1 Introduction

Developing human resources in science and technology has long been a priority for nations around the world. Preparing an educated workforce to enter science, technology, engineering and mathematics (STEM) occupations is crucial for scientific innovation, technological advancement and economic competitiveness (Darolia et al., 2018; Lichtenberger & George-Jackson, 2013; Moakler & Kim, 2014; Wiswall et al., 2014). Cambodia is no exception. Promoting STEM enrolment and enhancing students’ STEM competencies, particularly in higher education, are the priority areas of the Cambodian government as the country shifts from reliance on agricultural development to higher-value industries and sectors and more innovative technology that will usher in the Industrial Revolution 4.0 (MoEYS, 2019a; RGC, 2015, 2019).

To enhance competencies and encourage interest and enjoyment in science and mathematics at upper secondary schools, the Ministry of Education, Youth and Sport (MoEYS) has initiated several policy initiatives (MoEYS, 2014). One of those is the academic streaming system. The main objective of streaming is to strengthen students’ background in science and mathematics at upper secondary school as a foundation pathway to higher education and jobs in STEM (MoEYS, 2010). This requires all 10th-grade students to select and study either the science or social science stream in the 11th and 12th grades.

The science stream was a popular choice for upper secondary school students in Cambodia, but many changed their minds when they reached their final year. According to the MoEYS, from 2012/13 to 2017/18, about 80 percent of students enrolled in the science stream, while only 20 percent chose the social science (MoEYS, 2017). However, in recent years, there has been a significant drop in the number of students who continue the science stream in grade 12. For example, in 2019/20, only 33,227 out of 63,547 grade 11 science-stream students stayed on the same path in grade 12. The same pattern was observed in 2020/21, when only 35,394 out of 69,842 grade 11 science-stream students remained in the science track in grade 12 (MoEYS, 2021b). This trend has implications for the future of STEM education and careers in Cambodia.

Cambodia aims to become a higher-middle-income country by 2030 and a high-income nation by 2050. To achieve this goal, it needs to develop its STEM human capital. The Science, Technology, and Innovation Roadmap 2030 states that half of the university students should study STEM majors by 2030 (RGC, 2021). However, there is still a gap between STEM and non-STEM enrolments in higher education. Many students prefer non-STEM fields such as business, economics, humanities and accounting (Roth et al., 2021). This shows that students lose interest and confidence in science after finishing upper secondary school. There are two patterns to note. The science stream in high school was supposed to encourage more students to study STEM fields in higher education, but it did not work for two reasons. First, many science stream students did not choose STEM majors in higher education. Second, the number of science stream students in high school has been dropping over the years. Therefore, it is essential to explore the factors influencing students’ decisions to switch or stay in their academic majors and compare the switchers and non-switchers.

Studies have demonstrated that upper secondary school education is a critical period for attracting students to science as it is significantly correlated with their post-secondary educational choices (Dustmann, 2004; Li & Kuan, 2018; Maltese & Tai, 2011; Shim & Paik, 2014; Simpkins et al., 2015). Given that the aim of the upper secondary school science stream is to provide a pathway to higher education, particularly STEM majors, better understanding of why students switch from upper secondary to higher education will help ensure that students make well-informed decisions about upper secondary streaming and higher education majors. Therefore, this study investigates the factors influencing the decision to switch academic majors when students transition from upper secondary to higher education through the lens of the individual, family, upper secondary school, and HEI.

2 Literature review

2.1 Tracking and streaming: definition and types

Tracking has been defined in various ways. This study adopted the definition of tracking used by Oakes (1985, 3), which refers to the “process whereby students are divided into categories so that they can be assigned in groups to various kinds of classes”. For this study, streaming refers to the last two years of secondary education when students choose between science or social science classes based on their interests and strengths to prepare themselves for higher education majors. The main differences between the two streams concern compulsory instruction time, content coverage, and the subjects required for the national exit exam (details are discussed in Sect. 2.2). The definition, the form and the type of tracking in education differ around the world. Based on these multiple understandings, LeTendre et al. (2003) developed a typology of tracking and streaming, as shown in Table 1.

