Horizontal gender segregation persists and is higher in more developed countries (gender-equality paradox). According to the theory of gender essentialism, more developed countries are characterized by more individualism, which in turn pushes individuals to rely more on personal characteristics when deciding about their future careers. The presence of both essentialist beliefs on gender and the endorsement of the stereotypical beliefs that math and science are “for boys” could explain the so-called paradox. By exploiting the large dataset collected by Project Implicit, the study tested the different associations of gender essentialism and implicit and explicit gender stereotypes with attitudes toward science of young women aged 15–19 in 51 countries. The results confirmed that both essentialist and stereotypical beliefs are negatively associated with attitudes toward science, still with differences in the strength of the association. Furthermore, the association between gender essentialism and attitudes toward science is stronger in more developed countries.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
In 2004, Charles and Bradley published a book highlighting the persistence of horizontal gender segregation, i.e., the different distribution of women and men in job sectors, and the tendency of developed countries to have stronger levels of horizontal segregation compared to developing and transitional societies worldwide. A similar trend was later detected by Stoet and Geary (2018) who showed that more gender-egalitarian countries are those in which the underrepresentation of women in STEM (Science, Technology, Engineering and Math) is stronger. Due to the expected growth and its advantages in terms of salary (U.S. Bureau of Labor Statistics 2021), this sector – more than others in which men and women are not equally represented – has received significant attention from both researchers and institutions (United Nations 2015).
Modernization and economic development are associated with stronger support for gender equality (Inglehart and Welzel 2005). Furthermore, countries, such as the Scandinavian ones, are also those in which often girls outperform boys in math and science literacy (OECD 2023). Therefore, one would expect that wealthier and more developed countries, that excel in gender equality, should be those in which the issue of the unbalanced representation of women in STEM careers is closed. On the contrary, in Europe, the Scandinavian countries have larger gender gaps in STEM degrees than other, poorer countries and this pattern also extends outside the Western region. This counterintuitive result led Stoet and Geary to call this phenomenon the gender-equality paradox.1 For instance, in the Scandinavian countries where girls scored slightly better than boys in both math, science, and reading (OECD 2023), the presence of women among the STEM graduates in Norway (29%) and Finland (30%) was in 2021 below the OECD average, equal to 34% (OECD 2022). The representation of women was, instead, higher in countries such as Romania (43%) and Poland (41%).
Anzeige
Two main theories tried to explain this association, one relying on the concept of gender essentialism (Charles and Bradley 2009), and the other on the evolutionary approach (Geary 1998). According to the first theory, economic development and modernization push societies toward more emancipative, individualistic, and progressive values which attribute large importance to people’s self-expression. Primary identities – and gender among all – acquire more relevance as they help people express themselves, which in turn implies that people rely more on essentialist beliefs, i.e., the existence of innate sex differences influencing interests, preferences, and abilities. On the contrary, according to the evolutionary approach women and men have evolved developing innate sex differences in preferences, personalities, and interests. The rise of gender-egalitarian norms in modern societies has increased people’s freedom to express their intrinsic, distinct, inner preferences. This would thus explain why more gender-egalitarian societies are also those experiencing stronger gender segregation.
A few studies tested the hypotheses of the two theories to explain the differences in countries’ levels of gender segregation. Focusing on the gender-essentialist hypothesis, previous studies measured gender essentialism as the endorsement of gender stereotypes or beliefs about gender roles, rather than actual gender-essentialist beliefs. This study aims to test whether and how gender essentialist beliefs are associated with young women’s relationship with science above and beyond other gender stereotypical beliefs, thus using a proper measure for gender essentialism.
1.1 Gender segregation and the gender-equality paradox
As stressed by the work of Charles (Charles 2003; Grusky and Charles 1998), gender segregation, i.e., “the uneven distribution of women and men across occupations” (Charles 2003, 267) is not unidimensional. The author claims that we can distinguish between two related but distinct dimensions, i.e., vertical segregation – women’s and men’s differences in social status within the manual/non-manual sectors – and horizontal segregation – differences in the distribution of women and men across the manual/non-manual divide. This distinction is relevant and necessary especially when we try to understand why gender segregation persists despite the general improvement in women’s condition (Dilli and Rijpma 2021; England et al. 2020).
Thanks to the widespread increase in the number of women in tertiary education, vertical gender segregation in education remains an issue only in some world regions (UNESCO-IESALC, 2021). On the contrary, it is still present in occupations (Charles 2003; Longarela 2017). Women are frequently overrepresented in less remunerative sectors and in roles that are attributed lower social status and power, which contributes to explaining the persistence of the gender pay gap (Leythienne and Pérez-Julián 2022). Focusing on horizontal segregation, empirical evidence suggests that, despite the progress in gender-egalitarian norms, women are still largely represented in more “feminine” sectors but underrepresented in more “masculine” sectors (Cech 2013; Charles 2003; Charles and Grusky 2004; Cohen 2013). Furthermore, comparative studies highlighted a surprising trend associating economic development and gender egalitarian norms with horizontal gender segregation. More developed and gender-egalitarian societies are also those with higher levels of horizontal gender segregation, while transitional and developing societies, characterized by less gender equality, have lower levels of horizontal segregation (Charles 2003; 2017; Charles and Bradley 2002; 2009). This is similar also to another phenomenon known as the welfare state paradox, according to which high levels of female labor market participation are associated with higher levels of gender segregation in advanced post-industrial societies (Barth and Misje 2024).
