1863 Leadership · Issue Paper No. 9
The Conservative Case for Welfare Reform
The highest marginal tax rates in America are paid by people earning thirty thousand dollars a year. Four repairs, and a single schedule.
Abstract
A household climbing out of poverty in America faces dozens of separate programs, each with its own application, its own eligibility rule, and its own point at which the benefit disappears. Stack those withdrawals on top of payroll and income tax and a family in the phase-out range can face an effective marginal rate at or above 100 percent — poorer for having earned more. No billionaire in American history has faced a rate approaching it. This paper proposes replacing the categorical structure with a single graduated schedule that runs negative at the bottom, crosses zero somewhere in the middle, and rises from there: one curve, one taper, every benefit phasing out on the same slope, with a hard rule that no household's effective marginal rate may exceed the top statutory rate. It proposes paying it in cash without restriction, and housing the unhoused on the Finnish model — without preconditions and without force. Retirement is treated separately, in Issue Paper No. 11. We set out the arithmetic that has defeated every previous attempt, because it has not gone away.
Key findings
- The idea is not new and it nearly became law. President Nixon's Family Assistance Plan was a negative income tax with a 50 percent phase-out. The House passed it 243 to 155. It died in the Senate Finance Committee on a 10–6 vote, killed by conservatives of both parties.1
- America ran the experiment. Four randomized trials across 8,500 families between 1968 and 1982 measured the work response directly — the strongest evidence base any American antipoverty proposal has ever had. Results arrived too late to save the plan.2
- Cash does not go where critics claim. A World Bank review of 19 studies found no systematic increase in spending on alcohol or tobacco from cash transfers — often the reverse.3
- Finland abandoned the treatment-first model and rehoused its long-term homeless without preconditions. Long-term homelessness fell 68 percent between 2008 and 2022, and roughly 80 percent of the homeless population accepted housing.4
- Helsinki went from 2,121 shelter and hostel beds in 1985 to 52 in 2016, converting the shelters into apartments.4
Section 1The question before us
A woman earning thirty thousand dollars a year is offered more hours. She works them, earns another two thousand dollars, and ends the year with less money than she started with. Her food assistance fell, her housing subsidy fell, her child care subsidy fell, her health coverage moved a step closer to disappearing, and payroll tax took its share of the raise on the way past.
Nobody designed that. It is what happens when forty programs are written separately, each with a sensible phase-out of its own, and nobody adds them up.
The system taxes the transition out of poverty harder than it taxes anything else, and then we are surprised the transition does not happen.
A word on where this sits. Issue Paper No. 8 concerns what the tax code collects. This paper concerns what it pays out, and how those payments are withdrawn as earnings rise. The two share a mechanism — a single graduated schedule — and they are separate arguments, because a reader may accept either without the other.
This paper is not an argument about how generous the safety net should be. That is a real disagreement and this paper does not settle it. It is an argument that whatever we decide to spend should be delivered on one schedule rather than forty, because the interaction of the forty produces a result nobody would vote for if it were written down as a single sentence.
1.2 Our own share of the failure
People who argue for work incentives have spent decades attacking the size of the safety net and almost no time on its shape.
If the objection to welfare is that it discourages work, then the phase-out schedule is the whole ballgame — and the phase-out schedule is where nobody looked. We fought over eligibility, over time limits, over work requirements, and over the total dollars, while the marginal rate on the working poor climbed past anything we would tolerate on ourselves for a single afternoon.
There is a second thing to own. This author has argued for years that people respond to incentives. That argument cuts both ways. If a man declines the extra shift because the extra shift makes him poorer, he is not lazy. He is doing arithmetic, correctly, and the arithmetic was written by us.
Section 2How the trap is built
No single program is the culprit. Each withdrawal is individually defensible and the damage is entirely in the stacking.
Food assistance tapers. Housing assistance takes roughly thirty cents of every additional dollar. Child care subsidies phase out, often at a cliff rather than a slope. Health coverage has eligibility thresholds where a dollar of income costs thousands of dollars of benefit outright. The earned income credit phases out. The child credit phases out. Payroll tax applies from the first dollar. State programs layer their own schedules on top, and the schedules were not written to be added together.
Add them and you get bands of income across which a household keeps almost nothing from additional work, and narrower bands where it keeps less than nothing. A cliff is the extreme case: one dollar of earnings triggering the total loss of a benefit worth thousands.
