Guala appraises the hypothesis of strong reciprocity in the light of anthropological evidence from hunter-gatherer societies. Much of his discussion focuses on a particularly distinctive implication of this hypothesis, namely, the existence of altruistic punishment. An individual A engages in altruistic punishment when she incurs costs to punish some other individual B for an action by B which is contrary to social norms but not specifically directed at A.
The discovery of the mechanism of altruistic punishment is an achievement of behavioural and experimental economics. As Guala documents, there is now an influential literature arguing that altruistic punishment plays a fundamental role in stabilising cooperation in hunter-gatherer societies. If that claim were true, it would suggest the hypothesis that modern humans have hard-wired preferences for altruistic punishment; and were that hypothesis confirmed too, the methods of behavioural economics would have led to a major discovery in the domain of natural science. But is the claim justified?
Guala's review of the anthropological evidence suggests that it is not. I cannot claim expertise in anthropology, but I find the review convincing and consistent with my reading of a collection of research papers selected by leading advocates of strong reciprocity (Gintis et al. Reference Gintis, Bowles, Boyd and Fehr2005; Sugden Reference Sugden, Gintis, Bowles, Boyd and Fehr2007). It seems that economic cooperation in hunter-gatherer societies is mainly among kin and between directly reciprocating partners. Individual acts of “punishment” (perhaps better described as revenge) are usually in response to harmful behaviour directed at the revenge-taker, as in many sexual conflicts, and tend to be discouraged by third parties who fear the socially destructive effects of cycles of revenge. If one person violates a norm without directing the harm at particular others, punishment tends to be a collective act, structured so that the cost to any individual is very low. It is well known that gossiping and ridicule are common punishments in hunter-gatherer societies. Guala is particularly convincing in suggesting that these practices allow coalitions of punishers to form without there being any obvious ringleader, and hence with minimal cost to individuals.
In the rest of this commentary I accept Guala's assessment of the anthropological evidence, and focus on the apparent conflict between this evidence and experimental observations of altruistic punishment. Is this conflict symptomatic of flaws in the methodology of experimental economics?
It is useful to consider how the hypothesis of altruistic punishment originated. From at least the 1960s, economists recognised that public goods are sometimes supplied through voluntary contributions. Some explanations of this fact postulated that donors were motivated by individual incentives (e.g., Olson Reference Olson1965). Others postulated non-selfish motivations – usually altruism, but positive reciprocity was also proposed (Sugden Reference Sugden1984). With the development of experimental economics, the explanation of voluntary contributions to public goods became a prominent research programme. In the experimental design most commonly used in this programme, subjects interact anonymously in a game with real monetary payoffs. Each subject has an incentive to free-ride, but all subjects gain if all contribute. Face-to-face social pressures are screened out as far as possible, with the aim of isolating and investigating non-selfish motivations (or “social preferences”).
This line of research has led to three main conclusions. First, a significant proportion of individuals make positive contributions, contrary to the assumption of self-interest. Second, individuals' contributions tend to be positively correlated with one another, as implied by positive reciprocity. Third, as the game is repeated, the rate of contribution decays. The best explanation of these findings seems to be that they result from interaction between two types of individuals in the population – free-riders and positive reciprocators. The positive reciprocators gradually withdraw from cooperation as they find insufficient reciprocation from others (e.g., Bardsley & Moffatt Reference Bardsley and Moffatt2007).
Fehr and Gächter (Reference Fehr and Gächter2000a) initiated a new line of research by establishing that the tendency for contributions to decay can be overturned if, after each round of the public good game, each subject has the opportunity to impose costly punishments on individual others. For this mechanism to work, there must be some subjects with an “altruistic” preference for punishing free-riders, but Fehr and Gächter showed theoretically that there can be high and stable rates of contributions even if the proportion of such individuals is quite small and their preference for punishment is quite weak. There is now a large body of experimental evidence showing that, if the cost of punishing is low relative to the cost of being punished and if those being punished do not have the option to retaliate, high rates of contributions are sustained. The implication is that, if placed in this experimental environment, a sufficient proportion of individuals reveal a sufficiently strong preference for altruistic punishment for cooperation to be stabilised.
I agree with Guala that there has been real scientific progress here. The mechanism by which altruistic punishment can support cooperation is a genuine discovery which grew out of a programme of sound experimental research. But I agree too that there are many reasons for caution about extrapolating from these experiments to cooperation problems in real life. In particular, the anonymised experimental environment filters out mechanisms by which face-to-face contact might inhibit punishment; instead, it channels negative affective responses into punishment.
Understood as a brave conjecture inspired by experimental research, the hypothesis that hunter-gatherer cooperation relies on altruistic punishment is scientifically legitimate. But the hypothesis must be tested against real-world observations; and if it fails, it fails. Good science does not always succeed.
