# Food Security Fridays — Week 15: Charitable Food Access Reproducibility Package

## "Food Banks Will Pick Up the Slack"

**Author:** Matthew P. Rabbitt, PhD
**Date:** May 2026
**Series:** Food Security Fridays

---

## Overview

This post tests one of the most common claims raised whenever SNAP is cut:
that private charity will fill the gap. The One Big Beautiful Bill Act
(OBBBA, signed July 4, 2025) — a budget reconciliation bill, not a Farm
Bill reauthorization — cut SNAP by an estimated $186 billion over
2025–2034 (CBO score), so the substitution claim is no longer
hypothetical — millions of households will be displaced from SNAP
into whatever the charitable system provides.

The CPS Food Security Supplement asks two questions about charitable food
that test the substitution claim directly:

1. **Use** — Did the household get free groceries from a food pantry, food
   bank, or church in the past 12 months? (`HESC3`)
2. **Availability** — If not, is one available in the community where the
   household could get food if needed? (`HESC3A`)

The first question is regularly analyzed; the second — about perceived
availability among non-users — is rarely reported. The combination of the
two, restricted to in-universe households, lets us answer the question:
**among food-insecure households not currently using a food pantry, what
share say one is even reachable?**

> **Context: CPS-FSS cancellation.** USDA cancelled the CPS Food Security
> Supplement in September 2025. The December 2024 collection analyzed
> here is the **final CPS-FSS**. This package therefore documents the last
> nationally representative measurement of charitable food access in the
> United States until a successor data collection is fielded.

## Headline findings

- **Pantry reach.** Among in-universe food-insecure households, 31% used a
  food pantry in the last 12 months (36% among very-low-food-security
  households). Use rises with severity but caps below 40%.
- **Access gap among non-users.** Among food-insecure households that did
  not use a pantry, **29% report no pantry available in their community**
  — roughly 3.2 million households with no pantry available.
  The rate is essentially identical for VLFS non-users (29%); the
  infrastructure gap does not narrow as severity worsens.
- **Free-meal services have a wider gap.** 38% of food-insecure non-users
  report no free-meal service (church, shelter, Meals on Wheels) is
  available.
- **Geography is counterintuitive.** Metro food-insecure non-users report
  the larger access gap (31% vs. 18% in nonmetro) — population density
  alone doesn't translate into per-household availability.
- **Pantry visits are infrequent for most users.** Only 35% of pantry
  users go almost every month; 30% went only once or twice in a year.
  Pantries provide intermittent relief, not the consistent monthly
  support SNAP delivers. ("Chronic" and "episodic" are reserved here
  for food-insecurity persistence in the Nord 2013 sense; pantry-use
  frequency is a separate measure.)
- **Charitable food complements SNAP rather than substituting for it.**
  Among FI households, SNAP recipients use pantries at 42% vs. 25% for
  non-recipients.

## Unit of Analysis: Household

Every rate in this post is a household-level rate — the share of
*households* (not persons) in the relevant universe that fall into the
indicated category. The reproducibility file:

1. Loads the December 2024 CPS-FSS supplement-respondent file.
2. Keeps one record per household, restricting to the CPS reference person
   (`PERRP` ∈ {40, 41}). This yields N = 32,719 supplement-responding
   households (32,655 with a valid food security status).
3. Computes pantry-use and meal-use rates over all households except
   refused / DK / no response on the underlying CPS item. NIU
   (HESC3 == -1 / HESC4 == -1) households are coded as non-users
   (see "Universe rules" below). The pantry-use sample is N = 32,492.
4. Restricts availability rates to non-users with a substantive
   availability response (HESC3 == 2 and HESC3A ∈ {1, 2}, N = 10,990
   households for pantry; N = 11,893 for free meals; the FI subset is
   N = 2,452 / 3,109).
5. Weights all estimates by the household supplement weight (`HHSUPWGT`,
   stored with four implied decimals; divide by 10,000).

## Universe rules — read this before interpreting the rates

### Use questions (HESC3, HESC4)

The HESC3 / HESC4 universe is restricted by the CPS-FSS screener. A
household is **only asked** the charitable-food questions when at least
one of the following holds:

- `HRPOOR == 1` (income below 185% of the federal poverty line), **or**
- `HESS1 ∈ {2, 3, 4, -2}` (food-insufficient on the initial sufficiency
  screener, or screener missing), **or**
- `HES9 ∈ {1, -2}` (ran short of money for food at any point in the past
  year, or item missing).

Higher-income food-sufficient households are coded `HESC3 == -1` (not in
universe). **These households are coded as non-users** (`pantry_use = 0`,
`meal_use = 0`). The substantive justification is that NIU households
are higher-income and food-sufficient, and the maintained assumption is
that they are very unlikely to be using food pantries or free-meal
services. Refused, "don't know," and "no response" codes are kept
**MISSING** — these households were screened in and asked but did not
provide a substantive answer.

