The New Palgrave Dictionary of Economics

Living Edition
| Editors: Matias Vernengo, Esteban Perez Caldentey, Barkley J. Rosser Jr

Group Selection

  • Arthur J. Robson
Living reference work entry


‘Group selection’ is the biological term for the possibility that a characteristic that is beneficial to the group but possibly costly to the individual is evolutionarily successful. Although logically possible, it is generally viewed with scepticism by biologists. It is not problematic that group selection would favour an equilibrium whose payoffs dominated those of another, because there is then no conflict with individual selection within each group. Group selection might reject inefficient equilibria in a repeated game, for example. Since human societies can support rather arbitrary outcomes as equilibria, group selection could play a role in human evolution.


Altruism Assortative matching Cooperation and its evolution Cultural transmission Darwin, C. Evolution Exchange Group selection Hunter–gatherer economies Individual selection Kin selection Natural selection Other-regarding preferences Prisoner’s dilemma Punishment Repeated games Trust 

JEL Classifications

C71 C73 

It is uncanny how close Darwin (1859) came to the modern view of biological evolution, given that detailed understanding of the mechanics of genetic inheritance lay in the future. Darwin understood how mutations might arise randomly rather than in response to circumstances. Further, he espoused the modern dogma concerning the separation of the germ line (sex cells) and the somatic line (all other bodily cells). Under this dogma, which explicitly contradicts the earlier biologist Lamarck, only those characteristics present in the germ line, not those acquired during the individual’s lifetime, are genetically inherited by offspring. If a germ line characteristic produces a somatic or behavioural attribute that is best suited to the ecological or social circumstances of the individual (yielding the most offspring), then the attribute and the underlying genetic characteristic will become more common.

Darwin typically emphasized that a particular variation would spread if this variation led to greater reproductive success for individuals and were inherited by their descendants. However, Darwin also strayed occasionally into what would now be called ‘group selection’, especially when he was considering the implications of his theory for humans (Darwin 1871, p. 166). Thus, he thought an individual human might engage in behaviour that is beneficial to the survival of a group, even if this behaviour had a fitness cost to the individual. It is of interest that Darwin refers to humans in particular, since it is still sometimes argued that group selection might be important for our own species.

The Group Selection Debate in Biology

There is a ‘folk wisdom’ appeal to group selection in biology. This mechanism was once invoked in popular accounts of natural selection. For example, the idea that a predator species is doing a prey species a favour by eliminating its weakest members involves an egregious form of group selection. The English experimental biologist Wynne-Edwards provided an especially explicit manifesto on group selection and became thereby a favourite target for those wishing to argue against it. In Wynne- Edwards (1962), for example, he argued that birds limit the size of their clutches of eggs to ensure that the size of the population does not exceed the comfortable carrying capacity of the environment.

These particular group selection arguments were effectively devastated by Williams (1966). If a new type of individual does not so obligingly limit her clutch, why would this more fertile type not take over the population, without regard for the standard of living? Indeed, there are compelling arguments why it is in the interests of the individual to limit her clutch size. For example, it might well be that, beyond a certain point, an increase in the number of eggs reduces the expected number of offspring surviving to maturity, because each egg then commands a reduced share in parental resources. A finite optimum for clutch size is then to be expected.

Dawkins (1976 and 1982, for example) has been even more insistent than Williams on rejecting group selection, going further in arguing for the primacy of the gene as a still lower-level unit of selection. There certainly are phenomena best understood at the level of the gene. Consider, for example, sickle-cell anemia. At the relevant locus, there exists a particular variant gene, a particular allele, that is. If one of the two alleles that are present at this locus is this variant gene, the individual has improved resistance to malaria. However, if both alleles are of this variant type, the red blood cells have a characteristic sickle shape. Such cells malfunction by not carrying adequate oxygen, implying increased mortality. Under sexual reproduction, there is no way of maintaining only the individuals who have exactly one copy of the variant gene. Rather, both alleles are maintained as a stable mixture, where each allele is present in individuals who have quite different fitnesses.

