Social Welfare Policy

Nine hundred and seventy-three dollars

In fiscal year 2024, the maximum monthly SNAP allotment for a household of four in the forty-eight contiguous states was 973 dollars. That number is the end of a chain. Almost nobody who receives it knows the chain exists, and most people who argue about SNAP in public argue about the number without knowing what would have to change to move it. You are going to learn to walk the chain backward, because that skill is the whole of policy analysis and it takes about fifteen minutes once you know the moves.

Start at the card. A household applies at a county or state office. A worker enters income, household size, shelter cost, and a few other items. Software computes the benefit and it lands on an electronic benefit transfer card. The household experiences the policy as a balance.

One step back sits a formula, and it is short: the household gets the maximum allotment for its size, minus thirty percent of its net income. The thirty percent is not a natural constant. It is a legislative judgment, made decades ago, that a household should be expected to spend about thirty percent of its own money on food before assistance begins. Change that fraction to twenty-five and every SNAP household in the country gets more money without a single new eligibility rule.

Another step back sits the maximum allotment itself, and here is where it gets interesting. The maximum is not a figure Congress picks. It is set at the cost of a market basket called the Thrifty Food Plan, calculated by the United States Department of Agriculture, adjusted each October first for food price inflation.

The point: A benefit amount is almost never the policy. It is the visible output of a formula, and the formula is the policy.

The 2021 reevaluation, and why it was a fight

The Thrifty Food Plan is a real basket of real foods priced against national data, and for decades it was reevaluated only in the sense of being reindexed for inflation. Its 2006 version was still shaping benefits in 2020. The 2018 Farm Bill instructed USDA to reevaluate the plan by 2022 and every five years after that.

USDA delivered the reevaluation in August 2021, and it raised the real cost of the plan by roughly twenty-one percent. Because the maximum allotment is pinned to that cost, benefits rose for every SNAP household in the country at once. No new eligibility category was created. No caseload changed. One recalculated market basket moved billions of dollars.

The politics arrived immediately, and they were not about groceries. They were about whether an executive agency should be able to move an entitlement's cost that far without a new vote. In the debt limit negotiations that produced the Fiscal Responsibility Act of 2023, one live proposal was to constrain future reevaluations so that they could not increase real costs. That proposal did not survive in that form. What did pass in that law was a change to the age range subject to SNAP's time limit for able-bodied adults without dependents, raised in stages from forty-nine to fifty-four, alongside new exemptions for veterans, people experiencing homelessness, and young people aging out of foster care.

Notice what just happened. A single statute simultaneously narrowed eligibility for one group and widened it for three others, and both changes are invisible in the benefit formula. If you had studied only the formula, you would have missed the entire 2023 fight.

Why this matters: Policy changes arrive in at least three different places, the benefit formula, the eligibility rules, and the administrative process, and a reader who watches only one of them will be repeatedly surprised.

The six questions

Here is the anatomy. Ask these six questions of any program and you will know more about it than most people who have opinions about it. Try them on something you already half know, unemployment insurance or Medicaid, as you read.

One: who is eligible, and by what test? Categorical eligibility means membership in a defined group does it: you are over sixty-five, you are a veteran, you are a child. Means testing means an income or asset screen does it. Most American programs use both, which is why so many people who feel poor are told they do not qualify.

Two: what form does the benefit take? Cash, a restricted voucher like SNAP or a housing choice voucher, an in-kind service like Head Start, an insurance guarantee like Medicare, a tax reduction like the earned income tax credit, or a rule that costs the government nothing and costs somebody else money, like a minimum wage. These are not interchangeable. A voucher and cash of the same value produce different behavior, different politics, and different fraud narratives.

Three: how is it financed? An open-ended entitlement pays everyone who qualifies, so its cost rises automatically in a recession. A block grant sends a fixed sum, so its real value falls with inflation and it cannot expand when need does. A dedicated payroll tax, as with Social Security and Medicare Part A, creates a sense of earned ownership that general revenue funding does not. A federal-state match rewards states for spending, and its match rate is one of the most consequential numbers in American social policy.

Four: who administers it? A federal agency mailing checks, a state agency under a federal plan, a county office, a private contractor, or a nonprofit under contract. Each layer adds discretion and delay.

Five: is it indexed? Social Security benefits rise with a consumer price index each January. The SSI asset limit of 2,000 dollars for an individual has not moved since 1989 and is not indexed to anything. A program that is not indexed shrinks every year without anyone voting to shrink it, which is the quietest way to cut a benefit that has ever been devised.

Six: who is excluded, and by which specific rule? Not who is poor and unserved in general, but which sentence does the excluding. This is the question that turns compassion into analysis.

Gilbert and Terrell's four dimensions

The six questions above are a practitioner's version of a framework you will meet in most policy textbooks. Neil Gilbert and Paul Terrell organize any welfare policy along four dimensions of choice, and it is worth learning the vocabulary because committee reports and journal articles use it.

Dimension	The question it asks	SNAP's answer	Medicare's answer
Allocation	Who gets it?	Households under income and asset tests, with categorical routes in	Nearly everyone at sixty-five, plus certain disabled workers
Provision	What do they get?	A restricted electronic benefit usable only for food	Insurance coverage for defined services
Delivery	Who hands it over?	State agencies under federal rules, food purchased at private retailers	Federal program, care delivered by private providers and plans
Finance	Where does the money come from?	Federal general revenue, open-ended entitlement	Part A from a dedicated payroll tax, Part B from premiums and general revenue

Read across the SNAP row and you can predict its politics. Restricted provision means recurring arguments about what people buy. Private retail delivery means a grocery industry constituency that defends the program. Open-ended entitlement financing means the caseload and cost rise in recessions, which is a feature to its defenders and an alarm to its critics.

Key idea: Design choices are not neutral plumbing. Each one manufactures a constituency, a critique, and a failure mode, and you can often predict all three from the design alone.

The four places policy actually lives

Students reasonably assume that to know a policy you read the law. You should read the law. It will usually tell you less than you expected.

Statute. An act of Congress or a state legislature. It sets the frame, the money, and the boundaries, and then delegates. The Food and Nutrition Act does not say 973 dollars anywhere.

Regulation. An agency, acting under delegated authority, publishes a proposed rule in the Federal Register, takes public comment for a stated period, and then publishes a final rule with responses to the comments it received. Rules are law. They are also where most of the operational detail is, and they are far easier to change than statutes, which is why every change of presidential administration produces a wave of rulemaking.

Guidance and manuals. Below regulation sits an enormous layer of program instructions, state plan templates, information memoranda, and eligibility manuals. These are not law and cannot lawfully contradict the regulation, but they are what a caseworker actually reads. If you want to know why an applicant was denied, the answer is usually in a manual.

Street-level discretion. Michael Lipsky's 1980 book Street-Level Bureaucracy made the argument that has held up best in this field: public employees who work directly with people, under chronic resource scarcity and impossible caseloads, develop routines and rationing devices, and those routines become the policy as citizens experience it. Whether an office schedules interviews at nine in the morning or at four in the afternoon is nowhere in any statute, and it determines who can keep an appointment while holding a job.

Worth holding on to: The gap between a policy as written and a policy as delivered is not corruption or failure. It is a predictable product of discretion under scarcity, and it is where a great deal of social work practice happens.

Running the anatomy on a second program

Do it once more, quickly, on Supplemental Security Income, so you can see the frame travel. Eligibility: categorical, in that you must be aged sixty-five or older, blind, or disabled, and means-tested, in that countable resources must fall under 2,000 dollars for an individual. Benefit form: a monthly federal cash payment, which many states supplement. Financing: federal general revenue, not payroll taxes, which distinguishes it sharply from Social Security retirement and disability benefits despite both being run by the same agency. Administration: the Social Security Administration, federally, with state disability determination services making the medical decisions. Indexation: the benefit rises with the same cost of living adjustment as Social Security, but the asset limit does not move at all. Exclusion: most non-citizens, people with resources just over the limit, and, in practice, many people whose disabilities are real but hard to document.

Six questions, ninety seconds, and you now understand why a person on SSI can be afraid of a modest inheritance and why disability advocates have spent thirty years asking Congress to raise a number that has not moved since 1989.

Common misconceptions

• Social welfare policy means programs for poor people. It includes Social Security, Medicare, the mortgage interest deduction, and employer health insurance tax exclusion, all of which flow substantially to households well above the median. Restricting the term to means-tested aid quietly concedes the argument before it starts.
• If you read the statute you know the policy. The statute delegates. The regulation implements. The manual instructs. The worker decides. You need all four.
• Benefit levels are set by legislators voting on amounts. Most are set by formulas and indexes. The most consequential policy changes in the last fifty years include several that only altered how a number is calculated.
• An unindexed benefit is stable. It is a cut on a timer.
• Analysis means having a position. Analysis means being able to say which sentence would have to change, in which document, to produce a different result.

What to carry forward

• Every benefit amount is the output of a formula; find the formula before you argue about the amount.
• The six questions are eligibility, benefit form, financing, administration, indexation, and exclusion.
• Gilbert and Terrell's four dimensions are allocation, provision, delivery, and finance, and they predict a program's politics.
• Policy lives in statute, regulation, guidance, and street-level discretion, and the last one is where clients meet it.
• Entitlement, block grant, dedicated tax, and federal-state match are four financing designs with four different behaviors in a recession.

Sources

1. U.S. Department of Agriculture, Food and Nutrition Service. (n.d.). Supplemental Nutrition Assistance Program (SNAP). fns.usda.gov
2. U.S. Department of Agriculture, Food and Nutrition Service. (2021). Thrifty Food Plan, 2021. fns.usda.gov
3. Social Security Administration. (n.d.). Understanding Supplemental Security Income. ssa.gov
4. Lipsky, M. (1980). Street-Level Bureaucracy: Dilemmas of the Individual in Public Services. Russell Sage Foundation. en.wikipedia.org
5. Office of the Federal Register. (n.d.). Federal Register: The Daily Journal of the United States Government. federalregister.gov

43 Elizabeth I, chapter 2

In 1601 the English Parliament passed an act that did one plain administrative thing: it made every parish responsible for its own poor and gave the parish the power to tax local property to pay for them. Justices of the peace appointed overseers of the poor. The overseers levied a rate on landholders, collected it, and distributed it. There was no national fund and no national standard. Your relief depended on the parish you belonged to and on what its ratepayers were willing to bear.

That act is the ancestor of the American system, and you can find its bones in a state eligibility manual written last year. It is worth two hours of your attention not out of antiquarianism but because it installed three ideas that have never left: that relief is local, that the poor must be sorted before they are helped, and that help must not be more attractive than work.

The sort came in three parts. The impotent poor, meaning the aged, the sick, the blind, the lame, and children without support, were to be relieved. The able-bodied poor were to be set to work, with materials such as flax, hemp, and wool supplied by the parish. Those who refused work, described in the statutes of the period as sturdy beggars and vagabonds, were to be punished, whipped, and sent to a house of correction.

Look at that division and then look at any American benefit system. SSI and Social Security disability separate people who cannot work from people who can. TANF attaches work requirements to able-bodied adults. SNAP has a time limit for able-bodied adults without dependents. The vocabulary is different and the sort is the same one made in 1601.

Key idea: The oldest and most durable move in social welfare policy is not the giving of aid. It is the classification of the applicant into a category that determines what aid means.

Settlement, and the invention of the residency rule

If the parish pays, the parish will want to know who belongs to it. The Settlement Act of 1662 gave parish officers the power to remove a newcomer within forty days back to the parish where he had a legal settlement, unless he rented property of sufficient value. In practice this meant that a laborer who moved to find work could be picked up and carted back to a village he had left years earlier.

Adam Smith, who was not a sentimentalist about the poor, wrote in 1776 that there was scarcely a poor man in England of forty years of age who had not at some point been cruelly oppressed by this ill-contrived law of settlement. It restricted labor mobility, which is why the classical economists disliked it, and it separated families, which is why almost everyone else did.

The rule crossed the Atlantic. American colonies and then towns practiced warning out, in which a newly arrived family was formally notified that the town accepted no responsibility for them. State residency requirements for welfare survived until 1969, when the Supreme Court struck them down in Shapiro v. Thompson, holding that a one-year waiting period penalized the constitutional right to interstate travel. Three hundred and seven years passed between the Settlement Act and that decision.

Speenhamland, and an argument that will not die

In May 1795, in the middle of a wartime spike in wheat prices, the magistrates of Berkshire met at the Pelican Inn at Speenhamland and adopted a scale. Relief would top up a laborer's wages according to the price of a gallon loaf and the number of people in his family. Similar scales spread across southern England.

The system was attacked, then and later, on the grounds that it let farmers pay below-subsistence wages and hand the difference to the ratepayers, that it removed the incentive to bargain, and that it therefore pauperized the countryside. Karl Polanyi's 1944 account in The Great Transformation gave this reading enormous influence. Then economic historians went back to the parish accounts. Mark Blaug argued in 1963 that the standard indictment did not survive contact with the evidence, that the scales were less widespread and less generous than assumed, and that the real driver of southern rural distress was a collapse in demand for winter labor rather than the allowance system.

Why does a dispute about English parishes in the 1790s matter to you? Because the identical argument runs today about the earned income tax credit and about employer-sponsored wage subsidies. If the government tops up low wages, does it help workers, or does it let employers pay less and pocket the subsidy? That question is live, it has real economists on both sides, and you will meet it again in this course with modern data. Speenhamland is where it starts.

The upshot: A wage subsidy can raise a worker's income and depress the wage the employer offers at the same time. Which effect dominates is an empirical question, not a moral one, and it has been argued about for well over two centuries.

1834: less eligibility and the workhouse test

A royal commission reported in 1832, its investigation shaped heavily by the economist Nassau Senior and the sanitary reformer Edwin Chadwick. The Poor Law Amendment Act of 1834 followed. It grouped parishes into unions run by elected boards of guardians, created a central Poor Law Commission, and rested on a principle stated with real clarity.

The principle was less eligibility. The condition of the pauper receiving relief must be made less desirable than the condition of the lowest-paid independent laborer. Not less desirable in intention or in rhetoric, but actually, materially less desirable, because otherwise, the commissioners reasoned, relief would draw people off the labor market.

The instrument was the workhouse test. Rather than assess each applicant, offer relief only inside the workhouse, and make the workhouse unpleasant enough that only the genuinely desperate would enter. Families were separated on entry. Diets were monotonous by design. Labor was tedious by design: picking oakum, breaking stones, grinding bones for fertilizer.

In 1845 the design produced its own indictment. At the Andover union workhouse in Hampshire, inmates crushing bones were found to be fighting over and gnawing the marrow and gristle. A parliamentary inquiry followed, the master was dismissed, and the Poor Law Commission itself was replaced in 1847 by a Poor Law Board answerable to Parliament. The scandal did not end deterrence. It ended one commission.

It is worth being accurate about implementation. The 1834 act intended to abolish outdoor relief, meaning relief outside the workhouse, for the able-bodied. It largely failed to. Boards of guardians found outdoor relief cheaper than maintaining people in institutions, and for most of the nineteenth century the large majority of paupers were relieved outdoors. The workhouse mattered less as a place than as a threat.

What matters here: Less eligibility is not a historical curiosity. Every argument that a benefit must stay below the lowest wage so that work still pays is a restatement of it, and you will hear it in a legislative hearing this year.

America builds the same machine, then investigates it

The colonies imported the parish model as town responsibility, with overseers of the poor, outdoor relief in cash and kind, and, for those without local ties, removal. By the early nineteenth century, states were commissioning studies of the mess. Josiah Quincy's 1821 report in Massachusetts and John Yates's 1824 report in New York both concluded that outdoor relief was wasteful and demoralizing and that the almshouse was the better instrument.

The result was a century of institution building: almshouses, poor farms, orphan asylums, and county homes, in which the aged, the mentally ill, the disabled, and children were frequently housed together. Dorothea Dix spent the 1840s documenting the conditions of people with mental illness in Massachusetts jails and almshouses, presenting a memorial to the legislature in 1843 describing them confined in cages, closets, cellars, stalls, and pens, and chained, naked, beaten with rods. Her campaign built state hospitals across the country, which is a reminder that the reform of an institution often produces another institution.

Two roots: the visitor and the resident

Two movements arrived in the last third of the nineteenth century, and the American social work profession is the child of both. They were not friendly.

The Charity Organisation Society was founded in London in 1869 to bring order to a chaotic field of private charities. Its American beginning was in Buffalo in 1877, where the Reverend Stephen Humphreys Gurteen founded a society after studying the London model. The method was registration and investigation: keep a central case register so families cannot collect from several charities at once, investigate each application, and send a friendly visitor, usually an unpaid middle-class woman, to advise and to model habits. Relief was to be a last resort and character the primary target. Josephine Shaw Lowell, who founded the New York COS in 1882, wrote that the problem was to help without pauperizing.

The settlement house movement began at Toynbee Hall in London in 1884 and arrived at Hull House in Chicago in 1889, founded by Jane Addams and Ellen Gates Starr in a decaying mansion on Halsted Street. Lillian Wald opened the Henry Street Settlement in New York in 1893 and coined the term public health nurse. The method was the opposite of visiting: you moved in. Residents lived in the neighborhood, ran kindergartens, clubs, and classes, and above all collected data and pushed for law.

The data mattered. Hull-House Maps and Papers, published in 1895, mapped wages and nationalities block by block in the surrounding wards. Florence Kelley, a Hull House resident, became Illinois's first factory inspector in 1893 and then general secretary of the National Consumers League, and she spent her career on child labor law and hours legislation. Addams received the Nobel Peace Prize in 1931.

	Charity organisation	Settlement house
Unit of concern	The individual family	The neighborhood and the industry
Diagnosis of poverty	Character, habits, and circumstance, investigated case by case	Wages, hours, housing, sanitation, and law
Where the worker stands	Visits from outside, keeps records, refers	Lives in the neighborhood, organizes, testifies
Signature product	Casework method and the case record	Survey research and protective legislation
Descendant in the profession	Clinical and direct practice	Community organizing and policy practice

The caricature is that one tradition blamed the poor and the other did not. The reality is messier and more useful. COS investigation produced the first systematic case records and, through Mary Richmond's Social Diagnosis in 1917, the first attempt at a transmissible method of assessment. Settlement residents were capable of their own condescension toward immigrant neighbors. What separates them cleanly is the unit of analysis, and that is the split you inherit when you choose between a clinical and a macro concentration in a social work program.

The 1915 hinge

At the National Conference of Charities and Correction in 1915, the medical educator Abraham Flexner was asked whether social work was a profession. He said it was not. His reasoning was that a profession requires an educationally communicable technique of its own, and that social work mediated between other professions rather than owning a body of skill.

The field took the criticism seriously and moved toward the answer that would satisfy it: casework, a demonstrable individual method, taught in university schools. Richmond's Social Diagnosis appeared two years later. The consequence for policy was substantial. The tradition with the better claim to professional technique was the individual one, and the settlement tradition's strengths, organizing and legislating, drifted toward the edge of the field's self-definition for decades.

Remember: The profession's tilt toward individual method was not only an intellectual choice. It was partly an answer to a specific 1915 challenge about what made social work a profession at all.

Mothers' pensions, and a rule that outlived them

One more inheritance, and it leads directly into the next lesson. In 1911 Illinois passed the Funds to Parents Act, allowing juvenile courts in Cook County to pay a mother to keep her children at home rather than see them committed to an institution. It followed the 1909 White House Conference on the Care of Dependent Children, which had declared that children should not be removed from home for reasons of poverty alone. Within a decade most states had a mothers' pension law of some kind, and the federal Children's Bureau, created in 1912 under Julia Lathrop, promoted them.

The pensions were small, locally administered, and conditional on a suitable home. That last phrase did the work. Administrators used it to exclude unmarried mothers, divorced mothers, and, overwhelmingly, Black mothers. When Congress wrote Aid to Dependent Children into the Social Security Act in 1935, it wrote it in the image of the mothers' pension, including state discretion over suitability. The suitable home rule and its descendants, including the man in the house rule that the Supreme Court struck down in King v. Smith in 1968, shaped who actually received aid for the next three decades.

