Content uploaded by Yannan Gao
Author content
All content in this area was uploaded by Yannan Gao on Jan 14, 2025
Content may be subject to copyright.
Not a pipeline but a highway: Men’s and women’s STEM career
trajectories from age 13 to 25
Yannan Gao
a,b,c,*
, Jacquelynne S. Eccles
a
, Anna-Lena Dicke
a
a
School of Education, University of California, Irvine. 401 E. Peltason Drive, Suite 3200, Irvine, CA 92617, USA
b
Division of Arts and Sciences, New York University Shanghai. 567 Yangsi W Road, Pudong District, Shanghai 200124, China
c
Hector Research Institute of Education Sciences and Psychology, University of Tübingen, Europastraße 6, Tübingen 72072, Baden-Wüttemberg,
Germany
ARTICLE INFO
Keywords:
“Leaky pipeline”
STEM career trajectory
STEM attrition
STEM entry
Gender comparison
ABSTRACT
Concerns with diversifying and expanding the STEM workforce have elicited extensive efforts to
increase women’s adherence to a “no leak pipeline”to match that of men. However, is such a
trajectory optimal for boosting women’s STEM career attainment? If so, among which types of
STEM occupations? Studies often suggested that women are underrepresented in the pipeline of
“white-collar”, mathematical, engineering, physical, and computer science (MEPCS) occupations,
but to what extent does this conclusion hold true among other types of STEM careers? To answer
these questions, we plotted men’s and women’s STEM career trajectories from age 13 to 25 using
a U.S. national longitudinal sample and categorized occupations by domain-specic knowledge (i.
e., non-STEM, MEPCS, or LEHMS [life, ecological, health and medical sciences]) and by educa-
tional requirement (i.e., “blue-collar”non-STEM, "blue-collar" STEM, "white-collar" non-STEM, or
“white-collar”STEM). We found gender similarities in STEM attrition, gender differences in STEM
entry, and gender differences in STEM career attainment trajectories. For example, STEM workers
rarely took a “no leak pipeline”, except women in LEHMS occupations. Moreover, tracking the
size of STEM workforce longitudinally, we found that though small, women’s MEPCS workforce
expanded to nearly twice its size as participants grew from age 13 to 25; in contrast, the LEHMS
workforce shrank to less than a third of its size among both men and women. Our results specify
aspects in which STEM trajectories of men and women differ across various types of STEM oc-
cupations and thus provide an updated understanding of gendered STEM career trajectories.
1. Introduction
In 2023, the National Science Foundation (NSF) in the U.S. invested 95 % of its 9.5-billion-dollar budget in STEM (Science,
Technology, Engineering, and Mathematics) research, facilities and education (National Science Foundation, 2024). Promoting
women’s attainment of a STEM occupation is integral to the future STEM workforce. Thus, a detailed understanding of the STEM
training pipeline for both women and men is key to increasing gender inclusiveness in the STEM workforce, a major goal of the NSF.
Our study was designed to examine this question in a current national longitudinal sample.
Our study is driven by research development in three realms. First, women’s STEM participation pathways look different now from
* Corresponding author at: Hector Research Institute of Education Sciences and Psychology, University of Tuebingen, Europastrasse 6, Tübingen,
Baden-Wurttemberg 72072, Germany.
E-mail address: yannan.gao@uni-tuebingen.de (Y. Gao).
Contents lists available at ScienceDirect
Journal of Vocational Behavior
journal homepage: www.elsevier.com/locate/jvb
https://doi.org/10.1016/j.jvb.2024.104067
Received 19 January 2024; Received in revised form 18 November 2024; Accepted 21 November 2024
Journal of Vocational Behavior 156 (2025) 104067
Available online 7 December 2024
0001-8791/© 2024 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
( http://creativecommons.org/licenses/by/4.0/ ).
decades ago when the “leaky pipeline”metaphor came into use (Berryman, 1983). By 2020 there were both reductions in male
dominance in some STEM majors and ips from male to female majorities in others (e.g., the medical and biological sciences, National
Center for Education Statistics, 2019;National Science Foundation, 2023); on the other hand, small but sizable gender disparities
continued in physics majors and large gender gaps existed in mathematics, computer science and engineering majors (National Science
Foundation, 2023). Thus, an updated investigation is needed for specifying the STEM elds in which women’s high attrition rates exist.
Second, the denition of “STEM”has developed over time, extending beyond physics, engineering, mathematics, and computer
science elds to include medical, pharmaceutical and biological elds, as well as numerically demanding social sciences (e.g., eco-
nomics, some highly quantitative areas of psychology, sociology). Additionally, increasing attention has been paid recently to STEM
occupations that do not mandate a Bachelor’s degree but do require substantial technical and scientic knowledge and skills in applied
settings. These “blue-collar" STEM occupations recruit a higher share of underrepresented and underserved students than the "white-
collar" counterparts (National Science Board, National Science Foundation, 2021), providing individuals with opportunities for
lucrative employment and upward mobility.
Third, more recently, researchers have contended that the “leaky pipeline”is a biased and inaccurate representation of the
developmental trajectories associated with STEM labor supply. It implies that leaving STEM elds is a bad consequence rather than a
growth process to lead young adults into more personally rewarding occupations (Metcalf, 2010). Moreover, empirical evidence calls
for a shift of focus from STEM attrition to alternative, more developmentally accurate descriptions of current STEM career paths
(Cannady et al., 2014). For example, using a national longitudinal sample of college students, Xie and Shauman (2003, Chapter 4)
found that college students not only left STEM majors for non-STEM majors, but also entered STEM majors from non-STEM majors.
Others have advocated that STEM career trajectories are diverse, individualized, and recursive (Babarovi´
c, 2021;Blikstein &Worsley,
2016;Metcalf, 2010). In this regard, a singular, linear path as the “pipeline”metaphor implies may not only stigmatize individuals
taking alternative STEM career pathways, but also constrain or even mislead efforts to expand the STEM labor supply.
In this study, we aim to address these issues by describing the STEM career trajectories of a current longitudinal national sample of
U.S. adolescents/young adults. We differentiated career choices along two dimensions: domain-specic knowledge and educational
requirements. We then identied and analyzed STEM attrition, entry and career attainment trajectories of women and men as they
grew from age 13 to 25. Lastly, we calculated the change in the size of the STEM workforce by gender to examine the extent to which a
STEM “pipeline”shrinks (or expands) for different types of STEM careers among men and women respectively.
1.1. Limitations of the “leaky pipeline”metaphor
In the mid-twentieth century, the STEM labor policy community became concerned about the need to expand and diversify the
STEM workforce. The “leaky pipeline”metaphor was useful for highlighting the need to investigate gender differences in the pathways
to STEM employment (Berryman, 1983). It drew attention beyond the phenomenon of the gender disparity in the STEM workforce to
the developmental antecedents of the problem. However, criticisms have accumulated in recent decades that the metaphor is a biased
and oversimplied representation of STEM career trajectories.
The “pipeline”implies that STEM career trajectories are linear and one-directional –starting with a large group of individuals
aspiring to STEM careers but gradually and steadily shrinking as individuals grow through their adolescent and young adult years.
Given that the metaphor originated from concerns about increasing the STEM labor pool, leaving the potential pool was seen as bad. In
addition, the linear “pipeline”metaphor could not represent recursive career paths, such as re-entry or later entry into STEM elds
amidst education and training stage (Ma, 2011;Metcalf, 2010). Developmentalists interested in adolescence argue that shifts in career
goals represent an adaptive process of vocational exploration, planning, and selection. During adolescence and early adulthood,
changes in career choices can result from timely updates in adolescents’career perceptions and attitudes or useful explorations of
alternative careers as individuals learn more about themselves and the nature of the adult labor market towards which they are moving
(for theoretical conceptualizations, see Gottfredson, 1981;Malanchuk et al., 2010;Stephen et al., 1992). Thus, a linear and unidi-
rectional path is likely to be oversimplied and inaccurate. Specically, the emphasis of the metaphor on “leakage”overshadows
entrance into STEM eld. Due to differences in personal characteristics and social background, some teenagers may identify or specify
their career goals at a later time point among the various options that are available and comparably desirable to them (Messersmith
et al., 2008). In a study tracking high school students transitioning into universities across the US, Xie and Shauman (2003, Chapter 4)
found that a sizable portion of STEM baccalaureates changed their career paths from non-STEM to STEM by their sophomore year.
Therefore, the exclusion of later entries into STEM elds can lead to a biased representation of STEM workers by leaving out this
subgroup of adolescents.
Furthermore, the “leaky STEM pipeline”is unable to represent individual differences in STEM career trajectories. Sociologists and
developmental psychologists have underscored the individualized nature of career development, because of the unique combination of
personal characteristics and social context of each adolescent and the unique accumulation of such experience over time (Eccles, 2009;
Eccles &Wigeld, 2020;Hirschi, 2011). If the metaphor implies that an early and persistent trajectory is optimal for promoting STEM
career attainment or boosting STEM labor supply, a careful examination is required to specify among which group of individuals and to
which types of STEM careers this assumption applies. Xie and Shauman (2003, Chapter 4) found that among science/engineering
baccalaureates in the US, a persistence path was the most common career path of men, whereas an entry from a non-science into a
science major was the most common path of women. Two recent longitudinal studies supported gender differences in STEM career
attainment paths (Cimpian et al., 2020;Ma, 2011). However, such studies remain scarce, especially those covering a longer and earlier
developmental period on a more inclusive population beyond the college-attending individuals during their postsecondary education
years. In this study, we aim to address this issue by identifying individual differences in the longitudinal patterns of STEM career
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
2
choices, with a focus on gender differences in such patterns from age 13 to 25.