There are presumably a fair number of cases where the interests of the gene and the individual do not conflict. In any case, it is often difficult to give concrete form to the gene as the unit of selection, given our ignorance of the details of the transformation of genes into individuals, particularly for complex behavioural characteristics (Grafen 1991, advocates finessing such detailed questions on the genetic basis of individual variation. This is his so-called ‘phenotypic gambit’). Hence, despite the theoretical primacy of the gene, we restrict attention here to the comparison between the individual level and the group level of selection.

In order to fix ideas, consider the following outline of the classic model that addresses the issue of individual selection versus group selection.

The Haystack Model

The following is a simplified account of Maynard Smith (1964). There are a number of haystacks in a farmer’s field, each of which is home to two mice. Each pair of mice now play the Prisoner’s Dilemma game, with the usual two choices: cooperate or defect. Payoffs for each individual take the concrete form of the number of offspring, but reproduction is asexual, with offspring inheriting their mother’s choice of strategy. There are a number of subsequent stages of play, where the mice in each haystack are paired at random to play the Prisoner’s Dilemma game. The number of individuals within the haystack choosing each strategy then grows in an endogenous fashion, as does the overall size of the group. Every so often, once a year, say, the haystacks are removed, and the mice are pooled into a single large population. Now pairs of mice are selected at random from the overall population to re-colonize the next set of haystacks, and excess mice die.

Maynard Smith’s intention here was to give the devil his due, by building a model in which group selection might well have an effect. At the same time, he wished to show that the parametric assumptions needed to make group selection comparable in strength to individual selection would be unpalatable. In order for group selection to be effective there must be a mechanism that insulates the groups from one another. Only then can a cooperative group be immune to infection by a defecting individual, and maintain its greater growth rate. Even with the temporary insulation of each haystack in this model, cooperation will evolve only if there are sufficient rounds of play within each haystack, so that defectors from a particular haystack are likely to be eliminated when the groups are reformed.

A loose description of the problem with group selection is that it relies too heavily upon a group becoming extinct as a likely consequence of a choice that is bad for the group. There is clearly scope, in reality, for individual selection, since individuals die frequently; group selection is less plausible, since there may not be enough extinction of groups.

An Example of Group Selection?

Despite the disfavour into which group selection has fallen in biology, there remain examples of cases that are challenging to explain otherwise. One of these concerns the interaction of the myxoma virus and European rabbits in Australia (Lewontin 1970, proposed a group selection interpretation of this case. Sober and Wilson 1999, pp. 45–50, give a – somewhat partisan – view of the subsequent debate). English rabbits were introduced into Australia in a misguided attempt to recreate the English countryside in Australia, but their numbers grew out of control, causing massive economic damage to farms there. The myxoma virus was first identified in South American forest rabbits, where it was only mildly virulent. When this South American strain was originally tested on Australian rabbits, however, it seemed an ideal solution to the rabbit infestation there, since it killed nearly the entire sample. Unfortunately, in the long run, after the South American virus had been present in the Australian rabbit population for a while, the rabbit mortality rate fell dramatically.

Why? Perhaps the most obvious explanation is that the rabbit population had been selected to have greater resistance to the virus, consistent with individual selection of rabbits. That this was true to some extent was demonstrated by the finding that the new Australian strain of the virus had a greater effect on laboratory- bred Australian rabbits than on the feral stock. However, this effect was not sufficient to explain the entire drop in rabbit mortality. Indeed, both laboratory and wild strains of Australian rabbits were less susceptible to the new Australian strain of the virus than they were to the original strain imported from South America. The virus had evidently evolved to be less virulent as a result of its interaction with the Australian rabbit population.

