Loss Aversion

011 min

Where loss aversion shows up

An illustrative case, using a role rather than a person. A founder is deciding whether to move their team off a payment provider. The new provider charges 0.3 percentage points less per transaction, which on current volume is about $40,000 a year. Migration will take a quarter and carries a real chance of broken payouts during the switch.

They frame it in the meeting as: are we willing to risk outages to gain $40,000? Put that way, the answer is no, and it sounds prudent.

Change one input and run it again. Suppose the team had already migrated last year, and the question is whether to move back to the old provider. Same two providers, same fee difference, same migration risk. Now the framing is: are we willing to risk outages in order to lose $40,000 a year? Nobody argues for it, and the meeting is short.

The cash flows are identical in both runs. What differs is which provider counts as the starting point, and therefore which number appears as a loss.

That is the pattern. Loss aversion is not a preference for safety. It is an asymmetry in how the same difference is weighted depending on which side of the reference point it falls. The founder in the first run was not being careful. They were pricing a gain at a discount.

022 min

Why loss aversion happens

The mechanism is a claim about how outcomes are represented before they are compared.

Standard economic models assume people evaluate a choice by its effect on final wealth. Kahneman and Tversky's 1979 alternative, prospect theory, assumes something different: people evaluate changes relative to a reference point, which is usually the status quo. Losing $500 from $10,000 and gaining $500 from $9,000 leave a person in the same place, and prospect theory predicts they will not feel the same.

The shape of the value function

The paper states the shape precisely in its abstract. The value function "is normally concave for gains, commonly convex for losses, and is generally steeper for losses than for gains."

Three separate claims are packed into that sentence, and only the third is loss aversion:

• Concave for gains means each additional unit of gain adds less than the one before. The second $500 feels smaller than the first.
• Convex for losses means the same diminishing pattern applies downward. Losing $1,000 does not hurt twice as much as losing $500.
• Steeper for losses is the asymmetry itself. At any given size, the drop on the loss side is larger than the rise on the gain side.

The authors put the everyday version plainly: "A salient characteristic of attitudes to changes in welfare is that losses loom larger than gains." Their evidence includes a simple observation anyone can check on themselves. Most people decline a coin flip that wins $100 or loses $100, and the reluctance grows as the stake rises.

The reference point is the moving part

The asymmetry is fixed. What is not fixed is where zero sits. A quarterly target, last year's number, a competitor's price, what a colleague earns, the plan you already announced — each of these can become the reference point, and each turns different outcomes into losses. This is why the same result can be experienced as a win or a failure depending on what it is measured against, and why the choice of a target is never a neutral act of measurement.

On the adaptive explanation: the usual argument is that an organism can be ruined by one large loss but rarely transformed by one equivalent gain, so weighting losses more heavily is a reasonable default. That is a plausible story rather than a settled finding, and it is worth holding loosely.

031 min

Where loss aversion comes from

Daniel Kahneman and Amos Tversky introduced it in Prospect Theory: An Analysis of Decision under Risk, published in Econometrica in 1979. The paper is framed as a critique rather than a discovery. Its target was expected utility theory, the dominant account of decision-making under risk, which had been treated both as a normative model of rational choice and as a descriptive model of what people actually do.

Their method was to construct choices where the two came apart. Presented with a sure gain against a gamble of equal expected value, most people took the sure thing. Presented with a sure loss against a gamble of equal expected value, most people took the gamble. The same person was risk-averse in one frame and risk-seeking in the other, and expected utility theory has no room for that, because it evaluates final states rather than changes.

The paper names two further effects alongside loss aversion. The certainty effect is the tendency to overweight outcomes that are certain relative to those that are merely probable. The isolation effect is the tendency to discard components shared by the options under consideration, which produces inconsistent preferences when one choice is described in two ways.

Loss aversion did not arrive as a standalone bias. It arrived as one consequence of replacing final wealth with reference-dependent change.

041 min

Individual effects

For one person, the distorted decision is any comparison where the two options sit on opposite sides of a reference point.

The direction of the error is predictable: options framed as avoiding a loss are chosen too readily, and options framed as achieving a gain are chosen too rarely, even when the underlying numbers are identical.