Table 1 Typology of curricular differentiation (tracking/streaming) across nations.
Full size table

2.2 Tracking in Cambodia

In the Cambodian context, tracking/streaming falls into type 1 and type 3 of the typology. Type 1 tracking allows students to choose between technical/vocational or academic tracks after completing grade 9 (JICA, 2016; UNESCO, 2014). This type of tracking is not the focus of this study. Academic streaming divides students into science and social science streams after finishing grade 10. These streams last two years until the end of upper secondary education (MoEYS, 2010). Academic streaming was introduced to enhance students’ competencies in science and mathematics and to align their academic choices with their future careers (MoEYS, 2010). Students have to choose between the science stream or the social science stream at the end of grade 10, and they follow different curricula and assessment methods in grades 11 and 12. The science stream focuses on physics, chemistry, biology, earth and environmental science, and mathematics, while the social science stream covers Khmer literature, history, geography, morality, civics and citizenship.

The science stream has more instructional hours and higher scores for mathematics and science subjects than the social science stream. For example, mathematics has five hours per week and 125 points for each examination in the science stream, but only three hours per week and 75 points in the social science stream. The same applies to physics, chemistry, and biology. The scoring system reflects the importance of these subjects for the science stream students who aim to pursue majors and careers in STEM fields. The social science stream students, on the other hand, have more exposure to humanities and social sciences subjects that prepare them for different pathways.

In some so-called new generation schools (NGS), schools that aimed to increase skill levels in STEM subjects at upper secondary school levels through intensive capacity building in educational technology and STEM and inquiry and problem-based learning methodologies, the number of instruction hours might be higher. By contrast, in the social science streams, only three hours of instruction per week are allocated for mathematics and two for each science subject, with maximum scores of 75 and 50, respectively (MoEYS, 2010). Table 2 illustrates the different emphases (teaching hours) placed on each subject and the subjects that the students in the science and social science streams need to take in the baccalaureate exam.

Table 2 Weekly hours of instruction in science and social science subjects in traditional upper secondary and the new generation schools.
Full size table

2.3 Admission to higher education in Cambodia

The successful completion of a baccalaureate program is the primary route to higher education in Cambodia, but it is not the only pathway (). Students can be enrolled through either scholarship (tuition waiver) or fee-paying (MoEYS, 2002). Scholarships, particularly those sponsored by the government through MoEYS as a diversity and inclusion initiative, are provided based on four criteria: ability and academic merit, gender, socioeconomic status and geographical location (rural). To apply for a government scholarship, students must choose two majors of interest (as their first and second priorities) at any higher education institution (HEI) of their choice from those listed in the so-called annually-developed MoEYS booklet and submit their application to the Department of Higher Education in the second semester of grade 12. The application process starts in mid-March and lasts until the end of May each year.

Scholarship recipients are primarily selected based on their performance in the baccalaureate exam. However, to enrol in a STEM major at some prestigious HEIs—such as engineering at the Institute of Technology of Cambodia (ITC) or medicine at the University of Health Sciences (UHS)—students must also sit an entrance exam. To get into ITC, for instance, students need to take an exam in advanced mathematics, physics, chemistry and logic. To get into UHS, students must take a so-called national exam covering mathematics, chemistry and biology. Students can get only one government scholarship at a time to study either their first or second choice of major, depending on their baccalaureate or entrance exam results. In this regard, the stream pursued at upper secondary school could shed light on student choice in higher education.

2.4 Conceptual foundation

2.4.1 Major choice model

Hu’s (1996) major choice model argues that students’ choice of major aligns their personal goals with their social roles. According to Hu (1996), choosing a major involves two stages: initial and final. The initial choice is affected by factors such as individual ability (measured by high school performance), family characteristics (such as socioeconomic status, parental income, and educational aspirations), institutional attributes (such as type of HEI, school and class size, and geographic location), advice from significant others (such as relatives and peers), available financial support, perceptions of economic factors (such as career and job prospects), and perceived quality of the program. The final choice is influenced by similar factors, but also by the school attributes and economic perceptions (Hu, 1996).

