Content uploaded by David C Geary
Author content
All content in this area was uploaded by David C Geary on Jan 28, 2020
Content may be subject to copyright.
https://doi.org/10.1177/0956797617741719
Psychological Science
2018, Vol. 29(4) 581 –593
© The Author(s) 2018
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797617741719
www.psychologicalscience.org/PS
Research Article
The underrepresentation of girls and women in science,
technology, engineering, and mathematics (STEM)
fields is a worldwide phenomenon (Burke & Mattis,
2007; Ceci & Williams, 2011; Ceci, Williams, & Barnett,
2009; Cheryan, Ziegler, Montoya, & Jiang, 2017).
Although women are now well represented in the social
and life sciences (Ceci, Ginther, Kahn, & Williams, 2014;
Su & Rounds, 2016), they continue to be underrepre-
sented in fields that focus on inorganic phenomena
(e.g., computer science). Despite considerable efforts
toward understanding and changing this pattern, the
sex difference in STEM engagement has remained stable
for decades (e.g., in the United States; National Science
Foundation, 2017). The stability of these differences
and the failure of current approaches to change them
calls for a new perspective on the issue.
Here, we identified a major contextual factor that
appears to influence women’s engagement in STEM
education and occupations. We found that countries
with high levels of gender equality have some of the
largest STEM gaps in secondary and tertiary education;
we call this the educational-gender-equality paradox.
For example, Finland excels in gender equality (World
Economic Forum, 2015), its adolescent girls outperform
boys in science literacy, and it ranks second in European
educational performance (OECD, 2016b). With these
high levels of educational performance and overall gen-
der equality, Finland is poised to close the STEM gender
gap. Yet, paradoxically, Finland has one of the world’s
largest gender gaps in college degrees in STEM fields,
and Norway and Sweden, also leading in gender-equality
rankings, are not far behind (fewer than 25% of STEM
graduates are women). We will show that this pattern
extends throughout the world, whereby the graduation
gap in STEM increases with increasing levels of gender
equality.
741719PSSXXX10.1177/0956797617741719Stoet, GearyThe Gender-Equality Paradox
research-article2018
Corresponding Author:
Gijsbert Stoet, Leeds Beckett University, Psychology, Room CC-CL-919,
City Campus, Leeds, LS1 3HE, United Kingdom
E-mail: stoet@gmx.com
The Gender-Equality Paradox in Science,
Technology, Engineering, and Mathematics
Education
Gijsbert Stoet1 and David C. Geary2
1School of Social Sciences, Leeds Beckett University, and 2Department of Psychological Sciences, University of Missouri
Abstract
The underrepresentation of girls and women in science, technology, engineering, and mathematics (STEM) fields is a
continual concern for social scientists and policymakers. Using an international database on adolescent achievement
in science, mathematics, and reading (N = 472,242), we showed that girls performed similarly to or better than boys
in science in two of every three countries, and in nearly all countries, more girls appeared capable of college-level
STEM study than had enrolled. Paradoxically, the sex differences in the magnitude of relative academic strengths and
pursuit of STEM degrees rose with increases in national gender equality. The gap between boys’ science achievement
and girls’ reading achievement relative to their mean academic performance was near universal. These sex differences
in academic strengths and attitudes toward science correlated with the STEM graduation gap. A mediation analysis
suggested that life-quality pressures in less gender-equal countries promote girls’ and women’s engagement with STEM
subjects.
Keywords
cognitive ability, cross-cultural differences, educational psychology, science education, sex differences, open materials
Received 5/11/17; Revision accepted 10/12/17
582 Stoet, Geary
We propose that the educational-gender-equality para-
dox is driven by two different processes, one based on
distal social factors and the other on more proximal factors.
The latter is student’s own rational decision making based
on relative academic strengths and weaknesses as well as
attitudes that can be influenced by distal factors (Fig. 1).
Our proposal that students’ own rational decisions
play a key role in explaining the educational-gender-
equality paradox is inspired by the expectancy-value
theory (Eccles, 1983; Wang & Degol, 2013). On the basis
of this theory, it is hypothesized that students use their
own relative performance (e.g., knowledge of what
subjects they are best at) as a basis for decisions about
further educational and occupational choices, and this
has been demonstrated for STEM-related decision mak-
ing in the United States (Wang, Eccles, & Kenny, 2013).
The basic idea that individuals choose academic paths
on the basis of perceived individual strengths is reflected
in common practice by school professionals: When stu-
dents have the opportunity to choose their coursework
in secondary education, they are typically recommended
to make choices on the basis of their strengths and
enjoyment (e.g., Gardner, 2016; Universities and Colleges
Admissions Service, 2015).
Wider social factors may influence engagement in
STEM fields through students’ utility beliefs or the
expected long-term value of an academic path (Eccles,
1983; Wang & Degol, 2013). Social factors that might
influence STEM engagement are best assessed by com-
paring countries that vary widely in the associated costs
and benefits of a STEM career. One possibility is that
contexts with fewer economic opportunities and higher
economic risks may make relatively high-paying STEM
occupations more attractive relative to contexts with
greater opportunities and lower risks. This may con-
tribute to the educational-gender-equality paradox,
because economic and general life risks are lower in
gender-equal countries, which in turn results in greater
opportunity for individual interests and academic
strengths to influence investment in one academic path
or another, as demonstrated by Wang etal. (2013) for
the United States.
