Gendered Pathways to Career Exploration and Academic Persistence Among STEM Undergraduates in South Korea

Abstract

Gender disparities in STEM education continue to shape students’ academic persistence and career development. Identifying how psychological and contextual factors operate differently for male and female students is essential for understanding these disparities and designing targeted interventions. This study examines gender differences in the structural pathways linking contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions among STEM undergraduates in South Korea. Using data from 2393 students collected through a national institutional project, multi-group structural equation modeling (SEM) was conducted to compare path coefficients between male and female students. The results showed that contextual supports significantly enhanced engineering self-efficacy, which in turn predicted higher major motivation and stronger academic persistence intentions across both groups. However, gender-specific differences emerged. Major motivation had a stronger positive effect on persistence among male students, whereas career exploration behaviors were negatively associated with persistence intentions only among female students. In addition, career barriers exerted a stronger negative total effect on persistence among female students. These findings suggest that gender differences in STEM are reflected not only in overall levels but also in the structural mechanisms linking key variables. Practically, supporting female students in managing perceived barriers and reframing career exploration as an adaptive process may help strengthen their academic persistence.

1. Introduction

Science, Technology, Engineering, and Mathematics (STEM) fields are central to national competitiveness and technological advancement, and securing a highly skilled STEM workforce is essential for innovation and future growth [1,2,3] (Blickenstaff,). Career interest formed during secondary and postsecondary education significantly predicts later STEM engagement and persistence [4,5]. However, STEM interest frequently declines over time, contributing to reduced participation and the well-documented “leaky pipeline” [6,7,8,9]. This attrition is particularly significant among women and fuels persistent global gender disparities: women represent less than 30% of the scientific workforce and roughly 35% of STEM students worldwide [3]. South Korea reflects these global patterns, with several characteristics that warrant empirical attention. Women constitute approximately 31% of STEM undergraduates and only 25.8% of engineering majors [10]. At university entry, only 17% of newly admitted female students choose STEM, compared with 45% of men [11], and women comprise only 23.1% of the national R&D workforce [10,12]. These patterns underscore the urgent need for support systems that address both women’s entry into STEM and their long-term retention.

Social Cognitive Career Theory (SCCT) provides a central theoretical framework for understanding STEM career development and persistence, emphasizing the roles of self-efficacy, outcome expectations, and contextual influences [13,14]. Prior research shows that sociocultural norms, family and peer support, and gendered stereotypes shape self-efficacy and career decision-making across cultural contexts [4,15,16,17]. Teachers’ gender stereotypes and biased performance evaluations further reinforce these patterns [15,18]. Contextual barriers, particularly gendered expectations and STEM stereotypes, have been shown to undermine self-efficacy and diminish students’ STEM interests and intentions [17]. Gender has been widely recognized as a key moderator in SCCT processes: women in STEM typically report lower self-efficacy [4,19,20], different outcome expectations [21,22], and stronger negative effects from contextual barriers [23,24]. These gendered patterns shape career intentions [25], and although core SCCT pathways remain stable, consistent mean-level gender differences are observed [26,27,28]. Meanwhile, career exploration is a key component of early career development and plays an important intermediary role in linking initial STEM entry with long-term persistence [29,30]. Exploration behaviors support career identity formation, goal clarity, and sustained academic engagement [31,32,33], while also helping students evaluate barriers and mobilize resources, particularly among women and other underrepresented groups [26,34,35]. However, few studies have jointly examined exploration and persistence by gender within STEM fields [36,37]. These processes differ by gender, with women more affected by lower self-efficacy and sociocultural factors, whereas men’s persistence is more directly driven by engineering self-efficacy [4,8,19,20,38,39,40].

Although prior research has advanced understanding of STEM entry, development, and persistence by examining supports, barriers, self-efficacy, outcome expectations, gender differences, identity, and career aspirations [4,5,17,26,28,41,42,43], several limitations remain. Much of the existing literature focuses on adolescents, leaving limited empirical work on the structural determinants of undergraduate STEM students’ career exploration and persistence intentions. Although gender differences are frequently reported, few studies employ multigroup structural equation modeling (SEM) to compare gender-specific mechanisms among STEM undergraduates. Moreover, while research integrating contextual supports, barriers, and self-efficacy has grown, relatively few studies examine how these factors jointly predict both exploration and persistence within a single SCCT-informed model. Existing gender-difference studies have also been largely situated in Western contexts [1,8,44], despite evidence that East Asian educational systems, marked by competitive structures and deeply embedded sociocultural expectations, exhibit distinctive patterns, such as the “achievement–interest gap” documented in China [26]. South Korea shows similar dynamics, including persistent gender disparities throughout the STEM pipeline [10,45]. These contextual features underscore the need for cross-cultural evidence testing SCCT’s generalizability in East Asia. Rather than highlighting Korea’s particularities, the present study aims to address this Western-centric gap by empirically examining SCCT-based mechanisms of career development within an East Asian STEM context.

Within SCCT, the hypothesized paths linking contextual supports and barriers to self-efficacy, and subsequently to major motivation, exploration, and persistence, are grounded in its cognitive–motivational mediation process. This framework also enables the examination of gender-specific differences, given that women’s self-efficacy, motivations, and persistence are more strongly shaped by contextual and sociocultural factors than men’s. Based on this rationale, the present study investigates structural relationships among contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions among 2393 Korean STEM undergraduate students using the SCCT framework. The study additionally examines gender-specific differences in these structural pathways. The findings aim to deepen understanding of the gendered mechanisms underlying STEM career development and to inform STEM career-retention strategies, gender-equity policies, and higher education support programs. Accordingly, the research questions are:

RQ1. 

What are the structural relationships among environmental factors (contextual supports and career barriers), engineering self-efficacy, and major motivation that influence STEM undergraduates’ career exploration behaviors and academic persistence intentions?

RQ2. 

Do these structural relationships differ between male and female STEM undergraduates?

Figure 1 presents the research model developed based on prior research.

2. Literature Review

2.1. Gender Disparities in STEM

The low participation of women in STEM remains a persistent global concern [1], understood not only as a workforce issue but also as a structural challenge limiting innovation, economic productivity, and gender equity [46]. Gender disparities and women’s attrition, the so-called “leaky pipeline”, emerge at multiple stages, from university entry through advanced study and into the labor market [1,7,47,48]. These patterns reflect cumulative structural, cultural, and psychological constraints rather than a single point of departure [1]. Importantly, these disparities intersect with cultural, racial/ethnic, and policy contexts [49,50]. Although STEM graduates typically experience lower unemployment rates than non-STEM graduates, women in STEM encounter early-career inequities in salary and working conditions [51] and face additional barriers to entering, remaining in, and advancing within STEM occupations. These challenges make women more likely to leave STEM fields [7,46]. Understanding gendered patterns in STEM undergraduates’ career exploration and academic persistence is therefore critical for reducing attrition and promoting long-term labor market equity.