1.2. Specifying the denition of STEM careers
Existing STEM classications are inconsistent in their denitions of "STEM". Specically, some female-dominant scientic elds,
namely psychology, life/ecological sciences, health and medical sciences are included in some STEM classications (e.g., Department
of Homeland Security, 2022;National Science Board, National Science Foundation, 2021) but not others (e.g., National Science and
Mathematics Access to Retain Talent [the SMART Grant], U.S. Department of Education, 2009). Such inconsistency led to divergent
conclusions about the landscape of gender composition in STEM. Manly et al. (2018) compared women’s STEM degree attainment
based on ve STEM categorizations and found that women would hold 15 to 50 % of STEM Bachelor degrees across the different lists of
STEM majors. The STEM majors dened by the SMART grant, which excludes social science majors, consisted of the fewest women
baccalaureates, whereas the STEM elds dened by NSF constituted half of women. Thus, the inconsistent denitions could cause
mixed ndings of the magnitude of the gender gap in what is considered as “STEM”.
The inconsistency in STEM denitions is partially rooted in the lack of clear categorization criteria. STEM classications mentioned
above might have reected the demand for STEM immigration workers in the U.S. labor market, the structure of postsecondary
training programs, the disciplinary paradigms and skills, or a combination of these factors. However, the criteria of the STEM clas-
sications were not specied in the major STEM indexes, hindering reasoned uses of the existing STEM lists. Furthermore, the absence
of a clear classication methodology has created barriers for informed syntheses of the scientic literature. In their review of recent
quantitative journal articles about gender disparity in postsecondary STEM education, Manly et al. (2018) found that about 40 % of
papers did not report the classication criteria of STEM majors. Considering the pivotal role of a clear operationalization of STEM in
research transparency, in this paper, we attempt to offer one such example. Specically, we create our STEM classication based on a
priori research aims and specify our methodological details. Guided by our research aims to understand gender segregation and social
mobility in the STEM workforce through individuals’career choices, we differentiate STEM careers along two dimensions: domain-
specic knowledge and educational requirements.
1.2.1. By domain-specic knowledge: LEHMS versus MEPCS
A distinct contrast exists regarding the distribution and the career motivation of women and men in life, ecological, health and
medical sciences (LEHMS) versus mathematical, engineering, physical and computer sciences (MEPCS). In the past half-century,
LEHMS have achieved remarkable progress in reducing gender disparities, with 64 % Bachelor’s degrees awarded to women in
2020 (National Science Foundation, 2023). Meanwhile, gender disparities continue to persist in MEPCS, with 37 % Bachelor’s degrees
in physical sciences and 23 % in computer sciences awarded to women between 2011 and 2020. Additionally, women earned only 20
to 25 % of the two-year and four-year college degrees in engineering (National Center for Education Statistics, 2019), and this number
remained low by 2020 (National Science Foundation, 2023). Such gender aggregation within STEM exists not only in education but
also in the workforce (National Science Board, National Science Foundation, 2021). Psychologists found that the varied gender dis-
tribution within STEM could be explained by differences in men’s and women’s career motivation, such as the pursuit of altruistic
values, the goal to promote the interest of others or the interest of self, the preference for working with people or tools (Diekman et al.,
2017;Su &Rounds, 2015;Wegemer &Eccles, 2019). Such motivational underpinnings highlight women’s and men’s interest in
different content knowledge within STEM. Given the gender differences in career choices and motivation, we differentiate MEPCS
versus LEHMS domains in this study.
1.2.2. By educational requirement: “blue-collar”versus “white-collar”STEM careers
Within the STEM industry, “blue-collar STEM”occupations, which do not require a STEM Bachelor’s degree, earn 40 % less than
their “white-collar”counterparts (Rothwell, 2013). These “blue-collar STEM”jobs are often overshadowed by the education-
demanding, "white-collar" STEM jobs, despite that the former make up 30 to 50 % of employment opportunities in STEM-related
elds (e.g., Carnevale &Strohl, 2010;Fayer et al., 2017). Moreover, “blue-collar STEM”jobs play an integral role in the STEM in-
dustry (Rothwell, 2013). In manufacturing and engineering industries, “blue-collar STEM”workers are at the forefront of producing,
maintaining, and repairing machines and products; they are crucial for reducing design defects, innovating production procedures,
extending machine service life and improving production efciency. In the healthcare industry, “blue-collar STEM”workers take over
healthcare and hygienic routines, administer examinations, and supervise and educate patients, making them key to the imple-
mentation of treatments. Lastly, supporting the “blue-collar STEM”industry has consequential implications for underserved students.
The “blue-collar STEM”occupations mainly recruit students from two-year colleges, which serve a large portion of individuals from
underrepresented racial/ethnic groups and low-income families (Clotfelter et al., 2013). At the same time, “blue-collar STEM”jobs pay
higher salaries than “blue-collar non-STEM”jobs and were projected to grow faster than "white-collar STEM" jobs (Fayer et al., 2017;
Rothwell, 2013). Therefore, including the understudied "blue-collar STEM" careers as part of the STEM industry can better support the
career development of underserved youths in a promising sector. In this study, we include this set of STEM jobs, as well as non-college-
attending adolescents in our sample, to provide a broader picture of STEM career trajectories in the US today.
1.3. The present study
In this study, we examined the STEM attrition, STEM entry, STEM attainment trajectories, together with the change in the size of
the potential STEM workforce among men and women in different subtypes of STEM jobs. We investigated the attrition rate from, the
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
3
destination of the attrition into, and the trajectories of the STEM attrition over time to understand “how much”,“where”, and “when”
STEM attritions occur. Similarly, we analyzed the entry rate into, the origin of the entry from, and the trajectories of the STEM entry to
describe the various aspects of STEM entry. Next, integrating STEM attrition, entry and persistence, we tracked the change in the size of
the potential STEM workforce in both gender groups. Lastly, we identied trajectories of STEM workers to explicitly examine how
often STEM workers adhered to a "no leak pipeline" in their STEM eld of employment in adolescence.
We plotted career paths between age 13 and 25 with two additional time points around age 18 and 21 to capture meaningful
developmental trends. The starting time around age 13 was chosen because by early adolescence, teenagers become cognitively adept
to contemplate on abstract concepts, such as their future self (Nelson, 2017). Thus, probing career aspirations can yield more
meaningful responses than at a younger age (Auger et al., 2005). Additionally, adolescents typically start to participate in more
specialized STEM curriculum in middle school in the US (Xie et al., 2015), thus improving their understanding of various STEM
subjects. At age 18, most participants had nished their high school and were transitioning into their next stage of career, such as a
college major or a job in the workforce; hence, their career goals may further develop by this turning point. Age 21 marks the next
developmental stage when participants in postsecondary education or the workforce passed their transitional period, and those
attending a two-year college were near the end of their education. By age 25, the vast majority of participants had nished their
education and entered the workforce. Thus, these time points could represent major developmental milestones for both college-
attending and non-college-attending adolescents.
In this study, we aim to answer the following research questions:
1. What are men’s and women’s STEM attrition rates, directions and timing?
2. What are men’s and women’s STEM entry rates, directions and timing?
3. In what ways and to what extent does the size of potential STEM workforce change among men and women as they grew from age
13 to 25?
4. What are men’s and women’s STEM career attainment trajectories?
All analyses were carried out for men and women respectively followed by statistical tests for gender differences.
2. Method
2.1. Sample
Our sample consisted of 814 adolescents in the original cohort of the Child Development Supplement (CDS) of the Panel Study of
Income Dynamics (PSID) launched in 1997 (54 % women; 48 % White, 41 % Black, 6 % Latino, 5 % other; average age =13.12 years
old, SD =1.94 years old in 2002). They were children of families in the PSID main study, a national sample of US families with an
oversampling of low-income families. This sample provides us with a valuable opportunity to address the underrepresentation of non-
college-attending individuals and blue-collar STEM industry in the literature on STEM career development. Particularly, CDS effec-
tively retained disadvantaged adolescents with data collection through families because these teenagers tend to drop out of schools
early compared with their peers (Rumberger, 1983).
2.2. Procedure
PSID researchers interviewed participants in person at their home or via phone based on protocols and scripts. In 2002, adolescents
were interviewed about their career aspirations the rst time and were followed up repeatedly in 2005, 2007, 2009 and 2011. In 2013
or 2015, participants, who became young adults, were asked about their current employment. Open-ended responses were collected
and documented in the form of Census 2000 occupation codes by the PSID project team. The 1st author of this paper recoded the data
for analyses.
2.3. Measures
2.3.1. Career aspirations
One open-ended question was used to ask about participants’career aspirations: “What are the three kinds of jobs you would like to
have when you are done with school?”This question was only asked to CDS participants over the age of 12, who became cognitively
mature enough to reect on abstract concepts and got exposed to subject-differentiated STEM curricula (Nelson, 2017;Xie et al.,
2015). Up to three open-ended responses were recorded for each participant in the order of being mentioned. If multiple aspirations
were named, the partipant was asked to identify the most desired one. The only mentioned or the most desired occupation was used for
our analyses. For participants of age 18 or older, the question was worded as “What job would you most like to have when you are 30?”
Questions about career aspirations were asked in CDS and Transition to Adulthood Supplement (TAS) of the PSID project in 2002,
2005, 2007, 2009 and 2011. Responses were recorded in a 3-digit 2000 Census occupation code in the PSID dataset.
2.3.2. Main occupation
In 2013 or 2015, participants were asked about their current job or their most recent job in the past two years, including self-
employed jobs. If the participant worked in multiple jobs, they were asked to identify their main job. The questions were part of
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
4
the interviews of TAS and the PSID main study after participants reached age 18. Responses were recorded in 2000 Census occupation
codes the same way as career aspirations.
2.4. Data preparation
2.4.1. Census-O*NET translation
Original Census occupation codes documenting career aspirations and main occupations were translated into the eight-digit O*NET
occupation codes. This step links the reported career choices to the rich occupational information in the O*NET database, a national
database operated by the U.S. Bureau of Labor Statistics on all existing occupations in the U.S. economy. The 2000 Census codes were
translated into the 2019 O*NET codes (Version 24.0) by the 1st author, by using the occupation descriptions to identify the best
matching occupation in O*NET database. The author referenced a crosswalk between the 2000 Census codes and the 2018 Standard
Occupational Classication system, a set of six-digit occupation codes which the 2019 O*NET codes were based on. Up to six O*NET
occupations were matched for each Census code, listed in the order of the best to the least match.
