Interplay of socioeconomic status, cognition, and school performance in the ABCD sample

The study examines how socioeconomic status (SES) and cognitive ability jointly relate to children’s school performance and whether cognitive ability moderates SES effects, with implications for academic resilience. Prior literature shows robust associations between SES and academic achievement across countries and time, as well as links between SES and children’s mental and physical health. Cognitive abilities are strong predictors of academic performance and are themselves associated with SES. The purpose here is to test if interindividual differences in cognition alter the SES–grades relationship at the individual level, using two follow-up timepoints from the ABCD cohort to assess stability over time. Understanding these interactions could inform strategies to foster academic resilience among socioeconomically disadvantaged children.

The paper reviews extensive evidence that SES correlates with academic achievement across contexts, with some variation across countries and over time. SES-related disparities appear early and extend to mental health and quality of life. Research highlights cognitive abilities (e.g., language, executive function, working memory) as mediators or correlates of SES–achievement links, with home cognitive stimulation playing a role. Some studies report SES effects diminish when intelligence is controlled, suggesting complexity and the importance of non-cognitive factors (e.g., personality, motivation, gender, ethnicity, school type). Intelligence emerges as a robust predictor of grades, but interactions between SES and cognition are less explored, and prior country-level work found no moderation by cognitive ability. Concepts of academic resilience and buoyancy are introduced, with protective factors including self-efficacy, planning, low anxiety, persistence, school belonging, supportive teachers/classrooms, and positive home literacy activities. Cross-national differences in resilience mechanisms and the impact of COVID-19 on risk and resilience are also noted.

Design and data source: Observational analyses using publicly available ABCD Study data (Annual Release 4.0). Two timepoints of self-reported grades were used: 2-year and 3-year follow-ups after baseline. Participants: N = 5001 children (mean age = 9.98 years, SD = 0.6; 53% boys). Inclusion required complete baseline neurocognitive assessment; children’s self-reported overall school performance at both follow-ups; baseline parental education, household size, household income; and neighborhood SES. Measures: - Cognitive performance: NIH Toolbox Cognition Battery (7 iPad-based tests in English covering attention, executive function, episodic and working memory, language, processing speed). Composite cognition score (uncorrected standard score due to narrow age range) from baseline used as global cognitive ability. - SES: Three indicators from baseline: • Parental education (0–21 scale approximating years of education; treated as continuous). • Income-to-needs ratio (ITN): median of reported income bracket divided by 2017 U.S. Federal Poverty Guideline for household size; continuous (1 = poverty line; <1 below poverty; >1 above). • Neighborhood deprivation: Area Deprivation Index (ADI) national percentiles; higher = more deprived; continuous. - Grades: Children’s self-reported overall school performance in the last year. Original 1 (A+) to 12 (Fail) recoded into five ordered categories: A, B, C, D, Fail. Parent reports differed slightly but with negligible effect sizes; only child self-reports used. Analysis: - Two separate ordinal logistic regression models (proportional odds) predicting grades at each timepoint from SES indicators (parental education, ITN, ADI). - Post hoc model including interactions of each SES indicator with timepoint to assess SES effects over time. - Two additional models (one per timepoint) adding composite cognition main effect and interactions (cognition×parental education, cognition×ITN, cognition×ADI) to test moderation by cognition. - Predictors were scaled prior to analysis. Brant tests supported the parallel regression assumption for all models. R version 4.2.2 used. Pairwise model comparisons assessed added value of cognition (p < 0.001 at both timepoints).