Common misconceptions

• The poor laws were simply cruel and unthinking. They were often cruel and they were carefully reasoned. The 1834 commissioners published their logic, and versions of that logic are used today by people who would be appalled to know its source.
• The 1834 act abolished outdoor relief. It intended to for the able-bodied and largely failed. Guardians kept using outdoor relief because it cost less.
• Speenhamland proved that wage subsidies pauperize workers. That was the received view for a century and it was substantially challenged by later research on the parish records. Treat it as a contested case, not a settled lesson.
• Settlement houses were charities that ran programs. Their distinctive product was research and legislation. Hull-House Maps and Papers preceded the Chicago School of sociology, and Florence Kelley's factory inspection work changed Illinois law.
• Mothers' pensions were an early universal benefit for mothers. They were discretionary, small, and administered with suitability rules that excluded most of the mothers who needed them most.

Looking back

• The 1601 act made relief local, financed by a property tax, and conditional on being sorted into the impotent, the able-bodied, or the refusing.
• Settlement law tied relief to place of belonging; American residency requirements survived until Shapiro v. Thompson in 1969.
• Less eligibility, stated plainly in 1834, is the direct ancestor of every argument that benefits must stay below the lowest wage.
• The charity organisation tradition gave the profession casework and the case record; the settlement tradition gave it survey research and legislative advocacy.
• Mothers' pensions handed the Social Security Act both the idea of aid to children at home and the suitable home discretion that excluded Black and unmarried mothers.

Sources

1. Encyclopaedia Britannica. (n.d.). Poor Law. britannica.com
2. Encyclopaedia Britannica. (n.d.). Jane Addams. britannica.com
3. Library of Congress. (n.d.). Jane Addams Papers and the Hull-House settlement. loc.gov
4. Wikipedia contributors. (n.d.). Poor Law Amendment Act 1834. en.wikipedia.org
5. Wikipedia contributors. (n.d.). Speenhamland system. en.wikipedia.org
6. Social Security Administration. (n.d.). Historical background and development of Social Security. ssa.gov

A bottle of whiskey in Georgetown

In her memoir of the Roosevelt years, Frances Perkins describes calling the Committee on Economic Security to her house in Georgetown one evening in December 1934, producing a bottle of whiskey, and announcing that nobody was leaving until they had a plan. The committee had been created by executive order in June 1934 and had a report due to the President. Its members had spent months disagreeing about whether the country should build contributory insurance or expanded relief, whether states or the federal government should run any of it, and whether health insurance should be in the bill at all.

Perkins, the Secretary of Labor and the first woman to sit in an American Cabinet, had a social work background: she had lived at Hull House, she had watched the Triangle Shirtwaist fire from the street in 1911, and she had run New York's industrial commission. The plan that came out of her house, drafted into legislation by Edwin Witte and his staff and signed by Franklin Roosevelt on 14 August 1935, is the foundation of the American welfare state. Understanding what it included, what it deliberately left out, and who it did not cover explains most of the arguments in the rest of this course.

Key idea: The Social Security Act was not one program. It was a package of at least eight, built on two entirely different logics, and the difference between those logics organizes American social policy to this day.

The pressure that made it possible

Roosevelt did not propose old-age security into a vacuum. In September 1933, Francis Townsend, a physician in Long Beach, California, wrote to a local newspaper proposing that the federal government pay every citizen over sixty a fixed sum each month, on the condition that it be spent within thirty days. The plan was fiscally impossible and enormously popular. Townsend Clubs spread across the country and their petitions reached Congress by the sackful.

Huey Long, the senator from Louisiana, launched his Share Our Wealth program in 1934 with a proposal to cap fortunes and guarantee households a minimum income. Upton Sinclair nearly won the California governorship in 1934 on the End Poverty in California platform. None of these programs became law. All of them made a moderate contributory pension look like the cautious option, which is exactly how Roosevelt presented it.

Meanwhile the existing system had simply failed. Unemployment reached roughly a quarter of the workforce in 1933. Relief in America was a matter for counties, townships, and private charities, and those bodies ran out of money. The Federal Emergency Relief Administration, created in May 1933 under Harry Hopkins, sent federal money to state relief agencies for the first time in peacetime. It was explicitly temporary. The 1935 act was the permanent structure meant to replace improvisation.

What the act actually contained

The statute is organized by title, and the titles are worth knowing by number because policy documents still cite them that way. Title XIX and Title XXI, which you will meet later as Medicaid and the Children's Health Insurance Program, were added to this same act decades afterward.

Title	Program	Track	What became of it
I	Old-Age Assistance: federal grants to states for means-tested aid to the elderly poor	Assistance	Folded into SSI in 1972
II	Federal Old-Age Benefits: contributory pensions for covered workers	Insurance	Grew into OASDI, today's Social Security
III and IX	Unemployment compensation: administrative grants plus a payroll tax with a credit for state programs	Insurance	Still a federal-state hybrid, still uneven by state
IV	Aid to Dependent Children: grants to states for children deprived of parental support	Assistance	Became AFDC, then TANF in 1996
V	Maternal and child health, services for children with disabilities, child welfare services	Services	Still funds state maternal and child health work
VI	Public health work grants to states	Services	Foundation of federal public health funding
X	Aid to the Blind	Assistance	Folded into SSI in 1972

Title III and Title IX deserve a second look because their mechanism is clever and is still used. The federal government imposed a payroll tax on employers, then allowed employers to credit most of it against a state unemployment tax if their state had a qualifying program. No state was required to create unemployment insurance. Every state did within two years, because the money was going to be collected either way. That device, taxing and then offering a credit for state compliance, is one of the standard tools of American federalism, and you will see its cousins in Medicaid matching and in highway funding.

The two tracks, and why the difference never stops mattering

The act built social insurance and public assistance side by side, and they behave differently in every respect that matters.

	Social insurance	Public assistance
Basis of claim	Contributions from covered work	Demonstrated need
Test applied	Work history and age or disability	Income and assets
Who sets the rules	Federal, uniform nationwide	Largely states, within federal limits
How recipients are described	Beneficiaries who earned it	Recipients who are given it
Political durability	Very high; cuts are electorally dangerous	Low; benefits have been cut, capped, and block granted
Examples	Social Security retirement, SSDI, Medicare, unemployment insurance	TANF, SNAP, SSI, Medicaid, housing assistance

The contribution story is doing enormous political work in the left column, and it is worth being precise about it. Social Security has never been a savings account. Payroll taxes from current workers pay current beneficiaries. Ida May Fuller of Ludlow, Vermont, who received the first monthly benefit check in January 1940 for 22 dollars and 54 cents, had paid in about 25 dollars in total and lived to be one hundred. What the contribution buys is not an account balance. It is a claim that feels earned, and that feeling has protected the program through nine decades of budget politics while assistance programs with far smaller price tags have been repeatedly restructured.

What matters here: The perception that a benefit was earned, rather than any actuarial fact about accounts, is the strongest predictor of whether an American social program survives.

What was left out on purpose

Two omissions shaped the following century.

The first was health insurance. The Committee on Economic Security studied it and drafted material on it. It was left out of the bill. The reasoning, well documented in the committee's own record, was that the American Medical Association's opposition to compulsory health insurance was strong enough to endanger the entire package. Roosevelt chose the rest of the bill over that fight. National health insurance was proposed again in 1945 by Truman, defeated, and the next successful federal expansion did not arrive until 1965.

The second was occupational coverage. Title II old-age insurance excluded agricultural laborers and domestic servants, along with the self-employed, government employees, and workers for nonprofit organizations. Unemployment insurance excluded the same farm and household workers.

Those two occupational categories were where a very large share of Black workers were employed in 1935, particularly in the South. The consequence is not disputed: the new contributory system, the good track, largely bypassed Black workers, while the means-tested track, administered by states with discretion, was where many were left.

Why the exclusion happened, argued both ways

The motive is genuinely disputed among historians, and you should be able to state both cases.

Ira Katznelson's argument, developed most fully in When Affirmative Action Was White, points to the structure of Congress. Southern Democrats held committee chairmanships through seniority and were indispensable to Roosevelt's coalition. A federal pension paid directly to Southern agricultural and domestic workers would have undercut a low-wage labor system that depended on those workers having no alternative. Katznelson notes that the exclusions were added in committee, that Southern members had the power to insist, and that the pattern repeats across New Deal and postwar legislation, including the administration of the GI Bill.

The competing account, argued at length by Larry DeWitt in the Social Security Administration's own historical journal, is administrative. Collecting a payroll tax in 1935 required an employer with books, a payroll, and a stable place of business. A farmer with three hands and a household employing one cook had none of those things, and the Treasury had said so. DeWitt points out that the exclusion was proposed by Treasury officials on feasibility grounds, that it was not a Southern amendment added on the floor, and that in absolute numbers most excluded workers were white.

Here is what a careful reader takes from the two accounts. They are not symmetrical claims about the same fact, and they can both be partly right: administrative feasibility can supply a reason while the political usefulness of the result supplies the willingness not to look harder for a solution. What is not in dispute is the effect. Whatever the motive, a system built to reward covered employment gave its strongest protections to the workers already best positioned, and the repair took decades.

The upshot: Ask of any coverage rule not only what it says, but which existing labor market it maps onto, because a neutral-sounding rule laid over an unequal economy produces an unequal program.

The amendments that built the modern system

The 1935 act is less than half the story. Four sets of amendments made it what it is.

1939. Added benefits for the spouses and children of retired workers and for survivors of deceased workers, converting old-age insurance into old-age and survivors insurance. This changed the program from an individual pension into family protection, and it built in the assumption of a married couple with a single earner, which produces the spousal benefit rules that still puzzle two-earner couples today. It also moved the first monthly payments forward to 1940.

1950 through 1956. Coverage was extended in stages to most regularly employed farm and domestic workers, to the self-employed, and to many state and local employees, closing the largest 1935 gaps. Benefits were raised substantially in 1950 after a decade of inflation had eroded them. In 1956 Congress added disability insurance for workers aged fifty and older, extended to all ages in 1960 after a long fight in which the American Medical Association warned about federal control of medicine.

1965. Medicare and Medicaid, which get their own lesson.

1972. Two changes of the first order. Automatic cost of living adjustments were built in, with the first automatic increase paid in 1975, which removed benefit increases from the annual political calendar. And Titles I, X, and the aid to the permanently and totally disabled program were pulled out of state hands and federalized into Supplemental Security Income, effective in 1974, with a uniform national payment.

Notice what the 1972 federalization did and did not do. Aid to the elderly, blind, and disabled poor became a national program with one rule book. Aid to poor families with children stayed with the states. That divergence, taken in 1972, set up everything that happened to AFDC in the following twenty-four years.

Common misconceptions

• Social Security is a savings account with your name on it. Current taxes pay current benefits. The trust funds hold accumulated surpluses in Treasury securities, but no individual account exists.
• The 1935 act created welfare as we knew it. Title IV was small, expected to shrink as survivors insurance covered widows, and given to the states to run. It grew because the assumption behind it, that most poor single mothers were widows, was wrong.
• Health insurance was never seriously considered in 1935. It was studied and drafted, then removed because the committee judged that organized medical opposition would sink the whole bill.
• The occupational exclusions were an accident of drafting. Whatever the motive, they were deliberate, they were known to exclude most agricultural and domestic workers, and their consequences were foreseeable.
• Unemployment insurance is a federal program. It is a federal-state hybrid in which states set benefit levels, duration, and eligibility, which is why the same job loss produces very different outcomes in different states.

Putting it together

• The 1935 act bundled contributory insurance, means-tested assistance, and health and welfare services into one statute organized by title.
• The insurance track is federal, uniform, and politically protected; the assistance track is state-shaped, discretionary, and repeatedly restructured.
• Title IX's tax-and-credit device induced every state to build unemployment insurance without requiring any state to do so.
• Health insurance was dropped to save the bill; agricultural and domestic workers were excluded from coverage, and the reasons for that are argued as politics by Katznelson and as administration by DeWitt.
• The 1939, 1950s, and 1972 amendments added survivors, disability, automatic indexation, and the federalization of adult assistance into SSI, while aid to families with children stayed with the states.

Sources

1. Social Security Administration. (n.d.). Historical Background and Development of Social Security. ssa.gov
2. Social Security Administration. (n.d.). The Social Security Act of 1935. ssa.gov
3. DeWitt, L. (2010). The decision to exclude agricultural and domestic workers from the 1935 Social Security Act. Social Security Bulletin, 70(4). ssa.gov
4. Encyclopaedia Britannica. (n.d.). Frances Perkins. britannica.com
5. Encyclopaedia Britannica. (n.d.). New Deal. britannica.com
6. Wikipedia contributors. (n.d.). Townsend Plan. en.wikipedia.org

Card number one

On 30 July 1965, Lyndon Johnson flew to Independence, Missouri, to sign the Social Security Amendments at the Truman Library, and handed Harry Truman the first Medicare card. Truman had proposed national health insurance in 1945 and watched it die. Bess Truman received card number two. The staging was deliberate: Johnson was claiming a twenty-year-old defeat as a victory and, more practically, wrapping a contentious new entitlement in the memory of a president the country had grown fond of.

What he actually signed was stranger and more consequential than a health insurance bill. It was three different programs, built on three incompatible theories, stapled together by a committee chairman who had spent the spring letting each side believe he was on theirs.

Key idea: Medicare and Medicaid are not two halves of a plan. They are the surviving pieces of three competing plans that a legislative tactician combined rather than chose among.

The three-layer cake

By early 1965 there were three serious proposals on the table in the House Ways and Means Committee.

The administration's bill, called the King-Anderson bill, offered hospital insurance for the elderly financed by a payroll tax through Social Security. It covered hospitals, not doctors, on the theory that hospital bills were the ones that bankrupted families and that physicians would fight anything touching their fees.

The Republican alternative, from John Byrnes of Wisconsin and backed in outline by the American Medical Association, offered voluntary insurance for physician services, financed by beneficiary premiums and general revenue rather than payroll taxes.

The third option was expansion of the Kerr-Mills program of 1960, which gave states federal matching funds for medical assistance to the elderly poor, a means-tested approach that had been adopted unevenly and worked poorly in the states that most needed it.

Wilbur Mills, the chairman, was known for never bringing a bill to the floor he could not pass. In March 1965 he did something almost nobody expected. He took all three. The administration's hospital plan became Medicare Part A. The Republican physician plan became Medicare Part B. The means-tested expansion became Medicaid, Title XIX. Legend on Capitol Hill called it the three-layer cake, and it explains a structure that otherwise makes no sense: two Medicare parts with different financing, different enrollment rules, and different politics, plus a separate program for the poor operating under entirely different logic.

Medicare and Medicaid, side by side

	Medicare (Title XVIII)	Medicaid (Title XIX)
Who qualifies	People aged 65 and over with sufficient work credits, plus people receiving SSDI after a waiting period and people with end-stage renal disease	People meeting income and, historically, categorical criteria set within federal rules by each state
Track	Social insurance	Public assistance
Who runs it	Federal government, uniform nationwide	States, under federal minimums, with wide variation
Financing	Part A from a dedicated payroll tax; Part B and Part D from premiums and general revenue	Federal-state match, the federal share set by a formula based on state per capita income with a statutory floor of fifty percent
Long-term care	Limited skilled nursing after a hospital stay; no coverage of ongoing custodial care	The largest payer of nursing home care in the country

Two rows deserve emphasis because they generate constant confusion in practice. First, the financing row explains why Medicare feels earned and Medicaid does not: the payroll tax on the left, the welfare match on the right. Second, the long-term care row is the single most common surprise in family meetings. A retired machinist with Medicare who needs three years in a nursing home will find that Medicare pays for essentially none of it, and that the route to help runs through Medicaid, which requires spending down assets. You will meet this again in the long-term care lesson.

Why this matters: Medicare covers medical care for older people; Medicaid covers medical care and long-term care for poor people, including older people who have become poor by paying for care. Most families discover the difference at the worst possible moment.

Medicare desegregated American hospitals

Here is the part of the 1965 story that rarely appears in a health policy syllabus. Title VI of the Civil Rights Act of 1964 barred discrimination in any program receiving federal financial assistance. When Medicare began paying hospitals in 1966, hospitals that wanted the money had to certify compliance with Title VI, and the federal government sent inspectors.

Thousands of hospitals, concentrated in the South, desegregated wards, waiting rooms, and staffs in a matter of months. It was one of the fastest institutional desegregations in American history, and it happened because a payment stream was made conditional. The historian David Barton Smith documented the campaign in detail, including the improvised federal effort to inspect facilities in time for the program's launch.

The lesson generalizes past health care. Conditions attached to money move institutions faster than prohibitions attached to conduct, because the institution has to act to keep something it already wants.

The other 1965, and the other 1964

Medicare and Medicaid are the durable pieces. The War on Poverty was the more radical part, and it came first. Johnson declared unconditional war on poverty in his State of the Union address in January 1964. The Economic Opportunity Act followed in August, creating the Office of Economic Opportunity under Sargent Shriver, along with Job Corps, VISTA, and Community Action.

The Community Action Program contained six words that produced years of conflict: maximum feasible participation of the residents. Community action agencies were to be governed with real involvement by poor people in the areas served. In practice this meant federally funded organizations in cities across the country, sometimes staffed by organizers, challenging the mayors and agencies that had always controlled antipoverty money. Daniel Patrick Moynihan, who had been in the administration, published Maximum Feasible Misunderstanding in 1969 arguing that the phrase had been inserted with little thought about what it would mean once real money followed it. Mayors complained to the White House. The Green Amendment of 1967 gave local governments authority over community action agencies, and OEO itself was dismantled in the early 1970s with its programs parceled out to other departments.

What survived the dismantling is substantial: Head Start, begun as a summer program in 1965; the Job Corps; legal services for the poor; community health centers; and the food stamp program, made permanent by the Food Stamp Act of 1964. The Elementary and Secondary Education Act of 1965 created Title I funding for schools serving poor children. The Older Americans Act of 1965 created the aging services network you will meet in Module 5.

Did it fail? The argument, with the numbers

Ronald Reagan's 1988 formulation, that the federal government fought a war on poverty and poverty won, became the standard conservative verdict, and it rests on a real observation. The official poverty rate was 19.0 percent in 1964. It fell sharply through the 1960s, reaching 12.1 percent by 1969, and then moved within a band roughly between eleven and fifteen percent for the following half century. If you plot that series, the War on Poverty looks like a plateau with a lot of spending piled on it.

The counterargument is not that the series is wrong. It is that the series cannot see most of what was built. The official poverty measure counts pre-tax cash income only. It does not count SNAP, which is in-kind. It does not count housing assistance, which is in-kind. It does not count Medicaid at all. It does not count the earned income tax credit or the child tax credit, because those arrive through the tax system rather than as cash income. Almost every antipoverty program created after 1964 was designed in a form the official measure is structurally unable to register.

When researchers rebuild the historical series using the supplemental poverty measure's definition, which counts taxes and in-kind benefits, the picture changes. Work by Christopher Wimer, Liana Fox, Irwin Garfinkel, Neeraj Kaushal, and Jane Waldfogel at Columbia produced an anchored supplemental series back to 1967, and it shows poverty falling from roughly nineteen percent in 1967 to roughly sixteen percent in 2012, with the decline concentrated among the elderly and driven substantially by benefits the official measure ignores.

You should also hold the strongest version of the critique, which is not about measurement. It is that Community Action promised political power and delivered a decade of grants, that the programs which survived were the least threatening ones, and that a federal effort which never seriously touched wages, unions, or housing supply was never going to change the distribution of income. That argument does not depend on the poverty rate at all.

The upshot: Whether the War on Poverty worked is partly a factual question about which measure you use and partly a political question about what you thought it was for, and the two questions are almost always run together.

What happened to Medicare after 1965

Three later additions matter for practice. Part C, created as Medicare+Choice in the Balanced Budget Act of 1997 and renamed Medicare Advantage in 2003, lets beneficiaries take their Medicare benefit through a private plan; enrollment in these plans has grown to roughly half of all beneficiaries. Part D, created by the Medicare Modernization Act of 2003 and effective in 2006, added outpatient prescription drug coverage delivered entirely through private plans, and included a noninterference clause barring the federal government from negotiating drug prices. The Inflation Reduction Act of 2022 changed that in part, authorizing negotiation for a selected and growing list of high-spending drugs, capping insulin cost sharing at 35 dollars a month, and capping out-of-pocket Part D spending.

