Forty articles into a series about which findings survive scrutiny, the obligation runs backwards as well as forwards. This one applies the standard to something this publication has already told its readers.
Key Takeaway
Gal and Rucker (2018), in the Journal of Consumer Psychology: "Loss aversion, the principle that losses loom larger than gains, is among the most widely accepted ideas in the social sciences... The upshot of this review is that current evidence does not support that losses, on balance, tend to be any more impactful than gains."[1] A reply in the same dialogue: "Moderating loss aversion: Loss aversion has moderators, but reports of its death are greatly exaggerated."[2]
The Correction, Stated First
Before anything else, because a correction buried at the end is not a correction.
An earlier article on this site states that loss aversion "is among the most robustly replicated results in behavioural economics" and describes losses as felt "roughly twice as intensely as equivalent gains", in four separate places.
Three things about that.
The first clause is too strong. A 2018 paper in a peer-reviewed journal reviews the evidence and concludes it does not support the general claim, and that paper drew a formal published dialogue rather than being ignored.
The second clause is not sourced anywhere in this series. We have not obtained a meta-analytic estimate of the magnitude and, as this article explains, we do not assert one.
And we are not withdrawing the earlier article, whose subject is the disposition effect rather than the coefficient. What we are doing is qualifying a claim it makes in passing, publicly, in a piece of its own.
Our Grades For These Claims
Applying the scheme from the first article in this series.
Grade A that the question is genuinely contested. A review in a peer-reviewed journal, a formal research dialogue with named commentaries, and at least two meta-analyses.
Grade C for the general claim that losses systematically outweigh gains, which is precisely what the 2018 review disputes and what its critics defend.
Grade D for any specific coefficient, including the one this publication previously implied, because we could not source a meta-analytic estimate at all.
Grade B that moderators exist, on a paper whose title asserts them and which we did not obtain.
Our position: the direction may well be right and the magnitude is not established to our satisfaction, and the second matters more than the first for anything a business would do.
A Note On Method
Everything here is verified to August 2026.
We obtained the 2018 review's abstract verbatim from four independent sources, including the publisher and a hosted copy of the paper itself[1][3][4][7].
We did not obtain any paper in full, including the review, the commentaries, the defence, or either meta-analysis.
We could not obtain a meta-analytic point estimate of the loss aversion coefficient, and therefore assert none anywhere in this article. That is a significant gap and it has its own section.
All arithmetic is ours and uses a standard indifference formulation with invented figures.
This article reviews a scientific dispute. It is not investment, financial or risk management advice.
What We Published
The claim under correction, quoted from our own site.
The earlier article's key takeaway reads: "Loss aversion, the finding that losses are experienced roughly twice as intensely as equivalent gains, is among the most robustly replicated results in behavioural economics and forms the core of Kahneman and Tversky's prospect theory."
Its opening states that "losses hurt roughly twice as much as equivalent gains feel good" and that this asymmetry, "more than any other finding in behavioural economics, predicts what Canadian owners actually do."
Two observations, ours.
The phrase most robustly replicated is the specific problem. It is a claim about the state of the evidence, and the state of the evidence is disputed in print.
And the article's substantive subject is the disposition effect, being the tendency to hold losers and sell winners, which is a documented pattern in trading data independent of any particular explanation for it. That subject is not what we are correcting.
The Challenge
The paper, quoted at length because its own framing is careful.
Gal and Rucker published The Loss of Loss Aversion: Will It Loom Larger Than Its Gain? in the Journal of Consumer Psychology, 28(3), 497–516, in 2018, DOI 10.1002/jcpy.1047[1].
"Loss aversion, the principle that losses loom larger than gains, is among the most widely accepted ideas in the social sciences. The first part of this article introduces and discusses the construct of loss aversion. The second part of this article reviews evidence in support of loss aversion. The upshot of this review is that current evidence does not support that losses, on balance, tend to be any more impactful than gains."[1]
Three observations, ours.
Note how radical the claim is. It is not that the coefficient is smaller than believed. It is that the evidence does not support any general asymmetry at all, which is a different and stronger objection.
Note also the hedges: on balance, and tend to. The authors are making a claim about the aggregate of the evidence, not asserting that losses never loom larger.
