PISA reading achievement: identifying predictors and examining model generalizability for multilingual students

Reading research in the United States has mainly focused on early or, less frequently, middle grades and on monolingual (MN or English-only) rather than on multilingual (ML) students. To address these gaps, we focused on factors contributing to high school ML students’ reading achievement. In particular, we first used machine learning to identify predictors of high school students’ reading achievement on PISA 2018. We then conducted multilevel modeling on the entire sample (baseline model) and tested the model’s generalizability to ML and MN populations. Results suggest that ML students would benefit from instruction focused on enhancing their reading self-efficacy and increased learning opportunities for extracurricular reading activities. The results also suggest that students, especially ML students, would benefit from schools avoiding grade retention policies and focusing on minimizing truancy and supporting positive peer and teacher relationships. Limitations of the study and future directions are discussed.