Intelligence 48 (2015) 137151

Contents lists available at ScienceDirect

Intelligence

Sex differences in academic achievement are not related to

political, economic, or social equality
Gijsbert Stoet a,, David C. Geary b
a
School of Education, University of Glasgow, 11 Eldon Street, G3 6NH, Glasgow, UK
b
Department of Psychological Sciences, University of Missouri, Columbia, MO, USA

a r t i c l e i n f o a b s t r a c t

Article history: The national differences in gender equality in economic and political participation have garnered
Received 2 December 2013 considerable attention as an explanation of boys' better achievement in mathematics in most
Received in revised form 22 August 2014 countries, but the debate is far from resolved. Using data from four international assessments of
Accepted 11 November 2014 the academic achievement of 1.5 million 15 year olds (Programme for International Student
Available online 17 December 2014
Assessment, PISA), we demonstrate that the relation between sex differences in PISA achievement
and national measures of gender equality is not consistent across assessments, and several of the
Keywords: positive findings are confounded by outliers. Further, for overall achievement across reading,
Sex differences mathematics, and science literacy girls outperformed boys in 70% of participating countries,
Scholastic performance
including many with considerable gaps in economic and political equality, and they fell behind in
Reading
only 4% of countries. The results raise doubts about the relation between national equality policies
Mathematics
Science and mathematics achievement, and raise broader questions regarding women's underrepresen-
tation in political, economic, and academic leadership despite stronger academic skills and
regarding the long-term economic prospects and social stability of nations with many men who
are not competitive in the modern economy.
2014 Elsevier Inc. All rights reserved.

1 . Introduction although a recent meta-analysis of sex differences in school

The discussion of sex differences in educational attitudes and
achievement dates back hundreds of years. For example, the
British philosopher John Locke (1693) mentioned the relative
ease with which girls are able to learn a second language in his
treatise Some thoughts concerning education. The systematic
study of sex differences in educational achievement started in
the first half of the 20th century (Woolley, 1914). Overall, for the
past 100 years girls have been found to perform better than
boys in academic areas that involve language skills, including
reading and spelling (Burt & Moore, 1912; Halpern, 2012), and
boys in some mathematical areas, especially in adolescence and
adulthood (Geary, 1996; Stockard & Bell, 1916). The stability
and magnitude of these differences are vigorously debated,
Corresponding author.
E-mail address: Gijsbert.Stoet@glasgow.ac.uk (G. Stoet).
grades from 19142011 concluded that the gap remained stable
(Voyer & Voyer, 2014). This debate directly relates to a core
question in the field of differential psychology and cognitive
abilities, namely the degree to which sex differences in cognitive
abilities are related to environmental factors. Some researchers
have argued that these sex differences have been largely stable
over the decades (Ellis et al., 2008), while others have argued
they are disappearing (Feingold, 1988), or have essentially
disappeared (Hyde, 2005; Hyde & Linn, 1988; Hyde & Mertz,
2009).
The purported disappearance of these sex differences,
especially in mathematics, has been attributed to historical
changes in social roles and movement toward gender parity in
economic and political influence. This process is predicted by
the gender similarities and gender stratification hypotheses,
and some researchers have concluded that there is a linear
relation between the size of the sex difference in mathematics

