Introduction

To date, the Royal Government of Cambodia (RGC) has outlined a vision with the goal of transitioning from a lower-middle-income status to an upper-middle-income status by 2030 and ultimately achieving high-income status by 2050 (RGC, 2015, 2019). In alignment with this ambitious vision, the RGC formulated the Cambodia Industrial Development Policy (IDP) for 2015–2025, officially established in 2015 as a strategic roadmap to propel the nation’s industrial development. The primary objectives of the IDP are to promote sustainable and inclusive economic growth by diversifying the economy, enhancing competitiveness, and boosting productivity (RGC, 2015). Nonetheless, the shift from an agriculture-based economy to an industrialized one presents Cambodia with significant challenges in building a workforce equipped with the knowledge and skills necessary for Science, Technology, Engineering, and Mathematics (STEM) fields. The RGC (2015) underscores a major concern regarding the shortage of human resources possessing fundamental technical knowledge and skills, which are essential for transforming an unskilled labor force into a skilled workforce capable of assimilating new and high-value technical and technological skills. The urgent demand for skilled labor to drive the nation’s economy places Cambodia at a critical juncture, necessitating the implementation of various educational reforms. The Cambodian Ministry of Education, Youth, and Sport (MoEYS) (2016b) emphasizes that STEM subjects and competencies are crucial to realizing Cambodia’s future aspirations for 2030 and 2050, as outlined in the IDP 2015–2025. Consequently, the RGC has placed significant emphasis on these competencies, intensifying its efforts to enhance STEM education programs, align them with the country’s economic development objectives, and address the demand of the labor market.

To support its commitment, MoEYS has advocated major education reforms in secondary education. In 2016, MoEYS established and implemented the New Generation Schools (NGSs) (see MoEYS, 2016a) and Policy on STEM education (see MoEYS, 2016b). These initiatives are intended to promote science education and stimulate enrollment in the science track at upper secondary education and STEM-related majors in higher education. To support government policies, the Asian Development Bank (ADB) collaborates with MoEYS to formulate a support project called the “Second Upper Secondary Education Sector Development Program”. The project aims to provide financial and technical assistance to improve upper secondary education, particularly for students from financially disadvantaged backgrounds. The primary goal is to increase science-track enrollment, which is expected to increase STEM-major enrollment in higher education. Nationwide, it was reported that 4,334 students in the academic year 2019–2020 and 1,783 students in the academic year 2020–2021 received support (see ADB, 2020; MoEYS, 2020c).

Despite these efforts, the enrollment rates in the science track are significantly lower than the enrollment rates in the social science track at upper secondary schools. The enrollment rates in the science track declined from 96% in 2014 to only 34% in 2020, while enrollment rates in the social science track significantly increased from 4% in 2014 to 66% in 2020 (MoEYS, 2020a). The enrollment rates in STEM-related majors are also lower than the enrollment rates in non-STEM-related majors in higher education. For instance, according to MoEYS (2023), the enrollment rate in STEM-related majors was only 25.60%, while the enrollment rate in non-STEM-related majors was 74.40% in the academic year 2021–2022. The decline in science-track enrollment rates at upper secondary schools are associated with the decrease in STEM-major enrollment rates in higher education institutes (HEIs) (Kao & Kinya, 2020b). This has evolved into a global concern regarding the shortage of a STEM workforce to support a country’s economy (Le et al., 2014). In Cambodia, although science education is designed to cultivate students’ interest in pursuing STEM-related majors (MoEYS, 2010), it appears to have the unintended consequence of dissuading students from enrolling in the science track. There is a growing trend among upper secondary-school graduates to opt for non-STEM majors (see MoEYS, 2020a, 2023). This tendency raises concerns about Cambodian educational development in the long run. It highlights the paradox of track-division enactment at upper secondary schools, which requires a closer investigation and review.

Track division and STEM majors in higher education

In 2010, MoEYS issued the “Guidelines on the Practice of General Education Curriculum at Upper Secondary Schools” to strengthen students’ readiness for higher education and promote STEM education in Cambodia (MoEYS, 2010). These guidelines set forth a new paradigm for upper secondary education, introducing a division into two distinct tracks: the science track and the social science track. In 2020, a new guideline was issued pertaining to “Science and Social-Science Track Choices or Vocational Skills” for students in upper secondary schools (see MoEYS, 2020b). Students are required to choose a track in grade 10 and enroll in grade 11, although they can change their track choice in later grades. Each year, grade-10 students are informed about the track choices in April and register for a preferred track in May. Schools prepare and divide classes into two tracks and announce them in the next academic year (MoEYS, 2010, 2020b). According to the new guidelines of MoEYS (2020b), both tracks require students to take the same 14 subjects for 40 h per week. The main difference between the two tracks is the number of instruction hours for science and social science subjects. Social science-track students study more hours of social science subjects, while science-track students study more hours of science subjects. Additionally, the national examination is split into two tracks, with different tests and test scores for social science and science subjects, including mathematics and the Khmer (Cambodian language) tests (see MoEYS, 2010, 2020b). The implementation of track division aims to guide students toward their major choices in HEIs and future career decisions, with an emphasis on science and STEM-related fields. This initiative provides valuable insights to students regarding track choices and vocational skills, assisting them in selecting the appropriate track that aligns with their abilities and talents. It facilitates students’ understanding of subjects, encourages self-evaluation for track selection, and offers orientation for HEI majors (see MoEYS, 2020b). As Cambodia’s economy shifts toward the industrial sector, there is strong encouragement for students to enroll in the science track, with the expectation that they will pursue STEM-related majors in HEIs and STEM-related careers.

