Forms of Family Cultural Capital and Science Achievement: A Theory-Guided Machine-Learning Comparison of China (B-S-J-Z) and U.S. Student Samples in PISA 2025

Family cultural capital includes household educational resources, parental qualifications, and everyday family practices that may be relevant to students’ science learning. Using PISA 2025 data, we examined how these dimensions predicted science achievement among 12,258 students in China (B-S-J-Z: Beijing, Shanghai, Jiangsu, and Zhejiang) and 2,291 U.S. questionnaire respondents. We used the same twelve family cultural-capital and science-participation indicators in both samples and evaluated two predictive models—ridge regression and XGBoost—through school-grouped cross-validation and ten science plausible values. The full XGBoost models achieved held-out R2 values of .361 in China (B-S-J-Z) and .327 in the U.S. sample. Perceived parental educational expectations and books at home ranked first and second among the focal indicators in both settings. The additional predictive information associated with perceived expectations was greater in the China (B-S-J-Z) model, whereas family conversations and science-related participation added relatively more information to the U.S. model. Parental qualifications were also more prominent in the China (B-S-J-Z) results. Together, the findings identify both shared cultural-capital patterns and different profiles of family educational resources and practices across the two educational settings.