STEM career development research has traditionally emphasized individual psychological factors, such as self-efficacy [52], motivation [14], and decision behaviors. However, recent studies show that these mechanisms differ by gender [4,47] and may be mediated or moderated by gendered social processes [19,20,40,53,54,55]. These disparities reflect sociocultural gender-role expectations and stereotypical cues rather than personal competence alone [6,48,56]. Divergent goal orientations may further contribute to gendered pathways: men tend to prioritize agentic goals (e.g., status, income), whereas women more often endorse communal goals [8,57], influencing their persistence mechanisms in STEM fields [58]. Female STEM students typically perceive more barriers than male students, while also reporting stronger support from close social networks such as family and peers [54]. Gendered differences in learning environments further shape STEM career development. For women, research opportunities, structured curricula [59], and faculty or peer support [24,35,46,60] strengthen career commitment.

2.2. Gender Differences in Career Exploration Behaviors and Their Predictors

Career exploration is an enactment process that involves information seeking, self-exploration, and environmental exploration [36,61,62]. It constitutes a core developmental mechanism through which individuals gather and interpret information about themselves and the world of work, forming an integrated basis for career decision-making [34,61,63,64]. Career exploration is a central behavioral mechanism explaining gender differences in exploration and persistence among STEM majors [36,37], and predicts more effective decision-making, better person–job fit, and improved early career outcomes [51,65]. Recent studies indicate that exploration processes vary by gender and cultural context [37], underscoring the need to examine gender-differentiated mechanisms in STEM fields. For example, women’s exploration tends to be constrained by lower self-efficacy and ability-related anxiety, ultimately weakening their persistence in STEM [9,66]. Guided by SCCT [14] and subsequent empirical work [31,32,34,36,37,46,55], prior research identifies both cognitive–personal factors (e.g., self-efficacy, outcome expectations, career confidence, exploration self-efficacy, exploration goals, adaptability, and career preparation timing) and social–environmental factors (e.g., family and social support, contextual supports, and career barriers) as major antecedents of exploration. These influences vary by gender and developmental stage [16,32,33,37,55,67].

Research on gender differences in career exploration has yielded mixed results, though structural patterns are consistently observed [46]. While some studies suggest that women’s exploration may be constrained by limited access to gender-matched role models, others indicate that exclusion experiences can, in some cases, strengthen exploratory motivation. Overall, most findings converge on gendered determinants and pathways of exploration. For women, increases in career decision-making self-efficacy tend to reciprocally reinforce later exploration, whereas this reciprocity is weaker for men [55]. In STEM contexts, male students often show a direct and immediate link between self-efficacy and exploration, with exploration patterns tied closely to self-directed efficacy and academic achievement [19,53]. Female students, by contrast, typically require supportive contextual conditions, such as social support, belonging, and role models, before exploration is meaningfully activated [19,53,55]. Accordingly, mentoring and exposure to role models have been shown to enhance women’s exploration by elevating self-efficacy and interest [35,60,68,69,70,71], thereby increasing the likelihood of persisting in STEM majors [72]. These patterns suggest that female STEM students are more responsive to social comparison and stereotype threat, and that supportive mentoring and collaborative environments enhance exploration by strengthening self-efficacy and major-related interest.

Gender differences in perceived career barriers further shape exploration patterns. Female STEM majors report stronger confidence in career identity, commitment, learning relevance, and exploration than women in non-STEM fields, suggesting that STEM environments may bolster women’s exploratory intentions [46]. However, they show lower confidence than male students in reconsidering career commitment and navigating occupational mobility [46]. Importantly, female STEM students perceive greater career barriers than men [34,73], and such perceptions tend to suppress exploration [6,9,74]. In contrast, men’s exploration is more directly driven by performance- and achievement-oriented self-efficacy [5,40,75], and barriers appear less detrimental and may even reinforce confidence [76,77]. Gender differences also emerge in major interest and perceived goal–role fit. Men tend to prefer object- and system-oriented domains, whereas women gravitate toward people- and society-oriented interests; when goal–role congruity is low, women’s exploration declines [8,57,78,79]. These patterns underscore a structural mechanism in which social expectations and gender stereotypes shape exploration self-efficacy.

Studies of job search behaviors among STEM graduates show no major gender differences in overall job-search intensity. In fact, women often experience better early employment outcomes due to stronger academic performance and high-quality internships, which significantly increase their likelihood of receiving early job offers, an effect not observed for men [80]. However, gender differences emerge in exploration intensity, networking methods, and information-use strategies, shaped by structural and cultural forces [65,77,80]. Perceived employability and career readiness also exhibit meaningful gender differences, even after controlling for major, performance, and experiences [46]. Women’s perceived employability is significantly influenced by social support and access to resources, while men’s exploration is more directly predicted by self-efficacy and academic confidence. These findings suggest that perceived employability may function as an additional pathway influencing exploration beyond traditional SCCT variables.

Intersectional analyses further demonstrate that gender, race, and social class jointly structure exploration strategies and employment outcomes [80]. Female and minoritized students often have more limited networks and support portfolios, while male students tend to use more formal channels and performance-oriented strategies. Such disparities indicate that differences in exploration competence and social capital developed during university translate into unequal labor market outcomes, underscoring the importance of mentoring, networking, and field-based programs in STEM. Consistent with this, ref. [35] showed that female students’ exploration behaviors depend heavily on social support within structures of unequal resource access, highlighting that exploration gaps arise not solely from individual-level differences but from broader structural and environmental constraints.

2.3. Gender Differences in Academic Persistence Intentions and Their Predictors

Academic persistence intentions, defined as students’ intentions to remain in their major, develop through interactions among self-efficacy, outcome expectations (major satisfaction), and career exploration behaviors [16,22,81,82]. Because female STEM students are more likely than men to leave the field after graduation [7], persistence and retention are essential for sustaining the STEM workforce and reducing gender disparities [58]. Extensive longitudinal and multigroup SCCT research identifies self-efficacy, outcome expectations, major interest, contextual supports, and barriers as key determinants of persistence [4,14,21,55,83,84,85,86]. Across studies, the motivational pathway linking self-efficacy, outcome expectations, and major interest is central to explaining persistence. Self-efficacy significantly shapes outcome expectations and STEM career interest [87], predicts academic achievement, and supports long-term persistence [21,88,89]. Outcome expectations, consistent with expectancy–value theory, influence STEM motivation and engagement [21,86]. Major interest also drives persistence and exhibits gendered patterns [22,82]. Additional factors, including career motivation, identity, insight, and resilience, reinforce persistence [90], and major self-efficacy plays a decisive role by integrating personal interests and expected outcomes [21,86,89].

Numerous studies report that women in STEM exhibit significantly lower self-efficacy than men [19,30,35,71,91,92,93], even when their academic achievement is comparable or higher [46,59]. These gender gaps emerge early, during middle and high school, critical periods for career decision-making [4,59]. Women also tend to view themselves as less competent in STEM tasks [4,94], and lower self-efficacy mediates reduced STEM interest and career goals [4]. Although self-efficacy is central to STEM career commitment for both genders, the mechanisms differ [5], and gender stereotypes amplify these disparities. Professional role confidence, comprising expertise and career-fit confidence, also shapes persistence, reflecting individuals’ beliefs in their ability to enact professional roles and identities [66]. Lower levels of such confidence increase women’s attrition from STEM and hinder their academic persistence [20,66]. Belongingness and cultural cues further influence these patterns [6,48,66]. By contrast, men’s persistence intentions are more directly driven by achievement- and performance-oriented self-efficacy and engineering self-efficacy [20,40,95,96].