Conceptual framework

Comprehensive studies need to address what might be the reasons behind students’ decision to leave the science track for the social science track. A scarcity of literature emphasizes the urgent need for evidence to help policymakers and development partners find appropriate interventions. In Cambodia, only a study by Kao and Kinya (2020a) investigated this phenomenon in upper secondary education with some limitations discussed below. Several others tend to focus only on STEM major choices in higher education (see Eam et al., 2021; Kao, 2019; Kao & Kinya, 2019; Leng et al., 2021). Based on previous studies (i.e., Kao & Kinya, 2020a), more students continue to leave science-track classes for social-track classes. Several gaps denote a requirement for a more in-depth investigation into the issue and practical implications to improve the incidence. Therefore, a comprehensive study is needed to investigate this phenomenon from diverse perspectives regarding analytical methods, conceptual framework, and samples. For instance, Kao and Kinya (2020a) included grade-11 students from three provinces and cities as samples. It should be noted that students are allowed to choose either the social science track or the science track in grade 10 and change a track choice in later grades (see MoEYS, 2010). The guideline indicates that the decision in grade 11 is not the final decision for track choices because the final decision is made when they transition to grade 12. In grade 12, students must register for a specific track choice for the national examination.

Unlike previous studies, the current study included grade-12 students as samples. To ensure that the results can be generalized nationwide, it purposively selected 10 provinces that represent the four geographical areas of Cambodia as research sites (see the Ministry of Planning (MoP), 2020; Nesbitt, 1997). Then it also expanded to 20 schools, with ten schools in rural areas and ten others in urban areas (see Participants). Furthermore, previous studies commonly used a binary logistic regression for the analysis without considering the nested structure of the data (i.e., Kao & Kinya, 2020a). By considering the nested structure of the data, the current study alternatively used the two-level hierarchical linear modeling (HLM) for the analysis. Woltman et al. (2012) maintain that HLM analyzes data in hierarchical structure and precisely estimates the lower-level slopes and the estimation of the higher-level outcome. Furthermore, the current study adapted the instrument from previous studies with adjustments and added more important variables that were not included in those previous studies such as students’ perception of national examination, private tutoring, and scholarship (see Table 1). Among these factors, significant attention from policymakers and researchers has been directed toward shadow education regrading private tutoring practices within Cambodian public schools. These practices have far-reaching implications for promoting inclusive and equitable quality education for all Cambodian children, particularly at the secondary education level (i.e., Bray, 1999, 2006; Bray et al., 2016; Marshall & Fukao, 2019; Nhem & Kobakhidze, 2022; Soeung, 2021).

Nhem and Kobakhidze (2022) contend that, to enhance the quality of secondary education, the government has established NGSs, characterized as model secondary schools operating autonomously with funding from both private and public sources. The establishment of NGSs aims to address private tutoring practices, which have been identified as the root causes of social inequalities and substandard education. This issue is closely associated with the socioeconomic status (SES) of students’ families, as private tutoring represents a significant financial commitment for parents (Dawson, 2010; Palmer, 2020). In Cambodia, private tutoring places a substantial financial burden on parents and contributes to educational inequalities and social segregation within public school settings (Nhem & Kobakhidze, 2022). Previous studies have shown that students from financially disadvantaged backgrounds are less likely to engage in private tutoring classes than their peers and are often part of underachieving (Pov et al., 2021) and dropout groups (Pov et al., 2022). Additionally, students attending urban schools are more likely to participate in private tutoring compared to their rural school counterparts (Bray, 2006). Therefore, alongside other predictors, private tutoring was included and examined in the current study, with the hypothesis that it significantly influences students’ track choices.