Anzeige
Sikora and Pokropek (2012) used PISA data from 2006 to measure gender segregation in students’ career plans in science in 50 countries. They found that girls with stronger preferences toward science were planning a career in Biology, Agriculture, or Health, while their science-oriented male peers were planning a career in Computing, Engineering, or Mathematics. Furthermore, they confirmed the results found by Charles and Bradley on a smaller sample of countries (Charles and Bradley 2002; Charles and Bradley 2009) as they found that the gender gap in both the self-concept in science and preferences for science careers was larger in advanced industrial countries compared to developing or transforming societies.
This unexpected association has been identified especially for STEM, a sector that is still considered strongly masculine and, consequently, male-dominated (Nosek et al. 2002; UNESCO 2017). As found by Stoet and Geary (2018), more gender-egalitarian societies are those in which women’s underrepresentation in STEM is higher, an association now known as the gender-equality paradox. In a recent narrative review, Balducci (2023) revised and compared the results of 31 studies on the association between gender differences in basic skills and personality and gender equality, which is usually found to be negative. The author thus concludes that, overall, empirical evidence supports the hypothesis of a gender equality paradox.
In another recent review conducted using a systematic approach, Herlitz et al. (2024) summarized the empirical evidence on how sex differences vary based on countries’ living conditions, including wealth and gender equality. Results on the association were mixed. Sex differences in some cognitive functions, personality, and negative emotions were found to be larger in countries with higher living conditions, while others, e.g., cognitive abilities in math, self-efficacy in math and computer science, and sexual behavior, were instead negatively associated with living conditions. Furthermore, they also found that not all indicators of living conditions led to the same results, as some were more sensitive in predicting the magnitude of sex differences than others. This is the case, for instance of economic development that was more reliably associated with differences than gender equality. Finally, countries’ wealth and development were found to be associated also with other sex differences, e.g. in the proportion of female chess players (Vishkin 2022).
1.2 The underrepresentation of women in STEM
The underrepresentation of women in STEM is a well-known issue that affects most countries in the world, still with some heterogeneity. Being the technological sector one of the most promising in terms of salaries and employment, increasing the number of women working in this sector is among the main objectives of most countries (European Institute for Gender Equality 2017; U.S. Bureau of Labor Statistics 2021). While in most countries, the percentage of female graduates is higher than that of male graduates in many fields of study, i.e., Education, Health and Welfare, Humanities and Arts, Social Sciences, Business, and Law, still STEM programs are highly male-dominated (OECD 2021).
Research on the STEM gender gap has long tried to understand the reasons for the persistence of this gap. In compulsory education, differences between girls and boys emerge both in interests, behaviors, performance, and educational choices (Tsabari and Yarden 2011; Buser and Oosterbeek 2012; De Gioannis, Bianchi and Squazzoni 2023; Eccles 2011; Jones et al. 2000; Jugovic 2017; Microsoft Philanthropies 2017; OECD 2020; UNESCO 2017). As well synthesized by Dasgupta et al. «What seems like a free choice is constrained by subtle cues in achievement contexts, such as its sex composition, that signal who naturally belongs in STEM and is likely to succeed and who else is a dubious fit» (Dasgupta 2015, 4988). Most studies on the theme emphasized the pivotal role played by gender stereotypes in determining both the existence and persistence of the underrepresentation of women in STEM.
Stereotypes are defined as “general expectations about members of particular social groups […] that lead people to overemphasize differences between groups and underestimate variations within groups” (Ellemers 2018). Despite progress in attitudes toward women, gender stereotypes persist as they are reproduced, often subtly, in various ways, from media communications (Santoniccolo et al. 2023) to AI (Kotek and Sun 2023; Sterlie and Feragen 2024). In the case of math and science, these expectations regard performance, interests, and attitudes toward the scientific sector (De Gioannis 2022b). The endorsement of gender-science stereotypes was found to be associated with lower performance in math (Cvencek et al. 2015; Kiefer and Sekaquaptewa 2007; Ramsey and Sekaquaptewa 2011; Sanchis-Segura et al. 2018), negative attitudes toward math (Nosek et al. 2002; Nosek and Smyth 2011), educational/career intentions and aspirations (De Gioannis 2022a; Schuster and Martiny 2017; Steffens et al. 2010).
1.3 Gender essentialism and the gender-equality paradox
Going back to the gender-equality paradox, the cross-national variability in gender segregation has been explained by two main theories. The first applies the so-called evolutionary approach, according to which gender differences observed in many aspects of life derive from successful adaptations of women and men to environmental circumstances, which in turn determine genetic changes. As stated by Geary (1998, 326), “sex differences in occupational interests and achievement […] appear to be indirectly related to sexual selection, through the sex differences in orientation toward people, status striving, and cognitive pattern, among others”. On the contrary, the gender-essentialist hypothesis claims that modern societies increased the value attributed to individual self-expression and choice, which combined with gender-essentialist beliefs, strengthen the gender typing of curricula choice (Charles and Bradley 2009). While this emphasis on individualism was first present in Western societies, it has now crossed geographical boundaries, as it characterizes also other affluent late-modern societies (Inglehart and Welzel 2005).
According to the last theory, gender essentialism, i.e., the belief that women and men have innate predispositions, abilities, and preferences, is consistent across time and space and gender is the “most automatic, pervasive and earliest learned category used to classify people in social relationships” (Glick and Fiske 1999, 368). Therefore, the celebration of individual choice and the encouragement toward self-expression characterizing modern and developed societies indirectly push women and men to rely on cultural schemas related to their sex, thus strengthening the endorsement of self-stereotyping and encouraging gender segregation. Therefore, “the widespread taken-for-grantedness of essentialist beliefs and the centrality of gender as an axis of individual identity imply a preponderance of gender-typical choices” (Charles and Bradley 2009, 929). Scholars still debate on which of these two theories better explains the gender equality paradox. Just to mention one contribution, in his narrative review, Balducci (2023) concludes that the theory of the evolutionary explanation for the gender-equality paradox does not take into account the presence of socio-cultural elements in the evolutionary process. The current study focuses on the gender-essentialist hypothesis and, in particular, on the STEM sector.