Building on the notion that “choice is one of the major tenets of both a market economy and a democratic society” (Levin 1991 cited in Hu, 1996, p.2, p.3) contends that a student’s decision-making about choice of major “is an act of matching and combining individual goals with social roles”. The major choice model by Hu (1996) divides the decision-making process into the initial choice and the final choice.

The model integrates four primary models of students’ major choice: the econometrics model, sociological model, consumer model, and combined model. Students’ views of economic benefit are drawn from the econometric model, and the relationship between students’ educational experience/aspirations and socioeconomic status is derived from the sociological model. Personality-related factors of self-fulfilment are developed based on the consumer model with multiple stages and dynamic processes, and the availability of information in the decision-making process is consistent with the combined model (Hu, 1996).

2.4.2 Social cognitive career theory

Social cognitive career theory (SCCT) is a widely used framework to understand students’ choice of STEM major (e.g., Lee et al., 2015; Lent et al., 2018; Maltese & Tai, 2011; Moakler & Kim, 2014; Ruse & Xu, 2018; Sahin et al., 2017; Wang & Lee, 2019). Based on Bandura’s (1986) social cognitive theory, SCCT focuses on how self-referent thoughts and social processes influence behaviour. Lent et al. (1994) proposed that SCCT is a cyclical and longitudinal model that incorporates self-efficacy, outcome expectations, and interest as key factors in career choice. SCCT also considers personal inputs (such as gender, race, and predisposition), background contextual affordances (such as supports or barriers), and learning experiences from Krumboltz’s social learning theory as additional variables that shape career development (Kao, 2021). SCCT is similar to Hackett and Betz’s self-efficacy theory and Krumboltz’s social learning theory, but it also extends and integrates them in a comprehensive way (Lent et al., 2002 cited in Kao, 2021, p. 23).

2.4.3 STEM transfer model

The STEM transfer model developed by Wang (2103, 2017) integrates SCCT. It emphasises that students’ choice of STEM major is determined by not only the secondary but also postsecondary contexts. In other words, students’ intention to choose a STEM major is influenced by their achievement in mathematics and exposure to science and mathematics at upper secondary school, and their science and mathematics self-efficacy beliefs. In turn, prior achievements in and attitudes towards science and mathematics influence these variables. Furthermore, the choice of major is also influenced by postsecondary contextual supports (both academic interactions and financial aid, college readiness in mathematics and science, academic aspiration, and enrolment intensity) and barriers (remediation and external demands from the family) (Wang, 2013, 2017; Wang & Lee, 2019). In short, this model focuses on the influences of multi-dimensional factors—individual academic readiness and attitude, family support and challenges, and secondary and postsecondary support.

2.5 Factors influencing academic major choice: empirical evidence

Different theoretical perspectives inform the literature on students’ choice of science majors at the individual level. For example, behaviourists focus on the external factors that shape students’ behaviours and decisions regarding their academic majors. Academic achievement experts or psychologists examine this phenomenon from the lens of students’ psychological or academic factors. A key area of research in this field is how personal ability and affective factors influence students’ science outcomes, such as choosing a science major. Some of the factors that have been found to have significant effects on students’ choice of higher education majors in general and STEM majors in specific are gender (e.g., Gunderson et al., 2012; Kao & Shimizu, 2019), academic achievement in science and mathematics (e.g., Kao & Shimizu, 2019; Wang, 2013; Westrick et al., 2018), science and mathematics self-efficacy (e.g., Lent et al., 2018;Sahin et al., 2017; Wang, 2013; Wang & Lee, 2019), academic track (e.g., Li & Kuan, 2018; Shim & Paik, 2014; Wang & Lee, 2019), and outcome expectations (e.g., Nugent et al., 2015; Wang, 2013; Wang & Lee, 2019).