The scarcity of studies about track choices and their determinants threatens the policy and project review, monitoring, and evaluation. A comprehensive study must rigorously address the hidden reasons with the efficient analytical method and framework. Therefore, in combination with the emergence of some previous studies related to track choices at upper secondary education and STEM major choices at higher education in Cambodia (see Eam et al., 2021; Eng & Szmodis, 2016; Kao, 2019; Kao & Kinya, 2019, 2020a), this study developed and utilized a comprehensive conceptual framework for the analysis. It integrated the observed variables in previous studies into the two-level HLM. As mentioned earlier, some new variables were also added to the models to fulfil the gaps in previous studies, as shown in Fig. 1 and Table 1. To date, no previous studies in Cambodia have used multilevel analysis to examine factors influencing students’ track choices. In essence, HLM is a type of multilevel analysis that allows researchers to accurately analyze the nested structure data (Pov et al., 2021; Pov et al. 2022). Figure 1 illustrates the analytical framework of the two-level HLM examining the impact of individual-level and school-level factors on students’ track choices at Cambodian upper secondary schools. The extracted variables were included in different hierarchies and observed across individual (level 1) and school (level 2) levels. Students’ personal- and family-related factors, including the outcome variable, were nested in level 1 of the hierarchy. In principle, the outcome variable is always positioned in the lowest hierarchy of HLM (Castro, 2002; Pov et al., 2021). School-related factors were nested in level 2 of the hierarchy, as shown in Fig. 1 below.

Fig. 1
figure 1

The conceptual framework of the two-level HLM assessing factors influencing grade-12 students’ track choices in Cambodia

Full size image

Aim and research question

The aim of this study was to examine the factors influencing Cambodian students’ track choices of the social science track and the science track in upper secondary education, covering individual (student), family, and school aspects. It addressed the question, “What individual-level and school-level factors significantly influenced students to choose the social science track over the science track in at upper secondary schools in Cambodia?”.

Methodology

Research design

The current study employed a purely quantitative approach with a survey study design to identify factors that led students enroll in the social science track increasingly. A literature review has indicated that previous studies commonly used a survey study design to investigate track choices and STEM major choices in higher education in Cambodia (i.e., Kao, 2019, 2020; Kao & Kinya, 2020a). A survey study design allows researchers to collect a large number of samples by generalizing the findings to the entire population (Fowler, 2009). It also allowed this study to use the two-level HLM as the main analytical method. With a large number of samples and schools, HLM could estimate accurate results.

Participants

A cluster random sampling method was employed to select the samples and research sites for the study. Cambodia is divided into four regions: the central plain, Tonle Sap, coastal and sea, and plateau and mountains, as defined by MoP (2020). To ensure representation and generalization, all provinces were clustered within these four regions and conducted a random selection. Ten provinces were randomly chosen from these four regions. Secondly, the researcher requested a name list of upper secondary schools from provincial offices of education of the selected provinces and divided them into two clusters: urban and rural areas. Then 20 secondary schools were randomly selected, with ten schools from urban areas and ten others from rural areas. One science-track class and one social science-track class were selected from each school. Thus, 40 classes (20 social science classes and 20 science classes) were obtained. All students of each class were invited to participate and complete the questionnaire. In total, 696 students participated in the study. It should be noted that 46.10% of students were from science-track classes, while 53.90% of others were from social science-track classes. Approximately 56.30% of them were from urban upper secondary schools compared to 43.70% from rural upper secondary schools. Most of them were 17 (24%), 18 (34.90%), and 19 (20.50%) years old. About 16.9% were between 20 and 24 years old.

Questionnaire

The current study adapted the questionnaire from Kao and Kinya (2020a) for data collection. The questionnaire was adapted with some modifications and the addition of predictor variables to capture a holistic point of view from students about the reasons behind the decline of science-track choice. Some minor Khmer (Cambodian language) spelling mistakes were corrected. The final questionnaire consisted of 107 items divided into seven sections: demographic information, student-related factors, family-related factors, teacher-related factors, mathematics and science outcome expectations, encouragement to participate in science and mathematics, and scholarships.

Procedure

Approval was sought from MoEYS, provincial offices of education, and school principals of the target schools as a prerequisite to data collection. After official approval was granted, the questionnaire was distributed to all students in the selected schools. School principals and subject teachers were asked to assist with the distribution of the questionnaire to ensure that all students responded to the questionnaire. Each student was expected to spend 20 to 30 min completing the questionnaire. The valid 696 questionnaires were obtained.

The data was analyzed using various statistical methods, including descriptive statistics, as well as tests for validity and reliability. Moreover, the current study employed the two-level HLM as the main analytical method to examine the influential factors on track choices at two different hierarchies of HLM. In principle, HLM is used for nested data at different hierarchical levels. As mentioned earlier, the previous study by Kao and Kinya (2020a) employed a binary logistic regression to examine factors affecting students’ science-track choice. It did not consider the nested structure of the data and shared variance of variables. Unlike Kao and Kinya (2020a), this study employed the two-level HLM for analysis to generate more comprehensive results. HLM precisely estimates the lower-level slopes and the estimation of the higher-level outcome (Woltman et al., 2012).