A few studies verified the association between gender essentialism and the gender equality paradox. The theory was tested by its proponents first on advanced industrial societies and then also on transitional and developing countries (Charles and Bradley 2009). More specifically, Charles and Bradley (2009), tested the extent to which nine covariates could explain cross-national variability in horizontal gender segregation in 44 countries, including girls’ relative mathematical affinities. As argued by the authors, the disposition toward mathematics is more gender-typed in advanced, industrial than in developing and transitional societies and this could confirm the hypothesis that gender essentialism interacts with self-expressive ideals acting as a self-fulfilling prophecy. Recently, Breda et al. (2020) constructed a country-level measure of gender essentialism based on the belief that “math is not for girls” using PISA data from 2012 and tested the association between this measure and both GDP per capita and country-level gender difference in intentions to study math. They found that higher levels of GDP are associated with stronger stereotypes and larger gender differences in math intentions, thus confirming other evidence on the gender-equality paradox. However, when the measure of gender stereotypes is included as control, the association between countries’ economic development and students’ math intentions disappears. Finally, at the micro level and in a non-STEM sector, Joyce and Walker (2015) found that horizontal segregation pervades at different levels of insolvency practice and is associated with gender essentialism.
Other studies, instead, tested the correlation between gender differences and individualism, as measured by Hofstede’s cultural dimensions (Hofstede 2001; Hofstede et al. 1991). Among these, Kaiser (2019), Ott-Holland et al. (2013), Schmitt et al. (2008) and Costa et al. (2001) all found that individualism and gender differences in personality traits are positively correlated, consequently, more individualistic countries exhibit larger gender differences in personality traits.
1.4 Critiques against the gender-equality paradox and limitations of current studies
Studies on the gender equality paradox and especially the results reported by Stoet and Geary (2018) have extensively been criticized. The main criticism lies in the erroneous belief that gender equality means gender neutrality, i.e., that in countries where there is a stronger support for equality between women and men this holds in every domain (Noll 2020). As said before, gender stereotypes about math and scientific abilities persist even in countries where gender equality is at its peak. As claimed by the gender essentialist theory, in more developed countries, the persistence of gender stereotypes and the stronger emphasis on individualism can explain why in those countries horizontal segregation is stronger than in less developed countries.
The observed negative association with gender equality could, thus, depend on these other mechanisms that influence the factors of interest. For instance, Mann and DiPrete (2016) tested the association between national performance and self-assessment contexts on gender differences in students’ aspirations for a STEM career. Using country-level data of PISA 2006, they compared the aspirations and performance of 57 countries and found that in high-performance environments, competition is stronger and, consequently, the risk of failure increases for a given level of performance. In such environments, students are likely to require stronger evidence that they are good in math and science before deciding to pursue a career in STEM. However, girls and boys respond differently to competitive contexts, so that net of the powerful offsetting effects of the national performance environment, the gender gap in STEM orientations is generally smaller in countries with more gender-egalitarian cultures, as measured by the Gender Gap Index.
Another limitation regards how the indicators are measured. Among others, Richardson et al. (2020) argued that the association is sensitive to how both gender equality and STEM achievement are measured and that the indexes used by Stoet and Geary — such as the Global Gender Gap Index (GGGI) — are inappropriate, a choice later defended by the two authors in a reply to this critic (Stoet and Geary 2020). Being gender equality a complex phenomenon, the indicators include and collapse several dimensions that might be quite different one from the other (Boulicault 2020). Similarly, Reiches (2020) argued that the assumptions on which Stoet and Geary relied may be not founded, i.e., that (1) higher education systems globally offer the same range of options and similarly apply standardized definitions of STEM, (2) countries with low levels of GGGI are all low-income countries, and (3) STEM degrees pay off in all countries.
As explained by Maria Charles when commenting on the study by Stoet and Geary (Richardson and GenderSci Lab 2020), “[their] findings are consistent with mine as regards patterns of cross-national variability in the gendering of STEM degree programs: I have repeatedly reported that some of the most gender-segregated STEM workplaces are found in reputably gender-progressive countries. Where we seem to disagree is in the idea that gender-progressivism per se is responsible for this segregation”. Therefore, following Charles and colleagues, in this study we tested the gender essentialist hypothesis and verified the association between economic development and gender stereotypes.
However, a similar critique of the measures used could be also made about how previous studies measured gender essentialism. In the two studies mentioned before, gender essentialism was measured as girls’ relative mathematical affinities (Charles and Bradley 2009) and the endorsement of the belief that “math is not for girls” (Breda et al. 2020). Another study testing the relationship between societal affluence and the gender gap in STEM aspirations included a measure of implicit gender stereotypes and suggested that “career aspirations may be more strongly influenced by essentialist gender stereotypes in societies where occupational choice represents a vehicle for self-expression” (Charles 2017, 11). However, in psychology, essentialism refers to laypeople’s belief that social groups have essences that give rise to entities (Medin and Ortony 1989). This implies that gender essentialist beliefs are beliefs in innate, biological differences between men and women. As argued by De Gioannis (2022b), the measures of gender-science stereotypes rarely specify the cause determining the observed gender difference. Therefore, they are not a proper measure of what is defined as gender essentialism.
1.5 Aim and hypotheses
The study aims to contribute to the empirical evidence on the gender essentialist theory by Charles and colleagues, according to which developed countries attribute more relevance to individualism and self-expression and this, in turn, increases the influence of gender essentialist beliefs. Contrary to previous studies, here it was distinguished between gender essentialism, i.e., the belief about innate, biological gender differences, and the endorsement of gender-science stereotypes.