Also, social science researchers often explain the different students’ outcomes in science (students’ academic major in science versus social science) by utilising the deficiencies within the home environment of the students (e.g., Miller & Kimmel, 2012; Wang, 1995). According to human capital theory, students gain differential exposure to different cultural capital from their families and home and different access to social networks within their communities (e.g., Nui, 2017). Research studies have revealed several variables at home that could influence students’ choice of major in higher education, particularly in STEM fields. These include parental education (e.g., Crisp et al., 2009; Hodson & Freeman, 1983; Seymour & Hewitt, 1997), family socioeconomic status (e.g.,Niu, 2017; Xie et al., 2015), and relatives’ influence (e.g. Kao & Shimizu, 2019; Poeu, 2017; Seymour & Hewitt, 1997).

Last, school practices are also significant for understanding how students choose academic majors. Specifically, upper secondary school time enables students to decide whether to pursue STEM majors and careers after graduation (e.g.,Darolia et al., 2018; Lee et al., 2015; Maltese & Tai, 2011; Simpkins et al., 2015). Variables at upper secondary school can either motivate secondary school students to study and strengthen their background competence in science and mathematics and thus decide to pursue the STEM majors, or push students away from the STEM pipeline. The effects of educational institutions’ pull factors were also found to be significant (Hu, 1996; Wang, 2013, 2017; Wang & Lee, 2019).

3 Methodology

3.1 Conceptual framework

As framed by the theoretical foundations of one’s choice of major in higher education, the study analysed and synthesised the previously described three theoretical and conceptual models. Also framed by data availability, as a consequence, Fig. 1 draws the conceptual framework that provides the lens of investigation in the current explanatory study. In a broad sense, this synthesis illustrates the crucial dimensions and variables underlying each dimension that is common among the three theoretical models. The study synthesised the variables to give a holistic view of the variables. By and large, the study extracted and then classified the constructs of each conceptual model into the multi-dimensional variables of individual-level factors, family-level factors, upper secondary school factors and HEI-level factors. According to McGee (2020), gender, racism in policies, curriculum, pedagogies, rules/norms/values seem to play significant roles in STEM education, yet due to the availability of the data, the researchers could only measure the effect of gender on the disparity in switching the majors from upper secondary to higher education and leave out other structural racism variables from the conceptual framework. As in Fig. 1, the outcome variable is whether or not students switch academic majors in the transition from upper secondary school to higher education. The independent variables are broadly divided into four categories: individual, family, upper secondary school and HEI factors.

Fig. 1
figure 1

Conceptual framework of the current study. Source Developed by the authors based on the concepts discussed in Sect. 2 and available data

Full size image

3.2 Samples and sampling

The study employed a two-stage sampling method to select the student participants. The first stage involved the random selection of HEIs using systematic sampling with probability proportional to size, measured by total student enrolment. As of 2020, there were 124 HEIs in Cambodia, but not all of them are under the direct supervision of MoEYS. Sixteen different ministries have responsibility for overseeing HEIs, but the majority (80 and 25, respectively) of HEIs are under MoEYS and the Ministry of Labour and Vocational Training (MLVT). For logistics reasons, the research team sought permission to conduct research in the HEIs under these two ministries. Information on active HEIs under the two ministries was obtained from the MoEYS’s Department of Higher Education and MLVT’s Department General of TVET. Small-sized HEIs (fewer than 500 students) and branch campuses were dropped from the sampling frame, leaving 75 HEIs for first-stage sampling. A total of 21 HEIs, of which 12 are private, and four are located outside the capital, were selected. Next, student lists at the selected HEIs were obtained for second-stage simple random sampling. In the second stage, we calculated the number of students to sample from each HEI proportionally to their student population. Then we randomly chose students from the lists of each HEI.

Overall, the sample comprised almost equal proportions of male (49.8%) and female (50.2%) students, 60.8% of whom studied in the science stream and 39.2% in the social science stream at upper secondary school. The sample comprises more non-STEM than STEM students. A minority (15.4%) of the participants were studying for a STEM major, and 84.6% a non-STEM major. The largest shares of the sample attended upper secondary schools in district towns (36.3%), and urban areas (28.5%), and the smallest share (14.3%) went to rural upper secondary schools.