- SES and grades at each timepoint (without cognition): • 2-year follow-up (N = 5001): Parental education OR = 1.29 (95% CI 1.21–1.38, p < 0.001); ITN OR = 1.37 (1.26–1.49, p < 0.001); ADI OR = 0.91 (0.85–0.97, p < 0.01). Higher parental education and ITN associated with higher odds of better grades; lower ADI (less deprivation) associated with better grades. • 3-year follow-up: Parental education OR = 1.34 (1.26–1.43, p < 0.001); ITN OR = 1.36 (1.26–1.47, p < 0.001); ADI OR = 0.89 (0.84–0.95, p < 0.001). - SES effects over time: Timepoint main effect OR = 0.82 (0.76–0.89, p < 0.001), indicating lower odds of higher grades at 3-year vs 2-year follow-up. Interactions time×parental education, time×ITN, time×ADI were not significant. - SES, cognition, and grades (moderation models): • 2-year follow-up: Composite cognition OR = 1.72 (1.60–1.84, p < 0.001). Parental education OR = 1.22 (1.13–1.32, p < 0.001); ITN OR = 1.27 (1.17–1.39, p < 0.001); ADI not significant (OR = 0.94, 0.88–1.02, p = 0.12). Significant interaction cognition×parental education OR = 1.08 (1.00–1.16, p = 0.04); interactions with ITN and ADI not significant (ADI interaction marginal, p = 0.07). • 3-year follow-up: Composite cognition OR = 1.58 (1.48–1.69, p < 0.001). Parental education OR = 1.29 (1.20–1.69, p < 0.001); ITN OR = 1.27 (1.18–1.38, p < 0.001); ADI borderline/non-significant (OR = 0.93, 0.87–1.00, p = 0.05). Significant interaction cognition×parental education OR = 1.08 (1.01–1.16, p = 0.02); other interactions not significant. - Model comparisons showed improved fit when cognition was included (p < 0.001 at both timepoints). Overall: Parental education and ITN consistently predicted better grades; neighborhood deprivation’s effect attenuated when cognition was included. Cognitive ability interacted positively, but modestly, with parental education in predicting grades.

The study asked whether cognitive ability moderates the relationship between SES and school performance, potentially contributing to academic resilience. Consistent with prior research, parental education and income-to-needs ratio were robustly associated with higher odds of better self-reported grades at both follow-ups, while neighborhood deprivation showed smaller effects that diminished when cognition was modeled. Cognitive ability was strongly associated with grades and showed a small but significant interaction with parental education at both timepoints, suggesting that children with higher cognitive performance may benefit more from higher parental education. However, the moderation effect was small, indicating limited practical significance and implying that SES and cognition likely exert substantial independent influences with more complex interplays than captured by simple interactions. No SES×time interactions emerged, and grades were slightly lower at the later follow-up, suggesting that SES effects did not systematically strengthen over this short developmental window. Patterns align with a potential “Matthew effect,” where favorable SES co-occurs with higher cognition and better academic outcomes. These findings underscore persistent SES-related disparities in academic outcomes and suggest that while cognition contributes, it does not fully account for or eliminate SES–achievement associations.

In a large ABCD subsample with two follow-ups, parental education and income-to-needs ratio consistently predicted better self-reported grades, and neighborhood deprivation’s impact weakened when cognitive ability was considered. Cognitive ability had strong main effects and modestly interacted with parental education, offering limited evidence that cognition buffers SES disadvantages. The results highlight persistent SES influences on academic performance across timepoints and levels of cognitive ability. Future research should incorporate objective academic records, model school-level factors, follow trajectories across additional timepoints, and include more socioeconomically diverse samples to test generalizability and mechanisms. Targeted educational supports, especially for students with fewer home resources, remain warranted.

- Observational design precludes causal inference about SES, cognition, and grades. - Grades were self-reported by children; reliability is imperfect and may not align with objective records. - School identifiers were largely unavailable, preventing multilevel modeling to account for clustering and school-level grading practices/resources. - Sample is biased toward higher SES, metropolitan areas, and volunteer participants; low-SES families are underrepresented, limiting generalizability, especially to disadvantaged groups. - SES indicators are intercorrelated (multicollinearity), complicating estimation of unique effects, though unique contributions were still detected. - Cognition measured at baseline only; potential changes over time were not modeled; NIH Toolbox administered in English may have language-related biases for some participants.