Food aid is a “dependency trap.” A forgiven business loan is just good policy.
The same word, dependency, gets attached to the smallest transfers and withheld from the largest. It is aimed at the family on food stamps and never at the corporation, the large farm, or the top-bracket taxpayer collecting far more.
Every figure on this page is documented and graded FACT: what food assistance costs and who it serves, whether recipients work, what work requirements actually do, and how much larger the transfers are that flow the other way. We carry the honest caveats too, including that most of the 2008 bank bailout was repaid. What we pose, and do not stamp, is why one kind of help gets called dependency and the other never does.
What this page argues
The argument against programs like food stamps is that they breed dependency, so recipients should be pushed to work. The data cuts against the premise: in 2023 food assistance served 42 million people a month, four in five of them in a household with a child, an elderly person, or someone with a disability, and most working-age recipients already work. Where work requirements have been tried, as in Arkansas Medicaid, they cut thousands from coverage and produced no rise in employment.
Meanwhile the transfers that flow upward are far larger and never carry the label. Federal farm subsidies run over $30 billion a year, most of it to the biggest operations. Pandemic business loans were forgiven at over $757 billion, and the best peer-reviewed estimate is that only about a quarter to a third of that money reached workers. The tax break on capital gains and dividends alone was worth about $225 billion in one year, with 95 percent of the benefit going to the top fifth of earners. None of it is called dependency. That asymmetry, the same word applied to the smaller transfer and withheld from the larger one, is what this page documents.
The same investigation, restaged one beat at a time. Step through it here, or present it fullscreen.
Welfare Is “Dependency” for the Poor, a Subsidy for Everyone Else
The same word tracks the recipient's power, not the size of the transfer.
The record, claim by claim
SNAP cost $113 billion in FY2023 and served 42 million people a month, four in five in a household with a child, an elderly person, or someone disabled
FACTIn fiscal year 2023 the Supplemental Nutrition Assistance Program (SNAP, formerly food stamps) cost $113.2 billion total and served an average of 42.2 million people a month. Children were 39 percent of participants, older adults 20 percent, and people with disabilities 10 percent. Four in five SNAP households (79 percent) contained a child, an elderly person, or a nonelderly person with a disability, and 73 percent had gross income at or below the poverty line. The program overwhelmingly reaches children, the old, the disabled, and the working poor.
86 percent of non-disabled working-age SNAP households had earnings in 2021 — most recipients already work
FACTThe dependency framing assumes recipients do not work. Among 2021 SNAP households that included a non-disabled working-age adult, 86 percent had earnings during the year, per the Center on Budget and Policy Priorities' analysis of Census data. More than half work while receiving benefits, and over 80 percent work in the year before or after. Participation is typically temporary, filling gaps between or during low-wage jobs, and the share of SNAP households with earnings in a typical month rose from 19 percent in 1990 to 28 percent in 2020.
Arkansas's Medicaid work requirement cut about 18,000 people from coverage in months and produced no rise in employment (NEJM, 2019)
FACTArkansas became the first state to impose a Medicaid work requirement in June 2018. By April 2019, when a court halted it, about 18,000 adults had lost coverage. The peer-reviewed study by Sommers and colleagues in the New England Journal of Medicine (2019) found no increase in employment over 18 months. Coverage losses came largely from confusion and reporting burdens, even though most of the target group already met the requirement or qualified for an exemption. It is the strongest test we have of whether work rules produce work: they cut enrollment instead.
The 2023 debt-ceiling deal raised SNAP's work-requirement age to 54 — and CBO estimated it would slightly increase SNAP spending, not cut it
FACTThe Fiscal Responsibility Act of 2023 (H.R. 3746) raised the age at which able-bodied adults without dependents face SNAP time limits from 49 to 54, while adding exemptions for veterans, homeless people, and former foster youth. The Congressional Budget Office estimated the combined changes would slightly increase SNAP spending, by about $2.1 billion over ten years, because the new exemptions outweighed the age expansion. The savings rationale did not survive the agency's own scoring; the push to expand work rules is recurring regardless (the 2018 farm-bill fight, this deal, Medicaid waivers).
Federal farm subsidies run over $30 billion a year, and the top 10 percent of recipients collected roughly three-quarters of commodity payments
FACTFederal farm-support programs cost taxpayers more than $30 billion a year, the largest being crop insurance (about $17.3 billion in total program cost in 2022). The money concentrates at the top: from 1995 to 2021 the top 10 percent of recipients received over 78 percent of commodity subsidies and the top 1 percent got 27 percent, per the Environmental Working Group's database. The Government Accountability Office found the top 1 percent of crop-insurance policyholders collected 22 percent of premium subsidies in 2022. Many recipients are large, high-income operations. None of this is described as dependency.
PPP forgave about $757 billion in loans, and MIT's Autor et al. found only 23–34 percent reached workers — about three-quarters of the money went to the top income fifth
FACTThe Paycheck Protection Program disbursed roughly $790 to $800 billion, of which about $757 billion was forgiven, a forgiveness rate over 95 percent with minimal repayment scrutiny. The peer-reviewed study by David Autor and colleagues (Journal of Economic Perspectives, 2022) found only 23 to 34 percent of PPP dollars went to workers who would otherwise have lost pay. The rest flowed to business owners, shareholders, creditors, and suppliers, with about three-quarters of the funds accruing to households in the top income fifth. Loans were forgiven at scale; food assistance is means-tested to the dollar.
