Food stamps by the numbers: using USDA SNAP participation data to track hunger and benefit policy

The Supplemental Nutrition Assistance Program is the largest food assistance program in the United States, reaching 42 million people at its 2013 peak and touching that same ceiling again during the COVID-19 pandemic. Every month since 1969, the USDA Food and Nutrition Service has published state-level participation and benefit data—a 57-year unbroken series that records exactly how food assistance expands during recessions, contracts when eligibility tightens, and responds to emergency policy decisions made in Washington. The data is free, public, and almost always underused.

What SNAP is

SNAP is authorized under 7 U.S.C. 2011 et seq., the Food and Nutrition Act of 2008 (originally the Food Stamp Act of 1964). It is a federal-state partnership: the federal government funds 100 percent of benefit costs and splits administrative costs with states, while states handle eligibility determination, enrollment, and issuance. This structure means that federal policy sets the floor for who qualifies and how much they receive, while states retain discretion over categorical eligibility expansions, outreach intensity, and work requirement waiver applications.

For most of the program's history it was called Food Stamps. The 2008 Farm Bill renamed it the Supplemental Nutrition Assistance Program and replaced paper coupons with Electronic Benefit Transfer (EBT) cards. The name change was partly administrative and partly political: the word “stamps” carried stigma that advocates believed deterred eligible households from enrolling. The policy mechanics did not change.

At the program's statistical peak in December 2012 and January 2013, 47.8 million individuals participated. During the COVID-19 emergency, participation climbed back to 42.7 million by April 2020 and held elevated through the emergency allotment period. The program serves roughly one in eight Americans in an average non-recession year, making it the most data-rich window into food insecurity that the federal government produces.

The FNS data releases

The Food and Nutrition Service publishes several distinct data products, each with different geographic granularity and update frequency.

The primary product for historical analysis is the monthly state-level participation and issuance table. FNS posts this to its program data page as an Excel file or XLS download, updated roughly two months in arrears (January data appears in late March). The table covers every state and U.S. territory plus a national total row. It goes back to fiscal year 1969 for some series and to 1989 in full consistent form. This is the series most analysts use.

FNS also publishes annual summary reports aggregating participation by demographic category—household composition, income level, presence of elderly or disabled members, and gross income as a percent of the federal poverty level. These annual snapshots are more granular on the recipient side but do not have the state-by-state monthly resolution of the participation tables.

County-level data is available but substantially less granular. The county series covers annual averages rather than monthly snapshots, and several small counties are suppressed for privacy. For sub-state analysis, the county data is usable for broad patterns but not for tracking within-year variation.

A separate but related FNS release covers SNAP retailer authorization: the database of stores authorized to accept EBT cards. The retailer file includes store name, address, business type (supermarket, convenience store, farmers market, etc.), and authorization dates. It is a useful complement to participation data—retailer density determines whether benefits can actually be redeemed—and maps directly onto food desert research.

Data structure

The monthly participation table is organized as a panel: one row per state per month, with the following core fields:

Historical trajectory

The 57-year time series is a compressed history of American poverty policy. Each era of the program reflects specific legislative decisions and economic conditions.

The program launched federally in 1969 with 2.9 million participants. Participation grew steadily through the 1970s as eligibility rules were standardized nationally and the program expanded from its original pilot counties to all 50 states. By 1975, 17.1 million people were enrolled—a fivefold increase in six years, driven primarily by program expansion rather than economic deterioration.

The Reagan administration's Omnibus Budget Reconciliation Act of 1981 cut SNAP (then Food Stamps) by tightening eligibility and reducing benefit levels. Participation fell from 22.4 million in 1981 to 19.9 million in 1982. The 1981 cuts represent one of the few periods in the program's history where policy tightening drove a meaningful enrollment decline independent of economic improvement.

The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 introduced work requirements for able-bodied adults without dependents (ABAWDs), limiting their benefits to three months in any 36-month period unless they worked or participated in job training. Participation fell sharply—from 26.6 million in 1996 to 17.2 million by 2000. Both the requirements and the concurrent economic boom contributed to the decline. The 1996 PRWORA also made most legal immigrants ineligible for SNAP for five years after arriving, a restriction that was partially restored in subsequent Farm Bills.

