Academic, Social, and Economic Factors that Shape the Pursuit of Degrees in STEM Fields

Introduction

Research on participation in science, technology, engineering, and mathematics (STEM) majors consistently shows that entry into STEM fields is shaped by processes unfolding well before attending college. Rather than being determined solely by late-stage academic performance or individual preference, STEM major selection reflects a cumulative set of experiences involving academic trajectories, educational expectations, identity development, family resources, and institutional access. Using large-scale longitudinal datasets, researchers have examined how these factors interact across childhood, adolescence, and early adulthood to determine who enters STEM majors and who does not.

This Research Spotlight synthesizes findings from secondary analyses of data collected in nine datasets that are distributed at ICPSR, which contain a significant number of education variables. They include: (1) the Longitudinal Study of American Youth (LSAY); (2) the NICHD Study of Early Child Care and Youth Development; (3) the India Human Development Survey–II (IHDS-2); (4) the National Education Longitudinal Study (NELS); (5) the High School Longitudinal Study of 2009 (HSLS:09); (6) Metastereotypes of Women in STEM; (7) the Education Longitudinal Study of 2002 (ELS:2002); (8) the Advanced Placement Science Impact Study, United States, 2013–2016; and (9) the College and Beyond II Alumni Survey. The papers discussed here used one or more of these datasets to address STEM majors directly, through STEM degree completion or STEM major selection, or indirectly, by examining the academic, psychological, and structural conditions that precede entry into STEM fields. The Spotlight is organized into four themes around access to STEM education and career opportunities: early academic development and the formation of STEM pathways; psychological barriers and motivators in STEM education; unequal access and hidden diversity in STEM educational settings; and socioeconomic status (SES), institutional access, and STEM majors.

Early academic development and the formation of STEM pathways

A prominent theme across the literature is the importance of early academic development and educational expectations in shaping long-term STEM outcomes. Several works using the Longitudinal Study of American Youth (LSAY) demonstrated that middle school represents a critical period in the formation of STEM pathways. LSAY is a nationally representative, National Science Foundation (NSF)-funded longitudinal study that tracked students from middle and high school into adulthood to examine their achievement, attitudes, and career pathways in science, mathematics, and engineering. The data collected in LSAY enable research on how family, school, teachers, peers, and informal learning experiences shape long-term educational and occupational outcomes in STEM.

Using LSAY data from a 7th grade cohort followed into adulthood, Yeung et al. (2026) examined whether middle-school science performance predicts later STEM outcomes. The study found that both the initial level and the growth rate of science achievement across grades 7 through 9 predicted whether individuals later earned a STEM bachelor’s degree and entered a STEM profession. The authors also found that four social factors (parental encouragement, peer support, a positive school environment, and student self-esteem) measured in 7th grade each individually predicted students’ starting level of science achievement, while parental encouragement and school environment additionally predicted its rate of growth. These dynamic learning trajectories served as a critical bridge, sequentially linking early school experiences to STEM degree completion and, eventually, to entry into a STEM profession in adulthood, a process the authors described as a “positive cascade” initiated by a supportive middle school environment.

Science performance trajectories, however, do not develop in isolation from students’ broader academic motivations. Yeung et al. (2025), drawing on the same dataset, found that students’ educational expectations evolved alongside their science scores during middle school, and that both trajectories predicted long-term outcomes. Growth in educational expectations was associated with growth in science performance, and these linked trajectories together predicted the likelihood of completing a four-year STEM bachelor’s degree in adulthood.

While those findings traced STEM trajectories beginning in middle school, other research using a different national dataset suggests the relevant developmental window opens even earlier. Bustamante et al. (2023) utilized the NICHD Study of Early Child Care and Youth Development (SECCYD) to examine whether the quality of care children receive in their first five years predicts STEM success at age 15. The study found that two dimensions of caregiving quality (caregiver sensitivity and responsiveness) and cognitive stimulation were linked to STEM achievement and school performance in high school, primarily through an indirect pathway. It started with high-quality early care, which predicted stronger STEM proficiency in grades 3 through 5, which in turn predicted better STEM outcomes at age 15. These effects were particularly pronounced for children from low-income families, for whom exposure to high-quality early care helped reduce achievement gaps that would otherwise persist into adolescence. 