3.2.1 Data collection procedure

A survey questionnaire was designed and prepared in digital format using KoBoToolbox. The questions collected a wide range of information, encompassing students’ characteristics, family and educational backgrounds, and higher education experiences, including during the Covid-19 pandemic. In July 2020, a three-day training workshop was organised to familiarise 15 enumerators with the survey questions and instruct them on using the tablet-based KoBoCollect app. The questionnaire was pre-tested before using it for actual data collection. The survey was conducted from 20 July through 14 September 2020. All 1338 selected students were interviewed face-to-face (Table 3).

Table 3 Demographic information about the sample.
Full size table

3.2.2 Empirical strategy

The study sought to investigate the switch of academic majors from upper secondary school to higher education and the factors (individual, family, upper secondary school, HEI) influencing switching decisions. The collected data was analysed in probit modelling to identify the factors influencing the likelihood of switching. The dependent variable was coded dichotomously (0 = non-switcher, 1 = switcher).

Table 4 summarises the major switch variable and dependent variables used in the probit analysis. After dropping observations without complete information, the remaining 1281 observations were used for the probit analyses. The analyses were first conducted using the whole sample and then performed separately by the science stream and social stream subsamples. For each sample group, three models were run: (1) with all variables in the conceptual framework, (2) without mathematics and science as favourite subjects at secondary school, and (3) without perceived performance in mathematics and science at secondary school. The rationale for separating them in the models is that favourite subjects and perceived performance are likely to be highly correlated, causing multicollinearity issues. Splitting them can minimise the biased estimations, providing insight into whether mathematics and science are favourite subjects or whether perceived performances of mathematics and science have a more significant effect on the likelihood of switching.

Table 4 Summary statistics of variables included in the probit analysis.
Full size table

4 Results

4.1 Switching of academic majors

The results in Table 5 indicate that more students switched their majors on transition from upper secondary school to higher education than continued with similar subjects. Specifically, 54.4% of the 1338 participants were switchers and the other 45.6% were non-switchers. This apparent trend could be said to reflect the “swing from science”, a term coined by Dainton (1968) to describe the long-term and widespread decline in interest in science.

Table 5 Percentage of switchers versus non-switchers (n = 1338).
Full size table

As indicated in Table 6, of the 727 switchers, 93.1% switched from the science stream at upper secondary school to non-STEM majors in higher education. This is a worrying trend. A small minority (6.9%) switched from the social science stream to a STEM major at university, possibly related to individual ability and the requirement to have studied a science subject and/or maths at a higher level. Secondary school students who studied social sciences tended to choose a non-STEM major at university. Of the 611 non-switchers, 77.6% were social science stream students, and 22.4% were science-stream students. Only 14.0% of the 1,338 student participants had taken up a STEM-related major.

Table 6 Patterns of subject uptake in upper secondary to higher education (n = 1338).
Full size table

4.2 Factors influencing academic major switch

Individual performance and preferences, family background and HEI location have significant effects on a student’s decision to switch academic majors. Table 7 reports the marginal effects of three probit models on students’ decisions about whether or not to switch majors. Only selected variables with significant relationships are reported in the table; the full results can be found in Appendix A.

Table 7 Average marginal effects on the probability of major switch (selected variables).
Full size table

The model in column (1) includes all the variables, column (2) excludes favourite subjects (maths, physics, chemistry and biology) and column (3) excludes being good at the four subjects. In all three models, upper secondary academic stream (science), being female, family wealth index, HEI in Phnom Penh, and being a scholarship recipient displayed significant effects on the likelihood of switching majors.

In model (2), when science and mathematics as favourite subjects were excluded from the analysis, the model retained the significant variables of the first model and signified the influence of ability in mathematics. Last, in model (3) where perceived performance was excluded, influence was signified by mathematics and physics as favourite subjects.

