Reducing SNAP Payment Errors at Lower Cost With AI: A Simulation Study
Organizations: Department of Applied Statistics, Social Science, and Humanities New York University
Abstract
New federal legislation will require states with high payment error rates in the Supplemental Nutrition Assistance Program (SNAP) to pay a share of benefit costs, potentially costing some states over a billion dollars per year. These costs create substantial pressure to improve payment accuracy. One strategy to improve payment accuracy is preauthorization review, in which staff double-check applications and correct mistakes before benefits are issued. But limited staff capacity makes it important to target these reviews effectively. One leading targeting approach uses statistical risk models to prioritize cases likely to contain errors. Another novel approach uses AI to flag specific potential discrepancies for reviewers to check. For states under pressure to reduce error rates, it is unclear which approach-or combination of approaches-would reduce payment error rates most effectively. We compare these strategies through simulations using public SNAP quality control data. Under our main assumptions, risk-based review alone costs about seven times as much as combining AI flags with risk scores. This cost advantage persists across a wide range of plausible assumptions. Our findings suggest that AI-assisted preauthorization review could be a key tool to help states avoid substantial cost-sharing obligations.
Figures & tables
Appendix figures & tables14 assets
Supplementary material from the paper’s appendix.
Appendix
| Assumed detectable | Assumed not detectable |
| 2 : Incomplete or incorrect information provided; agency not required to verify. 10 : Policy incorrectly applied. 12 : Reported information disregarded or not applied. 14 : Agency failed to follow up on inconsistent or incomplete information. 15 : Agency failed to follow up on impending changes. 16 : Agency failed to verify required information. 17 : Computer programming error. 18 : Data entry and/or coding error. 19 : Mass change (error due to problem with computer-generated mass change). 20 : Arithmetic computation error. 21 : Computer user error. 22 : Agency budgeted an incorrect amount. 23 : Agency failed to follow recertification procedure related to notices/forms. 24 : Agency failed to follow recertification procedure related to interviews. 25 : Agency failed to follow recertification procedure related to time frames. Missing cause : Assumed detectable. | 1 : Information not reported. 3 : Information withheld by client (case referred for Intentional Program Violation [IPV] investigation). 4 : Incorrect information provided by client (case referred for IPV investigation). 7 : Inaccurate information reported by collateral contact. 8 : Acted on incorrect federal computer match information not requiring verification (such variance is excluded from error determination but must be recorded). 26 : Change was not required to be reported by the client or acted upon by the State agency based on time frames and reporting requirements. 99 : Other. |
| Review category | QC elements of variance |
|---|---|
| Earned income | 311 : Wages and salaries; 314 : Other earned income; 321 : Earned income deductions. |
| Unearned income | 331 : RSDI benefits; 332 : Veterans’ benefits; 333 : SSI and/or State SSI supplement; 334 : Unemployment compensation; 335 : Workers’ compensation; 336 : Other government benefits; 342 : Contributions; 343 : Deemed income; 344 : TANF, PA, or GA; 345 : Educational grants/scholarships/loans; 346 : Other unearned income; 350 : Child support payments received from absent parent. |
| Shelter and utilities | 362 : Homeless shelter deduction; 363 : Shelter deduction; 364 : Standard utility allowance. |
| Household composition | 140 : Residency; 150 : Unit composition; 170 : Social Security number. |
| Self-employment | 312 : Self-employment. |
| Medical expenses | 365 : Medical expense deductions. |
AI-Assisted Decision Making
AI-assisted decision making studies how computational systems can inform human choices while accounting for context, uncertainty, and consequences.
See all 157 papers