Contextual factors, namely contextual supports and career barriers, are central determinants of career decisions within SCCT [58]. Socioeconomic status, including parental education and income, also influences women’s academic persistence and degree completion [81,87]. Strong support structures such as mentoring, flexible policies, and inclusive practices enhance women’s academic and occupational retention [58]. Contextual supports, including social support and role models, reliably strengthen engineering self-efficacy and academic persistence [16,21,88], whereas internal and external barriers have negative effects [21,82,89]. Given the nature of STEM environments, women tend to perceive more career barriers than men [34,73] and often feel compelled to exert greater effort to succeed [59], contributing to decisions not to persist in STEM [16,22,81,82]. Gendered differences also appear in how barriers operate: barriers typically weaken women’s persistence indirectly by constraining exploration, whereas for men they exert more direct negative effects through engineering self-efficacy [46,58,97].

Value incongruity further contributes to gender disparities in persistence. STEM careers are often viewed as misaligned with communal goals, or values emphasizing social contribution [4,8,57]. Individuals endorsing communal goals show lower interest in STEM [8,39], and such incongruity is especially detrimental for women [46]. Additionally, gender bias undermines women’s belonging and participation in STEM [9]. Research on gender differences in contextual supports shows that women often benefit more from supportive environments [71,93], although findings remain mixed. Some studies find that teaching support exerts a more significant positive effect on men [97], consistent with studies suggesting that men may, in certain contexts, derive greater benefit from support [35,93]. Parental support positively influences self-efficacy and career expectations for both genders [54], while course resources and research opportunities tend to have greater effects on women, and peer interactions have stronger effects on men [35]. Women generally perceive more peer and family support and fewer family-related barriers, whereas teacher support appears more salient for men [24]. Mentoring, peer networks, and female role models also strengthen women’s persistence by enhancing self-efficacy and belongingness [42,60,68,98,99].

Inclusive instructional practices, such as low-stakes assessment and visibility of competencies, are likewise effective in supporting women’s engagement [30,59,100,101]. By contrast, men’s persistence is more directly driven by performance-based self-efficacy [20,40,96]. Educational experiences also shape persistence intentions: faculty and peer interactions reliably predict STEM persistence [81,102], and satisfaction with the learning environment affects persistence through self-efficacy [88]. Overall, women’s persistence is strengthened by social support, mentoring, and role models, and weakened by perceived career barriers [16,21,87,88,89]. The magnitude and pathways of these effects vary by gender and by cultural–institutional context [35,54,71,93]. Additionally, communal goal incongruity and gender bias undermine women’s sense of belonging and participation in STEM [4,8,9,57].

Gender differences in academic persistence intentions among STEM undergraduates parallel patterns observed in post-graduation retention. For early-career women, persistence is reinforced more by “sustains”, including mentorship, inclusive climates, peer and supervisor support, role models, professional development, belonging, and community, than it is undermined by “drains,” such as discrimination, bias, overload, isolation, invisible labor, work–life conflict, and lack of recognition [58]. Structural factors such as sociocultural expectations, impostor feelings, and work–family imbalance further weaken women’s persistence [7]. By contrast, men’s persistence intentions tend to align more closely with professional expertise and expectations of economic achievement [35,80].

3. Methods

3.1. Data Sources and Sample

The data analyzed in this study were collected as part of an institutional project conducted by the Korea Foundation for Women in Science, Engineering and Technology (WISET). A total of 2393 responses from STEM undergraduates enrolled in 51 universities across eight regions in South Korea, including three metropolitan regions in the capital area and five non-metropolitan regions, were validated and analyzed. The dataset includes information on individual backgrounds (gender and grade level) and university characteristics (major and university location). The sample comprised 1189 male students (49.7%) and 1204 female students (50.3%). First-year students constituted 11.2% of the sample, sophomores 26.39%, juniors 26.0%, and seniors 36.5%. Engineering students comprised 48.8% of the sample, while students in the natural sciences accounted for 51.2%. The target population included undergraduate students enrolled in four-year universities in both the metropolitan region (Seoul, Incheon, and Gyeonggi) and non-metropolitan regions. The sample was evenly divided by region (50% metropolitan, 50% non-metropolitan). A detailed summary of the undergraduates’ demographics is provided in

Table 1. Undergraduates’ demographics.

Demographics	Undergraduates n (%)
Sex	Male: n = 1189 (49.7%); Female: n = 1204 (50.3%)
Grade	Freshmen: n = 268 (11.2%); Sophomores: n = 629 (26.39%); Juniors: n= 622 (26.0%); Seniors: n = 874 (36.5%)
Major	Engineering: n = 1167 (48.8%); Natural sciences: n = 1226 (51.2%)
University location (Region)	Metropolitan regions: n = 1197 (50%); Non-metropolitan regions: n = 1196 (50%)

. The gender and regional distributions in the dataset reflect the characteristics of the original WISET data collection process and were not artificially balanced for the purposes of this study. These proportions should therefore be interpreted as properties of the dataset rather than as strictly representative of the population.

3.2. Measures

The study included the following research variables: career exploration behaviors, academic persistence intentions, contextual supports, career barriers, engineering self-efficacy, and major motivation among STEM students in South Korea. First, career exploration behaviors were measured using items based on the scale originally developed by Stumpf et al. (1983) [61], later adapted and culturally tailored for Korea by Kim (2019) [103]. The final items were validated through confirmatory factor analysis (CFA). The scale consists of 15 items across three subscales (media information search, professional information search, and interpersonal information search). Responses were rated on a 5-point Likert scale (1 = not at all; 5 = very much). The scale demonstrated good reliability in this study with a Cronbach’s alpha of 0.868. Second, academic persistence intentions were assessed using a shortened 4-item Korean scale [104] developed from prior research [105]. Items were rated on a 5-point Likert scale (1 = not at all; 5 = very much). The scale demonstrated good reliability in this study with a Cronbach’s alpha of 0.732. Third, contextual supports and career barriers were measured using shortened 8-item and 11-item scales, respectively, translated into Korean and previously validated [106]. These instruments are based on the widely used contextual supports and barriers framework by Lent and colleagues [14,85], which has extensive evidence of reliability and validity. Contextual supports included two subscales: external/environmental support and family/friend support. Career barriers comprised two subscales: discrimination barriers and competitive pressure barriers. All items were rated on a 5-point Likert scale. Both scales showed strong reliability in the current sample, with Cronbach’s alpha values of 0.828 and 0.907 for contextual supports and career barriers, respectively. Fourth, engineering self-efficacy and major motivation represented cognitive–motivational factors. Engineering self-efficacy was assessed using a shortened 19-item scale [107], covering effort and satisfaction, major aptitude, and goal setting and confidence. Items were rated on a 5-point Likert scale. The scale demonstrated good reliability in the current sample, with a Cronbach’s alpha of 0.91. Major motivation was measured using a shortened 14-item Korean-validated scale based on [85,104], comprising expected outcomes in the major and interest in the major. Items were rated on a 5-point Likert scale. The scale showed good reliability in the current sample, with a Cronbach’s alpha of 0.855. In line with SCCT, major motivation in this study integrates both outcome expectations and intrinsic interest in the major, reflecting the cognitive–motivational processes that guide academic engagement and persistence.