Each of these is a good test of the anatomy from Lesson 1. Part D is a federal entitlement delivered by private plans with premiums, an unusual hybrid; the noninterference clause was a financing choice with a twenty-year price tag; and the 2022 change is an example of a statute rewriting one sentence of an earlier statute and moving billions of dollars.

Common misconceptions

• Medicare covers nursing home care. It covers limited skilled nursing following a qualifying hospital stay. Ongoing custodial care is not a Medicare benefit, and Medicaid is the main public payer.
• Medicaid is a federal program. It is fifty-six state and territorial programs under federal minimums, with different eligibility levels, covered services, and provider payment rates.
• Medicare was designed as one coherent program. Parts A and B come from two rival bills that Wilbur Mills combined rather than reconciled, which is why their financing and enrollment rules differ.
• The War on Poverty had no measurable effect. It had a large measured effect on elderly poverty and a substantial one overall once in-kind and tax benefits are counted, though the official measure cannot show it.
• Community Action failed because poor people could not run programs. It was curtailed after mayors and members of Congress objected to federally funded organizations challenging local government, which is a political explanation rather than a competence one.

The short version

• Wilbur Mills combined a hospital insurance bill, a voluntary physician insurance bill, and a means-tested expansion into Medicare Part A, Medicare Part B, and Medicaid in one 1965 statute.
• Medicare is federal social insurance for older and disabled people; Medicaid is a federal-state assistance program and the country's largest payer for long-term care.
• Conditioning Medicare payments on Title VI compliance desegregated thousands of hospitals within months in 1966.
• The War on Poverty produced Head Start, community health centers, legal services, Title I, and the Older Americans Act; Community Action's maximum feasible participation clause produced a fight that ended in local government control.
• The official poverty series cannot register in-kind and tax-based benefits, which is why the failure verdict and the success verdict often rest on different measures rather than different facts.

Sources

1. Centers for Medicare and Medicaid Services. (n.d.). History of CMS. cms.gov
2. Social Security Administration. (n.d.). Social Security Amendments of 1965. ssa.gov
3. U.S. Census Bureau. (n.d.). Historical Poverty Tables: People and Families. census.gov
4. Encyclopaedia Britannica. (n.d.). Great Society. britannica.com
5. Medicaid.gov. (n.d.). Medicaid program history and financing. medicaid.gov

Two questions, two programs

To decide whether you can receive Supplemental Security Income, the Social Security Administration needs to know the face value of any life insurance policy you own, whether you have money set aside for a burial, what was in your bank account on the first of the month, and whether anyone gives you groceries or lets you live somewhere without paying full rent. To decide whether you can receive Medicare, it needs to know your date of birth and your work record.

Both programs are administered by the same agency, in the same buildings, often by the same staff. The difference is not bureaucratic personality. It is a design choice about what confers a claim, and that single choice determines how much a program costs, how many eligible people actually get it, how it is talked about at a dinner table, and how likely it is to survive the next budget.

The point: Whether a benefit is allocated by category or by demonstrated need is the most consequential design decision in social welfare policy, and almost every other feature of a program follows from it.

Four allocation designs

The vocabulary is not standardized across textbooks, so learn the underlying distinctions rather than the labels.

Design	Basis of claim	Examples	Characteristic weakness
Universal	Membership in the political community, or in a broad status	Public schools, public libraries, the fire department, Canada's Old Age Security in its basic form	Expensive per unit of poverty reduced; sends money to people who do not need it
Categorical	Membership in a defined group	Veterans benefits, Medicare at 65, universal school meals in participating districts	Boundary disputes about who is in the category
Social insurance	Prior contributions from covered work, plus a triggering event	Social Security retirement, SSDI, unemployment insurance	Reproduces labor market inequality; misses people with weak work histories
Means-tested	Income, assets, or both, verified	SNAP, SSI, Medicaid, housing vouchers, TANF	Administrative cost, incomplete take-up, stigma, benefit cliffs

Most American programs are hybrids. SSI is categorical and means-tested at once: you must be aged, blind, or disabled, and you must be poor. SNAP is means-tested with categorical routes in, meaning that receipt of certain other benefits can establish eligibility without a separate asset test. The hybrid character is why so many people fall between programs.

The case for means testing, made properly

The argument is about target efficiency, and it is not stupid. Suppose you have ten billion dollars to reduce poverty. Spend it universally across a population in which one household in eight is poor and roughly seven eighths of it goes to households that were not poor. Spend it through a means-tested program and nearly all of it reaches poor households. If your objective function is dollars delivered to poor people per dollar of budget, targeting wins by a very large margin, and it is not close.

Targeting also permits deeper benefits. A means-tested program can afford to raise a small number of households a long way. A universal program of the same total cost raises everyone slightly, which for a household well above the poverty line is a rounding error and for a household far below it is not enough.

There is a fairness argument too, and you should be able to state it without a sneer. A worker earning thirty-two thousand dollars who pays taxes into a program that sends money to households with more assets than she has is entitled to ask why. Vertical equity is a real value, not a cover story.

The case against, which is mostly empirical

The counterarguments are not about compassion. They are about what actually happens when you build the test.

Take-up is incomplete. Every means test loses eligible people. The Department of Agriculture estimates that a substantial share of people eligible for SNAP do not receive it, with the rate lowest among eligible older adults and working households. The IRS estimates that roughly one in five workers eligible for the earned income tax credit does not claim it, which for a benefit delivered through a tax return people already file is a striking figure. Participation among elderly people eligible for SSI has long been estimated well below full.

The test costs money. Verification, redetermination, appeals, fraud control, and the staff to run them are pure administrative overhead. Universal programs spend almost nothing determining eligibility because there is nothing to determine.

Means tests create marginal tax rates. If a benefit phases out at thirty cents per dollar earned, and a second benefit phases out at twenty cents, and payroll tax takes another seven and a half, a worker can face an effective marginal rate on additional earnings that a person with a much higher income would consider confiscatory. Congressional Budget Office analyses have found median effective marginal tax rates for low- and moderate-income workers in the range you would associate with the top of the income distribution, with much higher rates at specific thresholds. Where a benefit ends abruptly rather than phasing out, the result is a benefit cliff: a raise of a hundred dollars a month costs a family child care assistance worth far more, and the rational response is to refuse the raise.

The test itself is a burden. This is the deepest point, and it has a literature.

Administrative burden, in three costs

Pamela Herd and Donald Moynihan's 2018 book Administrative Burden gave the field a framework worth memorizing, because once you have it you cannot stop seeing it.

Learning costs. The effort required to find out that a program exists, whether you might qualify, and what it would give you. A benefit nobody has heard of has a take-up rate of zero regardless of its generosity.

Compliance costs. The documents, the forms, the appointments, the hours on hold, the interview during working hours, the pay stubs from an employer who does not answer the phone, the recertification every six months. These costs fall hardest on exactly the people the program is for, because a person working two jobs with no car and no printer has less capacity to absorb them than a person with a flexible salary and a scanner.

Psychological costs. Stigma, the sense of being suspected, the loss of autonomy in being asked to account for a bank balance, and the stress of an outcome you do not control. These are not soft costs. They change behavior, and they are frequently the reason someone eligible does not apply twice.

Herd and Moynihan's sharpest observation is political: burdens are often not accidents. They can be chosen. A legislature that wishes to reduce enrollment without repealing a program can add a verification requirement, shorten a recertification period, or require an in-person interview, and the caseload will fall without a single vote on eligibility.

The Medicaid unwinding that began in April 2023, when the pandemic-era continuous enrollment requirement ended, is the largest natural demonstration in recent memory. Millions of people lost coverage. Tracking by the Kaiser Family Foundation found that a large majority of disenrollments were procedural, meaning the state could not complete a renewal rather than the state determining the person was ineligible. Renewal packets went to old addresses. Call centers had hours-long waits. That is compliance cost operating at national scale.

Remember: Take-up is not a measure of need. It is a measure of the distance between eligibility on paper and enrollment in fact, and that distance is built out of learning, compliance, and psychological costs that policy chooses.

The paradox of redistribution, and its critics

Now the harder argument, because it flips the target efficiency case on its head.

In 1998, Walter Korpi and Joakim Palme published a comparative study of welfare states with a finding they called the paradox of redistribution: the more a country targets benefits at the poor alone, and the more it equalizes benefits across recipients, the less redistribution and poverty reduction it actually achieves. Their explanation was political. Targeted programs create small constituencies of poor recipients and large constituencies of taxpayers who receive nothing, so budgets stay small and are easy to cut. Broad programs create coalitions that include the middle class, which sustains far larger budgets, and a large budget imperfectly targeted can redistribute more in absolute terms than a small budget perfectly targeted.

Wilbur Cohen, who helped draft the 1935 act and later served as Secretary of Health, Education, and Welfare, put the same thought in a sentence that has been repeated ever since: a program for poor people will be a poor program.

The paradox is not settled. Ive Marx, Lina Salanauskaite, and Gerlinde Verbist reexamined the relationship using more recent data and found that the negative association between targeting and redistribution had weakened substantially and in some specifications disappeared. David Brady and Amie Bostic, working with the same data tradition, found that the size of transfers mattered far more than their targeting, and that some strongly targeted systems achieved considerable redistribution. The current state of the argument is roughly this: targeting is not automatically self-defeating, but the political mechanism Korpi and Palme identified is real, and the way a country targets, whether through a stigmatizing separate program or through a phase-out inside a broad one, matters more than whether it targets at all.

Theda Skocpol proposed a design that tries to have both, which she called targeting within universalism: build a program everyone can join, then layer additional help inside it for those who need more. Medicare with income-related premium adjustments is one example. A universal child benefit that is taxed back from high earners is another.

A natural experiment worth studying

For six months in 2021, the United States ran something close to a universal child benefit. The American Rescue Plan Act made the child tax credit fully refundable, raised it, removed the earnings requirement that had excluded the poorest families, and delivered it monthly to most households automatically.

The Census Bureau's supplemental poverty measure recorded child poverty at 5.2 percent in 2021, the lowest figure in the series. The expansion expired at the end of 2021. Child poverty on the same measure was 12.4 percent in 2022. Whatever else that sequence shows, it demonstrates that a near-universal, automatically delivered, unconditional benefit can move a poverty rate very fast, and that the effect goes away just as fast when the benefit does. You will work through the design details in Lesson 7.

Common misconceptions

• Means testing saves money. It reduces benefit outlays and raises administrative outlays, and part of the savings comes from eligible people who never enroll, which is a cost borne by them rather than a genuine efficiency.
• Low take-up means people do not want the benefit. Take-up tracks the burden of application far more closely than it tracks need, and burdens can be raised or lowered by policy.
• Universal programs are always more expensive. They are more expensive in gross outlays and cheaper in administration, and they can deliver more absolute redistribution if the resulting political coalition sustains a larger budget.
• Benefit cliffs are an unfortunate accident. They are a direct consequence of drawing an eligibility line, and they can be smoothed by phase-outs at real budget cost. The choice not to smooth them is a choice.
• The paradox of redistribution has been proven. It was an important 1998 finding that later work has substantially qualified. Treat it as a live argument with good evidence on both sides.

Where this leaves us

• Universal, categorical, social insurance, and means-tested designs differ in basis of claim, and every other feature of a program follows from that.
• Targeting delivers more dollars per poor household per budget dollar, which is a real and serious advantage.
• Targeting also costs take-up, administration, marginal tax rates, and dignity, and administrative burden decomposes into learning, compliance, and psychological costs.
• Burdens are frequently chosen; the 2023 Medicaid unwinding showed procedural disenrollment at national scale.
• Korpi and Palme's paradox of redistribution has been qualified by later comparative work, but the political mechanism it describes is real, and targeting within universalism is the standard attempt at a synthesis.

Sources

1. Herd, P., and Moynihan, D. P. (2018). Administrative Burden: Policymaking by Other Means. Russell Sage Foundation. russellsage.org
2. Korpi, W., and Palme, J. (1998). The paradox of redistribution and strategies of equality. American Sociological Review, 63(5), 661-687. doi.org/10.2307/2657333
3. U.S. Census Bureau. (n.d.). Supplemental Poverty Measure. census.gov
4. U.S. Department of Agriculture, Food and Nutrition Service. (n.d.). SNAP program participation rates. fns.usda.gov
5. Internal Revenue Service. (n.d.). Earned Income Tax Credit and Other Refundable Credits. eitc.irs.gov

Three resignations

In the weeks after Bill Clinton signed the Personal Responsibility and Work Opportunity Reconciliation Act on 22 August 1996, three senior officials at the Department of Health and Human Services resigned. Mary Jo Bane and Peter Edelman were assistant secretaries; Wendell Primus was a deputy assistant secretary and one of the government's most careful analysts of poverty data. Edelman published an essay in The Atlantic the following March under the title The Worst Thing Bill Clinton Has Done.

Their objection was not to work requirements, which all three had supported in some form. It was to the conversion of a legal entitlement into a fixed block grant, which meant that in the next recession the money would not be there no matter how many families needed it. Twelve years later the recession arrived and they turned out to be right about that specific mechanism. Whether they were right about the reform as a whole is the harder question, and it is the question this lesson is built around.

Why this matters: The 1996 law produced two findings that are both well supported and that point in opposite directions. A person who can hold only one of them has not understood the reform.

What AFDC actually was by 1996

Aid to Families with Dependent Children began as Title IV of the 1935 act, built in the image of state mothers' pensions, and expected to shrink. Instead it grew, because its founding assumption was wrong: most poor single mothers were not widows covered by survivors insurance but separated, divorced, or never married, and no insurance program reached them.

Three legal developments in the late 1960s changed the program's character. In King v. Smith in 1968, the Supreme Court struck down Alabama's substitute father rule, under which a woman with a male companion was treated as having a breadwinner. In Shapiro v. Thompson in 1969 it struck down residency requirements. In Goldberg v. Kelly in 1970 it held that benefits could not be terminated without a prior evidentiary hearing, making receipt a property interest protected by due process. Alongside those decisions the National Welfare Rights Organization, founded in 1966 and led by George Wiley with Johnnie Tillmon among its central organizers, ran campaigns to inform eligible women of benefits they had never been told about. Caseloads rose steeply.

Reform attempts followed almost immediately. Richard Nixon proposed the Family Assistance Plan in 1969, a guaranteed income for families with children that passed the House twice and died in the Senate Finance Committee, opposed from the right as a giveaway and from the left as too small. The federal government also ran a set of negative income tax experiments between 1968 and 1982 in New Jersey, Seattle, Denver, and elsewhere, which found modest reductions in work effort, larger for wives than husbands, and a finding about marital dissolution that was influential at the time and substantially disputed afterward.

By the early 1990s the operating reality was this. AFDC was an individual entitlement: a family that met the state's eligibility rules had a legal right to a benefit, and the federal government matched state spending without limit. Benefit levels varied enormously by state and had fallen sharply in real terms since 1970. And more than forty states were running experiments under section 1115 waivers, testing time limits, family caps, and work requirements, so much of what the 1996 law nationalized was already happening.

What PRWORA changed

Feature	AFDC	TANF
Legal character	Individual entitlement to a benefit if eligible	No individual entitlement; states decide whom to serve
Federal financing	Open-ended match, rising with caseload	Fixed block grant of 16.5 billion dollars a year, not indexed
Time limits	None	Sixty months of federally funded aid in a lifetime, with a hardship exemption for up to twenty percent of the caseload; states may set shorter limits
Work rules	JOBS program under the 1988 Family Support Act, weakly enforced	States must meet work participation rates or lose funds
State obligation	Match federal dollars	Maintenance of effort at a share of historic state spending
Use of funds	Cash assistance	Any activity meeting four broad statutory purposes

Two features of that table interact in a way that is easy to miss and that explains most of what followed.

The block grant is fixed in nominal dollars and was never indexed. Since 1997 it has lost a large share of its real value to inflation, and it does not grow when a recession arrives.

The work participation requirement came with a caseload reduction credit. A state's required participation rate is lowered by the amount its caseload has fallen since a base year. Read that carefully. A state that moved recipients into jobs got credit. A state whose recipients simply stopped receiving benefits, for any reason at all, including sanctions, procedural closures, or discouragement, got exactly the same credit. The measure of success written into the statute was the size of the rolls.

The core of it: PRWORA rewarded caseload reduction, not poverty reduction, and the two came apart almost immediately.

The first true statement: employment rose

Employment among single mothers rose sharply in the second half of the 1990s. Census and Bureau of Labor Statistics series show the employment rate of never-married mothers climbing by well over ten percentage points between the early 1990s and 2000, one of the largest movements in any demographic group's labor force participation in the postwar record. Earnings in the bottom quintile of families with children rose. Caseloads fell from roughly 4.4 million families in 1996 to about half that by 2000.

Defenders of the reform take this as vindication, and the movement is real. But causal attribution is contested, and honestly so. Three things happened at once.

First, the reform itself, with its time limits, work requirements, and sanctions. Second, the strongest low-wage labor market in three decades: unemployment fell below four percent by 2000 and wages at the bottom rose in real terms for the first time since the 1970s. Third, the earned income tax credit had been expanded dramatically in 1993, roughly doubling the credit for families with two or more children, which sharply increased the return to taking a job.

Bruce Meyer and Dan Rosenbaum's 2001 analysis attributed a large share of the employment increase among single mothers between 1984 and 1996 to EITC expansions rather than to welfare rules, since much of the rise predates the 1996 law. Rebecca Blank's 2002 review of the evidence concluded that all three forces mattered and that separating them cleanly was not possible with the available variation. The defensible summary is that the reform contributed to a real increase in employment, in an economy that was unusually favorable, alongside a large expansion of work-conditioned tax benefits.

The second true statement: deep poverty rose

Now the other finding, which is equally well documented.

Kathryn Edin and Luke Shaefer examined the number of households with children reporting cash income at or below two dollars per person per day, a threshold borrowed from international poverty measurement. Using Survey of Income and Program Participation data, they found that the number of such households roughly doubled between 1996 and 2011, from a few hundred thousand to over one and a half million, and they were careful to show what happens when you count noncash benefits: including SNAP as if it were cash cuts the figure substantially, roughly in half in their calculations, but does not eliminate the trend.

The mechanism is not mysterious. TANF stopped reaching poor families. The Center on Budget and Policy Priorities tracks a simple ratio, the number of families receiving TANF for every hundred families with children in poverty. In 1996 that ratio was around sixty-eight. Two decades later it had fallen to roughly a fifth of that level, and in a number of Southern states it is in the single digits. Meanwhile benefit levels, which states set, are in no state high enough to lift a family to the poverty line, and in many states a family of three receives less than a fifth of the poverty threshold.

The money did not disappear. It moved. Because TANF funds may be spent on any activity that meets four broad statutory purposes, states redirected the block grant into child welfare services, pre-kindergarten, refundable state tax credits, and in some cases substitution for existing state spending. Nationally, basic cash assistance now accounts for a modest minority of TANF spending. That is legal and in some cases defensible, and it means that a program described in public as welfare is largely not paying cash to anyone.

In short: The rolls fell far faster than poverty did, and a growing group of mothers ended up with neither earnings nor assistance, which is the population researchers call disconnected.

How both statements fit together

Think about the population as a distribution rather than a bloc. Reform pushed hard at one margin: it made receiving aid harder and made taking a job more rewarding. For a mother with a high school diploma, no disability, reliable child care, and a functioning labor market in her county, that push worked. She went to work, her income rose, and she is in the employment statistic.

For a mother with a serious health condition, a child with a disability, a criminal record, limited literacy, or no transportation, the same push produced a different outcome. She was sanctioned off, or timed out, or never applied because the diversion process told her not to bother. She is in the deep poverty statistic.

The reform's design had no mechanism for the second group. The block grant gave states a fixed sum and a caseload target, and serving people with barriers is expensive and lowers your measured work participation rate. The incentives pointed the other way.

Then the test arrived. In the 2008 recession, SNAP caseloads rose steeply and immediately, because SNAP remained an open-ended entitlement. TANF caseloads barely moved. That contrast, in one recession, is the cleanest available demonstration of what the entitlement-to-block-grant conversion actually changed.

