vol. 172, no.

5 the american naturalist november 2008

Natural Selection on Male Wealth in Humans
Daniel Nettle
*
and Thomas V. Pollet

Centre for Behaviour and Evolution, Newcastle University, Henry

Wellcome Building, Framlington Place, Newcastle NE2 4HH,
United Kingdom
Submitted March 31, 2008; Accepted June 17, 2008;
Electronically published September 22, 2008
Online enhancement: appendix.
abstract: Although genomic studies suggest that natural selection
in humans is ongoing, the strength of selection acting on particular
characteristics in human populations has rarely been measured. Pos-
itive selection on male wealth appears to be a recurrent feature of
human agrarian and pastoralist societies, and there is some evidence
of it in industrial populations, too. Here we investigate the strength
of selection on male wealth, rst in contemporary Britain using data
from the National Child Development Study and then across seven
other varied human societies. The British data show positive selection
on male income driven by increased childlessness among low-income
men but a negative association between personal income and re-
productive success for women. Across cultures, selection gradients
for male wealth are weakest in industrial countries and strongest in
subsistence societies with extensive polygyny. Even the weakest se-
lection gradients observed for male wealth in humans are as strong
as or stronger than selection gradients reported from eld studies of
other species. Thus, selection on male wealth in contemporary hu-
mans appears to be ubiquitous and substantial in strength.
Keywords: natural selection, natural populations, humans, selection
gradients, sexual selection.
Biologists have long been interested in measuring ongoing
natural selection in wild populations (Endler 1986; Grant
and Grant 1989; Hoekstra et al. 2001; Kingsolver et al.
2001). However, natural selection within human groups
is rarely quantied, despite clear molecular signatures of
ongoing selection in the human genomeselection that
appears to have accelerated in the past few thousand years
* E-mail: [email protected].

E-mail: [email protected].
Am. Nat. 2008. Vol. 172, pp. 658666. 2008 by The University of Chicago.
0003-0147/2008/17205-50349$15.00. All rights reserved.
DOI: 10.1086/591690
(Reed and Aquadro 2006; Voight et al. 2006; Wang et al.
2006; Hawks et al. 2007). One trait on which phenotypic
selection is likely to be acting is wealth. A consistent nd-
ing from many societies is that wealthier individuals, par-
ticularly men, have higher reproductive success than less
wealthy ones (Borgerhoff Mulder 1987; Voland 1990;
Cronk 1991; see Hopcroft 2006 for a review). Both an-
thropologists (Irons 1979) and historians (Clark and Ham-
ilton 2006) have thus concluded that a fundamental feature
of human societies is that men strive for cultural goals
such as wealth and status in order to convert these achieve-
ments into reproductive success.
Studies of contemporary developed economies have
produced an unclear picture of whether selection on male
wealth is still operative. Some studies have found a null
or even a negative relationship between socioeconomic
status and reproductive success in men (Vining 1986; van
den Berghe and Whitmeyer 1990; Perusse 1993), leading
to the general view that modern societies are paradoxical
from a Darwinian perspective (Vining 1986; Borgerhoff
Mulder 1998). However, there are methodological and
conceptual problems with the studies that demonstrate
null or negative associations. Some of these studies (Vining
1986; Kanazawa 2003) include individuals who are in the
midst of their reproductive careers and thus potentially
confound differences in the timing of reproduction with
differences in nal reproductive success. Also, the mea-
sures used in these studies often fail to distinguish the
effects of wealth, on the one hand, from education, on the
other (Hopcroft 2006). This is signicant because pro-
longing parental education may reduce the amount of time
that can be devoted to reproduction, and increasing off-
spring education increases the per-offspring investment
cost. Thus, one might expect smaller families in social
strata whose members either have prolonged education
themselves or who expect it for their children. Many mea-
sures of socioeconomic status are based on the prestige of
a persons occupation, and thus they confound educational
and wealth effects that may be operating in opposite
directions.
A small number of studies have overcome these limi-
tations. Hopcroft (2006) uses a probability sample of the
U.S. population and measures individual income and ed-
Natural Selection on Male Wealth in Humans 659
ucation separately. This study shows that education and
income do indeed work in opposite directions, with more
education reducing offspring number for both sexes but
a larger income causing offspring number to increase for
men. Fieder and Huber (2007) show, for a representative
sample of the Swedish population, that both education
and income increase reproductive success for men but that
increasing income is associated with fewer children for
women. Another study has shown that, when educationally
homogenous subsets of populations are considered (and
education is thus controlled for), signicant positive as-
sociations between income and reproductive success ap-
pear for men but not for women (Weeden et al. 2006).
