The earned income tax credit is the largest cash antipoverty program for working families in the United States, and its entire design lives inside a single provision of the federal tax code. Section 32 of the Internal Revenue Code, codified at 26 U.S.C. section 32, converts low earnings into a cash payment by applying a fixed percentage to the first dollars of earned income, holding the resulting credit flat across a plateau range of earnings, and then withdrawing it gradually as earnings rise. That three-part shape is the whole policy. The phase-in rewards the decision to work, the plateau delivers the maximum benefit to earners in the middle of the low-wage range, and the phase-out removes the benefit without imposing a cliff. Congress created the credit as a temporary measure in the Tax Reduction Act of 1975 and made it permanent three years later, then expanded it in 1986, 1990, 1993, 2001, and 2009. Understanding how this one section works means understanding how the federal government can run a cash transfer at welfare scale without a single benefits office, because every eligibility rule, every rate, and every limit is written as tax law.

This article applies a single test to itself: after reading it, a reader should be able to explain how the largest cash antipoverty program for working families in the United States operates entirely through the tax code, describe the phase-in, plateau, and phase-out structure and what each stage does to the incentive to work, name the political bargain that created the credit, and understand why the provision carries both the strongest employment evidence and the highest improper payment rate of any major federal program.

The Statutory Identity of the Earned Income Tax Credit

Every serious statement about the earned income tax credit begins with its address. The credit is section 32 of the Internal Revenue Code, cited as 26 U.S.C. section 32. Title 26 of the United States Code is the Internal Revenue Code itself, the codification of the federal tax statutes, so the citation places the credit inside the income tax rather than inside any welfare title. That placement is substantive rather than decorative. It determines which agency administers the program, which forms carry it, which definitions govern its terms, and which enforcement tools apply when claims go wrong. The operative rules occupy the subsections of the section, running from section 32(a) through section 32(i), and each subsection does distinct work: the benefit formula, the definitions of earned income and qualifying child, the special rules for filers without qualifying children, the bar on separate filers, and the investment income limit. When later Congresses expanded or tightened the program, they amended this section rather than building new machinery alongside it, so the full legislative history of the credit is the amendment history of one code section.

The enactments that created and changed the credit are cited by public law number, and the numbering convention deserves a word because it recurs throughout this article. A public law number names the Congress and the sequence of enactment: Public Law 94-12 was the twelfth public law enacted by the Ninety-fourth Congress. Statute names such as the Tax Reduction Act of 1975 are the short titles Congress assigns for convenience, but the public law number is the durable identifier, and this article uses both together at each enactment so the reader can trace the provision through the statute books.

The credit entered the law as an experiment. The Tax Reduction Act of 1975, Public Law 94-12, signed March 29, 1975, created a temporary refundable credit that was in effect for 1975 only. Congress designed it as a one-year provision, which meant the credit would vanish from the code unless a later Congress acted to keep it. The temporary form served two purposes at once. It let the credit’s proponents describe it as a modest stimulus measure folded into a tax reduction bill during a recession year, and it let skeptics consent without committing the Treasury to a permanent new outlay. The design also revealed what its authors thought they were testing: whether a work-conditioned payment delivered through the tax system would reach low-earning parents without reproducing the work disincentives that its opponents attributed to traditional welfare.

Congress kept the experiment alive by extending it on a temporary basis, repeatedly, which created the peculiar status the credit held in its first years. Each extension required fresh legislative action, and each one renewed the debate over whether the measure deserved permanence. The extensions also gave the provision a constituency: once low-earning households had received the payment for several filing seasons, repeal would have meant taking cash away from working parents, a political cost that grew with every year the credit survived. The temporary extensions thus functioned as a ratchet, making the credit harder to kill each time Congress renewed it.

Permanence arrived with the Revenue Act of 1978, Public Law 95-600. That statute made the credit a permanent feature of the Internal Revenue Code, removing the sunset that had forced the repeated extensions. The interval from 1975 to 1978 marks the credit’s probationary period, and the 1978 act closed it. From that point forward the credit no longer needed reauthorization to survive; changing it required affirmative amendment, which shifted the legislative burden from supporters to opponents. The provision’s subsequent history is a series of such affirmative amendments, each one enlarging the credit, and the permanence of 1978 is what made that one-directional growth possible.

The first great expansion came with the Tax Reform Act of 1986, Public Law 99-514. That statute raised the credit rate to 14 percent and permanently indexed the credit’s thresholds for inflation. Both changes mattered beyond their arithmetic. Raising the rate increased the reward for each dollar of earnings in the phase-in range, which sharpened the work incentive at the bottom of the earnings distribution. Indexing the thresholds for inflation protected the credit’s real value without requiring Congress to act: before indexing, inflation quietly shrank the credit’s reach every year by pushing nominal earnings past fixed thresholds, so permanence of the provision had not meant permanence of its generosity. The 1986 act made the credit self-maintaining in real terms, and that structural change did as much for the program’s long-run scale as the rate increase.

The Omnibus Budget Reconciliation Act of 1990, Public Law 101-508, split the formula into one-child and two-or-more-child schedules. Until 1990 the credit had used a single schedule, which meant a parent with three children received the same maximum as a parent with one. The 1990 act recognized that the costs of raising children rise with the number of children and that a flat benefit undercompensated larger households. By creating separate schedules, Congress allowed the phase-in rate, the maximum, and the phase-out range to vary with family size, and that differentiation became the template for every later redesign of the benefit structure. The one-child and two-or-more-child distinction introduced in 1990 survives in the modern schedules.

The Omnibus Budget Reconciliation Act of 1993, Public Law 103-66, substantially expanded the credit and made workers without qualifying children eligible for the first time. The 1993 expansion was the largest single enlargement of the program in its history, raising rates and maximums across the schedules and, crucially, extending a smaller credit to childless workers. Before 1993 the credit was strictly a benefit for parents; the 1993 act acknowledged that low-earning workers without children also faced payroll tax burdens the credit was designed to offset. The childless schedule carried lower rates and tighter eligibility rules than the family schedules, reflecting a compromise between extending the work incentive and limiting the fiscal cost, and those special rules remain part of section 32.