Other provisions worth knowing

PRWORA was a large statute and cash assistance was only part of it. It barred most legal immigrants from federal means-tested benefits during their first five years in the country, a provision partly softened in later legislation. It narrowed the childhood disability standard for SSI, removing benefits from a substantial number of children until Congress adjusted the rules. It strengthened child support enforcement, including paternity establishment requirements and the assignment of collected support to the state rather than the family in many cases. It formally delinked Medicaid and food stamp eligibility from cash assistance receipt, so families could keep those benefits after leaving welfare, though enrollment in both fell anyway because the welfare office had been the door.

Common misconceptions

• Welfare reform ended welfare. TANF still exists and still spends federal money. What ended was the individual entitlement and, in most states, the practical availability of cash assistance.
• Caseload decline measures success. The statute treats it that way through the caseload reduction credit, which is precisely the problem, since a family removed by sanction counts the same as a family that found a job.
• The employment gains prove the reform worked. They are real and they coincide with the strongest low-wage labor market in thirty years and a doubling of the EITC for larger families. Careful researchers attribute the change to all three.
• The deep poverty finding is refuted by counting SNAP. Counting SNAP cuts the estimate substantially and leaves a real increase, which Edin and Shaefer reported themselves.
• TANF money still mostly pays benefits to families. Basic assistance is a minority of TANF spending nationally, with the rest spread across child welfare, pre-kindergarten, tax credits, and other state purposes.

Pulling it together

• AFDC was an open-ended individual entitlement; TANF is a fixed, unindexed block grant with no individual right to aid.
• The caseload reduction credit made shrinking the rolls the statutory measure of success, whatever the reason for the shrinkage.
• Employment among single mothers rose substantially in the late 1990s, driven by the reform, an exceptional labor market, and the 1993 EITC expansion together.
• Households with children in extreme cash poverty roughly doubled between 1996 and 2011, and TANF now reaches a small fraction of poor families.
• The 2008 recession tested the design directly: SNAP expanded automatically and TANF did not.

Sources

1. U.S. Department of Health and Human Services, Office of Family Assistance. (n.d.). Temporary Assistance for Needy Families (TANF). acf.hhs.gov
2. Center on Budget and Policy Priorities. (n.d.). Policy Basics: Temporary Assistance for Needy Families. cbpp.org
3. Congressional Research Service. (n.d.). The Temporary Assistance for Needy Families Block Grant: An Overview. crsreports.congress.gov
4. Edin, K. J., and Shaefer, H. L. (2015). $2.00 a Day: Living on Almost Nothing in America. Houghton Mifflin Harcourt. en.wikipedia.org
5. Wikipedia contributors. (n.d.). Personal Responsibility and Work Opportunity Act. en.wikipedia.org

A rollout that became an experiment

Between 1961 and 1975, the food stamp program expanded county by county across the United States. It was not designed as a study. Counties adopted at different times for reasons that had little to do with local health, and by 1975 every county was in. That staggered rollout left behind one of the cleanest natural experiments in American social policy, because a child born in a county in 1968 might have had access from birth while a child born the same year one county over did not.

Douglas Almond, Hilary Hoynes, and Diane Whitmore Schanzenbach used it. Their first paper found that food stamp availability during pregnancy raised birth weights, with the largest effects at the bottom of the birth weight distribution. Their later work followed those children into adulthood and found that access in utero and in early childhood was associated with lower rates of metabolic syndrome, meaning obesity, high blood pressure, and diabetes, decades later, and for women with greater economic self-sufficiency.

That is an unusual kind of finding in this field. A food benefit paid to a household in 1968 shows up in a blood pressure reading in 2008. Keep it in mind as you read the rest of this lesson, because it sets the standard: what a program does to a person over a lifetime is a different question from what it costs this year.

The core of it: The best evidence on antipoverty programs is long-run and follows individuals, and it consistently finds effects that annual budget scoring cannot see.

SNAP: how it works

The Supplemental Nutrition Assistance Program, renamed from food stamps in 2008, is the largest food assistance program in the country and serves roughly forty million people in a typical recent year. Its rules follow the anatomy from Lesson 1 exactly.

Eligibility. Gross monthly income generally at or below 130 percent of the federal poverty guideline, net income at or below 100 percent after deductions for shelter, child care, and, for elderly and disabled households, medical expenses. Asset limits exist in statute but most states use broad-based categorical eligibility, which confers eligibility through receipt of certain other benefits and effectively waives the asset test.

Benefit. The maximum allotment for the household size, minus thirty percent of net income. Average benefits work out to a little over two hundred dollars per person per month in recent years, which is roughly seven dollars a day.

Financing and administration. Federal general revenue pays every dollar of benefits, states share administrative costs, and the program is an open-ended entitlement. That last feature makes SNAP the country's most responsive automatic stabilizer: caseloads and spending rise within months of a downturn without any legislation. The Congressional Budget Office has repeatedly rated increased SNAP benefits among the more cost-effective forms of fiscal stimulus per dollar, because recipients spend the money immediately and locally.

Work rules. Adults aged eighteen to fifty-four without dependents are limited to three months of benefits in a thirty-six month period unless they work or train twenty hours a week, with exemptions including veterans, people experiencing homelessness, and young people who have aged out of foster care. Studies of state-level reinstatements of this time limit, including work by Colin Gray and colleagues using Virginia data, have found large drops in participation with no detectable increase in employment, which is the central empirical objection to the rule.

Two arguments about SNAP worth knowing

Should it just be cash? The economist's standard point is that most SNAP households already spend more on food than their benefit, which makes them inframarginal: the benefit frees up money they were spending anyway, so SNAP behaves close to cash. Hoynes and Schanzenbach found spending patterns broadly consistent with that. If SNAP is nearly cash in effect, the restriction buys little except administrative cost and the appearance of control. The counterargument is political rather than economic: the restriction is why SNAP has survived at a size no cash program has ever reached, and the grocery and agricultural industries that defend it in every farm bill do so because it is a food program.

Should purchases be restricted further? Proposals to exclude sugary drinks recur. The public health case is straightforward. The objections are practical, that defining eligible foods at the register across hundreds of thousands of retailers and hundreds of thousands of products is genuinely difficult, and dignitary, that no other American consumer is told what to buy with their own money. The evidence base on whether restriction changes diet is thinner than either side usually admits.

The earned income tax credit

The EITC began in 1975 as a temporary offset to payroll taxes for working families, became permanent in 1978, and was expanded in 1986, 1990, and most substantially in 1993, when the credit for families with two or more children was roughly doubled. It is now among the largest antipoverty programs in the country, and it is administered by the Internal Revenue Service rather than by any welfare agency.

Its structure has three phases, and understanding them explains almost every argument about it.

Phase	What happens	Effect on the worker
Phase-in	The credit rises with earnings, at 34 percent for one child, 40 percent for two, and 45 percent for three or more	A pure wage subsidy; each hour worked is worth substantially more than the wage
Plateau	The credit stays at its maximum across a band of earnings	No effect on the return to an additional hour
Phase-out	The credit falls as earnings rise, at about 16 percent for one child and 21 percent for two or more	An implicit tax on additional earnings, stacked on payroll and income tax

Notice what this predicts. The credit should strongly encourage people to take a job at all, the extensive margin, and should mildly discourage additional hours for those in the phase-out range, the intensive margin. That is close to what the research finds. Nada Eissa and Jeffrey Liebman showed that the 1986 expansion raised labor force participation among single mothers. Meyer and Rosenbaum found the expansions explained a large share of the participation increase from the mid-1980s through the mid-1990s. Studies by Hilary Hoynes, Douglas Miller, and David Simon have also found improvements in infant health associated with larger credits.

Three honest weaknesses. First, the EITC does nothing at all for a person with no earnings, which by design excludes people who cannot work and people in a labor market with no jobs. Second, it arrives once a year as a lump sum, useful for a car repair or a deposit and useless for a Tuesday in October; an advance payment option existed and was eliminated for lack of use. Third, its improper payment rate is high, estimated by Treasury and the IRS at roughly a quarter to a third of claimed dollars in some years, driven mainly by the complexity of qualifying child residency rules in families where children move between households, not primarily by deliberate fraud. That complexity also produces an enforcement pattern in which low-income filers claiming the EITC have been audited at rates comparable to very high-income taxpayers, because correspondence audits are cheap.

The Speenhamland question, in modern dress

Recall the 1795 dispute from Lesson 2. If the government supplements low wages, does the employer capture part of the subsidy by paying less?

Jesse Rothstein's work on the incidence of the EITC argued yes, in part. Because the credit induces more people to enter low-wage labor markets, it increases labor supply, and in a market where employers have some wage-setting power, an increase in supply pushes wages down. Rothstein's estimates implied that a meaningful share of each dollar of EITC ends up with employers rather than workers, and that some of the cost falls on low-wage workers who do not qualify for the credit, since their wages fall too.

The response is not a denial. It is that the size of the effect depends on how competitive the low-wage labor market is and on what else is happening, and that a minimum wage set alongside the credit blocks the mechanism by putting a floor under the wage. That is why many analysts who support the EITC also support a minimum wage, and why treating the two as substitutes misunderstands both.

Worth holding on to: A wage subsidy and a wage floor are complements, not alternatives, because the floor is what prevents the subsidy from being partly captured by employers.

The child tax credit, and six months in 2021

The child tax credit was created in 1997 as a modest nonrefundable credit, raised to 1,000 dollars and made partly refundable in 2001, and raised to 2,000 dollars in the 2017 tax law with a refundable portion capped and available only above an earnings threshold.

That earnings threshold is the design feature that matters most. Because refundability phased in with earnings, families with the lowest incomes received a partial credit or none. Analysts at the Tax Policy Center and the Center on Budget and Policy Priorities estimated that roughly a third of children, on the order of twenty-seven million, lived in families receiving less than the full credit because their earnings were too low. A child benefit that gives least to the poorest children is an unusual design, and it was a deliberate one, intended to condition the benefit on work.

The American Rescue Plan Act changed all of it for one year. For 2021 the credit rose to 3,600 dollars for children under six and 3,000 dollars for children aged six to seventeen, became fully refundable with no earnings requirement, and was paid out monthly from July through December to most families automatically, reaching roughly sixty-one million children in a typical monthly payment according to Treasury and IRS reporting.

The Census Bureau's supplemental poverty measure put child poverty at 5.2 percent in 2021, the lowest ever recorded in that series. The expansion expired. Child poverty on the same measure was 12.4 percent in 2022.

The employment dispute, at full strength on both sides

Would a permanent version reduce work? This is the most substantive current argument in American antipoverty policy, and it is not a matter of good faith versus bad.

The case that it would not: several research teams examined the six months of monthly payments using contemporaneous survey data and found no statistically detectable reduction in employment among recipient parents. Work by Elizabeth Ananat and colleagues and by teams at Columbia's Center on Poverty and Social Policy reported this consistently. Recipients used the money for food, rent, utilities, school costs, and child care, and child care spending in particular can raise employment rather than lower it.

The case that it would: Kevin Corinth, Bruce Meyer, Matt Stadnicki, and Derek Wu argued that six months of a benefit announced as temporary is not a test of a permanent one, because people do not restructure their working lives around a payment they expect to end. Using a structural labor supply model calibrated to the existing literature on how people respond to unconditional income, they projected that a permanent, fully refundable credit would reduce employment by roughly 1.5 million workers and would therefore offset a substantial share, on their estimates around a third, of the measured poverty reduction.

Notice what actually separates them. It is not a disagreement about the 2021 data. It is a disagreement about whether short-run evidence from a temporary program can identify long-run responses to a permanent one, and about how much weight to give a model calibrated on older studies of very different programs. What would settle it is a permanent policy or a long-running randomized trial, and neither exists. A reader who wants a resolution now will have to accept an argument from prior beliefs, and should say so out loud rather than pretending the evidence decided.

Common misconceptions

• SNAP is mostly used by people who do not work. A large share of SNAP households with a nonelderly, nondisabled adult include a worker, and many households cycle on and off as hours fluctuate.
• The EITC helps the poorest families most. It gives nothing to those with no earnings and reaches its maximum well above the bottom of the distribution, which is why it and SNAP do different jobs.
• The EITC's improper payment rate reflects widespread fraud. Most of it traces to qualifying child residency rules in complicated households, which is an argument for simpler rules rather than more audits.
• The 2021 child tax credit was a new welfare program. It was an existing tax credit with its earnings requirement removed and its payment schedule changed from annual to monthly.
• The employment effect of a permanent child allowance is known. It is genuinely disputed, and the dispute is about the inference from six temporary months to a permanent policy, not about the 2021 numbers.

What to remember

• The staggered food stamp rollout showed effects on birth weight and on adult metabolic health decades later, setting the standard for what long-run evidence looks like.
• SNAP is an open-ended entitlement with a benefit equal to the maximum allotment minus thirty percent of net income, and it is the country's fastest automatic stabilizer.
• The EITC phases in as a wage subsidy, plateaus, and phases out as an implicit tax, which predicts its strong effect on whether people work and weak effect on how many hours.
• Rothstein's incidence work revives the Speenhamland question: part of a wage subsidy can be captured by employers unless a wage floor blocks it.
• The 2021 child tax credit removed the earnings requirement and paid monthly; child poverty on the supplemental measure fell to 5.2 percent and returned to 12.4 percent when it lapsed, and the employment effect of a permanent version remains contested.

Sources

1. U.S. Department of Agriculture, Food and Nutrition Service. (n.d.). SNAP eligibility and benefit rules. fns.usda.gov
2. Internal Revenue Service. (n.d.). Earned Income Tax Credit. irs.gov
3. Internal Revenue Service. (n.d.). Child Tax Credit. irs.gov
4. U.S. Census Bureau. (n.d.). Poverty in the United States: supplemental poverty measure reports. census.gov
5. Center on Budget and Policy Priorities. (n.d.). Policy Basics: The Earned Income Tax Credit. cbpp.org

16 March 1972, St. Louis

On that afternoon the St. Louis Housing Authority dynamited one of the thirty-three eleven-story buildings of the Pruitt-Igoe complex. The development had opened in the mid-1950s with roughly 2,870 apartments, designed by Minoru Yamasaki, and had been praised in the architectural press. Within fifteen years its vacancy rate was catastrophic, its elevators and heating had failed, and demolition was cheaper than repair. The architectural critic Charles Jencks later used the demolition as a marker for the death of modernist architecture, and the image of the collapsing tower became shorthand for the failure of public housing as an idea.

The shorthand is wrong, and unpacking why teaches most of what you need to know about American housing policy. Pruitt-Igoe did not fail because poor people cannot maintain buildings, and it did not fail because of its architecture. It failed because of its financing.

Federal law paid to build public housing. It did not pay to operate it. Operating costs, meaning heat, elevators, plumbing, staff, and repairs, were to be covered entirely by tenant rents. That arrangement can work only if enough tenants pay enough rent. St. Louis lost roughly half its population between 1950 and 1980. Employed families with any choice left the complex. The remaining households were poorer, so rent revenue fell while the buildings aged and costs rose. Then in 1969 the Brooke Amendment capped what any public housing tenant could be charged at a quarter of income, later raised to thirty percent, which was a humane and necessary protection and which also severed the last connection between operating costs and operating revenue. Congress created federal operating subsidies to fill the gap and never appropriated enough to fill it.

What matters here: Pruitt-Igoe is a story about a financing structure with no operating revenue, not a story about design or about tenants, and the housing scholar Katharine Bristol made that case in print as early as 1991.

How the federal government got into housing

The Housing Act of 1937 created public housing, and its terms were shaped by an industry that did not want it. Real estate and homebuilding interests accepted the program on conditions: it would serve only the very poor, it would be built and run by local housing authorities rather than by Washington, and each new unit would require the elimination of a unit of substandard housing, which capped net supply from the start.

The Housing Act of 1949 set a national goal of a decent home and a suitable living environment for every American family, and simultaneously created Title I urban renewal, which paid cities to acquire and clear land. Over the following two decades urban renewal and interstate highway construction demolished a great deal of occupied housing, disproportionately in Black neighborhoods, and replaced much less of it than was destroyed. Households displaced by clearance were frequently offered nothing, and where public housing was built to receive them it was sited to preserve segregation rather than reduce it.

HUD was created as a cabinet department in 1965. By the early 1970s the political appetite for building and owning housing had collapsed, and policy turned to a different instrument.

How a voucher works

The Housing and Community Development Act of 1974 created what became Section 8, and its existing-housing component grew into today's Housing Choice Voucher program. Learn the mechanics precisely, because families ask about them constantly.

A public housing agency issues a voucher to a household. The household finds a unit in the private market that meets housing quality standards and whose rent the agency approves. The household pays approximately thirty percent of its adjusted income toward rent and utilities. The agency pays the landlord the difference, up to a payment standard the agency sets in relation to HUD's fair market rent for the area, which is generally set around the fortieth percentile of local rents and in a growing number of metropolitan areas is calculated by ZIP code rather than for the whole region. A family may rent a more expensive unit and pay the excess itself, subject to a cap of forty percent of income at initial lease-up.

The design has real strengths. It uses existing housing rather than waiting for construction. It lets a family move without losing assistance. It costs less per household than new construction. And in principle it gives a family a choice of neighborhood.

Now the problems, in order of severity.

It is not an entitlement. This is the central fact of American housing policy and the one that surprises students most. SNAP pays every eligible household. Rental assistance is funded by annual discretionary appropriation, so it pays as many households as the appropriation covers and no more. Roughly one in four households eligible for federal rental assistance actually receives any, according to HUD and Center on Budget and Policy Priorities analyses. Waiting lists run for years and are frequently closed to new applicants entirely.

A voucher is not a home. A substantial share of families who receive a voucher never manage to use it, because they cannot find a landlord who will accept it within the search period. Refusal to rent to voucher holders is lawful in much of the country; a growing number of states and cities have passed source of income antidiscrimination laws, with mixed enforcement.

Vouchers cluster. Because payment standards were historically set for an entire metropolitan area, the units a voucher could actually reach were concentrated in lower-rent, higher-poverty neighborhoods, which undercut the choice the program was supposed to provide. Small area fair market rents, set by ZIP code, are the main policy response.

Moving to Opportunity, and why the answer changed

Between 1994 and 1998, HUD ran a randomized experiment in Baltimore, Boston, Chicago, Los Angeles, and New York. About 4,600 families living in high-poverty public housing were randomly assigned to one of three groups: an experimental group offered a voucher usable only in a census tract with a poverty rate under ten percent, plus search counseling; a comparison group offered a standard unrestricted voucher; and a control group offered neither.

The interim and final evaluations found substantial improvements for adults in the experimental group in mental health, in feelings of safety, and in physical health outcomes including obesity and diabetes. They found essentially no effect on adult earnings or employment, and no clear overall effect on children's test scores. For a program justified largely on economic mobility, that looked like a null result, and for roughly a decade it was reported as one.

Then Raj Chetty, Nathaniel Hendren, and Lawrence Katz linked the experiment's participants to federal tax records and looked at the children as adults. Age at move turned out to be decisive. Children who moved to a low-poverty neighborhood before about age thirteen had substantially higher earnings in their mid-twenties, on the order of thirty percent above the control group, along with higher college attendance and lower rates of single parenthood. Children who moved as older teenagers showed no gain and by some measures did slightly worse, consistent with the disruption of moving late in adolescence.

The methodological lesson is as important as the finding. The original evaluation measured the right people at the wrong time. Neighborhood effects on children accumulate with exposure and appear in adult outcomes, and an evaluation that ends four to seven years after random assignment cannot see them.

A follow-on experiment in Seattle and King County, Creating Moves to Opportunity, tested whether the barrier was information and search rather than money. Families offered housing search assistance, landlord recruitment, and short-term financial help moved to high-opportunity neighborhoods at a far higher rate than families offered the voucher alone, rising from around fifteen percent to over half. The voucher was not the binding constraint; the search was.

Key idea: A voucher buys purchasing power, not access. Whether it produces a move to a different kind of neighborhood depends on landlord acceptance, payment standards, and search support, all of which are separate policy levers.