3.3. Data Analysis

Data were analyzed within an SEM framework. Descriptive statistics and Pearson’s correlations were first calculated to examine the distributional characteristics of the variables and the initial associations among key constructs. Prior to estimating the structural model, a measurement model was tested using CFA to assess reliability and validity. Construct reliability was evaluated through Cronbach’s α and composite reliability (CR), and convergent validity was examined using average variance extracted (AVE), factor loadings, and model fit indices. After confirming the adequacy of the measurement model, the hypothesized structural model was estimated to identify the structural relationships among contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions. Standardized path coefficients, indirect effects, and overall model fit were evaluated.

To address the second research question, a multi-group SEM analysis was conducted to compare the structural paths between female and male students. Measurement invariance was examined sequentially (configural, metric, and scalar invariance), and nested model comparisons were performed using χ² difference tests and changes in CFI. Model invariance was primarily evaluated based on changes in fit indices, particularly ΔCFI, with ΔCFI ≤ 0.01 indicating acceptable invariance across nested models. Group-specific path coefficients were then interpreted to identify gender-based differences in structural relationships. All analyses were performed using AMOS 29.0 (IBM Corp., Armonk, NY, USA), and statistical significance was set at p < 0.05. Potential multicollinearity among latent constructs was considered to ensure the stability and interpretability of the estimated structural relationships. To assess common method bias, Harman’s single-factor test was conducted using an unrotated exploratory factor analysis including all measurement items. Although variance inflation factors (VIFs) were examined as an auxiliary check, multicollinearity in SEM is more appropriately evaluated using inter-construct correlations and HTMT ratios. All HTMT values were below the commonly accepted threshold of 0.90, with the highest value observed between engineering self-efficacy and major motivation.

4. Results

4.1. Descriptive Statistics and Correlation Analysis Results

Table 2. Descriptive statistics, reliability estimates, and correlations among study variables.

Latent Variable	Observed Variable	1		2		3			4		5			6
		1.1.	1.2.	2.1.	2.2.	3.1.	3.2.	3.3.	4.1.	4.2.	5.1.	5.2.	5.3.	6.1.	6.2.
1	1.1.	1
	1.2.	0.451 *	1
2	2.1.	0.040 *	−0.294 *	1
	2.2.	−0.221 *	−0.192 *	0.575 *	1
3	3.1.	0.476 *	0.575 *	−0.197 *	−0.164 *	1
	3.2.	0.501 *	0.381 *	−0.002	−0.094 *	0.638 *	1
	3.3.	0.466 *	0.479 *	−0.067 *	−0.101 *	0.708 *	0.574 *	1
4	4.1.	0.558 *	0.494 *	−0.093 *	−0.212 *	0.601 *	0.601 *	0.543 *	1
	4.2.	0.501 *	0.436 *	0.011	−0.094 *	0.645 *	0.701 *	0.580 *	0.623 *	1
5	5.1.	0.293 *	0.254 *	−0.069 *	−0.055 *	0.469 *	0.389 *	0.492 *	0.346 *	0.472 *	1
	5.2.	0.414 *	0.157 *	0.208 *	0.066 *	0.298 *	0.401 *	0.376 *	0.341 *	0.383 *	0.412 *	1
	5.3.	0.474 *	0.327 *	0.044 *	−0.052 *	0.476 *	0.404 *	0.534 *	0.423 *	0.471 *	0.535 *	0.595 *	1
6	6.1.	0.367 *	0.470 *	−0.216 *	−0.173 *	0.584 *	0.5 *	0.467 *	0.533 *	0.504 *	0.352 *	0.177 *	0.308 *	1
	6.2.	0.297 *	0.548 *	−0.325 *	−0.184 *	0.584 *	0.337 *	0.461 *	0.469 *	0.422 *	0.307 *	0.055 *	0.308 *	0.584 *	1
M		3.54	3.85	2.52	3.26	3.75	3.67	3.66	3.67	3.59	3.69	3.32	3.5	3.85	3.87
SD		0.7	0.67	0.95	0.86	0.56	0.62	0.62	0.62	0.66	0.66	0.79	0.69	0.67	0.7
Cronbach’s α		0.774	0.764	0.91	0.8	0.832	0.796	0.778	0.855	0.822	0.808	0.795	0.679	0.732	0.73
Skewness		−0.41	−0.3	0.22	−0.68	−0.17	−0.17	−0.25	−0.46	−0.55	−0.46	−0.48	−0.36	−0.38	−0.2
Kurtosis		3.28	3.13	2.25	3.35	3.16	2.91	3.02	3.76	3.78	3.89	3.07	3.26	3.31	2.76

1. Contextual supports, 1.1. external/environmental support, 1.2. family/friend support; 2. Career barriers, 2.1. Discrimination barrier, 2.2. Competitive pressure barrier; 3. Engineering self-efficacy, 3.1. Effort and satisfaction, 3.2. Major aptitude, 3.3. Goal setting and confidence; 4. Major motivation, 4.1. Expected outcomes in major, 4.2. Interest in major; 5. Career exploration behaviors, 5.1. Media information search, 5.2. Professional information search, 5.3. Interpersonal information search; 6. Academic persistence intentions, 6.1. Overcoming challenges, 6.2. Retention and graduation; * p < 0.05.

presents the descriptive statistics and correlations among the key variables. The mean (M), standard deviation (SD), skewness, and kurtosis values were examined to assess the distributional characteristics of the data. Pearson’s correlation coefficients were calculated to explore the relationships among contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions.

The correlation analysis results indicate that contextual supports were positively correlated with engineering self-efficacy (r = 0.381~0.575, p < 0.05) and major motivation (r = 0.436~0.558, p < 0.05). In contrast, career barriers showed a significant negative correlation with engineering self-efficacy (r = −0.067 to −0.197, p < 0.05). However, the correlations with career exploration behaviors were weak and mixed (r = −0.044 to 0.208, p < 0.05), indicating that the direction and strength of the association varied across subdimensions. Furthermore, engineering self-efficacy was positively associated with both academic persistence intentions (r = 0.337~0.584, p < 0.05) and career exploration behaviors (r = 0.298~0.534, p < 0.05), supporting the theoretical assumptions of the study. The results suggest that students with higher levels of engineering self-efficacy and major interest are more likely to persist in their STEM studies and actively engage in career exploration. To evaluate the normality of the data, skewness and kurtosis values were examined. The absolute values of skewness ranged from 0.22 to 0.68, which falls well within the commonly accepted threshold of |1.0| for approximately symmetric distributions [108]. The absolute values of kurtosis ranged from 2.25 to 3.89, remaining below the widely used criterion of |7.0|, which is considered acceptable for behavioral and social science data [109]. These results indicate that the data satisfied the normality assumptions required for regression-based path analysis.