**Why this choice does not move FI rates.** Empirically, **zero** food-
insecure households are NIU on HESC3 (verified in earlier QA passes;
the cross-tab of HESC3 cells against `fi == 1` shows the NIU row
contains no FI households). As a result, the FI use
rates in Tables 1, 2, 5, 6, 7, and 9 are invariant to the NIU coding
choice. Only the non-FI use rate and the population marginal rate
move:

| Rate | NIU as MISSING (in-universe only) | NIU as non-user (this file) |
|------|----------------------------------:|----------------------------:|
| Pantry use, all households | 16.3% | **7.2%** |
| Pantry use, non-FI HHs | 9.9% | **3.5%** |
| Pantry use, FI HHs | 30.9% | **30.9%** |
| Pantry use, VLFS HHs | 35.6% | **35.6%** |

The "this file" column is what the headline statistics in the post
reflect.

### Availability questions (HESC3A, HESC4A)

Availability questions are asked **only** when the corresponding use
question is == 2 (the household did not use the service). Households
that answered "yes" to HESC3 / HESC4 receive `-1` on the availability
item — but here -1 means **the household IS a user**, not "out of
universe in the screener sense." Coding such households as 0 ("no
pantry available") would be wrong: they are using the pantry, so by
revealed preference at least one is available. Coding them as 1 ("yes
available") would impute information they were never asked to provide.

Availability rates therefore stay conditional on being a non-user with a
substantive response: `pantry_available` and `meal_available` are
MISSING for HESC3A / HESC4A == -1 (users) and for refused / DK / NR.

## Contents

| File | Description |
|------|-------------|
| `fsf_week15_reproducibility.do` | Self-contained Stata analysis (10 tables) |
| `cpsdec2024.do` | CPS-FSS data cleaning program (snapshot) |
| `fsf_week15_README.md` | This file |

## Data Requirements

- **December 2024 CPS Food Security Supplement** raw data file
  (`dec24pub.dat`) — the **final CPS-FSS** before USDA cancelled the
  series in September 2025.
- Source: U.S. Census Bureau, Current Population Survey
- Available at: https://www.census.gov/data/datasets/time-series/demo/cps/cps-supp_cps-repwgt/cps-food-security.html

## Instructions

1. Download the December 2024 CPS-FSS raw ASCII file (`dec24pub.dat`) from
   the Census Bureau URL above.
2. Open `cpsdec2024.do` and update the two `local` path macros at the top
   (`indir` and `outdir`) to point to (a) the directory containing
   `dec24pub.dat` and (b) the directory where `cpsdec2024.dta` should be
   saved. Then run `cpsdec2024.do` to produce `cpsdec2024.dta`.
3. Open `fsf_week15_reproducibility.do` and update the `$rawdata` and
   `$projdir` globals (at the top of the file) to point to (a) the
   directory containing `cpsdec2024.dta` and (b) the project working
   directory for output and logs.
4. Run `fsf_week15_reproducibility.do`.

The analysis log lands in `$projdir/output/fsf_week15_analysis.log`.

## Analysis Tables

The package is streamlined to the analyses needed to reproduce every
number cited in the post body, infographic, and first comment. Earlier
drafts included additional methodology / sensitivity tables (T6
SNAP×pantry 2-cat; T8B NIU-coding sensitivity; T9 deff=1.5 CIs; T10
universe appendix; T11 4-cat HH typology) — dropped here because no
public-facing claim depends on them.

| Table | Description |
|-------|-------------|
| 1 | Charitable food use by 2-cat food security |
| 2 | Charitable food use by 4-cat food security (HRFS12MD) — severity gradient |
| 3 | Frequency of pantry use (HESCF3) and last-30-day use (HESCM3) |
| 4 | Perceived availability of pantries / free meals among non-users (HESC3A, HESC4A) |
| 5 | Access gap — FI / VLFS non-users with no pantry available, weighted counts (3.2M) |
| 6 | Charitable food use and access by metro status (GTMETSTA) |
| 7 | Sample sizes and weighted counts |
| 8 | SNAP receipt timing × charitable food use, FI HHs only (currently SNAP / lost SNAP during year / never SNAP); also reports 30-day pantry × 30-day SNAP cross-tab |
| 9 | OBBBA-targeted typology × pantry use, FI HHs only — splits ABAWD-like at age 55 (the OBBBA age-cap expansion threshold) |
| 10 | Census Region × pantry use and pantry availability, FI HHs only (Northeast / Midwest / South / West) — geographic heterogeneity |
| 11 | Pairwise significance tests at α=0.10 (Census Bureau-recommended 90% confidence level for CPS-FSS) for every comparison asserted in the post body and infographic. Comparisons with p ≥ 0.10 are not asserted as differences. |

