MaxGhenis · Oct 24, 2026
diff --git a/analysis/LITERATURE.md b/analysis/LITERATURE.md
@@ -1,6 +1,7 @@
 # Literature review — verified anchors and the questions each field must answer
 
-Status: living document, started 2026-08-13. Discipline: a work enters
+Status: living document, started 2026-08-13; developments section
+added 2026-10-24. Discipline: a work enters
 the VERIFIED table only after its existence, venue, and claim were
 confirmed against a primary source this session (link recorded); the
 TO-VERIFY queues hold candidate works named from memory, which MUST NOT
@@ -52,6 +53,22 @@ treatment forces:
 | DOJ False Claims Act settlements over SNAP QC manipulation (2016–2021): Virginia, Wisconsin, Alaska, Texas ($15M+), Mississippi, Louisiana, Florida ($17.5M), Tennessee ($6.85M); Osnes Consulting ($751,571); $67M+ recovered | A USDA-OIG nationwide audit found third-party consultants advised states to diminish identified QC errors (2008–2013); states and the consultant settled FCA liability | The paper's "documented manipulation history" now cites named federal records rather than a gloss; the bunching design (design 4) has its motivating episode — discretion demonstrably responded to the measurement system. https://www.justice.gov/opa/pr/texas-health-and-human-services-commission-agrees-pay-over-15-million-resolve-false-claims · https://justice.gov/opa/pr/florida-department-children-and-families-agrees-pay-175-million-resolve-false-claims-act · https://www.justice.gov/usao-edwa/pr/consultant-agrees-pay-751571-settle-false-claims-act-liability-alleged-falsification |
 | CRS, *Errors and Fraud in the Supplemental Nutrition Assistance Program* (R45147) | The Congressional Research Service synthesis of the error-measurement and manipulation episode | The neutral secondary source tying the QC system, the OIG audit, and the sanction history together for a policy audience. https://www.everycrsreport.com/reports/R45147.html |
 
+## Developments since 2026-08-14 (verified 2026-10-24)
+
+Policy, state-law, and tool developments after the anchor table was
+built. Same discipline: each row was read at the primary source linked
+in it; claims seen only in secondary coverage sit in queue 9 below.
+
+| Work | Verified claim | Bears on |
+|---|---|---|
+| USDA, *Supplemental Nutrition Assistance Program: Payment Error Rates, Fiscal Year 2025* (dated 2026-06-24) | National rate 10.62% (overpayments 9.28%, underpayments 1.33%); QC tolerance threshold $57 | The simulator's locked FY2025 rates. All 53 `official_fy2025` values and `national.fy2025` in `app/public/data.json` match the table to two decimals (differential check run 2026-10-24). https://www.fns.usda.gov/sites/default/files/resource-files/snap-qcfy25-per.pdf |
+| S. 5526, *Agricultural Act of 2026* (Boozman) | Original measure reported by Senate Agriculture on 2026-09-24 without written report; placed on the Senate calendar under General Orders (Calendar No. 683). Congress.gov lists no later action as of 2026-10-24 | Status record: pending, not enacted. https://www.congress.gov/bill/119th-congress/senate-bill/5526/all-actions |
+| CRS, *Senate Committee Reports a Farm Bill That Would Delay Onset of States Sharing SNAP Benefit Costs* (IN12723, updated 2026-10-05) | S. 5526 would start benefit cost sharing in FY2029, keyed to the lower of the FY2026 or FY2027 rate; delay to FY2030 states at 13.33% or higher in both FY2025 and FY2026; and raise the top share from 15% to 20% from FY2031. The House-passed H.R. 7567 keeps current law. USDA has issued no regulations or guidance, and CRS's figure assumes cost sharing keys to the point estimate | The statute the simulator encodes could change before the first bill: the election pair moves from {FY2025, FY2026} to {FY2026, FY2027}, the delay test becomes a two-year conjunction, and a 20% tier enters. CRS's point-estimate assumption is the open implementation question the tier-lottery argument addresses. https://www.everycrsreport.com/reports/IN12723.html |
+| Bergh & Rosenbaum (CBPP), "New Data Underscore SNAP Cost Shift's Harm to Low-Income Families and State Budgets" (2026-07-23) | On FY2025 point estimates: 9 states and the Virgin Islands below 6%; 6 states at 5%; 16 at 10%; 13 and Guam at 15%; 6 states and DC delayed. About $9 billion shifts to states in FY2028, priced on annualized first-half FY2026 issuance. New Jersey fell from 14.33% to 6.86%, leaving the delay group for a 5% share | Independent cross-check: CBPP's delay roster (AK, DE, DC, GA, IL, NM, OR) and New York at 13.18% match the paper's. The simulator prices on FY2024 issuance (`analysis/INTERVENTIONS.md`) and reports expectations over tier odds, so reconcile issuance bases before setting CBPP's $9 billion beside the paper's summed expected bill. https://www.cbpp.org/sites/default/files/7-23-26fa.pdf |
+| Corinth (AEI COSM), "How Fixing SNAP's Payment Error Rate Formula Would Change State Cost Shares" (2026-08-19) | The rate divides error dollars by benefits paid, so an overpayment also enlarges the denominator. For a household eligible for $300, a $400 payment scores 0.25 and a $200 payment 0.50; dividing by correctly computed benefits scores both 0.33. Approximated from published FY2025 figures, every state's rate rises (0.04–5.09pp) and 16 states' FY2028 shares change, 7 of them by crossing 13.33% into the delay | A definitional sensitivity separate from sampling noise, with a perverse sign at the delay cliff: the stricter definition lowers FY2028 state bills by a net $2.70 billion. The worked example and the approximation (corrected rate = rate / (1 − over + under)) were checked by hand. The simulator encodes the statutory definition; a definitional lens is a candidate extension. https://cosm.aei.org/how-fixing-snaps-payment-error-rate-formula-would-change-state-cost-shares/ |
+| North Carolina S.L. 2026-41 (S.B. 257), §9J.14, approved 2026-07-07, creating G.S. 108A-52.2 | Passes any state benefit cost share to counties by withholding sales-tax distributions. Each county pays a flat 0.75% of its base-year sales-tax allocation plus a proportional assessment scaled by its "penalty factor", its portion of the statewide PER dollar amount. §9J.14(b) writes the FY2028 election into statute: FY2025 unless FY2026 yields a lesser obligation. Nickodem (UNC SOG, 2026-08-17) puts the FY2025 rate at 7.36% and the FY2028 share at about $150 million | The first verified sub-state pass-through, and a second measurement layer. The statute does not say how PER dollars are attributed to counties; if from QC cases, the repo's FY2024 NC file has 955 cases and 127 with errors across 100 counties, so the tier lottery recurs at the county level on one or two error cases each. The statutory election matches the simulator's minimum-obligation election semantics. https://www.ncleg.gov/EnactedLegislation/SessionLaws/PDF/2025-2026/SL2026-41.pdf · https://canons.sog.unc.edu/blog/2026/08/17/state-appropriations-act-shifts-potential-federal-snap-benefit-cost-sharing-to-north-carolina-counties/ |
+| McClurg, SNAP payment error rate tool (released 2026-06-24) | State liabilities from point estimates on FY2023 issuance, with timeliness and PAI correlations and CAPER profiles; no sampling uncertainty, election, or delay modeling | The nearest public tool. Its gaps mark the simulator's distinct contribution: tier odds under sampling and process noise, and the election and delay semantics. https://snap-per.joelmcclurg.ai |
+
 ## Field map and TO-VERIFY queues
 