In addition, to assess the potential impact of common method bias, Harman’s single-factor test was conducted using an unrotated exploratory factor analysis including all measurement items. The results showed that the first factor accounted for 24.13% of the total variance, which is well below the recommended threshold of 50%, indicating that common method bias is unlikely to pose a serious threat to the validity of the findings. Although Harman’s single-factor test has known limitations, the relatively low variance explained by the first factor suggests that common method bias is unlikely to substantially affect the observed relationships. Future studies may further apply more advanced techniques, such as a common latent factor approach, to validate these findings.

4.2. Measurement Model

Before estimating the structural model, we evaluated the reliability and convergent validity of all latent constructs. CFAs were conducted for each construct, and all standardized factor loadings exceeded the recommended threshold of 0.50, indicating adequate item–construct relationships. Internal consistency reliability was satisfactory, with Cronbach’s alpha coefficients ranging from 0.73 to 0.91. CR values ranged from 0.777 to 0.937, exceeding the recommended cutoff of 0.70 for all constructs. Convergent validity was also supported, as AVE values ranged from 0.635 to 0.832—well above the recommended minimum of 0.50. These results confirm that the measurement scales exhibit strong reliability and convergent validity.

Table 3. Measurement model: Reliability and convergent validity.

Construct	Items (k)	Cronbach’s α	CR	AVE
Contextual supports	8	0.83	0.777	0.635
Career barriers	11	0.91	0.829	0.722
Engineering self-efficacy	19	0.91	0.937	0.832
Major motivation	14	0.86	0.931	0.818
Career exploration behaviors	15	0.87	0.867	0.688
Academic persistence intentions	4	0.73	0.857	0.749

summarizes these indices for each construct. In addition to convergent validity, discriminant validity was assessed using the Fornell–Larcker criterion. As shown in

Table 4. Discriminant validity (Fornell–Larcker Criterion).

Construct	1	2	3	4	5	6
1. Contextual supports	0.797
2. Career barriers	0.221	0.85
3. Engineering self-efficacy	0.575	−0.197	0.912
4. Major motivation	0.558	−0.212	0.701	0.904
5. Career exploration	0.293	−0.069	0.534	0.472	0.829
6. Academic persistence intentions	0.548	−0.325	0.584	0.533	0.308	0.865

, the square root of the AVE for each construct exceeded the corresponding inter-construct correlations, confirming adequate discriminant validity. Furthermore, discriminant validity was also assessed using the HTMT (Heterotrait–Monotrait) ratio. As shown in

Table 5. HTMT Ratios for Discriminant Validity Assessment.

Construct	1	2	3	4	5	6
1. Contextual supports	—
2. Career barriers	0.289	—
3. Engineering self-efficacy	0.751	0.178	—
4. Major motivation	0.714	0.206	0.866	—
5. Career exploration	0.551	0.160	0.663	0.656	—
6. Academic persistence intentions	0.687	0.365	0.785	0.790	0.456	—

, all HTMT ratios were below the commonly accepted threshold of 0.90, providing additional support for discriminant validity. The highest HTMT value was 0.866 between engineering self-efficacy and major motivation. Although this value slightly exceeds the more conservative 0.85 criterion, it remains below the widely accepted 0.90 threshold and thus supports acceptable discriminant validity. This pattern is theoretically consistent with the SCCT framework, in which self-efficacy and motivation are conceptually related yet empirically distinguishable constructs.

4.3. Path Analysis Results

Path analysis was conducted to examine the structural relationships among contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions. To identify gender-specific patterns, the analyses were conducted separately for female and male students, followed by a comparative examination of structural differences. Before interpreting the structural pathways, overall model fit was assessed.

The overall model demonstrated a generally acceptable, though not optimal, fit. Most incremental fit indices reached recommended thresholds (CFI ≈ 0.90, TLI ≈ 0.90, IFI ≈ 0.91, NFI ≈ 0.90), while the RMSEA value (0.09) indicated a marginal level of fit. Such patterns are commonly observed in complex models with a large number of observed indicators, as different fit indices capture distinct aspects of model fit [110,111]. In particular, RMSEA has been shown to overestimate model misfit in models with large numbers of indicators and degrees of freedom [112]. However, RMSEA values for both the female (0.112) and male (0.121) groups exceeded commonly recommended thresholds. These elevated RMSEA values suggest some caution in interpreting overall model fit. Nevertheless, other fit indices and the theoretical consistency of the model support its overall adequacy. In addition, the measurement model demonstrated adequate reliability and validity, providing further support for the interpretation of the structural relationships. No theoretically meaningful modifications resulted in substantively improved model fit; therefore, the current model was retained based on theoretical coherence. This limitation should be considered when interpreting the magnitude of the estimated structural relationships.

The direct, indirect, and total effects—including unstandardized coefficients (B), standardized coefficients (β), standard errors (SE), critical ratios (CR), and significance levels (p-value)—are summarized in

Table 6. Direct and indirect effects analysis for female students.

Path Between Variables			Direct Effect		Indirect Effect		Total Effect
			Unstandardized Coefficient	Standardized Coefficient	Unstandardized Coefficient	Standardized Coefficient	Unstandardized Coefficient	Standardized Coefficient
Contextual supports	→	Engineering self-efficacy	1.01	0.96	0	0	1.01	0.96
Career barriers	→	Engineering self-efficacy	0.11	0.09	0	0	0.11	0.09
Engineering self-efficacy	→	Major motivation	0.91	0.99	0	0	0.91	0.99
Engineering self-efficacy	→	Career exploration behaviors	0	0	0.9	0.71	0.9	0.71
Engineering self-efficacy	→	Academic persistence intentions	0.56	0.52	0.1	0.09	0.66	0.61
Major motivation	→	Career exploration behaviors	0.99	0.71	0	0	0.99	0.71
Major motivation	→	Academic persistence intentions	0.24	0.25	−0.02	−0.02	0.22	0.23
Career exploration behaviors	→	Academic persistence intentions	−0.11	−0.11	0	0	−0.11	−0.11
Contextual supports	→	Career exploration behaviors	0	0	0.91	0.68	0.91	0.68
Career barriers	→	Career exploration behaviors	0	0	0.1	0.06	0.1	0.06
Contextual supports	→	Academic persistence intentions	0.97	0.8	0	0	0.97	0.8
Career barriers	→	Academic persistence intentions	−0.28	−0.19	0	0	−0.28	−0.19

Note: The arrow symbol (→) indicates the directional path relationship between variables in the structural model.

(female students) and

Table 7. Direct and indirect effects analysis for male students.

Path Between Variables			Direct Effect		Indirect Effect		Total Effect
			Unstandardized Coefficient	Standardized Coefficient	Unstandardized Coefficient	Standardized Coefficient	Unstandardized Coefficient	Standardized Coefficient
Contextual supports	→	Engineering self-efficacy	1.18	1	0	0	1.18	1
Career barriers	→	Engineering self-efficacy	0.1	0.15	0	0	0.1	0.15
Engineering self-efficacy	→	Major motivation	0.9	0.98	0	0	0.9	0.98
Engineering self-efficacy	→	Career exploration behaviors	0	0	0.64	0.74	0.64	0.74
Engineering self-efficacy	→	Academic persistence intentions	0.48	0.42	0.24	0.21	0.72	0.63
Major motivation	→	Career exploration behaviors	0.71	0.75	0	0	0.71	0.75
Major motivation	→	Academic persistence intentions	0.34	0.34	−0.01	−0.01	0.34	0.34
Career exploration behaviors	→	Academic persistence intentions	−0.02	−0.02	0	0	−0.02	−0.02
Contextual supports	→	Career exploration behaviors	0	0	0.75	0.74	0.75	0.74
Career barriers	→	Career exploration behaviors	0	0	0.07	0.11	0.07	0.11
Contextual supports	→	Academic persistence intentions	1.05	0.89	0	0	1.05	0.89
Career barriers	→	Academic persistence intentions	−0.06	−0.08	0	0	−0.06	−0.08

Note: The arrow symbol (→) indicates the directional path relationship between variables in the structural model.