The other housing programs, briefly

Instrument	Created	How it works	Main limitation
Public housing	1937	Publicly owned and operated units, rent capped at 30 percent of income	Chronic operating subsidy shortfall; the 1998 Faircloth limit caps unit counts
Housing choice vouchers	1974	Tenant-based subsidy in private market units	Not an entitlement; landlord refusal; payment standards
Low-Income Housing Tax Credit	1986	State-allocated tax credits sold to investors to finance construction and rehabilitation	Rents generally target moderate rather than deep affordability; needs a voucher on top for the poorest
HOPE VI and RAD	1992 and 2012	Redevelopment of distressed public housing and conversion to project-based rental assistance	Redevelopment has often produced fewer deeply subsidized units than it replaced
Homeownership tax subsidies	Various	Mortgage interest deduction, property tax deduction, capital gains exclusion	Flow overwhelmingly to higher-income owners and dwarf direct rental assistance in cost

That last row is the one students most often have never considered. The largest federal housing expenditures in the United States are delivered through the tax code to homeowners, not through HUD to renters. Any statement about how much the country spends on housing assistance that omits tax expenditures is describing a fraction of the system.

Homelessness and Housing First

Homelessness policy deserves its own note because the evidence is unusually clear on one question. The traditional continuum of care required a person to achieve sobriety and treatment compliance before receiving permanent housing. Housing First, developed by Sam Tsemberis and Pathways to Housing in New York in the 1990s, reverses the order: provide permanent housing without preconditions, then offer voluntary services.

The strongest test is the Canadian At Home / Chez Soi trial, which randomized more than two thousand people with mental illness experiencing homelessness across five cities. Housing First produced large and durable improvements in housing stability. The evidence for effects on substance use and psychiatric symptoms is weaker and more mixed, and honest advocates say so. The policy conclusion that follows is narrower than the slogan: Housing First reliably houses people, and housing people is worth doing on its own terms, whether or not it also treats them.

Underneath all of it sits supply. Rental assistance helps a household compete for a unit; it does not create the unit. Where housing production is constrained by land use regulation, subsidies raise what families can pay without raising how much housing exists, and a share of the subsidy is absorbed into rents. That is the same incidence question you met with the EITC, in a different market.

Common misconceptions

• Public housing failed because of bad architecture or bad tenants. The 1937 design funded construction and not operations, and the Brooke Amendment capped the only revenue source, producing a structural deficit that Congress never fully filled.
• Everyone who qualifies for housing assistance can get it. Rental assistance is discretionary, not an entitlement, and roughly three in four eligible households receive nothing.
• A voucher means a family can move anywhere. Landlords may refuse vouchers in much of the country, and payment standards determine which neighborhoods are actually reachable.
• Moving to Opportunity showed that neighborhoods do not matter. It showed no adult earnings effect and large child effects that only became visible when researchers looked at those children as adults.
• Federal housing spending goes mainly to the poor. Homeownership tax expenditures exceed direct rental assistance and flow mostly to higher-income households.

What you now know

• Public housing's core flaw was financing construction without operations, made structural by the Brooke Amendment's rent cap and chronic subsidy shortfalls.
• A voucher pays the difference between thirty percent of a household's income and a payment standard tied to fair market rent.
• Rental assistance is rationed by appropriation, so eligibility and receipt are different things for most poor renters.
• Moving to Opportunity found no adult earnings effect and, for children who moved before roughly age thirteen, substantially higher adult earnings; search assistance, not the voucher alone, drives moves to high-opportunity areas.
• Housing First reliably produces housing stability; its effects on substance use and psychiatric symptoms are weaker and contested.

Sources

1. U.S. Department of Housing and Urban Development. (n.d.). Housing Choice Voucher Program Section 8. hud.gov
2. U.S. Department of Housing and Urban Development, Office of Policy Development and Research. (n.d.). Moving to Opportunity for Fair Housing. huduser.gov
3. Center on Budget and Policy Priorities. (n.d.). Policy Basics: Federal Rental Assistance. cbpp.org
4. Encyclopaedia Britannica. (n.d.). Pruitt-Igoe. britannica.com
5. Wikipedia contributors. (n.d.). Low-Income Housing Tax Credit. en.wikipedia.org

One sentence, written in 1956

Here is the legal test that determines whether roughly twelve million Americans receive a monthly check. A person is disabled if they are unable to engage in any substantial gainful activity by reason of any medically determinable physical or mental impairment which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than twelve months.

Read it again slowly, because every word is load-bearing and each one has produced decades of adjudication.

Any substantial gainful activity. Not your previous job. Not a job in your city. Any job that exists in significant numbers in the national economy. A former roofer who can no longer climb but could, in principle, do sedentary assembly work is not disabled under this test, whatever the local labor market looks like.

Medically determinable. Documented by objective medical evidence from an acceptable source. Pain is real, and pain without a documented underlying impairment does not establish disability.

Twelve months or death. There is no partial and no temporary disability in this system. A person who will be unable to work for nine months receives nothing at all.

The substantial gainful activity threshold is a specific dollar amount published each year and adjusted for wage growth, and in recent years it has sat in the range of fifteen hundred dollars a month for non-blind applicants. Earn above it and you are, by definition, not disabled.

Bottom line: American disability policy uses a single all-or-nothing definition built around long-term total incapacity for any work, which is why so many people with serious, partial, or fluctuating conditions fall outside it entirely.

Two programs, same definition, different worlds

	SSDI (Title II)	SSI (Title XVI)
Basis of claim	Insured status from covered work	Need, plus age, blindness, or disability
Financing	Payroll tax through the disability trust fund	Federal general revenue
Financial test	None on assets or unearned income	Countable resources under 2,000 dollars for an individual, 3,000 for a couple
Benefit size	Based on lifetime earnings using the same progressive formula as retirement	A flat federal benefit rate, reduced by countable income, sometimes supplemented by the state
Health coverage	Medicare, after twenty-four months of entitlement	Medicaid, automatic in most states
Waiting period	Five full months before the first payment	None

The Medicare waiting period deserves a moment. A worker whose condition has already lasted twelve months waits five more months for cash and twenty-four months from entitlement for Medicare. Someone disabled by a heart condition can therefore spend more than two years disabled, unable to work, and without the coverage that would treat the condition. The exceptions are amyotrophic lateral sclerosis and end-stage renal disease, which Congress carved out.

The SSI resource limits carry their own consequences. Two thousand dollars is not a great deal of money to hold across a lifetime, and the limit has not moved since 1989. It means a recipient cannot build an emergency fund, cannot accept a modest inheritance without losing benefits and often Medicaid, and faces a marriage penalty because the couple rate is one and a half times the individual rate rather than twice it. SSI also reduces the benefit for in-kind support and maintenance, meaning that a relative who provides free housing or regular groceries can reduce the check. Practitioners spend a remarkable amount of time explaining that last rule to families who were trying to help.

The ABLE Act of 2014 created tax-advantaged savings accounts for people whose disability began before a specified age, allowing savings above the SSI resource limit to be held without loss of eligibility, within annual and total caps. It is a genuine improvement and it does not solve the problem for people whose disability began later in life.

The five steps

Every claim runs through the same sequence, and knowing it lets you tell a client where their case actually sits.

Step one: are you working above substantial gainful activity? If yes, denied. The step is mechanical and ends many claims immediately.

Step two: is your impairment severe? Does it significantly limit basic work activities and meet the duration requirement? This is a low bar and screens out minor conditions.

Step three: does it meet or medically equal a Listing? The Listing of Impairments is a catalog of conditions with specified clinical findings. Meet a listing and you are allowed at step three without any assessment of your ability to work. Most claimants do not meet a listing.

Step four: can you do your past relevant work? The adjudicator assesses residual functional capacity, the most you can still do despite limitations, and compares it to jobs you have held. If you can still do one of them, denied.

Step five: can you do any other work? Here age, education, and transferable skills enter through a set of tables called the medical-vocational guidelines, known as the grids. The grids are why age matters so much: a fifty-five-year-old with limited education, a lifetime of heavy labor, and a sedentary residual capacity is directed to an allowance, while a thirty-five-year-old with identical medical findings is directed to a denial.

The procedural path is as consequential as the test. A state disability determination service decides the initial claim. Most are denied. The claimant may request reconsideration, then a hearing before an administrative law judge, then Appeals Council review, then federal district court. Allowance rates rise substantially at the hearing level, partly because claimants are usually represented by then and partly because time has passed and conditions have progressed. Hearing backlogs have at times exceeded a year and a half, during which an applicant by definition cannot work above the threshold and has no benefit.

The upshot: A disability claim is won or lost as much on documentation, representation, and endurance as on medical severity, which is precisely where a social worker can change an outcome.

Children, and what 1996 did

Children can receive SSI, and the standard is different because children do not have work histories. In Sullivan v. Zebley in 1990, the Supreme Court held that children whose impairments did not meet a listing were entitled to an individualized functional assessment comparable to the adult step-five analysis. Awards rose.

PRWORA in 1996 replaced that standard with a requirement of marked and severe functional limitations and eliminated a category of maladaptive behavior findings. Tens of thousands of children lost benefits. Congress later adjusted parts of the rule. The episode is a clean illustration of something you saw in Lesson 1: the eligibility criterion, not the benefit amount, was where the policy moved.

Why the rolls grew, argued both ways

The number of disabled workers receiving SSDI rose from under three million in the mid-1980s to over eight million by 2010, then declined. That growth drove a serious argument about whether the program had become something other than what it was designed to be.

David Autor and Mark Duggan made the case that the program had become an absorber of displaced low-wage workers. Their argument had three parts. Congress liberalized the medical criteria in 1984, giving more weight to pain, mental impairment, and combinations of conditions, and shifting many claims from the mechanical listings to the more judgment-dependent step five. The progressive benefit formula means the replacement rate for a low-wage worker is high, so as wages at the bottom stagnated, the value of the benefit relative to available work rose. And demand for the kind of physical labor those workers had done fell. Put together, they argued, a worker in his fifties whose plant closed faced a choice between a low-wage job and a benefit worth nearly as much, with Medicare attached.

The competing account, argued by the Social Security Administration's actuaries and by the Congressional Budget Office, is largely demographic. The baby boom generation moved into its fifties and sixties, the ages at which disability incidence rises steeply. Women's labor force participation had risen for decades, so far more women had the insured status needed to qualify, and their receipt rates converged toward men's. The full retirement age rose, which keeps people on disability longer before conversion to retirement benefits. Adjust for age and sex composition, they argue, and most of the growth disappears.

The evidence favors the demographic account for most of the trend and leaves room for the Autor and Duggan mechanism at the margin, particularly in regions with collapsing industrial employment. Notice also what happened after 2014: the rolls declined, which is difficult to reconcile with a story of ever-loosening standards and easy to reconcile with the baby boom aging into retirement benefits.

The other disability policy, which is not a benefit

Cash benefits are only half of American disability policy, and social workers sometimes forget the other half.

Section 504 of the Rehabilitation Act of 1973 barred discrimination on the basis of disability in federally funded programs, and then sat unimplemented for four years because no agency issued regulations. In April 1977, disabled activists occupied the San Francisco offices of the Department of Health, Education, and Welfare for twenty-six days, the longest occupation of a federal building in American history, supported by the Black Panther Party, which delivered food. The regulations were signed.

The Americans with Disabilities Act of 1990 extended nondiscrimination to employment, public services, and public accommodations. It is a civil rights statute, not a benefits statute, and it rests on the social model of disability: the claim that disability is produced by the interaction between an impairment and an environment built for other bodies, rather than residing in the body alone. In Olmstead v. L.C. in 1999, the Supreme Court held that unjustified institutional isolation of people with disabilities is discrimination under the ADA, which became the legal foundation for shifting long-term services from institutions to home and community-based settings.

Hold the tension between the two halves. The benefit system requires a person to prove total incapacity for any work. The civil rights system insists that people with disabilities can work if barriers are removed. A client can be told both things in the same week by two arms of the same government.

Common misconceptions

• SSDI and SSI are the same program. One is contributory insurance with Medicare attached; the other is means-tested assistance with Medicaid attached, and a person can receive both concurrently.
• You can qualify for partial disability. The federal definition is all or nothing. Partial and temporary disability exist in workers compensation and in some private insurance, not in Social Security.
• Most claims are approved at first. Most initial claims are denied, and a large share of eventual awards come after a hearing before an administrative law judge.
• Saving money will not affect benefits. For SSI it will, above 2,000 dollars for an individual, and gifts of housing or food can also reduce the check.
• Working ends benefits immediately. SSDI provides a trial work period of nine months and an extended period of eligibility, and SSI recipients can often keep Medicaid under section 1619(b) after cash benefits stop.

Summing up

• One statutory definition, built on inability to do any substantial gainful activity for twelve months or until death, governs both programs.
• SSDI is payroll-tax insurance with Medicare after twenty-four months; SSI is general-revenue assistance with a 2,000 dollar resource limit unchanged since 1989 and Medicaid in most states.
• The five-step sequence ends most claims at step one, three, four, or five, and age enters decisively through the medical-vocational grids.
• The growth of the rolls from the mid-1980s is best explained mostly by aging and by women's insured status, with the Autor and Duggan labor market mechanism operating at the margin.
• The ADA and Olmstead run on the opposite premise from the benefit test, and clients live inside that contradiction.

Sources

1. Social Security Administration. (n.d.). Disability Benefits. ssa.gov
2. Social Security Administration. (n.d.). Disability Evaluation Under Social Security (the Blue Book). ssa.gov
3. Social Security Administration. (n.d.). Understanding Supplemental Security Income: resources. ssa.gov
4. U.S. Department of Justice. (n.d.). Olmstead: Community Integration for Everyone. ada.gov
5. Congressional Budget Office. (n.d.). Social Security Disability Insurance. cbo.gov

The story everyone tells, and the part that is wrong

In 1874 a Methodist mission worker named Etta Wheeler learned that a girl of about nine, Mary Ellen Wilson, was being beaten and kept confined in a tenement on West 41st Street in New York. Wheeler could find no agency with authority to intervene. She approached Henry Bergh, founder of the American Society for the Prevention of Cruelty to Animals. The case went to court, Mary Ellen testified, her guardian was convicted, and in 1875 the New York Society for the Prevention of Cruelty to Children was founded, the first organization of its kind in the world.

Now the correction, because the version you have probably heard is a story about a legal absurdity. It is usually said that Mary Ellen was rescued under laws protecting animals, since no law protected children. That is not what happened. Bergh acted as a private citizen and through his lawyer Elbridge Gerry, and the case proceeded on a writ used to remove a person from unlawful custody. Children were not legally outside the protection of the law in 1874. What was missing was not a statute but an organization willing to use one.

That distinction is the theme of this lesson. Child welfare policy is rarely short of legal authority. It is short of the money, the alternatives, and the incentives that would let the authority be used well.

Key idea: Almost every failure in child protection is a failure of capacity and incentive rather than of law, which is why federal reform in this field works by changing what money pays for.

From orphan trains to a diagnosis

Before there was a system there were institutions and relocation. Charles Loring Brace's Children's Aid Society began sending city children west by rail in 1853, and over roughly seventy-five years the orphan trains placed something on the order of two hundred thousand children with rural families, with almost no screening and little follow-up. The 1909 White House Conference on the Care of Dependent Children declared that children should not be removed from their homes for reasons of poverty alone, a principle it has taken the field more than a century to approach. The Children's Bureau followed in 1912.

The modern system begins with a medical paper. In 1962, C. Henry Kempe and colleagues published The Battered-Child Syndrome in the Journal of the American Medical Association, describing a pattern of skeletal injuries in young children that radiologists had been seeing and not naming. The paper made physical abuse a diagnosable condition, and diagnosable conditions get reported. Within five years every state had passed a mandatory reporting law.

The federal framework arrived in 1974 with the Child Abuse Prevention and Treatment Act, sponsored by Walter Mondale. CAPTA set a federal definition of abuse and neglect, conditioned state grants on having reporting systems and investigation procedures, and required that a guardian ad litem be appointed for a child in every abuse or neglect proceeding, which is where the court appointed special advocate programs come from. Reports climbed steeply, as reporting laws with a hotline attached will always cause them to.

The four statutes that shape practice

Law	Year	What it required	Incentive it created
CAPTA	1974	State reporting systems, investigation, guardians ad litem	More reports, and an investigation-centered front door
Indian Child Welfare Act	1978	Tribal jurisdiction and notice, placement preferences, active efforts, higher evidentiary standards	Tribes as parties rather than interested observers
Adoption Assistance and Child Welfare Act	1980	Reasonable efforts to prevent removal and to reunify, case plans, periodic review, Title IV-E funding	Federal money followed the child into foster care, not into prevention
Adoption and Safe Families Act	1997	Child health and safety as paramount, termination petitions when a child has been in care fifteen of the most recent twenty-two months, adoption incentive payments	Faster permanency; pressure toward termination of parental rights
Family First Prevention Services Act	2018	Title IV-E funds usable for evidence-based prevention services; limits on federal payment for congregate care	For the first time, federal money for keeping a family together

Read the incentive column down and you can see the whole arc. For nearly forty years the main federal funding stream, Title IV-E, was an open-ended entitlement that paid for a child's maintenance in foster care and paid nothing for services that would have prevented the removal. Agencies were not perverse for placing children; the money was structured to pay for placement. Family First is the first serious attempt to change that, allowing IV-E funds to pay for mental health treatment, substance use treatment, and in-home parenting programs that meet evidence standards, while restricting federal reimbursement for group placements beyond a short period unless the setting is a qualified residential treatment program.

The Indian Child Welfare Act deserves its own paragraph. It passed in 1978 after congressional hearings established that a very large share of Native children, on the order of a quarter to a third depending on the state, had been removed from their families and placed overwhelmingly in non-Native homes and institutions. ICWA establishes tribal jurisdiction, requires notice to the tribe, sets placement preferences favoring extended family, other tribal members, and other Native families, requires active efforts rather than merely reasonable ones, and raises the evidentiary standard for removal and termination. It was challenged as unconstitutional and the Supreme Court rejected those challenges in Haaland v. Brackeen in 2023.

What the caseload actually looks like

Three facts organize everything else.

Most of the caseload is neglect, not abuse. Roughly three quarters of substantiated maltreatment findings in the federal Child Maltreatment reports are neglect. Physical abuse is a much smaller share, sexual abuse smaller still. Neglect is defined by state law in terms that frequently name conditions produced by poverty: inadequate food, unstable housing, lack of supervision, missed medical appointments, utility shutoffs.

Contact with the system is common, not rare. Research by Hyunil Kim, Christopher Wildeman, Melissa Jonson-Reid, and Brett Drake estimated that more than a third of American children experience a child protective services investigation at some point before their eighteenth birthday, with the figure above half for Black children. Whatever you think about the system, it is not a marginal institution.

Placement is racially patterned. Black children and Native children are represented in foster care at rates well above their share of the child population. Whether this reflects differential need arising from differential poverty, differential surveillance and reporting, differential decision-making at each step, or all three, is one of the central empirical questions in the field.

Why this matters: A system in which three quarters of findings are neglect and neglect tracks material hardship is, functionally, an income policy that operates through investigations.

The evidence that income changes the caseload

If neglect and poverty are entangled, then income policy should move maltreatment reports. It does.

Kerri Raissian and Lindsey Rose Bullinger examined state minimum wage increases and found that a one dollar increase was associated with a meaningful decline in reports of neglect involving young children. Other work has found reductions in maltreatment reports associated with more generous earned income tax credits, with expanded Medicaid eligibility, and with the receipt of emergency cash assistance. A randomized study of unconditional cash to low-income mothers, and evaluations of eviction prevention programs, point the same direction.

None of this says maltreatment is only poverty. Serious abuse occurs in affluent households and is under-detected there. What it says is that a substantial share of what the system currently addresses through investigation could be addressed through money, and that the choice of instrument is a policy decision rather than a fact of nature.

The argument about the system itself

You should be able to state both of the following positions well enough that someone holding them would recognize their own view.

The family policing critique. Dorothy Roberts, in Shattered Bonds and more recently Torn Apart, argues that the child welfare system functions as a mechanism of surveillance and control directed at poor Black families. Her evidence: the concentration of investigations in poor Black neighborhoods; the routine reporting of poverty as neglect; the fact that mandated reporting turns doctors, teachers, and social workers into agents of an investigative apparatus, which deters families from seeking help; the trauma of removal itself; and the outcomes of children who age out of foster care, which are poor by almost every measure. Her conclusion is not reform but abolition of the system in its current form and replacement with material support and voluntary services.

