Structural Inequality and Exogenous Shocks: A Data Mining Analysis of Rural Secondary Education in Honduras (2015-2023)

Abstract

The COVID-19 pandemic constituted an exogenous shock of systemic magnitude that destabilized the bases of human capital accumulation in Latin America. This research addresses the urgent need to quantify educational erosion at the level of Secondary Education (Grades 10 to 12) in rural areas of Honduras, territories characterized by a structural digital divide. An administrative database was processed using a longitudinal panel design covering the period 2015-2023, an administrative database of the Ministry of Education (SEDUC) was processed with a non-probabilistic dynamic sample of more than 280,000 accumulated records. The methodology transcends descriptive statistics to implement machine learning algorithms. The findings are conclusive: while the multivariate linear regression showed severe insufficiencies in modeling the volatility of the crisis (R2 ≈ 0.60), the Decision Tree model with Gradient Boosting achieved exceptional predictive accuracy on final enrollment (R2 = 0.9825, MSE = 0.37). Additionally, the Chi-Cuadrado independence tests confirm that gender is a determining variable in post-pandemic dropout (p < 0.001), revealing that the economic crisis disproportionately affected female students.