Guala appraises the hypothesis of strong reciprocity in the light of anthropological evidence from hunter-gatherer societies. Much of his discussion focuses on a particularly distinctive implication of this hypothesis, namely, the existence of altruistic punishment. An individual A engages in altruistic punishment when she incurs costs to punish some other individual B for an action by B which is contrary to social norms but not specifically directed at A.
The discovery of the mechanism of altruistic punishment is an achievement of behavioural and experimental economics. As Guala documents, there is now an influential literature arguing that altruistic punishment plays a fundamental role in stabilising cooperation in hunter-gatherer societies. If that claim were true, it would suggest the hypothesis that modern humans have hard-wired preferences for altruistic punishment; and were that hypothesis confirmed too, the methods of behavioural economics would have led to a major discovery in the domain of natural science. But is the claim justified?
Guala's review of the anthropological evidence suggests that it is not. I cannot claim expertise in anthropology, but I find the review convincing and consistent with my reading of a collection of research papers selected by leading advocates of strong reciprocity (Gintis et al. Reference Gintis, Bowles, Boyd and Fehr2005; Sugden Reference Sugden, Gintis, Bowles, Boyd and Fehr2007). It seems that economic cooperation in hunter-gatherer societies is mainly among kin and between directly reciprocating partners. Individual acts of “punishment” (perhaps better described as revenge) are usually in response to harmful behaviour directed at the revenge-taker, as in many sexual conflicts, and tend to be discouraged by third parties who fear the socially destructive effects of cycles of revenge. If one person violates a norm without directing the harm at particular others, punishment tends to be a collective act, structured so that the cost to any individual is very low. It is well known that gossiping and ridicule are common punishments in hunter-gatherer societies. Guala is particularly convincing in suggesting that these practices allow coalitions of punishers to form without there being any obvious ringleader, and hence with minimal cost to individuals.
In the rest of this commentary I accept Guala's assessment of the anthropological evidence, and focus on the apparent conflict between this evidence and experimental observations of altruistic punishment. Is this conflict symptomatic of flaws in the methodology of experimental economics?
It is useful to consider how the hypothesis of altruistic punishment originated. From at least the 1960s, economists recognised that public goods are sometimes supplied through voluntary contributions. Some explanations of this fact postulated that donors were motivated by individual incentives (e.g., Olson Reference Olson1965). Others postulated non-selfish motivations – usually altruism, but positive reciprocity was also proposed (Sugden Reference Sugden1984). With the development of experimental economics, the explanation of voluntary contributions to public goods became a prominent research programme. In the experimental design most commonly used in this programme, subjects interact anonymously in a game with real monetary payoffs. Each subject has an incentive to free-ride, but all subjects gain if all contribute. Face-to-face social pressures are screened out as far as possible, with the aim of isolating and investigating non-selfish motivations (or “social preferences”).
This line of research has led to three main conclusions. First, a significant proportion of individuals make positive contributions, contrary to the assumption of self-interest. Second, individuals' contributions tend to be positively correlated with one another, as implied by positive reciprocity. Third, as the game is repeated, the rate of contribution decays. The best explanation of these findings seems to be that they result from interaction between two types of individuals in the population – free-riders and positive reciprocators. The positive reciprocators gradually withdraw from cooperation as they find insufficient reciprocation from others (e.g., Bardsley & Moffatt Reference Bardsley and Moffatt2007).
Fehr and Gächter (Reference Fehr and Gächter2000a) initiated a new line of research by establishing that the tendency for contributions to decay can be overturned if, after each round of the public good game, each subject has the opportunity to impose costly punishments on individual others. For this mechanism to work, there must be some subjects with an “altruistic” preference for punishing free-riders, but Fehr and Gächter showed theoretically that there can be high and stable rates of contributions even if the proportion of such individuals is quite small and their preference for punishment is quite weak. There is now a large body of experimental evidence showing that, if the cost of punishing is low relative to the cost of being punished and if those being punished do not have the option to retaliate, high rates of contributions are sustained. The implication is that, if placed in this experimental environment, a sufficient proportion of individuals reveal a sufficiently strong preference for altruistic punishment for cooperation to be stabilised.
I agree with Guala that there has been real scientific progress here. The mechanism by which altruistic punishment can support cooperation is a genuine discovery which grew out of a programme of sound experimental research. But I agree too that there are many reasons for caution about extrapolating from these experiments to cooperation problems in real life. In particular, the anonymised experimental environment filters out mechanisms by which face-to-face contact might inhibit punishment; instead, it channels negative affective responses into punishment.
Understood as a brave conjecture inspired by experimental research, the hypothesis that hunter-gatherer cooperation relies on altruistic punishment is scientifically legitimate. But the hypothesis must be tested against real-world observations; and if it fails, it fails. Good science does not always succeed.