The Great Smoky Mountains Casino-Dividend Natural Experiment

Summary

The Great Smoky Mountains Study of Youth (GSMS) is a longitudinal survey that recruited 1,420 children aged 9, 11, and 13 from 11 counties in rural western North Carolina and gave them annual psychiatric and socioeconomic assessments beginning in 1993. Children of the Eastern Band of Cherokee Indians were oversampled to make up 25% of the sample (350 American Indian, 1,070 non-Indian).[1][2]

Midway through data collection, the study became the setting for one of social science's best-known natural experiments. In 1996 the tribe opened a casino on its reservation and began distributing a portion of the profits — on an equal, per-capita basis to every enrolled adult tribal member, regardless of employment, income, or other characteristics — in payments made every six months. The average disbursement has been approximately $4,000 per person per year;[2][4] children's shares were held in trust until age 18 (or paid at 18 if the child had finished high school).[2] Because eligibility depended only on pre-existing tribal membership, the income change was exogenous to family characteristics — non-Indian households in the same counties received nothing — letting researchers read the effect of an unconditional cash dividend against a built-in control group. (A figure of "$6,000 by 2001" circulated in an earlier draft of this page; it could not be confirmed in either primary text and has been removed — see Access Level below.)

Three papers mine this experiment. Costello, Compton, Keeler & Angold (2003), in JAMA, examine children's mental health before and after the dividend. Akee, Copeland, Keeler, Angold & Costello (2010), in the American Economic Journal: Applied Economics, follow the same children into young adulthood to measure education and criminal outcomes. Akee, Simeonova, Costello & Copeland (2018), in the American Economic Review, use the same panel to measure effects on personality traits and psychiatric-symptom counts through age 16, and test mechanisms.

What the Studies Found

Costello et al. (2003) — poverty, income, and children's mental health. The dividend moved 14% of study families out of poverty, while 53% remained poor and 32% were never poor.[1] Before the casino opened, persistently-poor and just-lifted ("ex-poor") children had markedly more psychiatric symptoms (mean frequency scores of 4.38 and 4.28) than never-poor children (2.75). After the payments began, symptom levels among the ex-poor children fell to the level of the never-poor (odds ratio 1.50, 95% CI 1.08–2.09, for symptom reduction), while the persistently-poor stayed high (odds ratio 0.91, 95% CI 0.77–1.07 — no significant change).[1] Crucially, the effect was specific to symptoms of conduct and oppositional-defiant disorders; anxiety and depression symptoms were unaffected. The same pattern appeared among non-Indian children whose families happened to move out of poverty in the same period. The authors read this as support for a social-causation explanation (poverty causing disorder) for behavioral problems, but not for emotional ones — an income intervention that removed poverty "for reasons that cannot be ascribed to family characteristics had a major effect on some types of children's psychiatric disorders, but not on others."[1]

Akee et al. (2010) — long-run education and crime. Exploiting the fact that younger cohorts were exposed to the dividend for more of their childhood than older cohorts, the authors run a difference-in-differences design on young-adult outcomes. Their headline result:

"An additional $4,000 per year for the poorest households increases educational attainment by one year at age 21 and reduces having ever committed a minor crime by 22% at ages 16−17."[2]

Effects were concentrated among children from the initially poorest households; children from households already above the poverty line showed little change (the coefficient for years-of-education "triples in size" for the previously-poor subsample, to a full 1.1 additional years at age 21, and is not statistically significant for the never-poor subsample).[2] The authors find "improved parental quality is a likely mechanism" — for example, fathers in dividend-receiving households were roughly half as likely to be arrested in a given year, and mothers' and fathers' self-reported supervision of their children improved by a "moderate to large" 3–5%.[2] They also rule out that the results are driven by the casino's local labor-demand boost rather than the cash itself: parental employment did not jump, the region's labor supply is elastic, and controlling for household distance from the casino does not diminish the effect.[2] Almost the entire per-capita transfer (about $3,900 of the ~$4,000 disbursed) showed up as additional household income at each wave, with no offsetting reduction in labor-force participation.[2]

Akee, Simeonova, Costello & Copeland (2018) — personality traits and psychiatric symptoms through age 16. The same panel, followed to age 16, lets the authors test effects on the Big Five personality traits (using the subset the GSMS instrument can proxy: conscientiousness, agreeableness, neuroticism) alongside psychiatric-symptom counts, with individual fixed effects. Relative to the American-Indian sub-sample, the casino transfer reduced behavioral-disorder symptoms by 26.7% of a standard deviation and emotional-disorder symptoms by 35.6% of a standard deviation, and increased conscientiousness by 42.8% of a standard deviation and agreeableness by 30.6% (the effect on neuroticism was positive but not statistically significant).[4] As in the 2010 paper, gains were concentrated among children who started furthest behind their peers on these measures. The authors test — and rule out as full explanations — changes in parental employment, marital status, national welfare reform, and other tribal-government programs; they instead find evidence for improved parent-child and spousal relationships as a mechanism, some (weaker, only partly significant) evidence of reduced parental drug/alcohol use, and — a channel not in the earlier papers — that a sub-sample of previously off-reservation households moved to census tracts with higher median income and more education, suggesting neighborhood effects contribute alongside the within-household channels.[4]

