Intergenerational effects of a casino-funded family transfer program on educational outcomes in an American Indian community

The study investigates whether a large, unconditional, casino-funded family cash transfer provided to members of a Southeastern American Indian Tribe during childhood has intergenerational effects on educational outcomes of their offspring. Motivated by extensive evidence that childhood poverty adversely affects cognitive and health outcomes and contributes to persistent intergenerational disadvantage in the United States, the authors test whether increased family resources in one generation translate into improved human capital in the next. Building on prior findings from the Great Smoky Mountains Study and related economic literature that early-life income gains yield long-term benefits and that impacts scale with duration of exposure, the research question centers on whether greater maternal exposure to cash transfers during her childhood is associated with improved third grade math and reading scores among her children.

Prior work documents strong links between parental income and children’s life chances, including education and health. Evidence from the Great Smoky Mountains Study (GSMS) and other quasi-experimental analyses shows that the introduction of casino per-capita payments to the Eastern Band of Cherokee Indians in the late 1990s improved recipients’ educational attainment, health, and financial well-being, with larger benefits for those exposed longer during childhood. Economic theory and empirical work emphasize high returns to early-life investments in human capital relative to later interventions. Third grade standardized test scores are established predictors of later educational outcomes such as high school graduation and college attendance, supporting their use as an early indicator of human capital. The literature also suggests that life-course decisions—such as maternal education, marriage, fertility timing, and health behaviors—may mediate the effect of increased resources on child outcomes. However, few studies directly assess whether benefits of childhood cash transfers persist into the next generation and affect offspring educational performance.

Design: Quasi-experimental difference-in-differences (DiD) with a time-varying treatment intensity. The treatment is a large, unconditional cash transfer financed by casino profits and distributed to enrolled tribal members beginning in 1996. American Indian (AI) race/ethnicity in Jackson, Swain, and Graham counties, North Carolina, is used as a proxy for tribal membership (treatment-eligible), while non-American Indian (non-AI) residents in the same counties serve as a comparison group exposed to the same regional secular changes but not to the transfers. Exposure: For mothers (G2), exposure is defined as the number of years of potential receipt of the cash transfer before age 18 (range 0–15). Mothers aged 18 or older in 1996 have exposure coded as 0. Duration functions as treatment intensity for AI mothers and as an age control among non-AI. Outcomes: Offspring (G3) third grade standardized end-of-grade math and reading Z-scores, standardized within each test year using statewide means and SDs to ensure comparability across time. Data sources and sample: Linked administrative records from the North Carolina Education Research Data Center (NCERDC) merging state education data with birth certificates from the NC Office of Vital Records. The analytic samples include children born in NC and enrolled in NC public schools with linked third grade scores (test years 2008–2017): math N=4,289; reading N=4,254. Mean AI third grade scores were about 0.39 SD below state means, whereas non-AI scores were slightly above the state mean. Statistical analysis: Generalized estimating equations (PROC GENMOD in SAS 9.4) with maximum likelihood estimators modeled continuous Z-scores. Core DiD specification regressed test scores on: (1) duration of maternal exposure before 18 (continuous), (2) AI status, and (3) their interaction (AI × duration), with controls for child age at testing (including quadratic and cubic terms) and sex. Models progressively added maternal and birth characteristics as potential pathways/mediators (maternal education, marital status, prenatal tobacco use, infant birth weight) and examined robustness. The DiD framework compares differences across maternal cohorts (younger vs older in 1996) between AI and non-AI, and tests the parallel trends assumption using pre-treatment interactions among mothers 18+ in 1996 (i.e., zero exposure). Robustness checks included: restricting exposure to 0–12 years; restricting maternal age at birth to 16–35; controlling for test year (continuous and indicators) to address secular trends and the Great Recession. Parallel trends tests did not reject the null in the pre-treatment period for either outcome.