Table 7 shows the individual-level factors that affect the switching decisions of science-stream students. The most important factor is the stream itself. Science-stream students have a 47% higher probability of choosing non-STEM majors than social science-stream students. This result is highly significant at p < 0.01 in all three models. The next important factor is gender. Female science-stream students are more likely to switch majors than their male counterparts. Holding other factors constant, the probability of switching is 12% higher for female students than for male students. Favourite subjects and perceived performance have mixed effects on switching decisions. When both variables are included in model (1), they are not significant. However, they become significant when they are separated in models (2) and (3). Students who like mathematics and physics and students who are good at mathematics are less likely to switch to non-STEM majors. The probabilities of switching are 5.5% and 4.4% lower for students who like mathematics and physics subjects, respectively. The likelihood of switching is also lower for students who are good at mathematics.

Among the factors related to family-level characteristics in the probit model, only the family wealth index had a significant effect on students’ tendency to change their academic majors. Students from more affluent families are more likely to change their academic majors. A rise of 0.716 (one standard deviation) in the family wealth index increases the chance of changing academic majors by about 4.0%.

The probit models show that upper secondary school factors have no significant impact on the likelihood of major switching. The location of the HEI, however, seems to affect this decision. Students who enrol in HEIs located in Phnom Penh have a 10% higher chance of switching majors than students living outside the Capital, and this result is significant at p < 0.01. Scholarship recipients are less likely to switch majors than non-recipients, and the relationship is significant at p < 0.10. A 0.146 increase (one standard deviation) in scholarship status reduces the probability of switching from science to social science by about 9.3%. This may be because scholarship criteria require high academic performance, and scholarship recipients are usually high achievers.

4.2.1 Subsample analysis

As mentioned earlier science-stream students are more likely to be switchers, therefore, it is intriguing to examine whether or not the factors associated with switching differ between science-stream and social science-stream students. Table 8 reports the marginal effects of probit models from the subsample analysis with selected variables that are found to be significantly associated with major switching (see Appendix A for the full results).

Table 8 Average marginal effects on the probability of major switch in subsample analysis (selected variables).
Full size table

The results indicate that female science-stream students are about 25% more likely than their male counterparts to switch their academic majors; in contrast, female social science-stream students are less likely to switch their academic majors. Although the analysis detected no significant association between the technology readiness index (TRI) and major switching in the whole sample, the TRI is negatively correlated at p < 0.05 with science-stream students’ decisions to switch majors. In other words, technology readiness helps students to stay in the STEM pipeline during their transition from upper secondary to higher education. An increase of 0.328 (one standard deviation) in TRI reduces the probability of switching academic majors by nearly 3.0%. In addition, the effect of the TRI is significant among science-stream students, but not social science-stream students. Another interesting finding is the effect of subject preference on students’ decisions about whether or not to switch academic majors. The subsample analysis shows that if students like mathematics or physics at upper secondary school, they are less likely to switch from the science stream to non-STEM majors and more likely to switch from the social science stream to STEM in higher education.

For family-level factors, wealth index, a measure of socioeconomic status, affects the likelihood of changing majors differently for science and social science students. Science students from wealthier families tend to switch from STEM to non-STEM fields in higher education, while social science students are not influenced by wealth index in their major choices. Father’s education level also matters for social science students, but not for science students. Social science students with fathers who have more than upper secondary education are more likely to switch to STEM fields. The location and year of upper secondary school graduation do not have any effect on major switching, even when the whole sample is separated into science and social science groups.

At the HEI level, the science-stream subgroup analysis indicates a strong positive relationship between enrolment in public HEIs and major switch as well as between enrolment in Phnom Penh-based HEIs and major switch. This means that switchers from the science track are likely to enrol in non-STEM majors at public HEIs in Phnom Penh. The social science-stream subgroup analysis indicates that the association between enrolment in Phnom Penh-based HEIs and major switch is statistically insignificant and that enrolment in public HEIs is negatively correlated with major switch. In other words, non-switchers from the social science track are likely to enrol in public HEIs.

