979.

As of 2026, the asymmetry shows up in the original 1979 choice problems across 19 countries and in brokerage and housing data, and it shrinks to almost nothing in laboratory designs where gains and losses are balanced and shown in random order. Depending on the method, the published estimates of lambda run from about 1.3 to about 2.

The disposition effect: selling winners too soon and holding losers too long

Hersh Shefrin and Meir Statman gave the pattern its name in a 1985 Journal of Finance paper: the disposition effect is the tendency to sell investments that have risen and keep those that have fallen. Prospect theory predicts it. Below the purchase price, an investor is in the convex, risk-seeking part of the value function, and selling turns a paper loss into a real one.

Terrance Odean measured it in 1998 in 10,000 accounts at a large US discount broker from 1987 to 1993. Odean compared, on every day an investor sold something, the share of gains available in the portfolio that were realized with the share of losses realized.

• Investors realized 14.8% of the gains they could have taken and 9.8% of the losses.
• That ratio of a little over 1.5 means a stock trading above its purchase price was more than 50% more likely to be sold on a given day than one trading below it.
• The winners they sold beat the losers they kept by 3.4 percentage points over the following year, measured against the market.
• In December the pattern reversed, with a ratio of 0.85, as investors sold losers to cut their tax bill.

Under US tax rules a realized loss offsets gains and lowers the tax owed, so tax considerations favor selling losers. The accounts did the opposite for eleven months of the year.

Take an investor who needs 10,000 dollars in cash and holds two positions of equal size, one up 30% and one down 30%. Selling the winner feels good and books a gain. It also triggers capital gains tax, and it keeps the type of stock that lagged on average in Odean's data. Selling the loser feels like admitting a mistake, yet it creates a deductible loss and keeps the type of stock that did better on average.

Nicholas Barberis and Wei Xiong showed in 2009 that a model in which people feel gains and losses on their portfolio once a year often fails to predict a disposition effect. A model in which the pleasure and pain arrive when a gain or loss is realized, at the moment of sale, predicts it much more reliably.

Our overview of behavioral finance biases every investor should know covers the related biases that often travel with this one, such as anchoring on the 52-week high and crediting wins to skill.

Myopic loss aversion and the equity premium puzzle

In 1985 Rajnish Mehra and Edward Prescott described what they called the equity premium puzzle. From 1889 to 1978 the S&P 500 returned about 7% a year after inflation and Treasury bills less than 1%. To justify a gap that large, a standard economic model needs investors with a coefficient of relative risk aversion above 30, while most estimates put it close to 1.

Shlomo Benartzi and Richard Thaler proposed an explanation in 1995 and named it myopic loss aversion: loss aversion combined with frequent evaluation of the portfolio. Stocks fall in many individual months and years even though they have risen over most long periods. An investor who checks often sees many of those losses, and with a lambda of 2.25 each one weighs heavily. Benartzi and Thaler asked how often an investor with the 1992 prospect theory parameters would need to look at a portfolio to be indifferent between stocks and bonds at the historical returns. The answer came out close to one year, between 9 and 13 months depending on the data used.

Their simulations also show what longer evaluation periods would do to the premium such an investor demands:

For someone with a 20-year horizon who still judges the portfolio every year, the gap between 6.5% and 1.4% is what Benartzi and Thaler called the price of excessive vigilance: 5.1 percentage points a year that go, in effect, to the investors able to look less often.

In 1997 Thaler, Tversky, Kahneman and Alan Schwartz had subjects split money between a volatile "stock" fund and a steadier "bond" fund, with feedback at different intervals. Subjects who saw results monthly ended with 59.1% in bonds. Those who saw yearly results kept 30.4% in bonds, and those who saw five-yearly results 33.8%. In the monthly condition 39% of the periods showed a loss on stocks, against 14% of the yearly periods. The monthly group also finished with the lowest earnings of the three.

In a 2005 experiment by Michael Haigh and John List, traders recruited from the Chicago Board of Trade behaved more in line with myopic loss aversion than students given the same task.

When Benartzi and Thaler showed employees in a pension plan the distribution of one-year returns, they chose to put about 40% of their retirement money in stocks. Employees from the same organization shown simulated 30-year returns chose about 90%.

Loss aversion outside the stock market

The same asymmetry appears in house prices, professional sport and simple trading experiments with everyday objects.

Condo sellers in Boston

David Genesove and Christopher Mayer studied sales of downtown Boston condominiums in the 1990s. Owners facing a nominal loss on their purchase price set asking prices higher by 25% to 35% of the gap between the expected market price and what they had paid, sold for 3% to 18% of that gap more, and were less likely to sell at all. The effect among owner-occupants was about twice that among investors.

