Moral licensing
Claim assessed: After a good deed, people allow themselves a less good one.
Published · how we work
In short
Moral licensing is the idea that after doing something good, you allow yourself something less good, as if the good deed had given you “permission”. The first studies reported the effect in many situations: hiring, donations, cheating. Later meta-analyses found a small effect, but one inflated by the fact that successful studies are published more easily. After correction, the average effect nearly disappears, and large replications either did not find it or found it much smaller.
What it says
The classic studies. Monin and Miller (2001) gave participants the chance to reject a few blatantly sexist statements. Afterwards, these participants more often said that a sales representative job at a cement company, with tough bargaining, would suit a man better. Mazar and Zhong (2010) reported that people who had had the chance to buy green products then shared less money in a game with another participant and stole more. We have not read these studies; we describe them from Kuper and Bott (2019) and Xiao et al. (2024).
The meta-analyses and the correction. The first meta-analyses found a small average effect, 0.31 across 91 studies (Blanken et al., 2015) and 0.32 across 106 studies. Kuper and Bott (2019) compared the two datasets and found coding errors, including an effect entered as 3.2 instead of 0.52. After fixing them, 76 independent results remained, with an average effect of 0.27. They then applied two methods that correct for unpublished studies. One brought the effect down to 0.18, the other to −0.05, practically zero. The authors say even the 0.18 estimate isn’t safe: it might still be too high.
The large replications. Kuper and Bott note that two large sets of studies, with 801 and 1,274 participants, did not find the effect. A project run across many labs, with 3,134 participants, found it, but at less than half the size estimated by the meta-analyses. Xiao et al. (2024) repeated Monin and Miller’s second study as a registered report: the analysis plan was reviewed and accepted before data collection. With 932 American participants recruited online, they did not find the effect. On the question about ethnicity, the effect was close to zero. On the one about gender it even went the other way: those who had previously hired a female candidate then preferred a man less often.
Only when you’re seen? Rotella et al. (2025) analysed 115 experiments, with 21,770 participants, following a plan registered in advance. After correcting for unpublished studies, no clear overall effect remained. The effect held up only in studies where participants knew others could see their choices, with a size of 0.51. In online studies, the effect was not different from zero even before correction. The authors propose that moral licensing is mainly about reputation: once you’ve shown others that you’re a good person, you allow yourself a small lapse. The paper is accepted for publication; we read the preprint.
Example
Someone donates to a fundraiser at the office, in front of colleagues, and the next day feels entitled to skip their share of the shared cleaning. If the current evidence is right, such a link is more likely when the good deed was seen by others. The example is an illustration, not a study cited here.
How to apply it
The steps below are a practical approach we propose, based on the studies cited.
- Don’t assume a good deed automatically leads to a bad one. On average, after correction, the effect is very small or zero.
- Don’t assume the opposite either. Sometimes people become more consistent after a good deed: in Xiao et al.’s replication, those who had hired a female candidate were then more balanced.
- Pay attention when your good deed is public. That’s where the effect has the most support, though from few studies.
- Ask for large, preregistered studies. When moral licensing is used as an argument, check what data it rests on.
Limits and nuances
- The corrections don’t agree. Kuper and Bott get 0.18 with one method and −0.05 with the other, and say they cannot state with certainty that the effect is zero.
- The observation result rests on little data. There are only 11 results, and the original studies weren’t designed to test this. The meta-analysis compares different studies, so it shows an association, not a cause.
- Xiao et al.’s replication was run online; the original was run in the lab. If the effect depends on being observed, that could explain part of the difference.
- Culture might matter. Kuper and Bott found larger effects in North American studies than in European ones, but the evidence is limited.
- What we read. We read Kuper and Bott (2019), from Meta-Psychology, Xiao et al.’s (2024) replication and Rotella et al.’s (2025) preprint in full, all open access. We have not read Monin and Miller (2001), Mazar and Zhong (2010), Blanken et al. (2015) or Ebersole et al. (2016); we describe them from the three papers we read.
Sources
- Benoît Monin, Dale T. Miller (2001). Moral credentials and the expression of prejudice
- Nina Mazar, Chen-Bo Zhong (2010). Do green products make us better people?
- Irene Blanken, Niels van de Ven, Marcel Zeelenberg (2015). A meta-analytic review of moral licensing
- Charles R. Ebersole et al. (2016). Many Labs 3: Evaluating participant pool quality across the academic semester via replication
- Niclas Kuper, Antonia Bott (2019). Has the evidence for moral licensing been inflated by publication bias?
- Qinyu Xiao, Lok Ching Li, Ying Lam Au, See Ngueh Tan, Wing Tung Chung, Gilad Feldman (2024). Licensing via credentials: Replication Registered Report of Monin and Miller (2001)
- Amanda Rotella, Jisoo Jung, Christopher Chinn, Pat Barclay (2025). Observation moderates the moral licensing effect: A meta-analytic test of interpersonal and intrapsychic mechanisms (preprint)
See also: Ego depletion, Power posing