# Food Security Fridays — Week 13: Additional Statistics

Supplementary material for "The Groups You Won't Find in the Annual
Report." All statistics come from the same December 2024 CPS Food
Security Supplement extract used in the main post; see
`fsf_week13_reproducibility.do` for the canonical code.

> **Context: CPS-FSS cancellation.** USDA cancelled the CPS Food
> Security Supplement in September 2025. The December 2024 collection
> analyzed in this package is the **final CPS-FSS**. Every disaggregated
> estimate below is therefore a *terminal* race/ethnicity food
> insecurity statistic from the CPS-FSS, not an interim number. Any
> successor data collection that replaces the CPS-FSS will need to
> benchmark against these final ERR-358 values.

---

## SPD 15: what changed in 2024

The March 2024 revision to OMB Statistical Policy Directive No. 15 is
the first substantive update to federal race/ethnicity standards since
1997. Three changes matter most for food security data:

1. **One combined question, not two.** Race and Hispanic origin move
   from separate questions into a single combined item. Respondents who
   identify as Hispanic no longer have to choose a race.
2. **MENA as a minimum category.** Middle Eastern or North African
   ancestry becomes its own reportable category, distinct from White.
3. **Disaggregation required.** Agencies must publish estimates within
   minimum categories when samples permit. Aggregated-only reporting is
   no longer compliant.

Federal agencies have until **March 2029** to implement. The December
2024 CPS still used the 1997 two-question format, so MENA is not
tabulable from that release. **Because USDA cancelled the CPS-FSS
in September 2025, there will be no further CPS-FSS collections under
either the 1997 or the 2024 standard.** A MENA food insecurity rate
will therefore depend on a successor survey that adopts SPD 15 — see
the concrete asks in the main post.

### MENA data gap

Roughly **3.5 million people** reported Middle Eastern or North African
(MENA) descent in the 2020 Census Detailed Demographic and Housing
Characteristics File A (DHC-A) — the first time the Census Bureau
specifically solicited MENA write-in responses, tabulated within the
White racial category per the 1997 SPD 15 standards (Marks, Jacobs, &
Coritz, 2023). On the December 2024 CPS, these respondents (and the
households they lead) are distributed across three annual-report
categories depending on how they answered the 1997-format race
question:

- Those who marked "White" only → inside "White, non-Hispanic"
  (the largest share).
- Those who marked "Other" or wrote in a MENA identifier under
  "Some other race alone" → inside "Other, non-Hispanic".
- Those who marked MENA alongside another race → inside "Multiracial,
  non-Hispanic".

No MENA food insecurity rate can be produced from CPS-FSS. With the
series cancelled, this allocation is **permanent in the historical
CPS-FSS record**: MENA households that experienced food insecurity
between 1995 and 2024 contribute to the White NH (and sometimes Other
NH) denominators without ever being separately visible. Any future
MENA-specific FI estimate will have to come from a successor survey or
from alternative sources (e.g., HPS, NHIS FSS modules, or a new ACS
subject inquiry) that adopt the 2024 SPD 15 combined question.

---

## The very-low-food-security picture

Aggregation hides severe food insecurity even more than it hides
overall rates. The VLFS rates inside the annual report's four lines:

| Group | VLFS rate |
|-------|----------:|
| **AIAN alone, non-Hispanic** | **14.5%** |
| Multiracial, non-Hispanic | 11.4% |
| Black, non-Hispanic | 10.5% |
| Central American (excl. Salvadoran) | 10.5% |
| Dominican | 10.1% |
| Puerto Rican | 9.9% |
| Other Spanish Hispanic | 8.0% |
| NHPI alone, non-Hispanic | 7.5% |
| Salvadoran | 5.9% |
| Cuban | 5.5% |
| Japanese alone, NH | 5.4% |
| Mexican | 5.2% |
| South American | 5.1% |
| White, non-Hispanic | 4.1% |
| Filipino alone, NH | 3.9% |
| Vietnamese alone, NH | 2.8% |
| Korean alone, NH | 1.9% |
| Chinese alone, NH | 1.6% |
| Other Asian, NH | 1.5% |
| Asian Indian alone, NH | 0.8% |

AIAN households face a higher VLFS rate (14.5%) than Black non-Hispanic
households (10.5%). This has never appeared in the annual report.

---

## Children's food insecurity (HRFS12MC)

Children's food insecurity is observed only in households with at least
one child under 18 (N = 8,454 households). The pattern largely mirrors
household-level FI with one important exception: **Dominican households
with children face a child-FI rate of 27.85%** — the highest of any
Hispanic origin and nearly 3x the national rate of 9.05%.

