Intelligence and Anti-Natalist Intentions on Dating Sites: An Analysis of the OKCupid Dataset

Abstract

A body of research indicates that people who are more intelligent tend to have fewer children than do those who are less intelligent, at least since around 1900 (Lynn, 2011). Nyborg (2012) has predicted that the consequent IQ decline will lead to the eventual decay of Western civilization. However, there is little research on fertility intentions and intelligence. Do smarter people end up with fewer children because they ideally desire fewer, or is it due to competing interests, such as a desire for money and status combined with more efficient use of contraception, as Nyborg (2012) observes? We analysed the OKCupid dataset of predominantly Western, English-speaking online users. Employing an ad hoc intelligence test composed of 14 160 Intelligence, Race and Sex questions on the dating service, we find that intelligence does indeed negatively relate to fertility intentions (β = -0.15, ordinal regression), even adjusting for age, sex, and race/ethnicity (β = -0.14). We also replicate the usual pattern of a negative association between intelligence and actual fertility, though the dataset was suboptimal for this analysis as fertility was only a binary outcome.

Full text


CHAPTER 7
INTELLIGENCE AND ANTI-
NATALIST INTENTIONS
ON DATING SITES:
AN ANALYSIS OF THE
OKCUPID DATASET
EMIL KIRKEGAARD AND EDWARD DUTTON
Abstract
A
   indicates that people who are more in-
telligent tend to have fewer children than do those who are less
intelligent, at least since around  (Lynn, ). Nyborg () has
predicted that the consequent IQ decline will lead to the eventual
decay of Western civilization. However, there is little research on fer-
tility intentions and intelligence. Do smarter people end up with fewer
children because they ideally desire fewer, or is it due to competing
interests, such as a desire for money and status combined with more
ecient use of contraception, as Nyborg () observes? We analysed
the OKCupid dataset of predominantly Western, English-speaking
online users. Employing an ad hoc intelligence test composed of 

 IN TELL IGENCE , RACE A ND SEX
questions on the dating service, we nd that intelligence does indeed
negatively relate to fertility intentions (β = -., ordinal regression),
even adjusting for age, sex, and race/ethnicity (β = -.). We also rep-
licate the usual pattern of a negative association between intelligence
and actual fertility, though the dataset was suboptimal for this analysis
as fertility was only a binary outcome.
1. Introduction
ere is a long-standing interest in the relationship between fertil-
ity and human capital traits and specically the relationship between
fertility and intelligence. Various scholars have come to the conclu-
sion that throughout much of history, there was a positive association
between socio-economic status and the number of surviving children,
such that the partly-genetic traits that led to socioeconomic status
were under positive selection (Clark, ; Lynn, ). Still, it has
been found that starting sometime in the s, these associations
started to reverse such that people lower in these traits had more (sur-
viving) children. is was later termed dysgenics as an antonym of eu-
genics. Nyborg () observed that this ‘Double Relaxed Darwinian
Selection’ — the collapse of conditions causing a positive IQ-fertility
nexus and the rise of innovations, such as contraception, which cause
a negative IQ-fertility nexus — will lead to the decay of civilization,
due to the high heritability of IQ, of around . (Lynn, , p.).
Interest in the subject began in earnest with the work and dis-
coveries of Francis Galton, a Victorian polymath and half-cousin of
Charles Darwin (Galton, ; Jensen, ). While there has been
some debate over the measurement, magnitude and consistency of
dysgenic fertility patterns, a modern meta-analysis shows a fairly
consistent pattern (Reeve et al., ). e overall eect size is weak, r
= -.. is varies with sexes; the pattern is almost always observed to
be more negative for women than men. In their meta-analysis, for the
samples with completed fertility and persons of at least  years of age,
the correlations were -. and -., for women and men, respectively.

