Moral licensing worked in 2001. In 2024 it ran backwards.

One good deed was meant to license the next bad one. A meta-analysis of 91 studies put it at d = 0.31. A 2024 replication with 932 people got the reverse.

In 2001 Benoît Monin and Dale Miller gave students a hiring problem. Some of them first got a chance to endorse a candidate from a group that rarely gets picked. Then everyone was asked about a second job — one described in terms that made a white man the obvious fit.

The students who had already made the fair choice were more willing to make the unfair one. Having proved something about themselves, they spent it. Monin and Miller called the currency moral credentials, and the paper landed with a phrase that outlived the data: doing good buys you permission to do bad.

The idea spread fast, because everyone recognises it

You have felt it. The gym session that justifies the dessert. The donation in December that makes January's ignored fundraiser feel already handled. Psychology gave it a name and a mechanism — a moral ledger you keep on yourself, running a balance you top up and draw down.

Sachdeva, Iliev and Medin published the sharpest version in 2009 under a title nobody forgets, "Sinning saints and saintly sinners." Write about yourself using positive words, and you donate less to charity afterwards. Write about yourself using negative words, and you donate more. The ledger appeared to work in both directions.

Then in 2010 Nina Mazar and Chen-Bo Zhong asked whether green products make us better people, and reported that they do the opposite. Shopping in a green online store made people more likely to cheat and steal in a later task. That one travelled furthest of all. It is the study behind every column arguing that recycling makes you smug and worse.

Someone added the whole literature up

In 2015 Irene Blanken, Niels van de Ven and Marcel Zeelenberg pooled 91 studies covering 7,397 participants. The licensing effect was there: Cohen's d of 0.31. Small, but real, and the field had a number at last.

The number underneath it was more interesting. Published studies showed d = 0.43. Unpublished ones showed less. That gap is the fingerprint of publication bias: the experiments that found the effect got written up, and the average everyone had been quoting was assembled out of the survivors.

The same team ran three fresh replications of their own. They found no evidence for the effect. Their conclusion was not that licensing is fake; it was that the effect is small enough that almost every study ever run on it had too few people to say anything reliable.

The green-shopping result did not hold

Jan Urban, Štěpán Bahník and colleagues went after Mazar and Zhong directly. Two conceptual replications and one close replication, 1,274 people in total. Green consumption did not make anyone cheat. They wrote that green consumption is unlikely to trigger cross-domain moral licensing at all.

The interesting part is that nothing appeared in the other direction either. Green shoppers were not more honest than anyone else. The result was flat.

And in 2024 the original went the other way

A registered-report replication of Monin and Miller ran 932 participants through the same design, with the analysis plan locked before the data came in. Working back from the 2001 paper, the replication team put the original effect at d = 0.44, with a confidence interval running from 0.09 to 0.79 — wide, and only just clear of zero at the bottom end.

In the ethnicity version, the replication got d = 0.08, with a confidence interval sitting comfortably across zero. Nothing. In the gender version it got d = −0.38, and that interval does not include zero. The effect ran backwards. People who had just made a fair choice were less likely to make an unfair one afterwards, not more.

The authors offered two readings and refused to choose between them. Either the 2001 findings were false positives, or the manipulation has lost its power in the twenty-three years since — a scenario in which endorsing a minority candidate no longer feels like proof of anything, because the world moved.

So what is left standing

Not nothing. A meta-analysis of 91 studies does not evaporate because three replications failed, and d = 0.31 corrected downward for publication bias is still a number above zero. In narrow settings, measured immediately, with the good deed fresh, something happens.

What has not survived is the version people actually believe. The idea that a good deed reliably licenses a bad one, across different domains, in a way you can plan around — that is the claim the last decade has been taking apart. Green shopping does not make you cheat. Hiring fairly does not make you discriminate. In one careful test it made people do the opposite.

Several competing accounts have been in the literature the whole time — moral credits, credentialing, self-signalling — and one of them predicts the opposite outcome entirely. Consistency: having done a thing, you become the kind of person who does that thing, and you do it again. The 2024 gender result points that way. One result is not a theory, and I would not build anything on it yet.

Why this matters if you are trying to be useful

Mostly because of what it stops you doing.

The licensing story is a very comfortable excuse. It says the impulse to help is a tank that empties, and that stopping for a while is what the science predicts you will do. If you have ever caught yourself thinking you have done your bit this month, you have used it.

That permission slip turns out to rest on much less than its popularity suggests, which leaves you where you started and without the alibi. Nothing you did last week has decided anything about this week. The ledger you thought you were keeping was probably never there.

Sources

Monin, B. & Miller, D. T. (2001), "Moral credentials and the expression of prejudice," Journal of Personality and Social Psychology 81, 33–43. Sachdeva, S., Iliev, R. & Medin, D. L. (2009), "Sinning saints and saintly sinners," Psychological Science. Mazar, N. & Zhong, C.-B. (2010), "Do green products make us better people?", Psychological Science 21, 494–498. Blanken, I., van de Ven, N. & Zeelenberg, M. (2015), "A meta-analytic review of moral licensing," Personality and Social Psychology Bulletin. Urban, J., Bahník, Š. et al. (2019), "Green consumption does not make people cheat," Journal of Environmental Psychology. Registered-report replication of Monin & Miller (2024), International Review of Social Psychology.