The child safety response. Its proponents point out that children die. Federal data record on the order of two thousand child maltreatment fatalities a year, concentrated overwhelmingly among children under four, and that most had prior contact with the system. They point to periods and jurisdictions in which reductions in removals were followed by increases in serious injury. And they note that the counterfactual to removal is not a supported family but, in many cases, the same household without intervention.

The best evidence sits uncomfortably between them. Joseph Doyle used the effectively random assignment of investigators with different removal tendencies to study children on the margin of placement, meaning children whose removal depended on which investigator drew the case. He found that marginal children placed in foster care had worse outcomes in adolescence and early adulthood than similar children left at home, including higher rates of juvenile delinquency and teen pregnancy and lower employment. Later work using similar designs in other states, including research by Max Gross and E. Jason Baron in Michigan, has found different results for some outcomes, including improvements in child safety and schooling. The honest summary is that for children at the margin, removal is not clearly protective, and that the marginal case is not the severe case.

What would settle it is better evidence about which children are on which margin, and that evidence is genuinely hard to get, because the decision to remove and the risk to the child are entangled by design.

Where the social worker stands

You will be a mandated reporter. In most states that obligation attaches to reasonable suspicion, not to certainty, and it is not discretionary. You will also be the person a family tells the truth to, and both of those things are true at once.

Practically, three things are within your control. First, tell families what your reporting obligation is before they tell you something, not after; concealing it costs trust you will need later. Second, know the difference in your state between a report and an investigation, and whether your jurisdiction offers a differential or alternative response track that connects a family to services without a formal finding. Third, document material hardship precisely, because the difference between a neglect finding and a service referral often turns on whether the record shows a parent who would not provide or a parent who could not.

Common misconceptions

• Mary Ellen Wilson was rescued under animal cruelty law. Bergh acted as a private citizen and the case used an ordinary writ; the missing element was an organization, not a statute.
• Most child welfare cases involve physical or sexual abuse. Roughly three quarters of substantiated findings are neglect, which overlaps heavily with poverty.
• ASFA simply sped up adoptions. It also created a fifteen of twenty-two month clock for filing a termination petition, which runs while a parent is on a waiting list for treatment or is incarcerated.
• Federal funding is neutral between prevention and placement. Title IV-E paid for foster care maintenance and not for prevention services for nearly forty years, and Family First is the first substantial correction.
• Research shows foster care harms children. Research using marginal-case designs shows worse outcomes for children at the margin of placement in some studies and better outcomes in others, which is a narrower and more useful claim.

What to carry forward

• CAPTA built the reporting front door, the 1980 act created reasonable efforts and IV-E foster care funding, ASFA imposed permanency timelines, and Family First finally allowed federal money for prevention.
• ICWA gives tribes jurisdiction, notice, placement preferences, and a higher evidentiary standard, and it survived constitutional challenge in 2023.
• Neglect dominates the caseload, more than a third of American children experience an investigation, and placement is racially patterned.
• Minimum wage increases, EITC expansions, and cash assistance are associated with reductions in neglect reports, which makes income policy a child welfare instrument.
• Roberts's abolition argument and the child safety response both rest on real evidence, and marginal-case studies of removal have produced results that point in different directions.

Sources

1. U.S. Department of Health and Human Services, Children's Bureau. (n.d.). Child Maltreatment reports. acf.hhs.gov
2. Child Welfare Information Gateway. (n.d.). Major federal legislation concerned with child protection. childwelfare.gov
3. U.S. Department of the Interior, Bureau of Indian Affairs. (n.d.). Indian Child Welfare Act. bia.gov
4. Encyclopaedia Britannica. (n.d.). Child abuse. britannica.com
5. Wikipedia contributors. (n.d.). Adoption and Safe Families Act. en.wikipedia.org

Day twenty-one

Medicare pays the entire cost of a semiprivate room in a skilled nursing facility for twenty days. On day twenty-one a daily coinsurance begins, reset each year and running above 200 dollars a day since 2023. On day 101 Medicare pays nothing. Those three sentences are the single most consequential thing an American family learns too late, and you will be the person explaining them in a hallway outside a hospital room.

Before any of that clock starts, the person has to clear a gate. Medicare covers skilled nursing facility care only after a qualifying inpatient hospital stay of at least three consecutive days, not counting the day of discharge. The word doing the work in that sentence is inpatient. A patient can occupy a hospital bed for four nights, receive tests and medication, be visited by physicians, and still be classified as an outpatient receiving observation services. Those nights do not count. Families discover this when the nursing home asks for a credit card.

Congress responded to the observation problem with a disclosure requirement rather than a coverage change. The NOTICE Act of 2015 obliges hospitals to hand a patient who has been under observation for more than twenty-four hours a written notice explaining that they are an outpatient and what that means for nursing home coverage. The patient is told. The patient is still not covered.

The point: Medicare is not a long-term care program and never was. Its nursing facility benefit is short-term rehabilitation attached to a hospitalization, gated by a status the patient cannot see and did not choose.

Skilled, custodial, and the line between them

Medicare pays for skilled care, meaning services that require a licensed nurse or therapist: wound care, intravenous medication, physical therapy after a hip replacement. It does not pay for custodial care, meaning help with the activities of daily living. Those are bathing, dressing, toileting, transferring, continence, and eating. Assistance with all six of them, seven days a week, for years, is what most people mean when they say long-term care, and it is precisely what Medicare excludes.

The same line runs through the home health benefit. Medicare will pay for home health when a physician certifies that the person needs intermittent skilled nursing or therapy and is homebound, meaning leaving home takes a considerable and taxing effort. Under that certification an aide may also help with bathing. Remove the skilled need and the aide goes too. A person who is stable, cognitively impaired, and unable to dress herself is not eligible for anything, because nothing about her situation is skilled.

Spend down: how Medicaid became the long-term care program

Medicaid is the largest single payer for long-term services and supports in the United States. It arrived at that position by default. Because Medicare stops and private insurance is thin, the way most Americans finance years of care is to exhaust their own money and then qualify for a means-tested program.

The mechanics matter, so learn them in order. Countable assets for an individual seeking Medicaid long-term care are generally limited to about 2,000 dollars in most states, the same figure inherited from SSI. A home is usually exempt while the person or a spouse lives in it, subject to an equity limit that the Deficit Reduction Act of 2005 set in the hundreds of thousands of dollars and that rises with inflation. Income above the state's limit generally has to be applied to the cost of care.

Giving assets away does not work, and the rule against it is more punishing than most people expect. States review transfers made for less than fair market value during a look-back period of sixty months, extended to that length by the Deficit Reduction Act of 2005. A disqualifying transfer produces a penalty period calculated by dividing the value transferred by the average monthly private-pay nursing home rate in that state. The penalty does not begin on the date of the gift. It begins when the person is otherwise eligible and already receiving care, which means the penalty lands at the exact moment the household has neither the transferred asset nor coverage.

Two protections soften this, and both came from a single statute. The Medicare Catastrophic Coverage Act of 1988 was repealed almost entirely within eighteen months after a revolt among older voters, but its spousal impoverishment provisions survived. A community spouse, meaning the husband or wife still living at home, may keep a share of the couple's countable resources up to a federal ceiling, and may keep a minimum monthly income allowance out of the institutionalized spouse's income. Without those rules, the standard path to nursing home coverage would leave the spouse at home destitute.

At the end there is estate recovery. Since a provision of the Omnibus Budget Reconciliation Act of 1993, states are required to seek recovery from the estates of deceased beneficiaries aged fifty-five and over for the cost of long-term care services. Recovery is deferred while a surviving spouse or a minor or disabled child is living, and states must have hardship procedures. For a family whose only asset is a house, this is the rule that determines whether anything passes to the next generation.

Bottom line: American long-term care financing is not insurance. It is asset depletion followed by a means-tested program, with a sixty month memory and a claim on the estate.

Why the money goes to the building

Ask a person where they would rather receive help with dressing, and almost all of them say at home. Ask where Medicaid was structurally easiest to spend, and for decades the answer was a nursing facility.

The reason is a distinction in the statute. Nursing facility services are a mandatory Medicaid benefit: a state that runs a Medicaid program must cover them for eligible adults. Most home and community based services are optional. States deliver them largely through waivers authorized under section 1915(c), created by the Omnibus Budget Reconciliation Act of 1981, and a waiver may cap the number of participants. That is the whole of the institutional bias in one sentence. Nursing home care is an entitlement for those who qualify; home care is a program with a fixed number of slots, and people sit on waiting lists for it, sometimes for years.

Litigation moved the line. Lois Curtis and Elaine Wilson were two women with mental illness and developmental disabilities confined in a Georgia state hospital after their own treatment professionals had concluded they could live in the community. In Olmstead v. L.C., decided in 1999, the Supreme Court held that unjustified institutional isolation of people with disabilities is a form of discrimination under Title II of the Americans with Disabilities Act. States must provide community-based services when treatment professionals determine placement is appropriate, the individual does not oppose it, and the placement can be reasonably accommodated given the state's resources and its obligations to others it serves.

That last clause, the reasonable modification standard, is why Olmstead did not empty the institutions. It gave advocates an enormously powerful tool and gave states a defense. What it produced in practice was two decades of consent decrees, state Olmstead plans, and rebalancing efforts, including the Money Follows the Person demonstration created in 2005 to fund transitions out of institutions. The share of Medicaid long-term care dollars going to home and community based services rose substantially over that period. The waiting lists did not disappear.

The network nobody has heard of until they need it

On 14 July 1965, sixteen days before he signed Medicare into law, Lyndon Johnson signed the Older Americans Act. It created what is now the Administration for Community Living, and the 1973 amendments built the structure you will actually call: fifty-six state units on aging and, beneath them, several hundred Area Agencies on Aging, each responsible for a defined planning and service area.

What the Act funds is unglamorous and specific. Title III B pays for supportive services: transportation to dialysis, case management, home repair, adult day care. Title III C pays for nutrition, both congregate meals at senior centers and home-delivered meals. Title III E, added by the 2000 amendments, created the National Family Caregiver Support Program, which funds respite and training for the relatives doing the work. Title V runs the Senior Community Service Employment Program. Title VI funds services through tribal organizations. Title VII houses elder rights protections, including the long-term care ombudsman program, which places an advocate with authority to enter nursing facilities and investigate complaints.

Two design facts about the Older Americans Act are worth holding separately. First, there is no income test. Anyone aged sixty and over is eligible, and the statute instead directs agencies to target services toward those with the greatest economic and social need, with particular attention to low-income minority elders and those in rural areas. Second, it is funded by annual discretionary appropriation, not as an entitlement. The consequence is exactly what you saw with housing vouchers: eligibility is universal and supply is not, so the rationing happens through waiting lists and reduced service levels rather than through a rule that tells anyone they do not qualify.

Who pays for what

Payer	What it actually covers	How it is rationed
Medicare	Post-hospital skilled nursing up to 100 days, skilled home health for the homebound, hospice	By the skilled care definition and the three-day inpatient rule
Medicaid	Nursing facility care, and home and community based services where the state offers them	By means test, then by waiver slots for home care
Older Americans Act	Meals, transportation, case management, respite, ombudsman advocacy	By annual appropriation; no income test, so waiting lists
Private long-term care insurance	A daily or monthly cash or reimbursement benefit after an elimination period	By medical underwriting at purchase and by premium cost
The family	Most hours of most care, unpaid	By whoever is available, usually a daughter or a wife

The insurance market that failed, twice

If long-term care is the risk that bankrupts households, why does no large private market cover it? Insurers tried. In the 1990s and early 2000s more than a hundred companies sold individual long-term care policies. Roughly a dozen still do.

The failure was actuarial rather than ideological. Insurers priced policies on three assumptions that all turned out wrong in the same direction: that a substantial share of buyers would lapse their policies before claiming, that investment returns on reserves would stay high, and that claims would be shorter. Lapse rates came in far below projection, because people who buy this product keep it. Interest rates fell. Claim durations lengthened. Companies raised premiums sharply on policies already in force, which is legal with regulatory approval and which is why the product acquired a reputation for changing its price after you had aged out of alternatives. General Electric took a charge of more than six billion dollars in 2018 to shore up reserves on long-term care business its insurance unit had written years earlier.

Underneath the pricing errors sits adverse selection. Voluntary insurance against a risk people can partially foresee attracts the people most likely to claim, which raises the price, which drives out the healthiest buyers, which raises the price again.

The federal government tried to solve exactly that with the CLASS program, Title VIII of the Affordable Care Act in 2010. It was to be a voluntary, payroll-deduction, publicly run long-term care benefit with no medical underwriting, a five-year vesting period, and a modest daily cash benefit. No underwriting was the point, and no underwriting was the problem: a voluntary program that cannot screen and cannot compel enrollment will be bought disproportionately by people who expect to need it. In October 2011 Secretary Kathleen Sebelius reported that the department could not identify a way to implement the program that would be solvent over the required seventy-five year horizon. CLASS was repealed in January 2013 without ever enrolling anyone.

Washington State took the other route. Its WA Cares Fund, enacted in 2019, is mandatory for most employees rather than voluntary, financed by a payroll premium of 0.58 percent, and pays a lifetime benefit initially set at 36,500 dollars and indexed thereafter. The benefit is small next to years of nursing home care. The design point is that compulsion is the answer to adverse selection, and that a program willing to be compulsory can be much cheaper per person than one that is not.

Worth holding on to: Voluntary long-term care insurance fails for the same reason in the private market and in the CLASS program. The only designs that have worked anywhere are compulsory ones.

The largest long-term care program in the country is unpaid

Tens of millions of Americans provide unpaid care to an adult relative. AARP's Valuing the Invaluable series has estimated the economic value of that labor in the hundreds of billions of dollars a year, far above total public spending on paid long-term services and supports. It does not appear in any budget, so it never appears in any debate about the cost of care.

It appears instead in the caregivers' own outcomes: reduced hours and earnings, interrupted careers, lower retirement savings, and higher rates of depression and physical strain. The burden is distributed unevenly by gender and by income, because a household that can buy help buys it. When a legislature declines to fund home care, it has not reduced the amount of care provided. It has transferred the cost to a daughter.

Common misconceptions

• Medicare covers nursing home care. It covers up to 100 days of post-hospital skilled care per benefit period, in full only for twenty of them, and only after a three-day inpatient stay.
• Time in a hospital bed is a hospital stay. Observation status makes the patient an outpatient, so those nights do not satisfy Medicare's three-day requirement.
• You can protect assets by giving them to your children before applying. The sixty month look-back captures transfers, and the resulting penalty period starts when the person would otherwise be covered.
• Medicaid will take the house from a surviving spouse. Estate recovery is deferred while a surviving spouse or a minor or disabled child is living, and states must offer hardship exceptions.
• Home care is simply cheaper, so states prefer it. Nursing facility services are a mandatory Medicaid benefit while most home and community based services are optional and capped, which is why waiting lists exist for the cheaper option.

The takeaway

• Medicare's skilled nursing benefit runs 100 days at most, is full only for twenty, and requires a three-day inpatient stay that observation status can silently defeat.
• Medicaid is the country's long-term care insurer, reached by spending down to roughly 2,000 dollars in countable assets, with a sixty month look-back and estate recovery at the end.
• Spousal impoverishment protections from the 1988 Medicare Catastrophic Coverage Act are what keep the spouse at home solvent.
• Nursing facility care is mandatory in Medicaid and home and community based services are mostly optional waivers, which is the institutional bias Olmstead attacked in 1999 without eliminating.
• The Older Americans Act funds meals, transportation, respite, and the ombudsman with no income test and no entitlement, so it is rationed by appropriation.
• Voluntary long-term care insurance collapses under adverse selection, which is what killed the CLASS program and what compulsory designs like WA Cares are built to avoid.

Sources

1. Centers for Medicare and Medicaid Services. (n.d.). Skilled nursing facility SNF care. medicare.gov
2. Centers for Medicare and Medicaid Services. (n.d.). Medicaid. medicaid.gov
3. Administration for Community Living. (n.d.). Older Americans Act. acl.gov
4. U.S. Department of Justice, Civil Rights Division. (n.d.). Olmstead: Community integration for everyone. ada.gov
5. Wikipedia contributors. (n.d.). Community Living Assistance Services and Supports Act. en.wikipedia.org

A lottery in Oregon

In early 2008 Oregon had money to add roughly ten thousand adults to its Medicaid program and no fair way to choose them. So the state opened a reservation list for a few weeks, took names, and drew from it at random. About ninety thousand people signed up. Roughly thirty thousand were selected, and of those about ten thousand completed the paperwork and enrolled.

Health economists recognized what Oregon had accidentally built. Every serious question about health insurance runs into the same problem: the insured differ from the uninsured in a hundred ways that also affect health, so a comparison of the two groups measures the difference between the people, not the effect of the coverage. A lottery removes that. Katherine Baicker, Amy Finkelstein, and their colleagues had a randomized controlled trial of health insurance in a rich country, which had not existed since the RAND experiment of the 1970s tested different levels of cost sharing rather than coverage itself.

Their results are the reason this lesson opens here rather than with the signing ceremony in 2010. The Oregon findings are cited by both sides of the American health policy argument, accurately, for opposite conclusions.

What the lottery actually found

Coverage increased use of care across the board: more primary care visits, more prescription drugs, more preventive screening, more hospital admissions. Nobody disputes this and it surprised nobody.

Coverage transformed financial security. Catastrophic out-of-pocket medical spending among the newly insured was reduced almost to nothing. Medical collections fell. The share of households reporting they had to borrow money or skip other bills to pay for care fell sharply. If you think insurance is a financial product, this is a large and unambiguous success.

Coverage improved mental health substantially. The rate of screening positive for depression fell by roughly a third among those who enrolled, an effect large enough that it would be considered a major result for a drug trial.

And then the finding that changed the politics. Two years in, the experiment measured blood pressure, cholesterol, and glycated hemoglobin, the standard markers for cardiovascular risk and diabetes control. It found no statistically significant improvement in any of them. Diagnosis of diabetes went up and use of medication went up; the measured biological markers did not move.

Read that result carefully, because both of the loud readings of it overreach. It does not show that Medicaid does not work; the study was designed with enough statistical power to detect large clinical effects over two years and would have missed small ones, and cardiovascular outcomes accumulate over decades rather than months. It also does not disappear because it is inconvenient; a well-run randomized trial found nulls where advocates had predicted gains, and the honest response is that the clinical case for coverage over a two-year horizon is weaker than the financial and mental health case.

So what?: The best experimental evidence says health insurance is, over a short horizon, a very effective financial and psychological intervention and an unproven clinical one. Most public argument treats it as the reverse.

Three legs, and what happens when you saw one off

Now the law. The Affordable Care Act was signed on 23 March 2010, and its individual market reform is best understood as a single structure with three parts that only work together.

Leg one: guaranteed issue and modified community rating. An insurer must sell to anyone who applies and may not vary the premium by health status or history. It may vary price only by age within a three to one band, by tobacco use, by geography, and by family size. This is the popular part. Polling has found protection for preexisting conditions to be the single most popular provision in the statute.

Leg two: a requirement that healthy people enroll. Guaranteed issue on its own invites a rational person to wait until they are sick to buy. If enough people do that, the risk pool concentrates, premiums rise, more healthy people leave, and premiums rise again.

Leg three: subsidies. If you require purchase, the product has to be affordable. Premium tax credits were set on a sliding scale tied to the second lowest cost silver plan in the area, originally for households between 100 and 400 percent of the federal poverty level. Cost sharing reductions lowered deductibles and copayments for silver plan enrollees up to 250 percent of poverty.

The three-leg claim is not a theoretical prediction. It was tested by states. Washington enacted guaranteed issue in the early 1990s and then dropped the accompanying enrollment requirement; within a few years carriers had stopped writing new individual policies in the state. New York adopted guaranteed issue and community rating in 1993 without an enrollment requirement, and its individual market shrank while its premiums became among the highest in the country. Massachusetts enacted all three legs in 2006 under Governor Mitt Romney, and its uninsured rate fell to the lowest in the nation. The federal law copied Massachusetts.