(male students). The final structural diagrams are presented in Figure 2 and Figure 3, respectively. Section 4.3.3 provides a comparative interpretation of gender similarities and differences in the structural pathways.

4.3.1. Path Analysis Results for Female Students

To examine the structural relationships among the selected variables for female students, the path model was estimated using the Maximum Likelihood Estimation (MLE) method. The model demonstrated an acceptable, though not optimal, level of fit, with several incremental fit indices approaching commonly recommended thresholds (GFI = 0.859, NFI = 0.876, CFI = 0.89, TLI = 0.89), whereas RMSEA (0.112) indicated marginal fit. Consistent with the overall model assessment reported above, the female-group model was retained based on its theoretical coherence and acceptable incremental fit indices. The standardized path estimates and significance levels are presented in Table 6, and the final structural model is illustrated in Figure 2.

For female STEM undergraduates, engineering self-efficacy emerged as the central mechanism through which contextual and motivational factors shaped persistence intentions. Contextual supports exerted a significant positive influence on engineering self-efficacy (β = 0.96), whereas career barriers showed only a weak positive direct effect (β = 0.09). Engineering self-efficacy, in turn, significantly predicted major motivation (β = 0.99), indicating that higher confidence in engineering abilities was closely tied to students’ expectations of positive academic and career outcomes. Self-efficacy also had a significant direct effect on academic persistence intentions (β = 0.52), and its indirect influence through major motivation (β = 0.09) contributed to a robust total effect (β = 0.61). Together, these findings underscore the significant mediating role of engineering self-efficacy in cultivating female students’ intentions to persist in STEM pathways.

Major motivation also directly enhanced academic persistence intentions (β = 0.25), although its indirect effect through career exploration behaviors was minimal. Interestingly, career exploration behaviors negatively predicted persistence intentions (β = –0.11) among female students. This suggests that, once self-efficacy and motivation are controlled, increased exploration may reflect uncertainty or reconsideration of one’s academic direction rather than a strong commitment to the STEM major.

Regarding antecedents of persistence intentions, contextual supports exerted a significant total effect (β = 0.80), while career barriers demonstrated a moderate negative total effect (β = –0.19). These findings indicate that supportive learning environments remain a critical foundation for promoting persistence among female students, whereas perceived barriers weaken their intention to remain in STEM. Neither contextual supports nor career barriers showed significant direct or indirect effects on career exploration behaviors, and the indirect influence of major motivation on exploration was similarly negligible. Overall, the model suggests that for female STEM undergraduates, persistence intentions are driven primarily by engineering self-efficacy and major motivation, whereas exploration behaviors may function differently from traditional SCCT expectations.

Bivariate correlations indicated that career barriers were negatively associated with engineering self-efficacy for both female and male students. However, in the multivariate SEM, the direct path from career barriers to self-efficacy became slightly positive, suggesting a suppression effect arising from shared variance among predictors. This coefficient should therefore be interpreted with caution, and we refrain from drawing substantive conclusions based on its sign. Similarly, although the bivariate correlations showed that career exploration behaviors were positively associated with academic persistence intentions for both genders, the direct path became negative for female students once self-efficacy and major motivation were included in the model. This again indicates a suppression effect: when central motivational predictors are controlled, greater exploration may reflect uncertainty-driven behaviors linked to lower persistence intentions. Accordingly, we interpret this path with caution while emphasizing that the zero-order association between exploration and persistence remained positive.

Notably, some standardized path coefficients were very high (approaching 1.0), particularly in the relationship from contextual supports to engineering self-efficacy and from engineering self-efficacy to major motivation. While such strong associations are theoretically consistent with SCCT, they may also reflect high shared variance among conceptually related constructs. Therefore, these estimates should be interpreted with caution. However, discriminant validity tests (Fornell–Larcker and HTMT) confirmed that the constructs are empirically distinct, suggesting that these high coefficients reflect strong theoretical relationships rather than multicollinearity or construct redundancy.

4.3.2. Path Analysis Results for Male Students

For male students, the path model was estimated using the Maximum Likelihood Estimation (MLE) method. The model also demonstrated an acceptable, though not optimal, level of fit, with several incremental fit indices approaching commonly recommended thresholds (GFI = 0.832, NFI = 0.847, CFI = 0.89, TLI = 0.89), whereas RMSEA (0.121) indicated a marginal fit. Consistent with the overall model assessment reported above, the male-group model was retained based on its theoretical coherence and acceptable incremental fit indices. The standardized path coefficients, standard errors (SEs), critical ratios (CRs), and significance levels are reported in Table 7, and the final structural model is presented in Figure 3.

For male STEM undergraduates, engineering self-efficacy again emerged as a significant predictor within the structural model. Contextual supports exerted a significant positive effect on engineering self-efficacy (β = 1.00), and career barriers showed a marginal positive effect (β = 0.15). Engineering self-efficacy also significantly predicted major motivation (β = 0.98), mirroring the pattern observed among female students. Engineering self-efficacy demonstrated a significant direct effect on academic persistence intentions (β = 0.42) and a meaningful indirect effect through major motivation (β = 0.21), resulting in a robust total effect (β = 0.63). This indicates that self-efficacy functions as a key mechanism through which male students develop stronger intentions to persist in STEM majors. Major motivation also had a significant direct effect on persistence intentions (β = 0.34), while its indirect effect through career exploration behaviors was negligible.

As in the female group, career exploration behaviors negatively predicted persistence intentions, although the magnitude was very small (β = –0.02). Unlike the pattern observed among female students, however, engineering self-efficacy exhibited a significant indirect effect on career exploration behaviors (β = 0.74), suggesting that higher self-efficacy among male students may encourage more active exploration of academic or career pathways in STEM. Contextual supports and career barriers also exhibited indirect effects on exploration (β = 0.74 and β = 0.11, respectively), primarily mediated through engineering self-efficacy and major motivation.

Contextual supports exerted a significant total effect on academic persistence intentions (β = 0.89), whereas career barriers demonstrated only a small negative total effect (β = –0.08). Compared with the female group, the negative influence of barriers on persistence was weaker among male students, implying that male undergraduates may perceive structural or psychological barriers as less detrimental to their academic commitment. Notably, some standardized path coefficients in the male group were very high (approaching 1.0), particularly in the relationships from contextual supports to engineering self-efficacy and from engineering self-efficacy to major motivation. While such strong associations are theoretically consistent with SCCT, they may also reflect high shared variance among conceptually related constructs. Therefore, these estimates should be interpreted with caution. However, discriminant validity tests (Fornell–Larcker and HTMT) confirmed that the constructs are empirically distinct, suggesting that these high coefficients reflect strong theoretical relationships rather than multicollinearity or construct redundancy. Overall, the findings suggest that among male STEM undergraduates, engineering self-efficacy and major motivation play dominant roles in shaping academic persistence intentions. Indirect pathways through career exploration behaviors were more significant for male students than for female students, indicating possible gender differences in the functional role of exploration within the SCCT framework.