In the present article, we report analyses of the aca-
demic achievement of almost 475,000 adolescents across
67 nations or economic regions. We found that girls and
boys have similar abilities in science literacy in most
nations. At the same time, on the basis of a novel
approach for examining intraindividual differences in
academic strengths and relative weaknesses, we report
that science or mathematics is much more likely to be a
personal academic strength for boys than for girls. We
then report that the relation between the sex differences
in academic strengths and college graduation rates in
STEM fields is larger in more gender-equal countries.
Finally, we conducted a mediation analysis that suggests
that the latter is related to overall life satisfaction, which,
in turn, is related to income and economic risk in less
developed countries (cf. Pittau, Zelli, & Gelman, 2010).
Method
Programme for International Student
Assessment (PISA)
PISA (OECD, 2016b) is the world’s largest educational
survey. PISA assessments in science literacy, reading
comprehension, and mathematics are conducted every
3 years, and in each cycle, one of these domains is stud-
ied in depth. In 2015, the focus was on science literacy,
which included additional questions about science learn-
ing and attitudes (see below). We used this most recent
data set, in which 519,334 students from 72 nations and
regions participated. In order to prevent double-counting
of samples, we excluded regions for which we also had
national data (Massachusetts and North Carolina, several
Spanish regions, and Buenos Aires, because we had data
from the United States, Spain, and Argentina as a whole);
this exclusion resulted in a sample of 472,242 students
in 67 nations or regions (Table S1 in the Supplemental
Material available online), which represents 25,141,223
students (i.e., the sum of weights provided by PISA for
each student). Our data set included the following
regions: Hong Kong, Macao, Chinese Taipei, and the
Chinese provinces of Beijing, Shanghai, Jiangsu, and
Guangdong (i.e., these four Chinese provinces were
combined into one sub-data set by PISA).
The PISA organizers selected a representative sample
of schools and students in each participating country
or region. Participating students were between 15 years
and 3 months and 16 years and 2 months old. All par-
ticipating students completed a 2-hr PISA test that
assessed how well they can apply their knowledge in
the domains of reading comprehension, mathematics,
Fig. 1. Schematic illustration of the factors influencing educational
and occupational choices. Distal factors, such as relatively poor living
conditions, might influence the development of personal academic
strengths and attitudes toward different academic fields, which in turn
result in choices individuals make in secondary education, tertiary
education, and occupations.
The Gender-Equality Paradox 583
and science literacy. The same (translated) test material
was used in each country.
PISA uses a well-developed statistical framework to
calculate scores for science literacy, mathematics, read-
ing comprehension, and numerous other variables
related to student attitudes and socioeconomic factors
(OECD, 2016a). The scores of each student in each
academic domain are scaled such that the average of
students in Organization for Economic Cooperation and
Development (OECD) countries is 500 points and the
standard deviation is 100 points.
The additional science literacy assessments in 2015
focused on attitudes, including science self-efficacy,
broad interest in science, and enjoyment of science. For
science self-efficacy,
PISA 2015 asked students to report on how easy
they thought it would be for them to: recognize the
science question that underlies a newspaper report
on a health issue; explain why earthquakes occur
more frequently in some areas than in others;
describe the role of antibiotics in the treatment of
disease; identify the science question associated
with the disposal of garbage; predict how changes
to an environment will affect the survival of certain
species; interpret the scientific information provided
on the labelling of food items; discuss how new
evidence can lead them to change their
understanding about the possibility of life on Mars;
and identify the better of two explanations for the
formation of acid rain. For each of these, students
could report that they “could do this easily”, “could
do this with a bit of effort”, “would struggle to do
this on [their] own”, or “couldn’t do this”. Students’
responses were used to create the index of science
self-efficacy. (OECD, 2016b, p. 136)
Broad interest in science was assessed as follows:
Students reported on a five-point Likert scale with
the categories “not interested”, “hardly interested”,
“interested”, “highly interested”, and “I don’t know
what this is”, their interest in the following topics:
biosphere (e.g., ecosystem services, sustainability);
motion and forces (e.g., velocity, friction, magnetic
and gravitational forces); energy and its transformation
(e.g., conservation, chemical reactions); the
Universe and its history; how science can help us
prevent disease. (OECD, 2016b, p. 284)
Enjoyment of science was assessed using the follow-
ing questions:
I generally have fun when I am learning <broad
science> topics; I like reading about <broad
science>; I am happy working on <broad science>
topics; I enjoy acquiring new knowledge in <broad
science>; and I am interested in learning about
<broad science>. (OECD, 2016b, p. 284; different
science topics were inserted in <broad science>
across questions)
In order to estimate whether a student would, in prin-
ciple, be capable of study in STEM, we used a proficiency
level of at least 4 (of a possible 6) in science, mathemat-
ics, and reading comprehension. For science literacy
for instance and according to the PISA guidelines,
At Level 4, students can use more complex or more
abstract content knowledge, which is either
provided or recalled, to construct explanations of
more complex or less familiar events and processes.