We want to be careful here, because the size of this effect is contested and varies enormously by state, by family composition, and by which programs a household actually receives. What is not contested is the mechanism, or that the highest effective marginal rates in the American economy occur somewhere between roughly fifteen and forty thousand dollars of earnings.5
The administrative side compounds it. Each program has its own application, its own documentation, its own recertification, its own caseworker, and its own definition of income and household. A family qualifying for six programs applies six times and reports changes six times. The cost of that is borne partly by the taxpayer and mostly by the applicant, in hours.
Section 3Reform I: One schedule, running negative at the bottom
Replace the categorical cash and near-cash programs with a single graduated schedule that pays out at low incomes, crosses zero, and rises from there.
A tax schedule need not begin at zero. It can run negative — the government pays you — cross zero at some income, and rise through the brackets above. One continuous curve, administered by one agency, with no caseworker deciding who deserves what.
3.1 The lineage, which is not what people assume
Milton Friedman proposed the negative income tax in Capitalism and Freedom in 1962, as a replacement for the patchwork of categorical programs. Hayek supported a guaranteed minimum. So did Martin Luther King, who wrote in 1967 that the solution to poverty is to abolish it directly through a guaranteed income.6 Few proposals in American life have drawn Friedman and King to the same page.
And a Republican president proposed it. Nixon's Family Assistance Plan, announced in August 1969, was a negative income tax with a 50 percent phase-out rate. The House of Representatives passed it 243 to 155. It died in the Senate Finance Committee on a 10 to 6 vote, with conservative Democrats joining conservative Republicans to kill it. A revised version cleared the Senate 68 to 5 in 1972 authorizing only a test, and the provision was stripped in conference before Nixon signed the bill.1
This is the second time in this series we have found a structural reform that passed a chamber of Congress and died procedurally. It is worth noticing how often the American difficulty is not that a good idea was never had.
3.2 And America tested it
Between 1968 and 1982 the federal government ran four randomized controlled trials of a negative income tax — in New Jersey, in rural Iowa and North Carolina, in Gary, and in Seattle and Denver — covering roughly 8,500 families. No other American antipoverty proposal has ever been tested this rigorously.2
Work effort fell, modestly. The Stanford Research Institute's analysis of the Seattle-Denver experiment found roughly a 9 percent reduction in hours for husbands and 18 percent for wives, with other estimates putting primary earners at 5 to 7 percent. The reductions concentrated among secondary earners and teenagers — wives with young children, and adolescents staying in school longer.2
The experiments also reported an increase in marital dissolution, which was politically devastating and has since been challenged as a statistical artifact. We report it because a reader deserves to know what the opposition will cite.2
The results arrived too late and too poorly understood to help the plan they were designed to inform.
Section 4Reform II: Every benefit on the same curve
This is the provision that distinguishes the proposal from a simple cash program, and it is the one we would defend hardest.
Assistance that survives as a voucher rather than cash — housing, and the health vouchers proposed in Issue Paper No. 10 — must phase out on the same schedule as the cash payment, not on a means test of its own.
A benefit that phases out is a tax on the next dollar earned, whatever the agency that administers it calls itself.
Consider what a separate means test does. Comprehensive health coverage is worth somewhere between eight and twenty-five thousand dollars a year. Phase that out across any band of income and the implicit marginal rate inside the band is enormous — worse than every cliff described in Section 2, and created by the very reform meant to eliminate them. A voucher that vanishes at sixty thousand dollars makes the dollar at fifty-nine thousand nine hundred and ninety-nine the most expensive dollar earned in America.
So the rule is arithmetic rather than sentiment. All withdrawals, summed, plus all taxes, must never exceed the top statutory rate for any household at any income. If the top rate the country is willing to impose on its highest earners is thirty-seven percent, that is the ceiling for the household at thirty thousand dollars as well.
4.1 The arithmetic that has defeated everyone
We are obliged to state the constraint plainly rather than discover it after publication.
A negative income tax has two dials: the guarantee at zero income, and the rate at which it phases out. Break-even income is the guarantee divided by the taper. A fifteen thousand dollar guarantee tapering at fifty percent breaks even at thirty thousand. Taper at thirty percent instead — much better work incentives — and break-even moves to fifty thousand, sweeping in a far larger share of the country at a far larger cost. Cut the guarantee to eight thousand and it is affordable and it is not a safety net.
