Large-scale Assessments in Education volume 7, Article number: 10 [2019] Cite this article
17k Accesses
8 Citations
7 Altmetric
Metrics details
Abstract
Understanding why women are consistently underrepresented in STEM fields has been a constant puzzle, with a consistent feature of the puzzle being performance in math. This study uses data from TIMSS exams to investigate cross-national gender differences in math-related affect, more precisely liking mathematics, confidence in mathematics, and valuing mathematics. We compared fourth and eighth graders to track any differences in these gender-related affective characteristics. Our findings suggest that despite the variability and some changes to the magnitude and direction of gender differences in math affect, boys and girls are similar. We also found that cross-national sociocultural, political, and educational equality of adults does not necessarily predict positive affect for both genders. In fact, the researchers found that some countries with a smaller adult gender gaps have students with higher gender differences in mathematics-relevant affect.
Introduction
The problem of female underrepresentation in mathematics and science related careers has been an ongoing concern despite the societies’ efforts to facilitate the process for women to hold “male-dominated” jobs [Ceci and Williams 2011; Frome et al. 2006; Wang et al. 2013]. In the United States, for instance, based on NGCP [National Girls Collaborative Project 2018], cited from Science and Engineering Indicators 2016 [National Science Board [US] 2018], women earned 50.3% of science and engineering bachelor’s degrees. However, women’s percentage in science and engineering at the undergraduate level significantly differs by the fields of study: women receive over half of bachelor’s degrees awarded in the biological sciences, they receive 17.9% of bachelor’s degrees in computer sciences, 19.3% in engineering, 39% in physical sciences, and 43.1% in mathematics.
Based on the same report by NGCP, however, women are still underrepresented in the STEM labor force. The most imbalance exists in the fields of engineering, computer science, and the physical sciences with only 29% of the science and engineering jobs held by women.
Literature review
There is a voluminous literature on possible reasons for the female underrepresentation in STEM fields. What follows is only some of the explanations for the issue. Relatively, some of the older studies relied on evidence related to physical and cognitive differences, such as differences in brain size [Romanes 1887] and assumed evolution-based explanations like male superiority in spatial skills [Levine et al. 1999; Linn and Petersen 1985; Voyer et al. 1995]. While there are some physical differences between the brains, the research findings on cognitive distinctions are inconclusive [Hill et al. 2010]. For instance, Lynn and Irwing [2004] in their study of gender differences in the Standard and Advanced Progressive Matrices found that there was no difference among children aged 6–14 years on the progressive matrices. They also found that males obtain just slightly higher means from the age of 15 through the old age on the progressive matrices.
Some other researchers found differences in cognitive skills with females outperforming males in certain tasks and vice versa. For example, Hedges and Nowell [1995] found that males, on average, are disadvantaged in reading and writing skills. They found that females slightly outperformed males in tests of reading comprehension, perceptual speed, and associative memory. On the other hand, males slightly, outperformed females in the tests of mathematics and social studies.
Still some other studies found that the existing gender differences, such as differences in spatial and visualization skills in which boys usually outperform girls, could decline and disappear with appropriate training [Baenninger and Newcombe 1989; Sorby and Baartmans 2000; Vasta et al. 1996].
More recent explanations relate these differences to socio-cultural environments and gender stereotypes that support the channeling of young girls away from STEM studies [Beede et al. 2011; Else-Quest et al. 2010; Guiso et al. 2008; Hyde and Mertz 2009]. For instance, Nosek et al. [2002a] found that gender stereotypes that mathematics is for males were related to identification with and attitudes towards mathematics, with stronger gender stereotypes corresponding with more negative mathematics attitudes for women but more positive attitudes for men. They suggest that fundamental categorization of ‘males’ and ‘females’ produces identification with one’s social group which consequently shapes and is shaped by experiences that are expected of that group by the society [p. 57]. The gender stereotypes are not necessarily explicit and could be subconscious hypotheses and expectations of men’s and women’s careers [Nosek et al. 2002b; Valian 1999].
Gender stratification of educational and occupational opportunities [also called gender segregation and gender inequality] has also been recognized as a social factor that could explain women underrepresentation in STEM fields. It has been argued that women and men equality in having access to higher education and job opportunities is positively related to mathematics achievement [Baker and Jones 1993]. Guiso et al. [2008] found support for the role of gender stratification in mathematics test performance cross-nationally. This hypothesis has also been supported strongly in reading achievement and partially supported in science achievement [Reilly 2012].