Congress expanded the credit again in 2001, through Public Law 107-16, and in 2009, through Public Law 111-5. Each of these enactments layered new generosity onto the existing schedules without disturbing the architecture that the 1975, 1978, 1986, 1990, and 1993 acts had built. The pattern of the provision’s history is therefore cumulative rather than episodic: one section of the code, amended seven times across three and a half decades, each amendment adding to what came before. That cumulative structure is why the statutory identity of the credit can be stated so compactly. The entire program, with all of its schedules, tests, and limits, is section 32 of the Internal Revenue Code, as amended by Public Law 94-12, Public Law 95-600, Public Law 99-514, Public Law 101-508, Public Law 103-66, Public Law 107-16, and Public Law 111-5.

The citation practice this article follows reflects a distinction the reader will need throughout. The United States Code states the law in its codified form, arranged by subject, while the public law number identifies the historical act that changed it. Section 32 of the Internal Revenue Code is therefore the living provision, and Public Law 94-12 is the 1975 event that brought it into being. The Statutes at Large, the chronological record of every enacted law, sit behind both citations as the authoritative source, but the working lawyer’s habit, and this article’s, is to cite the code for what the law is and the public law for when and how it became so. Keeping the two citation forms distinct prevents the common confusion between the provision’s current content and the particular Congress’s contribution to it.

The temporary form of the 1975 credit deserves attention as legislative technology rather than as mere caution. A temporary tax provision carries a sunset: the law states its own expiration, and after that date the provision ceases to exist unless Congress affirmatively extends it. Sunsets serve the enacting Congress in several ways at once. They reduce the provision’s apparent budget cost, because official estimates count only the years the provision is in effect. They force a future Congress to revisit the policy with evidence in hand, which gives skeptics a scheduled opportunity to kill it and supporters a scheduled opportunity to defend it. And they let legislators vote for an experiment without voting for a commitment, which is often the only way a novel benefit can assemble a majority. The one-year design of the 1975 credit used all three advantages: the credit could be scored as a single year’s outlay, its supporters promised a trial, and its skeptics accepted a test.

The vehicle mattered as much as the design. The Tax Reduction Act of 1975 was a recession-year stimulus bill, a broad package of tax cuts meant to lift a contracting economy, and the credit entered the law as one piece of that package rather than as a freestanding welfare bill. Placement inside a tax reduction bill gave the credit procedural advantages it could not have obtained on its own: it traveled under the Finance Committee’s jurisdiction, it benefited from the momentum of must-pass legislation, and it arrived in the statute books already framed as tax relief. The framing was not incidental. A benefit enacted as part of a tax cut is administered by the Internal Revenue Service, claimed on the tax return, and debated in the language of tax policy, and each of those facts shaped the credit’s later development.

Indexing, introduced in 1986, operates as quiet maintenance on the provision’s real value. Fixed dollar thresholds lose purchasing power every year that prices rise, because nominal earnings grow with inflation while the thresholds stand still. For a benefit aimed at low-earning households, that erosion is especially corrosive: inflation can push a worker’s nominal wages past a fixed phase-out threshold even when the worker’s real purchasing power has not changed, shrinking the credit for someone the statute meant to help. Permanent indexing answers the problem by adjusting the thresholds each year for inflation, so the credit’s real generosity persists without requiring Congress to legislate. The 1986 act’s indexing provision thus did something the rate increase alone could not: it made the credit’s scale self-preserving across decades of price change.

The schedule differentiation of 1990 and the childless extension of 1993 together completed the transformation of Long’s work bonus into a family policy. The single schedule of the early years treated all working parents alike; the split schedules recognized that the economic burden of low earnings falls harder on larger households and calibrated the benefit accordingly. The 1993 decision to cover workers without qualifying children closed the remaining gap in the payroll-tax-offset logic: if the credit exists to return payroll taxes to low-earning workers, there was no principled reason to exclude low-earning workers merely because they had no children. The childless schedule’s smaller rates and tighter rules reflected the fiscal compromise that made the extension possible, but the principle of the extension was the original one, applied universally at last.

Permanence changed the credit’s political economy in a way the temporary extensions never could. A temporary provision invites its opponents to wait it out; a permanent provision forces them to assemble a majority for repeal, which is a far heavier lift once households have organized their budgets around the payment. Permanence also changed behavior among the credit’s administrators and claimants: the Internal Revenue Service could invest in return-processing routines for the credit, tax preparers could build it into their standard workflows, and low-earning households could count on it when making work decisions. The 1978 act therefore did more than remove a sunset; it converted the credit from an experiment into an institution, and institutions develop constituencies that experiments never acquire.

The amendment pattern visible across the seven enactments carries the series thesis of this article in miniature. Each Congress that touched the credit amended the same section rather than creating parallel programs, so the provision’s history is legible as the growth rings of a single statute. A reader who wants the real policy does not consult committee reports about the credit’s purposes or agency guidance about its administration first; the reader opens section 32 and reads the text, because the text is where every bargain, every expansion, and every limit was written. The operative text is the location of the real policy, and the statutory identity section of this article has established the address.

The Political Bargain Behind the Credit

The earned income tax credit was conceived as the alternative to a guaranteed income. On August 8, 1969, President Nixon unveiled the Family Assistance Plan, a proposal to replace much of the existing welfare system with a federally guaranteed minimum income for families. The plan would have paid benefits to poor households whether or not anyone in the household worked, and Nixon presented it as a rationalization of a welfare system he considered fragmented and perverse. The proposal drew support from policy intellectuals who favored a clean income floor, but it collided in the Senate with a chairman who found its central feature morally and economically backwards.

That chairman was Russell Long of Louisiana, who led the Senate Finance Committee through the entire Family Assistance Plan debate. Long opposed unconditional payments on the ground that the plan gave its largest benefits to those without earnings and would discourage work. His objection was not a quibble about benefit levels or administrative design; it was a rejection of the principle that the federal government should pay households that did not work. Long argued that any federal income support for the able-bodied should reward employment rather than replace it, and he used his committee’s jurisdiction over tax and welfare legislation to block the guaranteed-income approach for as long as he held the gavel.

Long countered with a work bonus plan built on the opposite principle. Where the Family Assistance Plan paid whether or not the recipient worked, the work bonus would pay only on earnings, so that, in Long’s formulation, the more he works the more he gets. The design was deliberately tied to the payroll tax: the bonus was meant to offset social security taxes for low-earning parents, returning to workers a portion of the payroll taxes withheld from their wages. That framing gave the proposal two political advantages. It presented the payment as a tax refund rather than a welfare grant, which made it palatable to legislators who would never vote for expanded welfare, and it guaranteed that only workers could receive it, which answered the work-incentive objection at the level of the mechanism itself.

