State-level macro-economic factors moderate the association of low income with brain structure and mental health in U.S. children

Children raised in lower-income families exhibit disparities in education, physical and mental health, and differences in brain development, notably smaller hippocampal volume. Low income is linked to greater exposure to acute and chronic stressors (e.g., violence, food insecurity) that may affect neurodevelopment. Macrostructural factors such as state cost of living (COL) and the generosity of anti-poverty policies may shape the strength of associations between family income and neural/mental health outcomes. Because cost of living varies across states and anti-poverty benefits (e.g., EITC, TANF, Medicaid expansion) differ in generosity and implementation, these factors may either magnify (high COL) or mitigate (generous benefits) the effects of low income. Leveraging the multi-site ABCD study across 21 sites in 17 states, the authors test whether state-level COL and anti-poverty policy generosity moderate associations between family income and (a) hippocampal volume and (b) internalizing/externalizing problems in 9–11-year-olds. The authors hypothesized that lower income would be associated with smaller hippocampal volume and more psychopathology; that these associations would be stronger in high COL states; and that they would be weaker in states with more generous anti-poverty programs, particularly in high COL contexts. The study also evaluates whether moderation is specific to income levels eligible for benefits and assesses robustness to numerous alternative state-level social, economic, political, and educational characteristics.

Prior work consistently links lower family income with smaller hippocampal volume in children and adolescents and with higher internalizing and externalizing problems. Mechanistically, increased exposure to stressors in low-income contexts (e.g., neighborhood violence, caregiver separation) is associated with reduced hippocampal volume; animal models show chronic stress reduces hippocampal dendritic branching and neuron viability. Stress exposure mediates income–hippocampus associations. Macro-level context may shape these links: cost of living varies widely across states, potentially magnifying financial strain at a given nominal income. Generous anti-poverty programs (EITC, TANF, Medicaid expansion) improve various child and family outcomes and may reduce psychological distress among parents; program generosity varies by state. Evidence from Europe shows weaker income–brain/cognition associations than in the U.S., possibly reflecting macro-social/policy differences. A recent RCT providing unconditional cash payments to low-income families showed increased infant EEG high-frequency power, suggesting income supports can positively influence neurodevelopment. These findings motivate testing whether state policy generosity and COL moderate income–brain and income–psychopathology associations.

Design and data source: Cross-sectional observational analysis using ABCD Study curated release 3.0 baseline data. Participants were 9–11 years old (mean 9.91, SD 0.622), from 21 sites across 17 U.S. states. Sample sizes: 11,864 with parent-reported psychopathology; 11,533 with structural MRI after quality control; analyses restricted to those with reported family income yielded N=9,913 (hippocampal volume outcome) and N=10,633 (mental health outcomes). Informed consent/assent obtained; protocols approved by UC San Diego central IRB and/or local IRBs. Exposures and outcomes: - Family income: Parents reported income in 10 brackets; midpoints were used. Income-to-needs ratio computed using 2017 U.S. Census poverty thresholds adjusted for household size; natural log of income-to-needs used in analyses. - Hippocampal volume: Left and right hippocampal volumes from FreeSurfer 5.3 automatic segmentation were summed. Total intracranial volume (ICV) computed and included as covariate. Structural MRI data quality followed ABCD exclusion criteria; 471 participants excluded; exclusion rates varied by site but were unrelated to COL or policy generosity. - Psychopathology: Parent-reported CBCL internalizing and externalizing T-scores at baseline. State-level moderators: - Cost of living (COL): 2017 Regional Price Parity (RPP) from U.S. BEA; indexed relative cost level by state. - Cash assistance generosity: Mean of (i) average monthly EITC in each state, computed as $266 × (1 + state EITC proportion of federal EITC) and (ii) average monthly TANF benefit; reflects typical monthly cash support for low-income families (ranges approximately EITC $266–$399; TANF $286–$789). - Medicaid expansion status: Dichotomous indicator of Affordable Care Act Medicaid expansion adoption by end of 2017. Covariates: Age, sex, MRI scanner type (for hippocampal models), ICV (for hippocampal models), and proportions of White, Black, and Latinx participants at each site (site-level demographics). Statistical analyses: Linear mixed-effects models (R; lme4 and nlme reported across sections) with two-tailed tests. Random intercepts for study site in all models; random slopes for income-to-needs at site level included to test improvement in model fit (significantly improved at state level for hippocampal volume and internalizing). Fixed effects included income-to-needs, moderator(s), all two-way interactions among income, COL, and the policy moderator (cash assistance or Medicaid expansion), and the three-way interaction income × COL × moderator. For hippocampal models, ICV and scanner were included. COL and cash assistance were grand-mean centered. Assumptions (normality via Shapiro–Wilk, homoscedasticity, independence) were checked and met. Supplemental models examined two-way interactions without three-way terms. Sensitivity analyses adjusted separately for 10 alternative state-level characteristics (population density, inequality, unemployment, tightness/looseness, political preferences, women’s political participation, reproductive rights, incarceration rate, state-funded preschool enrollment, reading proficiency among low-income students). Additional sensitivity analyses: (a) dichotomized family income at federal poverty line vs ≥ poverty line and at ≥5× poverty line, (b) inclusion of individual-level race/ethnicity covariates, (c) limiting to one site per state to address multi-site states, and (d) alternative moderator: state minimum wage.