This situation might be roughly mapped onto the haystack model as follows. Consider a group of viruses to be those infecting a given rabbit. The evolutionary success of this group might best be promoted by settling for a moderate level of mortality for the host rabbit. Whatever the other advantages to the virus of strategies that induce high mortality in the rabbit, the early death of the current host makes transmission to a new host difficult. However, prolonging the life of the host is a public good from the point of view of the infecting viruses. A strain of virus with a strategy that was more lethal to the host could then grow as a fraction of the group of viruses. As in the haystack model, however, if there were enough generations of the virus within each rabbit’s lifetime, this conflict between the group and the individual might be resolved in favour of the group, as suggested by the data.

Selection Among Equilibria

When does group selection matter in biology? In theory, it could lead to different results than would individual selection, as in the Prisoner’s Dilemma, and as it may in above example. In practice, the above example is atypical, and group selection is usually a rather weak force. There is, however, one compelling scenario in which group selection would operate robustly, in any species. This is as a mechanism to select among equilibria (Boyd and Richerson 1990). Consider a population that is divided into various sub-populations, which are largely segregated from one another, so that migration between sub-populations is limited. Each sub-population plays the same symmetric game, which has several symmetric equilibria. Play of this game involves a random matching of the members of each sub-population. Individual selection ensures that some equilibrium is attained, within each sub-population. But group selection is then free to operate in a leisurely fashion to select the Pareto- superior equilibrium, since there is no tension here between the two levels of selection.

Group Selection and Economics

Why does group selection matter in economics? Group selection is the most obvious mechanism for generating preferences in humans that might make them behave in the social interest rather than that of the individual. At stake, then, is nothing less than the basic nature of human beings. As an economist, one should be sceptical of the need to suppose that individuals are motivated by the common good. Economic theory has done well in explaining a wide range of phenomena on the basis of selfish preferences, and so the view of the individual as the unit of selection is highly congenial to economists. Furthermore, to the extent that armchair empiricism suggests that non-selfish motivations are sometimes present, these seem as likely to involve malice as to involve altruism. For example, humans seem sometimes motivated by relative economic outcomes, which involve such a negative concern for others. Group selection is a blunt instrument that might easily ‘explain’ more than is true.

There are, nevertheless, some aspects of human economic behaviour that are tempting to explain by group selection. For example, we have a proclivity for trade that may go beyond myopic self-interest. As Darwin, an astute observer of both human beings and the natural world, observed on one of his visits to Tierra del Fuego,

Some of the Fuegians plainly showed that they had a fair notion of barter. I gave one man a large nail (a most valuable present) without making any signs for a return; but he immediately picked out two fish, and handed them up on the point of his spear. (Darwin 1845, ch. 10)

That is, human beings are often willing to trade with strangers they will likely never see again, as might be analogous to cooperating in the one-shot Prisoner’s Dilemma. There is no shortage of reliable data showing that human beings are capable of such apparently irrationally cooperative behaviour, in appropriate circumstances. Whatever the underlying reasons for this, it is clearly significant, and may even help account for the prodigious economic and biological success of humans.

Furthermore, identifying the underlying reasons would help shape a theory of such behaviour that remains falsifiable; such a theory might also predict what alternative circumstances might induce non-cooperative or antagonistic behaviour. Group selection is an obvious avenue to explore in this connection.

It is sometimes argued, in particular, that the structure of hunter–gatherer societies helps account for cooperative behaviour. Hunter–gatherer societies were composed of a large number of relatively small groups, and individuals within each group were often genetically related. Perhaps, so the argument goes, we acquired an inherited psychological inclination towards conditional cooperation in such a setting, partly perhaps as a result of group selection. These inclinations then carried over into modern societies, despite genetic relatedness now being essentially zero on average. Seabright (2004), for example, argues eloquently that human societies cannot function on myopic self-interest alone, but also that the trust needed for exchange in market economies sprang from adaptations to archaic small groups. It is hard to believe, however, that hunter–gatherers never encountered strangers. If there were good reasons to condition on this distinction, why would corresponding different strategies not have evolved?