This study included 24 independent or predictor variables for the analysis. Predictor variables were divided into two hierarchical levels: individual level (level 1) and school level (level 2). Individual-level factors were nested at level 1 and consisted of 18 variables, including the outcome variable (track choices). The outcome variable is typically at the lowest level of the hierarchy (Castro, 2002). Individual-level factors were gender, age, ease of national examination, high passing rates, consideration of good grades, spending on private tutoring, dividing track is good, consideration of a major at higher education, consideration of STEM majors in higher education, attitudes toward science, mathematics and science self-efficacy, parents’ encouragement, parents’ advice, family income, father’s education, mother’s education, father works in science fields, and mother works in science fields. School-level factors were considered at level 2 of the hierarchies and comprised six predictor variables: teachers’ encouragement, support from science and mathematics teachers, school principals’ encouragement, peer encouragement, scholarship, and school location.

The null model was computed to ensure that the two-level HLM was appropriate and reliable for the current study. The intraclass correlation coefficient (ICC) was also estimated in the null model and other models to determine (1) correlation coefficients confirming whether the models were needed and (2) the deviance statistics and other coefficients (Garson, 2013). The calculation method of ICC varies depending on types of outcome variable. In this study, the outcome variable is dichotomous or binary (0 = the science track, 1 = the social science track). Thus, to calculate ICC for a binary outcome variable, the formula is

$${\text{ICC = }}\left( {\frac{{{\uptau }_{{{00}}} }}{{{\uptau }_{{{00}}} { + }\pi^{2} {/3}}}} \right){ = }\left( {\frac{{{\uptau }_{{{00}}} }}{{{\uptau }_{{{00}}} { + 3}{\text{.29}}}}} \right)$$

where τ00 was the estimated variance component of the intercept (u0) in the models (Wu et al., 2012). To estimate the binary outcome variable, the Bernoulli distribution was chosen to estimate the two-level HLM. The analysis was conducted with the HLM software, version 8, with the standard license (see Scientific Software International, 2022). Beta coefficient (β) and odds ratio (OR) were generated and observed in the null model, model 1, model 2, and the final or mixed model. The Beta coefficient (β) and odds ratio (OR) of the final model were used for interpretation and discussion. The final model estimated all predictor variables altogether, as shown in Table 1.

Results

Validity and reliability tests

The exploratory factor analysis (EFA) using the generalized least squares with varimax rotation method was conducted to establish the construct validity of attitude toward science. The attitude toward science scale was adopted from Kind et al. (2007), comprising 39 items divided into six subscales: learning science at school, self-concept in science, practical work in science, science outside of school, importance of science, and future participation in science. The results of the EFA unveiled a total of six factors, explaining 62.11% of the total variance. The Kaiser–Meyer–Olkin (KMO) measure indicated the adequacy of the sampling for the analysis (KMO = 0.95). Bartlett’s test of sphericity was statistically significant (χ2 (528) = 15,231.98, p < 0.001), indicating that the variables’ correlations were sufficient for EFA. The communalities showed that a high percentage of the variance in the items was accounted for by the factors. Kaiser’s criterion of eigenvalues greater than 1 confirmed a six-factor solution as the best fit. The Cronbach’s Alpha values were 0.90 for learning science at school, 0.91 for self-concept in science, 0.92 for practical work in science, 0.83 for science outside of school, 0.81 for the importance of science, 0.89 for future participation in science, and 0.95 for the general attitude scale.

Effects of individual-level and school-level factors

Table 1 presents the results of four models of the two-level HLM: null model, level-1 model, level-2 model, and final model. The results of the null model revealed that there were statistically significant variations in track choices across schools (χ2 = 92.14, df = 19, p < 0.001). The reliability estimate was 67%. The results indicated that the ICC of the model was 0.15, suggesting that 15% of the variance was attributable to school-level factors at level 2, whereas the remaining 85% accounted for individual-level factors at level 1. It indicated that individual-level factors collectively observed much variance. The results justified that the two-level HLM was appropriate and reliable to use for the current study. It confirmed that much variance in track choices accounted for individual-level factors. Therefore, it was arguable that individual-level factors (level 1) were assumed to significantly influence students’ track choices more than school-level factors (level 2). It should be noted that the current study used the results of the final model for interpretation and discussion. The final model comprised all predictor variables at individual and school levels.

In the final model estimation, the results suggested that by considering all predictor variables at levels 1 and 2 altogether, the reliability estimate was 73%. The ICC was 0.24, meaning 24% of the variance was significantly explained by school-level factors at level 2. The remaining 76% was attributable to individual-level factors at level 1. Again, much of the variance was largely explained by individual-level factors. Predictor variables at level 1 were expected to have stronger effects on track choices than level-2 predictor variables in the final model.