### Child FI by ERR-358 4-category

| Group | Child FI rate |
|-------|--------------:|
| White, non-Hispanic | 6.36% |
| Black, non-Hispanic | 17.41% |
| Hispanic (any race) | 11.65% |
| Other, non-Hispanic | 6.38% |
| **Total** | **9.05%** |

### Child FI by 7-category disaggregation

| Group | Child FI rate |
|-------|--------------:|
| White alone, NH | 6.36% |
| Black alone, NH | 17.41% |
| **AIAN alone, NH** | **18.60%** (N = 90) |
| Asian alone, NH | 2.86% |
| NHPI alone, NH | 13.80% (N = 37) |
| Multiracial, NH | 13.85% |
| Hispanic (any race) | 11.65% |

### Child FI by detailed Hispanic origin

| Origin | Child FI rate |
|--------|--------------:|
| **Dominican** | **27.85%** (N = 42) |
| Cuban | 13.95% |
| Other Spanish | 13.95% |
| Central American (excl. Salvadoran) | 13.22% |
| Puerto Rican | 12.04% |
| Mexican | 10.69% |
| Salvadoran | 10.14% |
| South American | 10.13% |

The Dominican child-FI rate is notable even accounting for the small
cell (N=42, design-effect-adjusted SE ≈ 8.5 pp): the 95% CI is roughly
[11%, 44%], so the lower bound sits near the Hispanic average (11.65%)
but well above the national child-FI average (9.05%). This is a
directional finding that deserves multiyear pooling for confirmation.

---

## Income-controlled food insecurity (above 185% FPL)

One common interpretation of race/ethnicity gaps in food insecurity is
that they are artifacts of income differences. To test this, we
restrict the sample to households with **income at or above 185% of the
federal poverty line** — the threshold used by USDA for SNAP categorical
eligibility screens and for most federal nutrition-assistance income
tests. Household income is midpoint-imputed from the CPS `hefaminc`
brackets, the poverty threshold is assigned from the 2024 Census table
by household size, number of children, and reference-person age, and
the income-to-poverty ratio (`hhincpv = hhincome / hpvcut`) is compared
to 1.85. This is more precise than the CPS `HRPOOR` screener variable,
which only approximates income (and, per Census documentation, codes
`HRPOOR = 2` as "above 185% poverty **or** income not reported"). The
subsample drops most low-income households but preserves roughly
20,900 observations.

### FI rate above 185% FPL, by 7-category

| Group | FI rate (all HHs) | FI rate (above 185% FPL) | Ratio |
|-------|------------------:|-------------------------:|------:|
| White alone, NH | 10.1% | 6.10% | 0.60 |
| Black alone, NH | 24.4% | 17.11% | 0.70 |
| AIAN alone, NH | 30.9% | **20.64%** | 0.67 |
| Asian alone, NH | 6.6% | 3.13% | 0.47 |
| NHPI alone, NH | 18.4% | 14.16% | 0.77 |
| Multiracial, NH | 21.8% | 12.14% | 0.56 |
| Hispanic (any race) | 20.2% | 11.80% | 0.58 |

**Key finding.** Income controls narrow every group's FI rate, but the
racial/ethnic hierarchy persists — and under the income-to-poverty
measure, **AIAN households above 185% FPL face the highest food
insecurity rate (20.64%)**, exceeding Black non-Hispanic households
(17.11%). Above 185% FPL, AIAN households are food insecure at **~3.4x
the rate of White non-Hispanic households** (20.64% vs 6.10%), Black
households at nearly **3x** (17.11%), and Hispanic households at
**roughly 2x** (11.80%). Income differences are part of the story but
not all of it. *Small-cell caution:* AIAN alone above 185% FPL
N = 125, so the 20.64% point estimate has a wide CI and should be
interpreted as directional.

### FI above 185% FPL, detailed Hispanic

| Origin | FI rate above 185% FPL |
|--------|----------------------:|
| Puerto Rican | 12.68% |
| Central American (excl. Salvadoran) | 12.31% |
| South American | 12.27% |
| Cuban | 12.11% |
| Mexican | 11.67% |
| Other Spanish | 11.26% |
| Dominican | 10.48% |
| Salvadoran | 10.00% |

Under the income-to-poverty measure, Hispanic subgroup FI rates above
185% FPL cluster tightly in a 10.00%–12.68% band (a 2.68 pp range). No
single Hispanic origin subgroup is distinctly elevated once income is
controlled. The ranking also shifts relative to the unconditional
Hispanic ordering — a substantive narrowing that suggests most of the
Hispanic-subgroup variation in the headline rates is absorbed by
income differences, with the residual range small enough that it does
not cleanly track any single factor.