The work bonus passed the Senate in 1972, 1973, and 1974. It never became law under that name, because the House and the conference process would not accept it as a freestanding measure, but its repeated Senate passage kept the concept alive and refined its design through successive iterations. Each passage clarified the coalition behind the idea: liberals who wanted cash assistance for poor working parents, and conservatives who would support such assistance only if it were conditioned on work. The work bonus was the legislative expression of that coalition’s terms, and its three Senate victories demonstrated that the terms could command a majority even when the broader guaranteed-income proposal could not.

In 1975 the work bonus was renamed the earned income tax credit and enacted as part of the Tax Reduction Act of 1975. The renaming was more than cosmetic. Calling the measure a tax credit rather than a bonus relocated it conceptually from the welfare system to the tax system, which is where section 32 has lived ever since. Enactment inside a tax reduction bill reinforced the framing: the credit arrived as tax relief for workers rather than as a new spending program, and that framing shaped both its administration through the Internal Revenue Service and its political durability in the decades that followed.

The bargain that produced the credit therefore had a precise shape. Supporters of cash assistance for the poor got a federal payment that reached low-earning parents at meaningful scale; opponents of unconditional welfare got a payment that only workers could receive and that grew with earnings. Work conditionality was the price of the coalition, and the phase-in is its mechanical expression: the credit subsidizes the first dollars of earnings because Russell Long insisted that federal money should reward work rather than substitute for it. Every feature of section 32 that ties the benefit to earnings descends from that insistence.

The Family Assistance Plan debate produced a second statute alongside the credit, and the pairing illuminates what the debate was really about. The means-tested Supplemental Security Income program, enacted as Title XVI through the Social Security Amendments of 1972, Public Law 92-603, enacted October 30, 1972, with payments beginning in January 1974, emerged from the same failed proposal. Where the work bonus became the earnings-conditioned answer for working-age parents, Supplemental Security Income became the federally guaranteed answer for the aged, blind, and disabled, the populations for whom Congress accepted that a work requirement made no sense. The two statutes are products of the same history: the Family Assistance Plan died, and its carcass was divided between a tax credit for workers and a cash program for those not expected to work. Understanding the credit’s political identity requires holding both halves of that division in view, because the credit’s work condition only makes sense as the counterpart to a program that has none.

Nixon’s Family Assistance Plan needs a fuller statement, because the credit is best understood as the surviving answer to the question the plan posed. The plan proposed to replace the existing welfare structure, then dominated by Aid to Families with Dependent Children, with a federal guaranteed income: a minimum payment to poor families, varying with family size, available whether or not anyone in the household worked. Nixon’s stated rationale was administrative and moral at once. Administratively, the existing system was a patchwork of state-run programs with different benefit levels and eligibility rules, producing arbitrary differences between similarly situated families. Morally, Nixon argued that a system which paid only certain categories of the poor while excluding the working poor was indefensible. The guaranteed income would rationalize the patchwork and extend the floor to everyone below it, including households with earnings.

Long’s objection fastened on the word everyone. A benefit available without regard to work, he argued, would direct its largest payments to households with no earnings at all, precisely the households whose labor the economy most needed. The critique had an economic component and a moral one, and Long pressed both. Economically, he contended that paying households not to work would reduce work effort among the very population whose employment the nation sought to encourage. Morally, he held that federal support for able-bodied adults should reward the effort to support oneself rather than substitute for it. The two components reinforced each other: the economic argument gave the moral argument its urgency, and the moral argument gave the economic argument its constituency. Together they made Long immovable, and as chairman of the Finance Committee he held jurisdiction over both tax legislation and welfare legislation, which meant no guaranteed-income proposal could reach the Senate floor without his assent.

The work bonus was Long’s constructive alternative, and its design repays attention because every element of it survives in section 32. The bonus would pay a percentage of earnings, so the payment grew with work rather than shrinking with it. It was calibrated to offset payroll taxes, the social security taxes withheld from wages, which fall most heavily in proportional terms on low-earning workers because they apply from the first dollar of earnings. Framing the payment as a payroll tax offset did political work beyond its arithmetic: it presented the bonus as a refund of taxes the worker had already paid rather than as a grant the government had chosen to give. A refund carries none of welfare’s stigma in American political language, and Long understood that the difference between a refund and a grant could be the difference between enactment and defeat.

The bonus’s three Senate passages without enactment illustrate a basic fact of the legislative process that shaped the credit’s final form. Passage by one chamber is a position, not a law; enactment requires the concurrence of the other chamber and the President’s signature, and the House never accepted the work bonus as a freestanding measure. The Senate victories of 1972, 1973, and 1974 therefore functioned as rehearsals: they refined the design, demonstrated the coalition’s durability, and kept the concept available for the moment a suitable vehicle appeared. The Tax Reduction Act of 1975 was that vehicle, and the bonus’s incorporation into it shows how major policies often enter the law sideways, as provisions of larger bills rather than as the bills’ announced purposes.

Renaming the measure the earned income tax credit completed its institutional migration. A work bonus sounds like a spending program administered by a welfare agency; an earned income tax credit sounds like a feature of the tax system administered by the Internal Revenue Service. The rename determined the credit’s delivery mechanism, because tax credits are claimed on returns, and it determined the credit’s budgetary character, because the cost appears as reduced revenue and outlay through the tax system rather than as an appropriated program. Those institutional facts then shaped the politics: a benefit delivered through the tax code reaches its recipients without a welfare office, which spared the credit the administrative controversies that dogged traditional welfare, and it created a constituency of tax preparers and low-income filers whose interests aligned with the credit’s survival.

The Supplemental Security Income parallel deserves a final emphasis because it fixes the meaning of the bargain. Title XVI federalized cash assistance for the aged, blind, and disabled, guaranteeing a federal income floor to the populations Congress deemed unable to support themselves through work. The credit guaranteed no floor to anyone; it supplemented the earnings of those who worked. The two statutes divide the population that Nixon’s plan had treated as one: those expected to work receive a work-conditioned supplement through the tax code, and those not expected to work receive a guaranteed benefit through the Social Security Act. The division is the Family Assistance Plan debate’s lasting settlement, and the credit’s side of it explains why the provision looks the way it does. Every work test in section 32 is there because Russell Long won the argument that federal money for working-age adults must reward work, and every dollar of the credit is the price his opponents paid for his vote.