Individual-level factors

The results of the final model showed that at the level-1 hierarchy, age was a significant predictor. Most older students were more likely to choose the social science track over the science track than their younger peers (β = 0.22, OR = 1.25, p < 0.05). The odds ratio value indicated that a unit increase in students’ age increased the likelihood of choosing the social science track by 1.25 times. Most students reported that they decided to opt for the social science track over the science track because they believed that the national examination of the social science track was easier than that of the science track (β = 0.23, OR = 1.26, p < 0.05). A unit increase in such belief was likely to increase the chance of enrolling in the social science track by 1.26 times. Similarly, the high passing rate was a significant predictor variable influencing students to choose the social science track (β = 0.43, OR = 1.54, p < 0.001). Choosing the social science track improved the likelihood of passing the national examination 1.54 times. Moreover, most reported that they enrolled in social science-track classes because they only wanted to pass the national examination. They had no intention of getting good grades (β = − 0.31, OR = 0.73, p < 0.01). A unit increase of such intention increased the likelihood of choosing the social science track by 73%.

The results suggested that private tutoring significantly predicted students’ track choices (β = − 0.54, OR = 0.58, p < 0.001). Students were asked to rate their spending on private-tutoring fees for the social science track and the science track. Choosing the social science track significantly reduced the likelihood of spending on private tutoring classes by 58%. In relation to spending in education, family income (β = − 0.30, OR = 0.74, p < 0.05) and parents’ advice (β = 0.73, OR = 2.08, p < 0.001) significantly predicted students’ track choices. Students from family with lower SES were 74% more likely to choose social science than those from family with higher SES. Additionally, this study also found that most parents advised their children to select social science track over science track. A one-unit increase in parental advice increased the likelihood of social science-track choice by 2.08 times. It was also found that consideration of STEM majors in higher education negatively impacted students’ track choices (β = − 0.64, OR = 0.53, p < 0.01). Students who chose the social science track were 53% less likely to consider STEM majors in higher education compared to those who chose the science track. The results further revealed that students’ attitudes toward science significantly negatively impacted track choices (β = − 2.45, OR = 0.09, p < 0.001). A unit increase in attitudes toward science significantly reduced the likelihood of enrolling in the social science track by 9%.

Table 1 Factors influencing grade-12 students’ track choices at Cambodian upper secondary schools
Full size table

School-level factors

The results indicated that school location was a significant predictor variable at the school level, which positively affected track choices. Other factors, especially scholarship, did not impact track choices. In this study, students were chosen from an equal number of urban and rural schools. Students from rural secondary schools were more likely to choose the social science track than the science track compared to those from urban secondary schools (β = 1.37, OR = 3.95, p < 0.05). A unit increase in students from rural upper secondary schools significantly increased the likelihood of enrollment in the social science track by 3.95 times more than the science track, as shown in Table 1 below. Notably, 56.3% of students were from urban schools, with remaining 43.7% originating from rural schools.

Discussion

Individual-level factors

Overage enrollment

The findings indicated that most of older students opted for the social science track over the science track. It should be noted that the official school age for grade-12 students is 18 years old. However, about 37.6% of the participants were between 19 and 24 years old, and these students were considered overage students. This finding contrasts with Kao and Kinya (2020a) who found that students’ age did not significantly impact science-track choice in their study.

The observed phenomenon has implications across various facets, encompassing classic issues like dropout rates, academic repetitions, and especially, academic performance. Research has previously demonstrated that overage students often struggle to progress through their education, resulting in repeating grades and performing poorly in examinations (Grissom, 2004; Pov et al., 2021). Grisson (2004) argues that by the time students reach high school, being older does not confer any academic advantages. Overage students typically perform less satisfactorily than their peers, and this performance tends to worsen as they grow older. This detrimental association between age and academic performance persists consistently over time. Additionally, learning science poses even greater challenges for them due to the difficulty of science subjects (Palmer, 2020). Notably, previous studies have found that social science-track students face greater difficulties and exhibit lower performance in science subjects when compared to their peers in the science track (Kao et al., 2022, 2023). This factor significantly influences their choice of the social science track over the science track. These prior findings align with the results of this study, which indicate a higher percentage of overage students in the social science track, constituting approximately 44.9%, while the science track had only about 29% of overage students.

The current study suggests that the primary focus on these common issues should strongly consider the plan for students with special educational needs, especially those who are overage. To address whether students might need such special educational support, implementing the individualized education plan (IEP) is imperative. In Cambodia, IEP has escaped teachers’ focus. Although IEP is commonly used with students with disabilities (i.e., Christle & Yell, 2010; Musyoka & Diane Clark, 2017), it is an important scheme to support students with special educational needs or at-risk students.

Curriculum and national examination

The current study found that students did not prioritize achievement and grades as the forefront of their academic goals. Most students opted for the social science track with a strong belief that national examination tests for the social science track were easier than those of the science track. It has been evidenced that every year, most students who register for the social science track have higher passing rates than those who register for the science track (see MoEYS, 2020a). Moreover, students were not motivated by achieving high grades when opting for the social science track; their primary objective was merely to pass the examination. These findings are deemed a critical deliberation on the implementation of curriculum and national examination.

