Gender differences in educational outcomes and the effect of family background: A comparative perspective from East Asia

Abstract

There is wide variation in the degree of gender gap in test scores around the world, suggesting the strong influence of institutions, culture and inequality. We present comparative evidence on the gender gap in educational achievement in China, Japan, and the USA, with an emphasis on the gender-specific effect of parental income and education, and the child’s own preferences for study subjects. We used three major national representative longitudinal surveys with rich information about cognitive outcome measures of respondent children as well as educational investment and parental socio-economic status that allow us to analyze their inter-relationship. We found that low household income tends to have more adverse effects on language test scores for boys than for girls in the USA, as is consistent with previous studies. However, it does not have an impact on gender gap in test scores in China and tends to affect girls more adversely than boys in Japan.

Keywords

Education cognitive ability gender children

Introduction

There has been a considerable improvement in women’s educational attainment worldwide compared to men in the past half century. Today, there is little difference in the years of schooling between men and women in advanced countries, and in some countries, women have even more years of schooling (DiPrete and Buchmann, 2013; OECD, 2015). Studies, however, show that women are still at a disadvantage in the labor market, lagging behind in wages, employment status, and career paths (OECD, 2017).¹ Therefore, years of schooling as a measure of human capital is insufficient as an element to explain the gender gap in socio-economic outcomes still remaining in developed countries.

A strong candidate for the direct explanation of the gender gap is differences in the choice of academic major, especially regarding the so-called STEM fields (Science, Technology, Engineering, and Mathematics). In most countries, there are more male students majoring in STEM fields in higher education curricula than female students. If the wage level in the STEM fields is higher than in the non-STEM fields, with the years of schooling fixed, this gender difference in choice of major explains, to a certain extent, the existence of gender-based wage differences (Altonji and Blank, 1999).

Why are female students less likely to major in STEM fields of study than male students? If differences between men and women in biologically based abilities and preferences for the fields of math and science (assuming these exist) affect the choice of major, the gender differences in math test scores and preferences must be observed consistently in any society or time period. However, recent research increasingly challenges such a view. DiPrete and Buchmann (2013) found that males took more advanced math and science courses than females in secondary school in the 1970s, but today females take more advanced STEM courses than males. At the same time, while the gender difference in the average test score of STEM subjects is narrowing in the USA, there is still a gender-based difference in the test scores, at least at the upper end of the distribution (Pope and Sydnor, 2010; Xie and Shauman, 2003).

Interestingly, there exists wide variation in the degree of gender gap in math test scores across societies. The Organization for Economic Development (OECD) (2015: Figure 1.3) shows that, based on Program for International Student Assessment (PISA hereafter) 2012, although boys outperform girls in about three-fourths of the listed countries, the degree of such gap varies: some countries, such as Japan and Liechtenstein, still show a significant advantage for boys, but there is no statistically significant gender gap in other countries, including the USA. Among one-fourth of the countries with no advantages for boys, some countries, such as Thailand and Jordan, show girls have an advantage over boys in terms of math test scores. The degree of gender gap does not seem to be related to the overall level of educational achievement. High-achieving countries like Singapore and Finland have a very weak, even negative, gender gap, whereas Japan and Liechtenstein show a large gender gap.

In this paper, we explore a cross-national comparison between China, Japan, and the USA regarding the degree to which family socio-economic status (SES) influences the gender gap in test scores, participation in extracurricular activities, and preferences regarding study subjects by using three major longitudinal surveys of children. Our research addresses three broad questions. First, we ask how three countries can be compared in terms of gender differences in educational outcomes during the compulsory education stage. According to Wang (2013), math achievement and math self-efficacy beliefs in high school influence college STEM major entrance, and attitudes toward math affect math efficacy beliefs. Therefore, math achievement and attitudes toward math in earlier childhood may influence the decision regarding STEM major in college. The longitudinal nature of the data sets, with rich family background information, allows us to examine how a gender gap in education develops as children grow up and how such a gap is distributed within the society.

Second, we ask how much gender gap differs by the SES of the family. In recent years, there has been progress in research on the impact of expansion and persistence of economic disparities on children (Duncan and Murnane, 2011; Ermisch et al., 2012). Recent research suggests that family background variables such as household income and parents’ educational levels may have different impacts on children, depending on their genders, and many studies propose that an increase in poverty has a negative impact, especially on boys.² DiPrete and Buchmann (2013) and the OECD (2015) argue that widening inequality is likely to affect boys more adversely than girls; DiPrete and Buchmann (2013) provide an ‘institutional’ interpretation that an increasingly female-friendly school environment with stricter discipline towards misbehavior has advantaged girls more than boys, especially when they are from socio-economically disadvantaged families.³ It would be of great interest whether such a tendency can be observed in other countries with different levels of development and culture. According to the World Values Survey Round Six (Inglehart et al., 2014), the percentage of those opposing the opinion that ‘men should have the right to work rather than women’ was 69.3% in the USA, 37.5% in China, and 14.3% in Japan. Evidently, there is a high awareness of traditional gender role attitudes in Japan. Such differences by country may lead to gender differences both in parents’ educational investment and in children’s educational attainments.

Third, we examine whether the observed differences in parental investment across SES can explain the differences in gender gap in test scores. Although the gender disparity in the overall level of educational achievement narrows, there has been little research conducted on specific behaviors of parents, such as child-related investment. Rich information in the three data sets allows us to explore whether there are any gender differences in parental investment by parental SES across the three societies.