2.4.2. STEM classication
The STEM classication in our study is based on the importance of STEM-related domain-specic knowledge for an occupation. The
importance of STEM knowledge was operationalized with objective, statistically rigorous indices of the occupational reliance on nine
types of STEM knowledge in O*NET. One standardized score, ranging from 0 to 100, quanties how integral or trivial each kind of
knowledge is to an occupation in daily tasks. The nine STEM knowledge domains are 1) Biology, 2) Chemistry, 3) Computers and
Electronics, 4) Economics and Accounting, 5) Engineering and Technology, 6) Mathematics, 7) Mechanical, 8) Medicine and Dentistry,
and 9) Physics. This list was created by asking two raters to rate whether each of the 33 extant knowledge domains in O*NET was
important for STEM versus non-STEM occupations based on common sense. The two raters were counter-balanced on gender, eld of
study (i.e., STEM vs non-STEM) and whether their racial/ethnic group is underrepresented in STEM elds in the U.S. to reduce po-
tential biases in their ratings (see Appendix B for sociodemographic details). The nine knowledge domains above were consistently
labeled as STEM by two raters: the inter-rater reliability (Cohen’s Kappa) was .67, indicating substantial agreement between the two
raters (Warrens, 2015). We then inspected the absolute level and the relative ranking of the STEM knowledge scores and created cutoff
values for the two aspects. A job needs to have both a high “absolute importance”and a high “relative importance”of STEM knowledge
to be categorized as STEM. See Appendix B for detailed categorization criteria, our STEM classication list, and results of robustness
checks on our cutoff values.
2.4.2.1. LEHMS vs MEPCS. The highest ranked STEM knowledge domain determines whether a STEM job is LEHMS or MEPCS. We
used this criterion because it indicates the most important type of STEM knowledge for completing the job among all types of STEM
knowledge the job may require. MEPCS occupations are STEM jobs with the highest ranked STEM knowledge being 1) Computers and
Electronics, 2) Economics and Accounting, 3) Engineering and Technology, 4) Mathematics, 5) Mechanical, or 6) Physics. LEHMS
occupations are STEM jobs with the highest ranked STEM knowledge being 1) Biology or 2) Medicine and Dentistry. The same two
raters sorted the nine types of STEM knowledge into MEPCS or LEHMS (inter-rater reliability: Cohen’s Kappa =.73). The raters did not
agree on Chemistry; thus, for STEM occupations with the highest ranked STEM knowledge being Chemistry, the second highest ranked
STEM knowledge domain was referenced.
2.4.2.2. “Blue-collar”vs “white-collar”STEM. The classication differentiates STEM jobs based on whether the job requires a
Bachelor’s degree. The education requirement is indicated by the ve-point Job Zone index in O*NET database, from 1 “some of these
occupations may require a high school diploma or GED certicate”to 5 “Most of these occupations require graduate school”.“Blue-
collar”STEM jobs are STEM occupations that do not require a Bachelor’s degree (Job Zone 1 to 3), and “white-collar”STEM jobs are
STEM occupations that require at least a Bachelor’s degree (Job Zone 4 and 5).
2.5. Analyses
We used four time points for plotting trajectories: age 13 (year 2002), age 18 (year 2005 or 2007), age 21 (year 2009/2011), and
age 25 (year 2013/2015). We aggregated two neighboring waves between 2005 and 2015 to minimize missing data due to the study
design of PSID; data from the earlier wave were used unless missing.
We counted the frequency of STEM career choices at each time point cross-sectionally and the frequency of longitudinal patterns of
career choices. We then carried out these analyses for men and women separately. All existing paths in our data were identied.
We plotted the trajectories of career choices in Sankey trees (Sankey Tree, 2021). Sankey trees are a hybrid of Sankey diagrams and
tree graphs, which display the distribution of cases over the conditioned combinations of scenarios via weighted branches. For lon-
gitudinal patterns, the graphs show the temporal, rather than causal, sequence of career choices, with a later career choice conditioned
on an earlier one.
We then aggregated individuals with particular types of trajectories for further analyses based on our research questions. Spe-
cically, individuals with an aspiration in a STEM eld at age 13 but an occupation in a different eld at age 25 were the “STEM
attrition”subgroup, regardless of the types of their career choices at age 18 and 21; individuals with an employment in a STEM eld at
age 25 but an aspiration in a different eld at age 13 were the “STEM entry”subgroup. Similarly, individuals with an aspiration and an
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
5
employment in the same STEM eld were the “STEM persistence”subgroup; individuals with an aspiration and an employment in the
non-STEM eld were the “non-STEM persistence”subgroup. “STEM career attainment trajectories”were trajectories ending with an
employment in a STEM eld at age 25, irrespective of the patterns of prior aspirations. In other words, the “STEM career attainment
trajectories”consist entirely of STEM workers.
The Chi-square Test of Independence was used to examine the associations between gender and career paths, and the Fisher’s Exact
Test was used as needed to accommodate small cell sizes (Hess &Hess, 2017). Adjusted standardized residuals were used as post-hoc
analyses to identify the over- or under-representation of a gender group in a particular type of career path.
2.5.1. Missing data and analyses
The focus of our study is the development of STEM career aspirations in adolescence, which was only assessed for participants over
age 12. Younger teenagers and children tend to report superpower, fantastic career goals (e.g., superman) and may lack the cognitive
maturity to reect on and utilize abstract concepts (e.g., future goals. Auger et al., 2005;Nelson, 2017). Thus, our target sample
consisted of 1,784 eligible respondents after 1,125 respondents were excluded due to this age-based inclusion criteria. Out of the target
sample, respondents were included in the analyses if they participated in data collection at all four time points (752 respondents
excluded). This inclusion criterion was to ensure that STEM attrition and entry in our analysis reect changes in career choices, rather
than changes in study participation. An addition of 218 respondents were excluded due to other reasons (e.g., question not applicable,
see Appendix A for details). The nal analysis sample included 814 participants, 46 % of the target sample.
Missing data analyses were carried out to investigate potential differences in key variables between the analysis sample and the
excluded sample. The analysis sample was older than the excluded sample due to the age-based inclusion criteria, t(2601.72) =
−41.23, p=.001, Cohen’sd=1.34, and consisted of slightly more females, Х
2
(1, N=2907) =16.03, p<.001, Cramer’sV=.07. The
two samples did not differ in URM status, Х
2
(1, N=2900) =0.18, p=.67, or parent education (in year of formal schooling completed),
t(2752) = − 1.66, p=.10. Regression analyses showed that controlling for the sociodemographic characteristics, being in the analysis
sample was associated with higher reading and math achievements and higher math expectancy (Table A2).
3. Results
Cross-sectional frequencies of STEM career choices are listed in Table 1, and gender compositions of different types of career
choices are listed in Table 2. Statistics of the whole sample are discussed in the text of supplementary materials. Results by gender
group are discussed below. Men’s and women’s STEM career trajectories are visualized with Sankey tree diagrams in Fig. 1 (by content
domain) and Fig. S1 (by educational requirement).
3.1. What are men’s and women’s STEM attrition rates, directions, and timing?
3.1.1. Attrition rates
Across STEM careers of any type, women and men did not differ in their attrition rates (women: 72.0 %, men: 63.0 %, Table 3),
χ
2
(1, N=484) =3.09, p=.07, Cramer’s V =.10. Across content domains, men left LEHMS careers at a higher rate than women did
(women: 78.9 %, men: 95.4 %),
χ
2
(1, N=185) =5.16, p=.02, Cramer’s V =.18, and women left MEPCS careers at a higher rate than
men did (women: 80.8 %, men: 65.6 %),
χ
2
(1, N=145) =4.97, p=.03, Cramer’s V =.23. Across educational requirements, men and
women switched out of blue-collar STEM careers at comparable rates (women: 70.9 %, men: 74.2 %),
χ
2
(1, N=111) =0.17, p=.68,
Cramer’s V =.04, and the same held true for white-collar STEM careers (women: 85.1 %, men: 79.1 %),
χ
2
(1, N=219) =1.36, p=.24,
Cramer’s V =.08. See Table S1 in supplementary materials for statistics on the whole sample.
Table 1
Cross-sectional frequency of STEM career choices.
Occupation Age 13 Age 18 Age 21 Age 25
Non-STEM 484 (59 %) 474 (58 %) 452 (75 %) 602 (74 %)
STEM 330 (41 %) 340 (42 %) 362 (25 %) 212 (26 %)
Total 814 (100%) 814 (100%) 814 (100 %) 814 (100 %)
Non-STEM 484 (59 %) 474 (58 %) 452 (56 %) 602 (74 %)
LEHMS 185 (23 %) 167 (21 %) 187 (23 %) 57 (7 %)
MEPCS 145 (18 %) 173 (21 %) 175 (21 %) 155 (19 %)
Total 814 (100 %) 814 (100 %) 814 (100 %) 814 (100 %)
Blue-collar non-STEM 272 (33 %) 202 (25 %) 187 (23 %) 483 (59 %)
Blue-collar STEM 111 (14 %) 162 (20 %) 190 (23 %) 131 (16 %)
White-collar non-STEM 212 (26 %) 272 (33 %) 265 (33 %) 119 (15 %)
White-collar STEM 219 (27 %) 178 (22 %) 172 (21 %) 81 (10 %)
Total 814 (100 %) 814 (100 %) 814 (100 %) 814 (100 %)
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
6
3.1.2. Directions of attrition
Among individuals leaving LEHMS careers, women and men moved toward different domains,
χ
2
(1, N=105) =5.38, p=.02,
Cramer’s V =.19: 90 % of women switched into non-STEM careers and the rest 10 % into MEPCS careers, whereas 76 % of men
switched into non-STEM jobs and 24 % into MEPCS jobs (Table 4,Fig. 2). Among individuals who left MEPCS elds, women and men
did not differ in their destinations, Fisher’s exact test, p=1.00, Cramer’s V =.06: over 90 % of women and men moved into non-STEM
jobs, and the rest switched into LEHMS jobs. Across educational requirements, the direction of blue-collar STEM attrition did not vary
by gender, Fisher’s exact test, p=.15, Cramer’s V =.22, with over 80 % of women and men switching into blue-collar non-STEM
occupations (Table 4, Fig. S2). Gender was not related to the direction of white-collar STEM attrition either,
χ
2
(2, N=180) =2.20, p=
.33, Cramer’s V =.11: 65 % of women and 64 % of men moved into blue-collar non-STEM jobs, 13 % of women and 20 % of men
switched into blue-collar STEM jobs, and the rest 22 % of women and 16 % of men turned to white-collar non-STEM jobs.