They can conduct experiments involving two or
more independent variables in a constrained
context. They are able to justify an experimental
design, drawing on elements of procedural and
epistemic knowledge. Level 4 students can interpret
data drawn from a moderately complex data set or
less familiar context, draw appropriate conclusions
that go beyond the data and provide justifications
for their choices.” (OECD, 2016b, p. 60)
We believe that level 4 would be a minimal
req uirement.
At Level 3, students can draw upon moderately
complex content knowledge to identify or construct
explanations of familiar phenomena. In less familiar
or more complex situations, they can construct
explanations with relevant cueing or support. They
can draw on elements of procedural or epistemic
knowledge to carry out a simple experiment in a
constrained context. Level 3 students are able to
distinguish between scientific and non-scientific
issues and identify the evidence supporting a
scientific claim.” (OECD, 2016b, p. 60)
Publications further detailing the PISA framework
and methodology are available via http://www.oecd
.org/pisa/pisaproducts/.
STEM degrees
The United Nations Educational, Scientific and Cultural
Organization (UNESCO) reports national statistics on,
among other things, education. We used the UNESCO
graduation data (http://data.uis.unesco.org) labeled “Dis-
tribution of tertiary graduates” in the years 2012 to 2015
in natural sciences, mathematics, statistics, information and
communication technologies, engineering, manufacturing,
584 Stoet, Geary
and construction (Table S1). The percentage of women
among STEM graduates ranged from 12.4% in Macao to
40.7% in Algeria; the median was 25.4%.
Gender equality
The World Economic Forum publishes The Global Gen-
der Gap Report annually. We used the 2015 data (World
Economic Forum, 2015). For each nation, the Global
Gender Gap Index (GGGI) assesses the degree to
which girls and women fall behind boys and men on
14 key indicators (e.g., earnings, tertiary enrollment
ratio, life expectancy, seats in parliament) on a 0.0 to
1.0 scale, with 1.0 representing complete parity (or men
falling behind). For the countries participating in the
2015 PISA, GGGI scores ranged from 0.593 for the
United Arab Emirates to 0.881 for Iceland (Table S1).
Overall life satisfaction (OLS)
We took the OLS score from the United Nations Devel-
opment Programme (2016, pp. 250–253). The OLS ques-
tion was formulated as follows:
Please imagine a ladder, with steps numbered
from zero at the bottom to ten at the top. Suppose
we say that the top of the ladder represents the
best possible life for you, and the bottom of the
ladder represents the worst possible life for you.
On which step of the ladder would you say you
personally feel you stand at this time, assuming
that the higher the step the better you feel about
your life, and the lower the step the worse you
feel about it? Which step comes closest to the way
you feel?
This score was expressed on a scale from 0 (least
satisfied) to 10 (most satisfied; M = 6.2, SD = 0.9, ranging
from 4.1 in Georgia to 7.6 in Switzerland and Norway).
Analyses
For each participating student, the PISA data set pro-
vides scores for mathematics, science literacy, and read-
ing comprehension. We used these given scores to
calculate each student’s highest performing subject (i.e.,
personal strength), second highest, and lowest. To do
so, we needed to calculate each student’s average score
in these three subjects and then compare each subject
score to the calculated average score. In order to make
such calculations possible, we standardized data first.
In other words, we scaled the data into a common
format, namely z scores, which have a mean of 0 and
a standard deviation of 1.
We calculated each students’ relative strengths in
mathematics, science literacy, and reading comprehen-
sion using the following steps:
1. We standardized the mathematics, science, and
reading scores on a nation-by-nation basis. We
call these new standardized scores zMath, zRead-
ing, and zScience, respectively.
2. We calculated for each student the standardized
average score of the new z scores. We call this
zGeneral.
3. Then, we calculated each student’s intraindivid-
ual strengths by subtracting zGeneral as follows:
relative science strength = zScience – zGeneral,
relative math strength = zMath – zGeneral, rela-
tive reading strength = zReading – zGeneral.
4. Finally, using these new intraindividual (relative)
scores, we calculated for each country the aver-
ages for boys and girls and subtracted those
scores to calculate the gender gaps in relative
academic strengths.
To illustrate, one U.S. student had the following three
PISA scores for science, mathematics, and reading: 364,
411, and 344, respectively. After standardization (Step
1), these scores were zScience = −1.39, zMath = −0.69,
and zReading = −1.61. The student’s zGeneral score
was −1.27 (Step 2). His relative strengths were calcu-
lated by subtracting zGeneral from the standardized
scores and then again standardizing the difference
scores (because they are by definition not standardized).
Using this calculation, we obtained the following rela-
tive scores for this student: relative science strength =
−0.71, relative math strength = 2.23, and relative reading
strength = −1.34 (Step 3). Note that although this stu-
dent’s scores in all three subjects are below the stan-
dardized national mean (i.e., 0), his personal strength
in mathematics deviates more than 2 standard devia-
tions from the national mean of relative mathematics
strengths. In other words, the gap between his math-
ematics score and his overall mean score is much larger
(> 2 SDs) than is typical for U.S. students. Using these
types of scores, we could calculate the intraindividual
sex differences for science, mathematics, and reading
for the United States (and similarly for all other nations
and regions).