This paper is the first detailed empirical research article which examines the gender differences in both academic performance in and preferences for math and language while also considering the roles of family background and educational investment in Western and non-Western countries.⁴ If, in addition to the average gender difference, there are also gender differences in the ways that various family background and policy variables influence academic ability in some societies, it should be possible to suggest policy interventions that aim to narrow the gap based on that point. This is a key goal of this paper.

It has been found that girls score higher than boys on language tests in all three countries. However, girls score lower than boys in math only in the USA. It has also been found that low household income tends to have a more adverse effect on language test scores among boys than among girls in the USA, as is consistent with previous studies. However, it does not have an impact on the gender gap in test scores in China and tends to affect girls more adversely than boys in Japan. Among other findings, the effect of income on the participation in arts activities is larger for girls in Japan and the USA, but not for those in China.

Previous research

Studies on gender differences in the field of STEM cover a wide variety of research focuses, such as top performers, differences in attitudes towards competition between men and women, geographical differences, the STEM relationship with gender equality, and so on. In the USA, Ellison and Swanson (2010) analyzed data on male and female high school students. Their study confirmed that the gender difference among top students is increasing, and that the number of male students within the top 1% of grades is more than 10 times that of female students. This finding also appeared in Xie and Shauman (2003) and Pope and Sydnor (2010). In addition, among high school students who participate in international competitions on behalf of their country, gender differences are also observed in terms of the schools in which they are enrolled.

Pope and Sydnor (2010) analyzed gender differences in standardized testing in the USA from a geographical standpoint. They confirmed that the difference in test scores between male and female students in the most gender-equal states is less than half that of the most gender-unequal states, and this result is geographically dense. Some studies focus on the relationship between gender differences in educational outcomes and gender equality by using the Gender Gap Index (Guiso et al., 2008; Nollenberger et al., 2016; Rodríguez-Planas and Nollenberger, 2018).⁵ These studies suggest that the social setting and culture in which the child grows up may influence the educational gender gap.

There are not many studies that analyze gender differences in the investment behavior of parents in children’s education. Baker and Milligan’s study (2016) was an exception. They showed that gender differences in such factors as time to read books to children already exist from early childhood by using data from Canada, the United Kingdom and the USA, and they presumed that this leads to gender differences in language ability.

In China, Lai (2010) showed that girls surpassed boys in Chinese and English. Also, girls outperform boys in math until the end of middle school. Gong et al. (2014) examined the gender gap in the test scores in math and Chinese in a poor area in Southwestern China. They showed that in all grades, although the gap is small, boys surpass girls in math but girls surpass boys in Chinese. Also, these gaps become larger as grades become higher. Zhang and Tsang (2015) used data from the Chinese national university entrance exam and found that the average gender gap in math is not significant, but that boys significantly outperform girls in the top distribution. Lu (2018) examined the gender gap in math by using the data of elementary and middle school students in China and PISA and found that boys outperform girls in math in both urban and rural areas. Also, the gender gap in math expands as the grade goes higher, and such expansion is larger than in many other countries.

Previous research showed that there is a gender bias in the educational expectation in Japanese society. Brinton (1993) found that Japanese parents are more likely to want their sons to go to university than their daughters. She proposed two possible reasons for this. The first reason was the parents’ expectation of financial help from their sons in the future. Brinton showed that among parents with at least one son and at least one daughter, 74% expect financial help from their son, but only 6% expect it from their daughters. Her second reason was a gender difference in the value of acquiring higher education. If parents perceive gender discrimination in the Japanese labor market, they expect that the return on education of daughters will be lower than that of sons. Also, a man with more education has an advantage in the marriage market but a woman may not.

In Japan, only a few studies have focused on the gender gap in academic ability and motivation to learn.⁶ Using data from the Trends in International Mathematics and Science Study (TIMSS) and PISA, Hojo (2015) analyzed gender differences in the test scores and learning attitudes in math among fourth, eighth, and tenth graders. Among Japanese students, the study confirmed that gender differences in mean math test scores expanded between the fourth and eighth grades. Isa and Chinen (2014) analyzed academic ability in math and gender differences in motivation for math by using data from third-grade to ninth-grade children in one city. They rarely found significant gender differences in mean math ability at the time of elementary school. Female students also showed higher learning motivation than males until the fifth grade, but the opposite was the case in the sixth grade. Furthermore, male students’ average math ability and motivation to learn math was higher than those of female students in any grades in junior high school, and their learning motivation became significantly higher than that of female students from the eighth grade onwards. Finally, Yasuda (2015) analyzed the difference between male and female college students’ grades using a survey in 2005 and showed that in both humanities and science courses, female students had significantly better grades than male students. To summarize, although it has been pointed out that gender differences are observed in math test scores and learning motivation in Japan, a detailed analysis with controls for the child’s family background and parent economic situation has not yet been conducted.

Data⁷

In this study, we use longitudinal data on children in three countries. We use the China Family Panel Studies (CFPS), which Lyu et al. (2019) also used, for our analysis of Chinese children, and the Japan Child Panel Survey (JCPS), the Japan Household Panel Survey (JHPS), and the Keio Household Panel Survey (KHPS) for our analysis of Japanese children. We use the Early Childhood Longitudinal Study, Kindergarten Class of 1998–99 (ECLS-K 1998) for our USA analysis.

The CFPS has been conducted by the Institute of Social Science Survey at Peking University, China. This survey started in 2010 and is a nationally representative survey of Chinese communities, families, and individuals. Children aged 10 and over in selected households are also respondents in this survey. The collected information includes family background, education outcome, health, and so on.

The JCPS has been conducted by the Panel Data Research Center at Keio University, Japan. This survey is a panel survey targeting samples of children representing Japan as a whole, and it includes abundant information with respect not only to indicators of cognitive and non-cognitive abilities of children, but also to parents’ educational investment in their children and the parents’ SES. The JCPS is a bi-yearly survey of children that started in 2010 as a supplementary survey of the JHPS and KHPS, which targets children who go to elementary school or junior high school and their parents.