4.3.3. Gender Comparison of Structural Paths

The comparison between female and male students reveals several gender differences and is presented in

Table 8. Comparison of standardized total effects by gender.

Path Between Variables			Standardized Total Effect Coefficient			Difference
			Overall	Female (A)	Male (B)	(A-B)
Contextual supports	→	Engineering self-efficacy	0.98	0.96	1	−0.04
Career barriers	→	Engineering self-efficacy	0.12	0.09	0.15	−0.06
Engineering self-efficacy	→	Major motivation	0.99	0.99	0.98	0.01
Engineering self-efficacy	→	Career exploration behaviors	0.71	0.71	0.74	−0.03
Engineering self-efficacy	→	Academic persistence intentions	0.62	0.61	0.63	−0.02
Major motivation	→	Career exploration behaviors	0.72	0.71	0.75	−0.04
Major motivation	→	Academic persistence intentions	0.29	0.23	0.34	−0.11
Career exploration behaviors	→	Academic persistence intentions	−0.07	−0.11	−0.02	−0.09
Contextual supports	→	Career exploration behaviors	0.7	0.68	0.74	−0.06
Career barriers	→	Career exploration behaviors	0.08	0.06	0.11	−0.05
Contextual supports	→	Academic persistence intentions	0.84	0.8	0.89	−0.09
Career barriers	→	Academic persistence intentions	−0.16	−0.19	−0.08	−0.11

Note: The arrow symbol (→) indicates the directional path relationship between variables in the structural model.

.

Gender-based comparisons of the standardized total effects revealed significant differences between female and male STEM undergraduates, complementing the group-specific findings reported earlier. For female students, engineering self-efficacy served as the most significant predictor of academic persistence intentions (β = 0.61), driven by both significant direct and indirect effects through major motivation. Female students also exhibited a significant negative total effect of career exploration behaviors on persistence intentions (β = –0.11), suggesting that exploration may signal uncertainty or reconsideration of their academic trajectory when motivational resources are controlled.

For male students, engineering self-efficacy likewise represented the most significant predictor of persistence intentions, with a slightly higher total effect (β = 0.63) compared with female students (difference = –0.02). However, more significant gender-specific differences emerged in the role of major motivation and career exploration behaviors. Major motivation exerted a more significant total effect on persistence intentions among male students (β = 0.34) relative to female students (β = 0.23), indicating that anticipated academic and career outcomes play a comparatively greater role in shaping male students’ commitment to their STEM pathways. In contrast, the negative association between career exploration behaviors and persistence was significantly weaker for male students (β = –0.02), suggesting that exploration carries less interpretive weight regarding commitment in this group.

Gender differences also appeared in the antecedents of engineering self-efficacy. Contextual supports and career barriers had slightly more significant positive effects on self-efficacy among male students (β = 1.00 and β = 0.15) compared with female students (β = 0.96 and β = 0.09), although these differences were marginal and unlikely to reflect distinct underlying mechanisms. Similarly, the indirect effects of contextual supports and barriers on career exploration behaviors were more significant for male students (β = 0.74 and β = 0.11), whereas these pathways were negligible for female students. This pattern suggests that exploration functions more as an efficacy-driven behavior among male students, whereas for female students it may be more closely tied to uncertainty.

Together, although the overarching SCCT-based structure is consistent across genders, the strength and meaning of several pathways diverge. Female students show a more significant negative association between career exploration behaviors and academic persistence intentions, whereas male students’ persistence intentions are more significantly shaped by major motivation. These differences highlight the importance of adopting gender-responsive strategies when designing interventions to support STEM students’ academic persistence and career development.

5. Discussion

5.1. Theoretical Contributions

The present study makes several important theoretical contributions to the advancement of SCCT in STEM education. First, the findings contribute to the growing body of work examining the cross-cultural generalizability of SCCT. Although SCCT has been extensively validated in Western contexts [14,16], its structural mechanisms have been less frequently tested in East Asian STEM ecosystems characterized by competitive learning environments and socially embedded expectations [4,5]. The present study demonstrates that the core SCCT sequence remains conceptually robust among Korean STEM undergraduates. Specifically, contextual supports influence self-efficacy, which in turn predicts major motivation and academic persistence intentions. However, several pathway differences emerged that suggest cultural specificity. These results indicate that SCCT may not operate with uniform strength across sociocultural contexts and that theoretical refinements may be necessary to account for the role of structural pressures and gendered expectations in East Asian settings.

Second, the study offers a significant theoretical extension by identifying gender-differentiated SCCT pathways within the same cultural and educational system. Although existing research has documented gender disparities in STEM motivation and persistence [7,9], relatively few studies have empirically examined how SCCT constructs interact differently for male and female students at the structural level. The present findings reveal that major motivation exerted a more significant effect on academic persistence intentions for male students, whereas female students showed greater sensitivity to contextual supports and barriers. This pattern aligns with prior research suggesting that women’s efficacy and persistence in STEM are more significantly shaped by relational cues, belongingness signals, and perceived environmental constraints [8,20]. By demonstrating that the relative importance of SCCT components varies systematically by gender, this study enriches theoretical discussions regarding the contingent nature of motivational processes and highlights the need for gender-responsive extensions of SCCT.

Third, a particularly noteworthy theoretical contribution concerns the functional meaning of career exploration. SCCT and the broader career development literature typically conceptualize exploration as an adaptive behavior that facilitates informed decision-making [61,65]. Contrary to this assumption, the present study found that career exploration was negatively associated with persistence intentions for female students. This pattern suggests that in highly competitive and gendered academic environments, exploration may heighten awareness of structural barriers, anticipated discrimination, or misalignment between personal goals and institutional climates. Recent studies documenting women’s STEM disengagement in response to uncertainty or stereotype threat [28,58,113,114] support this interpretation. The present findings therefore offer a theoretically novel insight by demonstrating that the adaptive value of exploration is context-sensitive and that exploration can function as a risk signal rather than a developmental asset under specific sociocultural pressures.

Finally, this study extends theoretical understandings of SCCT by demonstrating that contextual supports do not operate uniformly across genders. Although supports positively predicted engineering self-efficacy for all students, the effect was significantly stronger for women. This aligns with prior research showing that women’s confidence and persistence in STEM are particularly responsive to environmental affordances, relational support, and perceptions of belonging [80,89]. Given that most research on gender disparities in STEM has been conducted in Western contexts [1,2,42,48] and that studies explicitly examining non-Western settings such as China and Chile have only recently begun to grow [17,26,94,97,115], this analysis of structural pathways linking contextual supports, career barriers, self-efficacy, and major motivation among Korean university students—where high academic achievement, intense educational competition, and strong gender norms intersect—helps fill a key gap in the international literature [10,11,45,50,95,116,117]. The findings emphasize the importance of treating contextual supports not as neutral background factors but as gender-sensitive mechanisms that shape career development in distinctive ways. Collectively, these contributions position this study as an important empirical extension of SCCT by offering pathway-level refinements and revealing gender-specific mechanisms in a non-Western STEM context.