And we did not obtain the paper's evidence review, only its abstract. We can report what they concluded and not how they got there.
The Third Part Of Their Paper
The section that makes this relevant to everything else in this series.
"The third part of this article aims to address the question of why acceptance of loss aversion as a general principle remains pervasive and persistent among social scientists, including consumer psychologists, despite evidence to the contrary. This analysis aims to connect the persistence of a belief in loss aversion to more general ideas about belief acceptance and persistence in science."[1]
Three observations, ours.
A third of the paper is about why the field believes something the authors think it should not. That is an unusual and confrontational structure, and it is why the paper drew the response it did.
It is also the structure this series keeps encountering from the outside. The nineteenth article found a refuting meta-analysis cited as support for the thing it refuted; the thirty-fourth found a debunking that stood for thirty-three years. Here a paper makes the persistence of belief its explicit subject.
And we note, since we are the ones being corrected, that this publication is an instance of the phenomenon they describe. We repeated a widely accepted claim without checking its evidentiary status, because it is widely accepted.
A Published Dialogue
What happened next, which is the system working.
The publisher records that the article "is part of a Research Dialogue" with named companion pieces: Shavitt (2018), DOI 10.1002/jcpy.1054; Simonson and Kivetz (2018), DOI 10.1002/jcpy.1046; and Higgins and Liberman (2018)[1].
We did not obtain any of the commentaries and report their existence and citations only.
Two observations, ours.
A journal commissioning a formal dialogue around a challenge to a central finding is the correction process working, in the same way the twentieth article's adversarial collaboration was.
And it means a reader can see the disagreement in print rather than in a corridor, which this series has repeatedly found to be the exception rather than the rule.
The Critics Who Partly Agree
The most useful sentence we found, because it is neither dismissal nor endorsement.
A commentary states: "Although we disagree with some of Gal and Rucker's (2018 – this issue) specific evidence and with their overstated conclusion regarding loss aversion, their overarching message makes a worthwhile" contribution, with our source truncating[3].
Three observations, ours.
That is three positions in one sentence: disagreement on specific evidence, a charge that the conclusion is overstated, and an acknowledgement that the overarching message is worthwhile.
The word overstated is the critics' assessment and we report it as theirs. A reader should weigh it against the abstract's own hedges.
And it is the position we would probably land on ourselves: the challenge is more right about the state of the evidence than about the state of the world.
Reports Of Its Death
The defence, whose title is its argument.
Mrkva, Johnson, Gächter and Herrmann published Moderating loss aversion: Loss aversion has moderators, but reports of its death are greatly exaggerated in the Journal of Consumer Psychology, 30(3), 407–428, in 2020[2].
We did not obtain it and report the title only.
Two observations, ours.
The title concedes moderators, which is itself a substantial concession. A finding with moderators is not a general principle; it is a conditional one, and the whole dispute is about whether the general version holds.
And the position it stakes out is the one this series has arrived at repeatedly: the third article found choice overload appearing only under specific conditions, the twentieth found a happiness plateau only in a subgroup, the twenty-ninth found expectancy effects concentrated rather than general. An average across heterogeneous conditions describes nobody in particular.
Small Losses
A more recent finding, with a title that states its result.
A 2022 paper in Judgment and Decision Making is titled Loss aversion (simply) does not materialize for smaller losses, and states: "These findings challenge the claim that loss aversion is reliable and robust."[5]
A separate source records the emerging position as "a more nuanced hypothesis: loss aversion depends on the magnitude of the stake"[6].
We obtained neither paper and report titles and one quoted sentence.
Three observations, ours.
If the effect appears at large stakes and not small ones, that is a boundary condition rather than a refutation, and it is a commercially important one.
Most decisions a firm makes are small relative to the firm. If loss aversion does not materialise at small stakes, then it is least relevant precisely where decisions are most frequent.
And it would explain the dispute's persistence. Two researchers studying different stake sizes would reach opposite conclusions honestly, which is a better explanation of a long disagreement than either side being careless.
Kahneman On The Limits
A detail worth recording, because it complicates the usual framing of this dispute.
Indexed text describing a commentary states: "In defining limits to loss aversion, Novemsky and Kahneman (2005) offer important new data and a needed summary of appropriate ways to think about loss aversion."[3]
We did not obtain either paper and report this description only.