Further, we calculated for each student the difference
between actual science performance and science self-
efficacy (i.e., self-perceived ability). For this, we used
the same method as reported elsewhere (Stoet, Bailey,
Moore, & Geary, 2016, p. 10): For each participating
nation, we first standardized science performance and
science self-efficacy scores. Then, we subtracted these
two variables for each student and then once more
The Gender-Equality Paradox 585
standardized the difference for the students of each
country separately. The resulting score is a measure of
the degree to which science self-efficacy is unrepresen-
tative of actual performance (i.e., underestimation of
own ability or exaggeration of own ability).
For correlations, we typically applied Spearman’s ρ
(correlation coefficient abbreviated as rs), because not
all variables were normally distributed. Throughout all
analyses, we used an alpha criterion of .05.
Results
Sex differences in science literacy
For each of the 67 countries and regions participating
in the 2015 PISA, we first tested for sex differences in
science literacy (i.e., average score of boys – average
score of girls, by country; Fig. 2a). We found that girls
outperformed boys in 19 (28.4%) countries, boys out-
performed girls in 22 (32.8%) countries, and there was
no statistically significant difference in the remaining 26
(38.8%) countries. The mean national effect size (Cohen’s
d) was −0.01 (SD = 0.13, 95% confidence interval, or
CI = [−0.04, 0.02]), ranging between −0.46 (95% CI =
[−0.50, −0.41]) in favor of girls (in Jordan) and 0.26 (95%
CI = [0.21, 0.31]) in favor of boys (in Costa Rica). The
relation between the effect size of the absolute science
gap and gender equality (GGGI) was not statistically sig-
nificant (rs = .23, 95% CI = [−.18, .46], p = .069, n = 62).
Sex differences in academic strengths
As we previously reported for reading and mathematics
(Stoet & Geary, 2015), there were consistent sex differ-
ences in intraindividual academic strengths across read-
ing and science. In all countries except for Lebanon
and Romania (97% of countries), boys’ intraindividual
strength in science was (significantly) larger than that
of girls (Fig. 2b). Further, in all countries, girls’ intrain-
dividual strength in reading was larger than that of
boys, while boys’ intraindividual strength in mathemat-
ics was larger than that of girls. In other words, the sex
differences in intraindividual academic strengths were
near universal. The most important and novel finding
here is that the sex difference in intraindividual strength
in science was higher and more favorable to boys in
more gender-equal countries, rs = .42, 95% CI = [.19,
.61], p < .001, n = 62 (Fig. 3a), as was the sex difference
in intraindividual strength in reading, which favored
girls in more gender-equal countries, rs = −.30, 95%
CI = [−.51, −.06], p = .017, n = 62.
Another way of calculating these patterns is to exam-
ine the percentage of students who have individual
strengths in science, mathematics, and reading, respec-
tively. To do so, we first determined students’ individual
strength. Next, we calculated the percentage of boys
and girls who had science, mathematics, or reading as
their personal academic strength; this contrasts with
the above analysis that focused on the overall magni-
tude of these strengths independently of whether they
were the students’ personal strength. We found that on
average (across nations), 24% of girls had science as
their strength, 25% of girls had mathematics as their
strength, and 51% had reading. The corresponding val-
ues for boys were 38% science, 42% mathematics, and
20% reading.
Thus, despite national averages that indicate that
boys’ performance was consistently higher in science
than that of girls relative to their personal mean across
academic areas, there were substantial numbers of girls
within nations who performed relatively better in sci-
ence than in other areas. Within Finland and Norway,
two countries with large overall sex differences in the
intraindividual science gap and very high GGGI scores,
there were 24% and 18% of girls, respectively, who had
science as their personal academic strength, relative to
37% and 46% of boys.
Finally, it should also be noted that the difference
between the percentage of girls with a strength in sci-
ence or mathematics was always equally large or larger
than the percentage of women graduating in STEM
fields; importantly, this difference was again larger in
more gender-equal countries (rs = .41, 95% CI = [.15,
.62], n = 50, p = .003). In other words, more gender-
equal countries were more likely than less gender-equal
countries to lose those girls from an academic STEM
track who were most likely to choose it on the basis of
personal academic strengths.
The above analyses show that most boys scored rela-
tively higher in science than their all-subjects average,
and most girls scored relatively higher in reading than
their all-subjects average. Thus, even when girls outper-
formed boys in science, as was the case in Finland, girls
generally performed even better in reading, which means
that their individual strength was, unlike boys’ strength,
reading. The relevant finding here is that the intraindi-
vidual sex differences in relative strengths in science and
reading rose with increases in gender equality (GGGI).
In accordance with expectancy-value theory, this pattern
should result in far more boys than girls pursuing a STEM
career in more gender-equal nations, and this was the
case (rs = −.47, 95% CI = [−.66, −.22], p < .001,
n = 50; Fig. 3b). And, similarly, girls will be more likely
than boys to choose options in which they can gain the
most benefit from their relative strength in reading.
Science attitudes and gender equality
Next, we considered sex differences in science atti-
tudes, namely science self-efficacy, broad interest in
586
Fig. 2. Sex differences in Programme for International Student Assessment (PISA) science, mathematics, and reading scores expressed as Cohen’s ds (see Table S2 in
the Supplemental Material for confidence intervals). Sex differences were calculated as the scores of boys minus the scores of girls. Thus, negative values indicate an
advantage for girls, and positive values indicate an advantage for boys. Results are shown separately for (a) sex differences in absolute PISA scores and (b) sex differ-
ences in intraindividual scores.