How Section 32 of the Code Operates

The operative text of section 32 decides who receives the credit by defining three things: what counts as earned income, who counts as a qualifying child, and which filers are excluded regardless of their earnings. The definitions are written with the precision of tax law because they carry the entire targeting burden of the program. There is no caseworker discretion, no state variation, and no separate eligibility interview; the return itself, processed against the definitions in the section, determines the payment. That administrative leanness is the credit’s great strength, and the complexity of the definitions is its price.

Earned income, defined in section 32(c), means wages, tips, and net self-employment earnings. The definition deliberately counts only income from labor and excludes every other kind of income at the threshold: interest, dividends, rents, royalties, and capital gains do not count as earned income no matter how large they are. The restriction serves the work-bonus logic directly. Because only labor income triggers the credit, a household cannot qualify through investment returns or passive receipts; the payment follows work and nothing else. Net self-employment earnings are included so that the self-employed, from day laborers to small contractors, are treated like wage earners, with the netting rule ensuring that the credit rewards actual earnings rather than gross receipts.

A qualifying child must pass four tests, and each test does distinct work in the targeting scheme. The relationship test limits qualifying children to the claimant’s own children, stepchildren, foster children, siblings, and their descendants, which keeps the credit within genuine family relationships. The age test requires the child to be young enough that the credit functions as support for raising children rather than as a general household subsidy. The residency test requires the child to have lived with the claimant for more than half the year, which ties the benefit to the household actually bearing the costs of raising the child and prevents two households from claiming the same child. The joint return test provides that a married child who files a joint return generally cannot be a qualifying child, which prevents the credit from flowing to children who have formed their own tax households. Together the four tests convert the abstract idea of a dependent child into an administrable rule, and their specificity explains both the credit’s targeting accuracy and its vulnerability to error: every test is a line that a filer must draw correctly on the return.

Section 32(i) bars the credit when disqualified investment income exceeds $10,000 as indexed. The barred income includes interest, dividends, rents, royalties, capital gains, and passive income, which is to say the provision looks past the earned income definition to the filer’s overall means. A household might have qualifying earnings and a qualifying child yet still be disqualified if its investment income crosses the indexed threshold. The limit expresses a judgment that the credit is for households whose resources are genuinely limited: earnings alone do not establish need when substantial capital income sits alongside them. Indexing the $10,000 threshold preserves the line in real terms so that inflation does not quietly disqualify filers the provision was meant to include.

Section 32(d) bars married-filing-separately filers from the credit entirely. A married couple that files separately is ineligible no matter how low the earnings or how many qualifying children are involved. The bar exists because separate filing can be used to manipulate the appearance of household income: a high-earning couple could file separately to make one spouse look like a low earner and claim the credit on that spouse’s return. By requiring joint filing from married couples, the section forces the credit to be measured against the household’s combined earnings, which is the economically relevant figure for a program meant to supplement low household resources.

Workers without qualifying children face their own set of boundary rules, added when the 1993 act first made them eligible. A childless claimant must be at least 25 but under 65, must not be a dependent of another taxpayer, and must have lived in the United States for more than half the year. The age floor excludes younger workers who are commonly claimed as dependents, which prevents the credit from subsidizing households that already receive support through the dependency system. The age ceiling reflects a judgment about the working-life population the provision targets. The dependent bar prevents double counting within families, and the residency requirement ensures the credit reaches workers with a genuine connection to the United States. These rules are narrower than the family schedules because the childless credit is smaller and Congress drew its boundaries more tightly.

Read together, the definitions form a single machine. Earned income determines whether the work condition is met, the four tests determine whether the family condition is met, the investment income limit determines whether the means condition is met, the filing-status bar determines whether the household is measured honestly, and the childless-worker rules determine whether the extended 1993 eligibility applies. Every one of these determinations happens on the tax return, under penalties of perjury, without human review in the ordinary case. That is what it means for the operative text to be the location of the real policy: the statute does not delegate the program’s substance to regulations or to administrators, because the substance is the text.

The decision to run the program through definitions rather than through discretion has consequences that run deeper than administrative convenience. A traditional welfare program employs caseworkers who interview applicants, verify documents, and exercise judgment at the margins; section 32 employs none of these, because the return is the application and the statute is the caseworker. That design eliminates an entire layer of bureaucracy and with it the delays, the regional variations, and the dignitary costs of the welfare office. It also concentrates every policy choice in the statutory language, which means the definitions must be precise enough to decide hard cases without human intervention. The precision is the program’s genius and its burden in equal measure.

The earned income definition illustrates the burden. Wages and tips present few difficulties, because employers report them and the figures on the return can be matched against information returns. Net self-employment earnings are harder, because the self-employed report their own income and expenses, and the netting of expenses against gross receipts requires judgments about what counts as a business expense. The statute’s inclusion of the self-employed was a deliberate choice to avoid penalizing workers whose labor takes non-wage forms, from construction day labor to childcare, but the choice opened the credit to the measurement problems that accompany all self-reported business income. The boundary between genuine self-employment and mischaracterized income is one of the provision’s persistent administrative pressure points, and it exists because the definition must cover real economic diversity with a single phrase.

The residency test’s more-than-half-the-year rule shows how a bright line does the work of a caseworker. Determining which household actually bears a child’s costs could in principle require investigation, but the statute substitutes a countable test: the child must have lived with the claimant for more than half the year. The rule is overinclusive in some cases and underinclusive in others, as all bright lines are, but it is administrable on a tax return, which is the only virtue that matters for a provision processed by machine. The joint return test performs a similar simplifying function for a different problem. A married child who files jointly has formed a separate tax household, and the statute treats that formation as conclusive evidence that the child is no longer the claimant’s dependent for credit purposes. The general disqualification avoids the need to inquire into the actual economics of the young couple’s household.

The investment income limit adds a means test to what is otherwise an earnings test, and the combination reveals the statute’s conception of need. Earnings alone do not establish that a household lacks resources, because a household can have modest wages alongside substantial capital. The $10,000 indexed threshold, applied to interest, dividends, rents, royalties, capital gains, and passive income, draws a second boundary around the eligible population: not only must the household work for its income, but its non-work income must also be limited. Indexing the threshold preserves the boundary’s real meaning across years, so inflation cannot silently move households across it. The limit is the provision’s answer to the objection that a pure earnings test would subsidize households that are asset-rich but income-poor by choice.