3.1.3. Timing of attrition
Two-way Chi-square Tests of Independence showed that men and women did not differ in the timing of their STEM attrition in any
domain (LEHMS:
χ
2
[3, N=153] =2.51, p=.27; MEPCS:
χ
2
[3, N=99] =1.71, p=.64) or at any educational level (blue-collar STEM:
χ
2
[3, N=78] =3.32, p=.35; white-collar STEM:
χ
2
[3, N =180] =0.43, p=.94) (Table 5). Across domains or educational re-
quirements, 50 to 60 % of STEM attrition occurred between age 13 and 18 (Table S2); this pattern was consistent between genders.
3.2. What are men’s and women’s STEM entry rates, directions, and timing?
3.2.1. Entry rates
AcrossSTEM careers of any type, STEM entry rates did not differ by gender (women: 49.5 %, men: 49.6 %),
χ
2
(1, N=212) =0.01, p
=.92, Cramer’s V <0.001. Across STEM content domains, men entered LEHMS careers at a higher rate than women did (women: 33.3
%, men: 83.3 %, Table 3),
χ
2
(1, N=57) =9.62, p=.002, Cramer’s V =.42, and women entered MEPCS careers at a higher rate than
men did (women: 89.6 %, men: 61.7 %),
χ
2
(1, N=155) =12.36, p<.001, Cramer’s V =.28. Across educational requirements, women
and men entered blue-collar STEM jobs at comparable rates (women: 77.9 %, men: 70.4 %),
χ
2
(1, N=131) =0.96, p=.33, Cramer’s
V=.09; the same held true for white-collar STEM careers (women: 56.4 %, men: 47.6 %),
χ
2
(1, N=81) =0.63, p=.43, Cramer’s V =
.09. The entry rates for the whole sample are listed in Table S1.
3.2.2. Directions of entry
Among individuals switching into LEHMS jobs, men and women differed in their aspirations, Fisher’s exact test p=.002, Cramer’s
V=.67: 93 % of women entrants had chosen non-STEM careers around age 13 and the rest 7 % had chosen MEPCS careers; in contrast,
30 % of men entrants had chosen non-STEM careers and the rest 70 % had chosen MEPCS careers (Table 6,Fig. 2). In MEPCS, men and
women entrants did not differ in their aspirations at age 13,
χ
2
(1, N =155) =1.82, p=.18, Cramer’s V =.13: 70 to 85 % of women and
men entrants entered from non-STEM elds, and the rest entered from MEPCS elds. No gender differences were found in the di-
rections of entry among blue-collar STEM entrants,
χ
2
(1, N =131) =1.94, p=.38, Cramer’s V =.14, or among white-collar STEM
entrants, Fisher’s exact test p=.29, Cramer’s V =.25 (Table 6, Fig. S2).
3.2.3. Timing of entry
Trajectories STEM entry in the whole sample are shown in Table S3. Two-way Chi-square Tests of Independence showed that men
and women differed in their timing of entry into LEHMS careers (Table 7), Fisher’s exact test p=.004, Cramer’s V =.68. Among
entrants into LEHMS occupations, more women than expected, based on the gender composition of all LEHMS entrants, switched into
the domain between age 18 and 21, whereas more men than expected made their entry between age 21 and 25. Among entrants into
MEPCS jobs, blue-collar STEM jobs or white-collar STEM jobs, the timing of their entry did not differ by gender (MEPCS:
χ
2
[3, N=
109] =1.42, p=.70, Cramer’s V =.11; blue-collar STEM:
χ
2
[3, N=98] =4.07, p=.25, Cramer’s V =.20; white-collar STEM: Fisher’s
Table 2
Gender composition of STEM career choices around age 13 and 25.
Field Age 13 Age 25
Frequency % of Women Frequency % of Women
Non-STEM 484 58 %
+
602 59 %
+
STEM 330 51 %
−
212 44 %
−
Non-STEM 484 58 %
+
602 59 %
+
LEHMS 185 77 %
+
57 79 %
+
MEPCS 145 18 %
−
155 31 %
−
Blue-collar non-STEM 272 47 %
−
483 57 %
Blue-collar STEM 111 49 %
−
131 41 %
−
White-collar non-STEM 212 73 %
+
119 70 %
+
White-collar STEM 219 52 % 81 48 %
Total 814 55 % 814 55 %
Note.
+
/
−
indicates over-/under-representation of women with standardized adjusted residuals greater than 1.96 or less than −1.96.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
7
Exact Test: p=.11, Cramer’s V =.37).
3.3. In what ways and to what extent does the size of potential STEM workforce change among men and women as they grew from age 13 to
25?
Changes in the quantity of the potential STEM workforce were indicated by a “change ratio”, which was the number of STEM
workers around age 25 divided by the number of STEM aspirants around age 13 in a given STEM eld. Because there was no addition or
attrition of participants in the sample, the change ratio reects the shrinkage/expansion of a workforce due to the ow of individuals
Fig. 1. Girls’and boys’career trajectories across LEHMS and MEPCS domains.
Note. The Sankey tree graphs are structured horizontally in temporal order in form of nodes connected via branches. The label of a node indicates the
type of career choices, and a branch represents a group of individuals with the same career choices at two successive time points. To the right of the
starting node labeled by the gender group (i.e., “girls”) do nodes align in four columns. Each column corresponds to one time point (i.e., age 13, 18,
21 and 25). The weight of branches indicates the relative size of the subgroup of individuals.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
8
across different types of careers. A change ratio greater than 100 % indicates an expansion of the workforce, and a ratio lower than 100
% indicates a shrinkage. For gender comparisons, change ratios were calculated for men and women separately. For example, 168 girls
aspired to STEM careers of any sort around age 13 and 93 women worked in STEM jobs around age 25. The change ratio of the women’s
STEM workforce was 55.4 % (93 divided by 168, Table 3). Among men, there were 162 STEM aspirants around age 13 and 119 STEM
workers around age 25, yielding a change ratio of 73.5 % (Table 3).
Fig. 1. (continued).
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
9
Table 3
Attrition rates, entry rates and change ratio of potential workforce by eld and gender.
Field Aspirations around age 13 Employment around age 25 Change ratio (%)
Total Attrition Persistence Attrition rate (%) Total Entry Persistence Entry rate (%)
W M W M W M W M W M W M W M W M W M
Non-STEM 282 202 46 59 236 143 16.3
a
29.2
b
357 245 121 102 236 143 33.9 41.6 126.6 121.3
STEM 168 162 121 102 47 60 72.0 63.0 93 119 46 59 47 60 49.5 49.6 55.4 73.5
Non-STEM 282 202 46 59 236 143 16.3
a
29.2
b
357 245 121 102 236 143 33.9 41.6 126.6 121.3
LEHMS 142 43 112 41 30 2 78.9
a
95.4
b
45 12 15 10 30 2 33.3
a
83.3
b
31.7 27.9
MEPCS 26 119 21 78 5 41 80.8
a
65.6
b
48 107 43 66 5 41 89.6
a
61.7
b
184.6 89.9
Blue-collar non-STEM 128 144 39 49 89 95 30.5 34.0 274 209 185 114 89 95 67.5
a
54.6
b
214.1 145.1
Blue-collar STEM 54 57 38 40 16 17 70.9 74.2 54 77 38 60 16 17 70.4 77.9 100.0 135.1
White-collar non-STEM 154 58 114 48 40 10 74.0 82.8 83 36 43 26 40 10 51.8
a
72.2
b
53.9 62.1
White-collar STEM 114 105 97 83 17 22 85.1 79.1 39 42 22 20 17 22 56.4 47.6 34.2 40.0
Note. W=women, M =men.
a
/
b
indicates gender differences in attrition/entry rates in the same eld (
α
=0.05). Attrition rate equals the total number of attrition divided by the total number of aspirants.
Entry rate equals the total number of entrants divided by the total number of workers. Change ratio equals the total number of workers divided by the total number of aspirants.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
10
Table 4
Gender comparisons in the directions of attrition.
Employment around age 25 Aspiration around age 13
Non-STEM LEHMS MEPCS
Everyone Women Men Everyone Women Men Everyone Women Men
Non-STEM – – – – – – 132 86 % 101 90 %
+
31 76 %
−
91 92 % 20 95 % 71 91 %
LEHMS 17 16 % 14 26 % 3 5 % – – – – – – 8 8 % 1 5 % 7 9 %
MEPCS 88 84 % 32 74 % 56 95 % 21 14 % 11 10 %
−
10 24 %
+
– – – – – –
Total attrition 105 100 % 46 100 % 59 100 % 153 100 % 112 100 % 41 100 % 99 100 % 21 100 % 78 100 %
Gender differences
χ
2
(1, N =105) =12.24, p<.001, Cramer’s V =.34
χ
2
(1, N =153) =5.38, p=.02, Cramer’s V =.19 Fisher’s exact test: p=1.00, Cramer’s V =.06
Empt. at age 25 Aspiration around age 13
BCNS BCS WCNS WCS
Everyone Women Men Everyone Women Men Everyone Women Men Everyone Women Men
BCNS – – – – – – 65 83 % 31 81 % 34 85 % 118 73 % 91 80 % 27 56 % 116 64 % 63 65 % 53 64 %
BCS 42 47 % 13 33 % 29 59 % – – – – – – 26 16 % 12 10 % 14 29 % 30 17 % 13 13 % 17 20 %
WCNS 27 31 % 16 41 % 11 23 % 8 10 % 6 16 % 2 5 % – – – – – – 34 19 % 21 22 % 13 16 %
WCS 19 22 % 10 26 % 9 18 % 5 7 % 1 3 % 4 10 % 18 11 % 11 10 % 7 15 % – – – – – –
Total attri. 88 100 % 39 100 % 49 100 % 78 100 % 38 100 % 40 100 % 162 100 % 114 100 % 48 100 % 180 100 % 97 100 % 83 100 %
Gender Diff.