587
Fig. 3. Scatterplots (with best-fitting regression lines) showing the relation between gender equality and sex differences in (a) intraindividual science performance and
(b) the percentage of women among science, technology, engineering, and math (STEM) graduates. Gender equality was measured with the Global Gender Gap Index
(GGGI), which assesses the extent to which economic, educational, health, and political opportunities are equal for women and men. The gender gap in intraindividual
science scores (a) was larger in more gender-equal countries (rs = .42). The percentage of women with degrees in STEM fields (b) was lower in more gender-equal
countries (rs = −.47).
588 Stoet, Geary
science, and enjoyment of science. Boys’ science self-
efficacy was higher than that of girls in 39 of 67 (58%)
countries, and especially so in more gender-equal coun-
tries, rs = .60, 95% CI = [.41, .74], p < .001, n = 61 (Fig.
4). Similarly, boys expressed a stronger broad interest
in science than girls in 51 (76%) countries, and again
this was particularly true in more gender-equal coun-
tries, rs = .41, 95% CI = [.15, .62], p = .003, n = 50. And
finally, the same was found for students’ enjoyment of
science; boys reported more joy in science than girls
in 29 (43%) countries, and more so in gender-equal
countries, rs = .46, 95% CI = [.23, .64], p < .001, n = 61.
Further, these attitude gaps were correlated with the
intraindividual science gap (self-efficacy: rs = .24, 95%
CI = [−.00, .46], p = .052, n = 66; enjoyment of science:
rs = .31, 95% CI = [.07, .52], p = .010, n = 66; broad
interest: rs = .27, 95% CI = [.01, .51], p = .043, n = 54).
Science self-efficacy was relatively weakly correlated
with science performance (across participating nations,
r = .17, 95% CI = [.16, .18], n = 472,242, p < .001). This
means that the deviation between science self-efficacy
and science performance is of interest (e.g., students
might under- or overestimate their own performance,
and this could influence later choices). We calculated
for each student the difference between standardized
science self-efficacy scores and standardized science
performance scores (this is a measure of the component
of self-efficacy that is independent from actual perfor-
mance; see Method). Using this metric, we found that
in 34 (49%) countries, boys overestimated their science
self-efficacy and deviated significantly from girls, in
comparison with 5 (7%) countries where girls overes-
timated their science self-efficacy and deviated signifi-
cantly from boys. Paradoxically, boys’ overestimation
of their competence in science was larger in countries
with higher GGGI scores (M = 0.739, SD = 0.06) relative
to countries in which there was no sex difference in
the estimation of science competence (M = 0.697,
SD = 0.04), t(54) = 2.66, p = .010.
Next, we used the science performance data and
attitude data (broad interest in science and enjoyment
of science) to determine the percentage of female stu-
dents who, in principle, could be successful in tertiary
education in STEM fields. For this, we defined suitability
as follows: A student would need to have a proficiency
level of at least 4 in all three PISA domains (science,
mathematics, and reading; see Method). Using these
ability criteria, we would expect far more women
among STEM graduates (international mean = 49%,
SD = 4) than are actually found in any country (inter-
national mean = 28%, SD = 6; Fig. 5a). In regard to
attitudes, we assumed that they should at least have the
Fig. 4. Scatterplot (with best-fitting regression line) showing the relation between
sex difference in science self-efficacy and the Global Gender Gap Index.
589
Fig. 5. Scatterplots showing the relation between the percentage of female students estimated to choose further science, technology, engineering, and math (STEM) study
after secondary education and the estimated percentage of female STEM graduates in tertiary education. Red lines indicate the estimated (horizontal) and actual (vertical)
average graduation percentage of women in STEM fields. For instance, in (c), we estimated that 34% of women would graduate college with a STEM degree (internationally),
but only 28% did so. Identity lines (i.e., 45° lines) are colored blue; points above the identity lines indicate fewer women STEM graduates than expected. Panel (a) displays
the percentage of female students estimated to choose STEM study on the basis of ability alone (see the text for criteria). Although there was considerable cross-cultural
variation, on average around 50% of students graduating in STEM fields could be women, which deviates considerably from the actual percentage of women among STEM
graduates. The estimate of women STEM students shown in (b) was based on both ability, as in (a), and being above the international median score in science attitudes. The
estimate shown in (c) is based on ability, attitudes, and having either mathematics or science as a personal strength.
590 Stoet, Geary
international median level of enjoyment of science,
interest in science, and science self-efficacy. Using these
additional criteria, the percentage of girls likely to
enjoy, feel capable of participating in, and be successful
in tertiary STEM programs is still considerably higher
in every country (international mean = 41%, SD = 6),
except Tunisia, than was actually found (Fig. 5b).
As argued above, we believe that factors other than
attitude and motivation play a role—namely personal
academic strengths. When we added this factor to our
estimate (Fig. 5c), we saw that the difference between
expected and actual STEM graduates became smaller
(international mean = 34%, SD = 6), although it is still
the case that in most countries women’s STEM gradu-
ation rates are lower than we would anticipate (see
Discussion).
Mediation model
Thus far, we have shown that the sex differences in
STEM graduation rates and in science literacy as an
academic strength become larger with gains in gender
equality and that schools prepare more girls for further
STEM study than actually obtain a STEM college degree.