CHAPTER 7
For men, sometimes in some developed countries, the pattern is ob-
served to be slightly positive, though probably not positive enough to
outweigh the female negative relationship. e best evidence of the
positive associations comes from Nordic register data studies, which
rely on near-complete population datasets from the army induction
testing (Barclay & Kolk, ; Bjerkedal et al., ; Kolk & Barclay,
, ). As women are not forced to participate in Nordic military
service, their intelligence data are rarer and of questionable represen-
tativeness. Women’s fertility is not analyzed in the published studies
of Nordic military data, so we do not know what the associations are.
ere is uncertainty about the causes of these patterns, and their
prevalence in poorer regions of the world. ere is emerging evidence
that dysgenic patterns are already visible outside of the Western
world. One study in Libya found a correlation of -. between the
number of siblings and scores on the Raven’s test (Al-Shahomee &
Lynn, ). Another study found correlations of -. for women and
. for men in the Dominican Republic (Meisenberg et al., ). Of
particular interest was that the correlation between the ideal number
of children and actual number of children was only r = .. us, the
ideal number of children does not seem to have much to do with the
actual number of children, at least for this country and time period.
Furthermore, the correlation between the ideal number of children
with intelligence was very weak, . for men and -. for women. It is
unknown whether these values will generalize to Western countries.
In recent decades, childlessness has gone from an unpopular rarity to
being explicitly embraced. It is conceivable this has altered the corre-
lations with intelligence, if smarter people tend to adopt popular ideas
more or faster than duller people. In this regard, it has been shown
that intelligence is associated with conformity to the dominant set of
societal values. is may be because intelligent people are better able
to understand what these values are, are better able to understand
the future benets of conformity, and have sucient eortful control

 INT ELLIG ENCE, RACE AND SEX
to force themselves to adopt the dominant set of values (Woodley of
Menie & Dunkel, ).
e purpose of this study is to examine the relationship between
the ideal number of children and intelligence in Western samples.
e goal is to see whether a decrease in the ideal number of children
may explain the decrease in the number of children among smarter
people. A prior study, Kanazawa (), has shown a very weak as-
sociation between childhood IQ and wishing to remain childless of
about ., though this did attain signicance. ose who, at age ,
wished to have children had an IQ of , while those who wanted
to be childless had an IQ of approximately . is was based on
Britain’s National Childhood Development Study, which interviewed
people born in  and  at the age of . However, an examina-
tion of the US National Longitudinal Study of Youth, , found no
relationship between fertility intentions and intelligence (Meisenberg
& Kaul, ). We aim to develop these analyses by drawing upon a
new and substantial data set.
2. Method
Data
We used data from the OKCupid dataset, which was published as
part of Kirkegaard and Bjerrekær (). is dataset was collected
by scraping (automated downloading) of information from the then
large dating service OKCupid (https://www.okcupid.com/). In total,
the dataset contains data from , users across , variables.
However, since all users only have partial data, and most users have
only very little data, the eective sample size is oen , to ,.
In demographic terms, the dataset is mostly Western (~) and
English-speaking (~), as well as  male. e age variation is
extensive, skewed towards younger people as would be expected from
a dating service.

CHAP TER 7
e main functionality of the dating service was to allow users to
create and ll out ,s of survey questions. Each of these allowed
for between  to  answer options, and users could also assign them
importance. e site’s algorithm would then use the collected data in
order to compute a match score with each other user, which is use-
ful for dating purposes due to the very strong assortative mating in
humans (Hur, ; Kirkegaard, ; Luo, ). Almost all users
allowed their answers to be visible to other users of the site, which
is what allowed the dataset to be collected. e questions on the site
were mostly user-generated, but site sta had created the rst  or
so questions (out of many ,s). Since most people grow tired of
lling out questions on a dating site, there is not full coverage for users
and questions. e variety of questions is extreme compared to most
surveys since there was apparently little oversight or quality control of
the questions.
Some of the questions are intelligence questions. Prior studies of
these questions have shown that they work as a primitive intelligence
test (Kirkegaard, ; Kirkegaard & Bjerrekær, ; Kirkegaard &
Lasker, ). e questions are shown below:
• Which is bigger, the earth or the sun?
• STALE is to STEAL as  is to what?
• What is next in this series? , , , , , __
• If you turn a le-handed glove inside out, it ts on your le or
right hand?
• In the line ‘’Wherefore art thou Romeo?’’ what does ‘’wherefore’’
mean?
• How many fortnights are in a year?
• Half of all policemen are thieves and half of all policemen are mur-
derers. Does it follow logically that all policemen are criminals?
• Which is longer, a mile or a kilometer?