The married-filing-separately bar and the childless-worker rules complete the anti-manipulation architecture. The separate-filing bar addresses a straightforward gaming strategy: without it, one spouse in a high-earning household could file separately, report only personal earnings, and claim a credit designed for low household resources. The bar forces married couples to be measured on their combined earnings, which aligns the credit’s unit of measurement with the household’s actual economic position. The childless rules, with their age band of 25 to 65, their dependent exclusion, and their residency requirement, perform the same function for the 1993 extension that the four tests perform for the family schedules: they draw a bright, return-administrable boundary around the intended population and exclude the cases, such as students claimed as dependents, where the credit’s rationale does not apply.

Taken together, the definitions constitute a complete targeting system written in a single section of the code. Each rule answers a specific evasion or misallocation risk, each is stated as a test applicable on the return, and none requires administrative judgment to apply. The system’s completeness is what allows the credit to operate at national scale through the tax pipeline, and its rigidity is what generates the errors that show up in the improper payment statistics. A machine that decides everything by rule will misdecide every case the rules describe imperfectly, and section 32’s rules, precise as they are, describe the complexity of American household arrangements imperfectly. The operative text is the location of the real policy, and the real policy includes the mistakes its own precision makes possible.

The Shape of the Credit: Phase-In, Plateau, and Phase-Out

The credit’s benefit formula has three stages, and the stages do opposite things to the incentive to work. In the phase-in range, the credit equals a fixed percentage of earned income, so each added dollar of earnings increases the credit by the phase-in rate. In the plateau range, the credit holds at its maximum regardless of earnings, so added dollars of earnings neither increase nor decrease it. In the phase-out range, the credit falls by a fixed percentage of each added dollar of earnings, so each added dollar of earnings shrinks the credit by the phase-out rate. The formula thus subsidizes the first dollars of earnings, ignores the middle dollars, and taxes the later dollars, and that negative-then-positive pattern is the single claim from which the credit’s entire empirical record follows.

The negative-then-positive rate deserves careful statement because it is easy to misread. Over the phase-in range the effective marginal tax rate on earnings is negative: earning another dollar does not merely add a dollar of wages but adds the dollar plus the credit increment, so the tax system pays the worker to earn more. Over the plateau the effective marginal rate is zero: the credit is flat, so the tax system neither rewards nor penalizes the next dollar of earnings through the credit. Over the phase-out range the effective marginal rate is positive: each added dollar of earnings reduces the credit, which functions exactly like a tax on those earnings. The credit therefore pulls people into work powerfully at the bottom of the earnings distribution while discouraging additional hours at the top of the credit’s range, and both effects are mechanical consequences of the same formula.

The rates differ by schedule, and the differences encode Congress’s judgments about family size. The phase-in rates are 7.65 percent for workers without qualifying children, 34 percent with one qualifying child, 40 percent with two, and 45 percent with three or more. The maximum scales with the number of qualifying children, and the thresholds are indexed for inflation annually, so the benefit grows with family size and holds its real value without further legislation. The phase-out rates are 7.65 percent for the childless, 15.98 percent for one child, and 21.06 percent for two or more. The phase-in rates are steep because their job is to make work pay at the bottom; the phase-out rates are gentler because their job is to withdraw the benefit without creating the cliffs that would punish earners for crossing a threshold.

Why does the credit stop growing once earnings reach the plateau?

The credit stops growing because Section 32 fixes a maximum amount for each schedule, and once earnings reach the plateau threshold the formula holds the credit flat instead of multiplying additional earnings by the phase-in rate. The plateau keeps the maximum benefit within a defined earnings band before phase-out begins.

The plateau exists to concentrate the credit’s full generosity on the earnings band where low-wage full-time work actually falls. If the credit kept growing with earnings indefinitely, the benefit would flow increasingly to households that needed it less, and the phase-out would have to start from a higher maximum, which would either extend the phase-out deep into the middle class or require a punitively steep withdrawal rate. By capping the credit at a maximum and holding it flat across a plateau, the statute delivers the largest payment to earners in the lower-middle of the wage distribution and begins the withdrawal from a fixed, knowable amount. The plateau is therefore a targeting device disguised as a flat line: it defines the population that receives the full benefit and sets the starting point for the phase-out’s arithmetic.

The plateau also serves the work incentive in a subtler way. During the phase-in, the negative effective marginal rate rewards each additional hour; on the plateau, the zero rate neither rewards nor punishes it. That neutrality matters because the workers on the plateau are typically working substantial hours already, and the statute’s judgment is that their labor supply decision has been made: the credit’s job at that point is to supplement their income, not to manipulate their hours. The phase-out then reintroduces a positive marginal rate, but only after the credit has delivered its maximum, so the disincentive at the top is the price of the incentive at the bottom. The three stages are a single bargain, and the plateau is the hinge between its two halves.

The credit structure table below gathers the four schedules, their phase-in and phase-out rates, the plateau’s behavior, the effective marginal tax rate at each stage, and the resulting incentive effect on additional hours of work.

Number of qualifying children	Phase-in rate	Plateau behavior	Phase-out rate	Effective marginal tax rate	What each stage does to the incentive to work more hours
0	7.65 percent	Credit holds at its maximum across the plateau range of earnings	7.65 percent	Negative during phase-in, zero on the plateau, positive during phase-out	Phase-in raises the reward for added hours, the plateau leaves the reward unchanged, and phase-out lowers the net reward for added hours
1	34 percent	Credit holds at its maximum across the plateau range of earnings	15.98 percent	Negative during phase-in, zero on the plateau, positive during phase-out	Phase-in raises the reward for added hours, the plateau leaves the reward unchanged, and phase-out lowers the net reward for added hours
2	40 percent	Credit holds at its maximum across the plateau range of earnings	21.06 percent	Negative during phase-in, zero on the plateau, positive during phase-out	Phase-in raises the reward for added hours, the plateau leaves the reward unchanged, and phase-out lowers the net reward for added hours
3 or more	45 percent	Credit holds at its maximum across the plateau range of earnings	21.06 percent	Negative during phase-in, zero on the plateau, positive during phase-out	Phase-in raises the reward for added hours, the plateau leaves the reward unchanged, and phase-out lowers the net reward for added hours

The table makes the negative-then-positive claim visible at a glance. Every schedule begins by paying workers to earn more and ends by taxing additional earnings, and the steepness of each effect rises with the number of qualifying children. The childless schedule is the limiting case: its 7.65 percent rates in both directions make it the smallest subsidy and the gentlest withdrawal, which is why its employment effects are the most modest in the empirical record. The family schedules, with phase-in rates from 34 to 45 percent, deliver the powerful employment pull that the literature associates with the credit, and their phase-out rates of 15.98 to 21.06 percent produce the hours disincentive at the top of the range. The empirical findings all follow from that single shape: strong effects on whether people work, weaker and more contested effects on how much they work, because the formula is designed to move the participation decision and merely tolerates its consequences for the hours decision.