Thus, the best studies suggest that selection on male in-
come is still occurring in contemporary industrial societies
and that women display an opposite pattern.
The association between male income and reproductive
success could arise in three ways. First, among those men
who have one family, those with higher incomes could
have more children. Second, higher-income men could be
more likely to form a family in the rst place, and thus
the association would be driven by a disproportionate level
of childlessness among low-income men. Third, men with
high incomes might have an increased probability of form-
ing second or subsequent families. In Fieder and Hubers
(2007) study, the association between male income and
reproductive success is accounted for entirely by the sec-
ond mechanism, disproportionate childlessness in low-
income men. They argue that this is due to low-income
men having difculty attracting spouses, which is plausible
given the well-documented female preference for male re-
sources (Buss 1989; Borgerhoff Mulder 1990; Pollet and
Nettle 2008). Hopcroft (2006) suggests that the third
mechanism (multiple family formation) might be impor-
tant for the association between male wealth and repro-
ductive success in her U.S. sample, but she is unable to
test the hypothesis directly.
This study has two parts. In the rst part, we investigate
whether there is phenotypic selection acting on wealth in
the contemporary British population, how the pattern dif-
fers between the sexes, and what mechanisms are involved.
We use data from the National Child Development Study
(NCDS), which is an ongoing longitudinal investigation
of all the individuals who were born in the United King-
dom in a single week in March 1958. This design has the
advantage of guaranteeing social representativeness of the
sample. Indeed, strictly speaking, the NCDS is not a sample
at all but is the entire British population of a particular
narrow age band. There has been considerable loss to fol-
low-up over the years, and this loss is not completely
random with respect to socioeconomic status (Nettle
2003). However, the data set remains sizable and broadly
representative. The NCDS also has the advantage that all
cohort members are exactly the same age, removing any
need to statistically control for age. We examine income,
education, and number of children at age 46 years. Fieder
and Huber (2007) estimate that, in their Swedish cohort,
more than 99% of all female reproduction and more than
96% of all male reproduction occurs before this age, and
so the number of children reported is close to nal lifetime
reproductive success.
The second part of the study is comparative, both within
and across species. If there is phenotypic selection on male
wealth in contemporary Britain, how strong is it compared
with the selection on male wealth in other industrial and
in preindustrial societies? Moreover, how strong is it com-
pared with selection that is typically observed by biologists
working with other species in the eld? Directional phe-
notypic selection can be usefully compared across studies
by considering b, the standardized linear selection gradient
(Kingsolver et al. 2001). This is the standardized slope of
the regression relationship of the tness measure (e.g.,
number of offspring) on the trait under selection (e.g.,
wealth). It is an important measure in evolutionary terms
because, together with the heritability, it determines the
response to selection. We compare estimated values of b
on male wealth across eight societies with signicant ac-
cumulable wealth (two ethnographic agrarian or pastor-
alist societies, three historical agrarian societies, and three
contemporary industrial societies) and set these alongside
recent estimates of the selection acting on male hunting
ability in hunter-gatherer populations (Kaplan and Hill
1985; Marlowe 1999; Smith 2004; Gurven and von Rueden
2006). We then compare the selection gradients for male
wealth in humans against a large set of selection gradients
from other taxa drawn from the literature (Kingsolver et
al. 2001).
Methods
NCDS
The NCDS began with a sociological and perinatal medical
investigation of all the babies born in Britain during the
period March 39, 1958 ( ). The cohort formed n p17,416
by these babies has been restudied in seven major sweeps
consisting of parental, medical, or teacher reports and,
especially more recently, self-reports by the cohort mem-
bers themselves. It is possible to combine information
from different sweeps using a unique identier for each
cohort member. Data reported here are derived mainly
from NCDS7 (2004; cohort age, 46 years; ), n p11,939
with additional material from NCDS6 (2000; cohort age,
42 years; ) and NCDS5 (1991; cohort age, 33 n p10,979
years; ). Degrees of freedom vary from analysis n p10,986
660 The American Naturalist
to analysis because of missing data points within a survey
or the individual not having participated in all relevant
sweeps. Descriptive statistics for key variables are given in
table A1 in the appendix in the online edition of the Amer-
ican Naturalist.