 INTELL IGENCE , RACE A ND SEX
• When birds stand on power lines and don’t get hurt, it’s most likely
because of what?
• Etymology is?
• If some men are doctors and some doctors are tall, does it follow
that some men are tall?
• A little grade  science: what is the Ideal Gas Law?
• If you ipped three pennies, what would be the odds that they all
came out the same?
• Which is the day before the day aer yesterday?
3. Results
e intelligence data was scored using a single factor model using the
mirt package as done in prior studies. We standardized the resulting
scores to mean of  and standard deviation of . ere are no norms
for these data, so we are unable to convert to the familiar /
scale. e sample is likely somewhat above average in the countries
of collection as indicated by studies of who employs dating sites (e.g.
Nam, ), so likely the mean IQ would be around  with a slightly
reduced standard deviation, around . Due to the fact that most us-
ers did not provide answers to all of the  questions, the scores are
based on the available data. is means the reliability is not as high
as is desired. Using all the available data as is, the estimated empirical
reliability is only ., or . if we assume a normal distribution with
full data. One solution to this is to subset the data by the number of
lled out questions, thus making a quantity-quality trade-o. Figure 
shows the eect of doing this on the test’s estimated reliability.

CHA PTER 7
Figure 1. Estimated test reliability as function of

out at least a given number of intelligence items.
e marginal (theoretical) reliability declines somewhat across sample
restrictions. is may seem odd, but it probably reects the fact that as
sample restrictiveness in terms of intelligence variation declines (be-
cause smarter people ll out more items), the g-loadings are deated,
and thus the reliability of the test is reduced. e empirical estimated
reliability, however, increases with sample restrictiveness, as expected
since each user provides more data.
In terms of fertility, there is no question about the number of chil-
dren subjects currently have. e prole, however, required people to
answer the question “Do you have a child or children?” (yes/no), thus
providing a crude binary measure. ere was a question about desired
fertility: “How many children would you ideally like to have?” with
options being: “None”, “-”, “-”, “ or more!”. Figures  and  show
the distributions of intelligence by responses to these questions.
We restricted our analysis of the dataset to subjects who were het-
erosexual, and who had lled out at least  items in the intelligence
test. is provided a reasonable trade-o between the quality of the

 IN TELL IGENCE , RACE A ND SEX
measure and ensuring sucient sample size to detect eects of con-
cern. is reduced the sample size to ,. Figure  displays intel-
ligence by parenthood status, while Figure  displays the relationship
between intelligence and desired number of children.
Figure 2. Violin plot of intelligence by parenthood status.
Figure 3. Violin plot of intelligence by ideal desired children number.

CHAP TER 7
Due to the large sample size, all dierences are much larger than ex-
pected purely by chance (i.e., very small p values). Table  shows the
eect sizes in terms of Cohen’s d.
Table 1. Intelligence differences between groups by ideal number of

output with pairwise sample sizes can be found in the R notebook.
How many children
would you ideally
like to have?
None   
None
-0.11 [-0.15
-0.076]
-0.24 [-0.29
-0.19]
-0.41 [-0.54
-0.28]
1-2
-0.11 [-0.15
-0.076]
-0.13 [-0.17
-0.081]
-0.30 [-0.42
-0.17]
3-4
-0.24 [-0.29
-0.19]
-0.13 [-0.17
-0.081]
-0.17 [-0.30
-0.042]
5 or more
-0.41 [-0.54
-0.28]
-0.30 [-0.42
-0.17]
-0.17 [-0.30
-0.042]
e eect sizes are fairly sizeable given the crude test employed here
(estimated reliability .). e eect size for the actual fertility (
vs. +) is d = -.. In terms of correlations, the Pearson correlations
are -. and -. (for desired and realized). However, due to the non-
normal nature of the data, these correlations are biased downwards.
e latent correlations (biserial; (Uebersax, )) are -. and -.
(computed using psych package’s mixedCor() function; (Revelle,
)). Even if we subset to subjects aged  or more (n = ,), the
correlations do not change much: -. and -.. e latent correlation
between desired and actual fertility was fairly weak as in prior studies:
Pearson r = ., latent r = ., and latent r among those aged  or more
..
Finally, we sought to examine whether the dierences were due to
obvious demographic confounders, such as age, race, and sex. To ex-
amine this, we t  ordinal logistic regression models to each outcome
with these predictors, as well as the interaction of sex and intelligence.
e models were t using the rms package’s lrm() function (Harrell,

