The marriage penalty buried in welfare programs, with the math shown

Suppose a government wanted to discourage marriage among the people who would benefit from it most. It would not need a speech or a campaign. It would need a spreadsheet. Build a benefit that shrinks as household income rises, count a married couple as one household, and wait. The couple that stays unmarried keeps two incomes on two forms. The couple that marries sees one large income on one form, and the check gets smaller. Nobody in Washington decided to punish wedding rings. The formulas did it on their own.

That is the marriage penalty in means-tested welfare, and it is one of the least discussed and most mechanical arguments in the debate over the family. This post does the math in the open. I will define the terms, work one labelled hypothetical couple through one tax year's actual parameters, state the best defence of the system, and then say what conservatives propose to do about it.

What a marriage penalty is, and what it is not

A marriage penalty exists when two people owe more tax or receive less in benefits as a married couple than the same two people would as unmarried individuals. A marriage bonus is the reverse. Both exist throughout the tax code and the benefit system, and the same couple can face a bonus in one program and a penalty in another.

Three clarifications keep the argument honest.

First, the penalty is not a claim about what couples feel. Most people do not marry or stay single because of a phase-out schedule. The claim is narrower: the rules attach a price to a decision that the culture and the evidence say is good for the adults and especially for children, and the price falls hardest on people with the least slack in their budgets.

Second, the penalty is not limited to the poor. The income tax has penalties and bonuses at higher incomes as well. But a tax penalty at high income reduces a surplus. A benefit penalty at low income reduces groceries, medical coverage and rent assistance, which is a different order of magnitude for the household.

Third, the penalty is not a bug in one program. It is a structural feature of means testing, which means tying a benefit to income and reducing it as income rises. Any such program must decide whose income counts. The moment it counts a couple's incomes together and treats them as one unit, the possibility of a penalty exists.

The three mechanisms

Four programs do most of the damage, and they work in three different ways.

One: the Earned Income Tax Credit. The EITC is a refundable credit for low-income workers, larger for those with children. It rises with earnings up to a maximum, plateaus, then phases out as income climbs. For a married couple filing jointly, the credit is computed on combined earnings. Congress did set a higher phase-out threshold for joint filers than for single filers, but the joint threshold is not double the single one. For the tax year used below, it is roughly seven thousand dollars higher, and that gap is the whole source of the problem.

Two: Medicaid and the Children's Health Insurance Program. Eligibility for children is set as a percentage of the federal poverty level for a household of a given size. When two adults marry, household income is combined, and household size changes by only one person if they were already living together with the children. The income test gets harder to pass faster than the poverty threshold rises. States set their own thresholds, which is why the exact breakpoints differ across the country and why this post does not quote one.

Three: the Supplemental Nutrition Assistance Program. SNAP defines a household by who buys and prepares food together, and spouses who live together are always one household. Income tests are applied to the household as a whole, so a married couple's wages are added together before the test. An unmarried couple sharing a kitchen can sometimes qualify as separate households, depending on the rules that apply to them, including whether they have a child in common. This is a place to check state and federal rules, not to assume.

Four: housing assistance and other programs count household income in similar ways. The pattern repeats: combine the incomes, apply the test, reduce the benefit.

Notice what the three mechanisms have in common. In each case the program does what it was designed to do, target aid by household income. The penalty is the side effect.

One worked example, with every step shown

The following is a hypothetical. The couple does not exist. The numbers are chosen to be plausible and the arithmetic is exact. It uses the Earned Income Tax Credit parameters for tax year 2024 for a filer with two qualifying children, as published by the IRS (Revenue Procedure 2023-34).

The 2024 parameters:

• Maximum credit with two qualifying children: $6,960.
• Phase-out rate: 21.06 percent.
• Phase-out begins at $22,720 of income for single and head-of-household filers, and at $29,640 for married couples filing jointly.

Check the parameters against each other, because they have to agree. The credit reaches zero when the reduction equals the maximum: $6,960 divided by 0.2106 is about $33,048 of income above the threshold. For a single filer that puts the end of the credit at $22,720 plus $33,048, which is $55,768. For a joint filer it is $29,640 plus $33,048, which is $62,688. Those match the IRS table. The difference between the two thresholds is $6,920, and that is the extent of the "marriage relief" built into the credit.

Now the family. Picture a mother of two young children, earning $24,000 a year in wages. Call her Dana. She lives with the children's father, whom we will call Marcus. He works and earns $30,000. They are not married. The children live with Dana, she files as head of household, and she claims the children for the EITC.