Adequate benefit, low marginal rate, affordable cost: any two. That is not a drafting problem and better language will not solve it. It is why no country has enacted a full negative income tax at scale, and any paper that pretends otherwise is selling something.
Our position is that the current system already fails all three — it is inadequate at the bottom, imposes the highest marginal rates in the economy on the working poor, and costs a great deal — and that choosing two deliberately is better than failing three by accident.
Section 5Reform III: Pay it in cash, without restriction
Payments should be made in money, usable anywhere, with no restriction on what may be purchased.
The obvious political answer is a restricted card — food, utilities, rent, clothing — because the attack writes itself otherwise. We considered it and rejected it for two reasons.
The first is evidence. A World Bank review of nineteen studies found no systematic increase in spending on alcohol or tobacco following cash transfers, and in many cases a decrease.3 The premise of the restriction is empirically weak.
The second is consistency. A restricted card requires merchant category codes, eligible-item lists, exception processes and enforcement — which is precisely the administrative apparatus this reform exists to eliminate, rebuilt inside the replacement. It is the current system with one card instead of six.
And there is a plainer point. A person who can be trusted to choose a doctor can be trusted to buy groceries. A framework that treats citizens as competent shoppers in one paragraph and as incompetent ones in the next is not a framework.
Section 6Reform IV: House people, without preconditions
Homelessness is where this framework reaches its limit, and we would rather say so than pretend a schedule solves it.
A curve delivers money to people who can receive it. It does nothing for a man with untreated schizophrenia, an addiction, or a criminal record that makes landlords refuse him. Vouchers and markets do not reach that population, and a paper claiming otherwise has not looked at it.
Homelessness is a pox on the free market. It is not possible to defend this economic system while people sleep behind dumpsters, and we should stop trying.
6.1 Finland stopped requiring people to earn housing
Until the 1990s Finland ran what it called the staircase model: housing was granted after treatment for mental illness or addiction. Finnish policymakers concluded that requiring people to be intoxicant-free and in control of their lives before receiving a home was an insuperable barrier for exactly the people who most needed one.4
They inverted it. Housing first, without preconditions, with services offered and not required. If a tenant does not wish to give up intoxicants, he is not forced to. Housing is the prerequisite that allows other problems to be solved, rather than the reward for having solved them.
They converted the shelters. Helsinki had 2,121 shelter and hostel beds in 1985 and 52 by 2016, while supported housing units rose from 127 to 1,309 and independent apartments for formerly homeless people from 65 to 2,433. Long-term homelessness fell 68 percent between 2008 and 2022, with roughly 80 percent of the homeless population housed through the programme.4
6.2 The provision Americans usually leave out
The Finnish model is more congregate than the American one — buildings with staff on site rather than scattered apartments — and that is the piece that answers the objection people raise most.
Some people genuinely cannot sustain an independent tenancy. The American answer has been to conclude that they must therefore remain outside. The Finnish answer is a building with a door that locks and somebody in the lobby. It is not an institution, nobody is committed, and nobody is required to be sober to stay.
We propose no encampments, sanctioned or otherwise, and no compulsion. A modelling study in a JAMA journal found that encampment sweeps produced more deaths, less treatment engagement and higher costs than leaving encampments alone, while housing without a treatment requirement produced the fewest deaths and the most treatment engagement at the highest cost.9 The ranking is clear and it does not have a tent in the middle of it.
6.3 And the part that is not a program at all
Houston cut chronic homelessness 68 percent using this approach. Utah and California adopted the same approach and saw chronic homelessness grow 95 and 93 percent.10
The difference is not the program. It is that Houston has abundant cheap housing and no traditional zoning code. An analysis by the Cato Institute — no friend of this approach — found that a state's relative land-use freedom explains roughly 38 percent of the variation in homelessness across states, and housing affordability another 34 percent.10
That is an uncomfortable finding for anyone who wants this to be a spending question, and it should be the first line of any serious proposal: the binding constraint is frequently the local zoning code, and no amount of federal money builds an apartment that a city has made illegal.
We want to be precise about what this does and does not say, because the sloppy version is a slander on markets. Housing supply is entirely capable of meeting demand — Houston demonstrates it. Where supply fails to respond, the impediment is zoning, permitting, minimum lot sizes and parking mandates. That is not a market failure requiring a program. It is a government failure requiring a repeal, and it happens to be the cheapest item in this paper.