Number of children is a composite variable derived
from responses to questions about biological children born
or fathered in 1991, 2000, and 2004. Income is from 2004
and is based on responses to questions about take-home
pay from employment or self-employment. The vast ma-
jority of cohort members had some form of employment
or self-employment (92% of men, 83% of women). Note
that individuals who are homemakers, are unemployed,
or are acquiring an education are absent from the re-
porting of the income variable rather than being entered
as zero. Income has been annualized (median income,
15,236) and is logged to reduce skewness.
The education variable is based on responses to ques-
tions in 2004 about highest academic qualication ob-
tained. Here, we distinguish three groups: those with only
academic qualications usually acquired at age 16 (i.e.,
General Certicate of Secondary Education; ), n p6,477
those with academic qualications usually acquired at the
age of 18 (i.e., A-levels; ), and those with a n p1,277
university degree or equivalent ( ). More ne- n p1,780
grained classications produce essentially the same results
(data not shown).
To investigate patterns of relationship formation, we
derive two additional variables. Marital situation is based
on marital status in 2004 and divides the cohort into three
groups: never married ( ), married once (still with n p691
rst spouse; ), and multiple marriages (either n p5,394
remarried or now cohabiting with someone other than
rst spouse; ). Although marriage was over- n p1,852
whelmingly the norm in this cohort, the marital situation
variable does entirely not account for unmarried cohab-
itations, so we also derive a second variable, cohabitations,
which is a sum over responses from 1991, 2000, and 2004
of all reported lifetime cohabitations with a partner that
lasted more than 1 month. We use general linear models
and least squares regression to examine the effects of
ln(income) and education on number of children, and we
use logistic regression for secondary analyses where the
outcome is categorical (e.g., childlessness or marital
status).
Comparative Analysis
The comparative analysis presented here has two parts.
First, we identied 10 other studies in the literature that
contain data that would allow for the estimation of a linear
selection gradient for male wealth or a related measure.
Two studies in addition to ours were from representative
samples of contemporary industrial populations (Hopcroft
2006; Fieder and Huber 2007). Three more were from
historical studies of European and European-descended
agrarian populations before industrialization (Mealey
1985; Roskaft et al. 1992; Clark and Hamilton 2006). Two
more were from ethnographic studies of African subsis-
tence societies, one agrarian and one pastoralist (Borger-
hoff Mulder 1987; Cronk 1991). The nal three studies
came from ethnographically studied hunter-gatherer pop-
ulations, where there is little accumulated wealth. How-
ever, for hunter-gatherers, there is a large quantity of lit-
erature showing relationships between male hunting ability
and reproductive success (Kaplan and Hill 1985; Marlowe
1999; Smith 2004; Gurven and von Rueden, 2006), so we
include estimated selection gradients on hunting ability
for comparison. Standardized linear selection gradients (b)
were taken directly from the papers, obtained from the
authors of the studies, or calculated ourselves from data
presented. All the coefcients were either from individuals
at the end of their reproductive careers or from mixed-
age samples for which age was statistically controlled. Sam-
ple size varied from a few dozen men in the ethnographic
studies to thousands in the contemporary cohorts. Wealth
was assessed in different ways and with varying precision
in the different studies as a result of the large cross-cultural
differences in economic system (see table A1; note that
large stochastic variation in hunting returns makes the
hunting ability b values particularly prone to error; K. Hill
and K. Kintig, unpublished manuscript). However, be-
cause these are unitless coefcients, they can reasonably
be informatively compared, despite the variation in study
design and statistical procedures.
In the second part of the comparative analysis, we used
the database of selection gradients observed in eld pop-
ulations provided by Kingsolver et al. (2001). These re-
searchers performed a search of the literature from 1984
to 1997 for all reported estimates of selection strength in
diverse species. The resulting database contains more than
2,500 estimates, from 62 species in 51 genera. Of these,
995 were standardized linear selection gradients and were
thus directly comparable to those reported here for human
male income. For these purposes, only the strength and
not the direction of the gradients is important, and so we
consider the absolute b values, henceforth FbF. First, we
calculated where in the overall distribution of FbF values
our observations for human male wealth were located.