A phenomenon that looms large in the case of human beings is culture, by which is meant the non-genetic transmission of behaviour, by imitation of peers, for example. Many attempts to derive cooperative behaviour have focused then on group selection in models of cultural transmission. We now turn to this literature.

Cultural Group Selection and Economics

A spectacular and famous example of cultural group selection features the Nuer and Dinka, who lived as neighbouring ethnic groups in 18th century southern Sudan (Kelly 1985). Despite the similarity of their environment, these two groups differed in various economic and political respects. Perhaps the key difference was that Dinka lived in small groups, the size of which was related to the needs of their economic activity. The Nuer, on the other hand, organized their society according to a patrilineal system that bound many such smaller units together, over a greater geographic area. Over a period of 100 years, the Nuer expanded their territory at the expense of the Dinka, killing or assimilating their rivals. Nuer culture, that is, was selected over that of the Dinka.

Despite the apparently strong military advantages of the Nuer political system over that of the Dinka, it seems plausible that any individual – Dinka or Nuer – would have had the incentive to play the appropriate role within their society. It would not have been possible for an individual Dinka to shift unilaterally to Nuer behaviour.

Human societies have the capacity to render a wide range of behaviour optimal for the individual. If a society wishes to adopt a rather arbitrary rule, it may well have adequate sanctions to enforce this (Boyd and Richerson 1992). As described above, group selection can then be relied upon to select between various equilibria. Boyd et al. (2003) make the important additional point that, because punishment is needed only infrequently near full cooperation, the weak force of group selection can support cooperation as an equilibrium, without the usual need for punishment of those who fail to punish, and so on.

Group selection is uncontroversial as a mechanism for selecting an efficient outcome within each group. But this does not directly explain observations, such as Darwin’s, of our apparent willingness to trade with strangers. The difficulty is stark: defection is a strictly dominant strategy in the one-shot Prisoner’s Dilemma game.

One stark option is that cooperation is hard-wired. Bowles et al. (2003) argue that, if behaviour is directly genetically controlled, cooperative behaviour may then be sustained in the presence of culturally maintained institutions. These institutions, food-sharing for example, serve to reduce the negative impact on individuals of cooperative behaviour. Group selection arises from conflict between groups, with more cooperative groups emerging victorious in such conflict.

However, human strategic behaviour is genetically mediated in a complex and poorly understood way, and is not always cooperative. Even if we did somehow acquire a genetic inclination to cooperate in archaic societies, shouldn’t we now be in the process of losing this inclination in modern large and anonymous societies?

A Recent Revival?

Sober and Wilson (1999) push energetically for a rehabilitation of group selection within biology. They argue that group selection is closely related to other well- accepted phenomena. For example, they argue that kin selection – the widely accepted notion that individuals are selected to favour their relations – should be regarded as a special case of group selection.

In its simplest form, kin selection is the argument that individuals should undertake actions that benefit a relation if this benefit, when deflated by the degree of relatedness, exceeds the cost to the first individual (This is ‘Hamilton’s rule’ as in Hamilton 1964). The empirical evidence in favour of kin selection is overwhelming mothers, human and non-human, routinely make large sacrifices in favour of offspring. Even economists would exempt such interactions from the general presumption of selfish behaviour.

Sober and Wilson certainly make the case that these phenomena can be viewed in an integrated fashion. Indeed, it is a consequence of the ‘Price equation’ (Price 1970) that what matters most fundamentally is the likelihood that altruistic individuals will be preferentially matched with other altruistic individuals. In the case of kin selection, this could be ensured by a mechanism to directly recognize relations by smell, for example – or by indirect but reliable methods, such as, for example, assuming that those who are found in proximity to one’s mother are relations.

Bergstrom (2002) provides the best introduction to the literature on group selection for economists. He presents a unified and intuitive treatment of the literature, and also stresses the key role of assortative matching. Thus, for example, if the subgroups in the haystack model are dispersed after one round of the game, and then randomly recombined, there is no force to group selection. Only if the subgroups remain together for repeated play of the game is there effective assortative matching. Such a structure seems less compelling for non-relations than it is for relations.