Section 7 · The strongest case against
7.1 People will work less
The experiments found it and we are not going to pretend otherwise: roughly 9 percent fewer hours for husbands and 18 percent for wives in the largest trial.2 A reader may reasonably regard a secondary earner leaving the workforce as a cost even if it concentrated among mothers of young children.
Our answer is comparative rather than absolute. The current system imposes marginal rates at or above 100 percent on the same households. Whatever work reduction a fifty percent taper produces, it is smaller than what a hundred percent taper produces, and the honest comparison is against the system we have rather than against a system with no disincentives at all, which does not exist.
7.2 The arithmetic may simply not close
Section 4.1 states the triangle and we cannot escape it. A guarantee adequate to live on, with a taper gentle enough to preserve work incentives, costs more than the programs it replaces — considerably more. A reader who concludes the numbers do not work has identified the reason this has never been enacted, not a flaw in our presentation.
7.3 Cash is not the right instrument for everything
Some needs are lumpy, time-inconsistent, and catastrophic to get wrong. Losing housing is far harder to recover from than losing a month's groceries. A person in active addiction may make choices with cash that a voucher would have prevented.
The evidence on temptation goods runs our way in the aggregate, and aggregates conceal individuals. We accept that some people will be worse off under an unrestricted payment than under a restricted one, and we think the alternative — treating every recipient as incapable because some are — costs more than it saves.
7.4 Finland is small, homogeneous, and differently housed
Finland has a fraction of the American population, a much larger social housing stock, and universal health coverage sitting underneath the housing programme. Only about 19 percent of Finnish housing is private rental, which makes the congregate model easier to build and harder to replicate.4
The Houston comparison suggests the approach can travel. The Utah and California comparison suggests it does not travel by itself.
7.5 A single schedule means a single point of failure
Forty programs are inefficient and they are also redundant. If one agency fails, the others continue. A unified schedule administered by one agency concentrates the risk — of a systems failure, of a political decision to cut the whole curve at once, of a payment infrastructure that cannot reach people without bank accounts or filed returns. The federal government struggled badly with exactly this problem in 2020 and 2021.
7.6 What we concede, and what we do not
We concede that work effort falls. We concede that the arithmetic may not close at any politically acceptable cost. We concede that unrestricted cash will be misused by some recipients. We concede that Finland's success rests partly on conditions America does not have, and that zoning may matter more than any program we propose.
We do not concede that the present arrangement deserves the benefit of the doubt. A system that imposes its highest marginal rates on people earning thirty thousand dollars, requires six applications for six benefits, and leaves several hundred thousand Americans sleeping outside is not a cautious default. It is a set of accidents nobody chose, defended mainly because it is what we happen to have.
Section 8What we are not claiming
We are not claiming this saves money. Administrative consolidation is worth something, but American transfer programs are administratively cheap already — single-digit percentages of benefits for most, well under one percent for Social Security. Anyone promising that eliminating agencies pays for the reform is misleading you. The case here is about incidence and incentives, not overhead.
We are not claiming this ends poverty. It ends one specific, self-inflicted feature of American poverty: the fact that escaping it is taxed more heavily than anything else a person can do.
We are not claiming the level is right. Where the guarantee should be set is a genuine political question on which this organization takes no position. What we claim is that whatever level is chosen should be delivered on one curve.
And we are not claiming this is easy. A Republican president proposed a version of it, the House passed it, four randomized trials tested it, and it still failed. That should be sobering to anyone who thinks the argument is obvious.
Section 9The argument you can carry
The paper compressed to what a person can remember and repeat.
And the one that is ours rather than theirs. People who argue for work incentives spent decades attacking the size of the safety net and almost no time on its shape — while the marginal rate on the working poor climbed past anything we would tolerate on ourselves for a single afternoon.
Section 10Conclusion
There is a version of this argument about generosity, and this paper is not it. Americans will go on disagreeing about how much a wealthy country owes its poorest members, and that disagreement is legitimate and old and will outlast everyone reading this.
What should not be in dispute is the shape. Whatever we decide to spend, it should arrive on one schedule, phase out on one slope, and never take more from a household's next dollar than we would take from anyone else's. That is not a position about the size of government. It is a position about arithmetic, and the arithmetic is currently indefensible.
The rest follows from the same principle. Pay it in money, because a citizen competent to choose a doctor is competent to buy groceries. And house people without requiring them to be well first, because Finland tried the other way for forty years, measured it, and stopped.