http://dx.doi.org/10.1016/j.intell.2014.11.006
0160-2896/ 2014 Elsevier Inc. All rights reserved.
138 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
performance and the extent to which men and women have
equal social, economic and political opportunities (Else-Quest,
Hyde, & Linn, 2010; Guiso, Monte, Sapienza, & Zingales,
2008; Hyde & Mertz, 2009; Reilly, 2012). We use four PISA
assessments and national gender equality indices developed by
the United Nations and the World Economic Forum to provide
rigorous tests of these hypotheses, and find that they are not
consistently supported.
The concept of gender stratification has its origin in sociology
(Collins, Chafetz, Blumberg, Coltrane, & Turner, 1993), referring
to the differentiation between men and women based on
income and power. Hyde's gender similarities hypothesis about
gender differences in psychological variables posits that males
and females are more alike than different (Hyde, 2005; Hyde,
Lindberg, Linn, Ellis, & Williams, 2008). On the basis of a meta-
analysis of some constructs, she concludes that there are only
few psychological variables on which men and women differ,
such as some motor and sexual behaviors (Hyde, 2005). One of
Hyde's main points is that the focus on sex differences results in
an underestimation of girls' potential in mathematics.
The gist of the stratification hypothesis is that sex dif-
ferences in economic and educational opportunities result in
sex differences in mathematics performance (Else-Quest et al.,
2010; Guiso et al., 2008). In short, this theoretical model
proposes a causal link between gender equality and education-
al outcomes. The gender similarities and gender stratification
hypotheses form a coherent theoretical model, which essen-
tially states that nearly all psychological sex differences are
the result of social and political factors. It thus follows that
implementing policies that create equality of opportunity will
ultimately eliminate the stratification of society by gender, and
so eliminate any sex differences in achievement. It should be
noted, though, that this theoretical model has been criticized
for overlooking many psychological constructs that are quite
large (e.g., occupational interests, Su, Rounds, & Armstrong,
2009), for creating the impression that psychological sex
differences are categorical, that is, an all or none issue (Lippa,
2006), and for being based on inadequate methodology (Del
Giudice, Booth, & Irwing, 2012). Nevertheless, the gender
similarities hypothesis has received and continues to receive
much attention in both academia and popular science and it is
therefore important to subject the associated predictions to
rigorous empirical tests. We specifically address the claim that
indicators of social, political, and economic equality are related
to the sex difference in mathematics achievement, and draw
the attention of the field to a wider problem, that of boys'
overall underachievement throughout most of the world.
The PISA is well suited for such an assessment, as it not only
measures achievement in mathematics, reading, and scientific
literacy in 15-year olds around the world, but also a consider-
able number of demographic and socioeconomic variables. The
first PISA assessment was conducted in 2000, and repeated
every three years since then, with an increasing number of
countries, participating schools, and students across assess-
ments; in the 2009 assessment, 515,958 students in 18,641
schools in 74 countries and economic regions participated (see
Methods and Supplementary online material [SOM] for details).
Performance on the PISA is also linked to national differences
in economic output, although it is unclear to what degree
the national differences on the PISA (across the mathematics,
reading, and science literacy scales) reflect general intelligence,
variation in educational systems, or most likely some combina-
tion (Hunt & Wittmann, 2008; Lynn & Mikk, 2009; Rindermann,
2008). In any event, PISA scores matter; for example, it has been
argued that an increase of 25 PISA points (in mathematics and
science) in the next twenty years would raise the GDP by 115
trillion US dollars across OECD countries (OECD, 2006, p.27). In
short, no other international instrument matches the PISA in
terms of assessed breadth of educational achievement and
related factors or in terms of assessed human capital and the
economic well-being of countries and individuals. The one
drawback is the narrow time frame of the PISA, which limits its
sensitivity to long-term secular changes in achievement and
factors that influence any such changes.
The results from previous comparisons of boys' and girls'
performance in PISA assessments, and especially how differ-
ences in performance might be related to differences in gender
equality measures (below) are confusing to say the least. Some
studies, published in highly visible journals, conclude that there
is a link between sex differences on the PISA and national
gender equality policies (Else-Quest et al., 2010; Guiso et al.,
2008; Hyde & Mertz, 2009; Reilly, 2012), whereas others do not
find such a link (Fryer & Levitt, 2010; Kane & Mertz, 2012; Stoet
& Geary, 2013).
Our approach to resolving the contradictory conclusions is
to analyze all of the PISA assessments between 2000 and 2010.
If the gender similarities and gender stratification hypotheses
are correct, countries with higher levels of gender equality have
smaller sex differences in educational achievement in mathe-
matics as well as reading and science literacy. Previous tests of
the hypotheses have focused primarily on mathematics and
thus we do as well. Further, we argue that the focus is often on
girls and mathematics, while boys' performance gets less
attention, despite boys falling behind in reading around the
world (Stoet & Geary, 2013).
We will show that when overall scholastic achievement
is calculated from the core competencies in mathematics,
reading, and scientific literacy, boys fell behind in the majority of
countries. This is not only relevant for the gender stratification
hypothesis, because it raises the question of why despite
educational opportunities and success women are under-
represented in leadership positions in politics, business, and
academia. We will present support for two factors that might
explain this. First of all, we show that the sex difference in
educational achievement depends on the overall level of
achievement, with boys at the highest levels doing equally
well or better than girls at the highest levels. The second factor
is relative academic strength. Students' relative academic
strengths and weaknesses are critical to understanding sex
differences in later schooling and occupational outcomes
(Humphreys, Lubinski, & Yao, 1993). Students who are relatively
better at language related competencies than mathematics are
more likely to choose humanities majors in college, whereas
students who are relatively better at mathematics than language
related competencies are more likely to choose physical
science, technology, engineering, and mathematics majors
(Park, Lubinski, & Benbow, 2007). These differences in turn
contribute to later occupational choices and success within
these occupations (Park et al., 2007). In other words, when it
comes to choices made by individual students, mean sex
differences in mathematics or reading may not be as important
as whether they are, as individuals, better at mathematics than
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
verbal competencies, including reading, or vice versa (Valla &
Ceci, 2014; Wai, Lubinski, & Benbow, 2005; Wang, Eccles, &
Kenny, 2013). We acknowledge there are other important
factors influencing women's underrepresentation in leadership
positions that are beyond the current research project,
especially factors that affect later career development, such as
the pipe-line issues (Jacobs & Simpkins, 2005; Jeffcoat, 2004;
Tucker, Hanuscin, & Bearnes, 2008). Further, there are a
number of relevant sex differences in cognitive (Hedges &
Nowell, 1995; Linn & Petersen, 1985; Voyer, Voyer, & Bryden,
1995) and personality (Del Giudice et al., 2012; Irwing, Booth,
& Batey, 2014; Lippa, Collaer, & Peters, 2010; Schmidt, 2011;
Su et al., 2009) traits (including interest differences) that
influence sex differences in career choices and are beyond the
scope of our data.
2 . Methods
The current study combines the analysis of data from the
PISA and equality data from a number of reliable international
institutions, such as the United Nations. We will describe both
types of data separately.
2.1. PISA data
The PISA data were collected in four international surveys
carried out by the Organisation for Economic Co-operation
and Development (OECD). The main aim of the surveys is to
evaluate how well 15-year old students (defined as between
15 years and 3 months and 16 years and 2 months) perform at
the end of compulsory schooling (which in most countries is
15 years) in topics essential for employment and day-to-day
functioning in modern economies, in particular reading,
mathematics, and science (for an indepth descriptions of the
PISA assessment framework, see OECD, 2003a, 2010). Data of
the four surveys were collected in 2000, 2003, 2006, and the
period 20092010. The PISA sampling and quality monitoring is
documented in high detail; documents can be freely downloaded
from the PISA website: http://www.oecd.org/pisa/
Since the first PISA survey in 2000, between 41 and 74
different countries and economic regions have participated
(Table 1 lists countries for each PISA and OECD membership).
The participating regions are parts of countries that did not
participate as whole nations. For example, three economic
regions in China participated: Shanghai, Hong Kong, and
Macao. It should be noted that the total population of Shanghai
(over 20 million) is larger than that of many participating
whole nations, and some whole nations have fewer than one
million inhabitants (e.g., Iceland). Similarly, in 2009 the two
Indian states Himachal Pradesh and Tamil Nadu participated as
economic regions while India as a whole did not participate.
In each participating country or region, a representative
sample of schools and then a representative sample of parti-
cipating students of each school are determined. The PISA
consortium determines which schools participate. Schools then
provide lists of students that are eligible to participate, from
which national project managers select students according to
standardized procedures (OECD, 2012, p.58). The OECD reports
that on average between 4500 and 10,000 students per country
are tested (OECD, 2003a).
139
From each survey, we have the data of each participating
student. Those data include the standardized achievement
scores on reading, mathematics, and science (and depending
on survey year some additional basic skills, such as problem
solving, or specific subtypes of mathematics, reading, or
scientific literacy). From each student, we have also detailed
information about background, including a number of socio-
economic variables (including a standardized composite
measure of socioeconomic status) and attitudes to learning.
Altogether, we have hundreds of variables per participating
student, including sample weighting variables.
Not every participating student completes all the test
material. Instead, different students complete different test
items such that each of them spends 2 h on the assessment.
The core PISA achievement (mathematics, reading, and
science) levels are estimated using a Rasch model (which is
commonly used in large educational surveys). For each of these
three domains, we have five plausible values and 80 replicate
weights. The values are based on the specific questions each
student answered. The final scales for mathematics, reading,
and science are transformed to have a mean value of 500 PISA
points in OECD countries, with a standard deviation of 100
points. Each analysis of the core domains was carried out
separately on each of the five plausible values and then
averaged (exactly in accordance with the PISA guidelines for
statistical analysis, OECD, 2003b). Country averages on the
relevant values used in the analyses of this study are listed in
the tables of the supplementary online material (SOM).
2.2 . Gender equality data
The PISA does not include national gender equality data but
such data are available from a number of other international
organizations, such as the United Nations (United Nations
Development Programme, 2011) and the World Economic
Forum (Hausmann, Tyson, & Zahidi, 2010). We used two
different types of national data reflecting gender equality from
these sources. These included specific indicators of gender
equality, such as the gender ratio in parliament or sex differences
in longevity and income, as well composite indicators that
combine a number of specific indicators. Among the latter, are
the frequently used Global Gender Gap Index (GGI) published
annually by the World Economic Forum in the Global Gender
Gap Report (Hausmann et al., 2010), and the Gender
Empowerment Measure (GEM), reported annually in the
United Nations' Human Development Report (United
Nations Development Programme, 2011). The GGI is con-
sidered to be an improvement over older composite
indicators such as the GEM, in part because it is more
comprehensive, and in part due to the way it corrects for
confounding factors, such as nations' affluence (Zentner &
Mitura, 2012).
The GGI measure was first introduced in 2006, and yet has
been used in the analyses of the 2003 PISA data (Else-Quest
et al., 2010; Guiso et al., 2008; Hyde & Mertz, 2009). Because
the purpose of this study is, in part, to evaluate the conclusions
drawn in those articles, we have also correlated GGI with PISA
data before 2006. Although this is not ideal, we believe it is
reasonable given the scores on these measure do not change
much from year to year. For example, for countries participating
140 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151

Table 1 Table 1 (continued)