The concrete cost is a portfolio of held positions that would not be bought today. An engineer keeps a deprecated dependency because removal might break something, while cheerfully accepting the equivalent risk of a new dependency that offers an upgrade. A manager does not move a report into a role that suits them better, because the team would lose capacity now and gain it later.

There is a second-order cost that shows up in how decisions get argued rather than in the decisions themselves. Because the asymmetry runs on framing, whoever describes the options first has already moved the answer. In a meeting, the person who says "if we do this we lose X" has done more work than anyone who follows them with a spreadsheet.

051 min

Systemic effects

Across an organisation, loss aversion compounds through the mechanism it is most often used to fix: the target.

A target creates a reference point by definition, and everything below it becomes a loss. That is exactly why targets motivate. It is also why they distort. Effort concentrates on the region immediately below the line, because that is where the steep part of the value function sits, and falls away above it, because gains above target are on the shallow part.

The pattern shows up wherever a salient threshold exists and outcomes are measured against it. Quarterly numbers, service level objectives, and sprint commitments all create the same geometry. A team that will work a weekend to avoid missing a commitment will not work that weekend to exceed it by the same margin, and the work is worth the same in both cases.

The compounding is hard to see in the records, because the metric that created the reference point is usually the metric used to judge whether the system is working. A dashboard showing most sprints landing just inside their commitment can mean the estimates are good, or it can mean the line is being defended.

062 min

Examples

A real case with 2.5 million observations. Devin Pope and Maurice Schweitzer set out to test whether experience, competition and high stakes eliminate the bias, which is the standard objection to laboratory findings. They chose professional golf on the PGA Tour, and published the result in the American Economic Review in 2011 under the title Is Tiger Woods Loss Averse?

The setting is well chosen and worth understanding, because the argument rests on it. A golfer's score is the total number of strokes across the tournament, so economically every stroke is worth exactly the same. But each individual hole carries a salient reference point: par. A putt to make par avoids a loss against that reference. A putt to make birdie achieves a gain. Both are worth one stroke.

Analysing over 2.5 million putts with precise laser measurements of distance, the authors find evidence that even the best golfers, Tiger Woods included, show loss aversion. These are competitors with enormous experience, direct competitive pressure and large amounts of money at stake, which is the combination that was supposed to make the bias disappear.

The finding that transfers is not about golf. It is that a reference point which carries no economic meaning still changes behaviour among experts who have every incentive to ignore it. Par is an accounting convention. It moves professionals anyway.

An illustrative contrast. A subscription product tests two messages on the same cancellation screen. One reads keep unlimited exports and priority support. The other reads you will lose unlimited exports and priority support. The features are identical and the price is identical. The second is describing the same state of the world from the other side of the reference point, which is the entire difference between them.

072 min

How loss aversion shows up in product, design, and AI

Deliberately used: framing what is at stake. Cancellation flows, renewal notices and downgrade screens all describe the same facts from the loss side. So do free trials, which is a mechanism the endowment effect entry covers in detail. The judgement is about accuracy rather than technique. Telling a user what they will lose is honest when the list is true and complete. It becomes manipulation when the loss is manufactured, for example by making export harder than it needs to be so there is something to lose.

Accidentally suffered: roadmap decisions. Almost every prioritisation argument has a loss side and a gain side, and they are rarely weighted evenly. Removing anything is a loss to somebody in the room. Adding anything is a gain to somebody who is not. This asymmetry is one reason backlogs grow and surface area never shrinks.

Accidentally suffered: pricing changes. A price increase is a loss to every existing customer and a neutral fact to every future one. Teams routinely underestimate the reaction, because they model the change against the new price rather than against what customers currently pay.

In evaluation and experiment design. Watch the reference point in how a test is reported. "The new model loses 2 points on this benchmark" and "the old model gains 2 points" describe one number, and the first will get more argument. When a team is deciding whether to ship a model, fix in advance what counts as the baseline, because whichever system holds that position is getting a weighting advantage that has nothing to do with quality.

082 min

How to guard against loss aversion

Knowing about the asymmetry does not remove it. What changes decisions is altering the frame or removing the reference point that created it.

Write both frames, then decide. For any consequential choice, state it once as a gain and once as a loss, in writing, before discussing it. If the two versions produce different intuitions, the difference is the bias, measured on your own decision.