A country that had never built a safety net would not build this one. We have the one we have because it accumulated, program by program, each addition sensible on its own. The accumulation is the problem, and no individual program is going to fix it.
Notes
- Family Assistance Plan, announced August 1969; H.R. 16311 passed the House of Representatives 243–155 in 1970 and was rejected by the Senate Finance Committee 10–6. See David R. Henderson, "The Failed Welfare Revolution," Regulation, Cato Institute (Spring 2018), cato.org, reviewing Brian Steensland's history; and the legislative record of H.R. 1 in the 92nd Congress. Cato and Steensland reach opposite conclusions about whether the defeat was fortunate; we cite the facts rather than either verdict.
- The four income maintenance experiments — New Jersey, Rural (Iowa/North Carolina), Gary, and Seattle-Denver — conducted between 1968 and 1982. Institute for Research on Poverty, University of Wisconsin, irp.wisc.edu; overview of the final SIME/DIME report at aspe.hhs.gov. The Stanford Research Institute estimates of 9 percent and 18 percent are reported in "Negative Income Tax," Concise Encyclopedia of Economics, econlib.org. On the marital dissolution finding and its subsequent challenge, see Robert Moffitt's review literature; the finding should be characterized as contested rather than established.
- David K. Evans and Anna Popova, "Cash Transfers and Temptation Goods," World Bank Policy Research (review of 19 studies). Most of the underlying studies are of developing-country programs; the applicability to American recipients is an inference rather than a direct finding, and readers should weigh it accordingly.
- On Finland's abandonment of the staircase model and adoption of Housing First: HUD PD&R, huduser.gov, and Marybeth Shinn and Jill Khadduri in Cityscape vol. 22 no. 2, huduser.gov, the source for the Helsinki bed and unit counts and for the observation that only 19 percent of Finnish housing stock is private rental. The 68 percent decline in long-term homelessness between 2008 and 2022 and the 80 percent housing rate are reported by Pathfinders, sdg16.plus. On respect of choice and non-compulsion, see the Finland chapter in Successful Public Policy in the Nordic Countries (Oxford, 2022).
- Effective marginal tax rates on low-income households vary substantially by state, family composition, and program participation, and the literature reports a wide range. The Congressional Budget Office and the Urban Institute have both published estimates. This paper deliberately does not cite a single figure, because the honest answer is a distribution rather than a number; any published version should present the range with its sources and state the assumptions.
- Milton Friedman, Capitalism and Freedom (University of Chicago Press, 1962); Martin Luther King Jr., Where Do We Go from Here: Chaos or Community? (1967). Confirm the King quotation against the original text before publication.
- Modelling study comparing encampment abatement strategies for people with opioid use disorder, finding that sweeps produced more deaths, fewer person-weeks in treatment, and higher costs than the status quo, while housing without a medication requirement produced the fewest deaths at the highest cost. ncbi.nlm.nih.gov. This is a simulation rather than a trial and should be described as such.
- Vanessa Brown Calder, "Housing Markets First: Housing Supply and Affordability Are Key to Reducing Homelessness," Cato Institute, cato.org. Source for the Houston, Utah and California comparisons and for the land-use and affordability variance figures. Cato is critical of the Housing First approach this paper endorses; we cite it against our own position and readers should note that its authors would not draw our conclusion from their data.
A note on the author
Issue papers are published under the name of 1863 Leadership rather than an individual byline. Where this paper speaks in the first person, the author is its founder, who served in the United States Marine Corps as a cryptologic Arabic linguist and spent fourteen years building and operating a multi-unit restaurant enterprise, employing several hundred people at wages inside the phase-out ranges this paper describes.
A note on sources
Two of the load-bearing citations here run against our own position and we have used them anyway: the Cato analysis in note 10, whose authors are critical of the housing approach we endorse, and the work-reduction findings in note 2, which are the strongest evidence against the reform in Section 3. Note 5 deliberately declines to cite a single figure for effective marginal rates because the honest answer is a distribution. Notes 3, 6, 8 and 9 flag verification still required before publication.
Recommended citation
1863 Leadership. "The Conservative Case for Welfare Reform." Issue Paper No. 9.
September 2026. 1863leadership.org
Corrections: None to date. Errors of fact are corrected on this page within one business day of notice, with a dated note describing the change.