List of countries participating in the PISA.
Country 2000 2003 2006 2009
Country 2000 2003 2006 2009
Singapore *
OECD countries Slovenia * *
Australia * * * * Tamil Nadu (India) *
Austria * * * * Thailand * * * *
Belgium * * * * Trinidad and Tobago *
Canada * * * * Tunisia * * *
Chile * * * United Arab Emirates *
Czech Republic * * * * Uruguay * * *
Denmark * * * *
Estonia * *
Finland * * * * in the 2006 or 2009 PISA, the 2006 GGI and the 2009 GGI are
France * * * *
strongly correlated, r(47) = .94, p b .01.
Germany * * * *
Greece * * * * Further, it is important to note that even though GEM and
Hungary * * * * GGI are different measures, there is a relatively strong linear
Iceland * * * * correlation between them. For the countries that participated
Ireland * * * *
in the 2006 PISA for which a GEM and GGI score was available,
Israel * * *
Italy * * * * the correlation was r(41) = .83, p b .001. For the countries that
Japan * * * * participated in 2009, the correlation was r(58) = .75, p b .001.
Luxembourg * * * * The use of composite gender equality indicators is relatively
Mexico * * * * novel and not without limitations (Klasen, 2006). Despite this,
New Zealand * * * *
the advantage of composite indicators over specific indicators
Norway * * * *
Poland * * * * is that they are the result of a careful and targeted approach to
Portugal * * * * measuring gender equality, whereas the specific indicators are
Slovak Republic * * * (necessarily) less reliable, and less likely to give a broad picture
South Korea * * * *
of national gender equality.
Spain * * * *
Sweden * * * *
The authors of one of the studies in which a link between
Switzerland * * * * gender equality measures and the mathematics gender gap
The Netherlands * * * * was found on the 2003 PISA stated: The gender differences on
Turkey * * * the PISA math composite showed some association with
United Kingdom * * * *
composite indices of gender equity; the GEM and GGI were
United States * * * *
non-OECD countries the best composite predictors of the gender gap in math
Albania * * achievement (Else-Quest et al., 2010, p.122). However, the
Argentina * * * same authors argued that more specific indicators of gender
Azerbaijan * *
equality, namely women's share of research positions, parlia-
Brazil * * * *
Bulgaria * * *
mentary seats, and the female economic activity rates were also
Chinese Taipei * * directly related to the gender gap in mathematics PISA1: when
Colombia * * women participate at the rate of men in the labor market
Costa Rica * (particularly in science jobs) and in national government, the
Croatia * *
gender gap in math achievement on the PISA is smaller.
Georgia *
Himachal Pradesh (India) * Therefore, we not only tested GEM and GGI, but also these
Hong Kong * * * three specific variables for all four PISA surveys.
Indonesia * * * * For women's share in research positions, we combined the
Jordan * *
data from the United Nations Educational, Scientific and
Kazakhstan *
Kyrgyzstan * *
Cultural Organization (UNESCO) (available through http://
Latvia * * * * stats.uis.unesco.org) and the OECD (available through http://
Liechtenstein * * * * stats.oecd.org/). We used both data sets, because they partially
Lithuania * * provide data for different countries. When the data were
Macao * * *
available in only one of the two sources, we used that
Macedonia *
Malaysia * information, and when information was available from both
Malta * data sets, we averaged them. Further, these data were not
Mauritius * always available for each year, and in those cases we used data
Miranda (Venezuela) * from the closest year available. If a preceding and succeeding
Moldova *
Montenegro * *
year was available we used the average.
Panama * The specific indicator of women's economic activity was
Peru * * taken from the Human Development Reports for the years
Qatar * *
Romania * * *
Russia * * * *
Serbia * * * 1
Reilly (2012) makes similar claims for the 2009 PISA data, but his data
Shanghai * analysis is confounded by the way the achievement gap in high achieving
children is calculated; see SOM for a detailed analysis.
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
2000 (United Nations Development Programme, 2002), 2003
(United Nations Development Programme, 2005),2 2005
(United Nations Development Programme, 2007) (the data of
2006 are not available), and 2009 (United Nations Development
Programme, 2011). Similarly, the seats of women in parliament
were taken from the Human Development Reports.
It should be pointed out that each Human Development
Report states explicitly that comparisons of labor statistics
across countries should be made with caution. Also, there is
the more general question of whether economically active
reflects increased opportunities and fairness in society. Argu-
ably, economic activity can also mean that some women have
to work for the mere survival of their family. Further, the
measures of the percentage of women in research are not as
complete as the composite indicators. As noted, we had to
combine some data sources, but do not have information on the
percentage of women researchers in some large countries,
including Australia, Brazil, Canada, Russia and the United
States. It is for these reasons that composite gender equality
indices have been created, and we feel that they are therefore
more suitable than these specific indices, although we report
the results for both composite and specific indices.
2.3 . Data analysis
We use Pearson correlations to analyze relations between
sex differences on the PISA and national equality scores (with
an criterion of 5%). In our analyses, we only include the data
of whole countries, and not for the few economic regions
(e.g., Hong Kong), because we do not know if the equality data
available for whole countries are representative for the selected
economic region it seems reasonable to assume that
especially the economic, social, and educational development
of urban areas such as Hong Kong and Shanghai is different
from that of other areas of China, in particular rural areas (and
the same is true for other economic regions in PISA).
Further, for the calculation of a correlation between GGI and
the mathematics gap, we use two different methods, following
a previous study (Guiso et al., 2008). Here, correlations were
based on data using children from families whose socioeco-
nomic status was higher than the median of their country. They
argued in their supplementary online material that there is the
potential that boys and girls drop out of school at different
rates, and that this is potentially more likely for children below
the median on socioeconomic status. However, they also stated
that their findings are robust when including all children. In any
case, we have reported both values for the GGI and mathematics
gap correlations. For outlier analysis in all correlations, we used
Cook's distance (Cook, 1977). We determined a data point as an
outlier when it was higher than the commonly used cutoff point
of 4/(n k 1).
3 . Results
We start with an analysis of the consistency of the relation
between gender equality measures on the one hand and the
2
Else-Quest and colleagues used the data of 2001 from HDR report 2003 to
correlate it with the 2003 PISA data (United Nations Development Programme,
2003).
141
sex differences in mathematics achievement on the other hand.
We then report on overall sex differences in achievement, and
end with an analysis of how overall academic achievement
levels and intra-individual differences in relative strengths in
reading or mathematics can lead to insights into the observed
sex differences in this study and in career choices.
The first part of our analysis challenges the conclusion that
sex differences in mathematics achievement across countries
are related to varying levels of gender equality. We start by
showing that the correlation between the level of gender
equality and the mathematics gap is not consistently found
across the four PISA assessments (20002009). Next, we show
that the observed correlations are very sensitive to the data of a
few countries that can be considered as outliers.
The relation between the GGI and the mathematics gap is
shown in Fig. 1. For each PISA assessment, we calculated the
correlation twice; first using the method of Guiso et al. (2008),
which removed school children from the lower half of each
country's socioeconomic status (see Methods), and second
including all participating school children.
Of the six correlations, the mathematics gap and gender
equality measures were only significant for the 2003 PISA,
r(35) = .41, p = .012, irrespective of whether the data of all
students are included or not (with all data: r[35] = .44,
p b .01). Although the GGI measures were first reported for the
year 2006, as noted above, we correlated the 2006 GGI with the
2000 PISA mathematics gap as well, and found no relation,
r(39) = .11, p = .49. Thus, the prediction that higher levels of
gender equality lead to a smaller sex differences in mathemat-
ics achievement was supported by the correlation between
2006 GGI and 2003 PISA data, but not supported for the 2000,
2006, or 2009 PISA data.
We carried out the same analyses using the main alternative
of GGI, namely the GEM, which is available for all four years the
PISA surveys were conducted (see Methods for explanation of
differences between GEM and GGI). As with the GGI, there was
a significant correlation between GEM and the mathematics
gap in the 2003 PISA data set, with higher levels of gender
equality associated with smaller sex differences in mathematics
achievement, r(31) = .367, p = .036. However, this relation
was not found for the 2000 PISA, r(30) = .005, p = .977, or the
2006 PISA, r(41) = .027, p = .863. In the 2009 PISA, we found
a correlation in the opposite direction, r(60) = .26, p = .043:
higher levels of gender equality were associated with larger sex
differences in mathematics achievement. Thus again, the
analysis shows that the prediction of greater gender equality
leading to smaller sex differences in mathematics achievement
was only supported for the 2003 PISA.
As explained in the Methods, there are three specific
indicators that have been described as being related to the
mathematics gap, and we tested these for all four PISA surveys
as well. We found that women's share of research positions
correlated negatively with the sex differences in mathematics
achievement, although not reliably so across all four PISA
assessments (Fig. 2). Note that a negative correlation means
that the larger the percentage of adult women in research
positions, the smaller the sex difference in mathematics
performance among school children. We observed this effect
for the 2000 PISA, r(31) = .68, p b .001, 2003 PISA, r(29) =
.35, p = .058, and 2006 PISA, r(40) = .29, p = .07 (trend),
but not for 2009 PISA data, r(46) = .16, p = .28.
142 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
The second specific indicator is female economic activity.
Here, there was a clear effect for the 2000 PISA, r(39) = .352,
p = .024 and 2003 PISA r(35) = .377, p = .021. These positive
correlations mean that the more adult women were partici-
pating in the economy, the smaller the sex difference among
school children. However, this was not found for the 2006 PISA,
r(49) = .046, p = .751, or for the 2009 PISA, r(62) = .002,
p = .986. This latter finding is also true using just the countries
that participated in 2000 and 2003 (for 2006: r[40] = 0.282,
p = 0.07; for 2009: r[42] = 0.089, p = 0.565). Thus, the
indicator female economic activity does not consistently relate
to the mathematics gap, with a clear distinction between
20002003 PISA and 20062009 PISA.
The final specific indicator is the percentage of women in
parliament. When applied to all available data, the correlation
between this indicator and the mathematics gap never reached
statistical significance, with correlations ranging between
r(37) = .233, p = .154 and r(64) = .181, p = .146. However,
for the 2003 data, the expected negative correlation of a smaller
mathematics gap with more women in parliament was found
when the non-OECD countries were excluded, r(28) = .398,
p = .030 (Fig. 3).
Thus far, we have shown that the relation between
composite and specific indicators and the mathematics gap
was never found in more than two of the four PISA surveys, and
even when found the pattern was not always the same. We
now address one of the reasons why this relation was
sometimes found and sometimes not.
A visual inspection of these correlational analyses suggests
that grouping and outliers complicate the interpretation of
statistically significant correlations. Before, we concluded that
the GGI correlated with the mathematics gap in 2003 (but not
other years). But even in 2003, this conclusion seems to be
strongly influenced by the data from the Nordic countries,
Iceland, Finland, Sweden, and Norway (Fig. 4). Iceland was a
clear outlier in terms of Cook's distance. For this dataset, a Cook's
distance value over 0.12 can be considered an outlier, and
Iceland's value of 0.38 was the only data point well over that
value. If the same outlier analysis is applied to only half the PISA
data, as in the paper by Guiso et al. (2008), then Iceland's Cook's
distance is 0.62. The correlation was reduced to r(34) = .33,
p = .053, without Iceland, and to r(31) = .25, p = .16
without the Nordic countries. Thus only a few data points might
be the reason why the GGI was correlated with the mathematics
gap in 2003 but not in any other PISA assessment. In the
Discussion, we return to the contribution of the Nordic countries
to the debate about sex differences in academic performance.
As is the case for the GGI data, the only correlation between
GEM and the gap in mathematics was found for the 2003 data,
r(31) = .367, p = .036. Iceland, Thailand, and Korea had a
Cook's distance larger than the cutoff point, but even without
these countries, the Pearson r correlation still approached
significance, r(28) = .34, p = .063 (Fig. 5). It is interesting to
note that the Nordic countries played an important role here as
with the GGI. Dropping these countries and outliers reduced
the correlation to r(25) = .25, p = .20. In 2009, the positive
correlation between the mathematics gap and GEM was
entirely driven by the group of non-OECD countries. Again,
we are not arguing that these countries should not be included,
but we are merely demonstrating that some data points or
groups of similar countries seem to have repeatedly played a
key role in driving the correlation between gender equality
measures and the PISA mathematics gap when it was found.
In regard to women's share of research positions (Fig. 2),
the 2003 results were strongly driven by four non-OECD
countries. The OECD and non-OECD countries differed consid-
erably in the percentage of female researchers, which partic-
ularly biased the data that Else-Quest et al. (2010) reported.
The four non-OECD countries (Latvia, Thailand, Tunesia, Serbia)
responsible for the highest percentages of female researchers
(ranging from 47% to 53%) had a strong influence on the
correlation. When excluding these four countries, the correla-
tion was reduced from r(29) = .35, p = .06 to r(25) =
.172, p =0.39. Thus, the two PISA surveys for which there
was a reliable relation between the mathematics gap and the
percentage of women researchers irrespective of OECD status
are 2000 and 2006. That in itself does not allow us to reject the
hypothesis of a link between this variable and the sex
differences in mathematics. However, using Cook's distance,
removing the outliers Iceland and Korea reduced the correla-
tion between women's share of research positions in 2003 and
the mathematics gap to r(27) = .25, p = .20.
In regard to the relation between the percentage of women
in parliament and the mathematics gap, the only meaningful
correlation was for the 2003 PISA when considering the OECD
countries only, r(28) = .40, p = .03. Removing the outlier
Iceland from these data did not change the correlation.
In regard to the third specific gender equality indicator, the
female economic activity rate, there was no indication that
outliers affect the statistically significant correlations found for
2000 and 2003.
Altogether, the relation between gender equality indicators
was unreliable across time, and when the relation existed, it
was repeatedly driven by some countries, in particular Iceland
and the nearby other Nordic countries.
3.1 . The overall gap
The focus on international variation in gender equality and
sex differences in achievement has largely been on the
mathematics gap. This gap is certainly important, but so is the
gap in overall achievement, that is the average achievement
in the main PISA domains of mathematics, reading, and science.
This is especially true given the strong correlations among
these domains (in the 2009 PISA assessment, average r for
mathematics and reading = .81, for mathematics and science =
.85, and for reading and science = .86, see SOM for further
details about these cross correlations), and the importance of
all three domains for success in employment and further
education.
We found that boys' overall score was lower than that of
girls in most countries (Fig. 6). In the 2009 PISA, girls performed
significantly better than boys in 70% of the countries (52 out of
the 74), whereas boys significantly outperformed girls in only
4% of the countries (3 out of 74). In 2003 and 2006, the pattern
was similar (in 2003, boys fell behind in 41% of the countries,
girls in none; in 2006, boys fell behind in 58% of the countries
and girls in 2%; the 2000 PISA data were organized differently
such that inferential statistics for overall scores could not be
calculated; see SOM for further details).
As with the mathematics gap, the overall gap was not
consistently related to relevant national equality indicators. For
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151 143