The finding that early disadvantage shapes later STEM participation is not unique to the American context, though the mechanism operates differently when access to quality early care is replaced by acute environmental shocks. Using the Young Lives Study (YLS) and the India Human Development Survey–II (IHDS-2), Dhamija and Sen (2022) examined the impact of early-life rainfall shocks, an exogenous source of malnutrition, on later educational pathways in India. The authors found that these shocks negatively affected cognitive development at ages 5 and 8, and modestly reduced the likelihood of completing primary, middle, and secondary school. Among students who remained enrolled long enough to reach the higher secondary level, no statistically significant direct effect on STEM subject choice was detected. However, the authors argued that this null result should not be interpreted as evidence of no harm, as their research showed that the children most damaged by early shocks had already exited the education system before STEM choices became available. Early disadvantages thus shaped STEM participation not by directly deterring subject selection among those who remained, but by determining who survived long enough in the education system to make that choice at all.

Psychological barriers and motivators in STEM education

Beyond academic preparation, research drawing on three independent US national datasets — NELS, HSLS:09, and LSAY — consistently identifies psychological factors, including self-efficacy, identity, math anxiety, and gender stereotypes, as significant forces shaping who enters and remains in STEM pathways. The fact that these patterns hold across different datasets and student populations makes them difficult to dismiss as a coincidence. Using the National Education Longitudinal Study (NELS), Dangur-Levy (2026) examined whether the gender gap in physical STEM fields could be explained by math self-efficacy. This refers to self-beliefs in learning mathematics and completing mathematics-oriented tasks, according to Dangur-Levy. The results showed that math self-efficacy accounted for about 10 to 12 percent of the gender difference in STEM enrollment and completion. However, this effect became insignificant once prior academic performance and family background (e.g., test scores and income) were taken into account, suggesting that confidence operates within broader structural and academic conditions. The study also found that higher math self-efficacy increased the likelihood of STEM outcomes, but this effect was stronger for young men than for young women. 

Research using a more recent dataset suggests self-efficacy is only one of several psychological factors at work, and not necessarily the most powerful. Findings using the High School Longitudinal Study of 2009 (HSLS:09) consistently highlight STEM identity as a particularly strong predictor of STEM pathways. HSLS:09 was a nationally representative longitudinal study that followed a cohort of US students from 9th grade (starting in 2009) into high school, postsecondary education, and the workforce. The survey focused especially on how students make decisions about courses, college, and careers, including pathways into and out of STEM fields. Using HSLS:09, Marsh et al. (2024) examined four key attitudes in mathematics and science, and their relationships with students’ achievement, course-taking, and intentions to major in STEM. The attitudes included “self-efficacy,” referring to students’ confidence in their ability to succeed in math or science tasks; “identity,” reflecting whether they saw themselves (and were seen by others) as a math/science person; “interest” captured their enjoyment of these subjects; and “perceived utility” referred to usefulness of these subjects for everyday life, education, and careers. While all four attitudes contributed to STEM outcomes, identity emerged as the most consistent and one of the strongest predictors, especially for STEM major choice, and played a key role in explaining gender differences. The study also showed that lower STEM participation among racially minoritized students and those with a lower socioeconomic status was not due to weaker attitudes. Rather, their findings suggested it was due to structural factors, such as differences in prior achievement and access to advanced coursework, that limit how positive attitudes translate into STEM pathways.

Negative experiences, like math anxiety, can derail STEM trajectories even among students with otherwise strong foundations. Ahmed (2018) utilized data from the Longitudinal Study of American Youth (LSAY) to examine math anxiety trajectories across adolescence (grades 7 to 12), identifying four distinct patterns associated with various levels of math anxiety, which was defined as a feeling of tension and anxiety that interferes with the manipulation of numbers and the solving of mathematical problems in both academic and everyday contexts. The study found that students in groups with consistently low or decreasing math anxiety were up to 7.4 times more likely to enter STEM occupations as adults compared to students with consistently high anxiety. 