Two observations, ours.
The framing of this dispute as challengers versus the originators is too simple. One of prospect theory's own authors co-wrote a paper defining limits to loss aversion in 2005, thirteen years before the challenge.
Which is the same pattern the twentieth article found, where Kahneman co-authored the adversarial collaboration that revised his own earlier result. The originators have been narrowing the claim for longer than the critics have been attacking it.
The Meta-Analyses We Could Not Obtain
The gap, and it is the most consequential one in this article.
Two meta-analyses exist. Reference records identify Walasek, L., Mullett, T. L., and Stewart, N., A Meta-Analysis of Loss Aversion in Risky Contexts, posted in June 2018 and last revised in April 2024[4]; and Brown, A. L., Imai, T., Vieider, F., and Camerer, C. F. (2021), Meta-analysis of empirical estimates of loss-aversion, CESifo Working Paper No. 8848[5].
We obtained neither, and therefore we do not report a coefficient anywhere in this article.
Three observations, ours.
This is why the earlier article's "roughly twice" is being withdrawn rather than revised downward. We are not replacing one unsourced number with another.
A reader who needs the magnitude must go to those two papers, and we would note the second was still being revised in 2024, six years after first posting, which suggests the estimate was not simple to pin down.
And this is the fourth time in this series that the single most decision-relevant document was the one we could not get. We record that as a limitation of our method rather than of the literature.
Why The Magnitude Decides Everything
The reason the missing number matters more than the direction. Ours.
Suppose someone faces a coin flip: win an amount, or lose $100. Under a loss aversion coefficient, the smallest acceptable win is that coefficient times the downside.
Three consequences.
Knowing only that losses loom larger tells you the coefficient exceeds one. It does not tell you whether it is 1.1 or 2.5, and those imply completely different behaviour.
Which means the directional claim is not actionable. A statement that cannot distinguish between accepting and declining a given opportunity cannot inform a decision about it.
And this is a general failing of how behavioural findings are transmitted, which this series has now hit repeatedly: the direction survives popularisation and the magnitude does not, and the magnitude is the part you need.
The Arithmetic Of The Disagreement
Making that concrete. Ours, with invented figures and a standard indifference formulation; no source states these numbers and we assert no coefficient.
On a $100 downside, the smallest acceptable win is $100 at a coefficient of 1.0, $150 at 1.5, $200 at 2.0, and $250 at 2.5.
Now suppose a firm faces twenty independent opportunities a year, each a coin flip paying +$150 or -$100, so each has an expected value of +$25.
At a coefficient of 1.0 or 1.5, the firm accepts all twenty, for an expected annual value of $500.
At 2.0 or 2.25, it accepts none, for an expected annual value of zero.
Two observations.
The entire difference between capturing every positive-expected-value opportunity and capturing none sits between 1.5 and 2.0, which is inside the range the dispute concerns.
And notice this is not an argument that loss aversion is irrational. A firm that cannot survive a run of losses is right to require better than fair odds, and the arithmetic above assumes the losses are survivable, which for a real business is a question about the balance sheet rather than about psychology.
What Survives
Our honest residue.
Four statements we think hold.
The question is live and in print. Not settled in either direction, with a formal dialogue and named commentaries on both sides.
Moderators are conceded by the defenders, on the title of the 2020 paper, which means the general version is not what is being defended.
Stake size looks like a real boundary, on a 2022 paper whose title asserts the effect does not materialise for smaller losses.
And the disposition effect is a separate matter. Holding losers and selling winners is a pattern observed in trading records. Whether loss aversion is its cause is a question about mechanism, and the pattern does not stop existing if the mechanism is wrong.
What We Say Now
The replacement wording, so that the correction is usable rather than merely apologetic. Ours.
What we would write today, in place of the earlier claim.
People often, but not always, weigh a prospective loss more heavily than an equivalent gain. How much more heavily is disputed, appears to depend on the size of the stake and other moderators, and is not a number this publication is in a position to give.
Two observations.
That sentence is less useful and more accurate, and we would rather publish it.
And it changes what a reader should do with the idea. It is a reason to look for the asymmetry in your own decisions, not a coefficient to apply to them.