Subsequently, because typical selection strength is different
for fecundity, survival, and mating success (Kingsolver et
al. 2001), we restricted the comparison to gradients where
the measured component of tness is fecundity or mating
success.
Natural Selection on Male Wealth in Humans 661
Figure 1: Plot of best-t linear regression equations of number of children
against natural log of income (ln[income]), National Child Development
Study data, for men and women of different education groups.
Results
NCDS
Full results of general linear models are presented in table
A2 in the appendix. For the whole cohort, using number
of children as the dependent variable, sex and education
as factors, and ln(income) as a covariate, there are sig-
nicant effects of sex ( , , F p104.29 df p1, 5,575 P !
) and education ( , , ) but .01 F p5.15 df p2, 5,575 P ! .01
not of ln(income) ( , , P value not F p0.18 df p1, 5,575
signicant). However, there are signicant interactions be-
tween sex and education ( , , F p3.61 df p2, 5,575 P !
) and between sex and income ( , .05 F p104.40 df p
, ). The sex effect is the result of women on 1, 5,575 P ! .01
average having more children than men (marginal means:
women, 1.83 children; men, 1.66 children), whereas the
education effect is the result of the least educated group
having more children than the other groups (marginal
means: 16 years of education, 1.82 children; 18 years of
education, 1.70 children; degree achieved, 1.71 children).
The sex-education interaction can be appreciated by com-
paring mean number of children in the most and least
educated groups for men and for women. The most ed-
ucated men have 0.09 fewer children than the least edu-
cated men, whereas the most educated women have 0.22
fewer children than the least educated women (table A3
in the appendix). Thus, an increased level of education
has a greater negative effect on womens fertility than on
mens.
The sex-income interaction is illuminated by regressing
the number of children on ln(income) separately in the
male and the female halves of the cohort. For men, the
relationship is signicant and positive ( , F p23.93 df p
, ; ), whereas for women it is sig- 1, 2,592 P ! .01 b p0.10
nicant and negative ( , , ; F p97.38 df p1, 2,987 P ! .01
). Because previous literature suggests that the b p0.18
effects of income may be seen most clearly within edu-
cationally homogenous subsets, we also carried out re-
gressions separately for each sex in each educational group
(table A3). The b values of ln(income) were signicant
and positive in all three male education groups and were
signicant and negative in all three female groups (see g.
1). Within each sex, none of the slopes differs signicantly
from the others and none differs from the slope obtained
in the sample not divided by education.
In order to investigate the roles of family size, child-
lessness, and multiple family formation, we rst repeated
general linear models for each sex with ln(income) and
education as predictors and number of children as the
dependent variable, but we excluded childless individuals.
For men, neither ln(income) ( , , P F p2.43 df p1, 2,073
value not signicant) nor education ( , F p1.27 df p
, P value not signicant) has a signicant effect. 2, 2,073
For women, ln(income) does have a signicant effect
( , , ), but education does not F p6.73 df p1, 2,505 P ! .01
( , , P value not signicant). This F p0.62 df p1, 2,505
suggests that the effects of education on number of chil-
dren in both sexes are driven by more education increasing
the likelihood of childlessness.
We veried that this is the case using binomial logistic
regression of childlessness on education and ln(income)
(table A4 in the appendix). More educated men were sig-
nicantly more likely to be childless (odds ratio for child-
lessness in the most vs. the least educated group, 2.17;
), and the same was true for women (odds ratio P ! .01
Wald
for childlessness in the most vs. the least educated group,
1.47; ). The same regressions showed that in- P ! .01
Wald
creasing income signicantly reduces the likelihood of
childlessness for men (odds ratio for childlessness with
each unit of ln[income], 0.43; ). For women, it P ! .01
Wald
has the opposite effect, signicantly increasing the likeli-
hood of childlessness (odds ratio for childlessness with
each unit of ln[income], 2.28; ). Figure 2 illus- P ! .01
Wald
trates the effects of income on childlessness for both sexes.
