# When good norms turn bad

(This post is about a paper I recently published, with co-authors, in Nature Communications. If you want to cut to the chase, here it is.)

## What is a bad norm?

Is giving gifts at Christmas time a good social norm? That might seem a strange question to ask. Social norms are widely credited with the promotion of cooperation and social efficiency. We might think that some societies have better social norms than others. We might think that some social norms are bad for some individuals, but they should at least be good for the group as a whole. Why would a society embrace a social norm that is downright “bad”?

But now consider some candidate bad norms: gender norms that deprive women of basic liberties; rites that require inflicting painful bodily mutilation on one’s children, such as footbinding, circumcision, tattooing, and female genital mutilation; or consider cultures of binge drinking among college students. All of these behaviours appear to be sustained by social norms; and it is plausible that they are not only bad for the individuals involved, but are bad for the group as a whole. There are some conceivable benefits, but they don’t seem to justify the costs. We certainly do not see other groups envying these practices and wishing that they could institute similar norms.

Of course, the material waste associated with gift giving is just one side of the story. What about the intangible benefits? Many of us retain sentimental attachments to gifts that we receive, and this is a type of value that Waldfogel’s estimates neglect. Maybe relationships are strengthened and sustained by gift giving?

There is no doubt a lot to this. But if we were to compare societies that do and do not have holidays that are associated with customary gift-giving, would we see any robust difference in the quality of relationships across types of society? Are the lives of those who do not receive Christmas gifts really poorer in some non-material sense? These hypotheses strain my credulity, but I suppose the question remains an open one.[1]

Suppose you grant my hypothesis, that a norm like Christmas gift-giving is bad for those societies that have it. What sustains it? Why don’t we move to a less wasteful practice? In a new paper – my first empirical paper, in fact – we report a simple laboratory experiment that vividly demonstrates a process that is probably implicated in sustaining norms of this sort.

## Bad norms in the lab

Subjects are put into groups of four. They cannot directly communicate with other group members, and they interact via computer terminals, so they won’t know who, in the lab of 20 or more subjects, they are grouped with. They play a simple economic game for 20 rounds with this group.

The game is a variant on what economists call a public goods game: Each round, subjects have one dollar that they can contribute to a common pool. They can contribute any amount they like from zero to the whole dollar. All the contributions are made simultaneously, pooled together, and then reduced by 20%. The resulting pool is then shared out equally to everyone in the group, whether or not they made a contribution.[2] Notice that this is much like the dynamics of giving Christmas gifts: the act of donating is generous – it makes the other group members better off, but it is a wasteful way of doing so. It does not benefit them as much as it costs the donor to give the gift.

Each contribution of $1 is effectively a gift of 20c to each other group member, and wastes 20c. In our experiment, the maximum endowment that could be spend each round was 80c, rather than a dollar, but the economic logic is the same. Our experiment had two treatments. In one treatment, players simply played this game as described, and were then paid in accordance with their earnings from the game. Very few subjects contributed on the first round (average per capita contribution was roughly 15% of the maximum possible), and soon thereafter – by round five – contributions effectively went to zero. This looks sensible: the custom of contributing is bad for the group, and it dies out. But in the other treatment, we provided subjects with the opportunity to punish each other after each round. They punish by paying to reduce the earnings of someone else. Each 10c spend on punishment would reduce the earnings of the punished person by 30c. The result: right out of the box, contributions in this treatment were a bit higher in the first round. But more impressively, contributions remained significantly higher over the 20 rounds. By round 20, the average per capita contribution was around 15% of the endowment. The result, however, was strikingly heterogeneous across groups. Roughly one third of groups sustained much higher contributions – around half of the maximum, while the rest of the groups went to zero. In reality, we cannot “punish” each other in such a clinical fashion as occurred in our experiment. But we can express our disapproval by dirty looks, by gossip, and all manner of other informal sanctions. Most people are strongly motivated to avoid social disapproval. So even if we know that a practice is wasteful, we might be happy to participate if it spares us the disapproval of our peers. To be really sure that what was occurring in our first experiment was mediated by normative attitudes, we then ran a second experiment, where we asked a fresh batch of subjects what they thought was the appropriate thing to do in a setting such as this. Using a clever technique that has been developed by other economists to elicit social norms, we asked “how appropriate is it to contribute x units?”, but we paid a bonus if subjects gave the answer to this question that was most common among all subjects. This incentivizes subjects not to report their idiosyncratic normative attitudes, but to try to work out what do most people think most other people think? That’s exactly what a social norm is: a widely shared belief about what other people believe. The result was overwhelming: contributing zero was inappropriate, and contributing positive amounts – especially amounts from half to all of the endowment – was regarded as appropriate. (What exactly was the norm involved? We couldn’t determine that from this study. My suspicion is that it was a notion of fairness. Imagine one participant contributes on the first round, because they mistakenly think it is good for the group, or because they think it is just a nice thing to do, even if it is bad for the group payoff. But then they observe that fellow group members have not contributed. They think: “Hey, you benefited at my expense, but you didn’t make a similar contribution to benefit me. That’s not fair!”. And so they punish the non-contributors, and the norm is on its way to being established. If this sort of story is right, then it is arguable that this study does not demonstrate an intrinsically bad norm is at work. It might be better described as a case where a normally good norm – fairness – is being applied with insufficient sensitivity to the context. Yes, there is something unfair about the fact that some have benefited at another’s expense. But it would be foolish – bad for both group and individuals – to insist on fairness at the expense of efficiency in a scenario like this.) So the story here is in one way quite unsurprising: people conform to social customs – even stupid customs – because they want the approval of their peers. Even if most of your peers know that the custom is bad for the group, there may be enough people who would disapprove of an individual deviation that compliance remains very high. OK, we probably knew that already. But what is still surprising is that we saw this effect, even in an artificial environment that lacked all of the social context that we might normally think is necessary to sustain a costly social norm. Once students walked out the lab, they would never be held accountable for their actions. We quizzed the subjects beforehand to check that they understood the implications of contributing, so they should all have had a reasonable understanding of the effect on the group. But still, some people felt so strongly that contributing was desirable that they were willing to take on costs to inflict their normative views on others. Norm psychology is powerful stuff! 1. It is also possible that there is no difference across societies, but that there is a confound. Perhaps even in societies that lack a holiday associated with customary gift-giving, there is still a comparable level of gift-giving overall – it just occurs in a less concentrated fashion, spread throughout the calendar. I should look further into this, but my understanding is that at least some societies differ from my own in having holidays that are associated with gifts of cash. And the benefit of cash is that there is none of the inefficiency of purchasing less fungible gifts. To receive a gift of$N in cash, everyone should be willing to pay up to \$N, obviously! So the value to the recipient at least matches the cost to the donor. This sort of difference, if it exists, should provide the necessary variation to put the hypothesis to the test.  ↩

2. In normal public goods games, the pool is multiplied by a factor greater than one, but less than the number of players. That means that the group is made better off by contributions, but the individual who contributes is made worse off. In our case, the group is made worse off too.  ↩