![](https://csdnimg.cn/release/download_crawler_static/5531333/bg4.jpg)
4 J.C.J.M. van den Bergh, J.M. Gowdy / Journal of Economic Behavior & Organization 72 (2009) 1–20
Group selection is often discusse d and judged in terms of its application to altruism. However, group phenomena can
be linked to a variety of social (other-regarding) preferences including reciprocity, reciprocal fairness, inequity aversion,
pure altruism, altruistic punishment, spite, envy and status-seeking (comparison). Second, group selection can address a
range of genetic or individual interactions, notably cannibalism (especially insects), conflict (fighting) and complementary
roles (labor division) such as in community selection (Goodnight and Stevens, 1997; Swenson et al., 2000). So any genetic
or cultural trait might influence within and between-group selection, even if the directions of the effects of these are not
necessarily opposite as in the case of altruism.
3. What economists can learn from the group selection debate
Natural selection above the level of the individual was a quite acceptable idea to Darwin, Wallace and Spencer, all of
whom believed in the differential survival of groups. Nevertheless, group selection was originally met with skepticism if
not outright hostility, partly due to the publication of Wynne-Edwards (1962) book Animal Dispersion in Relation to Social
Behavior, which was later evaluated as an incorrect statement of group selection and inconsistent with modern formulations
of it (Wilson, 1983).
5
Wynne-Edwards received strong adverse reactions from authorities in biology, notably Williams (1966).
Two widely accepted theories countered Wynne-Edward’s version of group selection. One was kin selection, the idea that
apparent altruism is genetically based because altruists are actually protecting their own genes (inclusive fitness) by helping
close relatives survive (Hamilton, 1964). The other was reciprocal altruism, the view that apparent altruism was based on the
expectation that favors would be returned (Trivers, 1971). These two theories formed the basis of a new field, ‘sociobiology’,
which initially was opposed to the idea of group selection, arguing that social behavior is entirely the result of gene-level
and individual-level selection (Ruse, 1979; Wilson (E.O.), 1975).
While the critique of Wynne-Edwards’s version of group selection was correct, it should not be misinterpreted as
undercutting later, more sophisticated, conceptualizations of group selection. Williams’ influential critique contained many
misperceptions of group selection, one being the averaging fallacy. This means that group selection effects are attributed to
individual-level (i.e. between-individual) selection due to simply calculating the average fitness of individuals across groups,
rather than decomposing it into average fitness per group and a between-group selection component depending on the
relative size of groups in the total population (Wilson, 1983). The averaging fallacy in fact merely defined group selection
away, rather than showing it to be incorrect.
Williams assumed that when a trait is decreasing in frequency in every group, its population frequency must also decrease.
Surprisingly, this is not necessarily true. The reason is differential growth at the group level which causes the composition
of the population in terms of relative group sizes to change. The net outcome may be that the trait will increase in fre-
quency in the overall population. In statistics this is known as Simpson’s (1951) paradox. Suppose a population consists
of two groups, then this paradox requires that the group with relatively many individuals with the trait grows relatively
fast compared to the groups with relatively few individuals with the trait; the proportion in the population of the first
group will then increase (note that this is a necessary but not sufficient condition). In mathematical terms this can be for-
mulated as follows. Suppose that population size at time t is p(t), the population consists of two groups with sizes g
1
(t)
and g
2
(t), the number of individuals with the trait is a(t), and a
i
(t) in group g
i
(t)(i = 1, 2). Then p(t)=g
1
(t)+g
2
(t). Now
even if the changes over time in frequencies of the trait are characterized for group 1 by a
1
(t + 1)/g
1
(t +1)<a
1
(t)/g
1
(t) and
forgroup2bya
2
(t + 1)/g
2
(t +1)<a
2
(t)/g
2
(t), then it is still possible that the trait’s frequency in the population increases:
that is, [a
1
(t +1)+a
2
(t + 1)]/[g
1
(t +1)+g
2
(t +1)]>[a
1
(t)+a
2
(t)]/[g
1
(t)+g
2
(t)] or a(t + 1)/p(t +1)>a(t)/p(t) holds. To illustrate
the numerical possibility: 1/4 < 2/7 and 5/7 < 3/4 while (1 + 5)/(4 + 7) = 6/11 > 5/11 = (2 + 3)/(7 + 4). The example suggests that
a rather special case is require d, which is not surprising, otherwise it would not have been calle d a paradox. The paradox just
shows the power of group selection, that in principle it can increase a trait’s frequency in a population even if its frequency
falls in all groups. The net effects of group selection can be much larger if the frequency of traits can also increase in certain
groups. The latter is the more common or general case of group selection, as will become clear later.
Especially in the context of altruism, the relationship between kin and group selection has been subject to much debate.
First, the observed presence of high kin relatedness in a population should not be confused, as is often done, with kin selection
acting as a crucial factor in the historical evolution toward the current system (Griffin and West, 2002). In the context of
eusociality, it is now considered likely that group selection gave rise to kin selection which in turn produced a high degree of
relatedness in colonies (Wilson and Hölldobler, 2005). Indeed, repeated group selection may increase relatedness in a group,
through limited dispersal or inbreeding (Hamilton, 1975), thus providing a basis for more effective kin selection. Things get
even more complex when group and kin selection function in tandem (Avilés et al., 2004). It is also often overlooked that
the effect of kin selection, such as altruism benefits to kin and thus selfish benefits from the shared gene’s perspective, is
opposed by competition between kin (Frank, 1998). Of course, seemingly altruistic behavior among kin-related individuals
can simply be the result of direct, reciprocal or general group (i.e. not kin-group specific) benefits (Griffin and West, 2002).
5
Wilson and Sober (1994) make a distinction between naïve and modern group selection theories. Wynne-Edwards (1962) is considered an example of
the first as his theory does not account for effects of individual selection and essentially assumes that higher level adaptation (even of a single, isolated
group) is a fact. Modern group (or multilevel) selection theory instead recognizes simultaneous processes of individual and group selection, as wellasgroup
selection requiring multiple groups.