Unmarried. Dana's earned income is $24,000. That is $1,280 above the $22,720 threshold. The reduction is $1,280 times 0.2106, which is $269.57. Her credit is $6,960 minus $269.57, or about $6,690. Marcus earns $30,000 with no qualifying child of his own for the credit, and at that income he receives no childless credit either. The household's combined credit is about $6,690.

Married, filing jointly. Combined earnings are $24,000 plus $30,000, which is $54,000. That is $24,360 above the joint threshold of $29,640. The reduction is $24,360 times 0.2106, which is $5,130.22. The credit is $6,960 minus $5,130.22, or about $1,830.

The difference: $6,690 minus $1,830 is $4,860 a year. That is roughly $405 a month, lost to the household because the two adults signed a marriage license.

Two caveats belong on the table. The calculation isolates the EITC and leaves out the child tax credit, payroll taxes, state taxes and the difference between head-of-household and joint-filer standard deductions, all of which also change when a couple marries. A complete comparison would net those out, and in some cases the joint standard deduction softens the blow. I am isolating one credit to show one mechanism, not predicting any real family's return. The second caveat is that the example assumes the couple's decision to remain unmarried does not change how they claim the children. The IRS has tie-breaker rules when more than one person could claim a child, and a real couple must follow them.

Now add Medicaid or SNAP to the same household. Dana's $24,000 against a poverty guideline for a household of three is one test. The couple's $54,000 against a household of four is another. The household's income more than doubles, the household's size grows by one, and the program thresholds do not double with it. Benefits that were available to Dana's household can disappear entirely above a cutoff, which makes the effective penalty much sharper than a gradual phase-out.

Why the penalty is steep at exactly the wrong place

Economists call the rate at which a benefit shrinks as income rises an implicit marginal tax rate. The EITC phase-out rate of 21.06 percent in the example sits on top of the income tax and the payroll tax, which is 7.65 percent of wages for the worker's share, and in some states a state tax. Layer SNAP, which phases out at a much higher rate than most people realise, and Medicaid cliffs, and the combined marginal rate on an additional dollar earned can climb steeply for a low-income family.

Marriage is the largest step change in household income a family ever takes. It is a jump rather than a gradual slope, so it is the decision most likely to land a family on the steepest part of the curve. That is why the penalty is a policy problem and not a curiosity.

What the Congressional Research Service says

The Congressional Research Service, the non-partisan research arm of Congress, has written on this subject for decades. Its reports on marriage penalties in the tax code and in means-tested programs describe the same structure in neutral language: programs that count income at the household level and phase benefits out will produce penalties for some couples and bonuses for others, and a policy cannot be fully neutral on marriage if it also targets benefits by income and gives adequate benefits to those with the lowest incomes. Congress.gov carries those reports, and I would encourage you to read them rather than take my word for it. They are the strongest evidence that this is a feature of design that both parties have known about, and not a talking point.

That structure is sometimes described as a trade-off among three goals: adequate benefits for the poor, targeting aid on those who need it, and neutrality toward marriage. A program can usually deliver two of the three. If you want generous benefits and tight targeting, you get marriage penalties. If you want marriage neutrality and tight targeting, you must accept stingier benefits. If you want marriage neutrality and generous benefits, you must spend more and aim at more of the population. The fight is over which of the three to give up.

The best case for the system as it stands

Here is the strongest defence, stated the way its defenders would want it stated.

Means testing exists because resources are finite. A couple earning $54,000 has more resources than a mother earning $24,000, and a program for low-income families should spend its money where need is greatest. If a couple's income is pooled in fact, as married couples' incomes generally are, then pretending that they are two separate low-income people produces a worse error: it gives a subsidy to a household that does not need it, and it takes the money from households that do. Meanwhile, the penalty is not guaranteed. Couples with a very unequal earnings split often get bonuses, because the lower earner's income no longer counts alone. And many unmarried couples who report themselves as unrelated adults are, in fact, partners living together, so some of the apparent penalty is the removal of an advantage they were not meant to have.

There is also a fairness argument from the other direction. A single mother cannot be told to marry. She may not have a partner. She may have a partner who cannot be relied on. The state should not design its aid on the assumption that a husband will appear.

Part of that defence is right, and I will say which part. Targeting does matter, and it is wrong to treat the pooled income of a household as if it belonged to two strangers. Where I part company is on the conclusion. The defence proves that some phase-out is unavoidable. It does not prove that Congress must draw the joint thresholds as narrowly as it does, or that states must write eligibility tests that treat a husband's paycheck as a disqualification. And the argument from fairness to single mothers shows only that aid must remain available to them. Nothing in it requires that marrying be the most expensive thing a low-income couple can do.