To examine whether income increases the probability
of multiple family formation, we used multinomial logistic
regression to test the effect of ln(income) on marital sit-
uation (never married, rst and only marriage, or multiple
marriages; table A4). For men, the overall effect of
ln(income) is signicant, which is accounted for by an
increased ln(income) sharply reducing the likelihood of
never marrying compared with being in the rst and only
marriage category (odds ratio, 0.27; ). Increasing P ! .01
Wald
ln(income) does not signicantly affect the odds of having
multiple marriages versus having only one marriage (odds
ratio, 0.85; P value not signicant). For women, too, the
662 The American Naturalist
Figure 2: Proportion of (A) men and (B) women in the National Child Development Study who were childless in 2004, by for-sex quartile of
income.
overall effect of ln(income) is signicant. This is due to
ln(income) increasing the likelihood of never marrying
compared with having only one marriage (odds ratio, 1.85;
) and also increasing the likelihood of having P ! .01
Wald
multiple marriages compared with having only one mar-
riage (odds ratio, 1.31; ). We performed the same P ! .01
Wald
analysis using reported cohabitations rather than mar-
riages and achieved essentially identical results (data not
shown).
Comparative Analysis
Table 1 shows the linear selection gradients (b) on male
wealth calculated for eight agrarian and pastoralist societies
plus the estimated b values on hunting ability for three
hunter-gatherer groups. The b values from the three con-
temporary industrial societies are in the range 0.100.17,
with a mean value of 0.13, which is lower than the his-
torical and ethnographic societies. The historical European
and European-descended agrarian societies have a mean
gradient of 0.24 and the two remaining ethnographic so-
cieties, the Mukogodo pastoralists and the Kipsigis farmers
of Kenya, produce the highest gradients, with a mean of
0.63 (see Discussion). The hunter-gatherer estimates for
hunting ability are in the range 0.220.36, with a mean
value of 0.30.
The b values for male wealth among individuals from
contemporary industrial societies fall in the center of the
cross-taxa FbF distribution (forty-sixth percentile),
whereas those from other societies are at the higher end
(historical European, sixty-eighth percentile; polygynous
African, ninety-fourth percentile; g. 3A). The mean
hunter-gatherer hunting ability gradient is at the seventy-
sixth percentile of the cross-taxa distribution. Restricting
the comparison to just the studies in Kingsolver et al.s
(2001) database, which measured fecundity as the com-
ponent of tness, does not dramatically change these con-
clusions (contemporary industrial, forty-fth percentile;
historical European, sixty-eighth percentile; polygynous
African, ninety-second percentile; g. 3B). We additionally
compared the FbF values for male wealth in humans with
those based on measurements of mating success in the
cross-taxa database, because although our measure, strictly
speaking, is one of fecundity, our interpretation is that
variation in male fecundity is driven primarily by variation
in mating success. This reduces the position of the con-
temporary industrial b values to the thirty-eighth percen-
tile of the cross-taxa distribution and the historical Eu-
ropean b values to the sixty-second percentile, with the
polygynous African b values once again at the ninety-
second percentile.
Discussion
NCDS
The results of the analysis of the NCDS show that in-
creasing income has a signicant positive effect on repro-
ductive success in contemporary British men. This is partly
offset by a negative effect on reproductive success of level
of education achieved. However, the effect of income is
signicant even without partialing out educational differ-
ences. Thus, this study conrms comparable investigations
demonstrating positive selection on male income in de-
veloped economies (Hopcroft 2006; Fieder and Huber
2007), contradicting earlier null or negative ndings.
The pattern for women is very different from that for
men. There is a strong negative relationship between ed-
ucation and reproductive success and also a negative re-
lationship between income and reproductive success. It is
Natural Selection on Male Wealth in Humans 663
Table 1: Linear selection gradients on male wealth or male hunting ability across 11 different human societies
Characteristic, source Population Measurement b Comment
Male wealth:
Borgerhoff Mulder
1987
Kipsigis
agriculturalists
Land ownership .68 Weighted mean for two oldest cohorts
Cronk 1991 Mukugodo pastoralists Livestock .58 Controlling for age
Rskaft et al. 1992 Norwegian farmers,
17001900
Agricultural
resources
.32 From table 2
Clark and Hamilton
2006
English testators,
15401850
Assets in will .18 Data (including some not in article)
supplied by G. Clark; assets log
transformed
Mealey 1985 Nineteenth-century
Mormons
Wealth .23 Middle value from three cohorts reported
Hopcroft 2006 Contemporary United
States
Income .12 Controlling for age
Fieder and Huber
2007
Contemporary
Sweden
Income .17 Number of children square-root
transformed
NCDS data (this
article)
Contemporary Britain Income .10 Income log transformed
Male hunting ability:
Kaplan and Hill
1985
Ache Hunting ability
(weighed returns)
.31 From data in table 1 controlling for age
Marlowe 1999 Hadza Hunting ability
(ranked by peers)
.36 Controlling for age
Gurven and von
Rueden 2006
Tsimane Amount of meat ac-
quired (kg)
.22 Controlling for age
Note: NCDS p National Child Development Study.