OECD countries
30 Finland, Sweden, Norway
Iceland
non OECD countries
Gender Math Gap (2003 PISA)

20
girls better boys better

10

10

r(35) =0.408, p =0.012 Iceland r(35) =0.435, p =0.007 Iceland

20

30
Gender Math Gap (2006 PISA)

20
girls better boys better

10

Iceland
10 Iceland

r(35) =0.202, p =0.232 r(35) =0.174, p =0.304

20

30
Gender Math Gap (2009 PISA)

20
girls better boys better

10

Iceland
Iceland
10

r(35) =0.236, p =0.16 r(35) =0.114, p =0.501

20
55 60 65 70 75 80 85 55 60 65 70 75 80 85
GGI GGI

Fig. 1. Correlations between the composite gender equality index GGI on the one hand and the sex differences in mathematics achievement in PISA (boy scoresgirl
scores). Top panels) The relation between the 2006 GGI and the 2003 PISA mathematics gap. Note that the 2006 GGI data set is taken because 2006 is the first year GGI
was published. The blue square indicates Iceland, and the three blue data points the three Nordic countries Norway, Sweden, and Finland. The left panel uses the data
from students that score higher than the median socioeconomic status of a given countries (like Guiso et al., 2008), whereas the right panel data uses the data from all
students. Middle panels) Same as top panels for the 2006 PISA gender gap in mathematics. Bottom panels) Same as top panels for the correlation between the 2009 GGI
and 2009 PISA data.