More specifically, it was hypothesized that (H1) gender essentialist beliefs contribute to explaining the low interest in science of young women over and above other gender-science stereotypes, both explicit and implicit. Furthermore, based on the gender essentialist theory (Charles and Bradley 2002; Charles and Grusky 2004), it was hypothesized that (H2) gender essentialist beliefs play a more important role in women’s interest in science in more developed and richer countries.
2 Materials and methods
2.1 Project implicit
The sample was retrieved from the data collected by Project Implicit between 2003 and 2022. Project Implicit is a non-profit organization managed by a group of researchers who are interested in implicit social cognition (Nosek and Greenwald 1998). One of the main activities of the organization is the provision of a “virtual laboratory” for collecting data on stereotypes, both implicit and explicit. The website allows people to take the Implicit Association Test, a psychological test designed by Greenwald et al. (1998) to measure implicit cognition, i.e., the automatic association between a target concept (e.g., gender) and an attribute dimension (e.g., preference for STEM vs liberal arts domains).
The data collected by Project Implicit are shared in open access and can be used to conduct academic research (Xu et al. 2020).2 The data shared in the database are collected in 239 countries and include both the results of the IAT and the answers to a survey that is displayed after the test. The survey includes several questions that are related to the theme of the IAT. In the case of gender and STEM/liberal arts (“Gender-Science IAT”), the survey collects information on demographic characteristics, explicit stereotypical beliefs about gender and science/liberal arts, and attitudes toward science and arts.
2.2 Sample
Since users of the platform are self-selected, the sample collected by Project Implicit cannot be considered representative of the population. Despite this weakness, the data collected by the platform provide unique access to information on gender stereotypes about science that are, so far, not retrievable for such a large and heterogeneous group of countries elsewhere. To test the hypothesis on a sample of countries as broad and heterogeneous as possible, we included all countries for which the platform allows to take the test. However, to ensure the reliability of the data, especially those on the implicit association test, following previous studies using this dataset, we excluded those who did not meet the following criteria (Ackerman and Chopik 2021; Greenwald et al. 2003), i.e., (1) those for whom it was not possible to compute an IAT score, (2) were younger than 15 years, (3) had missing data on country, gender, and on the IAT score, (4) were from countries with fewer than 100 participants, and (5) those for whom in the IAT more than 10% of trials had latencies faster than 300 ms.
Furthermore, since previous studies on the theme focused on educational rather than professional choices, we decided to focus on the younger cohorts, to have a sample similar in age to that of previous studies. This would also allow the exclusion of heterogeneity due to differences among generations in the endorsement of values and attitudes toward gender (Inglehart and Welzel 2005). Since according to the theoretical framework there are no reasons for excluding some countries from the study, we did not specify any inclusion criteria depending on this, and the final list of countries only results from the above-mentioned exclusion criteria. Finally, the analysis was conducted on the female sample only as the theoretical framework of the gender essentialist hypothesis focuses on women. The male sample was included in the sensitivity analysis.
The final sample consisted of 95,044 complete cases of young women aged 15 to 19 years old (M = 18.3, SD = 1.40) coming from 51 countries, as most of the countries included in the original sample collected too few cases. Nevertheless, most respondents came from the United States (83%) and the sample size for individual countries ranged widely, from 43 to 79,855 (mean = 1865; median = 143). The full list of countries and the sample sizes of each country can be found in the Supplementary Material (Table A1).
2.3 Procedure
The study aimed to test the association between the endorsement of stereotypical beliefs about gender and science and people’s relationship with the science domain. Due to the clustered nature of the data, a multivariate model was employed following a multilevel approach, incorporating random effects and robust standard errors to estimate the association of interest. The model estimated was specified as follows:
where y is the exploratory variable (attitudes toward science or liked science in school), x represents gender stereotypes (i.e., gender essentialist beliefs, implicit and explicit gender stereotypes), g represents the country-level indicators of development and C the control variables (i.e., race, age, and year of data collection). Since, according to the theoretical framework, the association between gender stereotypes and their relationship with science should vary based on the country, the model also included an interaction term between gender stereotypes and the country’s development, and a random slope for gender stereotypes (\({u}_{1j}\)).
The analysis was conducted on Stata (StataCorp 2021). To increase statistical power, data from 2007 and 2022 were pooled together. While there might have been a change in the endorsement of gender stereotypes over time (Charlesworth and Banaji 2019), this should not affect the association between stereotypes and attitudes toward science. The theory proposed by Charles and colleagues hypothesizes that the association between gender essentialism and horizontal segregation does not depend on the level of stereotype endorsement, but rather on the level of individualism in the country. Since such shifts in cultural norms typically take time, pooling data from thirteen years should not be problematic (Ackerman and Chopik 2021). Nevertheless, we checked whether the results were sensitive to time.
Furthermore, several additional analyses were conducted to check the robustness of the results. First, the model was estimated without the interaction term on three subsamples, identified by dividing the countries by income group. Furthermore, the model was estimated with several indicators of countries’ development. To avoid multicollinearity issues due to the high correlation among these indicators (see Table A3 in the Supplementary Material), the models were estimated separately for each indicator. Finally, some sensitivity checks were also performed, as reported in the results section.
2.4 Variables
The variables used were all taken from the Project Implicit survey. Their validity and reliability have been extensively tested (see for instance Greenwald et al. 2009; Nosek et al. 2005). While the data on the Project Implicit Demo Site has been collected since 2003, at the end of 2006, the questionnaire underwent major changes. Furthermore, data collected before 2006 almost entirely come from the US. This is why the analysis shown in the results section was performed on data from 2007 to 2022.