 IN TELL IGENCE , RACE A ND SEX
). Table  shows the results for desired fertility. e appendix
contains the results for actual fertility, which was of less interest here.
Table 2. Regression model results for desired fertility. Base groups are
Whites for race, and men for sex. Non-linearity was modelled using
natural splines. Values in parentheses are standard errors and p values.
Predictor / Model Basic Controls With interaction
Intelligence -0.1478 (0.0146;
<0.0001)
-0.1392 (0.0153,
<0.0001)
-0.141 (0.0174,
<0.0001)
race=Mixed 0.2608 (0.0552,
<0.0001)
-0.4255 (0.7751,
0.583)
race=Asian 0.3863 (0.0814,
<0.0001)
0.2608 (0.0552,
<0.0001)
race=Hispanic / Latin 0.4481 (0.0852,
<0.0001)
0.3863 (0.0814,
<0.0001)
race=Black 0.4817 (0.0873,
<0.0001)
0.448 (0.0852,
<0.0001)
race=Other 0.1669 (0.1078,
0.1218)
0.4818 (0.0873,
<0.0001)
race=Indian 0.0844 (0.1677,
0.6149)
0.1665 (0.1078,
0.1225)
race=Middle Eastern 0.5533 (0.3025,
0.0674)
0.0841 (0.1677,
0.6158)
race=Native American 0.3164 (0.3223,
0.3264)
0.5523 (0.3025,
0.0679)
race=Pacic Islander 0.6169 (0.4374,
0.1584)
0.3153 (0.3224,
0.3281)
sex=Woman -0.4243 (0.775,
0.5841)
0.6162 (0.4374,
0.1589)
sex * intelligence 0.0075 (0.0359,
0.8353)
Age (nonlinear) (nonlinear)
sex * age (nonlinear) (nonlinear)
Pseudo-R2 0.007 0.034 0.034
N16730 15859 15859
Since the model summary is dicult to intuitively understand when
an ordinal outcome, nonlinear terms, and interactions are used, we
plotted the predicted values from the model using the mean/modal

CHA PTE R 7
values for the non-varying variables (marginal eects). Figure  shows
the results.
Figure 4. Model predicted ideal number of children by predictor values.
Here we can see that it is mainly the dierence between the “-” and
“None” group that changes with intelligence. ere is no substantial
change predicted for the “-” group. Of interest also is the distribu-
tion of the ideal number of children by race, controlling for the mea-
sured covariates. Figure  shows this.

 IN TELL IGENCE , RACE A ND SEX
Figure 5. Model predicted ideal number of children
by race holding covariates constant.
e largest dierences appear to be due to dierences in the size of the
“None” group, which is largest for Whites and Indians, and smallest
for Middle Easterners and Pacic Islanders (but very sample size for
that group). ere are concomitant dierences in the “-” groups.
Discussion
All of the associations with intelligence would be considerably stron-
ger if we had a higher quality measure of intelligence. Estimated
reliability was only ., though this is likely an underestimate of the
test-rest reliability. Nevertheless, we can cautiously assert that intel-
ligence is negatively associated with desiring children in the OKCupid
sample. Unlike the NLSY-based study, we nd that this relationship
exists and we nd that it is much stronger than that unearthed by
Kanazawa () employing a British study.
In making sense of our results, it is worth noting that this does
not appear to be the rst time in history that intelligence has been
negatively associated with having or wanting children. By the 

