The Heterogeneity and Conditional Effects of Achievement Drivers: A Logistic Regression Analysis Based on Mathematics Scores of Portuguese Middle School Students

Identifying students at risk of academic failure remains a central challenge for education systems. Existing research assumes linear and homogeneous effects of family capital, overlooking context-dependent resource transformation. This study integrates Conservation of Resources Theory (COR) with the Process–Person–Context–Time (PPCT) model to propose that maternal education functions as a conditional resource, with its impact contingent on transformation through proximal processes like family interactions. Using data from 399 Portuguese students, logistic regression and generalized random forests were employed. Logistic regression identifies mid-term performance as the strongest predictor, family relationship quality as protective, and attendance and failure history as risk factors. Generalized random forest estimation reveals a non-significant average association of maternal education, yet conditional estimates fluctuate from −0.14 to 0.47, a pattern tentatively consistent with gain and loss spirals. These findings offer preliminary evidence for resource conditionality, are consistent with a conditional-resource interpretation of family capital effects, and generate hypotheses for future testing.