## Key Variables

### CPS source variables

| Variable | Description |
|----------|-------------|
| `HESC3` | Got free groceries from food pantry/bank/church past 12 mo (1=Yes, 2=No, -1=NIU) |
| `HESC3A` | Pantry available in community? Asked if HESC3==2 (1=Yes, 2=No, -1=NIU/user) |
| `HESCF3` | Pantry frequency: 1=Almost every month, 2=Some months, 3=1–2 months |
| `HESCM3` | Pantry use in last 30 days (asked of 12-month users): 1=Yes, 2=No |
| `HESC4` | Got free meal from church/shelter/Meals on Wheels past 12 mo |
| `HESC4A` | Free-meal service available in community (asked if HESC4==2) |
| `HESCF4` | Free-meal frequency |
| `HESCM4` | Free-meal use in last 30 days |
| `HRFS12M1` | 3-cat household food security (1=high/marg, 2=low, 3=very low) |
| `HRFS12MD` | 4-cat household food security (1=high, 2=marg, 3=low, 4=very low) |
| `HRPOOR` | Income below 185% FPL screener (1=Yes, 2=No) |
| `HESP1` | SNAP receipt past 12 months (1=Yes, 2=No, -1=NIU) |
| `GTMETSTA` | Metropolitan status (1=Metro, 2=Nonmetro, 3=Not identified) |
| `HHSUPWGT` | Household supplement weight (4 implied decimals) |
| `PERRP` | Relationship to reference person (40 = ref w/ relatives, 41 = ref alone) |

### Constructed variables

| Variable | Definition |
|----------|-----------|
| `pantry_use` | 1 if HESC3==1, 0 if HESC3 ∈ {2, -1}, missing if refused/DK/NR |
| `pantry_available` | 1 if HESC3A==1, 0 if HESC3A==2, missing if -1 (user) or refused/DK/NR |
| `meal_use` | 1 if HESC4==1, 0 if HESC4 ∈ {2, -1}, missing if refused/DK/NR |
| `meal_available` | 1 if HESC4A==1, 0 if HESC4A==2, missing if -1 (user) or refused/DK/NR |
| `pantry_freq` | HESCF3 if 1–3, missing otherwise (almost every month / some months / 1–2 months) |
| `meal_freq` | HESCF4 if 1–3, missing otherwise (free-meal frequency analog of pantry_freq) |
| `pantry_30d` | 1 if HESCM3==1, 0 if HESCM3==2, conditional on HESC3==1 |
| `any_charitable` | 1 if pantry_use==1 OR meal_use==1, 0 if both==0 (missing if both missing or one missing/other 0) |
| `fi` | 1 if HRFS12M1 in {2,3}, 0 if HRFS12M1==1 |
| `vlfs` | 1 if HRFS12M1==3, 0 otherwise (in-universe) |
| `fs12dcat` | HRFS12MD recoded 1–4 |
| `metro` | 1 if GTMETSTA==1, 0 if GTMETSTA==2 |
| `region` | Census Region from GEREG (1 = Northeast, 2 = Midwest, 3 = South, 4 = West) |
| `revhhid` | Household identifier: `hrhhid + hrhhid2` (string concatenation) |
| `numch` | Number of children in household: `bysort revhhid: egen numch = sum(prtage <= 17 & perrp > 47)`. PERRP > 47 excludes reference persons (40/41) and spouses/partners (42–47), so only minors who are own children, stepchildren, grandchildren, foster children, siblings, other relatives, or non-relatives are counted. Computed before restricting to reference persons so all HH members are on the file. Matches the standard USDA-ERS construction (CPS-FSS-GEN-VAR-YYYY.do) and Week 5 of this series. |
| `has_children` | 1 if `numch` > 0, 0 otherwise |
| `hh_obbba` | 5-cat OBBBA-targeted typology (1 = with children; 2 = elderly 65+, no children; 3 = working-age disability, no children; 4 = ABAWD-like 18–54 already subject; 5 = ABAWD-like 55–64 NEWLY subject under OBBBA) |
| `snap_recent` | 1 if HESP211==1 OR HESP212==1; 0 if HESP1∈{2, -1} or (HESP1==1 & neither month); missing if refused/DK/NR. NIU coded as non-recipient (consistent with `pantry_use` and `meal_use`). |
| `snap30` | 1 if HESP212==1; 0 if HESP1∈{2, -1} or (HESP1==1 & HESP212≠1); missing otherwise. NIU coded as non-recipient. |
| `snap_timing` | 3-cat SNAP receipt timing (1 = currently SNAP, received Nov/Dec 2024; 2 = lost SNAP during year; 3 = never SNAP or NIU) |