In a similar vein, self-perception of abilities in STEM school subjects, shaped by the children’s cultural and social milieu and gender stereotypes, could impact women’s interest in STEM careers [Correll 2001, 2004]. Correll [2001], analyzing National Educational Longitudinal Study dataset on high school students, found that males have higher assessment of their own mathematical competence than their female counterparts with the same math grades and test scores. The reverse was observed for verbal skills; female students had higher assessment of their verbal skills than male students. Moreover, she found that the self-assessments of mathematics competence had an impact on both males’ and females’ decisions to continue the path towards quantitative careers [e.g. enrolling in calculus and choosing quantitative college majores]. Ganley and Lubienski [2016] in their analysis of data from the a nationally representative sample of students in the United States over 5 years [3rd–8th grades] found that that girls are less confident and less interested in mathematics than boys across third through eighth grades. Similarly, Cvencek et al. [2011] claim that children as early as second grade demonstrate their understanding of the American cultural stereotype that math is for boys. They also found that elementary school boys identify with math more than girls.
Gender difference in science course selection patterns has also been observed for young children; Farenga and Joyce [1999] found that students perceived physical science and technology-related courses as appropriate subjects for boys and life sciences as appropriate for girls to study. Turner et al. [2008] in their study of gender differences in vocational personality types [Holland 1997] of eighth and ninth grade students found that boys had significantly greater Investigative vocational personality scores than girls associated with enjoying studying and solving mathematics and science problems and valuing science and mathematics.
Finally, some other researchers have focused on the society’s negative reactions to women in the workplaces that are perceived as male dominated [Heilman and Okimoto 2007; Heilman et al. 2004]. Heilman et al. [2004] found that women who were acknowledged to have been successful in the STEM workplace were less liked than equivalently successful men. Moreover, they found that being disliked can have impacts on the career in terms of overall evaluation and for recommendations for reward allocation. These findings further highlight the impact of social and cultural factors in women decision to choose or stay in ‘male gender-typed’ jobs.
Regional differences in TIMSS: outliers or a cluster
In IEA’s Trends in International Mathematics and Science Study [TIMSS] 2015 results, a few countries, mainly from the Middle East, are [apparently] outliers in terms of gender difference in achievement [International Association for the Evaluation of Educational Achievement [IEA] 2017]. The relationship between socio-economic and educational gender equities and girls’ achievement has been negative in this set of nations. In other words, while they are usually ranked low in terms of general gender gaps [e.g. World Economic Forum’s Global Gender Gap Report 2019], girls outperform boys in these countries. This observation is in contrast with gender stratification hypothesis which anticipates a positive relationship between general gender equity and scholastic achievement equity [for more information on gender stratification hypothesis see Fiorentine 1993; Kane 1992; Baker and Jones 1993].
The girls’ higher achievement in some of the Middle East nations also shows consistency across several waves of TIMSS. For instance, in TIMSS 2015, Saudi Arabi, Oman, Jordan, Bahrain, and Kuwait are among the countries with highest gender difference in mathematics achievement in favor of girls. In TIMSS 2011 [Arora et al. 2012], in addition to the above-mentioned nations, girls in Qatar, Yemen, and Palestinian Nat’l Auth outperformed boys. This trend also exists for the results of TIMSS 2007 [Gonzales et al. 2008] and to a considerably lesser extent for TIMSS 2003 [Mullis et al. 2004].
The literature related to some of these countries reveal that the gender gap for these countries is not limited only to international tests of mathematics and science. The girls have a better performance than boys in other subjects in secondary and tertiary education. There is also a big difference in the proportion of girls and boys at universities in these countries. For instance, 70% of students in tertiary education in the United Arab Emirates are women and at Jordan’s largest university, women to men ratio is two to one [Abdulla and Ridge 2011; AlSindi 2013; Ripley 2017; Ridge 2010]. Based on Ripley [2017], fewer than one in every five workers in most of these countries is woman which is, as she claims, against the conventional sense in the Western nations: more female graduates must result in more employment for women but it is not what happens in these countries. This highlights the importance of cultural and motivational considerations and a need for some form of classification in cross-national gender studies.
Tables 11, 12 in Appendix 1 represents the mathematics achievement of the 10 countries with the highest gender gap based on Global Gender Gap Report and achievement data availability in TIMSS 2015. Among fourth graders, in eight countries girls outperformed boys in math achievement tests from which four of the differences were statistically significant. For eighth graders, out of the ten high-gap countries, nine countries had girls outperforming boys with two of them significantly different. It is noteworthy to mention that in both grades, all countries except Republic of Korea are in the Middle East.
