Female peer mentors early in college have lasting positive impacts on female engineering students that persist beyond graduation

The study addresses persistent underrepresentation of women (and other minority groups) in STEM education and the workforce, particularly acute in engineering, despite national needs for a skilled STEM workforce. Prior randomized controlled interventions often focus on mental reappraisal (e.g., values affirmation, growth mindset, emotion regulation) to reduce achievement gaps, but these place the onus on students to adapt rather than on institutions to improve environments, and frequently assess short-term academic performance rather than subjective experiences (belonging, confidence, anxiety, motivation) that shape persistence. A complementary literature shows that exposure to own-group role models boosts performance and identification with STEM, consistent with the Stereotype Inoculation Model, but most studies are short-term and lab-based, lacking authentic, sustained mentoring relationships. This research tests whether authentic peer mentoring relationships with female mentors early in college causally improve female engineering students’ subjective experiences (motivation, confidence, well-being) and objective outcomes (internships, STEM persistence) over the long term. Building on prior work that showed first-year benefits of female peer mentors, the present study follows the full cohort through graduation and one year post-graduation to examine durability of effects, adds outcomes capturing work experience and emotional well-being, and measures major choice at graduation to more accurately assess STEM retention. The authors hypothesized that female peer mentors would act as “social vaccines” (per the Stereotype Inoculation Model), preserving motivation and confidence, and increasing persistence and professional development in STEM, with benefits persisting 3–5 years after mentoring ended.

The paper reviews two main intervention traditions. First, mental reappraisal interventions (self-affirmation, growth mindset, emotion reappraisal) can reduce disparities in academic performance but often: (a) emphasize individual cognitive change over environmental change; (b) lack measures of students’ subjective experiences; (c) are not STEM-specific and often target adolescents. Second, role model research demonstrates that exposure to successful own-group role models improves performance and STEM identification for women and racial/ethnic minorities; however, impacts are typically assessed at a single time point or short-term, and many studies are lab-based, limiting ecological validity and opportunities for authentic, sustained mentoring. Evidence on authentic mentoring (e.g., faculty mentoring) suggests improved persistence and well-being, particularly in supportive institutional contexts (e.g., HBCUs). Collectively, theory (Stereotype Inoculation Model) and evidence suggest that authentic mentoring relationships with own-group mentors may protect against stereotype threat and enhance long-term outcomes, but causal, longitudinal field experiments in STEM undergraduate settings have been scarce. The present study addresses these gaps by implementing a randomized, blinded, year-long peer mentoring intervention and tracking outcomes through and beyond graduation, also testing whether male mentors provide comparable benefits.

Design: Field-based longitudinal randomized controlled experiment at a large public university (UMass Amherst). Female undergraduates planning to major in engineering were randomly assigned during their first year to one of three conditions for one academic year: female peer mentor (n=52), male peer mentor (n=51), or no mentor (control; n=47). Participants and mentors were blinded to the study’s focus on mentoring and mentor sex. Mentoring interactions concluded after the first year. Participants were tracked annually through graduation and one year post-graduation (total data collection 2011–2019). Participants: N=150 female engineering students (80% first-year entrants, 20% transfer students). Mean age at baseline 18.34 (SD=1.34). Race/ethnicity: 67.3% white, 17.3% Asian, 5.3% multiracial, 2.7% Black, 2.7% Hispanic, 2% other. All consented to surveys and transcript access. Mentors: 58 undergraduate mentors (32 female, 26 male), primarily seniors, some juniors, in good standing and typically holding leadership roles in engineering student organizations. Mentors were matched by major with mentees and received a half-day training focused on challenges in early engineering coursework and support strategies. A mentoring guide (topics: coursework advice, tutoring, planning for research/internships, social support, networking) was provided. Mentors were asked to meet monthly; actual median meetings=4 (about one hour each). Mentors received $100 per mentee; mentees received survey incentives. Procedures and measures: Surveys administered at baseline (pre-assignment) and mid- and end-of-first year; then annually each subsequent year through one year post-graduation. Transcripts collected for all participants at endline. Continuous self-reported outcomes measured longitudinally:

• Confidence in engineering ability: 2 items (e.g., talent for engineering; confidence in engineering ability), 1–7 scale; alpha ~0.78–0.92.
• Motivation in engineering courses: 5 items (e.g., skills, preparedness, efficacy for engineering classes), 1–7 scale; alpha ~0.79–0.95.
• Graduate aspirations in engineering: likelihood of pursuing MS/PhD in engineering, 1–7.
• Emotional well-being: single item of overall psychological health/happiness over past year, 1–7, measured from end of first year annually through post-graduation. Objective/dichotomous outcomes:
• Engineering internship participation: self-reported annually; aggregated as any (1) vs none (0) during college (treated dichotomously due to variable time-to-degree and survey response variation).
• Degree at graduation: coded from transcripts for engineering degree (0/1) and STEM degree (0/1) per U.S. Department of Education definition (biological, physical, computer sciences, engineering, mathematics/statistics, science technologies; excluding social/behavioral sciences). Data and analysis: Multilevel models (Mplus 8) with full information maximum likelihood (FIML) handled missing data. Time centered at baseline and scaled in years (months/12); random intercepts and slopes estimated and allowed to covary. Level-2 predictors: mentor condition using two dummies (female mentor vs others; male mentor vs others; no mentor as reference). Tested: (a) within-condition change over time; (b) differences in change trajectories between conditions. Unstandardized coefficients reported; two-tailed p-values. Post hoc Monte Carlo simulations (1,000) provided power estimates for significant effects. Dichotomous outcomes analyzed via chi-square tests of percentage differences between conditions. Multilevel mediation (2-1-2 for internships and STEM degree; 2-1-1 for graduate aspirations) tested whether change trajectories in confidence, motivation, or emotional well-being mediated effects of mentor condition on academic choices. GPA analyses examined mean and trajectory differences by condition.

• Confidence in engineering: Participants with female mentors maintained confidence across time (B=0.03, SE=0.04, p=0.559), whereas no-mentor (B=-0.13, SE=0.05, p=0.007, power=0.92) and male-mentor groups (B=-0.10, SE=0.05, p=0.025, power=0.77) declined. Change trajectory for female-mentor condition differed significantly from both other conditions (B=0.14, SE=0.06, p=0.010, power=0.94).
• Motivation in engineering courses: Female-mentor group showed no change (B=-0.01, SE=0.04, p=0.782), while no-mentor (B=-0.12, SE=0.04, p=0.006, power=0.92) and male-mentor (B=-0.14, SE=0.04, p=0.001, power=0.98) groups declined. Female-mentor trajectory differed from others (B=0.12, SE=0.05, p=0.017, power=0.91); female vs male mentor contrast: B=0.13, SE=0.06, p=0.024, power=0.74.
• Graduate aspirations in engineering: Declines observed for no-mentor (B=-0.50, SE=0.14, p<0.001, power>0.99) and male-mentor (B=-0.44, SE=0.13, p=0.001, power>0.99); female-mentor group showed attenuated, non-significant decline (B=-0.16, SE=0.11, p=0.159). Female-mentor trajectory differed from others (B=0.31, SE=0.15, p=0.036, power=0.90).
• Emotional well-being: Declined for no-mentor (B=-0.32, SE=0.17, p=0.061, power=0.70) and male-mentor (B=-0.42, SE=0.14, p=0.002, power=0.87) groups; held steady for female-mentor group (B=0.20, SE=0.14, p=0.147). Female-mentor vs others difference significant (B=0.57, SE=0.18, p=0.001, power=0.97); differed from no-mentor (B=0.52, SE=0.22, p=0.017, power=0.78) and male-mentor (B=0.62, SE=0.19, p=0.001, power=0.89).
• Engineering internships: Any internship during college reported by 82% (female-mentor) vs 61% (no mentor) vs 65% (male mentor). Female-mentor condition exceeded others (χ²(1, N=125)=4.79, p=0.029); exceeded no-mentor (χ²(1, N=82)=4.58, p=0.032); male vs others not significant.
• Degree attainment: Engineering degrees: 79% (female mentor) vs 71% (male mentor) vs 66% (no mentor); differences not significant (χ²(1, N=150)=1.85, p=0.174). STEM degrees: 92% (female mentor) vs 78% (male mentor) vs 81% (no mentor); significant by condition (χ²(1, N=150)=4.09, p=0.043). Female mentor > male mentor (χ²(1, N=103)=3.99, p=0.046); female mentor vs no mentor trend (χ²(1, N=99)=2.84, p=0.092); no mentor vs male mentor ns (χ²(1, N=98)=0.09, p=0.767).
• Mediation: Longitudinal change in confidence in engineering skills mediated the positive effect of having a female peer mentor on academic choices (securing engineering internships, graduating with a STEM degree, maintaining graduate aspirations). Across the full college span, confidence emerged as a stronger mediator than belonging or other subjective measures, indicating shifting psychological mechanisms over time.
• GPA: No differences in engineering or STEM GPA by mentor condition (ps ≥ 0.457); no significant GPA trajectory changes over time (ps ≥ 0.161) or by condition (ps ≥ 0.504).