## Key Statistics

| Finding | Value | Source Table |
|---------|------:|--------------|
| Pantry use among FI HHs | 30.9% | Table 1 |
| Pantry use among VLFS HHs | 35.6% | Table 2 |
| Free-meal use among FI HHs | 9.8% | Table 1 |
| Any charitable food use among FI HHs | 33.1% | Table 1 |
| Pantry use among non-FI HHs (NIU treated as non-user) | 3.5% | Table 1 |
| Pantry use, all households | 7.2% | Table 1 |
| Pantry availability among FI non-users (Yes) | 70.7% | Table 4 |
| **No pantry available among FI non-users** | **29.3%** | Table 5 |
| **No pantry available among VLFS non-users** | **28.7%** | Table 5 |
| No free-meal service among FI non-users | 38.2% | Table 5 |
| Weighted FI HHs with no pantry available | 3.17M | Table 5 |
| Pantry use among FI metro HHs | 30.0% | Table 6 |
| Pantry use among FI nonmetro HHs | 35.4% | Table 6 |
| **No pantry available — metro FI non-users** | **31.5%** | Table 6 |
| **No pantry available — nonmetro FI non-users** | **18.4%** | Table 6 |
| Pantry frequency: almost every month | 35.3% | Table 3 |
| Pantry frequency: some months | 34.4% | Table 3 |
| Pantry frequency: only 1–2 months | 30.3% | Table 3 |
| Sample N: HHs with FS status | 32,655 | Table 7 |
| Sample N: pantry-use sample (non-missing pantry_use) | 32,492 | Table 1 |
| Sample N: FI HHs (non-missing pantry_use) | 4,074 | Table 7 |
| Sample N: FI non-users with availability data | 2,452 | Table 7 |
| **OBBBA-targeted typology (FI HHs only) — 5-cat (Table 9)** | | |
| Pantry use, FI elderly (65+) no-children HHs | 36.0% (N = 833) | Table 9 |
| Pantry use, FI ABAWD-like 18–54 (already subject) | **20.6%** (N = 869) | Table 9 |
| **Pantry use, FI ABAWD-like 55–64 (NEWLY subject under OBBBA)** | **26.9%** (N = 314) | Table 9 |
| Share of FI HHs newly subject (ABAWD 55–64) | 7.1% | Table 9 |
| **SNAP receipt timing × pantry use (FI HHs only) — Table 8** | | |
| Pantry use, currently SNAP (received Nov/Dec 2024) | 42.5% (N = 1,224) | Table 8 |
| **Pantry use, lost SNAP during year** | **36.9%** (N = 179) | Table 8 |
| Pantry use, never SNAP | 25.0% (N = 2,654) | Table 8 |
| Share of FI HHs currently SNAP | 29.9% | Table 8 |
| Share of FI HHs lost SNAP during year | 4.2% | Table 8 |
| Share of FI HHs never SNAP | 64.1% | Table 8 |
| Share of FI HHs missing SNAP timing (refused/DK/NR on HESP1) | 1.8% | Table 8 |
| (sums to 100.0% across the four categories) | | |
| **Census Region × charitable food access (FI HHs only) — Table 10** | | |
| Pantry use, Northeast | 29.8% (N = 548) | Table 10 |
| Pantry use, Midwest | 33.0% (N = 787) | Table 10 |
| Pantry use, South | 31.6% (N = 1,643) | Table 10 |
| Pantry use, West | 28.3% (N = 1,096) | Table 10 |
| **No pantry available among FI non-users, South** | **33.0%** (N = 1,643) | Table 10 |
| No pantry available among FI non-users, Northeast | 31.2% | Table 10 |
| No pantry available among FI non-users, West | 29.9% | Table 10 |
| **No pantry available among FI non-users, Midwest** | **19.8%** (best access) | Table 10 |
| Share of FI HHs in the South (weighted) | 42.7% | Table 10 |
| Share of FI HHs in the West (weighted) | 21.2% | Table 10 |
| Share of FI HHs in the Midwest (weighted) | 20.6% | Table 10 |
| Share of FI HHs in the Northeast (weighted) | 15.5% | Table 10 |
| (sums to 100.0%; unweighted N-shares would differ slightly because the South has a larger sample) | | |
| **Volume comparison** | | |
| OBBBA SNAP cut, total over 2025–2034 | $186 billion | OBBBA / CBO score |
| OBBBA SNAP cut, annual average | $18.6 billion / year | OBBBA / CBO score |
| Pantry-access welfare value, aggregate U.S. annual | $19–$28 billion / year | Byrne & Just (2023, AJAE) |
| Pantry visit value (per visit, willingness-to-pay) | $40–$60 / visit | Byrne & Just (2023, AJAE) |
| Pantry annual value per household | $600–$1,000 / year | Byrne & Just (2023, AJAE) |
| SNAP benefit per person per month, FY 2024 average | $187.20 | Jones, Todd, & Toossi (2025), USDA-ERS EIB-291, Table 1 |
| SNAP total spending, FY 2024 | $99.8 billion | Jones, Todd, & Toossi (2025), USDA-ERS EIB-291, Table 1 |
| SNAP avg monthly participation, FY 2024 | 41.7 million people | Jones, Todd, & Toossi (2025), USDA-ERS EIB-291, Table 1 |

