Published online 28 September 2009 | Nature | doi:10.1038/news.2009.955

Column: Muse

When does punishment work?

The outcome of games designed to tell us how we behave may depend on who 'we' are. Economists and evolutionary biologists should take note, says Philip Ball.

Doing business abroad is not like doing it at home. A tough, no-nonsense approach will get results in New York but may be seen as rude and aggressive in Tokyo, where business negotiations are a mixture of ceremony and courtship designed to avoid direct confrontation. In Japan, an apparent 'yes' can mean 'no', and there are sixteen ways of avoiding having to say 'no' directly1. So negotiations there are long-winded, ambiguous and, to outsiders, plain baffling.

Traditional economics makes no concessions to such cultural differences. Its models tend to adopt a 'one size fits all' view of human interactions, in which choices and strategies are based on the cold logic of 'utility maximization': whatever gets you the best deal. But behavioural economics, which sets out to investigate how people make their decisions, is undermining that picture. Now a paper in the Proceedings of the National Academy of Sciences reports that Chinese people respond quite differently from Americans in transactions that contain the threat of punishment for uncooperative behaviour2.

Man in prison cellThe prisoner's dilemma plays out differently in China.Jupiter images / Getty

Yi Tao of the Chinese Academy of Sciences' Institute of Zoology in Beijing and his co-workers have looked at how Chinese university students play a version of the prisoner's dilemma. This is the classic model system in game theory for studying two-person interactions that include a temptation to cheat. The game awards points according to whether the two players decide to cooperate or 'defect'. Two defections — when the players both try to cheat each other — elicit a low pay-off, whereas both players are better rewarded if they cooperate. However, if player 1 cooperates and player 2 defects, then player 1 is the sucker, getting the worst possible pay-off, while player 2 does best of all.

The 'logical' way to play this game is always to defect, because that gives you the best pay-off whatever your opponent does. Yet if the game is played repeatedly, players realize that they can both score more highly if they agree (tacitly) to cooperate. Thus cooperation can arise from self-interest.

Bad behaviour

The prisoner's dilemma — which was devised during the cold war in a US think tank but was anticipated in the political philosophies of Thomas Hobbes and Jean-Jacques Rousseau — has been used to show how altruism can develop in animal communities, including human societies. This kind of game theory is also central to economic analysis of competitive markets.

Generally, the players need to encounter one another repeatedly before the long-term benefits of mutual cooperation become apparent. That's why, the theory suggests, we cultivate good relations with our neighbours. But in 2002, behavioural economists Ernst Fehr and Simon Gächter in Switzerland showed that cooperation may also emerge in lone encounters if they happen in a context where defection can be punished3. In those experiments, groups of players were given a sum of money to invest in a project, and the group was jointly rewarded according to the level of investment. Freeloaders can enjoy the group reward without investing. But if such selfish behaviour can be punished with fines, it is suppressed. Punishment is typically used insistently: players will do so even at a cost to themselves.

Gächter and his co-workers have investigated whether attitudes to punishment in this 'public goods' game differ across cultures4. They compared the amount of expenditure on punishment for players in 15 countries, including the United States, China, Turkey and Saudi Arabia. They found that the cooperation-enhancing effect of punishment, and cooperation overall, tended to be strongest in democratic countries that had a long tradition of market economies. One exception to this trend, however, was that the behaviour of players in the United States and China were similar.

Tao and his co-workers now show this doesn't mean American and Chinese people regard or employ punishment in the same way. The participants in their study played the straightforward iterated version of the game, but with the added provision that each player could opt to punish rather than to cooperate or defect. Punishment deducts several points from the other player, but also costs the punisher. Ill-gotten gains from defection can be cancelled by being punished in the next round.

This might be expected to promote cooperation — and when other researchers tested the game on US students last year, that's what they found5. But in China, punishment made very little difference to the amount of cooperation: it was, if anything, slightly lower than in control games without that option.

Cultural complications

Tao's team says this probably reflects the fact that, in China, individuals conduct transactions by cultivating so-called guanxi (literally 'closed system') networks of two-person relationships based on empathy and mutual understanding. That might sound chummy, but in fact guanxi is a delicate dance in which feelings of friendship, obligation and guilt are patiently probed and manipulated to reach the desired goal. As the rules of prisoner's dilemma-with-punishment allow for nothing of this sort, there's nothing to link cooperation to punishment.


All the same, reputations are important to guanxi networks, so in the public-goods game4 where reputation (for trustworthiness, say) may come into play, punishment is restored as an operative force — making the US and Chinese behaviours more similar, although for different reasons.

One implication is that it's dangerous to extrapolate from lab tests of behavioural economics to evolutionary questions — for example, to 'explain' the adaptive role of punishment in human society. Economic behaviour too can evidently depend on the 'culture' in which it happens. For example, economists are divided on the issue of how incentives affect productivity, because real-world experience sometimes seems at odds with what behavioural tests imply. But that's no longer surprising if behaviour varies according to the system of norms within which it is enacted. It's a reminder that a search for economic and behavioural 'first principles' may be doomed to fail. 

  • References

    1. Hendon, D. W., Hendon, R. A. & Herbig, P. Cross-Cultural Business Negotiations (Praeger Publishers, 1996).
    2. Wu, J.-J. et al. Proc. Natl Acad. Sci. USA advance online publication doi:10.1073/pnas.0905918106 (2009).
    3. Fehr, E. & Gächter, S. Nature 415, 137-140 (2002). | Article | PubMed | ISI | ChemPort |
    4. Herrmann, B., Thöni, C. & Gächter, S. Science 319, 1362-1367 (2008). | Article | PubMed | ChemPort |
    5. Dreber, A., Rand, D. G., Fudenberg, D. & Nowak, M. A. Nature 452, 348-351 (2008). | Article | PubMed | ChemPort |
Commenting is now closed.