the 2003 PISA data a significant relation was found between GGI Perhaps more important than between-country gaps are
and the overall gap, r(35) = .403, p = .013, with boys falling the overall achievement gaps found within each country. We
behind more in gender equal societies. But also here Iceland's calculated these by dividing the scores of each country into 100
Cook's distance value was unusually high, and without Iceland percentiles, and then taking the difference (boysgirls). These
the correlation was r(34) = .31, p = .06. scores, averaged for OECD and non-OECD countries revealed a
144
2000 PISA
30
Mathematics gender gap
20
girls better boys better
10
10
20 r(24) = 0.479, p = 0.013 (OECD)
r(31) = 0.681, p <.01 (all countries)
2006 PISA
30
Mathematics gender gap
20
10
10
20 r(27) = 0.376, p = 0.044 (OECD)
r(40) = 0.286, p = 0.067 (all countries)
10
percentage female researchers
Fig. 2. Correlations between the percentage of female researchers and the sex difference in mathematics (boy scoresgirl scores) by country. The four panels show the
data for each of the four PISA surveys (2000, 2003, 2006, 2009).
quite remarkable pattern (Fig. 7). The achievement gap, favoring
girls, was larger among lower achieving children for both OECD
and non-OECD countries. The only exception to the pattern of
girls outperforming boys in overall achievement was found
among the highest performers in OECD countries.
3.2 . Achievement continuum and intra-individual differences
PISA data are well suited to assess intra-individual differ-
ences in reading and mathematics achievement, because scores
are standardized; as noted, the average score in OECD countries
is 500 (SD = 100) for mathematics and reading. We calculated
this intra-individual gap by subtracting the mathematics and
reading scores of each student (see SOM). Thus, if the value is
positive, the student was better in mathematics, and if the
value is negative the student was better in reading. When this
mathematics-minus-reading gap is plotted by country (Fig. 8),
a clear pattern emerges. In most countries, boys were better in
mathematics than in reading, whereas girls were better in
reading than in mathematics (that is, there was a common sex
difference in the asymmetry of cognitive profiles).
4 . Discussion
Our comprehensive analyses of all four PISA assessments
enabled us to resolve current debate regarding whether or not
G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
OECD countries 2003 PISA
non-OECD countries
r(25) = 0.172, p = 0.39 (OECD)
r(29) = 0.345, p = 0.058 (all countries)
2009 PISA
r(27) = 0.135, p = 0.484 (OECD)
r(46) = 0.158, p = 0.283 (all countries)
20 30 40 50 10 20 30 40 50
percentage female researchers
a nation's gender equality policies, customs, and women's
wider participation in politics and the economy is systemati-
cally related to sex differences in mathematics achievement
and overall achievement, as predicted by the theoretical
framework of the gender similarities and gender stratification
hypotheses (Else-Quest et al., 2010; Guiso et al., 2008; Hyde &
Mertz, 2009; Reilly, 2012).
Our main conclusions are as follows. First of all, there was
no reliable relation between gender equality measures and sex
differences in mathematics achievement or in overall achieve-
ment. It seems that the positive findings for the 2003 PISA data
were unusual in that a small number of countries strongly
contributed to the correlation; below we will discuss whether
there is something different about those countries. Altogether,
the results across the four very large data sets analyzed here are
not consistent with the gender similarities and gender
stratification hypotheses.
Our second conclusion is that there is a surprisingly large
number of countries, including highly developed countries
with progressive gender equality policies, in which boys fell
considerably behind girls in overall achievement. In the 2009
PISA, this was true for 70% of the countries. Even though boys
fell behind on average, this pattern was different among the top
performers, especially in OECD countries. Here, top performing
boys had higher achievement levels than did top performing
girls. The achievement gap, favoring boys, at the high end may
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151 145

30
2000 OECD countries 2003
non-OECD countries
Mathematics gender gap
20

girls better boys better

10

10

20 r(28) = 0.212, p = 0.26 (OECD) r(28) = 0.398, p = 0.03 (OECD)

r(40) = 0.024, p = 0.878 (all) r(37) = 0.233, p = 0.154 (all)

30 2006 2009
Mathematics gender gap

20

10

10

20 r(31) = 0.194, p = 0.281 (OECD) r(31) = 0.167, p = 0.353 (OECD)

r(52) = 0.087, p = 0.533 (all) r(64) = 0.181, p = 0.146 (all)

0 20 40 60 80 100 0 20 40 60 80 100

percentage of women in parliament percentage of women in parliament

Fig. 3. Correlations between the percentage of women in parliament and the sex difference in mathematics (boy scoresgirl scores) by country. The four panels show
the data for each of the four PISA surveys (2000, 2003, 2006, 2009). Note that this relation is statistically significant for 2003 if non-OECD countries are excluded.

be one of the factors contributing to the overrepresentation of

men in many leadership roles (e.g., academia, business), at
OECD least in OECD countries (of course, other factors play a role as
OECD & Nordic
M non-OECD well, as noted in the introduction). Our final conclusion is that
M
individual boys, on average, have better mathematics skills
Mathematics gender gap (2003 PISA)

M
20 M M M
than reading skills, whereas individual girls have, on average,
r(35
)= -. M
408
, p=
.0 12
M better reading skills than mathematics skills, as noted by Valla
M M M
M
M
M and Ceci (2014), Wang et al. (2013) and Wai et al. (2005). We
M M M

10
M
M M
M
M Finland believe that this is one of the factors that contribute to the sex
M
M
M
M r(31)=
-.249
, p=.1
differences in college majors and careers. These conclusions
M M 62
all contradict with the gender similarities and stratification
boys better

M Norway
M
M M M
M
Sweden
hypotheses.
0
girls better

M
4.1 . Relation between measures of gender equality and sex
differences in achievement are tenuous at best and likely spurious

10 M
Studies that have reported a relation between measures of a
M country's gender equality and sex differences in mathematics
Iceland
achievement have largely focused on the 2003 PISA and
0.60 0.65 0.70 0.75 0.80
concluded that countries with more equality have a smaller
2006 GGI gap (Else-Quest et al., 2010; Guiso et al., 2008; Hyde & Mertz,
2009; Reilly, 2012). We also found such a relation for the 2003
Fig. 4. The 2003 PISA sex differences in mathematics achievement and the 2006
Global Gender Gap Index. The correlation of r = .41 (black line) was reduced
PISA, but demonstrated that this was not consistently found in
to r = .33 without Iceland, and to r = .25 without the indicated Nordic the other PISA surveys, and further that some countries seemed
countries (blue line). to drive the correlations (more about this below). This was the
146 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151

Fig. 5. The relation between the mathematics gap and the Gender Empowerment Measure (GEM). Lines indicate regression lines (red line only of OECD data points).
The four panels show the data for the four PISA assessments and the GEM data of the corresponding year. Red circles represent OECD countries, and black circles of non-
OECD countries. In 2009, there was a statistically significant correlation in the opposite direction, suggesting that the mathematics gap was larger in countries with a
higher level of gender equality (but this was not the case for either the OECD or non-OECD countries taken apart).

case for composite gender equality measures, including the
widely used GGI, and for specific measures (i.e., women's
participation in research, economic activity, and political
participation).
Another issue highlighted by our re-analysis of previous
studies is that there are many potential variables that can be
correlated across the four different PISA assessments, and
beyond that, there are multiple ways of correlating variables
with the equality data of different years (e.g., in the reviewed
studies, both the 2006 and 2007 GGI data have been correlated
with the 2003 PISA). The large number of possible comparisons
increases the risk of statistical type-I errors (Simmons, Nelson,
& Simonsohn, 2011), that is, wrongly concluding that there is a
relation when there is not one. In this view, strong arguments
that national gaps in boys' and girls' achievement are related
to national gender equality policies and outcomes are only
supported if the relation is found consistently in each PISA
assessment. And, they are not.
4.2 . Are Nordic countries doing things differently?
Nordic countries (Iceland, Norway, Sweden, Finland, and
Denmark) score very high in national gender equality mea-
sures (Hausmann et al., 2010), and there is much literature
about the Nordic model in terms of equality policies. We have
shown that, apart from Denmark, they play an important role in
the correlations between equality measures and sex differences
in achievement, especially in 2003. Therefore, one might argue
that it is important to consider the approach of these countries
to gender equality and education, and how their approach
might be a model for other countries. However, we think that
the situation is more complicated, as our data have shown.
Finland, Iceland, and Norway are the top three OECD countries
in terms of the overall achievement gap (whereas Sweden and
Denmark are more similar to the other OECD countries, see
SOM). Thus, even though these countries have a strong
reputation for gender equality social policies, these policies
clearly do not automatically lead to a reduction of sex
differences in achievement. Alternatively, would there be
reasons to argue that the Nordic equality model is the cause
of boy's poor overall achievement relative to girls? Is it fair
to conclude that boys underperform at the cost of girls?
We think that this conclusion is as unwarranted as the
conclusion that gender equality leads to a reduction of
educational achievement gaps. We believe that there is
simply no systematic, causal relation. Our argument is
further strengthened by our findings for Muslim countries.
These countries generally have a poor gender equality
record (Mller, 2006), and the patriarchally organized
culture and Sharia inspired legal proceedings limit women's
G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151 147

