Loss Aversion: Why Losing 100 Dollars Hurts More Than Winning It

Decisions are made from a reference point rather than from a total. Prospect theory describes people as evaluating outcomes as gains and losses relative to where they currently stand, and the curve that turns outcomes into feelings is steeper on the loss side than on the gain side. Twenty dollars gained and twenty dollars lost are the same size on a bank statement and different sizes in a decision.

That asymmetry explains a pattern that pure arithmetic cannot. A fifty-fifty bet to win 110 dollars or lose 100 has a positive expected value and is still refused by people who understand the odds, because the prospective loss counts for more than the slightly larger gain. When Tversky and Kahneman fitted their 1992 model to the choices of 25 graduate students, the median coefficient on losses came out at 2.25, which is where the rule of thumb that losses weigh about twice as much comes from.

Because the reference point does the work, the same decision can be made to look like a gain or a loss. A 5 percent price increase presented as the loss of a discount is resisted more than the same number presented as a smaller gain. An employer offering a bonus for hitting a target gets a weaker response than one that hands it over up front and takes it back on a miss.

Loss aversion is also the simplest explanation for inertia. If giving something up hurts more than getting it would please, the current arrangement wins by default: the subscription stays, the portfolio is not rebalanced, the job is not left. The reluctance to sell a losing investment and the premium owners place on what they own follow from the same steepness.

The size and the generality of the effect are argued over, and what the asymmetry measures with them. Gal and Rucker reviewed the evidence in 2018 and concluded that losses are not reliably more impactful than gains once the studies that use status quo defaults and other confounds are set aside, and they describe the universal version of the claim as unsupported. They also question the gambles themselves: refusing a bet with a positive expected value is also consistent with ordinary risk aversion or a dislike of gambling, so a refused bet is not by itself proof of the asymmetry. Other work keeps the effect but draws boundaries around it: Novemsky and Kahneman proposed in 2005 that money given up as intended in an ordinary purchase is generally not subject to loss aversion, and set out other conditions under which the asymmetry weakens.

Daniel Kahneman and Amos Tversky set out prospect theory in Econometrica in 1979, in a paper that used hypothetical choice problems to document where expected utility theory fails. Respondents chose between pairs of gambles, and three patterns emerged: outcomes were valued as changes from a reference point rather than as final states, the pattern of choices reversed when gains were turned into losses, and small probabilities were overweighted. The paper's value function is concave for gains, convex for losses and steeper on the loss side, and the sentence that the steeper side produced, losses loom larger than gains, is the one that became the name of the effect.

The 1979 paper did not put a number on the steepness. That came in 1992, when Tversky and Kahneman published "Advances in prospect theory" in the Journal of Risk and Uncertainty. Twenty-five graduate students from Berkeley and Stanford, none with training in decision theory, worked through a long series of choices between gambles and sure amounts, and the authors fitted the theory's value and weighting functions to each person's answers. The median estimates were 0.88 for the curvature of the value function and 2.25 for the loss aversion coefficient. That 2.25 is an estimate from a small sample of students answering about hypothetical money, not a constant of human nature; later estimates vary.

Field studies then looked for the same asymmetry in real decisions, from investors who sell winners and hold losers to reluctance to trade goods people already own, and a long argument began about what the coefficient measures, which Gal and Rucker's 2018 review brought to a head by asking how much of the evidence rests on tasks where the alternative to a loss is doing nothing.

The pattern turns up wherever a choice can be framed around something a person already has. The rows below are illustrations, not findings from the studies cited on this page.

The counters all work by moving the reference point out of the way.

• Restate the options as final states. "End the year with 48,000 dollars or with 50,000 dollars" is the same choice as "lose 2,000" with the sting taken out of it.
• Decide once for many cases. A rule applied to a long run of similar bets collects the average, which for many favorable gambles is reliably positive even though each one feels risky.
• Use the ownership test. If you did not already hold this stock, this subscription or this plan, would you buy it now at today's price? If not, holding it is a decision you never actually made.
• Write the exit condition at the start. Rebalancing dates, stop rules and review dates are decisions taken before a loss exists to be avoided.
• Name the opportunity cost. Money left in a losing position is also money not working somewhere else, which is a loss that no reference point makes vivid.
• Expect the asymmetry in other people. Concessions, price rises and benefit changes land harder than the equivalent gains, so they need more notice, more explanation and sometimes a different design.

Loss aversion sits in a family of effects about holding, keeping and giving up, and the differences matter.

• It is not the endowment effect, which is the specific finding that people demand more to give up an object than they would pay to acquire it. Loss aversion is the general asymmetry that is usually offered as its explanation, and the two are tested in different ways.
• It is not the sunk cost fallacy, which is about past spending influencing a present decision. Loss aversion is about a prospective loss, and it can push someone to quit early as easily as to carry on.
• It is not status quo bias, the preference for leaving things as they are. Status quo bias describes the behavior; loss aversion is one of its proposed causes, and separating the two is the heart of the current argument about how large the effect really is.
• It is not risk aversion, the general reluctance to take a gamble over a sure thing. A refused even-money bet is consistent with either, so the demo above describes your answers without naming a cause.