The 2002 Farm Bill began reversing the 1996 restrictions, restoring eligibility for legal immigrants and making it easier for states to waive the ABAWD time limit in areas of high unemployment. Participation began rising from its 2000 low even before the 2007 recession began.

The Great Recession produced the program's steepest sustained growth: from 26.3 million in December 2007 to 47.6 million in December 2012. The 2009 American Recovery and Reinvestment Act temporarily increased benefit levels by 13.6 percent and suspended the ABAWD time limit nationally. Both changes were later allowed to expire: the benefit increase ended in November 2013, producing what advocates called the largest single benefit cut in SNAP history at the time.

The 2019 work requirement expansions under the Trump administration attempted to reinstate and narrow the ABAWD waiver criteria, tightening which counties qualified for exemptions based on unemployment rate thresholds. Implementation was partially blocked by federal courts. Participation had already been declining from the 2013 peak; by February 2020, enrollment stood at 36.9 million.

The COVID-19 emergency allotments, described below, pushed participation back up and expanded benefit levels dramatically. The February 2023 end of emergency allotments produced what was widely described as the largest single benefit cut in SNAP history—surpassing the 2013 ARRA expiration both in absolute dollar terms and in the speed of the reduction.

The COVID emergency allotment story

Section 2302 of the Families First Coronavirus Response Act (March 2020) authorized USDA to issue emergency allotments that brought every SNAP household up to the maximum benefit for their household size. Households already receiving the maximum were unaffected; households receiving less than the maximum—the majority, because SNAP benefits are income-scaled—received a supplement to reach the maximum.

The emergency allotments ran from March 2020 through February 2023—35 months. Total emergency allotment expenditure over this period was approximately $142 billion above what would have been issued under normal program rules. At peak, emergency allotments added roughly $95 per month to the average SNAP household's benefit.

The February 2023 expiration was not a gradual phase-out. On March 1, 2023, every SNAP household that had been receiving a supplement saw its benefits return to normal income-based levels. The average household lost approximately $95 per month. Food banks across the country reported sharp increases in demand within weeks of the cutoff. Several states had already ended their emergency allotments earlier—state participation was voluntary—providing a natural experiment: states that ended allotments early showed participation and food insecurity patterns that foreshadowed the national cliff.

The FNS monthly participation data captures this cliff precisely. Average benefit per person peaked in early 2021 (when the Thrifty Food Plan reevaluation also increased maximum benefits by 21 percent) and then dropped sharply in March 2023. Participation itself was slower to respond, as households remained enrolled even with reduced benefits, but the dollar series shows the cliff immediately.

State variation

SNAP is a federal program, but state policy choices create substantial variation in how it operates. Three levers drive most of the cross-state differences visible in the FNS participation table.

Work requirement waivers: states with areas meeting FNS unemployment thresholds can request waivers from the ABAWD time limit. The waiver coverage map changes monthly based on local unemployment data, and states vary substantially in how aggressively they apply for waivers. A state that covers only its highest-unemployment counties will see more ABAWDs lose benefits after three months than a state that covers its entire territory.

Broad-based categorical eligibility (BBCE): most states have adopted policies that extend SNAP categorical eligibility to households receiving any Temporary Assistance for Needy Families (TANF)-funded service, not just direct cash assistance. This effectively raises the gross income limit above 130 percent of the federal poverty level for some households and eliminates the asset test. States that have not adopted BBCE, or that adopted only the narrowest version, have lower effective eligibility ceilings.

State supplements: a handful of states add state-funded supplements on top of federal SNAP benefits. These are not captured in the FNS federal benefit tables, because they are state-funded separately. California has historically had one of the more substantial supplement programs. State supplements appear in state budget data rather than FNS tables, creating an undercounting of total food assistance in supplementing states.