### FI above 185% FPL, detailed Asian (NH)

| Origin | FI rate above 185% FPL |
|--------|----------------------:|
| Vietnamese | 7.47% |
| Japanese | 6.61% |
| Korean | 5.77% |
| Filipino | 5.55% |
| Other Asian | 4.48% |
| Chinese | 1.44% |
| Asian Indian | 0.03% |

Vietnamese households remain highest within Asian NH above 185% FPL,
consistent with the unconditional pattern. Asian Indian households
above 185% FPL show essentially no food insecurity (0.03%, N = 276).
Small-cell caution applies: the Vietnamese, Japanese, and Korean
sub-samples above 185% FPL are N = 76, 118, and 84 respectively.

---

## Multiracial detail

"Multiracial, non-Hispanic" is the fastest-growing race category in the
U.S. (21.8% food insecure, 2.2M households). Inside it:

| Combination | FI rate | VLFS rate | N |
|-------------|--------:|----------:|--:|
| Three or more races | 31.2% | 17.5% | 46 |
| White-AIAN | 27.4% | 14.5% | 172 |
| Black-AIAN/Asian/NHPI | 26.2% | 14.5% | 42 |
| Other two-race | 22.2% | 4.6% | 22 |
| White-Black | 18.6% | 10.7% | 107 |
| White-Asian | 13.4% | 5.9% | 93 |
| White-NHPI | 12.1% | 3.8% | 22 |

**White-AIAN multiracial households** show 27.4% FI (N=172), consistent
with the AIAN-alone point estimate (30.9%). The "Black-AIAN/Asian/NHPI"
cell (26.2%, N=42) pools PTDTRACE codes 10–12, only one of which
(Black-AIAN) includes AIAN; "Three or more races" (31.2%, N=46) includes
many non-AIAN combinations. The cleanest multiracial-AIAN signal is
therefore White-AIAN, which confirms elevated FI among AIAN-identifying
households is not a single small-sample artifact.

---

## SNAP take-up among food-insecure households

**Methodological note on selection.** These rates condition on the
household being food insecure. SNAP receipt is endogenous to food
security status: households that receive SNAP and become food secure
exit the denominator. The observed take-up rate among food-insecure
households therefore underestimates total SNAP reach and overrepresents
households for whom SNAP was insufficient to achieve food security.
The *between-group* comparisons are still informative (the selection
operates similarly across races), but the *levels* should not be read
as a straightforward measure of program access.

The annual report's four categories show modest SNAP participation
differences (32.6% White NH to 38.0% Black NH). Disaggregation reveals a
nearly 3x range:

| Group | % of FI HHs receiving SNAP |
|-------|---------------------------:|
| AIAN alone, NH | 52.1% |
| Dominican | 51.7% |
| Puerto Rican | 51.4% |
| Cuban | 38.6% |
| Black, NH (for reference) | 38.0% |
| Other Spanish | 37.0% |
| Asian alone, NH | 36.9% |
| Mexican | 34.0% |
| White, NH (for reference) | 32.6% |
| Multiracial, NH | 30.9% |
| NHPI alone, NH | 28.8% |
| Salvadoran | 23.7% |
| Central American (excl. Salv.) | 21.3% |
| South American | 18.1% |

Central and South American households face FI rates above 20% but
receive SNAP at less than half the rate of Puerto Rican or Dominican
households. The gap reflects several overlapping policy mechanisms:

- **Citizenship eligibility.** Non-citizens without a qualified status
  are ineligible for SNAP; qualified-status non-citizen adults are
  subject to a 5-year bar. Mixed-status households face chilling effects
  even when children or some adults are eligible.
- **Puerto Ricans are U.S. citizens by birth** — so the high Puerto
  Rican SNAP take-up reflects both high FI exposure and the absence of
  citizenship barriers. (Note: households surveyed on the mainland CPS;
  residents of Puerto Rico are covered by NAP, not SNAP.)
- **State administrative variation.** Several Southeast states have
  historically higher administrative denial rates for mixed-status
  households; Central and South American households are disproportionately
  concentrated in those states.
- **Program awareness and language access.** Outreach materials and
  application support remain uneven across immigrant-origin groups.

The policy takeaway: the same annual-report line hides both higher food
insecurity exposure *and* lower program access for several Hispanic
origin groups. Food security policy designed for a single "Hispanic"
denominator misses both halves of the pattern.

---

## Geographic concentration (context, not computed here)

The detailed race/ethnicity groups disaggregated in this post are
concentrated geographically in ways that matter for state and regional
food security policy:

- **AIAN alone, non-Hispanic:** concentrated in the Northern Plains
  (North Dakota, South Dakota, Montana), the Southwest (Arizona, New
  Mexico, Oklahoma), and Alaska. State-level USDA FI estimates in
  these regions will be materially understated when AIAN rates are
  aggregated into "Other, non-Hispanic."
- **Puerto Rican:** concentrated in New York, Florida, Pennsylvania,
  New Jersey, Connecticut, and Massachusetts.
- **Mexican-origin households:** concentrated in California, Texas,
  Arizona, Illinois, and New Mexico.
- **Cuban:** heavily concentrated in Florida (especially Miami-Dade).
- **Dominican:** concentrated in New York, New Jersey, Florida, and
  Massachusetts.
- **Central American and Salvadoran:** concentrated in California,
  Texas, Virginia, Maryland, and the DC metro area.
- **Vietnamese:** concentrated in California, Texas, Washington,
  Florida, and Georgia.