The negative marginal rate over the phase-in range can be stated in the concrete terms of an hour of work. When a worker in the phase-in range earns an additional dollar of wages, two payments respond: the employer pays the dollar, and the Treasury increases the credit by the phase-in rate. The worker’s total gain from the dollar of earnings is therefore the dollar plus the credit increment, which means the tax system is adding to the wage rather than subtracting from it. For a parent of two facing a 40 percent phase-in rate, each added dollar of earnings yields the dollar and an additional forty cents of credit, a combined return that no employer could match and that transforms the economics of taking a job. That transformation is the point of the phase-in: it makes the first hours of work the best-paid hours, reversing the usual pattern in which low-wage work is the worst-paid.

The steepness of the phase-in rates across the schedules reflects a deliberate calibration to family size. The childless rate of 7.65 percent offers a modest supplement, consistent with the schedule’s origin as a payroll tax offset for workers without the costs of raising children. The family rates rise from 34 percent for one child to 40 percent for two and 45 percent for three or more, which means the reward for each added dollar of earnings grows with the household’s needs. The progression embodies the judgment, first legislated in 1990 and deepened thereafter, that the work incentive should be strongest where the financial pressure to work is greatest. A parent of three receives nearly half again as much credit per phase-in dollar as a parent of one, and that difference is policy, not arithmetic accident.

The plateau’s flat line performs quieter but equally deliberate work. By holding the credit at its maximum across a defined earnings band, the statute identifies the population it most wants to help: workers whose earnings reflect substantial labor force attachment at low wages, the full-time and near-full-time low-wage workers whose paychecks cover the basics but leave no margin. The plateau delivers the full benefit to exactly those workers, neither increasing it for higher earners nor reducing it for those at the band’s lower edge. In doing so it also fixes the starting point of the phase-out, which matters because the phase-out’s arithmetic depends entirely on the maximum from which the withdrawal begins. A higher maximum would require either a longer phase-out reaching further up the income scale or a steeper withdrawal rate; the plateau’s fixed maximum lets Congress choose the phase-out’s length and gentleness with full knowledge of the cost.

The phase-out rates embody the gentleness side of that choice. At 15.98 percent for one child and 21.06 percent for two or more, the withdrawal takes back only a fraction of each added dollar, which means workers in the phase-out range keep most of the gain from additional earnings. The design avoids the cliffs that characterize many means-tested programs, where crossing an income threshold by a single dollar can cost a household its entire benefit. Cliffs punish work at the threshold with an effective marginal rate exceeding 100 percent; the credit’s gradual phase-out keeps the effective rate well below that level, preserving the reward for additional work even as the benefit shrinks. The trade-off is that a gentle phase-out extends over a longer earnings range, which brings more households into the credit’s reach and raises its cost. Congress accepted that trade-off in every expansion, preferring a longer, gentler withdrawal to a shorter, steeper one.

The distinction between the participation decision and the hours decision organizes the credit’s incentive effects into a coherent story. The participation decision is whether to work at all; the hours decision, for those already working, is how much to work. The phase-in’s negative marginal rate operates almost entirely on participation, because it makes the first dollar of earnings the most heavily subsidized and therefore pulls nonworkers into the labor force. The phase-out’s positive marginal rate operates on hours, because it reduces the net reward for additional earnings among workers already past the plateau. The plateau itself is neutral on both margins. This asymmetry explains the shape of the empirical record: the credit’s clearest effects appear in employment rates, where the phase-in does its work, while its effects on hours among existing workers are smaller and harder to isolate, because the phase-out’s gentle rates produce only mild disincentives spread across a wide range.

The childless schedule illustrates the same logic at reduced intensity. Its symmetric 7.65 percent rates make it the smallest subsidy in the phase-in and the gentlest withdrawal in the phase-out, which follows from its narrower purpose: offsetting payroll tax burdens for low-earning workers without children rather than supporting the costs of raising a family. The schedule’s modest scale means its participation effects are correspondingly modest, and its tight eligibility rules confine it to the population Congress meant to reach. The family schedules, by contrast, carry the credit’s full ambition, and their rates show it: phase-in rates high enough to transform the return to work, phase-out rates gentle enough to avoid punishing the workers the credit has drawn into employment. The negative-then-positive shape is not a compromise between competing goals but a sequence, first pulling workers in and then letting them go, and the sequence is the policy.

Refundability: The Rule That Turns Tax Law Into Cash

Refundability is the provision that converts section 32 from a tax reduction into a cash transfer. The credit is payable in excess of tax liability, which means the statute computes the credit amount from the filer’s earnings and qualifying children, subtracts it from the tax owed, and pays any remainder to the filer. A filer who owes no tax at all still receives the full credit as a payment. That single rule is what makes the credit a cash benefit rather than a tax cut, and it is the reason the program can reach households whose earnings are too low to generate income tax liability in the first place.

How does Section 32 turn a tax provision into cash for workers who owe no tax?

Section 32 computes a credit amount from the filer’s earnings and qualifying children, subtracts it from the tax owed, and pays any remainder to the filer. Because the statute makes the amount payable in excess of liability, a filer with no tax owed receives the full credit as a payment.

The mechanics repay close attention because they distinguish the credit from every nonrefundable benefit in the code. A nonrefundable credit can reduce tax liability to zero but not below: a filer who owes $500 in tax and holds a $2,000 nonrefundable credit uses $500 of it and forfeits the rest. A deduction is weaker still, because a deduction only reduces taxable income, so its value equals the filer’s marginal rate multiplied by the deducted amount, and it is worth nothing to a filer with no taxable income. The earned income tax credit escapes both limitations. Its refundability decouples the benefit from the filer’s tax liability entirely: the payment does not depend on owing tax, does not scale with the marginal rate, and does not vanish when liability hits zero. A low-earning parent who owes no income tax receives the same credit as an identical parent who owes some, because the statute pays the computed amount either way.