5 Discussion

A key finding from the study is that a majority of switchers are from the science track. To some extent, this reflects the social reality of the Cambodian context. More than 40% of the 1,338 students participated in this study took their baccalaureate exam in 2017/18 or earlier. In that academic year, albeit more upper secondary students were in the science track, the higher education landscape showed the opposite trend. This mismatch might be due to students’ academic performance in and their passion for science and mathematics. As evidence, during the 2015 national exam, out of 83,325 students, only 23.3% passed the mathematics portion, while 41.7% passed the biology portion. However, in order to pursue higher education in STEM, students need to have a strong academic background in science (physics, chemistry and biology) and mathematics (MoEYS, 2010). Moreover, according to the outcomes of the Program for International Student Assessment for Development (PISA-D), Cambodian 15-year-old (grade 7 and above) students outperformed those in Senegal and Zambia in all subjects, and their academic performance in mathematics was comparable to those in other PISA-D member states (Cambodia scored 325 and the PISA-D average was 324). However, their performance in science was significantly lower than that in other PISA-D and ASEAN member states (Vietnam, Thailand, Indonesia and Singapore). Their performance was especially lower than students in OECD member nations. Cambodian students scored 330 (out of roughly 700) in science, whereas students in PISA-D member countries scored 349 on average (MoEYS, 2018). This finding might perpetuate a false belief among Cambodian students that STEM is for “the brightest” students only. Students tend to pursue STEM majors if they believe they are good at maths and/or science (Eng & Szmodis, 2016).

Another interesting finding is the gender difference in patterns of switches, notably that female science-stream students are more likely to switch to non-STEM majors. Some anecdotal evidence might explain this phenomenon. First, at upper secondary school, there are more female students than male students. According to statistics from the Department of General Education (MoEYS, 2021b), in the academic year 2020/21, female students accounted for 60% of science-stream students. Science-stream students usually perform better and challenge themselves more than social science-stream students. The findings also confirmed that gender does matter when it comes to the choice of STEM majors. There are two pieces of evidence to support this phenomenon. First, according to Wiswall et al. (2014) STEM majors are characterised by a “chilly environment,” of STEM fields, which can make female students feel unwelcome or discouraged by the male-dominated culture. This environment might also be influenced by their family’s perception of STEM majors. Students might be more interested in STEM if they felt their parents value STEM disciplines (Eng & Szmodis, 2016). Second, female students seemed to perceive that women are suited to jobs in air-conditioned offices such as in accounting and finance but not outdoor work in engineering or electronics (Kao, 2019; Eam & Keo, 2022). McGee (2020) also claimed that gender racism is weaved through the selection process not only on those who are pushed out of STEM field but also those who are considered as highly successful. As our findings reveal, many female switchers opted for business-related institutions. For instance, among 140 female students from our sample who enrolled at the National University of Management, 80% had switched from the science track to non-STEM majors.

In upper secondary school, female students are more likely to choose the science track (Kim, 2006; Stokking, 2000). On its own, being female is not a significant predictor of persistence in science. Contextually, female students’ track choice is not strongly connected to their major or career choice, as in the case of male students. Female students’ choice of the science track might be due to the perception that it will provide them with an open choice to various higher education majors (Kao & Shimizu, 2019). However, female science-stream students will switch to non-STEM majors. Although students need to have passed the baccalaureate exam before they can enrol in higher education, those that do not have good enough grades in the required subjects can sit entrance exams in order to gain admission to non-STEM majors. This allows students a free choice about whether or not to switch.