The ECLS-K 1998 was conducted by the National Center for Education Statistics in the USA. This is a longitudinal study of children from kindergarten in 1998 to eighth grade, and the information was collected seven times over this period. The children’s parents, teachers, and schools also participated in this survey. The ECLS-K provides information on the development of children’s cognitive and physical abilities, family background, school and classroom environment, teacher’s classification, and so on.

This research study descriptively confirmed the gender differences in the test scores for math and language taking the child’s developmental stage, household income, and parents’ educational background into account. Through simple regression analysis, it will confirm the extent to which family background differences affect gender differences.

As an outcome of cognitive ability at ages 10–15, we use the math and language test scores, which we standardized with zero mean and a unity standard deviation by grade.⁸ We also focus on whether girls are more likely to be in the top (bottom) 10% of each test score or not. Gender difference in educational achievement can show up in both tails of the distribution rather than in the mean level. Urasaka et al. (2010) show that having chosen math in college entrance exams tends to have a positive effect on later income, especially when the university rank is middle or high. In addition, Urasaka et al. (2012) show that the incomes of college graduates who majored in science are higher than those of graduates who majored in social science or humanities. From these results, we infer that whether girls’ math scores are in the top 10% can be an important factor for gender wage differences in the top tail of the distribution. Therefore, we created dummy variables indicating whether each child respondent is in the top 10%/bottom 10% of the whole country. As a non-cognitive ability, in Japan and the USA only for children aged 13–15, a dummy variable (‘really like’/‘like’ = 1, ‘indifferent’, ‘dislike’, ‘hate’ = 0 for the JCPS and ‘very true’/‘mostly true’ = 1, ‘a little bit true’/‘not at all true’ = 0 for the ECLS) was constructed from a 5- or 4-level Likert scale for the question about preferences for math and language.

Previous studies have shown a positive relationship between participating in extracurricular activities and children’s outcomes. Broh (2002) showed that participating in interscholastic sports in the 10th and 12th grades is positively associated with math and English grade and math test score. In addition, participating in a music group has a positive effect on math test score. Lipscomb (2007) also showed that participating in athletics is positively associated with math and science test scores. These results indicate that participating in extracurricular activities can be an important factor for children’s outcomes. If there are gender differences in participating in extracurricular activities, this may cause a gender difference in children’s outcomes. In the CFPS, the JCPS and the ECLS, information on participation in various extracurricular activities was gathered separately by the type of activity, such as ‘study’, ‘arts’, and ‘sports’, at ages 13–15. We used dummy variables representing the participation in each extracurricular activity as the measure of educational investment in study, arts, or sports.⁹

As with other control variables, the parents’ logarithmic before-tax household income (previous year of the survey, in yuan in the CFPS, 10,000-yen in the JCPS, and dollars in the ECLS) was used as the overall economic condition for children; from the information on the educational background of the father and mother, a dummy variable was created to track whether they obtained a high degree of education,¹⁰ and the number of siblings (only for high school students and below), categorical dummy variables for region, a dummy variable for ethnicity (only for the CFPS and the ECLS), and a dummy variable for single father/mother (only for the ECLS) were used.

Results of empirical analysis

Descriptive statistical analysis

Table 1 shows the math and language average test scores by income quartiles. In math, only in the USA do boys perform better than girls in all income classes. In China and Japan, there is no statistically significant difference. In language, in the USA girls outperform boys in all income classes. However, the same result as in the USA is observed only in the higher income classes in Japan. In China, girls outperform boys except for the highest income class at ages 13--15.

Table 1.

The mean math and word test scores by gender and income class.

	China			Japan			USA
	Boy		Girl	Boy		Girl	Boy		Girl
(a) Math test scores
Age 10–12
Income class
1st Q	−0.742		−0.747	−0.743		−0.763	−0.704	>>>	−0.930
2nd Q	−0.586		−0.639	−0.501		−0.433	−0.342	>>>	−0.499
3rd Q	−0.495		−0.589	−0.403		−0.284	0.049	>>>	−0.146
4th Q	−0.373		−0.462	−0.214		−0.118	0.309	>>>	0.135
Total	−0.562		−0.619	−0.460		−0.417	−0.212	>>>	−0.408
N	831		772	473		452	5256		5188
Age 13–15
Income class
1st Q	0.306		0.345	0.030		−0.039	−0.010	>	−0.095
2nd Q	0.552		0.554	0.172		0.247	0.488	>	0.426
3rd Q	0.635		0.767	0.358		0.482	0.730	>	0.641
4th Q	0.774		0.724	0.496		0.664	0.909	>>>	0.792
Total	0.561		0.584	0.278		0.367	0.490	>>>	0.392
N	796		818	622		588	4166		4147
(b) Language test scores
Age 10–12
Income class
1st Q	−0.716		−0.617	−0.610		−0.590	−0.866	<<	−0.765
2nd Q	−0.489		−0.427	−0.363		−0.336	−0.484	<<<	−0.289
3rd Q	−0.301		−0.228	−0.521	<<<	−0.058	−0.104	<<<	0.029
4th Q	−0.216		−0.065	−0.362	<<	0.049	0.168	<<	0.262
Total	−0.450	<	−0.351	−0.468	<<<	−0.249	−0.363	<<<	−0.236
N	831		772	473		452	5256		5188
Age 13–15
Income class
1st Q	0.138	<	0.321	0.112		0.094	−0.249	<<<	−0.040
2nd Q	0.207	<<<	0.511	0.079		0.215	0.336	<<<	0.565
3rd Q	0.290	<<<	0.619	0.087	<<	0.408	0.607	<<<	0.819
4th Q	0.480		0.571	0.366	<<	0.624	0.897	<	0.963
Total	0.268	<<<	0.498	0.174	<<<	0.362	0.350	<<<	0.521
N	796		818	622		588	4166		4147

Note: >, ≫, and ⋙ indicate that the boy’s test score was higher than the girl’s test score at 0.05, 0.01, and 0.001 levels, respectively. Further, <, ≪, and ⋘ indicate the reverse order at each significance level.