- Main effects: Higher family income (log income-to-needs) was associated with larger hippocampal volume controlling for ICV, age, and sex: t(9,888)=7.78, p<0.001, B=62.71 mm³ (95% CI 46.91–78.52), β=0.079. Family income was negatively associated with internalizing problems: t(10,605)=-3.97, p<0.001, B=-0.59 (95% CI -0.88 to -0.030), β=-0.058; and with externalizing problems: t(10,605)=-6.46, p<0.001, B=-1.50 (95% CI -1.96 to -1.05), β=-0.152. - Moderation by cost of living and cash assistance (hippocampal volume): Significant three-way interaction income × COL × cash assistance: t(9,883)=-3.14, p=0.002, B=-3.64. In high COL states (≈1 SD above mean RPP ~1.07), low-income children in states with high cash benefits (~$531/month; 1 SD above mean) had hippocampal volumes on average 60 mm³ larger than those in high COL states with low cash benefits (~$317/month; 1 SD below mean). Income disparities in hippocampal volume were about 195 mm³ in high COL/low cash states vs 129 mm³ in high COL/high cash states, a 34% reduction. The COL × cash assistance interaction was significant only at low income (≈1 SD below mean; ~80% of poverty line), consistent with benefit eligibility. - Moderation by cost of living and Medicaid expansion (hippocampal volume): Significant three-way interaction income × COL × Medicaid expansion: t(9,883)=-3.31, p=0.001, B=-983. In high COL states, income-related disparities in hippocampal volume were 43% smaller in states that expanded Medicaid compared to those that did not. - Psychopathology (internalizing): Significant three-way interactions for internalizing problems with both cash assistance (t(10,602)=2.67, p=0.008, B=0.0615) and Medicaid expansion (t(10,602)=2.43, p=0.015, B=14.9). In high COL states, the internalizing T-score gap between low- and high-income participants was 2.11 in low cash benefit states vs 1.08 in high cash benefit states (≈48% reduction). Patterns similarly indicated weaker income–internalizing associations in high COL states with Medicaid expansion. - Externalizing problems: No significant three-way interactions with COL and anti-poverty program generosity. - Robustness and specificity: Findings were robust to controlling for 10 alternative state-level characteristics and their interactions with income. Effects were observed when income was dichotomized at the federal poverty line (benefit-eligible) but not when dichotomized at ≥5× poverty line (non-eligible). Results held when restricting to one site per state. Supplemental analyses using state minimum wage as an alternative policy indicator reproduced similar three-way interaction patterns for hippocampal volume and internalizing (and also for externalizing in supplemental results).

The study demonstrates that the strength of the association between family income and children’s hippocampal volume and internalizing symptoms varies systematically with macrostructural state-level characteristics. High cost of living exacerbates the negative impacts of low income, whereas more generous anti-poverty policies—cash assistance and Medicaid expansion—attenuate these associations, particularly among low-income children eligible for benefits. In high COL states with generous policies, income–hippocampus and income–internalizing associations resembled those observed in low COL states, indicating that structural supports can buffer children against the neurodevelopmental and mental health risks associated with economic hardship. These results were robust to numerous potential confounders reflecting social, economic, political, and educational contexts, strengthening the interpretation that social safety net generosity is a relevant contextual factor. While the observational design precludes causal inference, the findings align with emerging causal evidence from cash transfer interventions and suggest that policy-level interventions may reduce socioeconomic disparities in neural and mental health outcomes. The work underscores the importance of incorporating macro-social context in psychological and neurodevelopmental research, which often focuses solely on individual-level associations within single communities.

Lower family income is linked to smaller hippocampal volume and higher internalizing symptoms in early adolescence. These associations are moderated by state-level cost of living and the generosity of anti-poverty programs: in high cost of living states, more generous cash benefits and Medicaid expansion substantially reduce income-related disparities in hippocampal volume (by ~34%) and internalizing problems (by ~48%). In such states, associations resemble those in low cost of living contexts. The findings highlight macrostructural policy environments as relevant levers for mitigating socioeconomic disparities in children’s brain development and mental health. Future research should leverage quasi-experimental designs and longitudinal policy changes to strengthen causal inference, expand analyses as the ABCD cohort ages and policies evolve, examine additional policy domains (e.g., early childhood education, SNAP, child tax credit), and consider finer-grained measures of local cost of living.

- Cross-sectional, observational design limits causal inference; policy changes did not occur during ABCD data collection to enable within-state natural experiments. - ABCD sites span 17 of 50 states, reducing variability and power; cost of living and policy generosity are correlated within these states, limiting disentanglement. - Potential confounding with scanner differences and site-level demographics; models controlled for scanner and site-level racial/ethnic composition, but residual confounding is possible. - Use of state-level cost of living ignores within-state variation; site metropolitan COL generally tracked state COL (r=0.73), but mismatch is possible. - Parent-reported income may include cash benefits, potentially overlapping with moderators; results were consistent using a poverty dichotomy. - Age range limited to 9–11 years; generalizability to other developmental stages is uncertain. - Random effects for family (siblings) could not be modeled due to convergence; analyses restricting to one sibling per family yielded similar results.