CHAPTE R 7
century BC, Greek writers were noting that upper-class men did not
desire to have children and tended not to have many of them. By the
time of Christ, Roman commentators were observing that upper-class
males and females both did not desire children and, relative to mem-
bers of lower classes, did not have many of them (Dutton & Woodley
of Menie, ). And there are further parallels. Currently, there are
assorted anti-natalist social movements such as ‘Birth Strike’ (Hunt,
 March ) who aver that we should reject procreation in order
to save the planet from the terrible damage done to it by humanity.
In the rst century AD, many Gnostic movements averred that the
world was the Devil’s province and that you should focus on achieving
supreme knowledge, while certainly avoiding bringing children into
this Satanic world (see Gottfried, ). Both of these periods in his-
tory were marked by ‘individualizing moral foundations.’ Graham et
al. () have argued that humans dier on approximately  sets of
moral foundations, reecting the fact that we are group animals but
we also desire to attain status within the group. ere are the ‘Binding
Foundations’ of in-group loyalty, obedience to authority and concern
for sanctity/disgust and there are the individualizing foundations of
‘equality’ and ‘harm avoidance’. Individuals who lack power will co-
vertly play for status by signalling these foundations (Benenson, ).
Graham et al. () further show that we roughly divide between
‘conservatives’ and ‘liberals’, though there is much nuance in between.
Overall, conservatives are about the same on all ve foundations,
whereas liberals are very high on the individualizing foundations but
are very low on the binding foundations. e result is asymmetric em-
pathy between conservatives and liberals. is permits liberals to hi-
jack the culture and push it in an ever-more individualistic direction,
where people play for status by signalling ever more individualistic
values.
In a situation of runaway individualism, this could evolve into
bettering the last individualist by saying that you so committed to
harm avoidance, in this evil world, that you won’t have children. For

 IN TELL IGENCE , RACE A ND SEX
the reasons already discussed, we would expect those who were intel-
ligent to more strongly embrace and then compete for status within
the dominant set of values, helping to explain, in part, the negative
relationship between intelligence and fertility. e dierence between
the American and British data from the same cohort may reect the
fact that America, possibly for partly genetic reasons based on its hav-
ing been partly founded by Puritans, is more religious and thus more
group-oriented (Graham et al., ) than is Britain (see Dutton,
), meaning it took longer in the US for runaway individualism
to get to the point of anti-natalist attitudes. It should be noted, in this
regard, that the heritability of fundamentalism has been found to
be close to . (Bradshaw & Ellison, ). Under harsh Darwinian
conditions, this runaway individualism would be halted by the fact
that such a group would be low in group-orientation and it has been
demonstrated using computer-modelling that groups that are the
highest in positive and negative ethnocentrism — that is binding
values — tend to dominate and displace other groups (e.g. Hammond
& Axelrod, ). However, in a period of weakened selection pres-
sure, such as our low child mortality world or the Roman Warm
Period with its concomitant urbanization and relatively high levels of
development and of living standards, then this runaway individualism
could take place (see Dutton, ; Sarraf et al., ) and it is even
more extreme in our civilization than in Roman civilization.
A second, and related, possibility, proposed by Dutton and van
der Linden (), is that intelligence is associated with being less ‘in-
stinctive’ and more environmentally sensitive. ey argue that this is
because solving-problems involves rising above, and suppressing, in-
stinctive reactions in order to be able to calmly reason. ey maintain
that it follows that you would be better able to accomplish this if you
were attracted to non-instinctive possibilities, as these may be where
new knowledge and a more subtle understanding of reality — and so
the solving of your problem — lie. ey argue that this may explain
why intelligence appears to be associated with various non-instinctive,