Professional golfers on the PGA Tour

Devin Pope and Maurice Schweitzer analyzed more than 2.5 million putts measured by laser. From the same distance, golfers made birdie putts about 2 percentage points less often than par putts. The authors read this as loss aversion around par: a missed par putt adds a bogey, which registers as a loss, so players put more effort into it. Tiger Woods showed the pattern too. The authors estimated that putting for birdie as well as for par would have been worth about one stroke per tournament, and on average 640,000 dollars more in 2007 prize money for each of the top 20 players.

Coffee mugs in the classroom

In experiments published in 1990 by Kahneman, Jack Knetsch and Thaler, half of a Cornell class received mugs that sold for 6 dollars at the campus bookstore and markets were then run between owners and non-owners. Owners asked a median of 5.25 dollars to sell, while buyers offered 2.25 to 2.75 dollars, so few mugs changed hands. A follow-up with Simon Fraser students added a third group who chose between a mug and cash: sellers valued the mug at 7.12 dollars, choosers at 3.12 dollars and buyers at 2.87 dollars. Choosers stood to gain a mug or cash and lose nothing, and they valued it like the buyers. Economists call this the endowment effect.

How investors can limit the effect of loss aversion

The methods with evidence behind them change the conditions under which a decision is made, so that losses are seen less often, counted over a whole portfolio and settled by rules written in advance.

1. Look at the portfolio less often. The feedback experiments above cut the share left in bonds from 59.1% to 30.4% by moving from monthly to yearly results. Turning off daily balance notifications and reviewing on a fixed date each quarter or year reduces the number of losses an investor sees without changing the returns.
2. Judge the portfolio as a whole. Samuelson's colleague refused one bet and accepted a hundred. Two of his bets combined give a 25% chance of +400, 50% of +100 and 25% of -200 dollars, which keeps a positive value even with lambda at 2.25, because gains and losses start to offset.
3. Write sell and rebalancing rules before buying. A target allocation, rebalanced on a calendar date or when a weight drifts by a set amount, makes the decision before the loss exists. The rule then sells what has risen and buys what has fallen, the opposite of the disposition effect.
4. Ask whether you would buy the holding today at today's price. The purchase price is the reference point that loss aversion runs on, and it has no bearing on future returns. In Odean's data, the losers investors held on to went on to lag the winners they sold.
5. Treat realized losses as a tax asset. Selling a loser in a taxable account creates a loss that offsets capital gains. In the US, the wash sale rule disallows the loss if substantially identical securities are bought within 30 days before or after the sale, so any replacement has to be a different holding or wait out the window.
6. Commit future money in advance. Thaler and Benartzi's Save More Tomorrow plan asked employees to commit part of future pay raises to retirement saving, so take-home pay never fell and the increase never felt like a loss. 78% of those offered the plan joined, and their average saving rate rose from 3.5% to 13.6% over 40 months.

A rules-based index fund takes several of these steps out of the investor's hands, since it holds every stock in its index and the investor never picks which loser to sell. Our comparison of active and passive investing sets out what that choice costs and returns.

This article is general information and does not constitute investment advice.

Frequently asked questions

What is loss aversion in simple terms?

Loss aversion is the tendency to weigh a loss more heavily than a gain of the same size. Losing 100 dollars moves most people more than finding 100 dollars, so they turn down bets and hold on to positions that a neutral calculation would accept or sell. Gains and losses are measured from a reference point, usually the price paid or the amount a person expected to have.

Do losses really hurt twice as much as gains?

On average across published studies, roughly. Tversky and Kahneman estimated a loss aversion coefficient of 2.25 in 1992, and a 2024 meta-analysis of 607 estimates put the mean at 1.955. The figure moves a lot with the method, though. A 2024 meta-analysis that refitted prospect theory to individual risky choices found 1.31, and a 2025 reanalysis found about 1.07 in experiments where gains and losses were the same size and shown in random order.

What is an example of loss aversion in investing?

Holding a falling stock to avoid selling it below the purchase price, while selling rising stocks quickly to lock in the gain. In trading records for 10,000 US brokerage accounts from 1987 to 1993, investors sold 14.8% of the gains available to them against 9.8% of the losses, and the winners they sold went on to beat the losers they kept by 3.4 percentage points over the next year.

What is the difference between loss aversion and risk aversion?

Risk aversion means preferring a sure amount to a gamble with the same or slightly higher expected value. Loss aversion is about the asymmetry between losses and gains around a reference point, and it matters most for choices that mix the two. The two can pull in opposite directions: people who are risk averse over gains often become risk seeking once they face a sure loss.

How can investors reduce loss aversion?

The methods with evidence behind them change the setting of the decision. Checking the portfolio less often, judging it as a whole, writing sell and rebalancing rules before buying, and committing future savings in advance all reduce how often a loss is felt and acted on. In one experiment, people who saw monthly results kept 59.1% of their money in bonds, against 30.4% for people who saw yearly results.