thesisPrototype

Government data

US SNAP participation, September 2026

USDA FNS SNAP Data Tables national monthly Persons for September 2026, not seasonally adjusted, first print, converted to millions

Forecast

34.6M
34M–35.2M
80% prediction interval
gpt-5.5Run record ↗
33363942October 2025May 2026Feb 202734.6M
historyforecast path80% interval
Probability distribution

Cumulative probability

Chance that the outcome is at or below each value.

0%25%50%75%100%33.1M34.6M36.1MForecast value

Shaded band: 80% interval (34M–35.2M). Dashed line: point estimate (34.6M).

Derived from the point estimate and 80% interval; the agent did not report a full distribution.

Analysis

§September 2026 SNAP persons forecast

Framing and exact resolver: this targets the USDA FNS SNAP national monthly Persons field for September 2026, first print only, from the SNAP Data Tables. The current FNS PDF displays Persons as whole-person counts, for example 36,562,128 for May 2026; I convert those counts to millions for this cell while preserving the registered target unit and dataPointId.

▸ Reported tool use: official.lookupmodel report
Opened the current FNS SNAP Data Tables page for source identity and latest available monthly table.
↳ The official SNAP Data Tables page says the FY 23 through FY 26 National View Summary has Latest Available Month: May 2026, and the page was updated August 20, 2026.
▸ Reported tool use: official.lookupmodel report
Opened the August 2026 SNAP national monthly participation PDF and read FY2026 monthly Persons values.
↳ The PDF is Data as of August 14, 2026 and reports FY2026 Persons of 41,096,660 in Oct 2025, 39,999,890 in Nov 2025, 39,207,898 in Dec 2025, and 38,552,133 in Jan 2026.
▸ Reported tool use: official.lookupmodel report
Read the latest FY2026 monthly Persons values from the same SNAP national monthly participation PDF.
↳ The same PDF reports Persons of 37,947,123 in Feb 2026, 37,531,152 in Mar 2026, 37,078,596 in Apr 2026, and 36,562,128 in May 2026.
▸ Reported tool use: official.calendarmodel report
Checked official FNS data-and-research publication listings and the registered bounded target schedule for release timing.
↳ The official FNS listing shows May 2026 Keydata Report dated 08/10/2026, May 2026 Performance Report dated 08/12/2026, April 2026 Keydata Report dated 07/15/2026, and March 2026 reports dated 06/18/2026; 2027-02-15 is the registered resolve-by-bound date at the end of the September 2026 target's expected release window, not an FNS-announced exact release date.

The outside-view base rate/reference class is a short monthly persistence model on the same official FNS national Persons variant. The latest level is 36.562128 million in May 2026, with seven observed month-to-month changes from October 2025 through May 2026: -1.096770, -0.791992, -0.655765, -0.605010, -0.415971, -0.452556, and -0.516468 million.

Level effect starts at 36.562128 million. Momentum effect uses the more recent four-change average from Jan-May 2026, -0.497501 million per month, rather than the full seven-change mean of -0.647790 because the earliest FY2026 drops look unusually front-loaded. One-off/policy mechanism is set to 0.000 million because the draft did not establish a concrete public policy adjustment separate from the observed trend.

Prior/update/interval: persistence prior = latest official May 2026 level of 36.562128 million; historical sample = Oct 2025-May 2026 same-table Persons values listed above. Successive changes are -1.096770, -0.791992, -0.655765, -0.605010, -0.415971, -0.452556, and -0.516468 million; their sample standard deviation gives sigma = 0.235716 million. Point = 36.562128 + 4*(-0.497501) + 0.000 = 34.572124 million, rounded to 34.572. The one-month 80% half-width is 1.28*sigma = 1.28*0.235716 = 0.301717 million; for a four-month horizon I scale by sqrt(4), giving 0.603434 million. Bounds are 34.572124 - 0.603434 = 33.968690 and 34.572124 + 0.603434 = 35.175558, rounded to 33.969 and 35.176. The interval reflects recent monthly-change dispersion scaled over the four-month horizon and does not add a separate policy-adjustment term.

Upside risk is that the early FY2026 enrollment drop stabilizes quickly, state backlogs or recertifications raise the September count, or disaster-related reporting adds participants; that would land above the interval if September exceeds 35.176 million. Downside risk is faster eligibility loss, stronger work-requirement implementation, or administrative churn continuing near the Oct-Jan decline pace; that would land below the interval if September is under 33.969 million. The outside the interval cases are therefore a clear stabilization above 35.176 or an accelerated drop below 33.969.

Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

Key drivers

  • latest official May 2026 level was 36.562 million persons
  • recent monthly declines averaged about 0.498 million over January-May 2026
  • large FY2026 downtrend dominates short-horizon persistence
  • forecast point is driven by same-series persistence rather than an extra policy adjustment
  • first-print target uses unrevised FNS monthly national Persons table

Tool evidence

This run has no captured tool responses. “Reported tool use” in the analysis is the model’s account, not an archived tool response.

Sources and resolution

Official source
FNS SNAP data tables, national monthly participation (thousands)
Resolution date
February 15, 2027· outcome not recorded
Resolution rule
Resolve to the first value published by USDA FNS for September 2026 in the SNAP Data Tables national monthly Persons field in the Persons, Households, Benefits, and Average Monthly Benefit per Person and Household table. Express the published Persons count in millions, rounded to three decimals for display. Use the first official print only and ignore later revisions, backfills, same-day corrections, or subsequent workbook/PDF updates.
Run details

The analysis is the model’s written report. Tool-use descriptions in that report are model claims; the activity artifacts contain the execution record.

thesis.analyst · gpt-5.5 · fast · v2.5.11

pre-submit review · completed

The forecast is mostly coherent, but it needs source URL cleanup and better support for the policy-tightening adjustment before publication.