The variable of interest was the relationship of young women with science. This was measured using two variables from the Project Implicit survey, i.e.,
Attitudes toward science. Respondents were asked to rate on a 5-point Likert scale (1 = “Strongly dislike”, 3 = “Neither like nor dislike”, 5 = “Strongly like”) their attitudes toward science (variable science).
Liked science in school. Respondents were asked on a 5-point Likert scale (1 = “Strongly dislike”, 3 = “Neither like nor dislike”, 5 = “Strongly like”) “In your education experience, how much have you liked studying science?” (variable likesci).3
As regards gender stereotypes, three measures were used, i.e.,
Gender-essentialist beliefs about science. Respondents were informed that “Women hold a smaller portion of the science and engineering faculty positions at top research universities than do men. The following factors were typically included to explore possible explanations of these differences”. The statement was followed by a list of potential explanations for this gender gap. Among these, only one explicitly mentioned that the difference between the two genders is biological, i.e., “On average, men and women differ naturally in their scientific interest” (variable factorinterest).
Implicit gender stereotypes about science and liberal arts were measured by the IAT score. As suggested by Greenwald et al. (2003), the indicator used was the IAT D score, which can resist both the contamination due to response speed differences and the IAT-score-reducing effect of prior experience with the test. The IAT requires participants to assign the words that appear in the center of the screen to the correct category, either on the left or on the right, by clicking as quickly as possible on the letter “e” or “i”, respectively, on the keyboard. In the case of the gender-science IAT, two types of words (stimuli) appear on the screen, one related to gender (either male or female pronouns) and the other related to college majors (either STEM-related or humanities-related). The two genders and the two types of majors are shown either on the left or the right side of the screen in two combinations. In one round, the combination is “stereotypical”, i.e., male and STEM on one side and female and liberal arts on the other side. In another round, the combination is “counterstereotypical”, i.e., male and liberal arts on one side and female and STEM on the other side. The score is computed as the difference in the time needed to classify the stimuli to the correct side when the combination shown is counterstereotypical and when the combination is stereotypical. The idea is that people with stronger stereotypes require more time to assign the stimuli to the right side of the screen when the combination shown is counterstereotypical. The score ranges from − 2 to 2, where negative values indicate that for the respondent it is easier to associate men with liberal arts and women with STEM (counterstereotypical association) rather than the opposite. Positive values, instead, indicate that for the respondent it is easier to associate men with STEM and women with liberal arts (stereotypical association), while a value around or equal to zero means that the respondent has no preference between the two associations.
Explicit gender-science stereotypes. Respondents were asked to rate the extent to which they associated science with males or females, on a 7-point Likert scale where 1 was “Strongly female”, 4 was “Neither male nor female” and 7 was “Strongly male”.4
Finally, the estimated model also controlled for two demographic characteristics, i.e., age and respondents’ race, to control for potential heterogeneity in cultural norms (Inglehart and Welzel 2005). In the last case, the variable had eight categories, i.e., 1 “American Indian/Alaska Native”, 2 “East Asian”, 3 “South Asian”, 4 “Native Hawaiian or other Pacific Islander”, 5 “Black or African American”, 6 “White”, 7 “Other or Unknown” and 8 “Multiracial”.5 Please note that since the Project Implicit questionnaire began collecting gender data only in 2006 and changed its measurement approach in 2016, we chose to use sex as the indicator for consistency, as it has been measured in the same way across all years.
As regards the country-level data, we used several indicators of a country’s level of economic development, and equality i.e. the log of GDP per capita (World Bank 2023a), the Gini index (World Bank 2023b), the Human Development Index, HDI (UNDP 2023c), the Gender Development Index, GDI (UNDP 2023a) and the Gender Inequality Index, GII (UNDP 2023b). Furthermore, since the gender essentialist theory explicitly refers to the level of individualism, we also estimated a model including the classification of countries proposed by Hofstede on individualism and collectivism (Hofstede 2015).
3 Results
3.1 Descriptives
On average, participants from low or lower-middle-income countries held more positive attitudes toward science and also liked science more in school compared to those in upper-middle and high-income countries (Table 1). On the contrary, richer countries tended to endorse stronger implicit and explicit gender stereotypes about science. As regards gender essentialism, instead, the agreement with the statement that the underrepresentation of women is due to innate differences was higher in upper-middle income countries and similar in lower-middle and high-income countries. The low correlation between the endorsement of gender stereotypes and gender essentialist suggest that these three concepts did not overlap (Table A2 in the Supplementary Material). It is thus worth seeing how the three contribute to explaining young women’s relationship with science.