CHAPTE R 7
or evolutionarily novel, preferences, such as being nocturnal and even
with atheism. Congruous with an intelligence-sensitivity nexus, some
studies have found that people with high IQ are more environmentally
plastic. For example, in a sample from , twin pairs: ‘individuals
with high IQ show high environmental inuence on IQ into adoles-
cence (resembling younger children), whereas individuals with low
IQ show high heritability of IQ in adolescence (resembling adults),
a pattern consistent with an extended sensitive period for intellectual
development in more-intelligent individuals’ (Brant et al., ). is,
in turn, would mean that, in order to develop adaptively, intelligent
people would be more reliant on being placed, by their group, on
an evolutionarily adaptive road map of life and, in an evolutionary
mismatch, they would be more likely to suer from dysphoria due to
their environmentally sensitive nature. In that we are evolved to be
pack animals, a highly individualistic society would be an example of
a dysphoria-inducing evolutionary mismatch.
In this regard it has been shown that there is a relationship between
Machiavellianism and liberal values, possibly because, as discussed,
liberal values are individualistic (Ok et al., ). Anti-natalism,
like extreme liberalism more generally, has been found to be associ-
ated with Machiavellianism and also with suering from depression.
Indeed, depression mediates (causes) the relationship between anti-
natalism and Machiavellianism (Schönegger, ). So it may be that,
in a pronounced evolutionary mismatch, intelligent people — being
more environmentally sensitive — are more likely to become de-
pressed and thus to not want children. is would be a fascinating
possibility for future research. One problem with this explanation is
that Northern European countries (which are generally rated as the
most individualistic) currently have higher fertility than Southern and
Eastern European countries (ESHRE Capri Workshop Group, ;
Davis & Williamson, ). A possible explanation for this anomaly
may be that not only have these countries lost their religiousness in
recent decades — leading to an evolutionary mismatch — but they

 INT ELLIGENCE, R ACE AND SEX
are relatively poor and cannot fund a lavish welfare system. Welfare-
recipient females are the engine of fertility — only families where both
parents are on welfare have above replacement fertility, specically
those requiring social-worker and police interventions — and these
females have been shown to increase their fertility in response to in-
creased welfare payments. In addition, the fertility of many northern
European countries is substantially driven by immigrants (Perkins,
).
It is also noteworthy that we do not nd any sex dierences in
the relationship between fertility intentions and intelligence. is is
despite the fact that, as we have discussed, dysgenic fertility on intelli-
gence is higher among females than among males. ere are a number
of possible explanations for this. It may be that the pressures on intel-
ligent females to delay child-bearing, such as through lengthy educa-
tion and pursuing a career, mean that many end up, without truly
wishing to, trading fertility for education and careers. is would be
congruous with evidence that intelligence correlates with absorbing
and convincing yourself, through eortful control, of the veracity of
the dominant ideology, in order to attain the relevant social benets,
as already discussed (Woodley of Menie & Dunkel, ). It has been
found that females are more socially conformist than males (Eagly &
Chrvala, ) and, manifestly, their fertility period is more limited.
A related explanation is that proposed by Apostolou () who has
argued that females are strongly adapted to patriarchy (male control
of females, especially of their sexuality), as those who were not were
historically selected out, unable to nd a partner. Accordingly, females
are adapted, to a greater extent than males, to a situation in which
choices are made for them, usually by their parents, and particularly
by their fathers. Le to their own devices, they are, accordingly, more
likely to make choices that are more maladaptive. In this regard, it
has been found that societal individualism is correlated with gender
equality (Davis & Williamson, ).

CHA PTER 7
Supplementary Materials
e data are available in the supplementary materials of the original
study (Kirkegaard & Bjerrekær, ). e full code and other materi-
als are available at https://osf.io/twcf/. e R notebook can also be
found at https://rpubs.com/EmilOWK/.
References
Al-Shahomee, A. A. & Lynn, R. (). Intelligence and Family Size in Libya.
Mankind Quarterly, : .
Apostolou, M. (). Sexual Selection Under Parental Choice: e Evolution of
Human Mating Behaviour. Hove: Psychology Press.
Barclay, K. & Kolk, M. (). e Inuence of Health in Early Adulthood on Male
Fertility. Population and Development Review, 46: -.
Benenson, J.F. (). e development of human female competition: al-
lies and adversaries. Philosophical Transactions of the Royal Society B, :