likely that the male and female patterns arise in different
ways. The negative effects for women can mainly be in-
terpreted as the results of trade-offs that women have to
face if they desire childrenfor example, ceasing their
education or working part-time to allow for motherhood
(Marini 1984). For men, by contrast, a possible expla-
nation of the income effect is female choice. A large por-
tion of the literature that analyzes across many cultures
has documented female preferences for men with resources
(Buss 1989; Borgerhoff Mulder 1990; Pollet and Nettle
2008), and the increased probability of never being mar-
ried or of having no cohabitations among men with low
incomes in the NCDS is consistent with poorer men having
difculty attracting mates. Thus, the causal pathways to
childlessness do indeed seem to be rather different for the
two sexes (Keizer et al. 2007).
Note, however, that we cannot completely exclude the
possibility of reverse causality from male reproductive suc-
cess to income, with men with children seeking and gain-
ing higher wages than those without. To do so would
require ner-grained longitudinal analyses than are pos-
sible using the data we have here. However, given earlier
literature on wealth and marital transitions that point to
an inuence from income on marriage rather than the
other way around, we feel this is an unlikely scenario (Na-
kosteen and Zimmer 1997; Burgess et al. 2003).
Although we found negative selection gradients on in-
come for women, we note that this does not mean that
women in households with lower incomes have higher
reproductive success. Rather, women have to trade per-
sonal income generation for children. They may compen-
sate for this by seeking mates with relatively high incomes;
thus, the relationship between a womans reproductive
success and her households income could be positive.
In the NCDS sample, the positive effects of male income
on reproductive success were entirely due to a reduction
in the probability of childlessness. This is similar to the
ndings for Sweden reported by Fieder and Huber (2007;
cf. our g. 2 with their g. 1E). We did not nd any
evidence of high-income men increasing reproductive suc-
664 The American Naturalist
Figure 3: A, Position in the cross-taxa distribution of linear selection gradients (FbF) of selection on male wealth in contemporary industrial (CI),
historical European (HE), and African polygynous (AP) human societies. B, Comparison restricted to FbF values where the component of tness
measured is fecundity. Cross-taxa data are from Kingsolver et al. (2001).
cess with serial marriages. However, we note that there is
still scope for additional male reproduction in this cohort,
given that men generally have wives who are younger than
they are. Indeed, the fact that the women in the cohort
currently have more children than the men do suggests
that this will occur. Thus, the data do not preclude the
emergence of income-related multiple marriages in the
next few years, which will strengthen the selection on male
wealth.
Comparative Analysis
The intraspecic comparative analysis found that the
strength of selection on male wealth for contemporary
Britain is comparable to that of Sweden and the United
States, but selection in all three industrial societies appears
to be weaker than for preindustrial European populations
and for two African societies. A potential confounding
factor here is that, in the industrial societies, the measure
is annual income, whereas in the other societies it is ac-
cumulated wealth. Although it is likely that there is a cor-
relation between these two quantities, the former will be
much noisier than the latter because of year-to-year in-
come uctuations. However, we assume that this difference
is not entirely responsible for the lower coefcients in the
industrial societies. Instead, it seems likely that the com-
bination of effective monogamy and reduced variation in
family size weakens selection on male wealth. Thus, there
is some truth in the perception that the evolutionary dy-
namics of modern societies are different frompreindustrial
ones. However, the claim that modern societies represent
a theoretical problem for Darwinian approaches to be-
havior (Vining 1986) is not realized. Men pursue status
and resources in order to convert these into reproductive
success, just as is the case in preindustrial contexts, and
positive phenotypic selection on male wealth is still op-
erative. Our results suggest that selection on male wealth
is culturally and temporally ubiquitous in food-producing
societies and is generally of comparable magnitude to se-
lection on hunting ability found among hunter-gatherers.
Two selection gradients, from the Kipsigis and Muku-
godo, were remarkably strong. These are both resource-
based, polygynous African societies in which most of the
large variance in male reproductive success is explained
by variation in numbers of wives, which in turn is largely
explained by mens holdings of land and cattle (Borgerhoff
Mulder 1987; Cronk 1991). It is therefore not surprising
that the effective selection on male wealth is so much
higher in these groups than in the others, which are mo-
nogamous or more mildly polygynous.
