Senior Food Insecurity
Risk Atlas

Key terms

Senior poverty rate

The percentage of people aged 65 or older whose family income falls below the federal poverty line. ACS Subject Table S1701, variable S1701_C03_010E. This is the headline metric throughout the atlas.

Seniors living alone

The count of adults 65+ living in a one-person household — the sum of ACS detail table variables B09020_015E (males 65+ living alone) and B09020_018E (females 65+ living alone). The share is this count divided by B09020_001E (all people 65+ in the relationship-to-householder universe).

Senior SNAP take-up

The percentage of households containing one or more members aged 60 or older that received SNAP benefits in the past 12 months. ACS Subject Table S2201, variable S2201_C04_002E. SNAP defines a "senior household" by a 60+ threshold (not 65+), so we do likewise here.

Senior median household income

Median household income when the householder is aged 65 or older. ACS detail variable B19049_005E. Compared with B19049_001E (all householders) to compute the senior income gap.

Social isolation index

A blended 0–1 indicator that averages normalized senior poverty and normalized share of seniors living alone. Defined below under Social isolation index. This is a visualization and ranking aid, not a direct measurement of food insecurity.

Why the ACS 5-year dataset

Senior-specific metrics at the county level are only available from the ACS. SAIPE — which anchors our Child Poverty Atlas — publishes age-disaggregated data only for children (under 18 and ages 5–17). For 65+ we have to go back to the underlying survey.

• 5-year estimates, not 1-year. ACS publishes 1-year estimates only for areas with 65,000+ people, which excludes most rural counties. 5-year estimates cover every county in the United States and have smaller margins of error, at the cost of blurring year-over-year change.
• 2019–2023 vintage. The most recent 5-year release at the time of this build. It spans the pandemic and reflects both pandemic-era benefit expansions (expanded SNAP allotments, emergency CSFP flexibilities) and the 2023 roll-back.
• Subject tables + detail tables. We use S1701 and S2201 for the headline rates (poverty, SNAP), and B-tables for the demographic counts (household relationship, age-of-householder income). Subject tables are already expressed as percentages; detail tables are counts we derive shares from.

ACS variables used

All variables come from the ACS 5-year 2019–2023 release. Counties are identified by 5-digit FIPS codes (2-digit state + 3-digit county).

Derived metrics

• Senior alone share = (B09020_015E + B09020_018E) ÷ B09020_001E. A 0–1 fraction that is null if either numerator components or the denominator is missing.
• Senior income gap = B19049_001E − B19049_005E. Reported in dollars; positive values indicate that senior-headed households earn less than the overall household median, which is almost universally the case.

Social isolation index

A blended 0–1 indicator we compute from two normalized components:

1. Normalized senior poverty — a linear rescaling of senior_poverty_rate from [3%, 25%] to [0, 1], clipped outside that range. The anchors are chosen so the healthiest U.S. counties sit near 0 and the poorest sit near 1 without compressing the middle.
2. Normalized living-alone share — a linear rescaling of senior_alone_share from [18%, 40%] to [0, 1], clipped outside that range. The anchors reflect the empirical distribution in the ACS: few counties sit below 18% or above 40%.

The index is the simple average of the two. If either input is null the index is null; we do not impute. The index is a visualization and ranking aid — it is not a direct measurement of food insecurity, nor is it peer-reviewed. Treat it as a way to read two correlated signals at a glance, not as a prediction.

How state and national aggregates are computed

Rates cannot be summed across geographies — they are ratios. The pipeline handles aggregation as follows:

• State aggregates are fetched directly from the ACS 5-year API at the state level. This is the Census Bureau's published value — no re-weighting.
• National rankings weight state rates by the appropriate universe: senior_snap_rate weighted by senior_hh_total; all other rates weighted by seniors_total. This avoids small-population states dominating the top-10 lists.
• National aggregates are the weighted roll-ups of the 51 state values, not a separate ACS "us:*" call — because several of our derived metrics (income gap, alone share, isolation index) are computed post-fetch. State values themselves come straight from the ACS.

Handling suppression & sentinels

ACS occasionally returns sentinel values — negative flag codes, empty strings, or explicit non-numeric markers — for geographies where an estimate cannot be published (too few respondents, disclosure avoidance, etc.). Our pipeline treats every sentinel as null and propagates it through every downstream derived metric. No imputation, no interpolation, no invented values. Counties with missing values display "—" everywhere a number would otherwise appear.

Why "food insecurity risk"? Poverty & isolation as structural proxies

This atlas does not publish a direct county-level food-insecurity rate for seniors. USDA measures food insecurity through the Current Population Survey Food Security Supplement, which releases estimates only at the national and state level. Feeding America's State of Senior Hunger reports (Ziliak & Gundersen) provide the best state-level senior food-insecurity numbers; county-level estimates exist from proprietary models but restrict bulk reuse.

Senior poverty, living alone, and low SNAP take-up are the three most consistent structural predictors of senior food insecurity in the peer-reviewed literature. USDA ERR-337 (Household Food Security in the United States in 2023) documents that food insecurity among households with an elderly person is driven primarily by income; living alone amplifies the effect by removing household-level risk-pooling. We publish ACS numbers (measured annually, at the county level, with published margins of error) and explicitly label them a proxy for senior food insecurity risk, rather than publishing modeled food-insecurity estimates we did not produce ourselves.

Limitations

• Poverty + isolation ≠ food insecurity. Many food-insecure seniors live above the federal poverty line or with others. Poverty and living-alone share are structural risk factors, not direct measurements.
• 60+ vs. 65+. ACS S2201 uses a 60+ threshold for SNAP-household definitions, while S1701 poverty, B09020 relationship, and B19049 income use 65+. We follow each table's native cutoff and disclose it rather than force-align them.
• Isolation index is a visualization aid. The blended 0–1 score is not peer-reviewed and cannot be compared across studies. Use it to scan the map, not to benchmark.
• 5-year average. ACS 5-year estimates pool five calendar years of survey responses. Policy shifts that happened inside the window (pandemic SNAP allotments, end of emergency allotments in early 2023) are partially reflected.
• High SNAP take-up is not "worse." States where a large share of eligible senior households enroll in SNAP indicate strong program reach, not deeper food insecurity. The map flips the SNAP legend accordingly (higher = better).
• Territories excluded. Puerto Rico uses the Nutrition Assistance Program (NAP) rather than SNAP and is omitted from this atlas for cross-state comparability.

Reproducibility

The entire atlas is driven by one pipeline script: scripts/senior-hunger/build-counties.mjs. It fetches the ACS Subject and Detail tables, computes derived metrics (senior alone share, income gap, social isolation index), writes one JSON file per state and one per county collection, and rebuilds the national rankings. Emitted files live in src/data/senior-hunger/. Re-running the pipeline against a newer ACS vintage regenerates every page.

Pipeline metadata — dataset vintage, build timestamp, variable list — is embedded in every emitted JSON file and surfaced on the report's index page.