This study investigates the relationship between entrance exam scores and academic performance, as well as their implications for student retention, at the Jorge Basadre Grohmann National University (UNJBG). Using data from 1,526 students admitted in 2017 and tracked through 2024, the analysis explores entrance scores by subject area, average academic performance, and dropout rates across four study tracks. Results reveal that Verbal Reasoning (VR) and Language (LA) consistently exhibit moderate positive correlations with academic performance, while discipline-specific subjects such as Biology (BI) and Mathematical Reasoning (MR) show varying levels of influence depending on the study track. A strong negative relationship between entrance scores and dropout rates was identified, supported by Spearman correlations of -0.8 (average scores) and -0.77 (minimum scores). Based on these findings, a data-driven admission policy is proposed, introducing track-specific minimum score thresholds designed to reduce dropout rates while maintaining accessibility. The proposed threshold for Health and Biomedical Sciences would result in eliminate 24.29% of dropouts by reducing admissions by only 16.56%. The study underscores the value of aligning admission criteria with program-specific academic demands and highlights the potential of advanced data science techniques to refine predictive models and identify more nuanced patterns. These insights contribute to the development of evidence-based admission policies that optimize retention and support long-term academic success in the Peruvian higher education context.
29 September 2026 in Ingeniare. Revista chilena de ingeniería
Los exámenes de admisión universitaria se utilizan ampliamente para seleccionar candidatos a la educación superior, pero su validez predictiva sigue siendo objeto de debate, especialmente al considerar sesgos institucionales como las inconsistencias de calificación entre docentes. Este estudio evalúa en qué medida los puntajes por áreas de evaluación de los exámenes de admisión predicen el rendimiento académico del primer semestre en la Universidad Nacional Jorge Basadre Grohmann (UNJBG) en el Perú, incorporando un novedoso análisis visual de correlaciones que toma en cuenta el sesgo de calificación entre docentes. A partir de datos de tres cohortes consecutivas de ingreso (2017–2019) y más de 4,500 estudiantes en cuatro áreas académicas, se calcularon correlaciones de Spearman a nivel de sección entre las notas de admisión y las calificaciones de los cursos, lo que permite una visión más justa y controlada del desempeño. Los resultados muestran que ciertos dominios cognitivos, en particular Razonamiento Verbal y Razonamiento Matemático, presentan asociaciones consistentes y de moderadas a fuertes con el éxito académico temprano en las distintas áreas, mientras que otros evidencian relevancia específica según la trayectoria académica. Al visualizar estas relaciones y desagregarlas por área de estudio, el trabajo revela predictores matizados de la progresión académica y aporta evidencia empírica que respalda políticas de admisión más equitativas y sensibles a las disciplinas. Este enfoque contribuye tanto metodológica como prácticamente al campo de la evaluación educativa, resaltando el papel de la visualización de datos en la toma de decisiones institucionales.
University admission exams can be understood as information systems in which subject areas act as components intended to convey predictive signals about students’ future academic performance. However, the informational effectiveness of these subject areas is rarely evaluated using data-driven approaches. This study proposes a data-driven framework for assessing the informational effectiveness of admission exam subject areas by analyzing their empirical relationships with subsequent academic performance. Institutional data of 2197 students across 33 undergraduate programs from two cohorts after four semesters of study are used. Each academic program is represented as a vector of correlations linking performance in admission subject areas to long-term academic outcomes. The importance of each subject area in the admission exam is contrasted with empirically observed correlations to identify mismatches in informational effectiveness. Additionally, similarity analysis is applied to uncover affinities among academic programs. The results reveal substantial heterogeneity in the informational effectiveness of admission exam subject areas, indicating that predefined subject-area weightings do not consistently reflect their empirical contribution. Similarity patterns further identify groups of programs, suggesting opportunities for program-specific optimization of admission exam design. The proposed framework provides a replicable approach for evaluating and refining admission exams as information systems, contributing to data-driven decision-making in educational assessment design.