  • warning resolver: The main resolutionSourceUrl is correct, but sourceContext and reasoning repeatedly cite `fna.usda.gov` instead of the registered FNS host `www.fns.usda.gov`.
  • warning update: The -0.150 million policy/administrative tightening adjustment is not tied to a concrete cited policy, administrative change, or observed evidence beyond the recent trend.
  • info optional_suggestion: State explicitly that 2027-02-15 is the registered resolve-by-bound date at the end of the expected release window, not an official FNS announced release date.
  • info optional_suggestion: Clarify whether the interval only reflects monthly residual volatility, since it does not separately widen for uncertainty in the trend choice or policy adjustment.

disposition accepted: Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

disposition accepted: Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

disposition not applicable: Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

disposition not applicable: Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

Activity artifacts

Complete original trace

§September 2026 SNAP persons forecast

Framing and exact resolver: this targets the USDA FNS SNAP national monthly Persons field for September 2026, first print only, from the SNAP Data Tables. The current FNS PDF displays Persons as whole-person counts, for example 36,562,128 for May 2026; I convert those counts to millions for this cell while preserving the registered target unit and dataPointId.

▸ Reported tool use: official.lookupmodel report
Opened the current FNS SNAP Data Tables page for source identity and latest available monthly table.
↳ The official SNAP Data Tables page says the FY 23 through FY 26 National View Summary has Latest Available Month: May 2026, and the page was updated August 20, 2026.
▸ Reported tool use: official.lookupmodel report
Opened the August 2026 SNAP national monthly participation PDF and read FY2026 monthly Persons values.
↳ The PDF is Data as of August 14, 2026 and reports FY2026 Persons of 41,096,660 in Oct 2025, 39,999,890 in Nov 2025, 39,207,898 in Dec 2025, and 38,552,133 in Jan 2026.
▸ Reported tool use: official.lookupmodel report
Read the latest FY2026 monthly Persons values from the same SNAP national monthly participation PDF.
↳ The same PDF reports Persons of 37,947,123 in Feb 2026, 37,531,152 in Mar 2026, 37,078,596 in Apr 2026, and 36,562,128 in May 2026.
▸ Reported tool use: official.calendarmodel report
Checked official FNS data-and-research publication listings and the registered bounded target schedule for release timing.
↳ The official FNS listing shows May 2026 Keydata Report dated 08/10/2026, May 2026 Performance Report dated 08/12/2026, April 2026 Keydata Report dated 07/15/2026, and March 2026 reports dated 06/18/2026; 2027-02-15 is the registered resolve-by-bound date at the end of the September 2026 target's expected release window, not an FNS-announced exact release date.

The outside-view base rate/reference class is a short monthly persistence model on the same official FNS national Persons variant. The latest level is 36.562128 million in May 2026, with seven observed month-to-month changes from October 2025 through May 2026: -1.096770, -0.791992, -0.655765, -0.605010, -0.415971, -0.452556, and -0.516468 million.

Level effect starts at 36.562128 million. Momentum effect uses the more recent four-change average from Jan-May 2026, -0.497501 million per month, rather than the full seven-change mean of -0.647790 because the earliest FY2026 drops look unusually front-loaded. One-off/policy mechanism is set to 0.000 million because the draft did not establish a concrete public policy adjustment separate from the observed trend.

Prior/update/interval: persistence prior = latest official May 2026 level of 36.562128 million; historical sample = Oct 2025-May 2026 same-table Persons values listed above. Successive changes are -1.096770, -0.791992, -0.655765, -0.605010, -0.415971, -0.452556, and -0.516468 million; their sample standard deviation gives sigma = 0.235716 million. Point = 36.562128 + 4*(-0.497501) + 0.000 = 34.572124 million, rounded to 34.572. The one-month 80% half-width is 1.28*sigma = 1.28*0.235716 = 0.301717 million; for a four-month horizon I scale by sqrt(4), giving 0.603434 million. Bounds are 34.572124 - 0.603434 = 33.968690 and 34.572124 + 0.603434 = 35.175558, rounded to 33.969 and 35.176. The interval reflects recent monthly-change dispersion scaled over the four-month horizon and does not add a separate policy-adjustment term.

Upside risk is that the early FY2026 enrollment drop stabilizes quickly, state backlogs or recertifications raise the September count, or disaster-related reporting adds participants; that would land above the interval if September exceeds 35.176 million. Downside risk is faster eligibility loss, stronger work-requirement implementation, or administrative churn continuing near the Oct-Jan decline pace; that would land below the interval if September is under 33.969 million. The outside the interval cases are therefore a clear stabilization above 35.176 or an accelerated drop below 33.969.

Review disposition: accepted the source-host cleanup by replacing fna.usda.gov references with the registered USDA FNS host, accepted the critique that the draft's -0.150 million policy adjustment was not independently grounded and removed it from the point estimate, accepted the clarification that 2027-02-15 is the registered resolve-by-bound date, and clarified that the interval is based on recent monthly-change dispersion scaled to four months.

calibrated forecast · 80% CI
34.6M[34M · 35.2M]
Target metadata

Data point: usda.fns.snap.persons.september_2026

More government data forecasts