How to access the data

The canonical source is the USDA FNS program data page at fns.usda.gov/pd/supplemental-nutrition-assistance-program-snap. The page organizes downloads by data type: participation and costs, annual summary, state activity reports, and retailer data.

The participation and issuance data is distributed as Excel files with one sheet per fiscal year or in multi-year annual summary workbooks. The workbooks are not consistently formatted across years—column headers shift, merged cells appear in some versions, and territorial data is sometimes in a separate tab. Batch processing requires normalizing these inconsistencies before concatenating into a clean panel.

FNS releases the monthly participation table roughly 60 days after the reference month closes. January data is available in late March. This lag reflects the time states need to finalize their certification records and submit to FNS for aggregation.

For researchers who want the full history without manual spreadsheet stitching, the USDA Economic Research Service (ERS) maintains a SNAP data atlas and supplemental downloads at ers.usda.gov/topics/food-nutrition-assistance/snap-data-system. The ERS version is more consistently formatted than the raw FNS downloads and includes some derived variables (participation rate as a percentage of the poverty population, for example) not present in the raw FNS tables.

Research use cases

Recession response: participation rate vs. unemployment rate

SNAP participation does not respond to unemployment simultaneously. The lag is typically four to six months: a worker who loses a job in January may exhaust savings, qualify for unemployment insurance, and delay applying for SNAP until April or May. This 6-month lag structure is visible clearly in the 2007–2009 recession data: BLS unemployment crossed 5 percent in late 2007, while SNAP participation did not accelerate sharply until mid-2008.

Plotting state-level participation rate against BLS Local Area Unemployment Statistics (LAUS) with a 6-month lag on unemployment produces a reasonably well-fitted relationship at the state level—with the notable exception of states that changed their categorical eligibility rules during the period, which creates discontinuities in the participation series that are not driven by unemployment at all.

This analysis requires joining the FNS state participation table to LAUS state unemployment estimates by state and month. Both series are publicly available in consistent machine-readable form. The FNS table gives total participants; dividing by Census population estimates gives a participation rate. Dividing SNAP participants by the BLS estimate of state civilians in poverty (from the Current Population Survey) gives the poverty-population participation rate, a more targeted denominator.

The 2023 emergency allotment cliff

The March 2023 emergency allotment expiration produced an immediate and measurable signal in the FNS data. Average benefit per person dropped from approximately $230 in February 2023 to approximately $180 in March 2023 nationally—a 22 percent decline in one month. The dollar cliff is unambiguous; no seasonal adjustment is needed because it is a policy change, not a seasonal variation.

Participation responded more slowly. Some households left SNAP in the months following the emergency allotment end, but the exit rate was gradual rather than cliff-like. This asymmetry—benefits fell faster than participation—is consistent with what researchers call the “persistence” of SNAP participation: once households enroll, they tend to remain enrolled even when benefits decline, as long as they still qualify. The cliff is therefore most visible in the benefit-per-person series rather than the participation series.

States that had ended their emergency allotments before February 2023 (Arkansas, Florida, Missouri, and others that terminated early) provide comparison groups. Their benefit series show earlier but similar cliffs, and their participation series show the same slow-exit pattern.

SNAP participation rate as a share of poverty population by state

The most analytically informative measure of state SNAP performance is not raw participation but the share of the poverty population that is enrolled. A state with 1 million people in poverty and 850,000 SNAP participants is performing very differently from a state with 1 million in poverty and 550,000 SNAP participants. The denominator—the poverty population—comes from the Census Bureau's American Community Survey 1-year estimates or the Current Population Survey Annual Social and Economic Supplement.

Across the country, state SNAP participation rates as a share of poverty population range from roughly 55 percent (several Sunbelt and Mountain West states) to 90 percent or above (some Northeastern and Midwestern states). The variation is not fully explained by poverty rate, unemployment, or household composition differences. It reflects administrative factors: the ease of online enrollment, language access for immigrant communities, the aggressiveness of state outreach to eligible non-participants, and the stringency of income verification at renewal.