Despite the formal analogies between kin selection and group selection, acceptance of the former does not then require acceptance of the latter. In the end, a sceptical but not dogmatic view of the importance of group selection to human economic behavior seems warranted.

See Also

I received helpful comments from Ted Bergstrom, Lawrence Blume, Sam Bowles, Steven Durlauf and Peter Sozou. I thank them without blaming them.


  1. Bergstrom, T.C. 2002. Evolution of social behavior: Individual and group selection. Journal of Economic Perspectives 16(2): 67–88.CrossRefGoogle Scholar
  2. Bowles, S., J.-K. Choi, and A. Hopfensitz. 2003. The coevolution of individual behaviors and social institutions. Journal of Theoretical Biology 223: 135–147.CrossRefGoogle Scholar
  3. Boyd, R., and P. Richerson. 1990. Group selection among alternative evolutionarily stable strategies. Journal of Theoretical Biology 145: 331–342.CrossRefGoogle Scholar
  4. Boyd, R., and P. Richerson. 1992. Punishment allows the evolution of cooperation (or anything else) in sizable groups. Ethology and Sociobiology 13, 171–195.Google Scholar
  5. Boyd, R., H. Gintis, S. Bowles, and P.J. Richerson. 2003. The evolution of altruistic punishment. Proceedings of the National Academy of Sciences 100: 3531–3535.CrossRefGoogle Scholar
  6. Darwin, C. 1845. Journal of researches into the natural history and geology of the various countries visited by H. M. S. Beagle. London: Henry Colburn. Online. Available at Accessed 23 Oct 2006.
  7. Darwin, C. 1859. On the origin of species by means of natural selection. London: John Murray. Online. Available at Freeman_OntheOriginofSpecies.html. Accessed 8 Nov 2006.
  8. Darwin, C. 1871. The descent of man and selection in relation to sex. London: John Murray. Online. Available at Freeman_TheDescentofMan.html. Accessed 8 Nov 2006.
  9. Dawkins, R. 1976. The selfish gene. Oxford: Oxford University Press.Google Scholar
  10. Dawkins, R. 1982. The extended phenotype: The gene as the unit of selection. Oxford: Oxford University Press.Google Scholar
  11. Grafen, A. 1991. Modelling in behavioural ecology. In Behavioral ecology: An evolutionary approach, 3 ed., ed. J.R. Krebs and N.B. Davies. Oxford: Blackwell Scientific.Google Scholar
  12. Hamilton, W.D. 1964. The genetical evolution of social behavior. Journal of Theoretical Biology 7: 1–16.CrossRefGoogle Scholar
  13. Kelly, R.C. 1985. The Nuer conquest: The structure and development of an expansionist system. Ann Arbor: University of Michigan Press.Google Scholar
  14. Lewontin, R.C. 1970. The units of selection. Annual Reviews of Ecology and Systematics 1: 1–18.CrossRefGoogle Scholar
  15. Maynard Smith, J. 1964. Group selection and kin selection. Nature 201: 1145–1147.CrossRefGoogle Scholar
  16. Price, G. 1970. Selection and covariance. Nature 277: 520–521.CrossRefGoogle Scholar
  17. Seabright, P. 2004. The company of strangers. Princeton: Princeton University Press.Google Scholar
  18. Sober, E., and D.S. Wilson. 1999. Unto others: The evolution and psychology of unselfish behavior. Cambridge, MA: Harvard University Press.Google Scholar
  19. Williams, G.C. 1966. Adaptation and natural selection: A critique of some current evolutionary thought. Princeton: Princeton University Press.Google Scholar
  20. Wynne-Edwards, V.C. 1962. Animal dispersion in relation to social behavior. Edinburgh: Oliver and Boyd.Google Scholar

Copyright information

© The Author(s) 2008

Authors and Affiliations

  • Arthur J. Robson
    • 1
  1. 1.