Availability of STEM occupations.

Engineering and information science are two disciplines with strong growth potential [19]. Math-intensive occupations, such as mathematician, statistician, and physicist, also get high salaries [20]. The mean annual salary was higher in STEM jobs (95,350 USD) than all jobs (53,490 USD) in the US in 2019 [21]. Women in STEM jobs earn 33% more than women in non-STEM jobs. However, men still earn 14% more than women in the same STEM jobs [22]. In Japan, according to the results of the Basic Survey on Wage Structure in 2019 conducted by the Ministry of Health, Labour and Welfare, women account for about 24% of natural science researchers. The average monthly salary of natural science researchers (excluding overtime pay) is 422,000 yen, higher than the average for all occupations of 308,000 yen, but there is a large gap between that of male natural science researchers, 444,000 yen, and that of female natural science researchers, 355,000 yen [23]. In the US, women dominate close to half of all jobs. However, only 25% of women worked in STEM jobs in 2009. It has been reported that there are various possible reasons why women leave STEM careers: a lack of female role models, gender stereotyping, and less family-friendly flexibility [22].

Women’s underrepresentation in education.

Japan is a gender unequal country. The ranking of Japan in the Global Gender Gap Index based on health, education, economy, and politics was 121 out of 153 countries in 2019 [24]. Especially in higher education, the percentage of students in Japanese universities at the bachelor’s level in 2019 was 45.4% women and 54.6% men [25]. In other words, girls and women enter university in smaller numbers than boys and men. The percentages of the Japanese people over 16 years old who wanted a boy to get a university education was 72%, which was higher than that of 61% in girl [26], demonstrating that some people still consider that university education was more important for men than women. The gender gap still exists in Japanese society. Additionally, most STEM fields are dominated by men. The percentage of female university students studying the humanities in 2019 was 65.3%, with even fewer taking STEM subjects: 27.9% in the department of science and 15.4% in engineering. The percentage is also extremely low in mathematics (19.1%), physics (15.5%), and mechanical engineering (5.9%), in contrast to biology (38.8%), and chemistry (31.6%) [27].

Conventionally, Japanese women studying at university, especially STEM subjects, are likely to be recognized as intellectual. Some Japanese people still have a belief that intellectual women are not preferred in Japanese society. Several cases were reported where mothers disagreed with their daughters studying STEM. There was a mother who considered that this would prevent her daughter from getting married [28]. A study pointed out that the level of educational attainment for Japanese women can traditionally be used as an expression of the particular culture of the middle class: females were expected to be a good wife and a wise mother. This view was reflected in the low number of women in professional courses [29]. Parental belief also influences their children’s’ choice in science [30, 31]. Especially, the mother’s belief for her daughter’s success in STEM is related to her daughter’s future STEM career. This suggests that reducing the parents’ gender stereotyping would contribute to increasing the number of girls taking STEM.

For years, many initiatives to encourage girls to study STEM have been conducted at various levels. For example, a campaign called Rico-Challenge (rico means STEM in Japanese) targeting junior high school girls or high school girls has been hosted by the Gender Equality Bureau Cabinet Office of the Japanese government [32]. This campaign provides STEM information on the Internet, for example, messages from women in STEM and several offline events that introduce STEM-related occupations to girls. The Japan Science and Technology Agency (JST) financially supports universities and research institutions that have programs for girls to study STEM [33]. A Japanese publisher, Kodansha Ltd., has launched a website called Rikejo (science girls) that provides a range of information on STEM for girls, for example, a method to study science subjects, messages from STEM workers, a science column, and STEM events [34]. Google’s Mind the Gap program provides seminars and workshops on computer science for girls, with a version for Japanese girls [35]. It is assumed that these programs have directly or indirectly contributed to the increase in the number of girls studying STEM. The percentage of female university students increased from 18.3% in 1989 to 27.9% in 2019 in the department of science, and from 3.4% in 1989 to 15.4% in 2019 in the department of engineering [27, 36].

However, whether and how providing STEM information motivates girls to choose STEM for their future careers has yet to be investigated quantitatively.

Stereotype threat of math.