Relation to the Georgist Case

The Cherokee dividend is a near-textbook instance of the mechanism Georgists propose: a commonly-owned asset generates a rent, and that rent is returned to every member of the community as an equal per-capita cash dividend — the same design as the Alaska Permanent Fund Dividend. It supplies rare causal evidence, from a credibly exogenous income shock, on two questions the Georgist dividend literature cares about:

1. Does an unconditional dividend reduce poverty's harms? Yes, measurably: lifting families out of poverty improved children's behavioral health (Costello 2003) and their adult education and law-abidingness (Akee 2010) — strengthening the rent dividends reduce poverty claim with a design cleaner than the descriptive Alaska evidence. The 2018 follow-up (Akee, Simeonova, Costello & Copeland) extends this to a broader outcome set — personality traits and psychiatric-symptom counts through age 16 — with the same directional result and effects concentrated among the initially worst-off children.
2. Does the cash discourage work? No detectable labor-supply withdrawal appears — consistent with Jones & Marinescu's Alaska finding, reinforcing the durability-and-workability case.

Rent-gradient caveat (honest scope). This is not land-rent evidence. Casino gaming profits are best classified as a monopoly / regulatory-privilege rent (tribal gaming enjoys a legally protected market position) mixed with ordinary business return — a more contested rent category than location rent. What the experiment isolates cleanly is the distributive mechanism — an equal, unconditional per-capita payment from a co-owned asset — not the pure economics of the underlying rent. Extending its poverty and child-development findings to a dividend funded by land-value taxation is a reasonable analogy about how unconditional dividends affect recipients, but the studies do not test a land-rent source. The magnitude is also modest (~$4,000/person/year), so the evidence speaks to a supplemental dividend, not a full basic income.

Nuances and Limits

• Single reservation, small sample of "treated" poor. The strongest effects rest on a few hundred American Indian children, and the poorest-household subgroups are smaller still; precision on subgroup magnitudes is limited.
• Selective outcome effects. Costello (2003) finds the dividend moved behavioral (conduct/ODD) symptoms but not emotional (anxiety/depression) ones; Akee et al. (2018), consistently, find bigger, clearer effects on behavioral/conscientiousness-type traits than on neuroticism. A nuance often lost when the study is cited as blanket proof that "money fixes mental health."
• Mechanism is indirect and only partly pinned down. Akee (2010, 2018) attributes gains chiefly to improved parenting, supervision, and parental/spousal relationships, not to the cash mechanically buying better outcomes — and the 2018 paper adds a neighborhood-quality channel (moving to higher-income census tracts) that the 2010 paper does not test. Results therefore depend on how a dividend changes family functioning and residential choices, which may vary across settings.
• Working-paper vs. published-version caveat. The Akee, Simeonova, Costello & Copeland personality-traits results on this page are drawn from the freely available NBER Working Paper 21562 (September 2015); the final peer-reviewed version appeared as American Economic Review 108(3), 2018: 775–827 (DOI 10.1257/aer.20160133). The working paper and the published article report the same experiment and are very likely materially consistent, but this page has not cross-checked every coefficient against the final AER typesetting — treat exact decimal magnitudes as sourced to the 2015 draft.
• A competing casino literature exists. Aggregate census studies (e.g., Evans & Kim 2006) have found casinos on reservations associated with higher dropout rates — but those cannot follow the same individuals over time and are confounded by in-migration; Akee's panel design, tracking the same children before and after, is the more credible identification for the per-capita-transfer question.[2]

Access Level (honest grading — do not cite from memory)

This page was rebuilt from a genuine, source-by-source access check (2026-07-18), after an earlier draft's "fetched and read" claims for the JAMA paper could not be reproduced:

• Costello et al. (2003), JAMA — paywalled. The publisher page (jamanetwork.com) sits behind a Cloudflare bot-check that returns no article content to an automated fetch, and Unpaywall reports no open-access location for the DOI (oa_status: closed); a ResearchGate mirror PDF also returned HTTP 403. What is freely and reliably accessible is the full structured abstract (Context/Objective/Design/Population/Results/Conclusions), retrieved directly from the Europe PMC and PubMed record for PMID 14559956 — this abstract is unusually detailed for a JAMA paper (it includes the numeric results and odds ratios quoted above) but is not the full paper (methods detail, tables, discussion nuance beyond the abstract are not verified here).
• Akee et al. (2010), AEJ: Applied Economics — full text freely available as the NIH-deposited author manuscript on PMC (PMC2891175, PMID 20582231), fetched and read directly (introduction through conclusion, footnotes, and table text).
• Akee, Simeonova, Costello & Copeland (2018), AER — the final journal version is paywalled (AEA), but the pre-publication NBER Working Paper 21562 (2015) — the same underlying study — is hosted free on nber.org and was fetched and read in full (results, mechanism sections, and conclusion).