100 OECD countries

nonOECD countries

80

Achievement percentile
60

40

20

0 Girls better Boys better

30 20 10 0
Overall gap (in 2009 PISA)

Fig. 7. The sex difference in overall achievement for the 2009 PISA percentile.
The higher the performance, the better boys performed compared to girls. Boys
generally fell behind along the performance continuum, but this pattern was
reversed for the highest performing students.

and girls' opportunities. Yet, in many of these countries, girls'

overall achievement is better than that of boys (see also Fryer &
Levitt, 2010; Kane & Mertz, 2012).
There are important differences between the individual
Muslim countries, as there are between the Nordic countries.
We believe that these observations in regard to their educa-
tional achievement are most likely coincidences or at least not
related to overall gender-related policies or customs. Instead,
the finding that girls' overall achievement is better than that of
boys in countries as varied as Finland and Qatar demonstrates
that political, economic, and social gender equality is largely
uninformative about the sex differences in achievement.
At the other end of the spectrum are the countries in which
boys perform statistically better than girls (namely Costa Rica,
the Indian state Himachal Pradesh, and Columbia). The number
of these countries is too small to look for similarities that can
explain this, although you can always find something. For
example they are all coffee-bean producing nations, but that
does not justify the conclusion that coffee-bean production
helps boys' achievement; the same analogy-mistake might be
true for the equality policies of Scandinavia and their girls'
relatively high achievement.

4.3 . Girls outperform boys in most countries

As noted, most previous studies of sex differences in

achievement have focused on mathematics. One result, in
our opinion, is a comparative neglect of the gap in overall
achievement (across mathematics, reading, and science), in
particular the relatively poor achievement of boys in many
parts of the world. Issues related to the overall achievement

Fig. 6. The overall gap for sex differences (boy scoresgirl scores) in achievement
for the combined reading, mathematics, and science scores in the 2009 PISA.
Countries with a statistically significant sex difference are represented as a
square. In 70% of the countries, girls performed significantly (p b .05) better than
boys, whereas boys outperformed girls in only 4% of the countries.
148 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151

Fig. 8. Intra-individual gap between reading and mathematics. In each PISA assessment, boys were, on average, better in mathematics than in reading (x-axis), while
girls were, on average, better in reading than in mathematics (y-axis). Each data point represents the data of one country.