• Main DiD effects: A positive and statistically significant association between maternal childhood exposure duration to the cash transfer and offspring third grade scores among AI families.
• Model 1 (baseline): Interaction AI × duration significant for reading (p=0.0014; 95% CI: 0.013, 0.055) and math (p=0.0055; 95% CI: 0.009, 0.050). Coefficients reported in tables: math 0.029 (p=0.006, 95% CI: 0.009, 0.050); reading 0.034 (p≈0.001, 95% CI: 0.013, 0.055).
• Model 2 (adds child age terms and sex): Interaction remains significant: math 0.025 (p=0.015; 95% CI: 0.005, 0.045); reading 0.028 (p=0.007; 95% CI: 0.008, 0.049).
• Model 3 (adds maternal and birth characteristics as potential pathways): Interaction attenuated by ~20% but remains significant: math 0.020 (p=0.045; 95% CI: 0.001, 0.040); reading 0.024 (p=0.018; 95% CI: 0.001, 0.040).
• Magnitude: Compared to AI mothers with no exposure, having an AI mother with 10 years of exposure predicts offspring gains of 0.25 SD in math and 0.28 SD in reading. These gains are comparable to impacts of $1000 per pupil per year early childhood education investments in NC and approximate half a school year of learning in effect-size scaling.
• Baseline disparity: AI students’ average third grade scores are substantially lower than non-AI students (≈−0.39 SD), but the exposure-related gains narrow this gap.
• Robustness: Results persist when limiting exposure range (0–12 years), restricting maternal age (16–35 years), and controlling for test-year trends (including Great Recession). Pre-treatment parallel trends tests do not reject the null for either outcome.
• Exploratory pathways: Maternal decisions/behaviors (higher education, marriage, delayed fertility, reduced prenatal tobacco use) predict higher child test scores and explain part, but not all, of the exposure effect.

The findings support the hypothesis that large, unconditional family cash transfers during a mother’s childhood are associated with improved educational outcomes for her children. The DiD design, leveraging variation in maternal exposure duration and a non-AI comparison group within the same counties, suggests that intergenerational benefits to human capital arise even amid broader secular declines in rural test scores. Reading gains exceed math gains modestly, consistent with literature that reading development is more influenced by non-school environments, which aligns with mechanisms whereby increased family resources improve home learning conditions and parental investments. The results indicate narrowing AI/non-AI disparities in third grade performance among children whose AI mothers had greater exposure, though the absolute AI/non-AI gap remains. Exploratory analyses suggest that maternal education, marriage, fertility timing, and prenatal health behaviors partially mediate effects, but do not fully account for them, implying additional pathways such as improved maternal/infant health, parenting quality, or community-level improvements. While other tribal and regional investments (healthcare centers, educational programs) occurred, the DiD approach helps account for secular changes common to AI and non-AI residents. Nevertheless, unmeasured confounding unique to AI families that varied with time cannot be ruled out entirely.

This study provides population-based evidence of intergenerational human capital benefits from a large, unconditional cash transfer, showing that longer maternal exposure during childhood predicts higher third grade math and reading scores in the next generation. The magnitude of effects is educationally meaningful and comparable to notable early childhood interventions. Policy implications include the potential for sustained, sizable cash supports in childhood to break cycles of disadvantage across generations. Future research should: (1) examine cohorts of AI mothers who were exposed from infancy or early childhood (age <5) once their children reach testing age; (2) replicate in other settings to assess external validity; (3) incorporate richer school- and family-level contextual data and broader outcomes (attendance, later test scores, social-emotional well-being); and (4) identify mechanisms linking childhood income supports to later parental choices and investments that influence offspring achievement.

• The AI/non-AI achievement gap remains large; exposure-related gains narrow but do not close it, reflecting the limits of cash transfers in overcoming multigenerational structural inequities.
• Data linkage excludes private school students and those who moved out of state; NCERDC match rates (>74% from 2008) imply incomplete coverage, potentially introducing selection.
• Analyses are limited to public school third grade outcomes through 2017; many AI mothers with earliest-life exposure (infancy to age 5 in 1996) had children not yet observed, possibly underestimating effects.
• Missing/unknown paternity in birth records precluded assessing paternal AI status or dual-parent AI exposure.
• The design cannot disentangle effects of age at initiation versus total duration of exposure to transfers.
• Potential unmeasured confounding unique to AI families that varies over time cannot be definitively ruled out, though pre-trend tests and robustness checks mitigate concerns.
• Community-level improvements (healthcare, education, employment) coincided with casino opening and could contribute to outcomes; DiD controls for general secular trends but may not capture AI-specific time-varying factors.
• Outcome focus is limited to third grade math and reading; other domains of academic and socio-emotional development were not analyzed.
• Some prior work notes adverse outcomes temporally associated with large payment months, underscoring the need to weigh benefits and potential costs when considering transfer design (amount, frequency, in-kind vs cash).