## Methodological Notes

- **NIU treatment for use questions.** Higher-income food-sufficient
  households are coded NIU (HESC3 == -1, HESC4 == -1) on the use
  questions. This file treats them as non-users (`pantry_use = 0`,
  `meal_use = 0`) on the substantive grounds that NIU households are
  unlikely to be using food pantries or free-meal services. The choice
  is invariant for FI households: empirically zero FI HHs are NIU on
  HESC3 (Universe Appendix, Table 10), so the FI use rates and all
  cross-tabs that condition on FI = 1 are identical under either coding.
  Only the non-FI rate (3.5%) and the population marginal rate (7.2%)
  differ from the in-universe-only treatment.

- **SNAP timing alignment.** Charitable-food use questions HESC3 / HESC4
  use a 12-month reference period; HESCM3 / HESCM4 use a 30-day
  reference period. SNAP receipt is constructed at the same two
  resolutions:
  - 12-month: HESP1 — aligns with HESC3 / HESC4.
  - 30-day: HESP212 (December 2024 monthly indicator) — aligns with
    HESCM3 / HESCM4. The CPS-FSS does not have a single "past-30-days
    SNAP" item; it has 12 monthly indicators (HESP21–HESP212), each
    asked when HESP1 == 1.
  - SNAP timing (3-cat): currently SNAP = received in Nov or Dec 2024;
    lost SNAP during year = HESP1 == 1 but not received in Nov or Dec;
    never SNAP = HESP1 == 2 or NIU. The "lost SNAP during year" cell
    is the closest within-CPS-FSS proxy for the substitution mechanism.

- **OBBBA-targeted typology.** The One Big Beautiful Bill Act (signed
  July 4, 2025) expanded the ABAWD work-requirement age cap from 54
  to 64. Table 9 splits the ABAWD-like cell at age 55 to identify the
  newly subject population (working-age 55–64 with no children and no
  disability). This 7.1% of FI households is the demographic most
  directly affected by the OBBBA age-cap expansion.

- **Availability is conditional on being a non-user.** HESC3A measures
  what non-users *believe* about pantry presence in their community — not
  geocoded distance to the nearest pantry. The 29% rate is an
  *awareness/perceived-access* rate, which is the operative concept for
  the substitution claim: households that don't know a pantry exists
  cannot use one. Availability codes -1 (HH is a user) and refused / DK /
  NR are kept MISSING; users implicitly know a pantry is available, but
  the questionnaire does not record their answer.

- **SNAP benefit per person per month ($187.20, FY 2024).** Verified
  against Jones, Todd, & Toossi (2025), USDA-ERS EIB-291, Table 1.
  Replaces an unsourced "$250 per household per month" figure used in
  earlier drafts; ERS publishes per-person, not per-household, as the
  standard SNAP benefit measure. Used in the post draft as a comparison
  anchor for the consistency of SNAP versus the intermittent frequency
  of pantry use; the figure is not computed in this reproducibility .do.

- **Weights.** All estimates use household supplement weights (HHSUPWGT).
  Census stores HHSUPWGT as integer × 10,000; the analysis divides by
  10,000 to recover the actual weight. `svy:` linearization is used for
  Wald tests and rate CIs.

- **Reliability.** Headline rates are produced from samples of 2,000+
  households; under linearization SEs (svy: mean), the standard error
  on a 30% rate is roughly 1.0–1.2 pp at FI sample sizes around 4,000.
  An assumed design effect of 1.5 (a conservative bound for CPS subgroup
  proportions) would inflate the SE by sqrt(1.5) ≈ 1.22 — still well
  under 2 pp.

- **Replicate weights (planned enhancement).** The CPS-FSS December 2024
  collection ships with 160 BRR replicate weights in `dec24hhrep.dat`,
  alongside the public-use file. Replicate weights would tighten subgroup
  variance estimates relative to the linearization SEs reported here.
  The replicate-weight file is not bundled in this reproducibility
  package because reading it requires a Census Bureau dictionary file;
  users who need replicate-weight inference can download both files
  from the CPS-FSS Census page, build the dictionary from the
  documentation, and `svyset` with `vce(brr) brrweight(...)`. The deff
  = 1.5 approximation in Table 9 is uniformly conservative for CPS
  subgroup proportions.