gap have been examined previously (Gorard, Rees, & Salisbury,
2001; Hodgetts, 2008; Machin & McNally, 2005), but surpris-
ingly (given the wealth of data) have received little attention in
the context of PISA data. In every PISA assessment, the same
patterns emerged: girls had higher overall achievement than
boys, and the magnitude of the gap varied systematically across
the performance continuum. The gap was the largest at the
lowest levels of achievement and was closed in most developing
countries and reversed in developed countries at the highest
levels of achievement. Debates regarding sex differences in
mathematics achievement and in STEM fields were sparked in
these latter countries and for the highest achieving students.
We suggest that the focus on the mathematics gap in these
countries has contributed to the lack of attention to boy's
academic underperformance overall and especially for reading.
The finding that girls' overall achievement exceeds that of
boys' in most countries raises a question that has not yet been
adequately addressed; given that girls do so well academically,
why is it that women are underrepresented in leadership
positions in politics, business, and academia? This is a funda-
mentally important question, because answering it will not
only help to better understand the relation between cognitive
performance measures and real-world career outcomes, but it
will also have major implications for the effectiveness of equality
policies. There is a direct link between this question and the lack
of a link between gender equality and educational outcomes. For
example, despite the lack of opportunities for girls and women in
Muslim countries, girls' achievement is better than that of boys in
those countries. This implies that opportunities for girls and
educational outcome are unrelated (or if they are related it is
not in a transparent way), and if they are unrelated, it is not
surprising that their educational success does not directly
translate in full socio-political and academic participation.
One often proposed answer is that women's underrepre-
sentation is the result of explicit or implicit sexism, although it
is not the only factor (Ceci & Williams, 2011; Ceci, Williams, &
Barnett, 2009). On the basis of our results, we also have to
consider that the overall achievement gap between boys and
girls is different for low and high performing children. The gap
is large and favors girls at the lower end of performance, but it
is non-existent or favors boys at the high end of performance
(Fig. 7). Given that there is indeed evidence that a large
proportion of the best performing school children end up
in successful careers, some leaders in politics, business, and
academia (Kell, Lubinski, & Benbow, 2013; Wai, 2013, in press),
we might therefore expect an overrepresentation of boys just
based on performance data alone, especially in OECD nations.
We call this the high-achievers male advantage gap hypoth-
esis. We believe that this is only one factor among many of the
reasons for the leadership gap and clearly is not a viable factor
in non-OECD nations. There are, however, other factors at play
as well, and some of these start to come into play at later stages
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
of career development, such as pipe-line issues and other
cognitive or personality variables mentioned in the introduc-
tion. At the very least, our study helps to understand the role of
sex differences at the end of compulsory education. That helps
to constrain possible models of what causes the underrepre-
sentation of women in politics and some occupations. One of
the interesting outstanding questions for future research is
whether the increased opportunities for talent development in
richer countries explain why this gap is only found in the group
of OECD nations (Fig. 7). In OECD nations additional factors
likely include the different trade-offs that women and men
make with respect to investment in occupational attainment
and their families (Ferriman, Lubinski, & Benbow, 2009),
differences in career choices (Wang et al., 2013), and in
competitive striving ( Geary, 2010, pp.213235; Sax & Harper,
2007).
4.4 . Sex differences vary depending on where in the performance
distribution school children are
One of the findings of our study is that while, on average,
boys are outperformed academically by girls, the situation is
different at the extremes of the distribution. As we reported
previously for the PISA data and as is known from numerous
other studies (Benbow & Stanley, 1982; Benbow & Stanley,
1983; Stanley & Benbow, 1982; Stoet & Geary, 2013; Wai,
Lubinski, Benbow, & Steiger, 2010), the sex difference in
mathematics performance is considerably larger at the high
end of performance distributions. Interestingly, at least for
PISA data, we found no sex differences at the low end of
mathematics performance (Stoet & Geary, 2013). In regard to
reading differences, the pattern is exactly the opposite, with the
largest gap at the low end and the smallest gap at the high end
(Stoet & Geary, 2013). And indeed, in the current study we
found that while boys fell behind across subjects, it is the
opposite for the children at the top, where boys are in the lead
(Fig. 7).
4.5 . Relative intra-individual strengths play role
Finally, studies of predictors of college majors and career
trajectories have revealed the importance of intra-individual
differences, that is, students' relative academic strengths and
weaknesses (Humphreys et al., 1993; Park et al., 2007), an issue
that has not been assessed in previous studies of sex differences
using the PISA. For instance, it is not simply the absolute level of
mathematical competence that predicts entry into STEM fields,
but also the level of competence relative to other academic
skills. Individuals with strong mathematical abilities are more
likely to pursue non-STEM careers if they also have strong
verbal abilities, which include more women than men
(Lubinski & Benbow, 2006). Our findings across all four PISA
assessments revealed that in most countries boys have
relatively better mathematics achievement and girls relatively
better reading achievement, with implications for college and
later occupational patterns, in particular high achievers
entering STEM professions. These intra-individual differences
may in fact be relatively more important than mean sex
differences in pursuit of STEM careers, consistent with previous
studies (Humphreys et al., 1993; Park et al., 2007; Valla & Ceci,
2014; Wai et al., 2005; Wang et al., 2013) and current sex
149
differences in university enrollment (men are more strongly
represented in STEM, and women more strongly represented in
languages), especially when combined with large sex differ-
ences in vocational interests (Lubinski, 2000; Su et al., 2009).
Further, these intra-individual differences do not vary across
the achievement spectrum as do the general mathematics and
reading gaps (Stoet & Geary, 2013); that is, in most countries, it
is the case that boys are better in mathematics than in reading
while girls are better in reading than in mathematics, irrespec-
tive of whether they are middle or high attainers (see SOM).
4.6 . Outlook
We found no evidence for a reliable relation between PISA
performance and gender equality. This not only casts doubt on
the gender similarities and gender stratification hypotheses,
but also has important implications for educational policy.
Implications for educational policy go well beyond the data and
require taking value frameworks into account. The current
paper is not the right place to elaborate on such frameworks.
Nevertheless, it is possible that the current data will influence
how policy makers think about the options available, for
example, to increase levels of equal opportunities in education.
Therefore, we think it is important to raise a few issues that go
beyond the interpretation of the data. We believe that one
important implication of our analyses is that policy makers and
educators should not expect that broad progress in social
equality (e.g., increased political participation of women) will
necessarily result in educational equality. In fact, we found that
with the exception of high achievers, boys have poorer
educational outcomes than girls around the world, indepen-
dent of social equality indicators; a gap that has grown for
reading over the past decade (Stoet & Geary, 2013). This has
strong implications for national economic prospects as well as
social stability, as young men who are not well integrated into
the economic structures of their society are prone to crime and
violence (Wilson & Daly, 1985). It also seems that the majority
of research papers and interventions have focused on reducing
the mathematics gap. Our finding that the mathematics gap has
been stable for at least the first decade of the 21st century raises
questions regarding the efficacy of the interventions that
have been used thus far (Stoet & Geary, 2012), whether the
interventions focus on the right segment of girls (closing the
overall gap requires interventions with the highest achieving
girls in mathematics Stoet & Geary, 2013), or whether the
amount of resources devoted to these interventions has been
insufficient. We certainly believe that increasing the numbers
of women entering STEM professions and better understanding
why this has been difficult to achieve are important goals (Ceci
& Williams, 2011; Ceci et al., 2009), but the focus on this
issue has resulted in a relative neglect of boys' academic
underperformance. This underperformance has been debated
and solutions have been proposed (e.g., Hoff Sommers, 2001;
Weaver-Hightower, 2003; Yates, 1997; Younger & Warrington,
2004), but our impression of the literature is that this stands in
no comparison to the attention the underachievement and
underrepresentation of girls in STEM subjects has received. In
any case, psychological research focused on better understand-
ing the reasons for this phenomenon and addressing them have
the potential to improve the well-being of many people
throughout much of the world.
150 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
Acknowledgments
We thank Lynne Cooper, Ben Winegard and Drew Bailey for
insightful comments and discussion.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
http://dx.doi.org/10.1016/j.intell.2014.11.006.
References
Benbow, C.P., & Stanley, J.C. (1982). Intellectually talented boys and girls
Educational profiles. Gifted Child Quarterly, 26(2), 8288.
Benbow, C.P., & Stanley, J.C. (1983). Sex differences in mathematical reasoning
ability: More facts. Science, 222, 10291031.
Burt, C., & Moore, R.C. (1912). The mental differences between the sexes. Journal
of Experimental Pedagogy, 273284 (355388).
Ceci, S.J., & Williams, W.M. (2011). Understanding current causes of women's
underrepresentation in science. Proceedings of the National Academy of
Sciences of the United States of America, 108(8), 31573162.
Ceci, S.J., Williams, W.M., & Barnett, S.M. (2009). Women's underrepresentation