Bears On

• Outcome: Rent dividends reduce poverty and inequality — supplies causal evidence that an unconditional per-capita dividend improved poverty-linked child outcomes.
• Outcome: Resource-rent dividends are workable and durable — an independent, long-running per-capita dividend from a co-owned asset with no visible labor-supply cost.
• Concept: Citizen's Dividend — a real-world unconditional dividend whose downstream human-capital effects have been measured.
• Study: Jones & Marinescu, Alaska Permanent Fund — companion quasi-experimental dividend evidence.
• Study: Akee, Simeonova, Costello & Copeland (2018) — the same natural experiment, extended to personality traits and psychiatric-symptom counts through age 16 (NBER WP 21562, mined in full on this page).

See Also

• Rent dividends reduce poverty and inequality
• Resource-rent dividends are workable and durable
• Alaska Permanent Fund
• Citizen's Dividend · Resource Rents
• Jones & Marinescu (2022)

Sources

1. E. Jane Costello, Scott N. Compton, Gordon Keeler & Adrian Angold (2003), "Relationships Between Poverty and Psychopathology: A Natural Experiment," JAMA, 290(15): 2023–2029. DOI 10.1001/jama.290.15.2023. Publisher page (paywalled/bot-gated): JAMA Network. Structured abstract fetched and read via Europe PMC / PubMed (PMID 14559956), 2026-07-18 — used for the 1,420-child GSMS sample and design, the 14%/53%/32% poverty transitions, the pre/post symptom-frequency scores (4.38 / 4.28 / 2.75) and odds ratios (1.50 [1.08–2.09] ex-poor; 0.91 [0.77–1.07] persistently poor), the ex-poor-fall-to-never-poor finding, the conduct/ODD-specific (not anxiety/depression) result, the same-pattern-in-non-Indian-families note, and the verbatim Conclusions quote. Full paper body (Methods detail, tables, Discussion) not accessed — see Access Level. The earlier "$6,000 by 2001" figure and a claim of having read the paper's full Results section could not be substantiated on this pass and have been removed/corrected.
2. Randall K. Q. Akee, William E. Copeland, Gordon Keeler, Adrian Angold & E. Jane Costello (2010), "Parents' Incomes and Children's Outcomes: A Quasi-Experiment Using Transfer Payments from Casino Profits," American Economic Journal: Applied Economics, 2(1): 86–115. DOI 10.1257/app.2.1.86. AEA; full text fetched and read via PMC (NIH author manuscript) (PMCID PMC2891175, PMID 20582231), 2026-07-18 — used for the Eastern Band of Cherokee per-capita distribution design (~$4,000/person/yr since 1996, $3,900 of it passing through to household income), the sample (1,420 children, 350 American Indian / 1,070 non-Indian, 11 counties), the difference-in-differences result (+1 year education at 21 for the poorest households, −22% minor crime at 16–17), the improved-parental-quality mechanism (fathers ~half as likely to be arrested, 3–5% supervision improvement), the no-labor-supply-response finding, and the ruling-out of the casino labor-demand channel (distance-to-casino robustness check).
3. E. Jane Costello, Adrian Angold, Barbara J. Burns, et al. (1996), "The Great Smoky Mountains Study of Youth: Goals, Design, Methods, and the Prevalence of DSM-III-R Disorders," Archives of General Psychiatry, 53(12): 1129–1136. DOI — used for the original GSMS survey methodology and cohort design; referenced, not independently re-fetched this pass.
4. Randall Akee, Emilia Simeonova, E. Jane Costello & William Copeland (2018), "How Does Household Income Affect Child Personality Traits and Behaviors?," American Economic Review, 108(3): 775–827. DOI 10.1257/aer.20160133 (paywalled at AEA). Free pre-publication version fetched and read in full: NBER Working Paper 21562 (September 2015), 2026-07-18 — used for the personality-trait and symptom effect sizes (behavioral disorders −26.7% SD, emotional disorders −35.6% SD, conscientiousness +42.8% SD, agreeableness +30.6% SD, neuroticism not significant), the concentration of effects among initially-worst-off children, the parental-relationship and reduced-drug/alcohol-use mechanism evidence, and the residential-mobility-to-higher-income-tracts finding. See the working-paper-vs-published caveat above.