Abstract Students from socioeconomically disadvantaged backgrounds tend to show lower levels of mathematics engagement. However, some students demonstrate high mathematics engagement in the face of socioeconomic adversity. Understanding the factors that underpin such mathematics engagement resilience is crucial for advancing educational equity and quality. Drawing on the resilience framework of the Organisation for Economic Co‐operation and Development (OECD), this study investigates key personal, familial and school‐related predictors that distinguish engagement‐resilient students from their non‐resilient peers. Using the Programme for International Student Assessment (PISA) 2022 data from 80 economies, the analysis focuses on students in the bottom quartile of economic, social and cultural status ( N = 153,412). Three ensemble machine learning models (i.e., Random Forest, XGBoost and LightGBM) were employed to identify the most important predictors among 115 variables, followed by accumulated local effects analyses to examine relationships between key predictors and mathematics engagement resilience. The results indicate that perseverance, curiosity, mathematics self‐efficacy, lower mathematics anxiety and familiarity with mathematical concepts play central roles. Family support, teacher support, cognitively activating instruction, disciplinary climate and school size also emerge as important contextual factors. Accumulated local effects analyses further suggest pronounced non‐linear patterns for key predictors, such as threshold effects and diminishing returns. These findings extend resilience research beyond academic achievement to mathematics engagement and provide evidence to inform targeted interventions for disadvantaged students.
Wei et al. (Mon,) studied this question.