Furthermore, given that common method bias was assessed and found not to be a significant concern, it is unlikely that the observed strong structural relationships are merely artifacts of the measurement method. In this context, despite the strong structural relationships observed in this study, some standardized path coefficients approached the upper bound of 1.0. This pattern suggests a high degree of conceptual proximity between certain constructs, particularly engineering self-efficacy and major motivation, which are closely linked within the SCCT framework. Although discriminant validity was generally supported based on the Fornell–Larcker criterion, the possibility of shared variance or partial conceptual overlap among these constructs cannot be entirely ruled out. Therefore, the structural estimates should be interpreted with caution. Future research is needed to further refine construct boundaries and examine these relationships using alternative modeling approaches.

Finally, the RMSEA values for both groups were higher than commonly accepted cutoffs, which may reflect model complexity or sensitivity to sample size and degrees of freedom. Although multiple fit indices were considered to evaluate overall model adequacy, the elevated RMSEA values indicate that the model fit is not optimal. Therefore, the findings should be interpreted with caution, and future research should explore alternative model specifications or simplified structures to improve model fit.

5.2. Practical and Policy Implications

The gender-differentiated pathways identified in this study offer several practical implications for universities and policymakers seeking to strengthen the STEM pipeline. The finding that career exploration negatively predicts persistence intentions for female students indicates that exploration is not uniformly adaptive and may elicit anxiety or perceptions of constraint in gendered educational environments [118]. Prior research similarly shows that women often disengage from STEM when they anticipate discrimination or perceive structural inequities [9,58]. Institutions should therefore design career development programs that present exploration as a supported, scaffolded process rather than an isolated task. Structured advising, moderated career-information workshops, and guided reflection activities may reduce the cognitive load associated with uncertainty and prevent exploration from functioning as a deterrent. In the Korean higher education context, these interventions can be implemented through university-based career centers and national initiatives such as those led by WISET, which provide structured advising, curated career resources, and gender-sensitive mentoring to support female students’ exploration processes.

The significant influence of contextual supports on women’s self-efficacy also highlights the importance of relational and environmental interventions. Expanding mentoring networks that connect undergraduate women with graduate students and professionals, creating peer-based learning communities, and enhancing faculty accessibility can significantly strengthen women’s academic confidence and persistence [80,89]. In fields where women remain numerically underrepresented, visible role models and inclusive instructional climates are especially critical. For male students, the dominant role of major motivation in predicting persistence indicates the need for opportunities that sustain interest and deepen engagement. Undergraduate research experiences, industry-partnered capstone projects, and mastery-oriented feedback practices may reinforce motivational trajectories and strengthen their sense of professional identity [35]. These measures can also enhance the overall quality of STEM learning environments, promoting persistence across genders. At the institutional level, universities may benefit from systematically assessing environmental affordances using a contextual support index that monitors advising availability, departmental culture, inclusiveness, and perceived fairness. Given the persistent gender disparities in Korean STEM fields [10,11], national agencies such as WISET and MSIT could use such data to guide funding and policy decisions, support inclusive STEM program development, and expand mentoring and internship pipelines for underrepresented groups.

Together, the findings suggest that efforts to enhance STEM persistence should move beyond gender-neutral interventions and instead recognize the gendered structure of motivational and efficacy-related pathways. Women’s persistence is more significantly shaped by contextual supports and the reduction in career barriers, indicating the need for strategies that promote belonging, mentoring, and inclusive learning environments. By contrast, male students’ persistence appears more directly tied to performance-based self-efficacy and major motivation, suggesting that instructional approaches emphasizing mastery experiences may be particularly effective. Additionally, universities should expand structured career exploration opportunities, such as research participation, industry engagement, and mentoring networks, to reinforce self-efficacy and academic commitment across genders. These pathway-informed interventions may help reduce gender disparities, mitigate STEM attrition, and strengthen the long-term stability of the STEM talent pipeline.

5.3. Limitations and Future Research

Several limitations should be considered when interpreting the findings. First, the cross-sectional design limits causal inference. Longitudinal studies would enable researchers to examine how supports, self-efficacy, motivation, and exploration develop over time, particularly during critical transitions such as university entry or progression into advanced STEM coursework. Future research should examine when and how perceptions of uncertainty or constraint emerge during the undergraduate years and how these shape persistence over time. Second, although the dataset was nationally representative, the reliance on self-report measures may have introduced bias. Future studies could incorporate behavioral indicators, institutional records, or mixed-methods approaches to triangulate students’ career development experiences. Despite these limitations, the present findings offer a solid foundation for future research exploring how SCCT processes unfold across cultures, demographic groups, and STEM disciplines. In particular, the negative role of exploration for women and the gender-contingent function of contextual supports represent promising areas for theoretical refinement and intervention development.

6. Conclusions

This study examined the gender-differentiated structural mechanisms underlying STEM undergraduates’ academic persistence intentions in South Korea by applying SCCT to a large-scale national dataset. The findings revealed that while SCCT’s core pathways were broadly applicable, several relationships operated differently for male and female students. Contextual supports significantly predicted engineering self-efficacy and motivational processes for both groups, but the magnitude of these effects was significantly greater for women. Major motivation emerged as a significant predictor of persistence for men, whereas women’s persistence was more sensitive to perceived supports and barriers. Notably, career exploration was negatively associated with women’s persistence intentions, challenging traditional assumptions that exploration is universally adaptive.

These results offer several theoretical contributions. This study expands the cross-cultural generalizability of SCCT by demonstrating partial yet meaningful alignment with its core mechanisms in an East Asian STEM context. The identification of gender-specific pathways provides empirical support for refining SCCT to account for gendered experiences, particularly regarding the psychological salience of contextual supports and the context-dependent meaning of exploration. The findings contribute to ongoing discussions about how career development theories should be adapted to reflect cultural and structural realities beyond Western educational settings.

This study also offers actionable implications for practice and policy. For female STEM students, interventions should prioritize barrier management, structured career exploration, and relational supports through mentoring networks and inclusive departmental climates. For male students, strengthening motivation through research engagement, mastery-oriented learning environments, and industry-linked opportunities may enhance persistence. At the institutional and national levels, systematic monitoring of contextual supports and gender disparities can inform targeted initiatives aimed at improving equity in STEM participation. Although this study employs a cross-sectional design, which limits causal interpretation, the results provide a foundation for future research. Longitudinal analyses, mixed-methods approaches, and culturally comparative studies will be essential for deepening understanding of how supports, efficacy beliefs, and motivational processes unfold over time. In particular, the unexpected negative role of career exploration for women represents a promising direction for theoretical elaboration and intervention design. Overall, this study advances theoretical, empirical, and practical knowledge regarding gendered career development processes in STEM and underscores the importance of culturally grounded approaches to supporting students’ academic and professional trajectories.