We also confirmed a gender difference in the probability of earning the top 10% or bottom 10% test scores.¹¹ Regarding math test score, in the USA, it is confirmed that boys are more likely than girls to be in the top 10% at almost any age and at any income level and girls from low-income families are more likely to be in the bottom 10% than boys at ages 10–12. In China and Japan, there is little gender difference observed in math. If we look at language test scores for the top 10% or bottom 10%, in Japan, girls are more likely to be in the top 10%, being observed only in the third-quartile group of those aged 10–12. In the other two countries, there are no gender differences in language among those aged 10–12. On the other hand, among those aged 13–15, the probability of being in the top 10% in language is higher for girls in high-income groups in the USA and China. If we look at language test scores for the bottom 10%, at ages 10–12, in almost every group in the USA, boys perform worse than girls. There is no gender difference observed in China. In Japan, girls in the third quartile perform better only at a significant level of 10%. On the other hand, at ages 13–15 in the USA and China, girls outperform boys except those in the fourth-quartile income class. In Japan, girls outperform boys only in the higher income classes.

Table 2 shows the mean of extracurricular activities and preferences for study subjects by income classes. Regarding ‘study’ extra activities, there were gender differences in China and Japan in the third-quartile income class. In China, girls are more likely to attend ‘study’ extracurricular classes than boys, and in Japan, boys are more likely to attend ‘study’ extracurricular classes than girls. In the USA, no difference is observed. In the USA and Japan, girls more often participate in artistic extracurricular activities than boys in any income class. In China, such a trend is not observed. With regard to sports, on the other hand, it is observed that boys in the USA and Japan tend to learn sports after school more often than girls. In China, again, such a trend is not observed.

Table 2.

Gender gap in participation in extracurricular activity and preference for study subject.

	China			Japan			USA
	Boy		Girl	Boy		Girl	Boy		Girl
Study
Income class
1st Q	0.164		0.130	0.381		0.441	0.347		0.332
2nd Q	0.202		0.239	0.493		0.457	0.247		0.222
3rd Q	0.237	<	0.325	0.544	>	0.425	0.205		0.172
4th Q	0.387		0.401	0.537		0.516	0.232		0.209
Total	0.240		0.261	0.491		0.463	0.263	>	0.242
N	822		844	632		601	4137		4129
Art
Income class
1st Q	0.000		0.004	0.077		0.105	0.184	<<<	0.295
2nd Q	0.004		0.012	0.043	<<<	0.198	0.197	<<<	0.354
3rd Q	0.005		0.028	0.061	<<<	0.275	0.222	<<<	0.452
4th Q	0.024		0.054	0.042	<<<	0.335	0.288	<<<	0.512
Total	0.007	<	0.022	0.055	<<<	0.238	0.219	<<<	0.387
N	834		865	632		601	4114		4103
Sports
Income class
1st Q	0.000		0.000	0.142		0.119	0.547	>>>	0.467
2nd Q	0.000		0.000	0.214	>	0.121	0.644		0.624
3rd Q	0.000		0.009	0.170		0.119	0.684		0.667
4th Q	0.012		0.000	0.226	>>	0.104	0.691		0.698
Total	0.002		0.002	0.190	>>>	0.115	0.636	>>	0.603
N	834		865	632		601	4129		4096
Like math
Income class
1st Q				0.303		0.260	0.538	>>	0.477
2nd Q				0.369		0.328	0.576	>>>	0.499
3rd Q				0.440	>>	0.292	0.570		0.570
4th Q				0.458		0.375	0.615	>	0.563
Total				0.397	>>	0.316	0.574	>>>	0.518
N				638		607	4119		4099
Like language
Income class
1st Q				0.265		0.349	0.331	<<<	0.455
2nd Q				0.220	<<	0.371	0.354	<<<	0.566
3rd Q				0.260	<	0.379	0.350	<<<	0.605
4th Q				0.250	<	0.342	0.468	<<<	0.674
Total				0.249	<<<	0.359	0.373	<<<	0.564
N				638		607	4106		4097

Note: >, ≫, and ⋙ indicate that the boy’s estimated probability was higher than the girl’s estimated probability at 0.05, 0.01, and 0.001 levels, respectively. Further, <, ≪, and ⋘ indicate the reverse order at each significance level.

Regarding gender differences in preferences for each study subject—math and language—this comparison is made only between the USA and Japan. In the USA, more boys tend to prefer math than girls in almost all income classes. On the other hand, in Japan, the same result as in the USA is observed only with those in the third quartile. In both countries, girls tend to express that they prefer language more than male students.

