Predicting Student Performance Across Varying Family Structures: A Multigroup Analysis

• In Event: COVID-19 Experiences, Contexts, and Outcomes From The Nation’s Report Card

Abstract

The COVID-19 pandemic increased students' reliance on parental support for education when schools closed. Research on educational outcomes has primarily focused on socioeconomic status (SES), revealing that low-SES children tended to be more impacted (Moscoviz & Evans, 2022). Recent data from the Netherlands suggests that family composition was likely also a powerful factor in learning during the pandemic for children who attended school remotely (de Leeuw et al., 2023).

This study examines if students' background variables are associated differently with student performance across family structures, using the National Assessment of Educational Progress (NAEP) 2022 grade 8 mathematics assessment data. The analytic sample includes those students who indicated they learned remotely during the 2021-22 school year and who had no missing values in all the included variables. A multiple-group regression analysis by family structure was conducted to examine if the significance and magnitude of the regression coefficients for students' background variables would vary across different family structures. In essence, the approach took the interactions of family structure and all other predictors and compared the coefficients across family structures. The differences among background variables' coefficients across family types were examined using Wald tests.

The preliminary results indicated that the predictors were indeed associated differently with students' performance across family structures. We provide a few selected highlights based on the results presented in Table 1. When controlling for all other covariates, we found that:
• although the coefficients for SES were significant across all family structures, it was the largest in two-parent families, suggesting a similar level of increase in SES would be associated with greater benefits for student performance in two-parent families;
• the effect sizes of family remote-learning resources are larger in single-parent and foster families, suggesting that students in these families might benefit from more learning resources to make greater improvements;
• the Black-White gap (i.e., the coefficient of dummy-coded Black variable with White as reference group) is widest within step/blended families (20 for step vs. 13-18 for others); and
• the positive Asian-White gap was least pronounced within step families (8 for step vs. >14 for others), while it was about twice as high for other family structures.

Beyond showing that performance differences across family types persisted after controlling for other covariates, this study further revealed the differential predictabilities of the selected predictors across family types. It provides additional evidence of the importance of family structure as a factor in students' mathematics achievements post-COVID. Considering potentially more reliance on family support for younger students, the relationship might be stronger for grade 4 students. The analysis of the grade 4 data is underway, which results will also be presented in this session.

Authors

• Xiaying Zheng, American Institutes for Research

• Grace Ji Cheng, American Institutes for Research