Findings support the hypothesis that authentic mentoring relationships with own-group peers during the transition to college function as social vaccines, protecting female engineering students’ confidence, motivation, aspirations, and emotional well-being over multiple years. Despite the mentoring ending after the first year, benefits persisted through graduation and one year beyond, and extended to objective outcomes: higher likelihood of securing engineering internships and increased STEM degree completion. The engineering degree difference was not statistically significant at graduation, suggesting some mentees shifted into other STEM fields (e.g., CS, math, chemistry, biology), consistent with broader STEM persistence benefits. The preservation of emotional well-being may be a key protective factor against attrition. The results clarify mechanisms: while earlier short-term work highlighted belonging during the first year, the longer-term analyses show that sustained confidence in engineering skills better explains downstream academic choices. Male mentors did not produce harms but also did not match the benefits associated with female mentors, underscoring the importance of visible own-group representation in mentoring relationships. The durability and breadth of effects, alongside unchanged GPAs, imply that improving subjective experiences and professional development opportunities, rather than grades, drive long-term persistence and career-relevant outcomes.

This longitudinal randomized field experiment demonstrates that a low-cost, light-touch, first-year peer mentoring program pairing female engineering students with female peer mentors yields lasting benefits: maintained confidence and motivation, preserved emotional well-being, higher rates of engineering internships, and increased STEM degree completion, continuing up to one year post-graduation. The work advances understanding by providing causal, ecologically valid, long-term evidence for authentic own-group mentoring effects and by identifying confidence as a key mechanism over the college trajectory. Future directions include: (1) replication across other STEM disciplines where women are underrepresented (e.g., computer science, physics) and at later training/career stages (graduate, postdoc, early career); (2) testing generalizability to other underrepresented and negatively stereotyped groups in STEM (e.g., Black, Hispanic, Indigenous, first-generation/working-class students); (3) examining mentor-side outcomes and the role of mentor effectiveness; and (4) mapping how different psychological mechanisms (belonging vs confidence) operate at different developmental stages. Scaling research-driven mentoring solutions and tracking long-term impacts may be an important step toward improving representation in the STEM workforce.

• Generalizability: Single institution (large public U.S. university) and sample restricted to female engineering students; results may not generalize across institutions, disciplines, or demographic groups without replication.
• Measurement: Many outcomes were self-reported; emotional well-being was assessed with a single-item measure; internship participation was aggregated dichotomously due to variable time-to-degree and survey response patterns.
• Exposure dosage: Mentoring was relatively light-touch (median 4 one-hour meetings); variability in mentor-mentee contact and mentor effectiveness was not deeply modeled; mentor success might contribute to outcomes, and mentor-side outcomes were not assessed.
• Missing data/attrition: Although handled via FIML in multilevel models, not all participants completed every survey wave.
• Scope of outcomes: GPA and some objective academic measures did not change, suggesting effects were primarily on subjective experiences and professional development; the study did not include long-term labor market outcomes beyond one year post-graduation.
• Comparison mentors: Male mentors did not confer equivalent benefits; the study does not isolate which aspects of shared identity (e.g., role modeling vs network access vs psychosocial support) drove effects.