Abstract
Many recent studies emphasize how important the role of cognitive and social-emotional skills can be in determining people's quality of life. Although skills are of great importance in many aspects, in this paper we will focus our efforts to better understand the relationship between several types of skills with academic progress delay. Our dataset contains the same students in 2012 and 2017, and we consider that there was a academic progress delay for a specific student if he or she progressed less than expected in school grades. Our methodology primarily includes the use of a Bayesian logistic regression model and our results suggest that both cognitive and social-emotional skills may impact the conditional probability of falling behind in school, and the magnitude of the impact between the two types of skills can be comparable.
Introduction
Education research is increasingly shifting its focus from purely foundational literacies, such as mathematics and reading, to a broader set of "21st Century Skills." These skills encompass not only cognitive abilities but also competencies and character qualities that are essential for navigating complex challenges and changing environments.
While previous literature has extensively explored the impact of family background and income on school failure, there is a lack of granular analysis regarding specific social-emotional facets. This study addresses this gap by investigating how specific cognitive and social-emotional skills relate to academic progress delay—defined as progressing fewer than the expected number of grades over a five-year period—and comparing the relative magnitude of these impacts.
Methodology
The study utilizes a longitudinal dataset collected by LEPES/USP in Sertãozinho, Brazil. The sample consists of 1,780 students who were in the 5th and 6th grades in 2012 and were re-interviewed in 2017. The features analyzed include demographic data, cognitive scores (Language and Math) calculated via Item Response Theory, and ten social-emotional facets derived from the Big Five Inventory (Activity, Aesthetics, Altruism, Anxiety, Assertiveness, Compliance, Depression, Ideas, Order, and Self-Discipline). Notably, social-emotional scores were specifically corrected for acquiescence bias.
To model the likelihood of a student falling behind, we employed a Bayesian Logistic Regression framework. Inference was performed using the No-U-Turn Sampler (NUTS), a variation of Hamiltonian Monte Carlo, implemented via the rstanarm package in R. This Bayesian approach allows for the analysis of Highest Posterior Density (HPD) intervals, providing probabilistic bounds on the effects of specific traits rather than relying on brittle p-values.
Results
The Bayesian model demonstrated strong predictive performance, achieving an Area Under the Curve (AUC) of 0.74, outperforming a tuned Random Forest benchmark (AUC = 0.68). Descriptive analysis reveals significant differences in skill levels between students who fell behind and those who did not.
As shown in the table above, students who fell behind exhibited significantly lower mean scores in Language and Mathematics, and higher mean scores in Depression and Assertiveness. The regression analysis identified four primary predictors of academic delay:
- Cognitive Skills: Higher scores in Math and Language significantly reduce the odds of falling behind. A one standard deviation (SD) increase in Math reduces the conditional odds by 16% to 44%, while a similar increase in Language reduces the odds by 10% to 37% (with 95% probability).
- Social-Emotional Skills: Conversely, certain social-emotional traits are associated with an increased risk of delay. A 1 SD increase in Depression scores increases the conditional odds of falling behind by 17% to 56%. Similarly, higher Assertiveness scores increase the odds by 10% to 50%.
Crucially, the magnitude of the impact of these social-emotional traits is comparable to that of core cognitive skills, suggesting that character qualities are just as vital as foundational literacies in predicting academic success.
Discussion and Limitations
While the results provide robust evidence for the importance of social-emotional skills, several limitations must be noted. First, the study experienced a 34% attrition rate between 2012 and 2017. This dropout was non-random, skewing toward students with less-educated mothers and prior histories of school failure, which may affect the generalizability of the findings.
Furthermore, while the associations are strong, the current model does not fully disentangle causal pathways. It is possible that social-emotional skills indirectly affect academic progress by first impacting cognitive performance (e.g., depression leading to lower math scores). Future research should employ hierarchical modeling to explore these indirect effects and establish more rigorous causal links.
Summary updated