Cited by 4,547Experimental Behavioral Economics StudiesEvolutionary Game Theory and CooperationGame Theory and Voting Systems

The DOI matches its Crossref record; 73 of the 73 references it cites were found in a registry or an index; none of the 31 sources its claims cite is checked for support yet.

Cooperation and Punishment in Public Goods Experiments

Open access
Fehr, Ernst, Gächter, Simon·2000·American Economic Review

Abstract

Casual evidence as well as daily experience suggest that many people have a strong aversion against being the 'sucker' in social dilemma situations. As a consequence, those who cooperate may be willing to punish free-riding, even if this is costly for them and even if they cannot expect future benefits from their punishment activities. A main purpose of this paper is to show experimentally that there is indeed a widespread willingness of the cooperators to punish the free-riders. Our results indicate that this holds true even if punishment is costly and does not provide any material benefits for the punisher. In addition, we provide evidence that free-riders are punished the more heavily the more they deviate from the cooperation levels of the cooperators. Potential free-riders, therefore, can avoid or at least reduce punishment by increasing their cooperation levels. This, in turn, suggests that in the presence of punishment opportunities there will be less free riding. Testing this conjecture is the other major aim of our paper.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

Fehr and Gächter show experimentally that costly punishment of free riders is widespread and that the mere opportunity to punish dramatically raises cooperation in public goods games. The pattern of punishment and emotional responses supports the hypothesis that negative emotions make punishment threats credible.

Key findings

  • In the no-punishment condition, contributions collapse: in the final period, 75% of subjects in the Stranger-treatment and 53% in the Partner-treatment free ride completely. With punishment opportunities, average contributions are maintained at 50–95% of the endowment, and in the Partner-treatment final period 82.5% of subjects contribute the whole endowment.
  • Punishment targets free riding: the more a subject’s contribution falls below the others’ average, the more punishment points they receive. OLS regressions show absolute negative deviation is positive and highly significant (Stranger: 0.2428*; Partner: 0.4168*), while positive deviation has no significant effect.
  • Sanctions change behavior: in the Partner-treatment, 89% of punished below-endowment contributors increased their contribution next period by an average of 4.6 tokens; in the Stranger-treatment, 78% increased by an average of 3.8 tokens.
  • Payoff consequences: free riders’ overall payoff falls by 24% in the Stranger-treatment and 16% in the Partner-treatment relative to no punishment. By the final period, the punishment condition yields a relative payoff gain of roughly 20% in the Partner-treatment and 10% in the Stranger-treatment.

Methods

  • Computerized laboratory public goods experiment with 112 subjects across five sessions in January–February 1996; subjects were students from many fields except economics. Groups of n = 4, endowment y = 20 tokens, marginal public-good payoff a = 0.4, ten periods, N = 24 per session.
  • Four treatment conditions: Stranger vs. Partner group composition, each with and without punishment opportunities. In punishment conditions, subjects could assign up to 10 punishment points to each other after seeing individual contributions; each point reduced the target’s first-stage payoff by 10%, up to 100%, and imposed increasing costs on the punisher.
  • Analysis used nonparametric tests (Wilcoxon matched pairs, Mann-Whitney) and OLS/Tobit regressions of received punishment points on others’ average contribution, absolute negative deviation, positive deviation, and period/session/group controls. A separate emotion questionnaire (N = 33, non-experiment participants) measured negative feelings toward free riders on a 7-point scale.

Limitations

  • The paper does not include a dedicated limitations section. It states that the Stranger-treatment’s probability of future rematching is very low but not exactly zero, and addresses this with a Perfect Stranger robustness treatment where rematching probability is zero; results are similar.
  • The text makes evident that the experiment uses student subjects, neutral framing, specific parameters, and a finite ten-period horizon, so external validity to real-world settings is not directly established. The paper itself notes that real-world free riding is often described in value-laden terms, suggesting emotion-based punishment may be even more important outside the lab.
  • One Partner-session had only 16 subjects instead of 24, yielding 4 groups rather than the planned 6. The emotion questionnaire used a separate sample of 33 subjects who did not participate in the main experiment.