We will now consider one of the factors that might
explain why the graduation gap may be larger in the
more gender-equal countries. Countries with the high-
est gender equality tend to be welfare states (to varying
degrees) with a high level of social security for all its
citizens; in contrast, the less gender-equal countries
have less secure and more difficult living conditions,
likely leading to lower levels of life satisfaction (Pittau
etal., 2010). This may in turn influence one’s utility
beliefs about the value of science and pursuit of STEM
occupations, given that these occupations are relatively
high paying and thus provide the economic security
that is less certain in countries that are low in gender
equality. We used OLS as a measure of overall life cir-
cumstances; this is normally distributed and is a good
proxy for economic opportunity and hardship and
social and personal well-being (Pittau etal., 2010).
In more equal countries, overall life satisfaction was
higher (rs = .55, 95% CI = [.35, .70], p < .001, n = 62).
Accordingly, we tested whether low prospects for a
satisfied life may be an incentive for girls to focus more
on science in school and, as adults, choose a career in
a relatively higher paid STEM field. If our hypothesis is
correct, then OLS should at least partially mediate the
relation between gender equality and the sex differ-
ences in STEM graduation. A formal mediation analysis
using a bootstrap method with 5,000 iterations con-
firmed the mediational model path of life satisfaction for
STEM graduation (mean indirect effect = −0.19, SE = 0.08,
Sobel’s z = −2.24, p < .025, 95% CI of boot strapped
samples = [−0.39, −0.04]). The effect of the direct path
in the mediation model was statistically significant
(mean direct effect = −0.34, SE = 0.135, 95% CI of boot-
strapped samples = [−0.65, −0.02], p = .038), and the
mediation was considered partial (proportion mediated =
0.35, 95% CI = [0.06, 0.95], p = .013; Table S3 in the
Su p plemental Material). A sensitivity analysis of this
mediation (Imai, Keele, & Tingley, 2010; Tingley, Yama-
moto, Hirose, Keele, & Imai, 2014) showed the point
at which the average causal mediation effect (ACME)
was approximately zero (ρ = −0.4, 95% CI = [−0.11,
0.15],
RR
MY
2* 2*
= 0.16,
RR
MY
22
= 0.07; Fig. S1 in the Supple-
mental Material). The latter finding suggests that an
unknown third variable may have confounded the
mediation model (see Discussion).
Discussion
Using the most recent and largest international database
on adolescent achievement, we confirmed that girls
performed similarly or better than boys on generic sci-
ence literacy tests in most nations. At the same time,
women obtained fewer college degrees in STEM disci-
plines than men in all assessed nations, although the
magnitude of this gap varied considerably. Further, our
analysis suggests that the percentage of girls who would
likely be successful and enjoy further STEM study was
considerably higher than the percentage of women
graduating in STEM fields, implying that there is a loss
of female STEM capacity between secondary and ter-
tiary education.
One of the main findings of this study is that, para-
doxically, countries with lower levels of gender equality
had relatively more women among STEM graduates than
did more gender-equal countries. This is a paradox,
because gender-equal countries are those that give girls
and women more educational and empowerment oppor-
tunities and that generally promote girls’ and women’s
engagement in STEM fields (e.g., Williams & Ceci, 2015).
In our explanation of this paradox, we focused on
decisions that individual students may make and deci-
sions and attitudes that are likely influenced by broader
socioeconomic considerations. On the basis of expec-
tancy-value theory (Eccles, 1983; Wang & Degol, 2013),
we reasoned that students should at least, in part, base
educational decisions on their academic strengths.
Independently of absolute levels of performance, boys
on average had personal academic strengths in science
and mathematics, and girls had strengths in reading
comprehension. Thus, even when girls’ absolute sci-
ence scores were higher than those of boys, as in Fin-
land, boys were often better in science relative to their
overall academic average. Similarly, girls might have
scored higher than boys in science, but they were often
The Gender-Equality Paradox 591
even better in reading. Critically, the magnitude of these
sex differences in personal academic strengths and
weaknesses was strongly related to national gender
equality, with larger differences in more gender-equal
nations. These intraindividual differences in turn may
contribute, for instance, to parental beliefs that boys
are better at science and mathematics than girls (Eccles
& Jacobs, 1986; Gunderson, Ramirez, Levine, & Beilock,
2012).
We also found that boys often expressed higher self-
efficacy, more joy in science, and a broader interest in
science than did girls. These differences were also
larger in more gender-equal countries and were related
to the students’ personal academic strength. We discuss
some implications below (Interventions).
Explanations
We propose that when boys are relatively better in sci-
ence and mathematics while girls are relatively better
at reading than other academic areas, there is the
potential for substantive sex differences to emerge in
STEM-related educational pathways. The differences are
expected on the basis of expectancy-value theory and
are consistent with prior research (Eccles, 1983; Wang
& Degol, 2013). The differences emerge from a seem-
ingly rational choice to pursue academic paths that are
a personal strength, which also seems to be common
academic advice given to students, at least in the United
Kingdom (e.g., Gardner, 2016; Universities and Colleges
Admissions Service, 2015).