The findings also suggest if students have strong interests in mathematics and physic subjects, they are less likely to switch to non-STEM majors and that students ‘interests are more important than their academic performances when it comes to major decisions. The effects of these factors are even stronger, both in terms of significant level and magnitude, among the science-stream students who are very likely to switch majors. The findings are consistent with the conceptual framework, which postulates that an individual’s intention to pursue a certain field of study (non-switcher) is the consequence of the sequential cumulative effects of numerous learning experiences gained during science and mathematics classes (Lent et al., 2002; Wang, 2017). Students will be more likely to stay the course if their favourite subjects are mathematics and physics. So the question is why students are not interested or poorly performing in science and mathematics. Teaching quality and teaching approaches, availability of teaching resources, and school culture might also contribute to this phenomenon (Eng & Szmodis, 2016; Woolnough, 1994). Some Cambodian upper secondary schools might still be at a disadvantage in that they cannot access enough qualified teachers and teaching resources (Khieng et al, 2015). The learning gaps are also widened by social economic status between those who can and cannot afford private tutoring (Pov et al., 2020).

Science switchers are more likely to come from wealthier families. One interesting observation to help explain this phenomenon is that students from wealthier families might plan for a more “open choice of major” than their counterparts. They could, for instance, benefit from private tutoring at upper secondary school to enhance their academic achievement, especially in science and mathematics. This could lead to a wider range of higher education opportunities. Also, because wealthier students have better financial support, they could try out different majors or HEIs until they find a major or institution they really like. Students from low-income families may have less financial flexibility to change majors than their peers. Moreover, scholarship recipients tend to stick to their original majors, as the findings in the study show that having a scholarship reduces the probability of switching by 9.3%. This result is consistent with Chea, Tek, and Nok’s (2022) finding that STEM students were slightly more likely to rely on tuition waivers and scholarships as their main financial sources than non-STEM students.

6 Conclusion

The main finding of this study is that Cambodian upper secondary school students often change their majors when they enter higher education. This is especially true for female students who studied science in high school, but later chose non-STEM majors such as business management, finance and accounting. The study also explored the factors that influence the switching decision, including individual academic ability and preferences, family socioeconomic status and school and university supports. The results showed that female science students from wealthy families were more likely to switch to business-related majors in Phnom Penh-based HEIs.

From the findings, it is reasonable to conclude that Cambodian upper secondary school students are more likely to switch their majors when pursuing higher education. This is more likely for female science-stream students. Although female students choose the science track at upper secondary school, their interest in science tends to decline and they are more likely to choose non-STEM majors such as business management, finance and accounting. The findings also suggest that the decision to switch or not to switch is associated not only with individual academic ability and preferences, but also family socioeconomic status and school and university supports. Female science-stream students from high-wealth families are more likely to be science switchers, especially if pursuing business-related majors in Phnom Penh-based HEIs. However, the switching probability decreased if students enjoyed studying, had high self-efficacy in science and mathematics at high school, and received scholarships to pursue higher education.

The study suggests some strategies to address the issues of students switching majors in transition from upper secondary to higher education. First, science and mathematics teaching needs to be more engaging and relevant by using practical activities, inquiry-based methods and real-life applications, because students’ interest in science and mathematics matters. Science educators at the upper secondary school level and especially from early grades, should emphasise “how” rather than “what” when teaching science and mathematics. There should be more focus on applying science and mathematics knowledge and skills in real-life situations, particularly in reading and mathematics and problem-solving, rather than simply copying content from textbooks. The second strategy is to create interactive learning environments that foster enjoyment and cognitive activation in science and mathematics that promote students’ interest and enjoyment in learning mathematics and science. Simply put, increasing only teaching hours in the current science track would be misguided and unlikely to inspire learning without interactive teaching methods that involve inquiry-based and project-based activities. A third strategy is to provide more information and guidance about STEM majors and careers, especially for female students who are underrepresented in these fields. This would help enhance their perceived competencies and self-efficacy in science and mathematics, encouraging them to pursue future careers as scientists. Science teachers could also work to challenge and change parents’ preconceptions and gender stereotypes about mathematics and science. As quality matters, the higher education entrance examination criteria should also be reconsidered so that more qualified students can follow the same track from upper secondary school to higher education. Also, more scholarships should be available for those who want to pursue STEM majors in higher education to encourage students to stay in science rather than switch to the social sciences. The scholarship programs should also include guidance on the STEM fields in the application process.