These emerging issues strongly highlight shadow of secondary education reforms resulting from implementing various policies and projects at secondary education pertaining to its misalignments with curriculum and assessments, including national examination, for social and science tracks at secondary education (Brown & Niemi, 2007; Kurlaender & Larsen, 2013). Therefore, support projects necessitate a thorough reassessment of their inputs, outputs, outcomes, and impacts to ensure alignment with the intended curriculum and assessments of both tracks, with particular emphasis on the science track.

In the context of upper secondary education in Cambodia, the sole distinction between the curricula of both tracks lies in the allocated instruction hours (see MoEYS, 2010, 2020b). In this case, students of one track might lose interest in mastering the subjects regarded as least important of another track, which causes anxiety for both teachers and students (Silova, 2010). Such incidents constantly impose students to lose confidence in the efficacy of the track division. Especially the science track requires them to take more science subjects compared to social science track and unavoidably take more private tutoring classes. This incident makes students from low SES families unable to afford private tutoring fees (Pov et al., 2021) and consequently opt for the social science track.

Ensuring alignment between curriculum and assessments at the secondary and higher education levels is critical in preparing students for successful enrollment and the completion of certificates and degrees, especially in STEM-related majors (Brown & Niemi, 2007; Kurlaender & Larsen, 2013). This alignment can be understood as assessments (i.e., monthly and semester tests) and national examinations in secondary education defining the essential content and skills to be taught and learned at that level. These assessments can also serve as motivating factors for students pursuing further education. Moreover, college entrance exams serve as tools for evaluating a student’s potential for future success, regardless of their previous opportunities (Crouse & Trusheim, 1988; Kurlaender & Larsen, 2013). For instance, in California, a lack of alignment between high school coursework and university college requirements resulted in most enrolled students needing to take remedial or developmental courses in math and English (Brown & Niemi, 2007). Therefore, it is crucial to establish a strong link between secondary and postsecondary education through cohesive teaching methods, assessments, and policies to adequately prepare students for successful enrollment and degree completion (Kurlaender & Larsen, 2013) and reduce the tendency of STEM-major switching (Kao et al., 2022, 2023).

In this study, students reported that they opted from for social science track because examination tests of the social science track were easier than those of the science track. It is also suggested that an investigation into the efficacy of the implementation of curriculum and national examination of both tracks is imperative. The national examination tests of the social science track should not be easier than those of the science track. Test development should ensure that students of both tracks are treated fairly.

Private tutoring and socioeconomic status

This study revealed that the decline in enrollment in the science track was mainly attributed to private tutoring practices and family income. This phenomenon can be explained by the connection between the affordability of private tutoring classes and the financial constraints of families (Zhang & Xie, 2016). In this study, most students in the social science track came from families with lower incomes compared to their counterparts in the science track.

Private tutoring is a common practice in Cambodia, where students are often indirectly encouraged to participate in unofficial extra classes under the guise of “voluntary” activities (Bray, 1999, 2006; Bray et al., 2016; Brehm & Silova, 2014; Kinyaduka, 2014; Soeung, 2021). This practice is prevalent in science subjects. Students from financially disadvantaged backgrounds often cannot afford private tutoring fees, leading them to opt for the social science track over the science track. SES influences not only students’ performance in science (Smith & Gorard, 2011) but also their decision to pursue further studies in the field of science (Ainley et al., 2008). These issues contribute to social and educational inequalities in Cambodia (Bray, 2012; Dawson, 2010; Nhem & Kobakhidze, 2022). In Cambodia, Nhem and Kobakhidze (2022) assert that private tutoring continues to persist, even in the new model school system like NGSs. Students turn to private tutoring classes for various reasons, including the need to catch up with course content, additional practice, exam-related pressure, inadequate teaching quality, and lack of self-assurance (Nhem & Kobakhidze, 2022). The frequency of private tutoring classes students take varies depending on their family’s socioeconomic status (Pov et al., 2021). Higher family income is positively correlated with an increased likelihood of engaging in private tutoring (Zhang & Xie, 2016). Moreover, many students perceive science subjects as challenging and thus opt for supplementary practice or choose the social science track instead. Science-track students, particularly in subjects like mathematics, physics, chemistry, and biology, tend to enroll in more private tutoring classes (Kao, 2020; Kao & Kinya, 2020a).

Kinyaduka (2014) argues that private tutoring is a crucial means to attain high grades and prepare effectively for exams. However, in the Cambodian context, this study found that only science-track students valued private tutoring. In contrast, social science-track students, especially those from financially disadvantaged backgrounds, considered it costly and primarily focused on passing the examination, rather than achieving top grades. These findings align with studies conducted in Australia by Palmer (2020) and in three countries: Japan, Cambodia, and South Korea, as explored by Dawson (2010). For instance, Palmer (2020) identified factors like family socioeconomic status and the decision-making process as contributing to the decline in Australian students choosing science in their final years. Students often perceived science as expensive, difficult, and useful only for stereotypical scientific careers (Palmer, 2020).