The federal mandate was then removed anyway. The Tax Cuts and Jobs Act of 2017 set the penalty to zero beginning in 2019, and in California v. Texas in 2021 the Supreme Court disposed of the resulting constitutional challenge on standing without reaching the merits. The market did not collapse, which is a real data point against the strongest version of the three-leg argument. Enrollment held up, in large part because the subsidy leg was doing more work than the penalty leg ever had, and because a separate accident helped: when the federal government stopped reimbursing insurers for cost sharing reductions in 2017, insurers loaded the cost onto silver premiums, which raised the benchmark that premium tax credits are calculated from, which made bronze and gold plans cheaper for subsidized buyers.

What the law did, in four columns

Piece	Mechanism	Who it reaches	Main weakness
Individual market rules	Guaranteed issue, community rating, essential health benefits, no lifetime or annual limits, dependent coverage to age 26	Anyone buying their own insurance	Raises premiums for young and healthy buyers without subsidies
Marketplaces and premium tax credits	Sliding-scale subsidy pegged to the benchmark silver plan	Households above the poverty line without an affordable offer at work	Deductibles remain high in the cheaper metal tiers
Medicaid expansion	Coverage for adults under 138 percent of poverty at a 90 percent federal match	Poor adults without dependent children, previously ineligible in most states	Made optional for states by the 2012 decision
Insurer regulation and delivery	Medical loss ratio floors of 80 and 85 percent, preventive services without cost sharing, closing the Part D coverage gap	Nearly everyone with insurance	Does little about the price of care itself

The decision that created a hole in the middle

On 28 June 2012, in National Federation of Independent Business v. Sebelius, the Supreme Court did two things. It upheld the individual mandate, not as a regulation of interstate commerce but as an exercise of the taxing power, since the penalty was collected by the Internal Revenue Service and was less than the cost of insurance. And it held that Congress could not enforce the Medicaid expansion by threatening to withdraw a state's entire existing Medicaid funding, which the Court treated as coercion rather than persuasion. The expansion survived as an offer states could decline.

The consequence was an outcome nobody had drafted. Premium tax credits begin at 100 percent of the federal poverty level, because the drafters assumed everyone below that line would be covered by the expansion. In a state that declines to expand, an adult earning less than the poverty line is too poor for a marketplace subsidy and, in most such states, ineligible for Medicaid unless she is pregnant, disabled, or caring for a child and very poor indeed. She is in the coverage gap. Someone earning more than she does can get help; she cannot.

Forty states and the District of Columbia had adopted the expansion by 2024, most recently North Carolina. KFF has estimated that roughly one and a half million people remained in the coverage gap in the states that had not, concentrated heavily in the South and disproportionately Black.

Why this matters: The coverage gap is not a loophole and not an accident of drafting. It is what happens when a court converts one mandatory piece of an integrated design into an option.

Did coverage change?

Yes, substantially, and the direction is not seriously disputed. The uninsured rate fell from roughly 16 percent in 2010 to about 9 percent by 2016, and after the subsidy increases in the American Rescue Plan Act of 2021, which removed the 400 percent income cliff and capped premium contributions at 8.5 percent of income, and their extension in the Inflation Reduction Act of 2022, it reached a record low near 8 percent in 2023. Marketplace enrollment passed twenty million.

The people who gained coverage were mostly poor adults in expansion states and moderate-income buyers in the marketplaces. Employer-sponsored insurance, which covers roughly half of Americans, barely moved. The largest federal subsidy in health care is not any of the programs in this lesson: it is the exclusion of employer-paid premiums from taxable income, which is consistently among the largest tax expenditures in the federal budget and which, like the mortgage interest deduction you met in the housing lesson, is worth most to households in the highest tax brackets. Meanwhile the average annual premium for employer family coverage reached roughly 24,000 dollars in 2023 in KFF's employer survey, a cost split between employer and worker that economists generally treat as coming out of wages either way.

Mortality, and the argument about it

Does coverage save lives? This is where you must be able to state each position from its own evidence.

The case that it does. Benjamin Sommers, Katherine Baicker, and Arnold Epstein compared three states that had expanded Medicaid to adults before the ACA with neighboring states that had not, and found a significant reduction in adult mortality. Sarah Miller, Norman Johnson, and Laura Wherry linked survey respondents to death records and found lower mortality among low-income adults in ACA expansion states than in non-expansion states, with effects concentrated among people with conditions amenable to treatment. David Card, Carlos Dobkin, and Nicole Maestas exploited the sharp change in coverage at age sixty-five and found reduced short-term mortality among severely ill patients admitted through emergency departments just after their birthday.

The case for caution. Every study above is quasi-experimental, comparing states or moments rather than randomizing people, and states that expand Medicaid differ from states that do not in ways that also affect mortality. The one randomized study, Oregon, found no significant movement in the clinical markers most closely tied to mortality risk. And the mechanism matters: coverage buys access, access buys treatment, and treatment improves outcomes only for the conditions where treatment works and where the person actually receives it.

What would settle it is a large randomized trial with a decade of follow-up, which will not be run because a state cannot ethically or politically withhold coverage from a control group for ten years. The practical conclusion is that mortality effects are probable, concentrated in treatable conditions, and smaller than advocates claim and larger than critics allow.

Medical bankruptcy, measured twice

You will hear that medical bills cause most personal bankruptcies. That claim comes from work by David Himmelstein, Elizabeth Warren, and colleagues, who surveyed bankruptcy filers and classified a filing as medical when the debtor reported medical bills over a threshold, illness-related income loss, or similar factors, and reported figures above sixty percent.

Carlos Dobkin, Amy Finkelstein, Raymond Kluender, and Matthew Notowidigdo asked a narrower question: not how many filers had medical problems, but how many bankruptcies were caused by them. Using hospital admissions as a shock and comparing admitted adults to their own prior trajectory, they estimated that hospitalizations cause about four percent of personal bankruptcies among non-elderly adults.

Both numbers are defensible and they measure different things. The first counts filers with medical distress present. The second isolates a causal contribution from one type of medical event. If you use either number without saying which question it answers, you are misusing it.

Common misconceptions

• The Oregon experiment proved Medicaid does not improve health. It found large gains in financial protection and depression and no significant movement in three biological markers over two years, which is a narrower claim.
• The Affordable Care Act created government-run insurance. The marketplaces sell private plans; the public coverage in the law is the Medicaid expansion, run by states.
• Everyone below the poverty line is covered. In states that declined the expansion, adults below 100 percent of poverty are usually ineligible for both Medicaid and marketplace subsidies.
• Removing the individual mandate penalty destroyed the marketplaces. Enrollment held up, mainly because the subsidy structure absorbed the shock and silver loading made subsidized coverage cheaper.
• Government spends nothing on employer insurance. The tax exclusion for employer-paid premiums is one of the largest federal tax expenditures and flows mostly to higher earners.

Recap

• Oregon's 2008 lottery produced the only modern randomized evidence on coverage: large financial and mental health gains, higher utilization, and no significant two-year change in blood pressure, cholesterol, or glycated hemoglobin.
• Guaranteed issue, an enrollment requirement, and subsidies are one structure; state experiments in the 1990s showed what happens when the first is enacted without the others.
• NFIB v. Sebelius upheld the mandate as a tax and made the Medicaid expansion optional, creating a coverage gap for adults below the poverty line in non-expansion states.
• The uninsured rate fell from roughly 16 percent in 2010 to near 8 percent in 2023, helped by the 2021 subsidy expansion that removed the 400 percent cliff.
• Quasi-experimental studies find mortality benefits from coverage; the one randomized study found no clinical movement in two years, and both belong in an honest summary.
• Medical bankruptcy estimates above sixty percent and around four percent answer different questions and are not interchangeable.

Sources

1. KFF. (n.d.). Affordable Care Act. kff.org
2. Centers for Medicare and Medicaid Services. (n.d.). HealthCare.gov. healthcare.gov
3. U.S. Census Bureau. (n.d.). Health insurance. census.gov
4. Baicker, K., Taubman, S. L., Allen, H. L., Bernstein, M., Gruber, J. H., Newhouse, J. P., Schneider, E. C., Wright, B. J., Zaslavsky, A. M., and Finkelstein, A. N. (2013). The Oregon experiment: effects of Medicaid on clinical outcomes. New England Journal of Medicine, 368(18), 1713-1722.
5. Wikipedia contributors. (n.d.). National Federation of Independent Business v. Sebelius. en.wikipedia.org

A food budget multiplied by three

In 1963 Mollie Orshansky, an economist working in the Social Security Administration's research office, took the price of the Department of Agriculture's economy food plan, which was the cheapest of four plans and was designed for temporary or emergency use, and multiplied it by three. The multiplier came from a 1955 household survey which found that American families of three or more spent roughly a third of their after-tax income on food. If food is a third of a minimum budget, then a minimum budget is food times three.

She published the resulting thresholds in the Social Security Bulletin in 1963 and 1965 and was careful about what she had made. She described it as a measure of income inadequacy, not of adequacy, and wrote that it drew a line below which nobody could be said to be doing well rather than a line at which people were doing fine. In 1969 the Bureau of the Budget adopted her numbers as the federal government's official statistical definition of poverty. Since then they have been updated for the consumer price index and for essentially nothing else.

Every one of the components has since been overtaken. Food is now closer to a tenth of household spending than a third, so the multiplier of three is wrong. Housing, child care, and out-of-pocket medical costs, which the model did not treat separately, dominate the budgets of poor households. And the resource side of the calculation, what counts as the money a family has, was defined in a country where almost all assistance came as cash.

Remember: The official poverty measure is a 1963 food budget with sixty years of price adjustment bolted on, and its author never claimed it was more than that.

Debugging a claim you have certainly heard

Here is the argument, in the form it usually arrives. The official poverty rate was 19.0 percent in 1964, when Lyndon Johnson declared the War on Poverty. It has run around 11 percent in recent years. The federal government has spent enormous sums in between. Therefore the spending did not work.

The arithmetic is correct. The inference is broken, and it is worth tracing exactly where, because this is the single most consequential measurement error in American social policy.

Step one. The official measure counts only pretax cash income. Write that down and hold it.

Step two. List the major antipoverty programs built since 1964. Food stamps, made national in 1974 and now SNAP. Medicaid, 1965. Housing choice vouchers, 1974. The earned income tax credit, 1975, expanded repeatedly. The child tax credit, 1997, made partially refundable later. Low-income home energy assistance, school meals, WIC.

Step three. Ask which of those the official measure counts as income. None of them. SNAP is not cash. A voucher is not cash. Medicaid is not cash. Refundable tax credits arrive as a tax refund, which the measure does not count either, and the measure does not subtract the payroll taxes a working parent pays.

Step four. Draw the conclusion. Suppose a program lifted every poor family in America to a comfortable standard of living entirely through SNAP, housing vouchers, and refundable credits. The official poverty rate would not move by a single tenth of a point. The measure is structurally incapable of detecting the effect of the policies the country actually built.

That is not a small technical caveat. It means the most frequently cited evidence that the War on Poverty failed is evidence about the definition of a statistic. When Christopher Wimer, Liana Fox, Irwin Garfinkel, Neeraj Kaufman, and Jane Waldfogel constructed a historically consistent measure back to 1967 that did count noncash benefits and taxes, poverty had fallen substantially over the period, and most of the decline came from exactly the programs the official measure ignores.

Thresholds and guidelines are two different things

Before going further, clear up a confusion that appears in student papers and in newspapers constantly.

The Census Bureau publishes poverty thresholds. There are dozens of them, varying by family size, by the number of children, and for one and two person units by whether the householder is aged sixty-five or over. They are statistical instruments, used to count how many people are poor. They are the same in every state.

The Department of Health and Human Services publishes poverty guidelines, a simplified version issued each January, with one figure per family size and separate higher figures for Alaska and Hawaii. Guidelines are administrative instruments, used to set eligibility. When a program says it serves households up to 138 percent or 185 percent of poverty, it means the guidelines.

So the thresholds decide who is counted as poor and the guidelines decide who gets help, and the two are not the same numbers.

What the supplemental measure changed

In 1995 a National Academy of Sciences panel chaired through the volume edited by Constance Citro and Robert Michael proposed a rebuild. The Census Bureau began publishing the resulting supplemental poverty measure in 2011, alongside the official one rather than in place of it.

The threshold is built from what households actually spend on food, clothing, shelter, and utilities. Census takes spending on those four categories at roughly the thirty-third percentile among families with two children, multiplies by a small factor to cover other needs, and then adjusts the result twice: once for housing tenure, since an owner without a mortgage needs less than a renter, and once for geography, using local rents. A threshold in the San Francisco metropolitan area is therefore substantially higher than one in rural Mississippi, which is the point.

The resource side changes even more. Start with cash income, then add the value of SNAP, housing subsidies, school meals, WIC, and energy assistance, and add refundable tax credits. Then subtract federal and state income taxes and payroll taxes, work-related expenses, child care costs, child support paid to another household, and out-of-pocket medical spending. The unit itself is broader: cohabiting partners and unrelated children in the household are counted together, because they share a refrigerator.

Feature	Official measure	Supplemental measure
Threshold basis	1963 economy food plan times three, updated for prices	Recent spending on food, clothing, shelter, and utilities plus a multiplier
Geography	Identical everywhere in the contiguous states	Adjusted for local housing costs and for owner or renter status
Counts SNAP, vouchers, refundable credits	No	Yes
Subtracts taxes, work expenses, child care, medical costs	No	Yes
Family unit	People related by blood, marriage, or adoption	Includes cohabiters and unrelated children in the household
Moves with living standards	No, only with prices	Yes, since the threshold tracks current spending

The predictable consequences follow from the table. People aged sixty-five and over look poorer under the supplemental measure, because it subtracts the out-of-pocket medical spending they actually make, which the official measure ignores. Children usually look less poor under it, because it counts SNAP and refundable credits, which are aimed at families. And poverty looks higher in expensive metropolitan areas and lower in cheap rural ones than the official measure reports.

In short: The two measures do not disagree about the world. They disagree about what counts as money and what counts as a need, and both of those are policy questions wearing statistical clothing.

2021, and the clearest demonstration anyone has produced

For six months in 2021, the American Rescue Plan Act made the child tax credit fully refundable, raised it, and paid half of it out in monthly instalments from July to December. It was, briefly, something close to a child allowance.

The supplemental measure recorded a child poverty rate of 5.2 percent for 2021, the lowest ever measured. When the expansion lapsed and the credit reverted, the supplemental child poverty rate for 2022 was 12.4 percent, more than double the year before.

The official poverty measure did not register the 2021 decline, because a refundable tax credit is not counted in it. One country, one year, two official statistics, and only one of them could see the largest single-year change in child poverty in the history of the series.

The same accounting explains the standard finding that Social Security keeps more people out of poverty than any other program. Social Security arrives as cash, so both measures see it. SNAP and the earned income tax credit each move millions of people across the line, and only the supplemental measure can say so.

The serious objection, and it is serious

You should not leave this lesson believing the supplemental measure is simply correct.

Bruce Meyer and James Sullivan argue that both income-based measures overstate deprivation, and their reasoning has two parts. First, household surveys undercount transfers badly. When researchers link survey responses to administrative program records, they find that a substantial share of SNAP and cash assistance receipt goes unreported, which mechanically inflates measured poverty and, worse, inflates it most among the very poorest households. This bears directly on the deep poverty findings you met in the welfare reform lesson: some of the measured rise is real hardship, and some is missing data. Second, they argue that consumption, meaning what a household actually spends and uses, is a better proxy for material wellbeing than reported income, because households borrow, draw down savings, and receive help that never appears as income. Consumption-based measures show much larger declines in poverty since the 1960s than income-based ones.

The reply is that consumption data have their own defects, that the consumer expenditure survey underreports too, and that a household smoothing consumption by taking on debt or missing a rent payment is not thereby doing well. There is also a values disagreement underneath the technical one. An absolute measure asks whether a household can afford a fixed basket. A relative measure, of the kind the OECD and the European Union use when they set a line at half or sixty percent of median income, asks whether a household can participate in the society it lives in. Neither question is unscientific, and they will not give the same answer in a growing economy.

Why this is a political fight

If the poverty thresholds were only a statistic, nobody would fight about them. They are also the base for the guidelines, and the guidelines gate eligibility for a long list of programs.

Consider the price index. The thresholds rise each year with the consumer price index. A chained index rises more slowly, because it accounts for substitution between goods. Switching the update to a chained index would lower the thresholds a little every year, compounding quietly, and each year slightly fewer households would clear the eligibility line for programs pegged to the guidelines. Nobody would have voted to cut anything. In May 2019 the Office of Management and Budget published a request for comment on exactly this change, and the responses ran heavily against it, precisely because commenters understood the compounding.

The core of it: Choosing an index is choosing a benefit level. That is the general rule, and it is the same lesson as the Thrifty Food Plan reevaluation in the first lesson of this course, arriving from the opposite direction.

Common misconceptions

• The poverty line reflects what it costs to live. It reflects a 1963 emergency food budget multiplied by three and adjusted for prices, with no geographic variation and no accounting for housing, child care, or medical costs.
• The Census thresholds are what programs use for eligibility. Programs use the HHS poverty guidelines, a simplified annual version issued for administrative use.
• Poverty statistics show that antipoverty spending failed. The official measure counts none of the noncash and tax-based programs built since 1964, so it cannot register their effects.
• The supplemental measure always shows more poverty. It shows more among older adults, because of medical costs, and usually less among children, because of SNAP and refundable credits.
• Deep poverty figures are simply accurate. Linked administrative records show significant underreporting of transfers in surveys, concentrated among the poorest households, so part of measured deep poverty is missing data.

Where this leaves us

• Orshansky's thresholds are an economy food plan times three, adopted as official in 1969 and updated only for prices ever since.
• Census thresholds count the poor; HHS guidelines determine eligibility, and confusing them produces wrong answers about who qualifies for what.
• The official measure counts only pretax cash, which makes it blind to SNAP, vouchers, and refundable credits, and therefore blind to most of what the country actually did.
• The supplemental measure builds its threshold from current spending on food, clothing, shelter, and utilities, adjusts for geography and tenure, adds noncash and tax benefits, and subtracts taxes, work expenses, child care, and medical costs.
• Supplemental child poverty fell to 5.2 percent in 2021 under the expanded child tax credit and more than doubled to 12.4 percent in 2022 when it lapsed; the official measure saw neither.
• Meyer and Sullivan's consumption critique and the documented underreporting of transfers are real constraints on how confidently any income-based poverty number can be read.

Sources

1. U.S. Census Bureau. (n.d.). Supplemental Poverty Measure. census.gov
2. U.S. Census Bureau. (n.d.). Poverty. census.gov
3. U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation. (n.d.). Poverty guidelines. aspe.hhs.gov
4. Encyclopaedia Britannica. (n.d.). Poverty. britannica.com
5. Orshansky, M. (1965). Counting the poor: another look at the poverty profile. Social Security Bulletin, 28(1), 3-29.
6. Citro, C. F., and Michael, R. T. (Eds.). (1995). Measuring poverty: a new approach. National Academy Press.

The bill that could not pass on its own

In the early hours of 9 February 2018, after a government shutdown that lasted a few hours, Congress passed the Bipartisan Budget Act. Buried inside a package about spending caps and disaster relief was the Family First Prevention Services Act, the first federal law in forty years to let Title IV-E money pay for services that keep a child out of foster care rather than for the child's maintenance once removed.

Family First had been written years earlier. It had passed the House. It had bipartisan sponsors, endorsements from child welfare organizations, and a policy case almost nobody argued with in public. It could not get through the Senate on its own, because a handful of senators objected to its limits on federal payment for group placements and because several states worried about the cost of converting their systems. Holds are cheap and time is scarce, so it sat.

What finally enacted it was not persuasion. It was a vehicle. Something had to pass that night, and Family First was ready, scored, and small enough to ride.

The upshot: Most American social policy does not become law as the named bill you read about. It becomes law as a title inside something bigger that had to move anyway, which means a proposal's readiness matters as much as its merits.

Authorization is permission, not money

Congress acts twice on most programs and students routinely collapse the two acts into one.