States with low participation rates relative to their poverty population represent both a policy problem (eligible people are not receiving benefits they qualify for) and a research opportunity (the gap is measurable, it varies by county within states, and it is correlated with identifiable administrative variables).

Cross-references

SNAP data gains analytical power when joined to related federal datasets. Four cross-references are particularly productive.

USDA WIC data (Women, Infants, and Children) partially overlaps the SNAP population. Households with children under five are eligible for WIC based on income criteria that overlap but do not perfectly match SNAP eligibility. The USDA publishes WIC participation data at the same state-by-month granularity as SNAP. Because WIC covers a narrower demographic, changes in WIC participation that diverge from SNAP participation can signal shifts in the composition of the low-income population (more or fewer households with young children) that are not visible in SNAP alone.

Census ACS poverty data provides the denominator for participation rate calculations. The ACS 1-year estimates, available for states and counties above 65,000 population, give poverty counts by household composition type. Joining SNAP participation to ACS poverty by state and year produces the participation-rate-as-share-of-poverty series described above. The ACS poverty measure uses the official poverty line, not the Supplemental Poverty Measure, which means some SNAP-eligible households (particularly those above 100 percent but below 130 percent FPL) are not counted as poor in the ACS denominator even though they qualify for SNAP.

USDA SNAP retailer data provides the supply-side picture. The authorized retailer database—stores approved to accept EBT—covers approximately 260,000 retailers nationally. Joining retailer locations to participation data by ZIP code or county identifies areas where SNAP benefits are issued but few authorized retailers operate within a reasonable distance, a core operational definition of a food desert. Some Census Bureau food environment analyses use this join directly.

BLS CPI Food at Home component tracks the price of groceries over time and is the right deflator for converting nominal SNAP benefit levels to real purchasing power. The Thrifty Food Plan, which sets the maximum SNAP benefit, is periodically re-benchmarked to reflect actual grocery prices. But between benchmarking years, benefit increases may not keep pace with food inflation. Deflating the “average benefit per person” series by the CPI Food at Home index reveals that real purchasing power of SNAP benefits has fluctuated substantially even in years when nominal benefits appeared stable.

Limitations of the public release

The FNS monthly table has three significant limitations that affect what analyses are possible.

The data is state-level only. The public release does not include county or individual records. County-level annual averages are available separately but at lower frequency and with some suppression. This means sub-state analyses—looking at which neighborhoods or counties have high participation gaps relative to poverty rates—require either the county-level annual data (with its limitations) or supplemental data from state agencies, which is not consistently available across all 50 states.

Monthly averages mask within-month cycling. SNAP benefits are issued on EBT cards on staggered schedules—different households receive their benefits on different days of the month, typically keyed to Social Security number or case number. The FNS monthly participation count is an average across all reference dates in the month, which smooths over what is actually a highly lumpy benefit-issuance pattern. Research on food insecurity at the end of the SNAP month (the “SNAP cycle” phenomenon) requires administrative microdata held at the state level, not the FNS public release.

The participation count does not distinguish entrants from exits. A state with 100,000 average participants in January might have had 90,000 continuous participants plus 20,000 who enrolled and 10,000 who left, or it might have had stable enrollment with no movement at all. The gross entry and exit flows are not reported in the public table. This matters for evaluating the claim that work requirements reduce participation: if a requirement causes exits, participation falls; but if it also deters applications, participation falls differently, and the two mechanisms have different policy implications that cannot be distinguished without microdata.

Related writing: The Census data stack: using ACS tables to build neighborhood-level demographic profiles covers the American Community Survey—the primary source for poverty counts used as SNAP participation rate denominators.

Related writing: The $800 billion bailout: using SBA PPP data to trace who got pandemic relief covers the other major COVID-era federal benefit dataset—forgivable loans whose scale dwarfed even the SNAP emergency allotments.

Related writing: Counting the incarcerated: using BJS National Prisoner Statistics to track mass incarceration by state covers the BJS corrections dataset—another long-running federal series where state policy variation produces dramatically different outcomes within a shared federal framework.