- **Significance testing (Census Bureau standard for CPS-FSS).** The
  Census Bureau recommends 90% confidence (α=0.10, two-sided) as the
  default significance threshold for CPS-FSS comparisons given the
  supplement's complex sample design and sub-population sample sizes.
  Table 11 reports pairwise significance tests for every comparison
  asserted in the post body and infographic. Tests use linearization
  standard errors via `svy: mean` over the relevant grouping, with
  pairwise `lincom` contrasts. Comparisons with p ≥ 0.10 are not
  asserted as differences in the public-facing materials.

- **Substitution-test identification (limitation).** The headline
  finding — that lost-SNAP HHs use pantries at the same rate (37%) as
  current SNAP recipients (43%) — is a cross-sectional comparison
  inside a single survey wave. Strictly, a true substitution test
  would require a longitudinal design that tracks the same households
  before and after SNAP loss. The cross-sectional comparison rests on
  the assumption that lost-SNAP and current-SNAP HHs have similar
  unobserved characteristics relevant to pantry use. If lost-SNAP HHs
  systematically differ in baseline food insecurity (e.g., they
  "graduated off" SNAP because their incomes rose), the comparison
  could understate or overstate substitution. The interpretation in
  the public-facing materials — "pantries do not scale up to replace
  SNAP" — is defensible because the cross-sectional rates are
  monotonic in SNAP exposure (43%/37%/25% for currently/lost/never
  SNAP), are reinforced by the volume comparison ($18.6B annual SNAP
  cut, at $2–$3 per meal, translates to roughly 6–9 billion meals/year
  in lost food assistance — comparable to or exceeding the entire
  ~6 billion meals/year distributed by Feeding America's 200 food
  banks and 60,000 partner agencies; for FA's network to absorb the
  cut alone it would have to roughly double, and the broader U.S.
  charitable food system would have to grow by roughly 80–130%), and
  are consistent with the 2013 ARRA SNAP-cliff historical analog. A
  more measured academic framing would say "no evidence of pantry-use
  scaling up to replace lost SNAP benefits" rather than asserting
  causal substitution failure.

- **Subgroup-vs-overall SE approximation (Tests G1, G2, H in Table 11).**
  The tests of subgroup means against the FI overall mean
  (e.g., ABAWD 55–64 pantry use vs FI average pantry use) use
  SE_diff = sqrt(SE_subgroup² + SE_overall²), which assumes
  independence between the subgroup mean and the overall mean. Because
  the subgroup is a subset of the overall sample, the true covariance
  is positive, so this approximation OVERSTATES SE_diff and produces
  CONSERVATIVE p-values (more likely to fail to reject the null). The
  conclusions hold under the conservative test: ABAWD 18–54 vs FI
  overall (p < 0.10) passes with substantial margin; ABAWD 55–64 vs
  FI overall (p ≈ 0.18) does not, and would not pass under a proper
  test either. A more precise contrast can be implemented via
  `svy: mean over()` with explicit `lincom` or `test` of the difference
  using the full variance-covariance matrix; this is left for future
  work and is unlikely to change conclusions given the magnitude of
  the differences. The approximation is labelled "approx p" in the
  Stata output for transparency.

- **Multiple comparisons.** The pairwise tests are not adjusted for
  multiple comparisons in the headline rates. The region section
  involves the most pairwise comparisons (six pairs across four
  regions). Bonferroni-adjusted threshold at α=0.10 with six tests is
  α=0.0167. The substantive findings hold under the adjustment:

  | Region pair | Unadjusted *p* | Bonferroni-passes (α=0.0167)? |
  |-------------|----------------:|-------------------------------:|
  | Midwest vs South | < 0.001 | ✅ Yes |
  | Midwest vs Northeast | 0.002 | ✅ Yes |
  | Midwest vs West | 0.001 | ✅ Yes |
  | South vs Northeast | 0.610 | n/a (not asserted) |
  | South vs West | 0.294 | n/a (not asserted) |
  | Northeast vs West | 0.726 | n/a (not asserted) |

  All three "Midwest stands apart" claims survive Bonferroni adjustment
  for the six pairwise region tests. The other comparisons in the post
  (SNAP timing × pantry use, severity gradient, ABAWD-vs-FI-average,
  metro/nonmetro, SNAP recipients vs non-recipients) involve at most
  three pairwise tests within a family, and all asserted differences
  pass with substantial margin (p ≤ 0.01) so adjustment does not
  materially change the conclusions.