What Happens To The Earlier Article
Being specific about the remedy, since a correction that changes nothing is a gesture. Ours.
Three points.
The earlier article remains published. Its subject is the disposition effect, its account of that pattern does not depend on the coefficient, and removing it would destroy more than it fixes.
Its claim that loss aversion is among the most robustly replicated results is the part we are withdrawing, and this article is the withdrawal.
And we would rather correct in public than edit in silence. A quietly amended page leaves no record that the publication got something wrong, which is precisely the failure this series has criticised in others. The thirtieth article turned its four questions on this publication; this one turns the correction discipline on it.
The General Lesson
The transferable part, and it is not about loss aversion. Ours.
The error we made was specific and common: we repeated a claim about the state of the evidence, rather than a claim from the evidence.
Three points.
Saying a finding is robustly replicated is itself an empirical assertion, and it requires its own sourcing, which we did not have.
It is also the assertion least likely to be checked, because it sounds like context rather than content. Nobody demands a citation for a compliment paid to a literature.
And it is the assertion that ages worst, because the state of evidence changes while findings do not. A paper published in 1979 is what it was; whether it has held up is a live question in a way the paper itself is not.
What To Do
Distrust claims about how well replicated something is. That is an empirical assertion needing its own source, and it is the one least likely to be challenged.
Ask for the coefficient, not the direction. Knowing losses loom larger tells you a number exceeds one, which cannot distinguish accepting an opportunity from declining it.
Treat the general version as contested. A 2018 review in a peer-reviewed journal concluded current evidence does not support that losses are on balance more impactful, and drew a formal dialogue rather than silence.
Note that the defenders concede moderators. The 2020 reply's own title does, which means the conditional version is what is being defended.
Watch stake size. A 2022 paper's title asserts the effect does not materialise for smaller losses, which would make it least relevant where decisions are most frequent.
Separate the pattern from the mechanism. Holding losers and selling winners is observed in trading records whether or not loss aversion explains it.
Check whether your caution is psychology or solvency. Requiring better than fair odds is correct when a run of losses would be fatal, and that is a balance sheet question rather than a bias.
Correct in public. A silently edited page leaves no record that anything was wrong.
The Limits Of This Analysis
Several caveats matter. This article reviews a scientific dispute and is not investment, financial or risk management advice. Everything is verified to August 2026. We did not obtain any paper discussed in full. We have the 2018 review's abstract verbatim from four independent agreeing sources and none of its evidence review. We did not obtain any of the three named commentaries, the 2020 defence, the 2022 paper on small losses, the 2005 paper on limits, or either meta-analysis, and report titles, citations and single quoted sentences only. We could not obtain a meta-analytic estimate of the loss aversion coefficient and assert none anywhere, which is why the earlier claim is withdrawn rather than revised. All arithmetic is ours, uses invented figures and a standard indifference formulation, and the coefficients shown are illustrative values rather than estimates. The characterisation of the commentary as partly agreeing rests on a single truncated sentence reproduced by an indexing service. The replacement wording, the treatment of the earlier article, and the general lesson are our own reasoning. This article corrects a claim made in another article on this site; that article remains published and its principal subject, the disposition effect, is not what is being corrected.
Frequently Asked Questions
What are you correcting?
What does the challenge actually say?
Did anyone defend it?
So what is the number?
Why does the number matter more than the direction?
Does this mean the earlier article was wrong?