Current study
The current study examined gender differences in mathematics-related affect among fourth and eighth grade students cross-nationally using a recent international database. Large and nationally representative data can provide more reliable and generalizable findings to decide on gender similarities and/or differences compared to smaller and selective samples [Reilly 2012]. Moreover, the use of secondary national and international data has been a common practice in gender studies [e.g. Guiso et al. 2008; Wiseman 2008]. More precisely, we investigated gender difference in confidence in mathematics, liking mathematics, and valuing mathematics. We were interested in investigating the possible affective gender difference related to mathematics in both elementary and middle schools using large and representative samples.
Additionally, this study took a bird’s eye view of gender, affect, and math using meta-analysis procedures. Compared to the regular meta-analysis which endorses the use of studies as units of analysis, the present study used nations as the units of analysis [Else-Quest et al. 2010; Reilly 2012]. We meta-analyzed data from the International Association for the Evaluation of Educational Achievement’s TIMSS 2015 to assess the magnitude and direction of gender differences in mathematics related affect.
The analysis was conducted for both fourth and eighth grade students. Previous research found some evidence of gender difference in mathematics test performance in high school and college but not elementary school [e.g. Hyde et al. 1990]. One of our goals was to see if there was a gender difference in mathematics affect between fourth and eighth grade participants.
We were also interested in investigating how countries gender disparity in socio-economic and cultural areas may affect their students’ attitude towards math by putting countries in different categories based on the gender gap indices. We used World Economic Forum’s Global Gender Gap Report [GGGR] 2017 to achieve this goal. This report includes rankings and indices of 144 countries on four dimensions including economic participation and opportunity, educational attainment, health and survival, and political empowerment [see World Economic Forum 2019].
In sum this study was conducted to answer the following questions:
- 1.
Is there a gender difference in students’ interest in mathematics [Liking Mathematics] cross-nationally?
- 2.
Is there a gender difference in students’ mathematics self-confidence cross-nationally?
- 3.
Is there a gender difference in how much students value mathematics cross-nationally?
- 4.
How is the general gender gap index [including education, economic participation, political representation, and health] associated with the gender difference in interest in mathematics, mathematics self-confidence, and valuing mathematics?
- 5.
Is there a significant difference between fourth and eighth grade students in terms of mathematics-related affect?
Method
Data sources
The data for mathematics affect came from TIMSS 2015. TIMSS is a set of international examinations measuring achievement in mathematics and science. It is sponsored by International Association for the Evaluation of Education Achievement [IEA] and developed by TIMSS & PIRLS International Study Center at Boston College. It is administered every 4 years. At the time of conducting this research, the 2015 wave was the most recent administration. The data were based on students’ answers to three sets of items comprising scales in the TIMSS 2015 international database available to the public. The scales were Students Confident in Mathematics scale, Students Like Learning Mathematics scale, and Students Value Mathematics scale. Each of these scales had nine statements which were scored on a four-point scale. The choices for each statement were ‘Agree a lot’, ‘Agree a little’, ‘Disagree a little’, and ‘Disagree a lot’. The scale statements across grade levels were not completely identical [see Appendix 2 for the items of each scale at grade 4 and grade 8].
In TIMSS 2015, 57 countries and 7 benchmarking entities [regional jurisdictions of countries such as states or provinces] participated. In total, more than 580,000 students participated in TIMSS 2015. The number of countries and sample sizes for the purpose of the current study are summarized in Table 1.
Table 1 Sample size and number of countries in each analysis
Full size table
Gender parity measures came from Global Gender Gap Report 2017 [GGGR 2017]. It was established by World Economic Forum in 2006 to provide a picture of global gender equality. GGGR 2017 provides a Global Gender Gap Index [GGGI] for each of the major and emerging world economies that was utilized in the current study. GGGR 2017 included 144 countries. GGGR measures gender equality in four areas including economic participation, education [primary, secondary, tertiary], political representation, and health [life expectancy ratio and sex ratio at birth].
One of the relatively new concerns in the use of self-reported Likert-based scales data collected from different cultures is whether they have measurement invariance [e.g. Karakoc Alatli et al. 2016]. The measurement invariance refers to a statistical feature of a scale or measurement tool which ensures the same underlying construct is being measured across different groups or times [here across different cultures]. In other words, it is possible that a scale scores is affected by some culture-specific features that the researcher did not intend to assess; in that case the scale lacks measurement invariance. Moreover, the aggregation of the scale data from different cultures may result in puzzling patterns and erroneous inferences [He et al. 2017, 2018]. In the current study, the measurement invariance was assumed for the scales utilized and the inferences made from the findings.