χ
2
(2, N=88) =10.62, p=.005, Cramer’s V =.36 Fisher’s exact test: p=.15, Cramer’s V =.22
χ
2
(2, N=162) =10.63, p=.005, Cramer’s V =.26
χ
2
(2, N =180) =2.20, p=.33, Cramer’s V =.11
Note. Empt. =employment. BCNS =blue-collar non-STEM. BCS =blue-collar STEM. WCNS =white-collar non-STEM. WCS =white-collar STEM. Attri. =attrition. Gender diff. =gender differences. +/−
indicates over−/under-representation of women with standardized adjusted residuals greater than 1.96 or less than −1.96. ▴Fisher’s exact test (two-tailed) was used instead of the Chi-square Test of
Independence due to small cell sizes.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
11
Fig. 2. Transformation of STEM workforce across knowledge domains by gender.
Note. One branch from left to right represents one type of career path between age 13 and 25. The weight of branches and nodes represent the
relative size of a subgroup.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
12
Across STEM domains, the LEHMS workforce shrank to less than a third of its original size among both women (31.7 %) and men
(27.9 %, Table 3,Fig. 2). In MEPCS careers, women’s MEPCS workforce expanded to 184.6 % of its original size, whereas the men’s
MEPCS workforce shrank to 89.9 % of its original size. Across educational requirements, the potential workforce of blue-collar STEM
maintained its size among women but expanded among men (women 100.0 %, men: 135.1 %); the workforce of white-collar STEM
careers shrank considerably among both men and women (women: 34.2 %, men: 40.0 %, Fig. S2).
3.4. What are men’s and women’s STEM career attainment trajectories?
STEM career attainment trajectories were analyzed based on the number of prior choices in the same eld of employment around
age 13, 18 and 21 (Table S4) and examined in terms of gender differences (Table 8,Fig. 3). In other words, we investigated the in-
cidences of obtaining a STEM occupation with 0, 1, 2 or 3 prior aspirations in the same eld at age 13, 18 and 21 among men and
women. Men and women STEM workers took different trajectories to their LEHMS jobs, Fisher’s exact test p<.001, Cramer’s V =.63
(Table 8); specically, more men than expected worked in their job without LEHMS aspirations around age 13, 18 and 21, and more
women than expected aspired to LEHMS careers around age 13 and persisted around age 18 and 21 before their employment around
age 25. In particular, the pipeline-like trajectory, one with an early LEHMS aspiration around age 13 and persistent LEHMS aspirations
at the following three time points, consisted entirely of women. Gender differences were also found among MEPCS career attainment
trajectories,
χ
2
=8.71, p=.03, Cramer’s V =.23: among MEPCS STEM workers, more women and fewer men than expected worked in
their job without prior aspirations to the eld around age 13, 18 or 21.
No gender difference was found in attainment trajectories of blue-collar STEM careers,
χ
2
=6.53, p=.09, Cramer’s V =.22, or
white-collar STEM careers,
χ
2
=6.27, p=.10, Cramer’s V =.28 (Table 8). Although no main effect of gender was found, standardized
adjusted residuals indicated that among blue-collar STEM workers, more women and fewer men than expected took the “early entry
and persisting”path; among white-collar STEM workers, fewer women and more men than expected aspired to white-collar STEM
careers twice prior to their employment.
4. Discussion
4.1. LEHMS and white-collar STEM workforces shrank, but not others
We found that the STEM workforce shrank in LEHMS and “white-collar”STEM careers but maintained its size in MEPCS careers and
expanded in blue-collar STEM careers. Particularly, the LEHMS and the white-collar STEM workforce shrank to less than a third of their
original sizes. The desirability of these careers to adolescents might have contributed to the sizable shrinkage. From an early age,
children interact with LEHMS professionals in their daily lives; therefore, the tasks, setting, and impact of these jobs are particularly
familiar to children. In addition, many healthcare and “white-collar”STEM professions bear prestigious social status. Research sug-
gests that children could differentiate high-status occupations from low-status ones at a young age and preferred to interact with
individuals holding high-status jobs (Nesdale &Flesser, 2001). Adolescents’familiarity with healthcare professions and attraction to
prestigious white-collar STEM occupations might have contributed to the prevalence of career aspirations to these jobs in early
adolescence. These factors might have also led to the surplus of labor supply in the biomedical eld (Xue &Larson, 2015) as well as the
discrepancy between the career aspirations and the actual employment in the LEHMS and the white-collar STEM occupations in our
result.
Table 5
Gender differences in the timing of attrition.
Attrition trajectories Aspiration around age 13 (percentage of women in parentheses)
Non-STEM STEM LEHMS MEPCS BCNS BCS WCNS WCS
Leave by age 18: y-n-n-n 30 (47 %) 93 (56 %) 86 (69 %) 49 (24 %) 54 (54 %
+
) 47 (40 %) 73 (63 %) 106 (56 %)
Leave by age 21: y-y-n-n 29 (55 %) 37 (49 %) 18 (78 %) 21 (24 %) 17 (35 %) 8 (63 %) 27 (70 %) 32 (50 %)
Leave by age 25: y-y-y-n 37 (32 %) 72 (54 %) 35 (77 %) 19 (11 %) 10 (10 %
−
) 16 (63 %) 45 (80 %) 26 (54 %)
Recursive: y-n-y-n 9 (44 %) 21 (57 %) 14 (86 %) 10 (20 %) 7 (43 %) 7 (57 %) 17 (76 %) 16 (50 %)
Total attrition 105 (44 %) 223 (54 %) 153 (73 %) 99 (21 %) 88 (44 %) 78 (49 %) 162 (70 %) 180 (54 %)
Test statistics Chi-square Test of Association Between Gender and Trajectory Group Membership in Each Field
χ
2
3.57 0.64 2.51 – – – 4.20 0.43
df 3 3 3 – – – 3 3
p.31 .89 .47 .66
▴
.06
▴
.89
▴
.24 .94
Cramer’s V .18 .05 .13 .13 .29 .21 .16 .05
Note. BCNS =blue-collar non-STEM. BCS =blue-collar STEM. WCNS =white-collar non-STEM. WCS =white-collar STEM. Each letter in career path
represents one time point (age 13 –age 18 –age 21 –age 25). Y =choice in the eld of each column, n =choice in other elds. +/−indicates over−/
under-representation of women with standardized adjusted residuals greater than 1.96 or less than −1.96. ▴Fisher’s exact test (two-tailed) was used
instead of the Chi-square Test of Independence to accommodate for small cell sizes.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
13
Table 6
Gender comparisons in the directions of entry.
Aspiration around age 13 Employment around age 25
Non-STEM LEHMS MEPCS
Everyone Women Men Everyone Women Men Everyone Women Men
Non-STEM – – – – – – 17 68 % 14 93 % 3 30 % 88 81 % 32 74 % 56 85 %
LEHMS 132 59 % 101 83 % 31 30 % – – – – – – 21 19 % 11 26 % 10 15 %
MEPCS 91 41 % 20 17 % 71 70 % 8 32 % 1 7 % 7 70 % – – – – – –
Total entry 223 100 % 121 100 % 102 100 % 25 100 % 15 100 % 10 100 % 109 100 % 43 100 % 66 100 %
Gender differences
χ
2
(1, N=223) =64.55, p<.001, Cramer’s V =.54 Fisher’s Exact Test: p=.002, Cramer’s V =.67
χ
2
(1, N =109) =1.82, p=.18, Cramer’s V =.13
Asp. at age 13 Employment around age 25
BCNS BCS WCNS WCS
Everyone Women Men Everyone Women Men Everyone Women Men Everyone Women Men
BCNS – – – – – – 42 43 % 13 34 % 29 48 % 27 39 % 16 37 % 11 42 % 19 45 % 10 45 % 9 45 %
BCS 65 22 % 31 17 % 34 30 % ––––––8 12 % 6 14 % 2 8 % 5 12 % 1 5 % 4 20 %
WCNS 118 39 % 91 49 % 27 24 % 26 26 % 12 32 % 14 24 % ––––––18 43 % 11 50 % 7 35 %
WCS 116 39 % 63 34 % 53 46 % 30 31 % 13 34 % 17 28 % 34 49 % 21 49 % 13 50 % ––––––
Total entry 299 100 % 185 100 % 114 100 % 98 100 % 38 100 % 60 100 % 69 100 % 43 100 % 26 100 % 42 100 % 22 100 % 20 100 %
Gender Diff.
χ
2
(3, N=299) =19.98, p<.001, Cramer’s V =.26
χ
2
(3, N =98) =1.94, p=.38, Cramer’s V =.14
χ
2
(3, N=69) =0.66, p=.72, Cramer’s V =.10 Fisher’s Exact Test: p=.29, Cramer’s V =.25
Note. Asp. =aspiration. BCNS =blue-collar non-STEM. BCS =blue-collar STEM. WCNS =white-collar non-STEM. WCS =white-collar STEM. Gender diff. =gender differences. +/−indicates over−/
under-representation of women with standardized adjusted residuals greater than 1.96 or less than −1.96. ▴Fisher’s exact test (two-tailed) was used instead of the Chi-square Test of Independence to
accommodate for small cell sizes.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
14
4.2. STEM workforces did not shrink more among women
Although previous studies showed that the STEM attrition rates were higher among women than among men (Bieri Buschor et al.,
2014;Weeden et al., 2020), we did not nd that STEM workforces shrank more among women than men. Our results show that among
the LEHMS and the white-collar STEM careers, there were comparable shrinkages in both gender groups, and that the women blue-
collar STEM workforce maintained its size. Particularly, the woman MEPCS workforce nearly doubled its size. Although the group
size was small, more women entered the MEPCS careers than women leaving. Our study is the rst to report this nding to the best of
our knowledge. Considering the inuence of gender stereotypes on women’s physics career motivation (Starr &Leaper, 2022), our
nding highlights the resilience of the women entrants. Our results also highlight distinct dynamics of STEM attrition and entry among
different STEM occupations, which we discuss in depth below.