We posit that at least three stereotyping images and beliefs may prevent girls from studying STEM subjects. The first stereotyping image is that women by nature are not as good at mathematics as men. In Japan, the population of female university students is especially low in mathematics, physics, and engineering [27], which are generally considered to be math-intensive fields [37]. Although the results are inconsistent, several findings suggest that the gender difference in mathematics is lessening [16–18]. Regarding Japanese students, there was no significant gender difference in the scores for mathematics in TIMSS 2015 [16, 17]. On the other hand, in PISA 2018, a difference was reported in math literacy: boys scored higher (532 points) than girls (522 points). The 10-point difference was significantly higher than the world average of 5 points [18]. TIMSS targets primary and junior high school students, but PISA targets high school students. The difference of results between TIMSS and PISA might be due to the differences in the students’ grade level.

However, “stereotype threats” have been reported, or that negative stereotypes of the group that a person belongs to prevent him/her from reaching their full potential. One study showed that females, before taking a math test, scored the same as males if they read a text statement stating that there is no gender difference in math [38]. It is likely that girls who have a gender stereotype that girls are not good at mathematics score lower than boys in math tests. The gender gap in students’ performance varies among countries, and gender gaps are not considered to be innate [39].

Experimental design.

The online experiment had three stages: a pre-phase, information provision phase, and a post-phase. In the pre-phase, respondents replied to eight items (child) or five items (parent), and four quiz questions related to text paragraphs that we provided (child and parent). In the information provision phase, the respondents read the text. In the post-phase, the respondents replied to eight items (child) or five items (parent), and again answered the quiz questions (child and parent). The items and quiz questions in the pre- and post-phases were the same. The aim was to investigate whether their responses changed between the pre- and post-phases. The time required to complete all items and quizzes was about 10 to 15 minutes.

Questionnaire items.

The questionnaire consisted of the following items.

• Q1. Motivation to choose STEM. They were asked “If you want to go to university, do you want to think positively about going on to mathematics, physics, engineering, or information science (science and technology)? For science and engineering majors, physics and chemistry are the most common subjects for science examinations.”
• Q2. Motivation to participate in STEM events. They were asked “Do you want to participate in STEM events for junior high school students and/or high school students?”
• Q3. Motivation to encourage their children to choose STEM. They were asked “When your junior high school student is considering university, do you want to actively support your child to go on to a science course in mathematics, physics, engineering, or information (science and technology)? For science and engineering majors, physics and chemistry are the most common subjects for science examinations.”

Q1 to Q3 were asked in the pre- and post-phases. Q1 and Q2 were only for children, and Q3 was only for parents. Their responses were rated on a five-point scale from agree (= 1) to disagree (= 5). For analysis, the scale assignments were reversed to indicate that higher scores indicated higher motivation.

The perceptions of gender equality, of three non-stereotypical images and beliefs, and of the perception of STEM occupations and learning were measured using six items for children (Q4–Q9) and four items for parents (Q4–Q7).

• Q4. SESRA-S (Scale of Egalitarian Sex Role Attitudes) score. Individual egalitarian attitudes toward gender roles were measured using the short form of SESRA-S, which was developed in the field of psychology. SESRA-S consists of 15 items (for example, “women at home and men at work”), each rated on a five-point scale from strongly agree (= 1) to strongly disagree (= 5). The individual level of attitudes toward gender equality were shown by the total score for 15 items (from 15 to 75) as the SESRA-S score [40, 41]. The higher the total score, the higher the gender equality. The reliability coefficient of the items was reported as 0.91 [40].
• Q5. Non-stereotypical view of education. We asked the respondents how much they agreed or disagreed with the statement “a university education is more important for men than women” at the pre- and post-phases.
• Q6. Non-stereotypical view of math. We asked the respondents how much they agreed or disagreed with the statement “women are not as good at mathematics as men by nature” at the pre- and post-phases.
• Q7. Non-stereotypical view of women’s intellect. We asked the respondents how much they agreed or disagreed with the statement “women should be intellectual” at the pre- and post-phases.

Q5 to Q7 were asked in the pre- and post-phases. Q4-Q7 were for both children and parents. Their responses were rated on a five-point scale from strongly agree (= 1) to strongly disagree (= 5). For analysis, the scale assignments of Q7 were reversed. The higher scores indicate less stereotyping.

• Q8. Perception about occupations. The question was “Do you think it is important to earn a salary for a continuing occupation, such as being a permanent employee?”
• Q9. Perception about the importance of learning math. The question “Do you think that learning mathematics will give you more job opportunities when you become an adult?” was asked in the pre- and post-phases.