Regression analysis

The following regression equation (1) will be the base of the estimation:

y_{ijt} = α + δ_{1} Gir l_{i} + δ_{2} Z_{jt} + δ_{3} (Z_{jt} \times Gir l_{i}) + β X_{it} + ε_{ijt}

(1)

Here, $i$ is an index indicating a child, $j$ is a family, and $t$ is a year. $y_{ijt}$ is a variable expressing outcome and here represents academic ability in math/language, whether the child attends extracurricular activities (dummy), and whether the child likes the subject (dummy). $Z_{jt}$ is a variable that may influence children’s outcome, such as household income or parents’ educational background. Household income is a proxy for the capacity to invest in children’s education when the household faces credit constraint. We suppose that high household income has a positive effect on test scores and participating in extracurricular activities. Also, parental education may indicate the level of aspiration of educational investment. Brinton (1993) showed that more parents with a high level of education want their children to go to university relative to parents with a lower level of education.

$X_{it}$ is a vector of variables representing other attributes of children and includes a grade dummy. $ε_{ijt}$ is an error term, and it is assumed to be clustered at the household level. We estimate this equation using the ordinary least squares (OLS) model when the dependent variable is test score. When the dependent variable is the top 10% or bottom 10% of test scores, participation in extracurricular activities, and preferences for study subject, we use the logit model.

Table 3 presents the gender differences in the six indices of cognitive outcomes based on the regression. For each country, the results from the two models, with and without additional demographic controls other than age dummies, are presented. On average, girls tend to have significantly higher language test scores than boys. Similar results are also observed in both the top 10% and the bottom 10% groups. As for math, boys in the USA tend to do significantly better than girls in both the average and the top 10% groups. As for the bottom 10% group, in both the USA and China, girls are more likely to be in the bottom 10% in math, but in Japan, boys are more likely to be in the bottom 10%. This is in sharp contrast to the other two countries.¹²

Table 3.

Estimated coefficients of girl dummy variable.

	China		Japan		USA
	(1)	(2)	(3)	(4)	(5)	(6)
Outcome measures
Math score	−0.0076	0.0083	0.0562	0.0449	−0.1536***	−0.1432***
Language score	0.1766***	0.1922***	0.2128***	0.1963***	0.1456***	0.1566***
Math top 10%	0.0024	0.0077	−0.0053	−0.0102	−0.0413***	−0.0387***
Language top 10%	0.0471***	0.0499***	0.0393**	0.0351**	0.0144***	0.0171***
Math bottom 10%	0.0203*	0.0178	−0.0403***	−0.0381***	0.0116**	0.0096**
Language bottom 10%	−0.0421***	−0.0445***	−0.0564***	−0.0556***	−0.0375***	−0.0395***
Participation in extracurricular class
Study	0.0164	0.0206	−0.0383	−0.0395	−0.0208**	−0.0214**
Art	0.0124*	0.0141**	0.2036***	0.1920***	0.1647***	0.1672***
Sports	–	–	−0.0816***	−0.0815***	−0.0314***	−0.0298***
Preference for study subject
Like math	–	–	−0.0757**	−0.0828***	−0.0604***	−0.0590***
Like language	–	–	0.1144***	0.1118***	0.1871***	0.1891***
Additional control variables	Only grade/age dummies	Set 1	Only grade/age dummies	Set 1	Only grade/age dummies	Set 1

Note: ***, **, and * indicate that they are significant at the level of 1%, 5%, and 10% respectively. An additional control variable set 1 controls grade dummy, number of siblings, dummy for father being college/university graduate or over, dummy for mother being college/university graduate or over, and household income. The results of sports activities in China are dropped because the percentage attending a sports class is low (about 2%) and many in the sample dropped in the estimation.

Regarding the results for extracurricular activities and preferences for study subject, for ‘study’, boys are more likely to receive tutoring than girls in the USA. On the other hand, no significant gender differences were observed in China and Japan. As for arts, in every country, girls are more likely to attend arts lessons than boys. Regarding sports, boys in the USA and Japan tend to attend sports lessons significantly more often than girls. Concerning preferences for study subject, we confirmed the same results as the descriptive statistical analysis, that more girls than boys tend to prefer language and not to prefer math in the USA and Japan.

Table 4 shows gender differences in the effect of household income on cognitive outcomes. In the USA, language test score is lower for girls from high-income households, which means a low household income tends to have a more adverse effect on language test scores for boys than for girls. On the other hand, in Japan, math and language test scores are significantly higher for girls from high-income households. This trend is also observed in the top 10% and bottom 10% groups. However, it is rarely observed in the other countries, especially in China. Regarding extracurricular activities, in both the USA and Japan, girls from high-income households tend to study arts more significantly than others. On the other hand, in China, gender differences in participation in arts classes are narrower in students from high-income households. Only in the USA do more girls from high-income households attend sports lessons and prefer language than others.

Table 4.

The estimation results of interaction between girl dummy and household income.

	China	Japan	USA
	(1)	(2)	(3)
Explanatory variable	Girl X household income
Outcome measures
Math score	−0.0467	0.2004**	0.0034
Language score	−0.0262	0.3460***	−0.0327*
Math top 10%	0.0081	−0.0193	0.0112
Language top 10%	0.0118	0.0717*	0.0026
Math bottom 10%	0.0037	−0.0700**	−0.0059
Language bottom 10%	0.0143	−0.1289***	−0.0126**
Participation in extracurricular class
Study	−0.0039	−0.0828	−0.0098
Art	−0.0121*	0.1425***	0.0315**
Sports	–	−0.0563	0.0311**
Preference for study subject
Like math	–	−0.1007	0.0083
Like language	–	−0.0152	0.0333**
Additional control variables	Set 1	Set 1	Set 1

Note: ***, **, and * indicating that they are significant at the level of 1%, 5%, and 10% respectively. Additional control variable set 1 controls grade dummy, number of siblings, dummy for father being college/university graduate or over, dummy for mother being college/university graduate or over, and household income. The results of sports activities in China are dropped because the percentage attending sports class is low (about 2%) and many in the sample dropped in the estimation.