4.3. Entry impacts STEM workforce no less than attrition does
Compared with previous studies on STEM entry (Xie &Shauman, 2003, Chapter 4), our study extended beyond STEM major choices
in postsecondary education to include earlier STEM aspirations and actual occupation. We found that 47 to 90 % of STEM workers had
not aspired to their STEM eld of employment at age 13 but switched to their STEM eld at a later time point (except women LEHMS
workers). This nding indicates that STEM elds were characterized by frequent intake of workers from other elds. Integrating STEM
attrition and entry, our analysis highlight the bidirectional dynamics of STEM workforce. For instance, the LEHMS and the white-collar
STEM workforces featured frequent attrition and infrequent entry, leading to sizable shrinakges; in comparison, the women blue-collar
STEM and men MEPCS workforces featured frequent entry and frequent attrition, resulting in relative stability in their sizes. Theo-
retically, it might indicate that early STEM aspirations may function as a protective factor, rather than a risk factor, for individuals to
attain a STEM occupation (more in “More Evidence is needed for the ‘Cumulative Disadvantage’Assumption of the ‘Pipeline’
Table 7
Gender differences in the timing of entry.
Entry trajectory Employment around age 25 (percentage of women in parentheses)
Non-STEM STEM LEHMS MEPCS BCNS BCS WCNS WCS
Enter after age 13: n-y-y-y 93 (56 %) 30 (47 %) 10 (60 %) 18 (33 %) 38 (47 %
−
) 17 (59 %) 34 (76 %
+
) 14 (29 %
−
)
Enter after age 18: n-n-y-y 37 (49 %) 29 (55 %) 8 (100 %
+
) 28 (36 %) 46 (59 %) 20 (40 %) 10 (60 %) 10 (70 %)
Enter after age 21: n-n-n-y 72 (54 %) 37 (32 %) 6 (17 %
−
) 56 (45 %) 181 (65 %) 51 (31 %) 18 (56 %) 17 (59 %)
Recursive: n-y-n-y 21 (57 %) 9 (44 %) 1 (0 %) 7 (29 %) 34 (68 %) 10 (40 %) 7 (14 %
−
) 1 (100 %)
Total entry 223 (54 %) 105 (44 %) 25 (60 %) 109 (39 %) 299 (62 %) 98 (39 %) 69 (62 %) 42 (52 %)
Test statistics Chi-square Test of Association Between Gender and Trajectory Group Membership in Each Field
χ
2
0.64 3.57 –1.42 4.65 4.07 – –
df 3 3 –3 3 3 – –
p.89 .31 .004
▴
.70 .20 .25 .02
▴
.11
▴
Cramer’s V .05 .18 .68 .11 .13 .20 .38 .37
Note. Each letter in career path represents one time point (age 13 –age 18 –age 21 –age 25). Y =choice in the eld of each column, n =choice in
other elds. +/−indicates over−/under-representation of women with standardized adjusted residuals greater than 1.96 or less than −1.96.
▴
Fisher’s exact test (two-tailed) was used instead of the Chi-square Test of Independence to accommodate for small cell sizes.
Table 8
Gender differences in the types of STEM career attainment trajectories.
Number of prior choices in the eld Employment around age 25 (Total frequency in cell, percentage of women in parentheses)
Non-STEM STEM LEHMS MEPCS BCNS BCS WCNS WCS
0 72 (54 %) 37 (32 %) 6 (17 %
−
) 56 (45 %
+
) 181 (65 %
+
) 51 (31 %) 18 (56 %) 17 (59 %)
1 129 (54 %) 53 (45 %) 10 (90 %) 45 (27 %) 165 (57 %) 36 (39 %) 23 (43 %
−
) 20 (50 %)
2 201 (60 %) 62 (39 %) 17 (65 %) 34 (21 %) 99 (46 %
−
) 27 (48 %) 42 (79 %) 29 (31 %
−
)
3 200 (64 %) 60 (55 %
+
) 24 (100 %
+
) 20 (20 %) 38 (45 %) 17 (65 %
+
) 36 (83 %
+
) 15 (67 %)
Total 602 (59 %) 212 (44 %) 57 (79 %) 155 (31 %) 483 (57 %) 131 (41 %) 119 (70 %) 81 (48 %)
Test statistics Chi-square Test of Association Between Gender and Trajectory Group Membership in Each Field
χ
2
3.99 5.70 –8.71 11.09 6.53 13.94 6.27
df 3
p.26 .13 <.001
▴
.03 .01 .09 .003 .10
Cramer’s V .08 .16 .63 .23 .15 .22 .34 .28
Note. Each letter in career path represents one time point (Age 13 –Age 18 –Age 21 –Age 25), with the last letter always being “y”to indicate an
employment in the eld around age 25. Y =choice in the eld of employment, n =choice in other elds. BCNS =blue-collar non-STEM. BCS =blue-
collar STEM. WCNS =white-collar non-STEM. WCS =white-collar STEM. +/−indicates over−/under-representation of women with standardized
adjusted residuals greater than 1.96 or less than −1.96.
▴
Fisher’s exact test (two-tailed) was used instead of the Chi-square Test of Independence to
accommodate for small cell sizes.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
15
Metaphor”section). Practically, the result indicates that the absence of abundant STEM entrants could undercut the STEM workforce to
a great extent. We explain different barriers to STEM entry in detail below.
4.3.1. Fewer late entries into LEHMS than MEPCS careers
We extended Xie and Shauman (2003, Chapter 4) analyses on STEM entry by investigating the rates, directions, and trajectories of
entries into different STEM careers. We found that the LEHMS careers featured infrequent and early entry, whereas the MEPCS eld
featured frequent and relatively late entry. We speculate that the mandatory and sequential training programs of major healthcare
professions might be relevant. LEHMS occupations of high frequency in our data, such as physicians, nurses, and veterinarians, require
a specic series of undergraduate and post-graduate education in the U.S.. Such requirements are demanding for aspirants’
commitment, competence and resources. Switching into the eld halfway challenges the aspirants to gather the necessary preparations
promptly. For instance, our sample consisted of a high portion of adolescents from low-income families relative to the national
average, and an interview study with working-class aspiring medical students revealed multifaceted barriers and constraints these
adolescents face to enter the medical and healthcare eld (Packard &Babineau, 2009). Some students encountered pressure to provide
a stable income for their families as participants were typically older at the time of their later entry into STEM; others struggled to
complete prerequisite classes and apprenticeships. Family obligations, including caregiving duties, are another burden for adult
students. Thus, the costly and extensive training of healthcare professionals can be extremely difcult to fulll under the complex life
circumstances of the less privileged individuals. Our study started to identify STEM entrants in various STEM domains, and more
research is needed to investigate the developmental context of these subgroups.
4.3.2. Resilient, counter-stereotypical entries curbing gender aggregation in LEHMS and MEPCS STEM
Our ndings show that the counter-stereotypical entries, namely men’s entry into LEHMS careers and women’s entry into MEPCS
careers, reshaped the gender composition of the STEM workforces in these domains. Remarkably, more women entered MEPCS careers
than leaving, increasing women’s share in the MEPCS workforce from 18 % to 31 %. To our knowledge, our study is the rst to reveal
this critical dynamic of the MEPCS workforce. Relatedly, men’s entry into the LEHMS careers hindered the further intensication of the
gender disparity in LEHMS. These counter-stereotypical entries are worth noticing given the inuence of existing gender aggregations
in these elds. In our study, gender aggregation in STEM domains had formed by age 13 such that the majority of LEHMS aspirants
were girls, and the majority of MEPCS aspirants were boys. Gendered patterns of the attrition rates only exacerbated such aggregation.
These patterns in aspirations and attrition rates may partially results from traditional gender role socialization, which gives rise to
girls’and boys’differential preferences for LEHMS and MEPCS careers (Eccles &Wang, 2016;Su et al., 2009), as well as stereotypes
0
10
20
30
40
50
60
Total Women Men
MEPCS
0123
0
10
20
30
40
50
60
Total Women Men
LEHMS
0123
0
10
20
30
40
50
60
Total Women Men
Blue-collar STEM
0123
0
10
20
30
40
50
60
Total Women Men
White-collar STEM
0123
Fig. 3. Gender comparisons of STEM career attainment trajectories by number of prior choices in the eld of employment.
Note. Only workers in the given STEM eld were included in analysis. Three prior career choices are those at age 13, 18 and 21.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
16
differentially associating boys/men with MEPCS (Bond, 2016;Cvencek et al., 2011). We speculate that these men and women entrants
might have taken prolonged career exploration before afrming their counter-stereotypic career choices in LEHMS and MEPCS do-
mains respectively. Our results highlight these entry as exceptions in the development of STEM career motivation in both gender
groups. Considering the signicance of these counter-stereotypical entries for diversifying the STEM workforce, we call for more
research to gain an in-depth understanding of these career paths.
4.3.3. Education-related barriers possibly restricting entries into white-collar STEM careers
Entries into the blue-collar and the white-collar STEM careers possibly reect education-related barriers. We found that blue-collar
STEM careers featured frequent, early entries compared with white-collar careers. Moreover, downward shifts from white-collar STEM
aspirations to blue-collar STEM occupations were more common than shifts in the opposite direction. A previous study on the
development of college-associated career aspirations showed similar patterns (Gao &Eccles, 2020). Thus, white-collar STEM aspirants
face a two-fold challenge: one is to master STEM-specic knowledge and skills, and the other is to obtain a Bachelor’s degree.
Therefore, supporting individuals to enter and graduate from a four-year university and develop a STEM skillset might be two sides of
the same coin for expanding the white-collar STEM workforce.
4.4. STEM attrition was frequent, complete, early, and downward
We provided a multifaceted understanding of STEM attrition by examining not only the rates but also the direction and timing of
the attrition. Consistent with previous national studies (Chen &Soldner, 2013;Ma &Liu, 2017;Xie &Shauman, 2003, Chapter 4), our
analyses showed high STEM attrition rates, ranging from 65 to 95 % across gender and types of STEM careers. The generous income,
reputable social status, and opportunities for innovations and inventions of STEM occupations might have attracted many aspirants
initially (Nesdale &Flesser, 2001;Rothwell, 2013).
