Examining the impact of COVID-19 on student mathematics achievement across countries: A multilevel machine learning analysis of PISA 2022

The COVID-19 pandemic caused significant disruptions to education globally. This study examines and compares the impact of COVID-19 on students’ mathematics achievement across eight countries/regions, namely, Australia, Chinese Taipei, Hong Kong (China), Japan, Korea, Macao (China), Singapore, and the United States, using data from PISA 2022. It uses multilevel machine learning models that explicitly account for hierarchical data structures and incorporate a comprehensive set of national-, school-, family-, and individual-level predictors. Results show that school closures had the strongest negative effect on mathematics achievement, whereas school support and students’ self-directed learning exerted positive influences. Substantial cross-country and subgroup differences were also observed. Education systems with shorter school closures and more effective remote instruction were better able to mitigate pandemic-related learning losses. Male students, students from higher SES backgrounds, and second-generation immigrants generally demonstrated higher mathematics performance. The study concludes with methodological and practical implications.