in science: Sociocultural and biological considerations. Psychological
Bulletin, 135(2), 218261.
Collins, R., Chafetz, J.S., Blumberg, R.L., Coltrane, S., & Turner, J.H. (1993). Toward
an integrated theory of gender stratification. Sociological Perspectives, 36(3),
185216.
Cook, R.D. (1977). Detection of influential observation in linear regression.
Technometrics, 19(1), 1518.
Del Giudice, M., Booth, T., & Irwing, P. (2012). The distance betwen mars and
venus: Measuring global sex diferences in personality. PLoS ONE, 7(1).
Ellis, L., Hershberger, S., Field, E., Wersinger, S., Sergio, P., Geary, D., Palmer, C.,
Hoyenga, K., Hetsroni, A., & Karadi, K. (2008). Sex differences: Summarizing
more than a century of scientific research. New York, NY: Francis & Taylor.
Else-Quest, N.M., Hyde, J.S., & Linn, M.C. (2010). Cross-national patterns of
gender differences in mathematics: A meta-analysis. Psychological Bulletin,
136(1), 103127.
Feingold, A. (1988). Cognitive gender differences are disappearing. American
Psychologist, 43(2), 95103.
Ferriman, K., Lubinski, D., & Benbow, C.P. (2009). Work preferences, life values,
and personal views of top math/science graduate students and the
profoundly gifted: Developmental changes and sex differences during
emerging adulthood and parenthood. Journal of Personality and Social
Psychology, 97, 517532.
Fryer, R.G., Jr., & Levitt, S.D. (2010). An empirical analysis of the gender gap in
mathematics. American Economic Journal: Applied Economics, 2, 210240.
Geary, D.C. (1996). Sexual selection and sex differences in mathematical
abilities. Behavioral and Brain Sciences, 19, 229247.
Geary, D.C. (2010). Male, female: The evolution of human sex differences (2nd ed.).
Washington, DC: American Psychological Association.
Gorard, S., Rees, G., & Salisbury, J. (2001). Investigating the patterns of
differential attainment of boys and girls at school. British Educational
Research Journal, 27(2), 125139.
Guiso, L., Monte, F., Sapienza, P., & Zingales, L. (2008). Culture, gender, and
math. Science, 320(5880), 11641165.
Halpern, D.F. (2012). Sex differences in cognitive abilities (4th ed.). New York,
NY: Psychology Press.
Hausmann, R., Tyson, L.D., & Zahidi, S. (2010). Global gender gap report. Geneva,
Switzerland: World Economic Forum.
Hedges, L.V., & Nowell, A. (1995). Sex differences in mental scores, variability,
and numbers of high-scoring individuals. Science, 269, 4145.
Hodgetts, K. (2008). Underperformance or getting it right? Constructions of
gender and achievement in the Australian inquiry into boys' education.
British Journal of Sociology of Education, 29(5), 465477.
Hoff Sommers, C. (2001). The war against boys: How misguided feminism is
harming our young men. New York, NY: Touchstone.
Humphreys, L.G., Lubinski, D., & Yao, G. (1993). Utility of predicting group
membership and the role of spatial visualization in becoming an engineer,
physical scientist, or artist. Journal of Applied Psychology, 78, 250261.
Hunt, E., & Wittmann, W. (2008). National intelligence and national prosperity.
Intelligence, 36, 19.
Hyde, J.S. (2005). The gender similarities hypothesis. American Psychologist,
60(6), 581592.
Hyde, J.S., Lindberg, S.M., Linn, M.C., Ellis, A.B., & Williams, C.C. (2008). Gender
similarities characterize math performance. Science, 321(5888), 494495.
Hyde, J.S., & Linn, M. (1988). Gender differences in verbal ability: A meta-
analysis. Psychological Bulletin, 104(1), 5369.
Hyde, J.S., & Mertz, J.E. (2009). Gender, culture, and mathematics performance.
Proceedings of the National Academy of Sciences of the United States of
America, 106(22), 88018807.
Irwing, P., Booth, T., & Batey, M. (2014). An investigation of the factor structure
of the 16pf, version 5 a confirmatory factor and invariance analysis. Journal
of Individual Differences, 35(1), 3846.
Jacobs, J.E., & Simpkins, S.D. (2005). Leaks in the pipeline to math, science, and
technology careers, volume 110 of New directions for child and adolescent
development, chapter Mapping Leaks in the Math. Science, and Technology
Pipeline, 36 (San Francisco, CA).
Jeffcoat, M.K. (2004). It's a pipeline isue women in leadership. Journal of the
American Dental Association, 135(2), 142144.
Kane, J.M., & Mertz, J.E. (2012). Debunking myths about gender and
mathematics performance. Notices of the American Mathematical Society,
59(1), 1021.
Kell, H.J., Lubinski, D., & Benbow, C.P. (2013). Who rises to the top? Early
indicators. Psychological Science, 24(5), 649659.
Klasen, S. (2006). UNDP's gender-related measures: Some conceptual problems
and possible solutions. Journal of Human Development, 7(2), 243274.
Linn, M.C., & Petersen, A.C. (1985). Emergence and characteristics of sex
differences in spatial ability: A meta-analysis. Child Development, 56,
14791498.
Lippa, R.A. (2006). The gender reality hypothesis. American Psychologist, 61(6),
639640.
Lippa, R.A., Collaer, M.L., & Peters, M. (2010). Sex differences in mental rotation
and line angle judgments are positively associated with gender equality
and economic development across 53 nations. Archives of Sexual Behavior,
39(4), 990997.
Locke, J. (1963e). Some thoughts concerning education. New York: Bartleby
(Original 1693).
Lubinski, D. (2000). Scientific and social significance of assessing individual
differences: Sinking shafts at a few critical points. Annual Review of
Psychology, 51, 405444.
Lubinski, D., & Benbow, C.P. (2006). Study of mathematically precocious youth
after 35 years: Uncovering antecedents for the development of math
science expertise. Perspectives on Psychological Science, 1(4), 316345.
Lynn, R., & Mikk, J. (2009). National IQs predict educational attainment in math,
reading and science across 56 nations. Intelligence, 37(3), 305310.
Machin, S., & McNally, S. (2005). Gender and student achievement in English
schools. Oxford Review of Economic Policy, 21(3), 357372.
Mller, J. (2006). Equality under the sharia? Internationale Politik [International
Politics], 61(6), 124127.
OECD (2003a). The PISA 2003 assessment framework. Paris: OECD Publishing.
OECD (2003b). PISA 2003 data analysis manual. OECD (URL http://www.oecd.
org/dataoecd/35/51/35004299.pdf).
OECD (2006). The high cost of low educational performance. Paris: OECD Publishing.
OECD (2010). Pisa 2009 at a glance. OECD Publishing.
OECD (2012). PISA 2009 technical report. Paris: OECD Publishing.
Park, G., Lubinski, D., & Benbow, C.P. (2007). Contrasting intellectual patterns
predict creativity in the arts and sciences. Psychological Science, 18, 948952.
Reilly, D. (2012). Gender, culture, and sex-typed cognitive abilities. PLoS ONE, 7,
e39904.
Rindermann, H. (2008). Relevance of education and intelligence at the national
level for the economic welfare of people. Intelligence, 36(2), 127142.
Sax, L.J., & Harper, C.E. (2007). Origins of the gender gap: Pre-colege and colege
influences on diferences betwen men and women. Research in Higher
Education, 48(6), 669694.
Schmidt, F.L. (2011). A theory of sex differences in technical aptitude and
some supporting evidence. Perspectives on Psychological Science, 6(6),
560573.
Simmons, J.P., Nelson, L.D., & Simonsohn, U. (2011). False-positive psychology :
Undisclosed flexibility in data collection and analysis allows presenting
anything as significant. Psychological Science, 22(11), 13591366.
Stanley, J.C., & Benbow, C.P. (1982). Huge sex-ratios at upper end. American
Psychologist, 37(8), 972.
Stockard, L.V., & Bell, J.C. (1916). A preliminary study of the measuring of
abilities in geometry. Journal of Educational Psychology, 7, 567580.
Stoet, G., & Geary, D.C. (2012). Can stereotype threat explain the gender gap in
mathematics performance and achievement? Review of General Psychology,
16(1), 93102.
Stoet, G., & Geary, D.C. (2013). Sex differences in mathematics and reading
achievement are inversely related: Within- and across-nation assessment
of 10 years of PISA data. PLoS ONE, 8, e57988.
Su, R., Rounds, J., & Armstrong, P.I. (2009). Men and things, women and people:
A meta-analysis of sex differences in interests. Psychological Bulletin,
135(6), 859884.
Woolley, H. Thompson (1914). The psychology of sex. Psychological Bulletin,
11(10), 335379.
 G. Stoet, D.C. Geary / Intelligence 48 (2015) 137151
Tucker, S.A., Hanuscin, D.L., & Bearnes, C.J. (2008). The pipeline igniting girls'
interest in science. Science, 319(5870), 16211622.
United Nations Development Programme (2002). Human development report
2002. New York, NY: Oxford University Press.
United Nations Development Programme (2003). Human development report
(2003). New York, NY: Oxford University Press.
United Nations Development Programme (2005). Human development report
2005. (Unknown (not listed), New York, NY).
United Nations Development Programme (2007). Human development report
2007/2008. New York, NY: Palgrave Macmillan.
United Nations Development Programme (2011). Human development report
2011. New York, NY: Palgrave Macmillan.
Valla, J.M., & Ceci, S.J. (2014). Breadth-based models of women's underrepre-
sentation in stem fields. Perspectives on Psychological Science, 9(2), 219-214.
Voyer, D., & Voyer, S.D. (2014). Gender differences in scholastic achievement: A
meta-analysis. Psychological Bulletin, 140, 11741204.
Voyer, D., Voyer, S., & Bryden, P.M. (1995). Magnitude of sex differences in
spatial abilities: A meta-analysis and consideration of critical variables.
Psychological Bulletin, 117(2), 250270.
Wai, J. (2013). Investigating America's elite: Cognitive ability, education, and
sex differences. Intelligence, 41, 203211.
Wai, J. (2014s). Investigating the world's rich and powerful: Education,
cognitive ability, and sex differences. Intelligence, 46, 5472.
151
Wai, J., Lubinski, D., & Benbow, C.P. (2005). Creativity and occupational
accomplishments among intellectually precocious youths: An age 13 to age
33 longitudinal study. Journal of Educational Psychology, 97, 484492.
Wai, J., Lubinski, D., Benbow, C.P., & Steiger, J.H. (2010). Accomplishment in
science, technology, engineering, and mathematics (STEM) and its relation
to STEM educational dose: A 25-year longitudinal study. Journal of
Educational Psychology, 102(4), 860871.
Wang, M.T., Eccles, J.S., & Kenny, S. (2013). Not lack of ability but more
choice individual and gender differences in choice of careers in science,
technology, engineering, and mathematics. Psychological Science, 24(5),
770775.
Weaver-Hightower, M. (2003). The boy turn in research on gender and
education. Review of Educational Research, 73(4), 471498.
Wilson, M., & Daly, M. (1985). Competitiveness, risk taking, and violence: The
young male syndrome. Ethology and Sociobiology, 5973.
Yates, L. (1997). Gender equity and the boys debate: What sort of challenge is
it? British Journal of Sociology of Education, 18(3), 337347.
Younger, M., & Warrington, M. (2004). Raising boys' achievement. Nottingham,
UK: Department for Education and Skills.
Zentner, M., & Mitura, K. (2012). Stepping out of the caveman's shadow :
Nations' gender gap predicts degree of sex differentiation in mate
preferences. Psychological Science, 23(10), 11761185.