Table 5 shows gender differences in the effects of mother’s and father’s education on both cognitive outcomes. In the USA, female students whose fathers are university graduates or higher tend to have significantly lower language test scores in the overall sample than others. In contrast, in China and Japan, no significant difference in test scores was observed. In the USA, more girls with highly educated mothers tend to attend arts extracurricular activities and prefer language than others. On the other hand, in China and Japan, there are no significant gender-specific differences by parental educational background in participation in extracurricular activities, and more girls with highly educated fathers tend not to prefer math than others.

Table 5.

The estimation results of interaction between girl dummy and parental educational background.

	China		Japan		USA
	(1)	(2)	(3)	(4)	(5)	(6)
	Girl X parental educational background		Girl X parental educational background		Girl X parental educational background
Explanatory variable	Girl X father high school or above	Girl X mother high school or above	Girl X father college or above	Girl X mother college or above	Girl X father college or above	Girl X mother college or above
Outcome measures
Math score	−0.0275	0.0669	−0.0846	0.0206	−0.0349	0.0203
Language score	0.0256	−0.1314	0.0764	−0.0301	−0.0686**	0.0063
Math top 10%	−0.0045	0.0244	−0.0470	0.0358	0.0130	0.0011
Language top 10%	−0.0115	0.0271	0.0226	−0.0331	−0.0004	−0.0014
Math bottom 10%	0.0344	−0.0129	0.0116	−0.0453	0.0086	−0.0045
Language bottom 10%	−0.0644	0.0763	−0.0159	−0.0492	−0.0007	−0.0102
Participation in extracurricular class
Study	0.0111	−0.0848	−0.0597	−0.0242	−0.0117	0.0158
Art	0.0054	−0.0088	0.0545	0.0178	0.0137	0.0599**
Sports	–	–	0.0176	0.0008	0.0354	0.0387
Preference for study subject
Like math	–	–	−0.1125*	0.0424	−0.0026	0.0247
Like language	–	–	0.0758	0.0225	0.0130	0.0423*
Additional control variables	Set 1	Set 1	Set 1	Set 1	Set 1	Set 1

Note: ***, **, and * indicate that they are significant at the level of 1%, 5%, and 10% respectively. An additional control variable set 1 controls grade dummy, number of siblings, dummy for father being college/university graduate or over, dummy for mother being college/university graduate or over, and household income. The results of sports activities in China are dropped because the percentage attending sports class is low (about 2%) and many in the sample dropped in the estimation.

Conclusion

We presented comparative evidence about the gender gap in educational achievement in China, Japan, and the USA, with an emphasis on the gender-specific effect of parental income and education, and the child’s own preferences for study subjects. We used three major nationally representative longitudinal surveys with rich information about cognitive outcome measures of respondent children as well as educational investment and parental SES that allow us to analyze their inter-relationships.

This paper presents the first detailed empirical study which examines the determinants of gender differences in both academic performance and preferences for math and language while also taking into consideration the role of family background and educational investment in Western and non-Western countries.

The main results obtained from the analysis are as follows. We found that girls score higher than boys in language in all three countries. However, girls score lower than boys in math only in the USA. Concerning the preference for study subjects, in the USA and Japan, more girls prefer language than boys; on the other hand, the opposite is the case for math. We also found that low household income tends to have a more adverse effect on language test scores for boys than for girls in the USA, as is consistent with previous studies. However, it does not have an impact on the gender gap in test scores in China and tends to affect girls more adversely than boys in Japan. Among other findings, the effect of income on the participation in arts activities is larger for girls than for boys in Japan and the USA, but not for those in China.

Focusing on the result of gender differences in household income, in the USA, as household income increases, the gender difference in language test score results narrows. In the USA, as DiPrete and Buchmann (2013) point out, the family background is likely to have a strong influence on boys. On the other hand, Japan still has a strong sense of traditional gender role attitudes, and it is thought that educational investment for girls, especially in low-income groups, is less than for boys. This result suggests that educational support for low-income children by the government such as vouchers and educational subsidies is necessary.

One prominent explanation about the results for China is gender egalitarianism in China since 1949. The Chinese government’s social, economic, and educational policies supported similar education for males and females. It is clearly defined in China’s Education Law, Compulsory Education Law, and Vocational Education Law, that women enjoy the same rights and opportunities as men to receive education. A study by Hannum and Xie (1994) showed, for example, that sharp declines in gender gaps emerged during the Economic Recovery and Cultural Revolution, with a strong emphasis on equality, while increases or slow decreases in gender stratification occurred in the Great Leap Forward and Economic Reform Era, with a focus on economic development.

The other explanation emphasizes the role of falling fertility, and especially the one-child policy. A great many scholars have posited that fertility decline reduces sibship size and changes sibship configuration, thus prompting parents to invest in their one child’s education regardless of gender (Hannum, 2005; Tsui and Rich, 2002; Veeck et al., 2003; Yeung, 2013).¹³

It is still surprising that China, which is well-known for its male-favored culture, has shown a negligible gender gap across families of different socio-economic backgrounds. At the same time, China has increasingly become a gender-imbalanced society exactly for this reason (Hesketh and Xing, 2006). As a result, Chinese women have been shown to have more egalitarian divisions of housework (Qian and Sayer, 2016; Yu, 2014) and greater autonomy in marriage (Qian and Qian, 2014; Yu and Xie, 2015) and fertility (Qian and Jin, 2018), especially those women with greater bargaining power who have higher levels of education and income. A more recent study by Hu and Yeung (2019) pointed out that, relative to Japan, Chinese fathers are relatively more involved in the childrearing decision-making because of high female labor force participation and gender equity ideology. With such a change, it is possible that the resource allocation within the family is increasingly favoring girls rather than boys, although there is little evidence for this.