In the context of Japan, Cambodia, and South Korea, Dawson (2010) argues that government efforts to address private tutoring through various policies have often proven ineffective and, in some cases, have unintentionally fueled its expansion (Dawson, 2010). In Cambodia, there is a strong emphasis on track division as a pivotal policy reform aimed at promoting science education at the secondary level and encouraging students to pursue STEM majors in Higher Education Institutions (HEIs) (MoEYS, 2010, 2020b). However, the implementation of this policy faces significant challenges, particularly considering its declining popularity among students, especially those from low-income households. This shift in student preferences is mirrored in teaching practices. Dawson (2010) rightly points out that pedagogical and curricular practices within the private tutoring systems have contributed to heightened levels of anxiety and insecurity related to the formal education system, ultimately serving the purpose of expanding the tutoring market (Dawson, 2010). Some teachers have reported not having sufficient time to cover the curriculum and, as a result, have offered additional teaching in private tutoring sessions (Bray & Lykins, 2012). In some cases, private tutoring sessions organized by classroom teachers can turn into a form of coercion when these teachers solely cover critical topics during these private sessions rather than during regular class hours (Bray, 2006). These practices not only exacerbate social inequalities but also introduce distortions into the curriculum (Silova, 2010). Curriculum content should be accessible to all students without necessitating private sessions, and schools must ensure equity in pedagogical and curricular practices. To engage students and encourage their participation in learning science subjects, innovative pedagogies, including the use of technology, should be integrated into teaching practices (Syed Hassan, 2018).

STEM major choice in higher education

The findings showed that most social science-track students did not plan to pursue STEM majors in higher education. This highlights a concerning trend of declining STEM-major enrollment in higher education in Cambodia (i.e., Eam et al., 2021; Kao, 2019; Kao & Kinya, 2019; Kao et al., 2022, 2023). Previous studies found that only 30% of science-track students in Cambodia registered for STEM majors at HEIs, and they were more likely to switch from STEM to non-STEM majors than social science-track students (Kao et al., 2022, 2023).

To address this issue, the study suggests strengthening the entrance examination for STEM majors at HEIs to encourage students to choose the right track at upper secondary schools and decrease the rate of switching. Cambodian universities should contemplate enhancing the quality and alignment of their entrance examinations to better align with students’ abilities and track choices. The entrance examination plays a pivotal role in determining which students are admitted to specific colleges or majors (Mori, 2002). Simultaneously, teaching, learning, and assessments within each track should align with related course entrance requirements, curriculum standards, and diploma policies in HEIs. Schools should implement and strengthen academic advising programs that emphasize well-defined and consistent intellectual and career interests, particularly aligning track choices with intended majors in higher education (Shaw & Barbuti, 2010). Science education should prioritize hands-on practical experiences both within and beyond the classroom (Kao, 2020). It is crucial to make science teaching and learning activities stimulating and captivating, as finding enjoyment in science education can foster a growing interest in STEM disciplines in higher education (Osborne et al., 2003). These initiatives will enhance the readiness of students in each track to advance further in their postsecondary education and reduce the rate of major switching (Kurlaender & Larsen, 2013; Shaw & Barbuti, 2010).

Furthermore, this study strongly advocates for an increased level of parental involvement in the domain of career guidance and counseling. The establishment of school-based counseling, involving not only management committees but also science teachers and faculty members from local HEIs, is crucial. These counseling sessions should be initiated at the start of each academic year, particularly targeting students in grades 10 and 12. Such initiatives play an instrumental role in igniting students’ interests in STEM-related majors and concurrently elevating their awareness of the multitude of available career prospects (Yesenia, 2016). Additionally, the study recommends a collaborative approach to STEM education at the school level, which has not been previously discussed in any previous studies (see Kao, 2020; Kao et al., 2022, 2023; Kao & Kinya, 2020a) or guidelines by MoEYS (see MoEYS, 2010). The lack of collaboration between school committees, parents, teachers, school principals, and students is a significant hindrance to promoting science learning and teaching activities in and outside schools. Developing a STEM education model with a collaboration paradigm can be a promising initiative for Cambodia.

Attitudes toward science

The current study revealed that the attitude toward science significantly predicted track choices. Social science-track students had extremely negative attitudes toward science. This finding is consistent with Kao et al. (2022) and Syed Hassan (2018). It was found that science-track students perceived higher attitudes toward science compared to social science-track students.

The decision to study science and pursue science-related careers heavily relies on attitudes toward science (Syed Hassan, 2018). A substantial body of literature has addressed this critical concern, offering insights to enhance students’ attitudes toward science through practical, policy, and theoretical approaches (refer to George, 2000; Gibson & Chase, 2002; Kao, 2020; Kind et al., 2007; Movahedzadeh, 2011; Nieswandt, 2005; Osborne et al., 2003; Syed Hassan, 2018). For instance, a study conducted in Malaysia, which measured attitudes toward learning science, identified a decline in science class enrollment linked to a lack of interest in science teaching (Syed Hassan, 2018). In Cambodia, Kao et al. (2022) similarly discovered that social science-track students exhibited poor performance and interest in science subjects. This problem arises from inadequate teaching quality in science, characterized by outdated methodologies (Tytler, 2007; Tytler & Osborne, 2012), which in turn impacts the decline in the choice of the science track (Palmer, 2020).