Geographic concentration is not computed from CPS-FSS in this file; the
above summary draws on 2023 ACS one-year estimates from the Census
Bureau's Hispanic Origin and Race iterations tables. A state-level
extension would use the CPS-FSS state identifier (GESTFIPS) in a
multiyear pool; single-year state estimates are too noisy for most
disaggregated groups.

---

## Prior USDA/ERS acknowledgment

The gap between the four annual-report categories and the full CPS
detail is not a new observation. USDA-ERS researchers have acknowledged
it in several places:

- **Hales & Coleman-Jensen (2024), EIB-269.** The most directly relevant
  precedent. This standalone ERS data brief disaggregated food insecurity
  across nine race/ethnicity categories — including AIAN, Asian, NHPI,
  and three multiracial groups — using six pooled years of CPS-FSS
  (2016–21, N = 214,370). It also reported Hispanic and Asian origin
  subgroup rates. Key findings: AIAN FI 23.3% (highest), Asian 5.4%
  (lowest), MR/American Indian–White 21.7%; Hispanic range from Cuban
  11.4% to Dominican 21.0%; Asian range from Japanese 1.7% to Other
  Asian 11.4%. Laura Hales is a co-author on both EIB-269 and ERR-358.
  The disaggregation has been demonstrated as feasible by ERS itself —
  it simply has never been incorporated into the annual report series.
- The ERR-358 technical appendix notes that Native American and other
  "small population groups" are not separately reported because of
  sample-size concerns with a single-year CPS-FSS.
- Bleich et al. (2021) and several ERS working papers have called for
  multiyear pooling of CPS-FSS to produce publication-grade estimates
  for subgroups including AIAN, NHPI, and detailed Asian origin.
- The 2024 SPD 15 revision was explicitly anticipated in internal USDA
  statistical guidance documents; CPS-FSS data have been recoded in
  preparation for compliance.

The point of this post is not to fault ERS for past aggregation — with
a 1997-standards CPS and a strict single-year sample-size rule, the four
categories were defensible. The point is that **the data now support
disaggregation**, and the 2024 SPD 15 revision now *requires* it. The
ERR-358 appendix is the natural home. EIB-269 proved this is feasible;
the ask is to make it standard in the final annual report.

**The cancellation raises the stakes.** USDA's September 2025 decision
to end the CPS-FSS means ERR-358 — the report series that has published
annual U.S. food insecurity rates since 1999 — is itself in its final
cohort. If disaggregated race/ethnicity tables are not added to the
remaining ERR-358 output, they may never be published under USDA ERS
auspices at all. Separately, the 2020–2024 pool is now the **definitive
five-year CPS-FSS window** for subgroup analysis; any larger pool will
require cross-walking to a different data source.

---

## Method notes

Race and Hispanic origin are assigned at the household level from the
CPS reference person. The 7-category typology (`raceeth8`) is mutually
exclusive: Hispanic is assigned first across any race, then single-race
non-Hispanic households, then multiracial non-Hispanic. This follows
SPD 15 tabulation conventions.

All 95% confidence intervals reported in Table 9 of the reproducibility
`.do` use a design-effect-adjusted approximation:

```
SE = sqrt( p * (1 - p) / N * deff ),  deff = 1.5
```

The 1.5 design effect is used here as a conservative approximation for
CPS subgroup proportions in the absence of a group-specific generalized
variance parameter. For the AIAN point estimate of 30.9% (N = 272), the
adjusted 95% CI is [24.1, 37.6] percentage points.

Aggregate totals reproduce **ERR-358 Table 1A** exactly: 134,062K
households, 18,341K food-insecure, 13.7% food insecurity rate. The
4-category race/ethnicity breakdown reproduces **ERR-358 Table 2** to
within rounding.

---

## 280-character tweet version

> 30.9% of AIAN households were food insecure in the final 2024 CPS-FSS — highest point estimate of any single-race group. Never in the USDA annual report: AIAN gets aggregated into "Other, non-Hispanic." USDA cancelled the CPS-FSS in September 2025. #FoodInsecurity #SPD15

(266 chars.)

---

## Source

Author's calculations from the December 2024 CPS Food Security
Supplement (U.S. Census Bureau). All estimates use household supplement
weights. Full reproducibility package available at:
https://matthewpatrickrabbitt.com/food-security-fridays.html