Table 1
Descriptive statistics
Income group
Low or lower-middle
Upper-middle
High
Total
2049
3305
129,363
134,717
Attitudes toward science
4.10 (0.96)
3.90 (1.00)
3.85 (1.12)
3.85 (1.12)
Liked science in school
4.06 (1.02)
3.89 (1.09)
3.76 (1.21)
3.76 (1.20)
Gender essentialist beliefs
2.31 (1.30)
2.40 (1.28)
2.32 (1.22)
2.32 (1.23)
Explicit gender stereotypes
4.36 (1.30)
4.58 (1.29)
4.57 (1.25)
4.57 (1.25)
IAT D score
0.22 (0.40)
0.32 (0.39)
0.33 (0.40)
0.33 (0.40)
Age
18.40 (1.44)
18.50 (1.43)
18.31 (1.42)
18.31 (1.42)
Ethnicity
American Indian/Alaska Native
2 (0.1%)
21 (0.7%)
791 (0.7%)
814 (0.7%)
East Asian
261 (14.1%)
1164 (38.6%)
5566 (4.9%)
6991 (5.9%)
South Asian
1041 (56.2%)
161 (5.3%)
4159 (3.7%)
5361 (4.5%)
Native Hawaiian or other Pacific Islander
39 (2.1%)
4 (0.1%)
670 (0.6%)
713 (0.6%)
Black or African American
94 (5.1%)
46 (1.5%)
5464 (4.8%)
5604 (4.7%)
White
141 (7.6%)
788 (26.1%)
83,615 (73.9%)
84,544 (71.6%)
Other or Unknown
227 (12.3%)
639 (21.2%)
5769 (5.1%)
6635 (5.6%)
Multiracial
47 (2.5%)
193 (6.4%)
7181 (6.3%)
7421 (6.3%)
Mean and standard deviation in parentheses, except for ethnicity for which number of observations per category and percentage are reported
Table 2 shows the results of the estimated model for both attitudes toward science and liked science in school. Both factors were negatively associated with the endorsement of implicit gender-science stereotypes and the association did not change depending on the economic development of the country, as suggested by the interaction term. On the contrary, the association between the two exploratory variables and explicit gender stereotypes was not confirmed. The same resulted also for the endorsement of gender essentialist beliefs, still in that case the interaction term suggests a different pattern depending on the income of the country. It is thus necessary to look at the marginal effects to interpret the results. These are shown in Figs. 1 and 2, respectively.
Table 2
Regression results
Attitudes toward science1
Liked science in school2
B
SE
95% CI
B
SE
95% CI
Gender essentialist beliefs
0.08
0.08
[− 0.071, 0.234]
0.15
0.14
[− 0.129, 0.438]
Explicit gender stereotypes
0.07
0.07
[− 0.063, 0.193]
0.13
0.15
[− 0.160, 0.426]
IAT D score
− 0.20*
0.08
[− 0.357, − 0.036]
− 0.33*
0.16
[− 0.640, − 0.029]
GDP per capita (log)
− 0.03
0.02
[− 0.070, 0.015]
− 0.02
0.03
[− 0.074, 0.027]
Gender essentialist beliefs # GDP per capita
− 0.02*
0.01
[− 0.034, − 0.004]
− 0.03*
0.01
[− 0.053, − 0.000]
Explicit gender stereotypes # GDP per capita
− 0.01
0.01
[− 0.023, 0.001]
− 0.02
0.01
[− 0.052, 0.002]
IAT D score # GDP per capita
− 0.01
0.01
[− 0.022, 0.009]
0
0.01
[− 0.024, 0.032]
Ethnicity (ref. White)
American Indian/Alaska Native
− 0.10*
0.04
[− 0.183, − 0.022]
− 0.24**
0.09
[− 0.409, − 0.072]
East Asian
0.08***
0.02
[0.050,0.115]
0.17***
0.03
[0.110,0.234]
South Asian
0.18***
0.02
[0.146,0.216]
0.18***
0.04
[0.106,0.248]
Native Hawaiian or other Pacific Islander
− 0.05
0.04
[− 0.133, 0.042]
0.02
0.09
[− 0.151, 0.185]
Black or African American
− 0.23***
0.02
[− 0.263, − 0.199]
− 0.22***
0.03
[− 0.285, − 0.156]
Other or Unknown
− 0.16***
0.02
[− 0.194, − 0.135]
− 0.14***
0.03
[− 0.208, − 0.082]
Multiracial
− 0.01
0.01
[− 0.039,0.017]
0
0.03
[− 0.056, 0.046]
Age
0.01***
0
[0.007,0.016]
0
0
[− 0.008, 0.010]
Year
0.01***
0
[0.005,0.009]
0.02***
0
[0.016,0.026]
Random-effects parameters
Var(gender essentialism)
0.000
0.000
[0.000,0.008]
0.000
0.000
[0.000,0.000]
Var(explicit stereotypes)
0.000
0.000
[0.000,0.000]
0.000
0.000
[0.000,0.000]
Var(implicit stereotypes)
0.027
0.007
[0.015,0.046]
0.019
0.011
[0.007,0.054]
N
95,042
30,819
Gender stereotypes’ indicators standardized by sex, Number of clusters: 51 countries
*p < 0.05, **p < 0.01, ***p < 0.001
1Data from 2007 to 2022
2Data from 2007 to 2016
Fig. 1
Average marginal effects of gender stereotypes on attitudes toward science
Fig. 2
Average marginal effects of gender stereotypes on liked science in school
×
×
Marginal effects suggest that, as already observed from the results of the regression model, the association between the two exploratory variables and implicit gender stereotypes did not vary depending on countries’ wealth. On the contrary, the association varied for both gender-essentialist beliefs and explicit stereotypes. In that case, it was on average around zero for poorer countries, while it increased for wealthier countries. In the case of attitudes toward science, as income increased, gender essentialist beliefs became more relevant than explicit stereotypes, while the strength of the association remained the same for liked science in school. Regardless of gender, among the three indicators of stereotypes, those measured by the Implicit Association Test were the most relevant.
To check the robustness of the results, the model was also estimated separately on three subsamples, classifying countries based on their income (complete results are reported in Supplementary Material, Table A4). As shown in Figs. 3 and 4, the strength of the association between gender stereotypes and attitudes toward science increased as income increased (high-income countries are represented by round points, upper-middle-income countries by square points, and lower-middle-income countries by diamond points). However, we did not find such a clear pattern for liked science in school. In general, the association between the two exploratory variables and both gender essentialist beliefs and explicit stereotypes was not statistically significant in lower-middle-income countries. This confirmed previous results. Furthermore, the size of the coefficient of essentialist beliefs in the low-income group was almost half the size that of high-income countries. Finally, there were some differences in the estimated coefficients for implicit stereotypes in the case of liked science in school. However, contrary to what was found for essentialist beliefs, the income group exhibiting the strongest association between liking science and implicit stereotypes was that of countries with upper-middle income, followed, by high-income countries and lower-middle income countries.