Syed Hassan (2018) suggests that effective science teaching involves teachers’ ability to connect theories with real-life contexts, foster supportive learning environments, and implement teaching strategies aligned with attitudes regarding affective, cognitive, and behavioral aspects. Moreover, beyond these previous studies, negative attitudes toward science may be attributed to the findings of the current study, which indicate that most students chose the social science track due to the ease of the national examination, reduced expenditure on private tutoring, high passing rates, and a lack of intent to achieve good grades. Therefore, the implications discussed in the “Curriculum and national examination” section should be taken into account to enhance students’ attitudes toward science.

Parents’ advice on track choices

Earlier studies conducted in Cambodia have overlooked the influence of parental guidance on students’ selection of academic tracks. The current research not only supports the idea that there is a lack of cooperation between parents and schools, but it also underscores the significance of parental advice. It found that most Cambodian parents encouraged their children to choose the social science track over the science track, with approximately 58.90% of parents making this recommendation.

The involvement of parents is an important factor affecting students’ academic outcomes, as noted in prior research (Boonk et al., 2018; Fan & Chen, 2001; Wilder, 2014). The results of this study reveal a decline in the extent to which parents are collaborating and actively promoting enrollment in the science track. This indicates a lack of coordination between parents and schools. While some schools conduct career guidance sessions for students to explain track choices and potential career paths, parents are not typically included in these sessions. Hence, it is plausible that parents guided their children to select the social science track solely based on their own preferences and insights, potentially lacking a comprehensive understanding of the distinct characteristics of track choices and their connection to higher education and future career prospects. They may have been unaware of the different academic benefits and career opportunities associated with each track. It is suggested that schools should involve parents in career counselling and guidance sessions to increase their awareness of track choices and career pathways. The changing trend of globalization toward STEM majors and government’s policy on STEM should be explicitly introduced to parents. The collaboration and involvement of parents are crucial to increasing enrollment in the science track (Kao & Kinya, 2020a).

School-level factors

School location

The study revealed that students from rural schools displayed a stronger preference for the social science track over the science track compared to their urban counterparts. Prior research has identified that many rural secondary schools in Cambodia lack the necessary infrastructure and resources to support effective teaching and learning of science subjects (Keo et al., 2021). To address this challenge, MoEYS has established Secondary Resource Schools (SRS) that are well-equipped with facilities and infrastructure, particularly for enhancing the quality of science education. These SRSs serve as central hubs for facilitating collaboration, sharing knowledge, and fostering unity among network schools within their respective communities (MoEYS, 2008). Unfortunately, these SRSs are predominantly located in urban areas. While network schools have the option to utilize the science materials available in the resource rooms of SRSs, students from rural schools face challenges in accessing these resources. The limited support for transporting students from rural schools to SRSs results in reduced access to science materials and facilities.

To address this issue, it is crucial to establish a close collaboration between SRSs and network schools to provide hands-on science learning experiences for students in rural secondary schools. Expanding the reach of SRSs or science resource rooms is a pivotal step in increasing access to science facilities and alleviating the accessibility constraints faced by network schools. Simultaneously, efforts should be directed at enhancing the effectiveness of utilizing the science resource rooms to positively influence students’ attitudes toward learning science.

Conclusion

This study highlights significant challenges and implications for improving policy and practices in secondary education in promoting science-track enrollment in Cambodia. It contributes novel insights to a relatively scarce literature on track-choice studies in Cambodia. The research has identified a range of individual- and school-level factors that strongly influence students’ preference for the social science track over the science track. It found that much variance in track choices accounted for individual-level factors, suggesting that individual-level factors had significantly stronger influence on students’ track choices than school-level factors. At the individual level, factors such as overage enrollment, the perceived ease of national examinations, high passing rates, a lack of pursuit for good grades, private tutoring expenses, choices of STEM majors, attitudes toward science, parental guidance, and family income significantly influenced students to choose the social science over the science track. At the school level, only the geographical location of the school has a significant impact on track choices. Students from rural schools expressed a greater preference for the social science track compared to their counterparts from urban schools. To improve the implementation of track division and increase enrollment in the science track, it is advisable to address issues related to dropout rates, academic repetitions, academic performance, and late school entry. Refinement of the curriculum and assessments, particularly national examinations, is essential to provide equitable opportunities and align with higher education policies. Schools should ensure fairness in teaching practices and introduce innovative teaching methodologies in science subjects. Parental involvement in career guidance is crucial. Effective science teaching should focus on linking theory with real-life contexts, creating supportive learning environments, and aligning teaching strategies with students’ attitudes toward learning science, addressing affective, behavioral, and cognitive aspects.