Predicting entrepreneurial competencies from cognitive and dispositional indicators: an elastic net regularization approach in Peruvian university students

Introduction Research on university entrepreneurship has concentrated on broad explanatory models of entrepreneurial intention, whereas entrepreneurial competencies have rarely been modeled as a predictable outcome. This study examined whether entrepreneurial competencies can be predicted from specific cognitive, dispositional and contextual indicators in Peruvian university students. Methods A quantitative, cross-sectional predictive design was applied to 98 Peruvian university students who completed a 28-item Likert-scale questionnaire measuring cognitive decision making, entrepreneurial mindset and entrepreneurial competencies, together with selected contextual variables. An Elastic Net regularized regression model was estimated with cross-validation and evaluated through out-of-sample performance, permutation importance and bootstrap confidence intervals. Results The model showed strong predictive accuracy in the test set ( R 2 = 0.797), retaining 16 relevant predictors while shrinking non-informative variables to zero. The strongest predictors belonged to the entrepreneurial mindset domain, especially anticipatory proactivity, self-efficacy to initiate ventures, opportunity seeking and perceived ability to overcome obstacles; cognitive decision-making indicators showed secondary contributions and demographic variables were negligible. The convergence between model coefficients, permutation importance and bootstrap intervals supported the robustness of the retained predictors. Supplementary psychometric diagnostics indicated inadequate fit for the proposed three-factor measurement structure and insufficient discriminant validity between entrepreneurial mindset and entrepreneurial competencies (HTMT = 0.913). Discussion The findings suggest that entrepreneurial competencies among Peruvian university students are mainly associated with action-oriented dispositions rather than demographic or curricular characteristics, and they support university interventions focused on proactivity, self-efficacy and opportunity orientation. The study introduces item-level predictive, non-causal modeling into entrepreneurship education research; given the measurement diagnostics, the results should be interpreted as exploratory prediction among closely related self-report measures rather than as evidence of fully distinct latent constructs.