TARP disbursed $443.5 billion to rescue banks and firms in 2008 — a backstop no household gets, though most of it was repaid
FACTIn response to the 2008 financial crisis, the Treasury disbursed $443.5 billion through the Troubled Asset Relief Program to stabilize banks, the insurer AIG, and automakers. The honest caveat, carried here rather than buried: TARP was largely repaid, with the Treasury recovering $425.5 billion and the GAO putting the lifetime net cost around $31.1 billion. So this is best read as socializing risk and providing a backstop unavailable to individuals, not as a giveaway on PPP's scale. The point that survives is the double standard in terms: when a large firm faces ruin, the state absorbs the risk; when a family does, it gets a work requirement.
The capital-gains and dividend tax break was worth about $225 billion in 2024, with 95 percent of the benefit going to the top income fifth
FACTTax breaks that flow upward dwarf means-tested aid and are never called dependency. The preferential rate on capital gains and dividends was about a $225 billion tax expenditure in 2024. CBO's distributional analysis found 95 percent of its benefit went to the top income fifth, and 75 percent to the top 1 percent. The mortgage-interest deduction ($25.4 billion in FY2024) sends roughly 60 percent of its benefit to households earning over $200,000. These are federal money spent through the tax code, and they tilt to the top, but the word dependency never attaches to them.
The label tracks the recipient's power, not the size or the economics of the transfer
SOME SMOKEWhere it lands. The asymmetry above is documented fact: the largest transfers flow up and carry no stigma, while the smallest flow to the poor and get branded dependency and policed with work rules that do not produce work. What we stop short of is stamping a motive on any one official. A defender can argue the cases differ, that a business loan preserves jobs, that farm supports stabilize the food supply, that tax preferences encourage investment. Those arguments exist, and we carry them. But they are arguments about economics, and dependency is a claim about character, applied to one group and not the others. When the character claim tracks the recipient's power rather than the dollars or the incentives, a reasonable person is entitled to ask whether the principle was ever about self-reliance at all. We grade that SOME SMOKE: a documented double standard, and a motive that is posed, not proven.
- Synthesis of the USDA, CBO, GAO, Autor et al., and Treasury figures above; motive is posed, not asserted
Where the evidence is strong, and where it stops
- The dollar figures are not in dispute. SNAP's cost and caseload, the farm-subsidy concentration, PPP forgiveness, TARP, and the capital-gains preference all come from USDA, CBO, GAO, the Treasury, and a peer-reviewed study. These are the government's own numbers.
- Error rate is not fraud rate. SNAP's FY2025 payment-error rate was 10.6 percent, but that measures administrative accuracy, mostly unintentional over- and under-payments, not theft. Retailer trafficking is about 1.6 percent of benefits. We do not claim a single precise fraud number, because none exists.
- TARP was mostly repaid. Its net cost was on the order of $31 billion, not $443 billion. It belongs here as a contrast in rescue terms, not as a giveaway matching PPP. The strongest upward-transfer evidence is PPP forgiveness, farm-subsidy concentration, and the capital-gains break.
- The motive is the open question. That the largest transfers escape the dependency label while the smallest attract it is documented. Why is what we pose. We do not declare it a conscious strategy on any individual's part.
The word decides who gets help and who gets a work requirement
“Dependency” is not a neutral description. It is the word that justifies cutting one kind of help and the silence that protects another. When it is aimed only at the poor, it turns a question about economics into a judgment about character, and the judgment always runs downhill. That is why this page belongs in The Cover Story: a principle of self-reliance that exempts the powerful is not a principle, it is a cover for who gets to keep the money. The scale of the upward transfers is the subject of What did we get for $40 trillion?, which totals the bailouts, subsidies, and impunity built for the top. For a straight defense of food assistance on its own terms, see our SNAP page. Read together, they answer the question this one only poses.
Questions worth taking seriously
Isn't there real fraud and waste in food stamps?
Some, and we say so. Retailer trafficking is about 1.6 percent of benefits, and the payment-error rate is 10.6 percent, but that error rate measures administrative accuracy, mostly unintentional over- and under-payments, not theft. Even taking the worst reading, the dollars involved are a fraction of what flows, unscrutinized, through forgiven business loans and upward tax breaks. The point is not that SNAP is flawless; it is that the scrutiny is aimed at the smallest transfers and waved through on the largest.
Aren't business loans and tax breaks different from welfare? They keep people employed.
That is the argument, and we carry it, but it is an argument about economics, and it cuts both ways. The best study of PPP found only about a quarter to a third of the money reached workers; the rest went to owners and shareholders. Farm subsidies concentrate at the top, and the capital-gains break flows 95 percent to the top fifth. You can defend any of these on incentive grounds. What you cannot do is call them self-reliance while calling a working parent's food aid dependency. The label is doing moral work the economics does not support.
If you are named on this page
If you are named on this page, or are a party materially affected by the claims made here, and you wish to respond, correct the record, or add context, use the Contact page. Responses are published verbatim alongside the original claim, with the sender identified and the date of receipt. The channel stays open for the life of the page.
This site aggregates and grades a record that other outlets and primary sources have already put on the record. Every FACT-graded claim above is sourced to court filings, government reports, sworn whistleblower disclosures, published investigative journalism, or named-source statements. The citations are the accountability mechanism; this section is how you get on the record too.
The record
- USDA FNS — Characteristics of SNAP Households, FY2023 (report)
- CBPP — Most working-age SNAP participants work, but often in unstable jobs
- NEJM — Sommers et al., Medicaid work requirements in Arkansas (2019)
- Congress.gov — Fiscal Responsibility Act of 2023 (H.R. 3746)
- EWG — Farm Subsidy Database (concentration of payments)
- GAO — Crop insurance premium-subsidy concentration
- Autor et al. — The $800 Billion Paycheck Protection Program (JEP, 2022)
- U.S. Treasury — TARP program data
- GAO — TARP status and lifetime net cost
- CBO — The Distribution of Major Tax Expenditures
- Treasury — Tax Expenditures, FY2025