Exploring Predictors of Reading Achievement in Macao's Primary Students via Social-Ecological Theory: Machine Learning

Recent studies in Macao have focused primarily on the international assessments of higher-grade students, with relatively little attention given to the performance of younger learners in primary school. This research seeks to fill that gap by examining data from 4,059 fourth-grade students who participated in the Progress in International Reading Literacy Study (PIRLS) and identifying the key factors influencing their reading achievements. On the basis of Bronfenbrenner's (1979) social-ecological theory, we explore the roles of individual-, family-, classroom-, and school-level variables in shaping students’ reading achievements. Employing the random forest algorithm along with a tenfold cross-validation technique, we assessed the relative importance of 18 distinct predictors. Our findings reveal that family-related factors, particularly parents’ attitudes toward reading, are significant predictors of reading performance. Further validation through additional analyses strengthens the robustness of these results and highlights the effectiveness of the machine learning approach used. Building on these insights, this study offers a set of policy recommendations aimed at enhancing the primary school educational environment.