Furthermore, we highlight three novel ndings. First, we found that the vast majority of people who left LEHMS or MEPCS careers
switched into non-STEM jobs, instead of a job in a different STEM domain. It suggests that there might be distinct job requirements (e.
g., domain-specic knowledge) or personal preferences for LEHMS and MEPCS occupations that limit the exchanges of workers be-
tween the two domains. Second, we found that half or more STEM aspirants who left their eld changed their path by age 18. In other
words, those who left their STEM eld after age 18 constituted less than half of the attrition population, with STEM attrition in college
majors among the college-attending subgroup covering even less of the total attrition. Thus, more research and practice efforts to
reduce STEM attrition at an earlier stage, such as high school or middle school, can not only address a larger share of the attrition but
also reach a more diverse population by including individuals who do not attend college. Third, we found that blue-collar non-STEM
jobs were the most common destination of attrition from white-collar STEM aspirations. In our sample, over 60 % of white-collar STEM
aspirants who left the eld worked in job of different content knowledge and at a lower education level, experiencing a sharp contrast
between their career aspiration and attainment (the “aspiration-attainment gap”[Carr, 1997]). Evidence suggested that such gaps
often signaled a lack of information and support for career planning and preparation due to family disadvantages and predicted lower
job satisfaction and well-being in adulthood (Hardie, 2014;Perry et al., 2016). Follow-up studies are needed to reveal the develop-
mental antecedents and the consequences of such aspiration-attainment gaps in STEM elds.
4.4.1. Gender similarities surpassing gender differences in STEM attrition
Our investigations into variable aspects of STEM attrition highlight gender similarities. We found men and women had similar
attrition rates, destinations and timing when leaving blue-collar and white-collar STEM careers. Existing gender differences were
limited to small to moderate discrepancies in the rates of the LEHMS attrition and the MEPCS attrition and in the directions of LEHMS
attrition. These results support the context-dependence of gender differences in Hyde’s gender similarity hypothesis (Hyde, 2005). By
clarifying where gender differences in STEM attrition existed, our study helps channel future research and policy efforts to address the
gender gaps.
We found that in the LEHMS domain, the attrition rate was higher among men than among women, and the reverse was true in the
MEPCS domain. Relatedly, among individuals leaving LEHMS careers, more men than expected changed into MEPCS jobs whereas
more women than expected changed into non-STEM jobs. These trends perpetuate or strengthen the existing gender aggregations in
LEHMS and MEPCS elds. Previous research indicates that cultural values of different STEM domains (e.g., interpersonal pursuits in
life science domains versus self-directed pursuits in engineering domains [Diekman et al., 2017;Eccles &Wang, 2016]), the affordance
for work-life balance (Bieri Buschor et al., 2014;Frome et al., 2006), and existing gender aggregations in and stereotypes about
different STEM careers (Starr &Simpkins, 2021) might have contributed to such patterns. We note that existing gender differences in
attrition were of medium effect sizes, indicating that gender and STEM attrition were only moderately related. By specifying the
aspects and extent of the gender effects on STEM attrition, we hope our ndings can contribute to a more accurate understanding of the
specic differences and similarities in the patterns of men’s and women’s STEM career choices.
4.5. More evidence is needed for the “cumulative disadvantage”assumption of the “pipeline”metaphor
Early STEM experience is often discussed as a “capital”within the “leaky pipeline”metaphor such that teenagers lacking early
STEM engagement are disadvantaged in their STEM career pursuits. This disadvantage cumulates, resulting in weaker STEM per-
formance and career motivation to further pursue and complete STEM education and training. It is assumed that this disadvantage in
STEM capital undermines STEM career attainment in adulthood. Based on this assumption, children should identify their STEM career
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
17
goals as early as possible and engage with STEM elds continuously to maximize their chances of attaining a STEM job. Longitudinal
studies support that earlier STEM experiences predicted increased STEM performance and career choices (e.g., Sadler et al., 2014;
Tyson et al., 2007), but it is unclear how strongly such connection between earlier and later participation holds. In our study, we found
that it is common for people to attain a STEM job without persistent prior STEM choices. In particular, we showed a “no leak pipeline”
was not a predominant STEM career attainment path (except among women LEHMS occupations): less than 20 % of workers in MEPCS,
blue-collar STEM, and white-collar STEM followed the "no leak pipeline". Our study supports frequent changes in career choices across
time, and more research is needed to examine whether earlier STEM experience functions as a necessary condition (that is, an individual
cannot attain a STEM job without early STEM experience), a risk factor (i.e., an individual is less likely to attain a STEM job without
early STEM experience) or a protective factor (i.e., an individual will benet from the advantage of having early STEM experience but
can still attain a STEM job without such experience) for attaining a STEM job.
In terms of gender differences, discussions based on the “leaky pipeline”metaphor attributed women’s underrepresentation in
STEM occupations, especially in the MEPCS domain, to the lack of early STEM career choices and engagement in the eld (e.g., Sadler
et al., 2014;Tyson et al., 2007). Our ndings did not support this notion. We found that in the MEPCS domain, more women than
expected by chance attained their jobs without any prior aspirations to the eld. In other words, late entrants into MEPCS jobs between
age 21 and age 25 were particularly likely to be women. The fact that these women successfully worked in their MEPCS occupations
might indicate that that women’s MEPCS career choices did not develop like the accumulation of capital; instead, the career paths
might be driven by the overcoming of gender stereotypes or intermittent access to career opportunities (Ma, 2011). We hope this result
afrms girls’condence to carve out their unique career trajectories to MEPCS occupations, and more research is needed to understand
girls’, or any adolescents’STEM career paths that look different from a “pipeline”.
4.6. Limitations and future directions
In this study, we differentiated STEM careers by domain-specifc knowledge (LEHMS vs MEPCS) and by educational requirement (e.
g., blue-collar vs white-collar STEM). Due to the limit of our sample size, we lack the statistical power to investigate the intersection of
these two dimensions. Relatedly, we aggregated different kinds of non-STEM careers into one category, but the career trajectories to
the job of an attorney could be quite different from those of an athlete or a librarian. Therefore, future studies could include subtypes of
non-STEM careers to investigate the career trajectories in more detailed occupation categories. National longitudinal studies, such as
High School and Beyond, Add Health and Early Childhood Longitudinal Studies could offer a valuable subject pool to study these
questions by adding questions about career choices in longitudinal data collection.
We believe that STEM teachers in elementary and secondary schools are an integral part of the STEM workforce, because STEM
content knowledge is essential to their job and because they provide our children with their rst shared STEM experiences. Unfor-
tunately, elementary and secondary school teaching occupations were aggregated into a general category by school stage without
differentiations of the teaching subject (e.g., “Secondary School Teachers, Except Special and Career/Technical Education”) in 2000
Census occupation codes and 2019 O*NET database. Thus, we were unable to identify STEM teachers in elementary and secondary
schools in our analyses. This constraint could lead to an underestimation of women in the MEPCS workforce in our results, considering
that most of the elementary school and middle school STEM teachers were women (Nguyen &Redding, 2018). More broadly, we call
for an update in the Census occupation codes and the O*NET database to recognize elementary and secondary STEM teachers as STEM
workers.
CRediT authorship contribution statement
Yannan Gao: Writing –review &editing, Writing –original draft, Visualization, Software, Methodology, Investigation, Formal
analysis, Data curation, Conceptualization. Jacquelynne S. Eccles: Writing –review &editing, Visualization, Supervision, Resources,
Methodology, Funding acquisition, Formal analysis, Conceptualization. Anna-Lena Dicke: Writing –review &editing, Visualization,
Supervision, Resources, Funding acquisition, Formal analysis.
Funding information
This work was supported by the Institute of Education Sciences (grant number R305A170160, 2017)] awarded to Dr. Jacquelynnne
Eccles (PI) and Dr. Anna-Lena Dicke (Co-PI). The collection of data used in this study was partly supported by the National Institutes of
Health [grant numbers R01 HD069609 and R01 AG040213, 2011 and 2011], the Eunice Kennedy Shriver National Institute of Child
Health and Human Development (grant number R01 HD052646 [CDS] and grant number P01 HD087155 [TAS]), and the National
Science Foundation [award numbers SES 1157698, 2013] awarded to the PSID project team.
Declaration of competing interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to
inuence the work reported in this paper.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
18
Acknowledgements
We thank Drs. Judy Harackiewicz and Fani Lauermann for their feedback on an earlier draft of this manuscript.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jvb.2024.104067.
Data availability
The data is publicly available via https://simba.isr.umich.edu/data/data.aspx
References
Auger, R. W., Blackhurst, A. E., & Wahl, K. H. (2005). The development of elementary-aged children’s career aspirations and expectations. Professional School
Counseling, 8(4), 322–329.
Babarovi´
c, T. (2021). Development of STEM Vocational Interests During Elementary and Middle School: A Cohort-Sequential Longitudinal Study. Journal of Career
Development, 08948453211036986.https://doi.org/10.1177/08948453211036986
Berryman, S. E. (1983). Who Will Do Science? Trends, and Their Causes in Minority and Female Representation among Holders of Advanced Degrees in Science and
Mathematics. A Special Report.https://eric.ed.gov/?id=ED245052.
Bieri Buschor, C., Berweger, S., Keck Frei, A., & Kappler, C. (2014). Majoring in STEM—What accounts for women’s career decision making? A mixed methods study.
The Journal of Educational Research, 107(3), 167–176.
Blikstein, P., & Worsley, M. (2016). Children are not hackers: Building a culture of powerful ideas, deep learning, and equity in the maker movement. In Makeology
(pp. 64–79). Routledge.
Bond, B. J. (2016). Fairy godmothers >robots: The inuence of televised gender stereotypes and counter-stereotypes on girls’perceptions of STEM. Bulletin of Science,
Technology &Society, 36(2), 91–97. https://doi.org/10.1177/0270467616655951
Cannady, M. A., Greenwald, E., & Harris, K. N. (2014). Problematizing the STEM pipeline metaphor: Is the STEM pipeline metaphor serving our students and the
STEM workforce? Science Education, 98(3), 443–460.