Also using HSLS:09, Starr and Simpkins (2021) examined one key social force that shapes math anxiety: the gender stereotypes students encounter and internalize during adolescence. They found that adolescents’ stereotypical beliefs became more traditional between 9th and 11th grade and that parental stereotypes were significantly associated with those beliefs. Math and science identity was understood as the extent to which students see themselves, and believe others see them, as a “math” or “science” person. These stereotypes were linked to identity in gendered ways. Endorsing traditional stereotypes was associated with lower identity among girls and higher identity among boys. In turn, identity was related to key STEM outcomes, including advanced coursework and STEM career goals.

Unequal access and hidden diversity in STEM education settings

Unequal access to STEM varies by race, gender, ethnicity, and context. Focusing specifically on Black female participants in the HSLS:09, Edosomwan et al. (2026) examined how perceptions of teacher treatment relate to math and science identity. They found that when Black female students perceived their math and science teachers as fair and respectful, they reported stronger STEM identities. They also found that Black female students were often overlooked in STEM research, and creating supportive classroom environments where teachers actively value and respect Black female students was a critical “social justice” intervention for keeping them in the STEM pipeline.

Kang et al. (2023) also used HSLS:09, and they found that treating Asian Americans as a single group can conceal substantial variation in STEM participation across ethnic subgroups. Focusing on five Asian ethnic subgroups—Chinese, Filipino, Vietnamese/Thai, Indian/Sri Lankan, and Korean/Japanese—Kang and colleagues found that STEM major selection varied substantially across subgroups and that these patterns shifted in meaningful, sometimes opposing, directions depending on college selectivity. For instance, Vietnamese/Thai students were among the most likely to choose STEM majors at two-year colleges, but showed comparatively low rates of STEM selection at four-year institutions. The authors also found that differences across subgroups were only partially explained by prior math achievement, indicating that additional structural and contextual factors shape STEM pathways. They argued that treating Asian Americans as a monolithic group also actively reinforced the “model minority” myth, making underrepresented Asian ethnic subgroups, particularly Southeast Asian students, invisible within broader conversations about equity and access in STEM.

Hsieh and Simpkins (2022), drawing on HSLS:09 data, found that even a factor as consistently positive as parental support operates very differently depending on both race and parental STEM background. They found that the influence of a family background that included STEM motivation was far from uniform and depended heavily on both parental STEM background and race. They showed that parental STEM support positively predicted motivational beliefs for adolescents whose parents did not hold STEM degrees or occupations, but was not significantly associated with motivational beliefs among those whose parents did have STEM backgrounds. The racial pattern added another layer of complexity. Among students in non-STEM families, parental support predicted motivational beliefs for Asian, Latina/o, and White adolescents, but these associations did not emerge for Black adolescents, regardless of whether their parents had STEM backgrounds. The authors interpreted this not as evidence that parent support was unimportant for Black families, but as a signal that Black adolescents may face structural barriers, stereotypes, and microaggressions so pervasive that they suppress the influence of parental support of STEM in ways that did not apply to other groups. 

Using the Metastereotypes of Women in STEM, Gilrane et al. (2019) examined how women faculty in STEM experienced gender stereotypes through metastereotypes, or beliefs about how others perceive women in their profession, and how these beliefs related to impression management behaviors such as self-promotion and ingratiation. The study found little evidence that metastereotypes predicted these behaviors, but it did find that supervisor perceptions of these behaviors shaped likability evaluations. Specifically, when female faculty were ingratiating, supervisors rated them as likable, but self-promoting female faculty were less likely to be rated as likable, suggesting a backlash effect. Neither behavior significantly improved supervisor ratings of competence. Gilrane et al. argued that institutions should address unfair likability standards and gendered expectations rather than placing the burden on women to manage these biases themselves. 

Socioeconomic status, institutional access, and STEM majors

Research drawing on ELS:2002, HSLS:09, the Advanced Placement Science Impact Study, and the College and Beyond II Alumni Survey traced STEM access barriers across multiple stages of the educational pipeline. Niu (2017), using ELS:2002, found that family SES did not directly predict whether students chose a STEM major, but it powerfully shaped how other factors operated. For higher-SES students, math preparation and STEM enrollment aligned closely: stronger math scores meant a higher likelihood of choosing STEM. For students from the lowest-SES families, that alignment disappeared entirely. SAT math scores stopped predicting STEM enrollment, suggesting these students were making major choices without the guidance needed to accurately assess their own readiness. Niu attributed this not to lack of motivation but to lack of resources, including informed parents, knowledgeable counselors, and role models who could help students with preparation and decision-making.