That decoupling is what makes the credit the largest need-tested cash antipoverty program in the federal repertoire. The Congressional Research Service, in its report on the credit designated R43805, describes it as the largest need-tested antipoverty program that provides cash benefits, a formulation this article adopts because it states the program’s scale without importing the measurement disputes that surround any particular year’s figures. The phrase carries three distinct claims. Need-tested means the benefit flows to households selected by the statute’s income and family tests rather than to the general population. Cash means the benefit arrives as money the household can spend, not as a voucher, a service, or a reduction in a bill. Largest means that, measured by the dollars reaching low-income households, no other federal program that combines those two features moves more resources. For an article dated in 2013, the CRS formulation states the credit’s standing in the antipoverty system as the research service described it in that era, and it does so without leaning on any single year’s expenditure total.

Refundability also explains the credit’s most persistent administrative problem. Because the benefit is paid on the basis of facts the filer reports on the return, including earnings, the qualifying child’s residency, and the child’s relationship to the claimant, the program depends on self-certification to a degree unusual even among tax benefits. The Internal Revenue Service processes the credit through the ordinary return pipeline, which means eligibility determinations that would receive caseworker review in a traditional welfare program are made by the claimant, under penalties of perjury, with verification occurring after payment rather than before. Complexity compounds the risk: the four qualifying-child tests, the investment income limit, the filing-status bar, and the earned income definition create multiple points where an honest filer can err and where a dishonest one can misrepresent. The result is the highest improper payment rate of any major federal program, a record that is the direct counterpart of the credit’s design virtues. The same refundability that lets the provision reach workers who owe no tax also lets erroneous claims convert immediately into cash, and the same definitional precision that targets the benefit accurately creates the lines across which errors occur. The improper payment record is not an accident that befell the credit; it is the shadow of the mechanism that makes the credit work.

The phrase that does the work is payable in excess of tax liability, and its operation inside return processing deserves a precise description. When a return claims the credit, the Internal Revenue Service computes the credit amount from the earnings and qualifying children reported, applies it against the tax the return shows as owed, and treats any excess as an overpayment to be refunded. The sequence matters because it shows that refundability is not a separate program grafted onto the credit; it is the credit’s own arithmetic carried past zero. A nonrefundable credit stops at zero liability and discards the remainder; the earned income tax credit continues past zero and pays the remainder. The difference between stopping and continuing is the difference between a tax reduction and a cash transfer.

The three-way comparison among deductions, nonrefundable credits, and refundable credits clarifies what is at stake. A deduction reduces the income subject to tax, so its cash value depends on the filer’s marginal rate: the same deduction is worth more to a high-rate filer than to a low-rate filer, and worth nothing to a filer with no taxable income. A nonrefundable credit reduces the tax itself dollar for dollar, which improves on the deduction’s regressivity, but it cannot exceed the tax owed, so it too is worth nothing to the filer with no liability. A refundable credit breaks the link to liability entirely: its value is the computed amount, paid in full whether the filer owes tax or not. The progression from deduction to nonrefundable credit to refundable credit is a progression toward decoupling the benefit from the tax system that delivers it, and the earned income tax credit sits at the far end of that progression. Its value to a low-earning household does not depend on the household’s tax position at all.

That decoupling is what allows the credit to reach its intended population, because the intended population substantially overlaps the population with no income tax liability. Low-earning workers often owe no federal income tax: their earnings fall below the thresholds at which liability begins, particularly once the standard deduction and personal exemptions are accounted for. A nonrefundable credit aimed at such workers would be a promise the tax system could not keep, since there would be no liability for the credit to offset. Refundability closes the gap between the credit’s ambition and the tax system’s reach by paying the benefit as cash rather than as liability reduction. The design choice reflects an understanding that the workers whose payroll taxes the credit was created to offset are often the same workers who owe no income tax, and a credit confined to liability reduction would miss them.

Cash, as the benefit’s form, carries its own significance. A cash payment is fungible: the household decides whether it covers rent, food, transportation, childcare, or debt, allocating the money to its most pressing need without seeking permission. In-kind benefits and vouchers restrict that choice, directing the household’s spending toward categories the program’s designers selected. The credit’s designers selected no categories, which is consistent with the provision’s origin as a tax refund rather than a welfare grant: a refund returns the worker’s own money, in the political framing Long established, and refunds do not come with spending instructions. The fungibility of the payment is therefore not an oversight but a feature of the bargain, and it distinguishes the credit from the need-tested programs that deliver specific goods.

The Congressional Research Service formulation adopted earlier in this article states the credit’s scale with deliberate neutrality. Describing the credit as the largest need-tested antipoverty program that provides cash benefits makes a comparative claim about the program’s size without endorsing any particular year’s expenditure figure, which is the appropriate posture for an article whose date fixes its evidentiary horizon. The formulation’s three terms each do work: need-tested identifies the targeting mechanism, cash identifies the benefit’s form, and largest identifies the program’s rank among federal efforts sharing those features. Together they explain why the credit occupies the central position it does in discussions of American antipoverty policy: it is the biggest thing the federal government does that both selects by need and pays in money.

The improper payment record, the shadow side of refundability, follows from the same features that produce the program’s reach. Eligibility turns on facts the filer supplies: the amount of earned income, the child’s relationship and residency, the absence of disqualifying investment income, the correctness of the filing status. The Internal Revenue Service verifies these facts after payment rather than before, because pre-payment verification of tens of millions of returns would require the very caseworker apparatus the credit was designed to avoid. Complexity multiplies the opportunities for error at every step: a grandparent unsure whether the residency test is met, a self-employed worker uncertain about net earnings, a separated couple confused about filing status. Each uncertainty is a potential improper payment, and the aggregate across the filing population produces the highest improper payment rate of any major federal program. The rate is not evidence that the credit’s recipients are unusually dishonest; it is evidence that a program which pays cash on self-certified facts through the tax pipeline will convert every ambiguity in its definitions into dollars. Refundability makes the credit a cash transfer, and the cash transfer’s accuracy is bounded by the precision of the text that defines it.

Expansion history

The benefit Congress wrote in 1975 did not stay small. Five rounds of legislation between 1986 and 2009 enlarged the formulas, added family-size tiers, and eventually extended eligibility to earners without children at home. Each round moved through the ordinary tax-legislation process: the House Ways and Means Committee and the Senate Finance Committee drafted the language, the Joint Committee on Taxation scored the revenue effects, and budget reconciliation carried the measure whenever leaders chose that vehicle. Because Section 32 lives in the Internal Revenue Code, no enlargement could travel as an appropriations rider; every change had to survive the committees that guard the tax law, a procedural point explained in how tax bills move through Congress.

