Child Poverty in
America

Key terms

Federal poverty line

The Census Bureau's statistical definition of poverty — a set of money-income thresholds that vary by family size and composition. A family is "in poverty" if its pre-tax cash income falls below the applicable threshold. SAIPE uses this official definition (not the Supplemental Poverty Measure).

Child poverty rate (under 18)

The percentage of related children under age 18 living in families whose income falls below the federal poverty line. SAIPE variable SAEPOVRT0_17_PT. This is the headline metric throughout the atlas.

School-age poverty rate (5–17)

The percentage of related children ages 5 to 17 in families in poverty. SAIPE variable SAEPOVRT5_17R_PT. This is the specific denominator the U.S. Department of Education uses to allocate Title I, Part A funding to school districts.

All-ages poverty rate

The percentage of the total county population (children + adults) in poverty. SAIPE variable SAEPOVRTALL_PT. Reported alongside the child rate so the gap between the two is visible.

Child-to-adult concentration ratio

Child poverty rate divided by the all-ages poverty rate. A ratio above 1.0× means children are poorer than the overall population in that geography; this is almost universally the case in the United States. We compute and publish this as child_adult_ratio.

Median household income

Median money income of households. SAIPE variable SAEMHI_PT. Reported for context alongside the poverty rates.

Why SAIPE and not the ACS

Both SAIPE and the American Community Survey (ACS) publish county-level poverty statistics, and both are produced by the Census Bureau. We chose SAIPE for three reasons:

• Single-year timeliness. SAIPE produces single-year estimates for every county every year. ACS county-level estimates are 5-year averages, which blur year-over-year change.
• Model-based precision for small counties. SAIPE combines direct ACS survey data with administrative records (IRS tax returns, SNAP participation counts, Supplemental Security Income records, and population estimates) through a statistical model. This yields more reliable estimates for low-population counties than the raw ACS 5-year data can.
• Federal program alignment. SAIPE is the official source the Department of Education uses to allocate Title I funds to school districts. Our numbers match the numbers that drive real-world funding decisions.

SAIPE variables used

All variables come from the SAIPE 2023 single-year release (the most recent available vintage). Counties are identified by 5-digit FIPS codes (2-digit state + 3-digit county).

How state and national aggregates are computed

Rates cannot be summed across geographies — they are already ratios. Instead:

• State aggregates are fetched directly from the SAIPE API at the state level. This matches what the Census Bureau, Department of Education, and Annie E. Casey Foundation (KIDS COUNT) publish.
• National aggregates are fetched directly from the SAIPE API at the U.S. level (for=us:*). We do not sum or weight-average the state totals; we use the Census Bureau's published national estimate verbatim.
• Rankings (highest / lowest child poverty rate by state, by county) are computed at build time on the fetched values with no further smoothing.

Handling suppression & sentinels

SAIPE occasionally returns non-numeric sentinel values — empty strings, ., N, X, or negative flag values — for geographies where an estimate cannot be published. 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 they would otherwise render a number.

Why "food insecurity risk"? Child poverty as a structural proxy

This atlas is titled a child poverty atlas, but it is framed as a child food insecurity risk atlas. The distinction matters. Direct county-level child food insecurity data is not publicly available; the USDA measures food insecurity at the national and (with limits) state level through the Current Population Survey's Food Security Supplement, and Feeding America's Map the Meal Gap publishes modeled county-level estimates but restricts bulk reuse.

Child poverty is the strongest structural predictor of child food insecurity. USDA ERR-337 (Household Food Security in the United States in 2023) reports that 17.3% of households with children were food insecure in 2023, versus 13.5% for all households — and the gap is driven by income. A county with twice the child poverty rate of another county will, in expectation, have substantially higher child food insecurity. We publish poverty numbers (which are measured precisely, annually, at the county level, and carry federal funding consequences) and explicitly label them a proxy for food insecurity risk, rather than publish modeled food insecurity estimates we did not produce ourselves.

Limitations

• Poverty is not food insecurity. Many food-insecure households are above the federal poverty line; many households below the line are not food insecure in a given month. Poverty is a structural predictor and a strong proxy, not a direct measurement. See above.
• Official poverty measure. SAIPE uses the federal poverty line, which many researchers argue understates hardship because it does not adjust for regional cost of living, housing, medical out-of-pocket, or in-kind transfers. The Supplemental Poverty Measure (SPM) addresses some of these but is not available at the county level.
• Model-based estimate. SAIPE blends ACS survey responses with administrative records through a statistical model. Estimates carry a margin of error; published rates for small-population counties in particular should be read as indicative rather than precise.
• 2023 vintage. SAIPE releases run roughly 16–18 months behind the reference year. Post-pandemic benefit changes (end of expanded Child Tax Credit, SNAP emergency allotments ending in March 2023) are partially but not fully reflected in the 2023 estimate.
• Territories excluded. SAIPE covers the 50 states, the District of Columbia, and Puerto Rico. Puerto Rico uses distinct safety-net programs (NAP rather than SNAP) and is omitted from this atlas to keep cross-state comparisons apples-to-apples.

Reproducibility

The entire atlas is driven by one pipeline script: scripts/child-poverty/build-counties.mjs. It fetches the SAIPE API, computes derived metrics (child-to-adult ratio, severity index), writes one JSON file per state and one per county collection, and rebuilds the national rankings. Emitted files live in src/data/child-poverty/. Re-running the pipeline against a newer SAIPE 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.