The greater realization of these potential sex differ-
ences in gender-equal nations is the opposite of what
some scholars might expect intuitively, but it is consis-
tent with findings for some other cognitive and social
sex differences (e.g., Lippa, Collaer, & Peters, 2010;
Pinker, 2008; Schmitt, 2015). One possibility is that the
liberal mores in these cultures, combined with smaller
financial costs of foregoing a STEM path (see below),
amplify the influence of intraindividual academic
strengths. The result would be the differentiation of the
academic foci of girls and boys during secondary edu-
cation and later in college, and across time, increasing
sex differences in science as an academic strength and
in graduation with STEM degrees.
Whatever the processes that exaggerate these sex dif-
ferences, they are abated or overridden in less gender-
equal countries. One potential reason is that a well-paying
STEM career may appear to be an investment in a more
secure future. In line with this, our mediation analysis
suggests that OLS partially explains the relation between
gender equality and the STEM graduation gap. Some
caution when interpreting this result is needed, though.
Mediation analysis depends on a number of assumptions,
some of which can be tested using a sensitivity analysis,
which we conducted (Imai, Keele, & Yamamoto, 2010).
The sensitivity analysis gives an indication of the correla-
tion between the statistical error component in the equa-
tions used for predicting the mediator (OLS) and the
outcome (STEM graduation gap); this includes the effect
of unobserved confounders. Given the range of ρ values
in the sensitivity analysis (Fig. S1), it is possible that a
third variable could be associated with OLS and the
STEM graduation gap. A related limitation is that the
sensitivity analysis does not explore confounders that
may be related to the predictor variable (i.e., GGGI).
Future research that includes more potential confounders
is needed, but such data are currently unavailable for
many of the countries included in our analysis.
Relation to previous studies of gender
equality and educational outcomes
Our current findings agree with those of previous stud-
ies in that sex differences in mathematics and science
performance vary strongly between countries, although
we also believe that the link between measures of gen-
der equality and these educational gaps (e.g., as dem-
onstrated by Else-Quest, Hyde, & Linn, 2010; Guiso,
Monte, Sapienza, & Zingales, 2008; Hyde & Mertz, 2009;
Reilly, 2012) can be difficult to determine and is not
always found (Ellison & Swanson, 2010; for an in-depth
discussion, see Stoet & Geary, 2015).
We believe that one factor contributing to these
mixed results is the focus on sex differences in absolute
performance, as contrasted with sex differences in aca-
demic strengths and associated attitudes. As we have
shown, if absolute performance, interest, joy, and self-
efficacy alone were the basis for choosing a STEM
career, we would expect to see more women entering
STEM career paths than do so (Fig. 5).
It should be noted that there are careers that are not
STEM by definition, although they often require STEM
skills. For example, university programs related to
health and health care (e.g., nursing and medicine)
have a majority of women. This may partially explain
why even fewer women than we estimated pursue a
college degree in STEM fields despite obvious STEM
ability and interest.
Interventions
Our results indicate that achieving the goal of parity in
STEM fields will take more than improving girls’ science
education and raising overall gender equality. The gen-
erally overlooked issue of intraindividual differences in
academic competencies and the accompanying influ-
ence on one’s expectancies of the value of pursuing one
type of career versus another need to be incorporated
into approaches for encouraging more women to enter
592 Stoet, Geary
the STEM pipeline. In particular, high-achieving girls
whose personal academic strength is science or math-
ematics might be especially responsive to STEM-related
interventions.
In closing, we are not arguing that sex differences
in academic strengths or wider economic and life-risk
issues are the only factors that influence the sex differ-
ence in the STEM pipeline. We are confirming the
importance of the former (Wang etal., 2013) and show-
ing that the extent to which these sex differences mani-
fest varies consistently with wider social factors,
including gender equality and life satisfaction. In addi-
tion to placing the STEM-related sex differences in
broader perspective, the results provide novel insights
into how girls’ and women’s participation in STEM
might be increased in gender-equal countries.
Action Editor
Timothy J. Pleskac served as action editor for this article.
Author Contributions
G. Stoet and D. C. Geary collaboratively designed the study
and contributed equally to the writing of the article. G. Stoet
analyzed all data, which were processed and interpreted by
both authors. Both authors approved the final version of the
manuscript for submission.
ORCID iD
Gijsbert Stoet https://orcid.org/0000-0002-7557-483X
Declaration of Conflicting Interests
The author(s) declared that there were no conflicts of interest
with respect to the authorship or the publication of this
article.
Supplemental Material
Additional supporting information can be found at http://
journals.sagepub.com/doi/suppl/10.1177/0956797617741719
Open Practices
All materials are publicly available (for links, see the Open
Practices Disclosure form). The complete Open Practices Dis-
closure for this article can be found at http://journals.sagepub
.com/doi/suppl/10.1177/0956797617741719. This article has
received the badge for Open Materials. More information about
the Open Practices badges can be found at http://www.psycho
logicalscience.org/publications/badges.
References
Burke, R. J., & Mattis, M. C. (2007). Women and minorities
in science, technology, engineering, and mathematics:
Upping the numbers. Cheltenham, England: Edward
Elgar.
Ceci, S. J., Ginther, D. K., Kahn, S., & Williams, W. M. (2014).
Women in academic science: A changing landscape.
Psychological Science in the Public Interest, 15, 75–141.
Ceci, S. J., & Williams, W. M. (2011). Understanding current
causes of women’s underrepresentation in science. Pro-
ceedings of the National Academy of Sciences, USA, 108,
3157–3162.