  **Comparisons that fail at α=0.10 (NOT asserted as differences):**

  | Comparison | Diff (pp) | p-value |
  |------------|----------:|--------:|
  | Currently SNAP (43%) vs Lost SNAP during year (37%) | 5.6 | 0.208 |
  | South (33%) vs Northeast (31%) — no pantry available | 1.8 | 0.610 |
  | South (33%) vs West (30%) — no pantry available | 3.1 | 0.294 |
  | Northeast (31%) vs West (30%) — no pantry available | 1.3 | 0.726 |
  | VLFS (29%) vs FI (29%) — no pantry available | 0.6 | 0.796 |
  | ABAWD-like 55–64 (27%) vs FI overall (31%) — pantry use | 4.0 | ≈0.18 |

  **Comparisons that pass at α=0.10 (asserted as differences):**

  | Comparison | Diff (pp) | p-value |
  |------------|----------:|--------:|
  | Lost SNAP (37%) vs Never SNAP (25%) — pantry use | 11.9 | 0.005 |
  | Currently SNAP (43%) vs Never SNAP (25%) — pantry use | 17.5 | <0.001 |
  | Midwest (20%) vs South (33%) — no pantry available | 13.1 | <0.001 |
  | Midwest (20%) vs Northeast (31%) — no pantry available | 11.4 | 0.002 |
  | Midwest (20%) vs West (30%) — no pantry available | 10.1 | 0.001 |
  | Metro (31%) vs Nonmetro (18%) — no pantry available | 13.1 | <0.001 |
  | ABAWD-like 18–54 (21%) vs ABAWD-like 55–64 (27%) — pantry use | 6.3 | 0.064 |
  | ABAWD-like 18–54 (21%) vs Elderly 65+ (36%) — pantry use | 15.4 | <0.001 |
  | ABAWD-like 18–54 (21%) vs FI overall (31%) — pantry use | 10.3 | <0.001 |
  | All severity-gradient adjacent categories (HRFS12MD) | 7.7+ | <0.001 |
  | SNAP recipients (42%) vs non-recipients (25%) — pantry use | 16.8 | <0.001 |

- **Rounding convention.** Headline rates in the post and infographic
  are reported to whole percentage points to match ERR-358's
  presentation convention. Underlying log values are reported to two
  decimal places (e.g., 31.45% / 18.39% for the metro / nonmetro
  contrast in Table 7; 30.92% for FI pantry use; 29.30% for FI
  no-pantry-available). Either presentation reproduces from the .do.

- **Verification against USDA ERS Annual Food Security Report (ERR-358).**
  Aggregate household-level food security categories (134.06M households,
  18.34M FI, 13.7% FI rate) reproduce ERR-358 Table 1A. Charitable food
  use rates are not directly published in ERR-358; for cross-check,
  Coleman-Jensen et al. (2024) and earlier ERS analyses report FI pantry
  use in the 25–35% range, consistent with this file.

## Software

- Stata 14 or later
- No additional packages required

## External Citation Verification (pre-publication checklist)

The post and infographic cite several external figures that should be
verified against primary sources before publication. Status as of the
last QA pass:

| Figure | Source | Status |
|--------|--------|--------|
| **47.6 million SNAP recipients at the 2013 ARRA cliff** | USDA-FNS State Activity Reports (Nov 2013) | ✅ Verified — Wikipedia citing USDA-FNS confirms SNAP peaked at 47.6M in 2013 |
| **OBBBA $186 billion over 2025–2034** | CBO publication 61570, *Estimated Budgetary Effects of Public Law 119-21, ... Relative to CBO's January 2025 Baseline* (July 21, 2025), Table 1, "Subtotal, Subtitle A" row. https://www.cbo.gov/publication/61570 | ✅ **Verified directly against CBO.** The downloadable Excel workbook for publication 61570 reports −$186,650 million in net Estimated Outlays for Title I, Subtitle A (Nutrition) over 2025–2034 — i.e., $186.65 billion in SNAP/nutrition spending reductions, rounded down to $186B in public-facing materials. Component breakdown (in $ millions): ABAWD work requirements −68,600; Thrifty Food Plan re-evaluation −37,300; State Matching Funds Requirements −40,810; Administrative Cost Sharing −24,660; Internet Expenses Restrictions −10,980; Standard Utility Allowances −5,940; Education/Obesity Prevention Grant −5,470; Alien SNAP Eligibility −1,904; Total Interactions (offset) +9,014. Earlier drafts cited "CBO publication 61461" (the June 4, 2025 score of the House-passed version, which scored Subtitle A at $287B) — corrected to publication 61570 (the post-enactment July 21, 2025 score of P.L. 119-21 as signed). Earlier $187B / FY 2026–2035 framing also corrected to verified $186B / 2025–2034. ABAWD age-cap expansion (54 → 64) verified against Wikipedia and the bill text. |
| **SNAP $187.20 per person per month, FY 2024** | Jones, J. W., Todd, J. E., & Toossi, S. (2025). *The Food and Nutrition Assistance Landscape: Fiscal Year 2024 Annual Report* (Report No. EIB-291). USDA-ERS. Table 1. | ✅ **Verified.** This figure replaces the earlier "$250 per household per month" figure that could not be sourced. EIB-291 reports the FY 2024 SNAP benefit per person averaged $187.20/month (an 11.5% decline from $211.45 in FY 2023, reflecting the expiration of emergency allotments). The report does not separately publish a per-household average; per-person is the standard ERS measure. Materials updated accordingly. |
| **Pantry-visit welfare value: $40–$60 per visit** | Byrne, A. T., & Just, D. R. (2023). "What is free food worth? A nonmarket valuation approach to estimating the welfare effects of food pantry services." *American Journal of Agricultural Economics*. DOI: [10.1111/ajae.12355](https://onlinelibrary.wiley.com/doi/10.1111/ajae.12355) | ✅ **Verified.** Travel-cost methodology applied to administrative pantry-client data from a Colorado food bank; per-visit willingness-to-pay $40–$60; annual per-household value $600–$1,000; aggregate national pantry-access welfare value $19–$28 billion/year (assuming Colorado parameters generalize). Public-facing materials cite Byrne & Just (2023) for the **per-visit value** ($40–$60) only. The volume comparison ("the cut would require pantries to roughly double") is grounded in Feeding America's published meal-distribution figure rather than B&J's welfare-value estimate, because mixing welfare-value (consumer-surplus-inclusive) on the denominator with SNAP face value on the numerator produces an apples-to-oranges ratio that under-states the required physical/retail growth. |
| **Feeding America 2024 distribution: ~6 billion meals/year** | Feeding America 2024–2025 annual reports; corroborated by Charity Navigator and multiple press accounts. https://www.feedingamerica.org/about-us/financials | ✅ **Verified.** The Feeding America network — 200 member food banks and 60,000 partner agencies — distributed approximately 6 billion meals in 2024. Used as the volume-comparison denominator: $18.6B/yr SNAP cut at SNAP's typical per-meal cost ($2–$3) translates to 6–9 billion meals/year in lost food assistance, comparable to or slightly exceeding FA's entire annual distribution. Required charitable-food growth to absorb the cut: roughly double in retail/operational capacity. |

