CoChat

Talking the talk, not walking the walk: The coevolution of overconfidence and loss aversion

Open access
Dawson, Chris, de Meza, David · 2026 · Psychological Review

The DOI matches its Crossref record; 159 of the 162 references it cites were found in a registry or an index; 9 of the 35 sources its claims cite were checked for support.

Abstract

A puzzle for evolutionary theory is the existence of two seemingly offsetting behavioral "biases," overconfidence and loss aversion. Overconfidence is a call to action, while loss aversion curbs initiative. The most prominent evolutionary explanation of overconfidence, proposed by Trivers (1976), is that self-deceit arises to better deceive others. Missing from this account is why sincere messages are believed, especially given the widespread prevalence of self-deception. Moreover, if overconfidence is adaptive, why is it at least partially canceled by loss aversion? We propose a signaling theory according to which the role of self-deception is to better inform others. Since the decision error associated with high self-belief is less burdensome for the more able, and the benefit of being perceived as able increases with ability, hardwired overconfidence is a credible signal of true ability. Evidence supports this interpretation. A further implication of signaling is that loss aversion is part of the equilibrium. It partially ameliorates the decision costs of overconfidence, but as it is usually hidden, it does not eliminate its signaling role. "Biases" are thus symbiotic-the payoff to agents from an integrated set of biases is higher than would be the case in their absence. From this perspective, Kahneman's advice that individuals eliminate both overconfidence and loss aversion is poorly founded. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Overview

CoChat Summary

TL;DR

The paper proposes that overconfidence and loss aversion coevolve as a symbiotic signaling equilibrium: overconfidence credibly signals ability because decision-error costs are lower for more able individuals, while hidden loss aversion partially offsets those costs without destroying the signal. It argues against eliminating both biases, since their integrated payoff can exceed the no-bias benchmark.

Key findings

  • The model yields a separating equilibrium in which all types except the lowest are overconfident, and hidden loss aversion is part of that equilibrium; visible loss aversion would be pointless, and no single loss-aversion parameter can completely offset the aggregate decision-error cost of overconfidence (Lemma 1).
  • In the worked example (Table 1), equilibrium choices are γ: a = 0.1, b = 0.1, λ = 1.0, C = 0.00; β: a = 0.2, b = 0.4, λ = 2.2, C = 0.01; α: a = 0.6, b = 0.6, λ = 1.0, C = 0.00. A γ-type mimicking β at b = 0.4 incurs C = 3.23, and a β-type deviating to b = 0.6 incurs C = 0.27.
  • Welfare comparison (Table 2): unbiased decisions = $80.00 (100% of optimal); loss aversion with λ ≈ 3.5 = $68.00 (85%); risk aversion with σ ≈ 2.5 = $35.00 (44%); full bias = $22.50 (28%).
  • Empirical evidence from Jansen et al. (2021), approximately 4,000 Mechanical Turk participants: grammar mean correct = 10.17 (SD = 3.40) vs mean expectation = 12.49 (SD = 3.91); logic mean correct = 9.45 (SD = 3.59) vs mean expectation = 10.86 (SD = 4.05). Regressing actual on expected performance gave grammar β = 0.25, 95% CI [0.22, 0.28], t = 16.26, p < .001; logic β = 0.18, 95% CI [0.15, 0.21], t = 11.24, p < .001. Reduced major axis slopes were 1.15 (grammar) and 1.13 (logic); 25.35% (grammar) and 35.42% (logic) underestimated their performance.

Methods

  • Theoretical signaling model using perfect Bayesian equilibrium, single-crossing, continuous and discrete type distributions, and an extension with hidden loss aversion λ; numerical examples and simulations illustrate equilibrium and welfare comparisons.
  • Empirical evidence is a secondary analysis of Jansen et al. (2021): about 4,000 Mechanical Turk participants completed 20 grammar and 20 logic multiple-choice questions and estimated their correct answers; forecast accuracy was not incentivized.
  • Main empirical analysis regresses actual performance on expected performance via ordinary least squares, with reduced major axis regression and reliability corrections for measurement error; analysis script is openly available on OSF, and the study was not preregistered.

Limitations

  • The paper states that evolutionary arguments are difficult to test, unambiguous natural experiments are often challenging, human adaptation is slow, and purposeful experiments on human evolution are neither feasible nor ethical; it also notes the “Just-So” criticism of evolutionary theorizing.
  • The paper states that the framework does not yield unambiguous predictions for all relationships among key variables: the relation between overconfidence and ability need not be monotonic, unconditional predictions across groups are impossible, loss aversion need not be higher where overconfidence is greater, and no systematic correlation between loss aversion and overconfidence is predicted.
  • Evident from the text: the empirical support is a secondary analysis of one non-incentivized forecasting dataset rather than a direct test of coevolution or loss aversion; the model relies on assumptions such as single crossing, hidden loss aversion, and a non-domain-specific λ, and the study was not preregistered.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1037/rev0000644

Cites

159 of 162 references verified1 citation errors · 2 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

7 Valid · 1 Not valid · 13 Not checked · 104 Own results and interpretations

61 more claims · 1 with their cited sources checked

Create a free account to see all 125 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 50
  • Probably not a substantial claim. 15

Create a free account to see these claims.