 Nothing below may be cited until verified and promoted to the table.
@@ -81,7 +98,8 @@ Nothing below may be cited until verified and promoted to the table.
    Peeters–Widlak 2018 and 2023 (promoted). Queue: healthcare.gov
    retrospectives; GAO IT-modernization high-risk series; academic
    post-mortems of UHIP/Benefind if any exist beyond press.
-6. **Fiscal federalism and intergovernmental penalties.** Queue:
+6. **Fiscal federalism and intergovernmental penalties.** Verified:
+   North Carolina's county pass-through (G.S. 108A-52.2). Queue:
    Medicaid FMAP incentive literature; TANF penalty design; UI
    administrative-funding incentives — the cross-program frame for
    what error-rate-keyed cost shares do.
@@ -93,6 +111,17 @@ Nothing below may be cited until verified and promoted to the table.
    JEL 2021 survey; Abadie–Diamond–Hainmueller JASA 2010. Queue:
    inference refinements (conformal/placebo-based) if the RI/KY runs
    need them.
+9. **Implementation and state responses (opened 2026-10-24).**
+   Verified: the developments table above. Queue: FFIS's reported
+   roughly $9.4 billion annual cost-shift estimate (seen only in
+   secondary coverage); Illinois DHS's "up to $80 million" against the
+   Civic Federation's roughly $550 million on a 14.67% FY2025 rate that
+   qualifies for the delay; APHSA's survey of state root-cause work;
+   any USDA implementation guidance (none as of IN12723, 2026-10-05);
+   NC DHHS's county attribution method and its first report (due
+   2027-02-01); other states' pass-through or appropriation statutes;
+   Corinth's companion piece, "SNAP's New Cost Share Requirement Is a
+   Blunt but Necessary Tool".
 
 ## Rival predictions to state before fiscal 2027 outcomes
 
@@ -130,6 +159,16 @@ rests only on the anchor table above plus this repository's artifacts.
    adjudication-stage shifts (the wedge components) without matching
    movement in file-computable error.
 
+4. **Sub-state pass-through (G.S. 108A-52.2; added 2026-10-24).**
+   Where a state bills counties by attributed error dollars,
+   prediction 3's measurement-side pressure gains a county margin:
+   county agencies have a direct budget stake in how QC errors are
+   attributed and contested. Distinguishing observable: in North
+   Carolina, county QC practice and attribution disputes change after
+   the first assessments; the statute's annual DHHS report on
+   assessment amounts by county (due each February 1) supplies the
+   county series to read them against.
+
 Adjudication protocol: the fiscal 2026 pre-registration's committed
 nulls arbitrate 1; 2 and 3 are directional and need the FY2026-27
 files plus elections data — state them now, score them when the files