References
- 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, publisher record reproducing the abstract, on loss aversion, the principle that losses loom larger than gains, being among the most widely accepted ideas in the social sciences; on the first part of the article introducing and discussing the construct and the second part reviewing evidence in support of it; on the upshot of that review being that current evidence does not support that losses, on balance, tend to be any more impactful than gains; on the third part addressing why acceptance of loss aversion as a general principle remains pervasive and persistent among social scientists despite evidence to the contrary, connecting that persistence to more general ideas about belief acceptance and persistence in science; and on the article being part of a Research Dialogue with companion pieces by Shavitt (2018), DOI 10.1002/jcpy.1054, Simonson and Kivetz (2018), DOI 10.1002/jcpy.1046, and Higgins and Liberman (2018). Note: the publisher's record. We obtained the abstract in full and none of the paper's evidence review, and none of the three commentaries. myscp.onlinelibrary.wiley.com
- Reference list on a behavioural economics reference site, identifying 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; Gal, D., & Rucker, D. D. (2018), Journal of Consumer Psychology, 28(3), 497–516; Kahneman, D., & Tversky, A. (1979), Prospect theory: An analysis of decision under risk, Econometrica, 47, 263–291; and Wang, M., Rieger, M. O., & Hens, T. (2017), The impact of culture on loss aversion, Journal of Behavioral Decision Making, 30(2), 270–281. Note: a reference website; citations only. We did not obtain the 2020 paper and report its title, which states its argument. behavioraleconomics.com
- Bibliographic record for Gal and Rucker (2018), reproducing the abstract and carrying indexed text from citing works, including a commentary stating that although its authors disagree with some of Gal and Rucker's specific evidence and with their overstated conclusion regarding loss aversion, their overarching message makes a worthwhile contribution, our source truncating; and a description of Novemsky and Kahneman (2005) as offering, in defining limits to loss aversion, important new data and a needed summary of appropriate ways to think about loss aversion. Note: a bibliographic service reproducing text from citing papers. The partly-agreeing characterisation rests on a single truncated sentence; we did not obtain the commentary or the 2005 paper. semanticscholar.org
- Working paper repository record for Gal, David, and Rucker, Derek, The Loss of Loss Aversion: Will It Loom Larger Than Its Gain?, dated 30 September 2017 and listed as forthcoming in the Journal of Consumer Psychology, with the authors' institutional addresses in Chicago and Evanston; and its related listings identifying Walasek, Lukasz, Mullett, Timothy L., and Stewart, Neil, A Meta-Analysis of Loss Aversion in Risky Contexts, 46 pages, posted 17 June 2018 and last revised 11 April 2024, and Mrkva and colleagues, Moderating Loss Aversion: Loss Aversion Has Moderators, But Reports of its Death are Greatly Exaggerated, then forthcoming in the Journal of Consumer Psychology. Note: a working paper repository record. We did not obtain the meta-analysis and note only that it was still being revised in 2024, six years after first posting. papers.ssrn.com
- Article in Judgment and Decision Making titled Loss aversion (simply) does not materialize for smaller losses, stating that its findings challenge the claim that loss aversion is reliable and robust; and its reference list identifying Brown, A. L., Imai, T., Vieider, F., & Camerer, C. F. (2021), Meta-analysis of empirical estimates of loss-aversion, CESifo Working Paper No. 8848; Ert, E., & Erev, I. (2013), On the descriptive value of loss aversion in decisions under risk: Five clarifications, Judgment and Decision Making, 8, 214–235; and Gal, D. (2006), A psychological law of inertia and the illusion of loss aversion, Judgment and Decision Making, 1, 23–32. Note: we obtained the title and one sentence together with the reference list; we did not obtain the paper or the meta-analysis it cites. sas.upenn.edu
- Publisher record with third-party citing text for Gal and Rucker (2018), on the authors having highlighted the variability of loss aversion's degree and direction by context and thus the need for empirical models considering cultural, economic and psychological variables; on the statement that losses loom larger than gains having become accepted as a fundamental principle of human behavior, built on the general notion of bad being stronger than good and on prospect theory's value function; and on a citing paper proposing, instead of arguing for or against loss aversion, a more nuanced hypothesis that loss aversion depends on the magnitude of the stake. Note: third-party citing text on a publisher page; we obtained none of the papers it describes. researchgate.net
- Hosted copy of Gal and Rucker (2018), reproducing the abstract identically and showing section material on status quo bias, the endowment effect, retention paradigms and willingness-to-pay conditions, together with a figure caption on status quo bias in the absence of loss or gain coding with n = 149. Note: a hosted copy of the paper; we obtained the abstract and fragmentary section headings only, and none of the evidence review or its results. statmodeling.stat.columbia.edu
This article reviews a scientific dispute and is not investment, financial or risk management advice. No paper discussed was obtained in full. No meta-analytic estimate of the loss aversion coefficient could be obtained, and none is asserted anywhere in this article. All arithmetic is the authors' own and uses illustrative coefficient values rather than estimates. This article corrects a claim made in another article on this site, which remains published.