Citations

Cites

Show all 162 references

Research works printed but not linked: 16

  1. Anderson, C., Brion, S., Moore, D. A., & Kennedy, J. A. (2012). A statusenhancement account of overcon fi dence. Journal of Personality and Social Psychology 103 -
  2. Grafen, A. (1984). Natural selection, kin selection and group selection. In J. R. Krebs & N. B. Davies (Eds.), Behavioural ecology (pp. 62 -84). Blackwell Scienti fi c Publications.
  3. G ü th, W., & Yaari, M. (1992). Explaining reciprocal behavior in simple strategic games: An evolutionary approach. In U. Witt (Ed.), Explaining forces and change: Approaches to evolutionary economics (pp. 23 -34). University of Michigan Press.
  4. Heller, Y. (2014). Overcon fi dence and diversi fi cation. American Economic Journal: Microeconomics , 6 (1), 134 -153. https://doi.org/10.1257/mic.6 .1.134 · doi:10.1257/mic.6
  5. Hume, D. (1896). A treatise of human nature . Clarendon Press.
  6. Hvide, H. K. (2002). Pragmatic beliefs and overcon fi dence. Journal of Economic Behavior and Organization , 48 (1), 15 -28. https://doi.org/10 .1016/s0167-2681(01)00221-9
  7. Mas-Colell, A., Whinston, M. D., & Green, J. R. (1995). Microeconomic theory (Vol. 1). Oxford University Press.
  8. Mijovi´ c-Prelec, D., & Prelec, D. (2010). Self-deception as self-signalling: A model and experimental evidence. Philosophical Transactions of the Royal Society B: Biological Sciences , 365 (1538), 227 -240. https://doi.org/10 .1098/rstb.2009.0218
  9. P á l, C., & Papp, B. (2017). Evolution of complex adaptations in molecular systems. Nature Ecology & Evolution , 1 (8), 1084 -1092. https://doi.org/10 .1038/s41559-017-0228-1
  10. Rabin, M. (1998). Psychology and economics. Journal of Economic Literature , 36 (1), 11 -46. https://www.jstor.org/stable/2564950
  11. Rayo, L., & Becker, G. S. (2007). Evolutionary ef fi ciency and happiness. Journal of Political Economy , 115 (2), 302 -337. https://doi.org/10.1086/ 516737
  12. Spence, M. (1974). Market signaling: Informational transfer in hiring and related screening processes . Harvard University Press.
  13. Trivers, R. (1985). Deceit and self-deception. In Social evolution (pp. 395 -420). Benjamin/Cummings.
  14. Trivers, R. L. (1972). Parental investment and sexual selection. In B. Campbell (Ed.), Sexual selection and the Descent of Man (pp. 136 -179). Aldine.
  15. Trivers, R. L. (1976). ' Foreword. ' In R. Dawkins (Ed.), The selfish gene (pp. 19 -20). Oxford University Press.
  16. von Hippel, W., & Trivers, R. (2011a). The evolution and psychology of selfdeception. Behavioral and Brain Sciences , 34 (1), 1 -16. https://doi.org/10 .1017/S0140525X10001354

Metadata

Authors
Dawson, C., & de Meza, D.
Venue
Psychological Review
Published
2026-07-20
ISSN
  • 0033-295X (ISSN-L)
  • 1939-1471
OpenAlex
W4385837823
Cited by
2
References
66
Citations by year
2026: 1
FWCI
11.22
Type
journal article
Language
en
Open access
Locations
  • Psychological Review (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • London School of Economics and Political Science Research Online (London School of Economics and Political Science) (submittedVersion)
  • OSF Preprints (OSF Preprints) (submittedVersion)
  • PsyArXiv (OSF Preprints) (submittedVersion, cc-by)
  • PsyArXiv (OSF Preprints) (submittedVersion)
Topics
  • Social Sciences › Decision Sciences › General Decision Sciences › Decision-Making and Behavioral Economics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Evolutionary Psychology and Human Behavior
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Experimental Behavioral Economics Studies
Keywords
overconfidence, loss aversion, self-deception, signaling theory, evolutionary psychology, decision-making, perceived competence, cognitive biases, Kahneman

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.