Fig. 3
Coefficient plot for attitudes toward science by income group
Fig. 4
Coefficient plot for liked science in school by income group
×
×
3.2 Sensitivity analysis
The model estimation was replicated using different samples from Project Implicit and different variables, to check its robustness. First, it was replicated on the data collected between 2003 and 2006. In those years, the questions in the survey were formulated differently from those used in the analysis. Still, it is possible to identify a measure for essentialist beliefs. Participants were asked to rate on a 7-point Likert scale the extent to which they agreed with the statement “Males perform better than females in science because of greater natural ability”. There is, thus, an explicit reference to the fact that the difference is due to biological reasons, “is natural”, but in this case, the belief is on ability rather than interest. The association between gender stereotypes and attitudes toward science was confirmed still, it was not possible to test the difference between countries due to the lack of enough observations from countries other than the US (Table A5 in the Supplementary Material).
Second, it was also checked whether the association was present among male participants. The same model was estimated on the male sample (N = 42,700). The results are reported in the Supplementary material (Figure A1) As expected, young men exhibited a positive association between implicit gender stereotypes and attitudes toward science. However, the association was stronger in wealthier countries than in poorer countries. As regards explicit gender stereotypes, the association was negative and not statistically significant except for the richest countries in the sample, where the association was positive and statistically significant. Finally, there was a negative association between gender-essentialist beliefs and attitudes toward science also for young men. Nevertheless, this was statistically significant only in richer countries.
As done in most studies using the Project Implicit dataset, the model was estimated on data collected over a long period, between 2007 and 2022. It was thus checked whether the association changed depending on the period on which it was estimated. Note that here the indicator for economic development was adapted accordingly and computed as the average of that period. Table A6 and Figures A2–A4 in the Supplementary Material show that the pattern was confirmed for all three measures of stereotypes in the three periods, i.e., from 2007 to 2011, from 2012 to 2016, and from 2017 to 2022. The association between attitudes toward science and implicit gender stereotypes was negative and homogenous regardless of the countries’ economic development. On the contrary, it was negative but heterogeneous depending on income for both explicit gender stereotypes and gender essentialism. The only exception was in the period from 2017 to 2022, where the association with explicit stereotypes did not change based on income. Interestingly, there was quite a homogeneity also in the strength of the associations over time.
Furthermore, it was also checked whether the results were different when using other indicators for development. Following previous studies on the theme (Breda et al. 2020; Stoet and Geary 2020), we tested the same regression model substituting GDP per capita with the Gini index, Human Development Index (HDI), and gender equality indicators, i.e., Gender Inequality Index (GII) and Gender Development Index (GDI). The results are reported in Figures A5‐A8 in the Supplementary Material. The strength of the association between gender essentialist beliefs and attitudes toward science increased as the indicator increased when we used HDI and GDI. On the contrary, it decreased when we used the GII and the Gini index. Moreover, since the theoretical frameworks by Charles (Charles and Bradley 2009) explicitly refer to individualism, it was also checked how the associations changed when classifying countries based on the individualism scale proposed by Hofstede (2015). Note that this information was not available for all countries, therefore the sample used was restricted (see Table A1 for a complete list). As shown in Figure A9 the difference in the strength of the association based on individualism was almost null and not statistically significant.
It was also checked whether the negative association between stereotypical beliefs and attitudes toward science remained even when the gender gap in science was attributed to social, rather than biological reasons. We exploited the fact that the question used for measuring gender essentialism also asked participants to rate how much discrimination against women and differences in encouragement could explain the underrepresentation of women in science (for a validation of the measure see De Gioannis 2023). The results are shown in the Supplementary material (Table A7) and suggest that the attribution of the gender gap in STEM to biological or social, external causes mattered. In fact, contrary to what was found with gender essentialism, the attribution of the gender gap to social causes was positively associated with attitudes toward science. Finally, given the strong unbalance in countries’ representation within the sample, the model was also estimated on a subsample that excluded the United States. The results did not change, as shown in Table A8 and Figure A10 of the Supplementary Material.
4 Discussion
The study aimed to test the association between gender-essentialist beliefs, i.e., the belief that gender differences are innate, and the attitudes and interests of young women in science. Based on the gender essentialist hypothesis theorized by Charles and colleagues (Charles and Bradley 2009; Charles and Grusky 2004), it was hypothesized that essentialist beliefs on gender are negatively associated with the outcome of interest but that this association is stronger for young women living in more developed and richer countries compared to those coming from countries with lower income.
The results confirmed both hypotheses. Even controlling for both explicit and implicit gender stereotypes, a stronger endorsement of essentialist beliefs on the gender gap in science was associated with both worse attitudes toward science and a lower level of liking science in school. Furthermore, as hypothesized by Charles’ theory on gender essentialism, more developed and richer countries were also those in which young women relied more on essentialism about gender. This is coherent with what has already been found in previous studies on the theme (Breda et al. 2020). However, contrary to previous studies, here a proper measure for gender essentialism was included. The distinction between stereotypical beliefs also revealed interesting patterns. Implicit gender-science stereotypes contributed the most to the attitudes of young women towards science. Still, gender-essentialist beliefs seemed to play a more relevant role than explicit gender-science stereotypes. This suggests that not all stereotypical beliefs are equal and therefore, a distinction should be made.