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

Integrity signals

Registry

Crossref · 10.1257/aer.90.4.980

Cites

73 of 73 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 of 31 cited sources checked · 0 supported · 0 flagged

Show all 155 claims

  • It may state two separate claims as one. 37
  • Probably not a substantial claim. 29
  • The quote only partly states this claim. 1
59 of 59
  • The authors state this as an assumption.

    All subjects are rational money maximizers, and this is common knowledge.

    “If all subjects are rational money maximizers, and if this is common knowledge, the subgame perfect equilibrium prediction with”
  • The paper relies on this without support.

    Subjects strongly dislike being the 'sucker' (cooperating while other group members free ride).

    “We believe, in particular, that subjects strongly dislike being the 'sucker', i.e., being those who cooperate while other group members free ride.”
  • The authors state this as an assumption.

    There is a mix of selfish, conditionally cooperative and punishing subjects in our experiments.

    “In view of the evidence mentioned above it is reasonable to assume that there is a mix of selfish, conditionally cooperative and punishing subjects in our experiments.”
  • The paper relies on this without support.

    According to the standard approach, full defection is the unique subgame perfect equilibrium in the Partner-treatment without punishment.

    “full defection is the unique subgame perfect equilibrium.”
  • The paper relies on this without support.

    A frequent or modal individual contribution in the final period indicates the behavioral standard or focal action for that condition.

    “Thus, full free riding clearly emerges as the behavioral regularity in this condition.”
  • The paper relies on this without support.

    Negative emotions are more intense the more a free rider deviates from the others' average contribution.

    “because negative emotions are the more intense the more the free rider deviates from the others' average contribution.”
6 of 6

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Perfect Stranger subjects could not meet every other subject in six periods

    Minor

    With 24 subjects in groups of 4, each subject meets only 3 others per period. Six periods therefore allow meetings with at most 18 distinct people, not every one of the other 23 subjects as claimed.

  • Partner-treatment regression reports too many observations

    Minor

    The Partner-treatment has 10 groups of 4 subjects observed for 10 periods, giving 400 subject-period observations, not the reported 440. The regression's residual degrees of freedom, 378, also match 400 observations with 21 regressors and an intercept.

    Reported 440 · recomputed 400

2 of 2

Citations

Cites

Flagged first, then by citations

Show all 40 references

Cited by

Sorted by citations

Items 701–800 of 4,322

Metadata

Authors
Fehr, E., & Gächter, S.
Affiliations
University of Zurich (CH)
Venue
American Economic Review
Published
2000-09-01
Volume
90
Issue
4
Pages
980-994
ISSN
0002-8282 (ISSN-L)
OpenAlex
W3122867100
Cited by
4,547
References
40
Citations by year
2026: 195 · 2025: 235 · 2024: 225 · 2023: 204 · 2022: 176 · 2021: 217 · 2020: 240 · 2019: 228 · 2018: 202 · 2017: 203 · 2016: 190 · 2015: 215 · 2014: 223 · 2013: 224 · 2012: 217
FWCI
145.58
Type
journal article
Language
en
Open access
Locations
  • American Economic Review (publishedVersion)
  • SSRN Electronic Journal (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • Alexandria (UniSG) (University of St.Gallen) (submittedVersion)
  • Zurich Open Repository and Archive (University of Zurich) (submittedVersion, other-oa)
Funders
  • National Science Foundation
  • John D. and Catherine T. MacArthur Foundation
  • Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Experimental Behavioral Economics Studies
  • Social Sciences › Sociology and Political Science › Evolutionary Game Theory and Cooperation
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Game Theory and Voting Systems
Keywords
costly punishment, free-riding, public goods, cooperation, credible threats, emotions, negative emotions, group norms

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.

A language model drew the claims from the paper on 2026-11-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

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