One concrete mechanism through which SES shapes STEM access is advanced coursework. Jewett and Chen (2022), using HSLS:09, found that advanced placement (AP) STEM course-taking significantly increased the likelihood of choosing a STEM major in college, and that this relationship was notably stronger for female students than for male students. As the number of AP STEM courses taken increased, the gender gap in STEM major selection narrowed and eventually reversed, suggesting that AP exposure may be a particularly meaningful lever for women considering STEM. However, no comparable moderating effect was found for race or ethnicity, indicating that expanding AP access alone would be unlikely to close racial disparities in STEM representation. 

The impact of AP science on students, not just their major choices, is a harder question, and one that Conger et al. (2021) investigated. Drawing on the Advanced Placement Science Impact Study, they found evidence suggesting that AP science improved students’ scientific inquiry skills and increased their interest in pursuing a STEM degree. But the same students also reported higher stress and lower confidence in their ability to succeed in college science, and lower grades than peers in other science courses. 

The uneven benefits of STEM coursework access extend beyond gender and race. Peterson (2021), using HSLS:09, found that disability status interacted with family income in ways that made the relationship between support structures and STEM course completion far more complex than either factor alone suggested. Peterson examined whether receiving accommodations under an Individualized Education Program (IEP) and coming from a low-income family predicted the number of STEM courses students completed in high school. When the IEP status was considered alone, it showed no significant effects. But students who had an IEP and qualified for free or reduced-price lunch completed fewer STEM courses, while those with IEPs from higher-income families actually completed more. The two groups effectively canceled each other out in models that did not account for income. Only when the interaction between IEP and family income was included did the pattern become visible. In other words, having an IEP did not guarantee equal access to STEM coursework. The financial circumstances surrounding a student’s education shaped how much support was actually delivered.

Quadlin et al. (2021), using HSLS:09, examined parental college savings through an intersectional lens of race and gender, finding that both race and gender shaped not only whether parents saved for college but also how much they saved. But while parental savings tended to rise alongside academic achievement for White boys, White girls, and Black boys, this relationship was virtually absent for Black girls. Even the most academically qualified Black girls received savings comparable to the least academically qualified, a pattern the authors traced to the fact that high-achieving Black girls tended to come from families that were considerably less well-off than equally qualified peers in other race-gender groups. As a result, families of Black girls were more likely to rely on private loans with less forgiving repayment terms, potentially deepening financial inequalities across generations. 

The financial pressures that shape access to STEM do not resolve themselves at graduation. Agbonlahor (2025), using the College and Beyond II Alumni Survey, found that debt burden had no significant effect on STEM graduates’ civic engagement in either direction, while liberal arts graduates with moderate to high debt actually showed increased political engagement. The author attributed this difference not to debt levels but to disciplinary culture: liberal arts education appeared to provide frameworks for understanding financial strain as a systemic issue worth engaging politically, while STEM education did not cultivate the same response. For students who navigated significant economic barriers to reach STEM fields, the discipline they worked so hard to enter may offer little in the way of tools for connecting their financial experience to civic life.

Conclusion

This Research Spotlight does not reflect all of the existing research regarding factors that shape the pursuit of degrees in STEM fields. To see how each of the ICPSR studies mentioned in this Spotlight has been examined in other scholarly literature, to gain ideas for extending prior research, or to conduct a larger literature review, you can search the ICPSR Bibliography of Data-related Literature. Using search terms like “STEM” will lead you to search results containing publications that are linked to the study data analyzed in them. Discovering data via the literature in this way can begin your investigation of the existing and potential uses of the data distributed by ICPSR.

When authoring publications that include your secondary analysis of study data downloaded from ICPSR, be sure to cite the study data in the publication’s references section, using the provided data citation and unique identifier (in the form of a URL containing a DOI). Once your paper is published, submit its citation to the ICPSR Bibliography via this form, so it can be added to the ICPSR Bibliography of Data-related Literature, enabling others to find, learn from, and cite your work.