Data analysis
This study is based on meta-analysis procedures and techniques proposed Lipsey and Wilson [2001] for the computation of the effect sizes, the overall effect size, and calculation of heterogeneity [i.e. the variation of effect sizes]. The mean comparison was utilized as the statistic for the calculation of the effect sizes. Microsoft Excel, IBM SPSS, and IEA’s IDB Analyzer were utilized for the purpose of meta-analysis. To calculate the mean effect sizes both Wilson’s SPSS macros [Wilson 2017] and hand-written SPSS syntax were utilized.
To check the homogeneity of effect sizes across nations Q and τ2 [Tau-squared] statistics were calculated. Q values were assessed using the χ2 distribution with k − 1 degrees of freedom, where k stood for the number of countries in the analysis. The τ2 statistic, was assessed from Q random effects analysis. Based upon a rule of thumb, τ2 values greater than 1 indicated heterogeneity.
To merge and calculate the mean and standard deviation of the mathematics data into a single file IDB analyzer was used. IDB analyzer generates SPSS syntax for both merging and the analysis of the data in the TIMSS 2015 international database. It considers sampling design and standard errors using the jackknife repeated replication [JRR] method. It also makes appropriate use of sampling weights in the analysis.
Meta-analysis is a statistical method for synthesizing the results of multiple mean difference calculations to increase the power of estimates from different studies [Lipsey andWilson 2001]. For this analysis, the researchers treated each nation in the TIMSS database as a separate study of gender differences. For each nation, the researchers calculated the effect size [ES] as the mean difference between girls and boys divided by within-country pooled standard deviation. Negative ES values represented the boys’ advantage over girls regarding the construct. Effect size refers to the difference of two groups in the standard deviation unit. As a generally accepted guideline, the effect size below 0.2 is considered negligible, effect size of 0.2 is small, 0.5 is moderate, and 0.8 and higher are large [Cohen 1988].
Besides the meta-analysis of all nations together to answer research questions 1–3 [i.e. gender difference in students’ interest in mathematics, gender difference in students’ mathematics self-confidence, and gender difference in how much students value mathematics], the researchers also decided to select two sets of countries as high-gap and low-gap countries and do separate meta-analysis on each to answer the research question 4. We thought this procedure could provide a better picture of the role of socioeconomic and cultural differences related to mathematics affect. The selected countries, with lowest gender gap, based on the GGGR 2017 report with TIMSS 2015 data available, are as follows: Norway [2], Finland [3], Sweden [5], Slovenia [7], Ireland [8], New Zealand [9], France [11], Germany [12], Denmark [14], and Canada [16]. The numbers in the parentheses are the countries global ranking of gender equality. The selected countries with the highest gender gap are: Iran [140], Saudi Arabia [138], Morocco [136], Jordan [135], Turkey [131], Qatar [130], Kuwait [129], Bahrain [126], United Arab Emirates [120], and Republic of Korea [118].
Results
The meta-analysis of Student Like Mathematics scales revealed that there was almost no gender difference in interest in mathematics between fourth graders [grand random mean ES = − 0.073, grand fixed mean ES = − 0.065]; the effect sizes were heterogeneous [Q[47] = 1596.04, p < 0.001] and the τ2 [between nations true heterogeneity] was 0.024 [see Table 1].
The meta-analysis of gender difference in interest [i.e. Student Like Mathematics scales] for eighth graders showed that there was a slight gender difference favoring male students [Grand Mean ES = − 0.106, see Table 2]. The effect sizes were heterogeneous [Q [39] = 935.85, p < 0.001]. The τ2 value was .014.
Table 2 Overall effect size of gender differences in “Liking Mathematics” for fourth and eighth graders
Full size table
Regarding the unweighted effect sizes, for fourth graders, 15 countries out of 48 countries [31%], and for eighth graders 10 out of 40 countries [25%] had small to medium gender disparity in “liking mathematics” [Tables 3 and 4]. In other words, they had ds of 0.2 or more but less than 0.5 which were either positive or negative [Cohen 1988].
Table 3 Unweighted effect sizes and descriptive statistics for “Liking Mathematics” for fourth graders
Full size table
Table 4 Unweighted effect sizes and descriptive statistics for “Liking Mathematics” for eighth graders
Full size table
Regarding mathematics self-confidence both fourth and eighth grade boys were slightly superior than girls and this difference were statistically significant [p