The cross-species comparisons show that, even in con-
temporary industrial societies, the strength of selection on
male wealth falls in the middle of the distribution of se-
lection strengths ever measured by biologists, with the his-
torical European societies above the middle of the distri-
bution and the African polygynous societies in the top
10%. Thus, we conclude that ongoing phenotypic selection
on male wealth is of substantial strength even in the most
developed societies and has historically been rather strong
in comparative terms.
General Implications
This study raises some general evolutionary issues. First,
it shows that modern societies are only quantitatively, not
Natural Selection on Male Wealth in Humans 665
qualitatively, different from preindustrial ones. That is, the
selective gradients on resources, at least for men, are at-
tenuated but not abolished or reversed. This means that
ultimate tness-maximizing explanations of the kind es-
poused by behavioral ecologists may be more relevant to
the dynamics of contemporary societies than is commonly
assumed. The data also suggest considerable continuity
between modern human societies and those of hunter-
gatherers, where selection is on male hunting ability, and
indeed between human societies and other species, where
male rank is generally positively related to reproductive
success (Ellis 1995).
The cross-cultural results reveal that phenotypic selec-
tion on human male wealth is pervasive wherever there
are accumulable resources. The data presented here show
this phenomenon for African herders and farmers, for 300
years of preindustrial English history, and for contem-
porary Britain, Sweden, and the United States. Of course,
phenotypic selection produces evolutionary change only
if the trait has some heritable basis. Information is lacking
about whether genetic variation affects income-generating
propensities across cultures, but estimates using twin and
adoption studies suggest a moderate genetic heritability
within industrial populations (Bowles and Gintis 2002;
Bjorklund et al. 2006). This raises the interesting possibility
that ongoing genetic evolution, as well as phenotypic se-
lection, is related to male wealth in humans.
Acknowledgments
The NCDS is performed by the Centre for Longitudinal
Studies, Institute of Education, London, which makes the
data available to interested researchers. The cross-taxa da-
tabase has kindly been made available by J. G. Kingsolver
and colleagues. We are grateful to G. Clark, M. Fieder, and
R. Hopcroft for their assistance in comparing our results
with theirs and to K. Hill for sharing a prepublication
manuscript with us.
Literature Cited
Bjorklund, A., M. Lindahl, and E. Plug. 2006. The origins of inter-
generational associations: lessons from Swedish adoption data.
Quarterly Journal of Economics 121:9991028.
Borgerhoff Mulder, M. 1987. On cultural and reproductive success:
Kipsigis evidence. American Anthropologist 89:617634.
. 1990. Kipsigis womens preferences for wealthy men: evi-
dence for female choice in mammals. Behavioral Ecology and
Sociobiology 27:255264.
. 1998. The demographic transition: are we any closer to an
evolutionary explanation? Trends in Ecology & Evolution 13:266
270.
Bowles, S., and H. Gintis. 2002. The inheritance of inequality. Journal
of Economic Perspectives 16:330.
Burgess, S., C. Propper, and A. Aassve. 2003. The role of income in
marriage and divorce transitions among young Americans. Journal
of Population Economics 16:455475.
Buss, D. M. 1989. Sex differences in human mate preferences: evo-
lutionary hypotheses tested in 37 cultures. Behavioral and Brain
Sciences 12:149.
Clark, G., and G. Hamilton. 2006. Survival of the richest: the Mal-
thusian method in pre-industrial England. Journal of Economic
History 66:707736.
Cronk, L. 1991. Wealth, status and reproductive success among the
Mukogodo of Kenya. American Anthropologist 93:345360.
Ellis, L. 1995. Dominance and reproductive success among non-
human animals: a cross-species comparison. Ethology and Socio-
biology 16:257333.
Endler, J. A. 1986. Natural selection in the wild. Princeton University
Press, Princeton, NJ.
Fieder, M., and S. Huber. 2007. The effects of sex and childlessness
on the association between status and reproductive output in mod-
ern society. Evolution and Human Behavior 28:392398.
Grant, B. R., and P. R. Grant. 1989. Natural selection in a population
of Darwin nches. American Naturalist 133:377393.