Q8 and Q9 were asked in the pre- and post-phases for children. Their responses were rated on a five-point scale from agree (= 1) to disagree (= 5). For analysis, the scale assignments were reversed to indicate that higher scores indicated higher motivation.

Information.

We prepared four types of information. Each consisted of short paragraphs of texts in Japanese with an illustration of a woman. The first information was on social media usage in Japan (referred to as “social media”, 413 words in Japanese). The second was on STEM occupations in Japan (referred to as “occupations”, 300 words). The third was entitled gender equal society in Japan (referred to as “society”, 448 words). The fourth was girls are good at math in Japan (referred to as “math”, 381 words). The three information (“occupations”, “society”, and “math”) mentioned gender equality and included some sentences focusing on girls, which encouraged them, but it was not included in the information on “social media” (Examples in English are shown in S1 Text, and in Japanese in S2 Text). Information on “social media” was used as the control information, and other three information were used as the treatment or experimental information.

Also, we prepared four quiz questions corresponding to each information to make sure that the respondents had read the information. As the limitation of online survey, we always have unconcentrated careless readers. The quizzes were not to test the respondents’ reading comprehension, but rather to confirm that they had read the information correctly. All the respondents were asked to answer the four quiz questions (either “yes” or “no”) regardless of which information they were given. The quizzes:

• Quiz question for “social media”. “The most used social media in Japan is Twitter.” (correct answer = “no”)
• Quiz question for “occupations”. “One of the areas of expertise that companies are looking for is mechanical engineering.” (correct answer = “yes”)
• Quiz question “society”. “In the Gender Gap Index (2019), which measures gender equality, Japan was in the top 100 out of 153 countries.” (correct answer = “no”)
• Quiz question “math”. “Female students in Japan perform lower in mathematics than boys.” (correct answer = “no”)

We provided four groups with combination of those information as shown in Table 1. The existing campaigns to encourage girls to choose STEM in Japan have conventionally provided girls with information of STEM occupations. But the information on gender equal society and girls’ math ability have not rarely provided to girls. We thus decided to investigate what additional information of “society” and “math” would be effective, on the basis that information “occupations” is always given. The first Treatment Group (occupations) read the information only on “occupations”. The second Treatment Group (occupations & society) read the information both on “occupations” and “society”. The third Treatment Group (occupations & math) read the information both on “occupations” and “math”. Control Group (social media) read the information on “social media”.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Each group and combinations of information.

https://doi.org/10.1371/journal.pone.0252710.t001

Groups.

A pair (one child and their parent) was randomly assigned to one of the four groups (Control Group (social media), Treatment Group (occupations), Treatment Group (occupations & society), Treatment Group (occupations & math)). A child and their parent were always assigned to the same group. The percentages of correct answers to each quiz question in pre-phase by children and parents are shown in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Correct answer (%) to each quiz questions in pre-phase.

https://doi.org/10.1371/journal.pone.0252710.t002

We regressed each demographic variable on the four groups (Control Group (social media) served as the baseline) to investigate whether the control and treatment groups were balanced. Age, gender of the children and parents, education of the parents, and major subjects at university (only respondents with an undergraduate degree or higher) of parents were well balanced among the four groups (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Profiles of the four groups of parents and children.

https://doi.org/10.1371/journal.pone.0252710.t003

We regressed each item (Q1–Q9) in the pre-phase for the four groups (Control Group (social media) served as the baseline) to investigate whether the four groups were balanced (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Motivations and attitudes in children and parents in the pre-phase.

https://doi.org/10.1371/journal.pone.0252710.t004

For children, the motivation to choose STEM (Q1) and participate in STEM events (Q2) was balanced between the control and each treatment group. A significant difference was found in the SESRA-S score (Q4) in the Treatment Group (occupations & math) (p = 0.043), suggesting that the egalitarian attitude toward gender roles in Treatment Group (occupations & math) was higher than that in Control Group (social media). Thus, while the information was provided randomly, their egalitarian attitude was not balanced.