The second question deletes the reference point rather than arguing against it.

Name the reference point out loud. Most arguments about whether something is a loss are actually arguments about where zero is. Making that explicit converts an unresolvable disagreement into a specific one.

Aggregate decisions rather than taking them one at a time. A series of independent bets each of which is favourable will usually be rejected individually and accepted as a portfolio, because the loss side of each one is evaluated separately. Review a class of decisions together where you can.

Build experience deliberately. Mrkva, Johnson, Gächter and Herrmann tested what moderates loss aversion across five samples totalling 17,720 participants. More domain knowledge and experience were associated with lower loss aversion, though people at every level of knowledge and experience were still loss averse. Among car buyers, those who knew more about a specific attribute were less loss averse for that attribute and not for others. The practical reading: familiarity with the particular decision reduces the effect, and general cleverness does not.

092 min

Common misunderstandings

"Loss aversion means people are risk-averse." The opposite happens on the loss side. Prospect theory's whole argument rests on people being risk-seeking when choosing between a sure loss and a gamble. A team that is behind on a quarter takes bigger swings, not smaller ones, and that is the theory working rather than failing.

"Losses count exactly twice as much." A specific ratio is frequently quoted as if it were a constant of nature. The 1979 paper claims the function is steeper for losses, not that it has a fixed multiplier, and the estimates that produced a single number came later and from particular samples. Treat the asymmetry as real and its magnitude as situational.

"It is the same as the endowment effect." The endowment effect is one specific consequence of loss aversion, about owned objects and their prices, and it has its own entry. Loss aversion is the broader asymmetry, and it appears in choices where nothing is owned at all.

"Framing it as a loss is a dark pattern." Sometimes, and not inherently. If a user genuinely will lose something, the loss frame is the accurate description. The honest test is whether the statement would survive the user checking it.

Where the effect is useful rather than a defect. A strong weight on losses is a reasonable policy when losses are genuinely harder to recover from than gains are to repeat, which describes a lot of operational reality. Production outages, data loss and reputational damage are not symmetric with their upside. The error is applying the asymmetry to decisions where the two sides really are reversible.

101 min

Loss aversion vs. nearby concepts

The distinguishing question when these get tangled in a meeting: is the reluctance about something already spent, something currently owned, or the shape of the prospective outcome? Those are sunk cost, endowment effect and loss aversion in that order.

112 min

Where the evidence is contested

Loss aversion is among the most cited findings in behavioural science and it has been directly challenged, which is worth knowing before quoting it as settled.

David Gal and Derek Rucker published the challenge in the Journal of Consumer Psychology in 2018, under the title The Loss of Loss Aversion: Will It Loom Larger Than Its Gain? Their argument is that the evidence does not support a general principle that losses loom larger than gains, and that the effect is contingent on context rather than a basic law of decision-making. They treat the widely repeated two-to-one framing as a generalisation that outran what the underlying studies established.

Stated at its strongest, the objection is not that the demonstrations are fake. It is that a real effect observed under particular conditions was promoted to a universal principle, and that the conditions were dropped along the way.

The substantial reply came in the same journal in 2020 from Kellen Mrkva, Eric Johnson, Simon Gächter and Andreas Herrmann, whose title makes their position clear: Loss Aversion Has Moderators, But Reports of its Death are Greatly Exaggerated. Testing across five samples with 17,720 participants in total, they report that domain knowledge and experience reduce loss aversion, while participants at all levels of knowledge and experience remained loss averse. Their position is that the critics identified real moderators and drew too strong a conclusion from them.

Where this leaves a practitioner is usable. Expect the asymmetry, expect it to be weaker among people who know the domain well, and stop quoting a fixed multiplier. For designing a decision process, the disagreement barely matters, because the fix — write both frames and compare them — costs almost nothing and is worth doing even if the effect on a given decision turns out to be small.

?8 questions

Questions people ask

▶3 videos

Watch

• 

  Understanding Loss Aversion

• 

  CRITICAL THINKING - Cognitive Biases: Reference Dependence and Loss Aversion [HD]

  Wireless Philosophy

• 

  The psychology behind irrational decisions - Sara Garofalo

  TED-Ed

§6 sources

Sources