Bjerkedal, T., Kristensen, P., Skjeret, G.A. & Brevik, J.I. (). Intelligence test
scores and birth order among young Norwegian men (conscripts) analyzed
within and between families. Intelligence, : -.
Bradshaw, M. & C. Ellison. (). Do genetic factors inuence religious life?
Findings from a behavior genetic analysis of twin siblings. Journal for the
Scientic Study of Religion, : -.
Brant, A., Munakata, Y., Boomsma, D. et al. (). e nature and nurture of high
IQ: an extended sensitive period for intellectual development. Psychological
Science, : -.
Clark, G. (). A Farewell to Alms: A Brief Economic History of the World.
Princeton, NJ: Princeton University Press.
Davis, L. & Williamson, C. (). Does individualism promote gender equality?
World Development, : .
Dutton, E. (). Religion and Intelligence: An Evolutionary Analysis. London:
Ulster Institute for Social Research.
Dutton, E. (). Witches, Feminism, and the Fall of the West. Whitesh, MT:
Washington Summit.
Dutton, E. & Woodley of Menie, M. A. (). At Our Wits’ End: Why We’re
Becoming Less Intelligent and What It Means for the Future. Exeter: Imprint
Academic.

 IN TELL IGENCE , RACE A ND SEX
Dutton, E. & Van der Linden, D. (). Why is Intelligence Negatively Associated
with Religiousness?Evolutionary Psychological Science,: -.
Eagly, A. & Chrvala, C. (). Sex Dierences in Conformity: Status and Gender
Role Interpretations. Psychology of Women Quarterly, : -.
ESHRE Capri Workshop Group. (). Europe the continent with the lowest fertil-
ity. Human Reproduction Update, : -.
Galton, F. (). Hereditary Genius: Its Laws and Consequences. London:
Macmillan and Company.
Gottfried, P. (). Multiculturalism and the Politics of Guilt: Towards a Secular
eocracy. Columbia, MO: University of Missouri Press.
Graham, J., Haidt, J. & Nosek, B. (). Liberals and Conservatives Rely on
Dierent Sets of Moral Foundations. Personality Processes and Individual
Dierences, : -.
Hammond, R. & Axelrod, R. (). e Evolution of Ethnocentric Behavior.
Journal of Conict Resolution, : -.
Harrell, F. E. (). rms: Regression Modeling Strategies (.-.) [Computer so-
ware]. https://CRAN.R-project.org/package=rms
Hunt, E. ( March ). Birth strikers: meet the women who refuse to have
children until climate change ends. e Guardian, https://www.theguardian.
com/lifeandstyle/2019/mar/12/birthstrikers-meet-the-women-who-refuse-to-have-
children-until-climate-change-ends
Hur, Y.-M. (). Assortative Mating for Personality Traits, Educational Level,
Religious Aliation, Height, Weight, and Body Mass Index in Parents of a
Korean Twin Sample. Twin Research and Human Genetics, 6: 467-470.
Jensen, A. R. (). Galton’s Legacy to Research on Intelligence. Journal of
Biosocial Science, : -.
Kanazawa, S. (). Intelligence and Childlessness. Social Science Research, :
-.
Kirkegaard, E. O. W. (). Self-reported criminal and anti-social behavior on a
dating site: e importance of cognitive ability. Open Dierential Psychology,
1(). https://openpsych.net/paper/
Kirkegaard, E. O. W. (). Are there Complex Assortative Mating Patterns for
Humans? Analysis of  Spanish Couples. Mankind Quarterly, : -.
Kirkegaard, E. O. W. & Bjerrekær, J. D. (). e OKCupid dataset: A very large
public dataset of dating site users. Open Dierential Psychology. https://open-
psych.net/paper/