Gurven, M., and C. von Rueden. 2006. Hunting, social status and
biological tness. Biodemography and Social Biology 53:8199.
Hawks, J., E. T. Wang, G. M. Cochran, H. C. Harpending, and R.
K. Moyzis. 2007. Recent acceleration of human adaptive evolution.
Proceedings of the National Academy of Sciences of the USA 104:
2075320758.
Hoekstra, H. E., J. M. Hoekstra, D. Berrigan, S. N. Vignieri, A. Hoang,
C. E. Hill, P. Beerli, et al. 2001. Strength and tempo of directional
selection in the wild. Proceedings of the National Academy of
Sciences of the USA 98:91579160.
Hopcroft, R. L. 2006. Sex, status and reproductive success in the
contemporary United States. Evolution and Human Behavior 27:
104120.
Irons, W. 1979. Cultural and biological success. Pages 257272 in N.
A. Chagnon and W. Irons, eds. Evolutionary biology and human
social behavior: an anthropological perspective. Duxbury, North
Sciutate, MA.
Kanazawa, S. 2003. Can evolutionary psychology explain reproduc-
tive behavior in the contemporary United States? Sociological
Quarterly 44:291302.
Kaplan, H., and K. Hill. 1985. Hunting ability and reproductive
success amongst male Ache foragers. Current Anthropology 26:
131133.
Keizer, R., P. A. Dykstra, and M. D. Jansen. 2007. Pathways into
childlessness: evidence of gendered life-course dynamics. Journal
of Biosocial Science 35:213226.
Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N.
Vignieri, C. E. Hill, A. Hoang, et al. 2001. The strength of phe-
notypic selection in the natural populations. American Naturalist
157:245261.
Marini, M. M. 1984. Womens educational attainment and the timing
of entry into parenthood. American Sociological Review 49:491
511.
Marlowe, F. W. 1999. Showoffs or providers? the parenting effort of
Hadza men. Evolution and Human Behavior 20:391404.
Mealey, L. 1985. The relationship between social and biological suc-
cess: a case study of the Mormon religious hierarchy. Ethology
and Sociobiology 6:249257.
Nakosteen, R. A., and M. A. Zimmer. 1997. Men, money and mar-
666 The American Naturalist
riage: are high earners more prone than low earners to marry?
Social Science Quarterly 78:6682.
Nettle, D. 2003. Intelligence and class mobility in the British pop-
ulation. British Journal of Psychology 94:551561.
Perusse, D. 1993. Cultural and reproductive success in industrial
societies: testing the relationship at the proximate and ultimate
levels. Behavioral and Brain Sciences 16:267322.
Pollet, T. V., and D. Nettle. 2008. Driving a hard bargain: sex ratio
and male marriage success in a historical U.S. population. Biology
Letters 4:3133.
Reed, F. A., and C. F. Aquadro. 2006. Mutation, selection and the
future of human evolution. Trends in Genetics 22:479484.
Roskaft, E., A. Wara, and A. Viken. 1992. Human reproductive suc-
cess in relation to resource access and parental age in a small
Norwegian farming parish during the period 17001900. Ethology
and Sociobiology 13:443461.
Smith, E. A. 2004. Why do good hunters have higher reproductive
success? Human Nature 15:343364.
van den Berghe, P. L., and J. Whitmeyer. 1990. Social class and
reproductive success. International Journal of Contemporary So-
ciology 27:2948.
Vining, D. R. 1986. Social versus reproductive success: the central
theoretical problem of human sociobiology. Behavioral and Brain
Sciences 9:167187.
Voight, B. F., S. Kudaravalli, X. Q. Wen, and J. K. Pritchard. 2006.
A map of recent positive selection in the human genome. PLoS
Biology 4:446458.
Voland, E. 1990. Differential reproductive success within the Krumm-
horn population (Germany, 18th and 19th century). Behavioral
Ecology and Sociobiology 26:6572.
Wang, E. T., G. Kodama, P. Baidi, and R. K. Moyzis. 2006. Global
landscape of recent inferred Darwinian selection for Homo sapiens.
Proceedings of the National Academy of Sciences of the USA 103:
135140.
Weeden, J., M. J. Abrams, M. C. Green, and J. Sabini. 2006. Do high-
status people really have fewer children? education, income, and
fertility in the contemporary U.S. Human Nature 17:377392.
Associate Editor: Edmund D. Brodie III
Editor: Michael C. Whitlock