An authorization is a statute that creates a program, says what it may do, and states a level of funding that may be provided. It is written by a policy committee, such as Ways and Means or Health, Education, Labor and Pensions. An authorization by itself buys nothing.

An appropriation is a separate statute, written by the Appropriations Committees, that actually provides money for a fiscal year beginning 1 October. Programs funded this way are discretionary. The Older Americans Act is authorized at one level and appropriated at another, usually lower, which is why the meals program has a waiting list in your county.

Some programs skip the appropriations step entirely. Mandatory or direct spending flows from the authorizing statute itself, to everyone who meets the eligibility rules, without an annual vote. Social Security, Medicare, Medicaid, SNAP, and Title IV-E foster care and adoption assistance work this way. Family First amended Title IV-E, which is why it did not have to win an appropriation afterward to become real.

What matters here: Ask of any program whether it is mandatory or discretionary before you ask anything else about its politics. Mandatory programs are hard to cut and expand automatically in a recession. Discretionary programs are cut by not being increased.

The number that decides: scoring

The Congressional Budget and Impoundment Control Act of 1974 created the Congressional Budget Office and the modern budget process. Before a bill of any size moves, CBO produces a cost estimate against a projected baseline of what would happen under current law. That estimate is not advisory in practice. It determines whether a bill is affordable inside a budget resolution, whether it trips pay-as-you-go rules, and whether wavering members will vote for it.

Scoring shapes the policy itself. Family First's costs were offset in part by delaying a scheduled expansion elsewhere in the same title, which is the ordinary way an expansion of social policy is financed: with another piece of social policy. When you read that a bill was scaled back, the usual cause is not a change of heart. It is a score.

Two chokepoints in the Senate

The first is cloture. Under Senate Rule XXII, ending debate on most legislation takes sixty votes, a threshold lowered from two thirds in 1975. In a chamber where the majority rarely holds sixty seats, this means most social policy needs bipartisan support or another route.

The other route is budget reconciliation, created by the same 1974 act. A reconciliation bill gets limited debate and passes with a simple majority. That is why the amendments finishing the Affordable Care Act moved through reconciliation in March 2010, after the majority lost its sixtieth vote in a Massachusetts special election, and why major tax and spending changes cluster in reconciliation bills.

Reconciliation has a gatekeeper. Senator Robert Byrd secured a rule, now section 313 of the Congressional Budget Act, that strips provisions whose budgetary effects are merely incidental to their policy purpose, along with provisions that change Social Security or that increase deficits beyond the budget window. The Senate parliamentarian advises on whether a provision survives. In February 2021 the parliamentarian advised that a fifteen dollar federal minimum wage could not remain in the American Rescue Plan, and it came out, even though the majority wanted it and it plainly affected the budget. The test is whether the budgetary effect is the point or a side effect.

Two practical consequences follow. Money can move through reconciliation; rules generally cannot. And a policy you care about may be shaped less by what is popular than by whether it can be written as a spending or revenue provision.

A statute is not a rule

A statute rarely tells an agency what to do in enough detail to run a program. Congress delegates, and the agency fills in the detail through rulemaking governed by the Administrative Procedure Act of 1946.

The sequence is fixed. The agency publishes a notice of proposed rulemaking in the Federal Register, with the proposed text and its reasoning. A comment period opens, usually thirty to sixty days, and anyone may file. The agency must consider significant comments and respond to them in the preamble of the final rule, which is published with an effective date. Significant rules also pass through review by the Office of Information and Regulatory Affairs under an executive order issued in 1993.

Comments are the most underused lever available to a social worker, and most comments are wasted. A form letter counts as one voice and adds nothing to the record. What moves an agency, and what matters later in court, is a comment that supplies something the agency does not have: caseload data, a description of how a proposed verification requirement will actually operate at a county office, a documented error rate, an estimate of how many of your clients would lose coverage and why. When the Department of Homeland Security proposed expanding the public charge rule in 2018, more than 260,000 comments were filed. The volume was a political signal. The specific comments describing families disenrolling from programs their children were lawfully entitled to were what built the record.

Congress can also undo a rule. Under the Congressional Review Act of 1996, a joint resolution of disapproval passed within sixty legislative days can void a final rule, and it cannot be filibustered in the Senate.

Courts, and what changed in 2024

For forty years, when a statute was ambiguous, courts generally deferred to an agency's reasonable interpretation of it under Chevron v. Natural Resources Defense Council, decided in 1984. In Loper Bright Enterprises v. Raimondo in 2024, the Supreme Court overruled Chevron and held that courts must exercise independent judgment in deciding what a statute means. Alongside it sits the major questions doctrine, applied in West Virginia v. EPA in 2022, under which an agency claiming power over a question of vast economic or political significance must point to clear congressional authorization.

For social welfare policy the implication is direct. A great deal of what governs your clients is regulation and sub-regulatory guidance rather than statute. That body of policy is now more vulnerable to challenge, in both directions, and the precise words of the authorizing statute matter more than they did. Litigation has become a more central instrument of policy advocacy, not a last resort.

The same goal by the other route

Now the second worked example, which reaches the opposite result.

Section 1115 of the Social Security Act lets the Secretary of Health and Human Services waive certain Medicaid requirements for experimental, pilot, or demonstration projects likely to assist in promoting the objectives of the Act. It is the authority under which states have tested managed care, benefit redesign, and eligibility changes for decades, and it comes with a federal expectation of budget neutrality that is administrative policy rather than statute.

In January 2018 the Centers for Medicare and Medicaid Services issued guidance inviting states to propose work and community engagement requirements as a condition of Medicaid eligibility. Kentucky and Arkansas were approved. Arkansas implemented in June 2018, requiring reporting of qualifying hours through an online portal. More than 18,000 people lost coverage within months, and studies found that most of them were already working or should have been exempt and had been defeated by the reporting requirement rather than by the work requirement.

Litigation followed immediately. A federal district court vacated the approvals, holding that the Secretary had failed to consider whether the demonstrations would promote the central objective of the Medicaid statute, which is furnishing medical assistance. The court of appeals affirmed in 2020. The Supreme Court agreed to hear the case, then removed it from its calendar after the change in administration, and the approvals were withdrawn.

Compare the two routes honestly. The administrative route is fast: guidance in January, approvals within months, implementation the same year. It is also reversible by the next administration and vulnerable in court, and after Loper Bright it is more vulnerable still. The legislative route took Family First the better part of a decade and required a vehicle, and what it produced is a statute that a change of administration cannot undo.

Where a person can actually push

Leverage point	What it takes	What it realistically achieves
Committee testimony and staff briefings	A short written statement and one specific case you can describe accurately	Shapes bill text at markup, where most real drafting happens
Appropriations report language	A relationship with a member's staff and a modest, plausible ask	Directs an agency's attention without changing statute
Federal rule comments	Data or operational detail the agency lacks, filed before the deadline	Changes final rule text and builds the record for later litigation
State waiver and state plan comment	Watching your state Medicaid agency's public notices	Often the highest-value target, because most implementation is state level
Administrative fair hearings	Appealing individual denials rigorously	Reverses individual cases and, in volume, exposes systemic error
Litigation	A legal services or civil rights partner and a clean factual record	Can vacate a rule or a waiver outright, as in the work requirement cases

John Kingdon's account of agenda setting explains the timing you will observe. Three streams run separately: problems that get recognized, policy solutions circulating among specialists, and politics, meaning elections, moods, and who controls what. A window opens when the streams join, usually briefly, and what gets enacted is whatever solution was already drafted and scored when the window opened. Family First is that model in one sentence. The work was done years before the night it passed.

Bottom line: Advocacy is mostly preparation. You cannot open a window, but you can be the person with a finished, costed proposal when one opens.

Common misconceptions

• A bill becomes law by being persuasive. Most social policy is enacted as a title inside a must-pass vehicle, and readiness at the moment a vehicle moves matters enormously.
• Authorizing a program funds it. Discretionary programs need a separate annual appropriation, which is usually below the authorized level.
• Reconciliation lets a majority pass anything. The Byrd rule strips provisions whose budgetary effect is merely incidental, which is why a minimum wage increase was removed from the 2021 relief bill.
• Comments on a proposed rule are just a petition. Volume registers politically, but what changes rules and wins later lawsuits is specific data and operational detail the agency does not already have.
• Courts defer to agencies. Chevron deference was overruled in 2024, so courts now decide statutory meaning independently, and regulatory social policy is correspondingly less stable.

The short version

• Authorization creates a program, appropriation funds it, and mandatory spending skips the second step, which is the first thing to establish about any program.
• CBO scoring against a baseline shapes what a bill contains, and expansions are typically financed by trimming something else in the same policy area.
• Cloture requires sixty votes; reconciliation avoids it with a simple majority but is policed by the Byrd rule and the parliamentarian.
• The Administrative Procedure Act's notice and comment process is where statutory language becomes operational rules, and a comment with data has leverage that a form letter does not.
• Loper Bright ended Chevron deference in 2024, making regulatory policy more contestable in court in both directions.
• Section 1115 waivers move fast and can be undone fast, as Arkansas work requirements were after more than 18,000 people lost coverage; statutes like Family First move slowly and stick.

Sources

1. U.S. Congress. (n.d.). The legislative process. congress.gov
2. Office of the Federal Register. (n.d.). Federal Register. federalregister.gov
3. U.S. General Services Administration. (n.d.). Regulations.gov. regulations.gov
4. Centers for Medicare and Medicaid Services. (n.d.). Section 1115 demonstrations. medicaid.gov
5. Kingdon, J. W. (2011). Agendas, alternatives, and public policies (updated 2nd ed.). Longman.

A road through Fells Point

In the late 1960s Baltimore planned an expressway that would have run east and west through the city, taking parts of Fells Point and Canton and cutting through Black neighborhoods to the west. Barbara Mikulski was a young caseworker with a master's degree in social work from the University of Maryland, employed first by Catholic Charities and then by the city's social services department. Her clients were about to lose their houses.

She organized. What made the campaign work was not a petition drive in one neighborhood but a coalition between the white ethnic neighborhoods in the southeast and the Black neighborhoods in the west, two groups the road would have harmed in different ways and who had not previously worked together. The coalition fought the plan through hearings, historic district designations, and litigation. The expressway through Fells Point was never built. In 1971 Mikulski was elected to the Baltimore City Council. She went to the House in 1976, to the Senate in 1986, and became the longest-serving woman in the history of Congress.

The detail worth keeping is that she did not stop being a social worker in order to do that. She started with a caseload, noticed that the harm her clients faced was arriving through a public decision rather than through private misfortune, and acted at the level where the decision was being made.

Key idea: Policy practice is not a career path away from clients. It is what you do when you notice that the same problem is arriving in your office repeatedly and the cause is a rule.

What the Code actually requires

Students sometimes treat the social and political action language in the NASW Code of Ethics as decoration at the back of the document. Read section 6 again. It sets out obligations to the broader society: to promote the general welfare and the development of people and their communities, to facilitate informed public participation in shaping social policy, to provide professional services in public emergencies, and, in standard 6.04, to engage in social and political action aimed at ensuring people have access to the resources and opportunities they need, with particular attention to vulnerable and oppressed groups and to expanding choice for all people.

The Code's ethical principles put the same thing more bluntly: social workers challenge social injustice, and they pursue social change particularly with and on behalf of oppressed people. That is a statement about the content of the job, not a preference about how to spend a weekend.

Accreditation follows the same line. The Council on Social Work Education's educational standards make engaging in policy practice one of the core competencies every accredited program must teach and assess. That is why this course exists as a requirement rather than an elective.

Four scales, and the one most people skip

When students imagine policy work they imagine Congress. Congress is the least accessible of four scales and usually the least productive place to start.

Federal. Statutes, and the rules that implement them. Slow, national in effect, worth engaging mainly through coalitions and through the rule comment process you learned in the previous lesson.

State. This is where most of American social welfare policy is actually decided, because most federal programs are administered by states with substantial latitude. Your state chooses TANF benefit levels, work requirement definitions, and time limit exemptions. It chooses whether to expand Medicaid, what optional Medicaid benefits to cover, how many home and community based waiver slots to fund, and what its child welfare policy manual says. A state legislature is smaller, its hearings are easier to reach, and a single well-prepared witness can matter in a way that is impossible in Washington.

Agency. Much of what your clients experience is not law at all. It is the intake form, the verification checklist, the appointment scheduling system, the wording of a denial notice, and the unwritten practice of a particular supervisor. Agency policy can often be changed in weeks by someone who documents the problem and proposes a specific alternative.

The individual case. This is policy too, and the first lesson of this course explained why. Michael Lipsky's street-level bureaucrats make policy by aggregation: the discretionary choices of frontline workers, repeated across thousands of cases, become the policy the public actually meets. You will exercise that discretion whether or not you think of it as policy. The only question is whether you exercise it deliberately and consistently.

The point: Start at the scale where you can see the rule, name the person who controls it, and describe the change in one sentence. That is almost never Congress.

Your caseload is a dataset nobody else has

Here is the single most practical technique in this lesson, and it takes about ten minutes a week.

Pick one recurring problem. Keep a tally. Not a narrative, a count: how many applications in the last quarter were denied for a missing document that the state does not actually require, how many clients lost coverage at renewal and how many of those were still eligible, how many families were referred for a service that has a six-month waiting list. Add dates and a one-line description of each.

What you now have is something almost nobody in the policy conversation possesses. Agencies have administrative data, which shows outcomes and not causes. Researchers have surveys, which arrive years later. Legislators have anecdotes. You have a systematic count of a specific failure, collected at the point where the failure happens, with the operational detail that explains it.

That record is what turns testimony from an opinion into evidence. It is what a rule comment needs in order to change a rule. It is what a legal services lawyer needs to decide whether there is a case. And it protects you: a documented pattern is much harder for an administrator to dismiss than a strongly felt impression.

Two cautions. Client confidentiality governs everything you record and everything you disclose, so aggregate and de-identify from the start, and get consent before telling anyone's individual story. And check your own denominator: a count of the cases that reach you tells you about the people who got through the door, not about the people who gave up before it.

The hearing that exists because someone sued

In 1970 the Supreme Court decided Goldberg v. Kelly. New York had been terminating welfare benefits and offering a hearing afterward. The Court held that because those benefits were a matter of statutory entitlement for people who qualified, due process required an evidentiary hearing before termination, with notice, the chance to appear and present evidence, and a decision by an impartial officer.

That decision is why the fair hearing exists in every state and across most benefit programs. It is also the most underused advocacy tool available to a caseworker, because appeals are individually unglamorous and collectively powerful. Appealing rigorously reverses individual denials. Appealing rigorously in volume produces a record of which denial reason is generating reversals, which is direct evidence that a rule or a practice is wrong rather than that applicants are.

What you are legally allowed to do

Two sets of limits apply to most social workers, and both are narrower than the rumors about them.

If you work for a 501(c)(3) nonprofit. Lobbying is permitted; it simply may not be a substantial part of the organization's activities. Because substantial is vague, the tax code offers a bright line: an organization may make the 501(h) election and be measured by an expenditure test instead. Under it, permitted lobbying spending is twenty percent of the first 500,000 dollars of exempt purpose expenditures, then fifteen, ten, and five percent of successive tranches, with an overall ceiling of one million dollars, and grassroots lobbying limited to one quarter of that amount. Several things do not count as lobbying at all, including nonpartisan analysis and research, testimony given in response to a written request from a legislative committee, and communications about matters affecting the organization's own existence or funding. What is absolutely prohibited, with no threshold, is intervening in a political campaign for or against a candidate.

If you work for a government. The Hatch Act restricts partisan political activity by federal employees and extends to some state, county, and local employees whose principal employment is connected to federally financed programs. Most covered employees may register, vote, contribute, and express opinions off duty. They may not use official authority to affect an election, solicit political contributions, engage in political activity while on duty or in a government building, or, for covered employees, run for office in a partisan election. Advocating for a policy is not the same as campaigning for a candidate, and the Act draws the line at the candidate.

Why this matters: The most common reason social workers do not engage in policy is a belief that they are forbidden to. Almost everything described in this lesson, comments, testimony, data, coalition work, and advocacy for a bill, is permitted for both nonprofit and public employees.

Where loyalty splits

The Code also says you have obligations to your employer, to work within agreed commitments and to work toward improving employer policies. Sooner or later those two obligations point in different directions, and no course can resolve that for you in advance.

What is useful is a sequence. Document the pattern before raising it. Raise it internally in writing, with a specific proposed alternative rather than a complaint, because a written proposal creates a record and an oral complaint does not. Escalate one level at a time. Use external structures that are designed to absorb this, meaning professional associations, coalitions, and licensing or ombudsman bodies, rather than making yourself the sole channel. Learn what whistleblower protections apply in your jurisdiction before you need them, not after. And recognize that policy practice inside an agency can carry professional cost. Pretending otherwise would be dishonest, and the reason to say so plainly is that the people who are surprised by it are the ones who get hurt by it.

Aim at the burden

If you want the highest ratio of results to effort, target administrative burden. Pamela Herd and Donald Moynihan's framework names three costs a person pays to obtain a benefit they are entitled to: learning costs, meaning finding out the program exists and whether you qualify; compliance costs, meaning forms, documents, appointments, and travel; and psychological costs, meaning stigma, loss of autonomy, and the stress of being investigated.

These costs are policy choices, and unusually often they can be reduced without new money or a new statute. States can renew Medicaid coverage ex parte, using data they already hold, instead of mailing a form that a household never receives. Twelve-month continuous eligibility for children in Medicaid and the Children's Health Insurance Program became a national requirement in 2024 under the Consolidated Appropriations Act of 2023, which removes an entire annual opportunity for eligible children to fall off. States may use broad-based categorical eligibility to drop the SNAP asset test, so a household is not required to spend down a modest savings account to eat. Applications can be combined across programs. Notices can be rewritten at a readable grade level, which is an agency decision requiring nobody's permission but a supervisor's.

None of these change who is eligible. They change how many eligible people actually receive what the legislature already decided to give them, which is the gap between a policy on paper and a policy in a household, and closing that gap is available to you on a Tuesday.

Common misconceptions

• Policy practice means leaving direct practice. The Code and accreditation standards treat it as part of the job at every level, and the most useful advocacy evidence comes from people carrying caseloads.
• Nonprofit employees cannot lobby. A 501(c)(3) may lobby within limits and may elect a bright-line expenditure test; only campaign intervention for or against a candidate is absolutely barred.
• The Hatch Act forbids public employees from advocating on policy. It restricts partisan campaign activity, not advocacy on legislation or rules.
• Federal legislation is where change happens. Most social welfare policy is set or implemented at state and agency level, where a single prepared witness or a documented pattern can move a decision.
• Take-up problems are the client's failure. Learning, compliance, and psychological costs are designed into programs and can usually be reduced administratively.

Looking back

• Mikulski began as a caseworker who noticed that her clients' problem was a public decision, and organized at the level where that decision was made.
• The NASW Code obliges social workers to engage in social and political action, and policy practice is a required competency in accredited programs.
• Federal, state, agency, and individual case are four scales of policy, and the middle two are usually the most movable.
• A de-identified count of one recurring failure, collected from your own caseload, is evidence that agencies, legislators, and lawyers do not otherwise have.
• Goldberg v. Kelly created the pre-termination fair hearing, and rigorous appeals reverse individual cases while exposing systemic error.
• Lobbying within limits is lawful for nonprofit employees under the 501(h) election, and the Hatch Act restricts partisan campaigning rather than policy advocacy.
• Reducing learning, compliance, and psychological costs raises take-up without changing eligibility, and is the change most often available without new legislation.
• Across this whole course the pattern has repeated: the benefit amount is a formula, the formula is a rule, the rule was chosen, and somebody can choose differently.

Sources

1. National Association of Social Workers. (n.d.). Code of Ethics. socialworkers.org
2. Council on Social Work Education. (n.d.). Educational Policy and Accreditation Standards. cswe.org
3. Internal Revenue Service. (n.d.). Lobbying by section 501(c)(3) organizations. irs.gov
4. Wikipedia contributors. (n.d.). Goldberg v. Kelly. en.wikipedia.org
5. Herd, P., and Moynihan, D. P. (2018). Administrative burden: policymaking by other means. Russell Sage Foundation. russellsage.org