CH APT ER 7
Kirkegaard, E. O. W. & Lasker, J. (). Intelligence and Religiosity among Dating
Site Users. Psych, : -.
Kolk, M. & Barclay, K. (). Cognitive ability and fertility among Swedish men
born –: Evidence from military conscription registers. Proceedings
of the Royal Society B: Biological Sciences, 286(), . https://doi.
org/./rspb..
Kolk, M., & Barclay, K. (). Do income and marriage mediate the relationship
between cognitive ability and fertility? Data from Swedish taxation and con-
scriptions registers for men born –. Intelligence, : .
Luo, S. (). Assortative mating and couple similarity: Patterns, mechanisms, and
consequences. Social and Personality Psychology Compass, (), e. https://
doi.org/./spc.
Lynn, R. (). Dysgenics: Genetic Deterioration in Modern Populations. London:
Ulster Institute for Social Research.
Meisenberg, G. & Kaul, A. (). Eects of Sex, Race, Ethnicity and Marital Status
on the Relationship between Intelligence and Fertility. Mankind Quarterly, :
-.
Meisenberg, G., Lawless, E., Lambert, E., & Newton, A. (). e Social Ecology
of Intelligence on a Caribbean Island. Mankind Quarterly, : –.
Nam, T. (). Who is dating and gaming online? Categorizing, proling, and pre-
dicting online daters and gamers. Computers in Human Behavior, : -.
Nyborg, H. (). e decay of Western civilization: Double relaxed Darwinian
Selection. Personality and Individual Dierences, : -.
Ok, E., Qian, Y, Strejcek, B. & Aquino, K. (). Signaling Virtuous Victimhood
as Indicators of Dark Triad Personalities. Journal of Personality and Social
Psychology, https://doi.org/./pspp
Perkins, A. (). e Welfare Trait: How State Benets Aect Personality.
Basingstoke: Palgrave Macmillan.
Reeve, C. L., Heeney, M. D., & Woodley of Menie, M. A. (). A systematic review
of the state of literature relating parental general cognitive ability and number
of ospring. Personality and Individual Dierences, 134, –. https://doi.
org/./j.paid...
Revelle, W. (). psych: Procedures for Psychological, Psychometric, and
Personality Research (...) [Computer soware]. https://CRAN.R-project.
org/package=psych
Sarraf, M. Woodley of Menie, M.A. & Feltham, C. (). Modernity and Cultural
Decline: A Biobehavioral Perspective. Basingstoke: Palgrave Macmillan.

 IN TELL IGENCE , RACE AND SEX
Schönegger, P. (). What’s up with anti-natalists? An observational study on
the relationship between dark triad personality traits and anti-natalist views.
Philosophical Psychology, doi: ./..
Uebersax, J. S. (). Introduction to the Tetrachoric and Polychoric Correlation
Coecients. http://john-uebersax.com/stat/tetra.htm
Woodley of Menie, M.A. & Dunkel, C. (). Beyond the Cultural Mediation
Hypothesis: A reply to Dutton (). Intelligence, : -.
Appendix
Model results for realized fertility.
Table S1. Regression model results for realized fertility. Base groups
are Whites for race, and men for sex. Nonlinearity was modelled using
natural splines. Values in parentheses are standard errors and p values.
Predictor / Model Basic Controls With interaction
Intelligence -0.3551 (0.0170,
<0.0001)
-0.3589 (0.0193,
<0.0001)
-0.3484 (0.0219,
<0.0001)
race=Mixed -0.0335 (0.0707,
0.6353)
-0.0328 (0.0707, 0.6423)
race=Asian -1.1707 (0.1538,
<0.0001)
-1.1714 (0.1539,
<0.0001)
race=Hispanic / Latin -0.0273 (0.1139,
0.8106)
-0.0268 (0.1139, 0.8140)
race=Black 0.2118 (0.1093,
0.0526)
0.2123 (0.1093, 0.0521)
race=Other -0.0709 (0.1343,
0.5977)
-0.0684 (0.1343, 0.6103)
race=Indian -1.1109 (0.3402,
0.0011)
-1.1102 (0.3401, 0.0011)
race=Middle Eastern -0.8829 (0.4524,
0.0510)
-0.8778 (0.4523, 0.0523)
race=Native
American
-0.5326 (0.4320,
0.2176)
-0.5250 (0.4319, 0.2241)
race=Pacic Islander -0.8568 (0.6192,
0.1664)
-0.8502 (0.6189, 0.1695)
sex=Woman 3.4758 (1.8247,
0.0568)
3.4401 (1.8259, 0.0596)
sex * intelligence -0.0462 (0.0457, 0.3124)
Age (nonlinear) (nonlinear)
sex * age (nonlinear) (nonlinear)
Pseudo-R2 0.038 0.245 0.245
N17554 16637 16637