While several studies have already tested the pivotal role played by gender stereotypes, only rarely the measured beliefs were essentialist, i.e., specified that the cause for the gender differences in science is due to biological or innate predispositions (De Gioannis 2022b). This is however relevant. As shown in the results, the belief that the underrepresentation of women in the scientific sector is due to social, thus external causes did not have the same effect as the belief that this gap is due to innate differences. In the first case, the association was positive, i.e., young women had more positive attitudes toward science, while in the second case, the association was negative, i.e., young women exhibited more negative attitudes toward science. This implies that attribution matters in the context of gender-science stereotypes. However, current measures of gender stereotypical beliefs only rarely specify the cause for the gender gap in science, and, more importantly, existing studies are not even interested in distinguishing between the types of attribution. In this regard, it is worth mentioning here the existence of a few scales that were developed to measure essentialism (Haslam et al. 2002; Rangel and Keller 2011) and essentialism on gender (Lee et al. 2020).
Given the opposite effects that the attribution of the gender gap proved to have, future research should refine how stereotypical beliefs are measured. As suggested here, the distinction between different attributions, as well as different types of stereotypical beliefs (as done in this study) could shed light on the complex effects of stereotypes in the context of STEM. Furthermore, it could also inform future interventions aimed to increase women’s representation in this sector. For instance, previous studies on the influence of gender essentialism found that the endorsement of essentialist beliefs about gender influences both interest in STEM (Donovan et al. 2019) and performance in a math-related task (Dar-Nimrod and Heine 2006; Moè 2012).
The results also contribute to the ongoing debate on the gender equality paradox and its strong dependence on the indicators that are used to measure countries’ development (Boulicault 2020; Richardson et al. 2020). As suggested by the additional analysis including multiple indicators of both development and gender equality, results were sensitive to the indicator. Further research could shed light on how we should interpret this heterogeneity, and what are the implications for this. It is worth noting that when individualism was taken into account, the heterogeneity in the association between stereotypes, including essentialist beliefs, and attitudes toward science was not confirmed. This in part contradicts the hypothesis of Charles and colleagues (Charles and Bradley 2009; Charles and Grusky 2004) that explicitly linked heterogeneity to individualism.
Having said this, the study also has some limitations. First of all, data from Project Implicit are not representative of a country’s population because respondents were not randomly assigned to the survey, rather they somehow got to know the Project Implicit website and voluntarily decided to take the test. Therefore, the scores might reflect this self-selection of respondents. Furthermore, the measures of both essentialist beliefs and explicit stereotypes were constructed on a single item, due to the limitations of the survey. A battery of items would allow the construction of a more reliable measure. The question that is used to assess explicit gender stereotypes, i.e., the extent to which the respondent associates science with males and females is also quite generic and ambiguous (De Gioannis 2022b). This could explain the inconclusive results on explicit stereotypes that emerged in the analysis. Despite its limitations, data from Project Implicit provide unique insights into implicit cognition worldwide that would be difficult to collect. Furthermore, since the questionnaire and the test are the same, the responses collected by Project Implicit allow us to compare different countries.
Finally, the dataset of Project Implicit mainly consists of responses coming from the US. The representation of other countries is strongly unbalanced, and this could have affected the results of the regression analysis. As shown in the sensitivity analysis, when we remove the US from the sample, the results suggest that all three measures of stereotypical beliefs were negatively associated with the outcomes of interest, but the association can be confirmed only for essentialist beliefs. However, these results should be interpreted with caution, as they could be driven by the reduction in the sample size. Furthermore, there is a long debate on what the implicit association test is actually measuring (Jost 2019) and this could also depend on the country and the cultural context in which it is measured. Indeed, previous studies conducted on the dataset from Project Implicit mainly focused on the US.
5 Conclusion
To conclude, the current study confirmed the association hypothesized by Charles and colleagues between gender essentialism and the attitudes of women towards a masculine sector such as that of science. Young women endorsing stronger essentialist beliefs on gender were characterized by worse attitudes toward science, above and beyond the endorsement of other, implicit and explicit stereotypical beliefs. However, this association was stronger in more developed countries.
The findings also suggest that more attention should be put on how gender stereotypes are measured. The nuances in the endorsed beliefs may have different consequences on girls’ and women’s attitudes and behaviors. Furthermore, the recognition of the distinction between essentialist and non-essentialist beliefs could inform the design of future interventions aimed at reducing the underrepresentation of women in STEM. As shown in the study, a different approach should be adopted based on the economic and cultural characteristics of the country.
Declarations
Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Here we adopted the definition of gender proposed by Heidari et al. (2016) as the "socially constructed roles, behaviours and identities of female, male and gender-diverse people". On the contrary, "sex is a multidimensional biological construct based on anatomy, physiology, genetics, and hormones. (These components are sometimes referred to together as “sex traits.”)" (Bates, Chin, and Becker 2022). While we recognize that gender is not only binary, the theoretical framework the study referred to often considered sex rather than gender and does not discuss how other gender categories could be affected by these cultural norms. Making clear hypothesis regarding these categories would have thus be complicated. The same applies to the issue of intersectionality. We hope that further research in this area will also consider non-binary people and—more broadly—other elements of identity using an intersectional approach.
Before 2007, the question was evaluated on a 5-point Likert scale where -2 was "Strongly male" and 2 "Strongly female". This variable was used in the regression model on data from 2003 to 2006, however, the coding was reversed to be equal to the 7-point scale, i.e., higher values represent stronger explicit stereotypes.
The categories of the question about race changed in 2016. The variable included in the model combines the variable raceomb and raceomb_002. The classification was manipulated to be coherent, i.e., "More than one race—Black/White" and "More than one race—Other" were classified as "Multiracial".