Ceci, S. J., Williams, W. M., & Barnett, S. M. (2009). Women’s
underrepresentation in science: Sociocultural and biologi-
cal considerations. Psychological Bulletin, 135, 218–261.
Cheryan, S., Ziegler, S. A., Montoya, A. K., & Jiang, L. (2017).
Why are some STEM fields more gender balanced than
others? Psychological Bulletin, 143, 1–35.
Eccles, J. (1983). Expectancies, values, and academic behav-
iors. In J. T. Spence (Ed.), Achievement and achievement
motives: Psychological and sociological approaches (pp.
75–146). San Francisco, CA: W. H. Freeman.
Eccles, J., & Jacobs, J. E. (1986). Social forces shape math
attitudes and performance. Signs, 11, 367–380.
Ellison, G., & Swanson, A. (2010). The gender gap in sec-
ondary school mathematics at high achievement levels.
Journal of Economic Perspectives, 28, 109–128.
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, 103–127.
Gardner, A. (2016). How important are GCSE choices when
it comes to university? Retrieved from http://university
.which.co.uk/advice/gcse-choices-university/how-impor
tant-are-gcse-choices-when-it-comes-to-university
Guiso, L., Monte, F., Sapienza, P., & Zingales, L. (2008).
Culture, gender, and math. Science, 320, 1164–1165.
Gunderson, E. A., Ramirez, G., Levine, S. C., & Beilock, S. L.
(2012). The role of parents and teachers in the develop-
ment of gender-related math attitudes. Sex Roles, 66,
153–166.
Hyde, J. S., & Mertz, J. E. (2009). Gender, culture, and mathe-
matics performance. Proceedings of the National Academy
of Sciences, USA, 106, 8801–8807.
Imai, K., Keele, L., & Tingley, D. (2010). A general approach
to causal mediation analysis. Psychological Methods, 15,
309–334.
Imai, K., Keele, L., & Yamamoto, T. (2010). Identification,
inference and sensitivity analysis for causal mediation
effects. Statistical Science, 25(1), 51–71.
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 develop-
ment across 53 nations. Archives of Sexual Behavior, 39,
990–997.
National Science Foundation. (2017). Women, minorities,
and persons with disabilities in science and engineering:
2017 (Special Report NSF 17-310). Arlington, VA: National
Center for Science and Engineering Statistics.
OECD. (2016a). PISA 2015 assessment and analytical frame-
work: Science, reading, mathematic and financial liter-
acy. Paris, France: Author.
The Gender-Equality Paradox 593
OECD. (2016b). PISA 2015 results: Excellence and equity in
education (Vol. 1). Paris, France: Author.
Pinker, S. (2008). The sexual paradox: Men, women and the
real gender gap. New York, NY: Simon & Schuster.
Pittau, M. G., Zelli, R., & Gelman, A. (2010). Economic dis-
parities and life satisfaction in European regions. Social
Indicators Research, 96, 339–361.
Reilly, D. (2012). Gender, culture, and sex-typed cognitive
abilities. PLOS ONE, 7(7), Article e39904. doi:10.1371/
journal.pone.0039904
Schmitt, D. P. (2015). The evolution of culturally-variable sex
differences: Men and women are not always different, but
when they are . . . It appears not to result from patriarchy
or sex role socialization. In T. K. Shackelford & R. D.
Hansen (Eds.), The evolution of sexuality (pp. 221–256).
London, England: Springer.
Stoet, G., Bailey, D. H., Moore, A. M., & Geary, D. C. (2016).
Countries with higher levels of gender equality show
larger national sex differences in mathematics anxiety
and relatively lower parental mathematics valuation for
girls. PLOS ONE, 11(4), Article e0153857. doi:10.1371/
journal.pone.0153857
Stoet, G., & Geary, D. C. (2015). Sex differences in academic
achievement are not related to political, economic, or
social equality. Intelligence, 48, 137–151.
Su, R., & Rounds, J. (2016). All STEM fields are not cre-
ated equal: People and things interests explain gender
disparities across STEM fields. Frontiers in Psychology, 6,
Article 189. doi:10.3389/fpsyg.2015.00189
Tingley, D., Yamamoto, T., Hirose, K., Keele, L., & Imai, K.
(2014). Mediation: R package for causal mediation analysis.
Journal of Statistical Software, 59(5). doi:10.18637/jss
.v059.i05
United Nations Development Programme. (2016). Human
development report 2016. New York, NY: Palgrave Mac-
millan.
Universities and Colleges Admissions Service. (2015). Tips
on choosing A level subjects. Retrieved from https://www
.ucas.com/sites/default/files/tips_on_choosing_a_
levels_march_2015_0.pdf
Wang, M., & Degol, J. (2013). Motivational pathways to STEM
career choices: Using expectancy-value perspective to
understand individual and gender differences in STEM
fields. Developmental Review, 33, 304–340.
Wang, M., 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, 770–775.
Williams, W. M., & Ceci, S. J. (2015). National hiring experi-
ments reveal 2:1 faculty preference for women on STEM
tenure track. Proceedings of the National Academy of
Sciences, USA, 112, 5360–5365.
World Economic Forum. (2015). The global gender gap report
2015. Geneva, Switzerland: Author.