What the penalty does to the choice

Do penalties change behaviour? The honest answer is that the evidence is mixed, and I do not think the case needs more than the mixed evidence supports.

Researchers have found modest effects on marriage in some studies and small effects in others. The reason is plausible: most couples weigh love, children, religion, family expectations and trust well above a phase-out schedule. But modest effects at the margin still matter for tens of thousands of families, and the rule has other consequences beyond the decision itself. Couples who are advised, or who work out for themselves, that marriage would cost them thousands of dollars may marry later, may marry informally, or may marry without telling the benefit office. That last category creates a legal risk for the family and a credibility problem for the program. A rule that people evade is a bad rule.

The deeper argument is about what the law signals. Government sends messages through its default settings. A tax and benefit system that is neutral between married and unmarried parents might be acceptable. A system that charges a measurable premium for marrying tells low-income families that the state's priorities lie elsewhere. The nation's own history puts the point sharply: the 1996 welfare reform law listed, among the purposes of its block grant, the encouragement of the formation and maintenance of two-parent families. Congress understood then that welfare design bears on family structure. It follows that welfare design should be judged by that standard.

Where a penalty can be fixed and where it cannot

Conservatives offer several reforms. I will list them, say what each costs and say honestly where each falls short.

1. Widen the joint thresholds. Set the EITC phase-out start for joint filers closer to double the single threshold. This reduces the penalty directly. It costs money, because more households qualify for larger credits, and it creates new bonuses for some couples whose spouses both work.
2. Disregard part of a second earner's income. Some proposals would count only a portion of a spouse's wages in a benefit test. That targets the penalty at the point where it starts, but it complicates administration.
3. Align program definitions of household. Programs use different definitions, and families must track them separately. A single definition would reduce both the penalty and the paperwork. States and federal agencies would have to agree, which is the hard part.
4. Smooth the cliffs. Replace the hard cutoffs in Medicaid and similar programs with gradual phase-outs. This reduces marginal rates on additional earnings and lowers the penalty at the same time. It also extends benefits further up the income scale, which costs more.
5. Give states room to experiment. The federal structure means states can run pilots, design their own work and family incentives and report results. That is the argument from Federalism Lets Fifty States Experiment, and the Constitution Planned It, and it applies here as well as anywhere.
6. Simplify. Part of the problem is that no family can compute its own marginal rate. The case in tax complexity is a subsidy for lobbyists, and the code shows how it works applies to benefits too: complexity hides penalties from the people who pay them and from the voters who might object.

Each reform has a price, and I will not pretend otherwise. The conservative position is not that marriage should be subsidised without limit. It is that the state should stop taxing the thing it says it wants more of, and that any added cost should be set against the long-run cost of family breakdown, which falls on taxpayers through the very programs we are describing.

The measure of the principle

The family is older than any program and older than the republic. The question for a free people is whether the state treats the family as a partner or as an administrative problem. The argument in the family came before the state, so why does the state act like a parent? is that the family is the first institution, and the government's role is to protect it, not to price it. A system that raises the cost of marriage for exactly the households least able to bear it has the order of priorities reversed.

The evidence for why this matters runs through the research on 'The success sequence is just correlation,' answered point by point, which I take up at length in a companion post. Whatever one concludes about causation, the path of finishing school, working and marrying before having children is the path the culture endorses, and no one should be fined for taking it. And the fact that the penalty sits in programs run by churches' secular rivals reminds us of the point in why a church charity knows things a federal program never will: a local charity can tell the difference between a couple that needs help and a couple that has found it, and no formula can.

The case in brief

1. A marriage penalty arises whenever a benefit is tied to household income and a married couple's incomes are combined, and it is built into the structure of means testing, not added by anyone's malice.
2. In the 2024 EITC example, a hypothetical mother of two earning $24,000 who marries a partner earning $30,000 sees her credit fall from about $6,690 to about $1,830, a loss of about $4,860 a year, before other programs are counted.
3. The best defence of the system is that targeting is necessary and that pooled income is real, and that defence justifies some phase-out but not narrow joint thresholds or cliffs that penalise the decision to marry.
4. Conservative reforms include wider joint thresholds, partial income disregards, common household definitions, smoother phase-outs and state experiments, each with a cost that should be weighed against the cost of weaker families.
5. A government that says it values the family should stop charging the poorest couples for forming one.

What you can do: look up the current EITC parameters on the IRS website, work the same arithmetic for a family you know in the abstract, and write one sentence to your representative asking where the joint-filer threshold sits in the next tax bill, using the representatives page to find the office.