If any of these change before publication, update the post body, infographic source footer, website article block, and the relevant entry below.

## References

Byrne, A. T., & Just, D. R. (2023). What is free food worth? A
nonmarket valuation approach to estimating the welfare effects of food
pantry services. *American Journal of Agricultural Economics*. DOI:
[10.1111/ajae.12355](https://onlinelibrary.wiley.com/doi/10.1111/ajae.12355).
(Source for per-visit value $40–$60, annual per-household value
$600–$1,000, and aggregate U.S. pantry-access welfare value
$19–$28 billion/year.)

Congressional Budget Office (2025). *Estimated Budgetary Effects of
Public Law 119-21, to Provide for Reconciliation Pursuant to Title II of
H. Con. Res. 14, Relative to CBO's January 2025 Baseline*. CBO
publication 61570 (released July 21, 2025).
https://www.cbo.gov/publication/61570 (Source for the $186 billion over
2025–2034 SNAP/nutrition reduction figure; specifically the Title I,
Subtitle A "Subtotal, Subtitle A" row in the downloadable Excel
workbook.)

Feeding America (2025). *2024 Annual Report*.
https://www.feedingamerica.org/about-us/financials (Source for the
"~6 billion meals" annual distribution figure used in the volume
comparison; corroborated by Charity Navigator nonprofit profile and
multiple 2024–2025 press accounts.)

Jones, J. W., Todd, J. E., & Toossi, S. (2025). *The Food and
Nutrition Assistance Landscape: Fiscal Year 2024 Annual Report*
(Report No. EIB-291). U.S. Department of Agriculture, Economic
Research Service. (Source for FY 2024 SNAP benefit per person per
month: $187.20; total SNAP spending: $99.8 billion; average monthly
participation: 41.7 million.)

Rabbitt, M.P., Reed-Jones, M., Hales, L.J., Suttles, S., & Burke, M.P.
(2025). *Household food security in the United States in 2024* (Report
No. ERR-358). U.S. Department of Agriculture, Economic Research Service.
https://www.ers.usda.gov/publications/pub-details?pubid=113622

U.S. Census Bureau (2024). *Current Population Survey, December 2024:
Food Security Supplement Technical Documentation*.
https://www.census.gov/data/datasets/time-series/demo/cps/cps-supp_cps-repwgt/cps-food-security.html

USDA Food and Nutrition Service (2024). *SNAP Data Tables: Average Monthly
Participation and Benefits, FY 2024*. https://www.fns.usda.gov/pd/supplemental-nutrition-assistance-program-snap

## Citation

Rabbitt, Matthew P. 2026. "Food Security Fridays — Food Banks Will Pick
Up the Slack." Food Security Fridays series.
https://matthewpatrickrabbitt.com/food-security-fridays.html