For parents, a significant difference was found in the SESRA-S score (Q4) between the Control Group (social media) and Treatment Group (occupations & math) (p = 0.006), meaning that the egalitarian attitude towards gender roles in Treatment Group (occupations & math) was higher than that in Control Group (social media). A significant difference was also found in the non-stereotypical view of math skills (Q6) between Control Group (social media) and Treatment Group (occupations & math) (p = 0.030), meaning that the Treatment Group (occupations & math) disagreed that women were not as good at mathematics as men by nature more than Control Group (social media). These results suggest that parents were balanced between the control and treatment groups across observable characteristics, but not balanced in their attitude of gender stereotypes. Since the differences may be related to the outcomes, we controlled for the SESRA-S score in the pre-phase for estimating the effect of providing STEM-related information.

Change in motivations for STEM (Q1–Q3).

The coefficient of children’s motivation to choose STEM (Q1) was positively significant in Treatment Group (occupations) (p = 0.029), Treatment Group (occupations & society) (p = 0.008), and Treatment Group (occupations & math) (p < 0.001). This suggests that providing information on “occupations”, information on “society” and “occupations” together, or information on “math” and “society” together increased their motivation more than information on “social media”. The coefficient of children’s motivation to participate in STEM events (Q3) was positively significant in Treatment Group (occupations & math) (p < 0.014).

For parents, there was a positive significant difference in the motivation to encourage their children to choose STEM (Q3) in Treatment Group (occupations & math) (p = 0.032). These results demonstrate that providing information on “occupations” and “math” together increased both children’s and their parent’s motivation for STEM.

Change in stereotypical views and other factors (Q4–Q9).

For children, there was a positive significant difference in learning math (Q9) in Treatment Group (occupations & math) (p = 0.003). This suggests that providing information on “occupation” and “math” together, which is about there being no gender gap in mathematics, improved their attitudes to math. For the parents, positive significances were found in the coefficients of SESRA-S (Q4, p = 0.009) and the non-stereotypical views of math skills (Q6) in Treatment Group (occupations & math) (p = 0.021), suggesting that providing information on “occupation” and “math” together increased their egalitarian attitudes and decreased math stereotyping.

Change in motivations for STEM (Q1–Q3).

There was not a significant difference in the interaction between the four groups and gender in motivation to choose STEM (Q1) and motivation to participate in STEM event (Q2) in children, and motivation to encourage their children to choose STEM (Q3) in parents. This suggests that providing gender equality information does not discourage STEM choices for boys.

Change in stereotypical views and other factors (Q4–Q9).

For children, there was a positive significant difference in learning math (Q9) in the interaction between Treatment Group (occupations & society) and sex (girls) (p = 0.046), meaning that the information on “occupations” and “society” together improved girls’ attitudes to math. For parents, there were not a significant difference in the interactions in Q4 to Q7.

Change in motivations for STEM (Q1–Q3).

Coefficient of children’s motivation to choose STEM (Q1) were positively significant in the interaction between Treatment Group (occupations) and the correct quiz answer (p = 0.022), suggesting that children’s motivations in the Treatment groups (occupations) increased especially in those who correctly answered the quiz question on “occupation” compared with children in the same group who answered the quiz incorrectly. There were no significant differences in the interaction between Treatment Group (occupations & society) and the correct quiz answer (p = 0.166) and between Treatment Group (occupations & math) and the correct quiz answer (p = 0.138), but the coefficients were positive values.

The coefficient of parents’ motivation to encourage their children to choose STEM (Q3) was positively significant in the interaction between Treatment Group (occupations & math) and the correct answer (p = 0.012), indicating that the parents in Treatment Group (occupations & math) who answered quiz question on “math” correctly increased their motivation compared with those who incorrectly answered the same quiz question.

Change in stereotypical views and other factors (Q4–Q9).

For children’s learning math (Q9), a positive significance was found in Treatment Group (occupations) and the correct quiz answer (p = 0.023); Treatment Group (occupations & society) and the correct quiz answer (p = 0.046); Treatment Group (occupations & math) and the correct quiz answer (p = 0.036) meaning that children who answered the quiz question corresponding to the information previously provided to them correctly increased their desire to learn math compared with those who answered quiz question incorrectly.

For parents, regarding the non-stereotypical view of women’s intellect (Q7), a positive significance in the interaction between Treatment group (occupations & math) were found (p = 0.040). This suggests that the parents in Treatment Group (occupations & society) who answered the quiz question on “occupation” and “math” decreased their stereotypical view compared with those who answered the quizzes incorrectly.