One remaining question is why the same patterns as in China have not significantly appeared in Japan, where the total fertility rate has dropped sharply to near one. Perhaps this is why low-income families in Japan, who tend to have more children than high-income families, tend to show more gender-biased educational outcomes. Clearly more research is needed to explore this possibility, taking the effects of siblings and fertility decisions into account in comparative studies of the gender gap in educational outcomes.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is financially supported by JSPS KAKENHI Grant Numbers JP16H06323 and JP17H06086.

Acknowledgements

We would like to thank the Institute of Social Science Survey of Peking University for providing the data from the China Family Panel Studies. Also, we thank the Panel Data Research Center at Keio University for providing the data from the Japan Child Panel Survey, Japan Household Panel Survey, and Keio Household Panel Survey, and the National Center for Education Statistics for providing the data from the Early Childhood Longitudinal Study, Kindergarten Class of 1998–99. We also would like to thank the participants of SLLS 2018, RC28 2019, and the East Asia Child Well-being Project, especially James Raymo, Yu Xie, Ryosuke Nakamura, and Jun Yamashita.

ORCID iD

Kayo Nozaki

Notes

References

Akabayashi

Naoi

Shikishima

(eds) (2016) An Economic Analysis of Academic Ability, Non-Cognitive Ability, and Family Background: Primary Findings from a Panel Survey of Japanese School-Age Children. Tokyo: Yuhikaku (In Japanese).

Altonji

Blank

(1999) Race and gender in the labor market. In: Ashenfelter

Card

(eds) Handbook of Labor Economics Vol 3C. Amsterdam: Elsevier Science B.V., 3143–3259.

Autor

Figlio

Karbownik

, et al. (2017) Family disadvantage and the gender gap in behavioral and educational outcomes. NBER Working Paper Series 22267. Available at: http://www.nber.org/papers/w22267 (accessed 13 February 2020).

Ayalon

Livneh

(2013) Educational standardization and gender differences in mathematics achievement: A comparative study. Social Science Research 42(2): 432–445.

Baker

Milligan

(2016) Boy-girl differences in parental time investments: Evidence from three countries. Journal of Human Capital 10(4): 399–441.

Bedard

Cho

(2010) Early gender test score gaps across OECD countries. Economics of Education Review 29(3): 348–363.

Brenøe

Lundberg

(2018) Gender gaps in the effects of childhood family environment: Do they persist into adulthood? European Economic Review 109: 42–62.

Brinton

(1993) Women and the Economic Miracle: Gender and Work in Postwar Japan. Berkeley: University of California Press.

Broh

(2002) Linking extracurricular programming to academic achievement: Who benefits and why? Sociology of Education 75(1): 69–95.

10.

Cobb-Clark

Moschion

(2017) Gender gaps in early educational achievement. Journal of Population Economics 30(4): 1093–1134.

11.

DiPrete

Buchmann

(2013) The Rise of Women: The Growing Gender Gap in Education and What it Means for American Schools. New York: Russell Sage Foundation.

12.

Duncan

Murnane

( eds) (2011) Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances. New York: Russell Sage Foundation.

13.

Ellison

Swanson

(2010) The gender gap in secondary school mathematics at high achievement levels: Evidence from the American mathematics competitions. Journal of Economic Perspectives 24(2): 109–128.

14.

Ermisch

Jantti

Smeeding

(eds) (2012) From Parents to Children: The Intergenerational Transmission of Advantage. New York: Russell Sage Foundation.

15.

Fortin

Oreopoulos

Phipps

(2015) Leaving boys behind: Gender disparities in high academic achievement. Journal of Human Resources 50(3): 549–579.

16.

Fryer

Levitt

(2010) An empirical analysis of the gender gap in mathematics. American Economic Journal: Applied Economics 2(2): 210–240.

17.

Gong

Ding

Tsang

(2014) Gender differences of academic performance in compulsory education in rural Southwestern China. International Journal of Educational Development 39: 203–214.

18.

Guiso

Monte

Sapienza

, et al. (2008) Culture, gender, and math. Science 320(5880): 1164–1165.

19.

Hannum

(2005) Market transition, educational disparities, and family strategies in rural China: New evidence on gender stratification and development. Demography 42(2): 275–299.

20.

Hannum

Xie

(1994) Trends in educational gender inequality in China: 1949–1985. Research in Social Stratification and Social Mobility 13: 73–98.

21.

Hesketh

Xing

(2006) Abnormal sex ratios in human populations: Causes and consequences. Proceedings of the National Academy of Sciences 103(36): 13271–13275.

22.

Hojo

(2015) Comparison of gender difference in learning math in Japan and the US. The Journal of Economics, Niigata University 99: 1--15 (In Japanese).

23.

Yeung

(2019) Education and childrearing decision-making in East Asia. Chinese Sociological Review 51(1): 29–56.

24.

Inglehart R, Haerpfer C, Moreno A, et al. (eds) (2014) World Values Survey: Round Six - Country-Pooled Datafile. Madrid: JD Systems Institute. Available at: www.worldvaluessurvey.org/WVSDocumentationWV6.jsp (accessed 21 November 2019).

25.

Isa

Chinen

(2014) Gender gaps of achievement and aspirations in mathematics. The Japanese Journal of Labour Studies 648: 84–93 (In Japanese).

26.

Lai

(2010) Are boys left behind? The evolution of the gender achievement gap in Beijing’s middle schools. Economics of Education Review 29: 383–399.

27.

