This research full paper examines how socio-cognitive constructs relate to academic commitment in undergraduate engineering students, addressing the ongoing challenge of student retention and timely graduation. This study, focusing on Chile's largest engineering school, examines on-campus undergraduates across majors. It highlights academic commitment as key to persistence and progress. Using social-cognitive self-efficacy, expectancy–value theory, and belonging literature, we provide a reproducible workflow for institutional decision-making. A questionnaire measured self-efficacy, self-regulation, motivation, belonging, employment expectations, perceived value, and academic commitment (N ≈ 950). After distributional checks (skewness/kurtosis) for approximate normality, we conducted correlations and estimated hierarchical linear regression models, comparing them using ΔR², RMSE, and AIC/BIC. We performed diagnostics (tolerance/VIF, residuals, influence statistics) and reported robust BCa bootstrap confidence intervals, with sensitivity analyses excluding high-influence cases. This pipeline clarifies what to check, how to compare models, and how to document robustness when assumptions are only approximately met. The parsimonious model explained approximately 42–44% of the variance in academic commitment. Program-level belonging exhibited the most significant standardized effect (~.30), followed by employability expectations (~.21), self-efficacy (~.18), self-regulation (~.10), and motivation (~.09–.11). STEM-field belonging increased predictive capacity (ΔR² ≈ .02) yet showed a negative partial coefficient—consistent with a suppression pattern when distal identity is modeled alongside proximal program constructs. These findings suggest that students’ daily academic integration and perceived program fit are more influential for academic commitment than broad STEM identification, once motivation and efficacy are accounted for. They point to actionable levers for improving retention in complex engineering programs: strengthening program-level belonging through cohort structures, advising, and faculty–student connections, and enhancing employability signaling via industry projects and mentoring. The study also offers a clear, adaptable workflow for diagnostics, model comparison, multicollinearity checks, and bootstrap inference. Finally, it provides a foundation for future analyses, such as latent profile analysis to identify student subgroups and SEM to test pathways, advancing our understanding of how multiple factors shape academic engagement.
http://orcid.org/https://0000-0002-7248-4492
Universidad Andres Bello, Viña del Mar, Chile
[biography]
http://orcid.org/0000-0002-0383-0179
Universidad Andres Bello, Santiago, Chile
[biography]
http://orcid.org/0000-0001-6066-355X
Universidad Andres Bello, Chile; Tecnologico de Monterrey, Mexico
[biography]
http://orcid.org/https://0000-0001-5880-1124
Tecnologico de Monterrey, Monterrey, Mexico; Universidad Andres Bello, Santiago, Chile
[biography]
The full paper will be available to logged in and registered conference attendees once the conference starts on June 21, 2026, and to all visitors after the conference ends on June 24, 2026