Carnevale, A. P., & Strohl, J. (2010). How increasing college access is increasing inequality, and what to do about it. Georgetown University Center on Education and the
Workforce.
Carr, D. (1997). The fulllment of career dreams at midlife: Does it matter for women’s mental health? Journal of Health and Social Behavior, 38(4), 331–344. https://
doi.org/10.2307/2955429
Chen, X., & Soldner, M. (2013). STEM attrition: College students’paths into and out of STEM elds (NCES 2014–001). National Center for Education Statistics, Institute of
Education Sciences, US Department of Education.
Cimpian, J. R., Kim, T. H., & McDermott, Z. T. (2020). Understanding persistent gender gaps in STEM. Science, 368(6497), 1317–1319.
Clotfelter, C. T., Ladd, H. F., Muschkin, C. G., & Vigdor, J. L. (2013). Success in community college: Do institutions differ? Research in Higher Education, 54(7),
805–824.
Cvencek, D., Meltzoff, A. N., & Greenwald, A. G. (2011). Math–gender stereotypes in elementary school children. Child Development, 82(3), 766–779.
Department of Homeland Security. (2022). STEM designated degree program list. https://studyinthestates.dhs.gov/stem-opt-hub/additional-resources/eligible-cip-
codes-for-the-stem-opt-extension.
Diekman, A. B., Steinberg, M., Brown, E. R., Belanger, A. L., & Clark, E. K. (2017). A goal congruity model of role entry, engagement, and exit: Understanding
communal goal processes in STEM gender gaps. Personality and Social Psychology Review, 21(2), 142–175. https://doi.org/10.1177/1088868316642141
Eccles, J. (2009). Who am I and what am I going to do with my life? Personal and collective identities as motivators of action. Educational Psychologist, 44(2), 78–89.
Eccles, J. S., & Wang, M.-T. (2016). What motivates females and males to pursue careers in mathematics and science? International Journal of Behavioral Development,
40(2), 100–106.
Eccles, J. S., & Wigeld, A. (2020). From expectancy-value theory to situated expectancy-value theory: A developmental, social cognitive, and sociocultural
perspective on motivation. Contemporary Educational Psychology, 61, Article 101859. https://doi.org/10.1016/j.cedpsych.2020.101859
Fayer, S., Lacey, A., & Watson, A. (2017). BLS spotlight on statistics: STEM occupations-past, present, and future. U.S. Bureau of Labor Statistics.
Frome, P. M., Alfeld, C. J., Eccles, J. S., & Barber, B. L. (2006). Why don’t they want a male-dominated job? An investigation of young women who changed their
occupational aspirations. Educational Research and Evaluation, 12(4), 359–372.
Gao, Y., & Eccles, J. (2020). Who lower their aspirations? The development and protective factors of college-associated career aspirations in adolescence. Journal of
Vocational Behavior, 116, 103367. https://doi.org/10.1016/j.jvb.2019.103367
Gottfredson, L. S. (1981). Circumscription and compromise: A developmental theory of occupational aspirations. Journal of Counseling Psychology, 28(6), 545.
Hardie, J. H. (2014). The consequences of unrealized occupational goals in the transition to adulthood. Social Science Research, 48, 196–211.
Hess, A. S., & Hess, J. R. (2017). Understanding tests of the association of categorical variables: The Pearson chi-square test and Fisher’s exact test. Transfusion, 57(4),
877–879. https://doi.org/10.1111/trf.14057
Hirschi, A. (2011). Career-choice readiness in adolescence: Developmental trajectories and individual differences. Journal of Vocational Behavior, 79(2), 340–348.
https://doi.org/10.1016/j.jvb.2011.05.005
Hyde, J. S. (2005). The gender similarities hypothesis. American Psychologist, 60(6), 581.
Ma, Y. (2011). Gender differences in the paths leading to a STEM baccalaureate. Social Science Quarterly, 92(5), 1169–1190.
Ma, Y., & Liu, Y. (2017). Entry and degree attainment in STEM: The intersection of gender and race/ethnicity. Social Sciences, 6(3), 89.
Malanchuk, O., Messersmith, E. E., & Eccles, J. S. (2010). The ontogeny of career identities in adolescence. New Directions for Child and Adolescent Development, 2010
(130), 97–110. https://doi.org/10.1002/cd.284
Manly, C. A., Wells, R. S., & Kommers, S. (2018). The inuence of STEM denitions for research on women’s college attainment. International Journal of STEM
Education, 5(1), 45. https://doi.org/10.1186/s40594-018-0144-1
Messersmith, E. E., Garrett, J. L., Davis-Kean, P. E., Malanchuk, O., & Eccles, J. S. (2008). Career development from adolescence through emerging adulthood: Insights
from information technology occupations. Journal of Adolescent Research, 23(2), 206–227.
Metcalf, H. (2010). Stuck in the pipeline: A critical review of STEM workforce literature. InterActions: UCLA Journal of Education and Information Studies, 6(2).
National Center for Education Statistics. (2019). Digest of education statistics (2018–070). https://nces.ed.gov/programs/digest/2017menu_tables.asp.
National Science Board, National Science Foundation. (2021). The STEM Labor Force of Today: Scientists, Engineers, and Skilled Technical Workers (NSB-2021-2;
Science and Engineering Indicators 2022). https://ncses.nsf.gov/pubs/nsb20212.
National Science Foundation. (2023). Diversity and STEM: Women, minorities, and persons with disabilities (NSF 23-315). https://ncses.nsf.gov/pubs/nsf23315/.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
19
National Science Foundation. (2024, July). Fiscal year 2023 appropriations. https://new.nsf.gov/about/budget/fy2023/appropriations.
Nelson, K. (2017). Cognitive development and the acquisition of concepts. In Schooling and the acquisition of knowledge (pp. 215–239). Routledge.
Nesdale, D., & Flesser, D. (2001). Social identity and the development of children’s group attitudes. Child Development, 72(2), 506–517.
Nguyen, T. D., & Redding, C. (2018). Changes in the demographics, qualications, and turnover of American STEM teachers, 1988–2012. AERA Open, 4(3), Article
2332858418802790.
Packard, B. W.-L., & Babineau, M. E. (2009). From drafter to engineer, doctor to nurse: An examination of career compromise as renegotiated by working-class adults
over time. Journal of Career Development, 35(3), 207–227. https://doi.org/10.1177/0894845308327270
Perry, B. L., Martinez, E., Morris, E., Link, T. C., & Leukefeld, C. (2016). Misalignment of career and educational aspirations in middle school: Differences across race,
ethnicity, and socioeconomic status. Social Sciences, 5(3), 35.
Rothwell, J. (2013). The hidden STEM economy. Metropolitan Policy Program at Brookings.
Rumberger, R. W. (1983). Dropping out of high school: The inuence of race, sex, and family background. American Educational Research Journal, 20(2), 199–220.
Sadler, P. M., Sonnert, G., Hazari, Z., & Tai, R. (2014). The role of advanced high school coursework in increasing STEM career interest. Science Educator, 23(1), 1–13.
Sankey Tree. (2021). [JavaScript]. Displayr. https://github.com/Displayr/rhtmlSankeyTree (Original work published 2016).
Starr, C. R., & Leaper, C. (2022). Undergraduates’pSTEM identity and motivation in relation to gender-and race-based perceived representation, stereotyped beliefs,
and implicit associations. Group Processes &Intergroup Relations, 13684302221128235.
Starr, C. R., & Simpkins, S. D. (2021). High school students’math and science gender stereotypes: Relations with their STEM outcomes and socializers’stereotypes.
Social Psychology of Education.https://doi.org/10.1007/s11218-021-09611-4
Stephen, J., Fraser, E., & Marcia, J. E. (1992). Moratorium-achievement (Mama) cycles in lifespan identity development: Value orientations and reasoning system
correlates. Journal of Adolescence, 15(3), 283–300.
Su, R., & Rounds, J. (2015). All STEM elds are not created equal: People and things interests explain gender disparities across STEM elds. Frontiers in Psychology, 6.
https://doi.org/10.3389/fpsyg.2015.00189
Su, R., Rounds, J., & Armstrong, P. I. (2009). Men and things, women and people: A meta-analysis of sex differences in interests. Psychological Bulletin, 135(6),
859–884. https://doi.org/10.1037/a0017364
Tyson, W., Lee, R., Borman, K. M., & Hanson, M. A. (2007). Science, technology, engineering, and mathematics (STEM) pathways: High school science and math
coursework and postsecondary degree attainment. Journal of Education for Students Placed at Risk, 12(3), 243–270.
U.S. Department of Education. (2009). National SMART Grant Program –List of Eligible Majors for Award Year 2009-2010 (pp. 1–18) (GEN-09-09 https://fsapartners.ed.
gov/sites/default/les/attachments/dpcletters/GEN0909NationalSMARTGrantFieldslistofmajorscorrection.pdf.
Warrens, M. J. (2015). Five ways to look at Cohen’s kappa. Journal of Psychology and Psychotherapy, 5.
Weeden, K. A., Gelbgiser, D., & Morgan, S. L. (2020). Pipeline dreams: Occupational plans and gender differences in STEM major persistence and completion. Sociology
of Education, 93(4), 297–314.
Wegemer, C. M., & Eccles, J. S. (2019). Gendered STEM career choices: Altruistic values, beliefs, and identity. Journal of Vocational Behavior, 110, 28–42.
Xie, Y., Fang, M., & Shauman, K. (2015). STEM education. Annual Review of Sociology, 41, 331–357.
Xie, Y., & Shauman, K. A. (2003). Attainment of a Science/Engineering Baccalaureate. In Women in Science: Career Processes and Outcomes (pp. 72–96). Harvard
University Press.
Xue, Y., & Larson, R. C. (2015). STEM crisis or STEM surplus? Yes and yes. Monthly Labor Review, 2015.https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5800410/.
Y. Gao et al. Journal of Vocational Behavior 156 (2025) 104067
20