Lipscomb

(2007) Secondary school extracurricular involvement and academic achievement: A fixed effects approach. Economics of Education Review 26(4): 463–472.

28.

(2018) The gender gap in math performance, self-concept, and anxiety: Rural and urban China in an international context. In: International Association of Agricultural Economists 2018 conference, Vancouver, Canada, 28 July–2 August 2018, pp.1–28.

29.

Lummis

Stevenson

(1990) Gender differences in beliefs and achievement: A cross-cultural study. Developmental Psychology 26(2): 254–263.

30.

Lyu

Xie

(2019) The influences of family background and structural factors on children’s academic performances: A cross-country comparative study. Chinese Journal of Sociology 5(2): 173–192.

31.

Nollenberger

Rodríguez-Planas

Sevilla

(2016) The math gender gap: The role of culture. American Economic Review Papers & Proceeding 106(5): 257–261.

32.

OECD (Organization for Economic Development) (2015) The ABC of Gender Equality in Education: Aptitude, Behaviour, Confidence. Paris: OECD Publishing. Available at: http://dx.doi.org/10.1787/9789264229945-en (accessed 13 February 2020).

33.

OECD (Organization for Economic Development) (2017) The Pursuit of Gender Equality: An Uphill Battle. Paris: OECD Publishing. Available at: https://doi.org/10.1787/9789264281318-en (accessed 13 February 2020).

34.

Pope

Sydnor

(2010) Geographic variation in gender differences in test scores. Journal of Economic Perspectives 24(2): 95–108.

35.

Qian

Jin

(2018) Women’s fertility autonomy in urban China: The role of couple dynamics under the universal two-child policy. Chinese Sociological Review 50(3): 275–309.

36.

Qian

(2014) The gender divide in urban China: Singlehood and assortative mating by age and education. Demographic Research 31: 1337–1364.

37.

Qian

Sayer

(2016) Division of labor, gender ideology, and marital satisfaction in east Asia. Journal of Marriage and Family 78(2): 383–400.

38.

Riphahn

Schwientek

(2015) What drives the reversal of the gender education gap? Evidence from Germany. Applied Economics 47(53): 5748–5775.

39.

Rodríguez-Planas

Nollenberger

(2018) Let the girls learn! It is not only about Math … It’s about gender social norms. Economics of Education Review 62: 230–253.

40.

Tourangeau

Nord

Lê

, et al. (2009) Early Childhood Longitudinal Study, Kindergarten Class of 1998–99 (ECLS-K), Combined User’s Manual for the ECLS-K Eighth-Grade and K–8 Full Sample Data Files and Electronic Codebooks (NCES 2009–004). National Center for Education Statistics, Institute of Education Sciences, US Department of Education. Washington, DC.

41.

Tsui

(2007) Gender and mathematics achievement in China and the United States. Gender Issues 24(3): 1–11.

42.

Tsui

Rich

(2002) The only child and educational opportunity for girls in urban China. Gender & Society 16(1): 74–92.

43.

Urasaka

Nishimura

Hirata

, et al. (2010) Mathematics education and human capital accumulation: Empirical study in Japan. Journal of Quality Education 3: 1–14 (In Japanese).

44.

Urasaka

Nishimura

Hirata

, et al. (2012) Comparison in the annual income between graduates of social sciences and humanities and those of sciences. Journal of Quality Education 4: 1–10 (In Japanese).

45.

van Hek

Kraaykamp

Wolbers

MHJ

(2016) Comparing the gender gap in educational attainment: The impact of emancipatory contexts in 33 cohorts across 33 countries. Educational Research and Evaluation 22(5–6): 260–282.

46.

Veeck

Flurry

Jiang

(2003) Equal dreams: The one child policy and the consumption of education in urban China. Consumption Markets & Culture 6(1): 81–93.

47.

Wang

(2013) Why students choose STEM majors: Motivation, high school learning, and postsecondary context of support. American Educational Research Journal 50(5): 1081–1121.

48.

Xie

Shauman

(2003) Women in Science: Career Processes and Outcomes. Cambridge, MA: Harvard University Press.

49.

Xie

Zhang

, et al. (2017) China Family Panel Studies User’s Manual 3rd edition. The Institute of Social Science Survey, Peking University, pp.1–141.

50.

Yamaguchi

Shikishima

Hoshino

, et al. (2019) Vertical scaling of academic ability tests for elementary school first year students through junior high school third year students. Japanese Journal of Psychology 90(4): 408–418 (In Japanese).

51.

Yasuda

(2015) The gender differences of test scores in the fourth-grade university students. Journal of Tokyo Keizai University: Economics 285: 127–153 (In Japanese).

52.

(2011) Fertility decline and the trend in educational gender inequality in China. Sociological Studies 26(5): 153–177 (In Chinese).

53.

Yeung

(2013) Higher education expansion and social stratification in China. Chinese Sociological Review 45(4): 54–80.

54.

(2014) Gender ideology, modernization, and women’s housework time in China. Chinese Journal of Sociology (Chinese version) 34(2): 166–192.

55.

Xie

(2015) Changes in the determinants of marriage entry in post-reform urban China. Demography 52(6): 1869–1892.

56.

Zhang

Tsang

(2015) Gender gap in the National College Entrance Exam performance in China: A case study of a typical Chinese municipality. Asia Pacific Education Review 16(1): 27–36.

Gender differences in educational outcomes and the effect of family background: A comparative perspective from East Asia

Abstract

Keywords

Introduction

Previous research

Data 7

Results of empirical analysis

Descriptive statistical analysis

Regression analysis

Conclusion

Footnotes

Declaration of conflicting interests

Funding

Acknowledgements

ORCID iD

Notes

References

Data⁷