The first great enlargement arrived inside the Tax Reform Act of 1986, Public Law 99-514. That act lifted the credit rate to 14 percent and, just as consequentially, permanently indexed the income thresholds to inflation. Indexing solved a quiet erosion problem. Before 1986, rising prices pushed low-wage earners into higher nominal income bands each year, shrinking the real value of the payment unless Congress acted. After 1986 the parameters adjusted on their own, so the real value held steady between legislative rounds. The 1986 enlargement also fit the reform’s political logic. Lawmakers who were cutting marginal rates wanted to demonstrate that the overhaul protected low-wage earners, and expanding a work-conditioned benefit let legislators of both parties make that claim. Researchers later used this very expansion as identifying variation: Nada Eissa and Jeffrey Liebman compared single mothers with single childless women across the 1986 change to measure the employment response, a study examined in the evidence section below.

The Omnibus Budget Reconciliation Act of 1990, Public Law 101-508, reshaped the schedule around family size. Where the original formula paid one benefit regardless of household size, the 1990 law split the schedule into a one-child formula and a two-or-more-child formula, paying larger amounts to bigger families. The reasoning was plain: a second child raises household need without raising the parent’s wage, so a flat payment left larger families deeper in poverty. The 1990 changes took effect in stages across 1991 through 1994, which is why the provision’s cost curve bends sharply upward in the early 1990s. By 1995, Representative Dave Camp could tell the House that three legislative revisions had expanded the program’s annual cost tenfold, to almost 25 billion dollars a year, a statement recorded in the Congressional Record of October 25, 1995. The figure captured how thoroughly the measure had outgrown its 1975 origins, when the first-year cost ran about 1.2 billion dollars and a decade later about 2.5 billion, according to the same floor remarks.

The largest single enlargement came with the Omnibus Budget Reconciliation Act of 1993, Public Law 103-66. The 1993 law substantially expanded the credit across the board and, for the first time, made workers without qualifying children eligible. The childless extension was modest by design: a small supplement meant to offset payroll taxes for low-wage earners with no dependents at home, rather than a full antipoverty benefit. Its symbolic weight was nonetheless large. For eighteen years the provision had served only households with children; after 1993 it reached childless adults as well, and later debates about adequacy would start from that baseline. The 1993 expansion phased in over 1994 through 1996, and its scale showed up quickly in participation figures and in the employment literature. Bruce Meyer and Dan Rosenbaum’s later work on single mothers’ labor supply treats the 1993 enlargement as one of the central policy shocks of the period, alongside welfare reform and a strong labor market.

Two later rounds refined rather than reinvented the structure. The Economic Growth and Tax Relief Reconciliation Act of 2001, Public Law 107-16, raised the income thresholds at which married couples filing jointly began to lose the benefit, supplying the first dedicated marriage-penalty relief in the credit’s history. The American Recovery and Reinvestment Act of 2009, Public Law 111-5, added a third family-size tier with a higher rate for households with three or more children and lifted the joint-filer thresholds further, extending the marriage-penalty relief begun in 2001. Both changes reflected a maturing view of the provision. No longer a small offset tucked into a tax bill, the EITC had become one of the federal government’s principal antipoverty tools, and fine-tuning its edges had become routine tax-policy business conducted through the same committees and scorekeepers as any other revenue measure.

One procedural thread ties the five rounds together. Expansions of this kind never moved as standalone EITC bills. They rode inside broader vehicles, were scored as combinations of revenue loss and outlay increase, and were debated under reconciliation rules when leaders invoked them. That is why each enlargement above carries the public law number of a larger tax or budget package. The expansion story is therefore also a tour through three decades of tax legislation, with the credit traveling inside, growing a little with almost every major revenue bill that left the Finance and Ways and Means committees between the Reagan and Obama years.

The 1986 changes deserve a closer look because they set the template for later rounds. The rate increase to 14 percent was large enough to matter in household budgets, but the permanent inflation indexing was the more durable reform. Before indexing, every year of inflation quietly shrank the real benefit by pushing workers into higher nominal income bands, and only fresh legislation could restore the lost value. Indexing broke that cycle by tying the thresholds to price changes automatically, which meant the credit’s real value no longer depended on Congress remembering to act. The politics of the moment helped: with lawmakers cutting marginal rates across the board, expanding a benefit aimed at low-wage earners let the reform’s authors answer the charge that the overhaul favored the affluent.

The 1986 act paired the credit expansion with other low-income relief that is easy to overlook. By raising the standard deduction and personal exemptions substantially, the reform removed millions of low-income filers from the income-tax rolls entirely, which meant the refundable portion of the benefit did more of the antipoverty work while the nonrefundable portion mattered less. The combination reflected a coherent strategy: take the poorest filers off the tax rolls, then use the refundable payment to supplement the wages of those who remained. Later expansions would build on that foundation rather than revisit it, which is why the 1986 round reads in retrospect as the moment the modern credit took shape.

The 1990 round showed how the provision rides inside larger bargains. The Omnibus Budget Reconciliation Act of 1990 was principally a deficit-reduction package negotiated at the 1990 budget summit, and the EITC enlargement traveled inside it alongside tax increases and spending cuts. The law also created two temporary supplemental credits available from 1991 through 1993, according to the Congressional Research Service’s legislative history: a young-child supplement that added five percentage points to the credit rate for families with infants, and a health-insurance supplement worth up to six percentage points for families purchasing coverage. The 1993 law ended both supplements and folded their value into the larger base expansion, which is why the family-size schedule looks cleaner after 1994 than during the transition years.

The 1993 expansion provoked the sharpest debate of any round. Supporters cast it as the work-based centerpiece of the administration’s antipoverty strategy, a way to lift full-time workers above the poverty line without expanding cash welfare. Opponents warned about the cost and about extending the benefit beyond families with children. Representative Dave Camp captured the latter view on the House floor in October 1995, arguing that Congress should not give the credit to childless workers because for eighteen of the program’s nineteen years both parties had agreed the benefit belonged to working families with dependent children, as recorded in the Congressional Record of October 25, 1995. The childless extension survived, but the objection explains why its parameters stayed so modest in every Congress that followed.

The 2001 round, though smaller in dollars, fixed a structural complaint that had been building for years. Before the Economic Growth and Tax Relief Reconciliation Act of 2001, the joint-filer phase-out thresholds sat only modestly above the single thresholds, so marriage could cost a two-earner couple a noticeable share of its combined benefit. The 2001 law raised the income level at which married couples filing jointly begin to lose the credit, delivering the first dedicated marriage-penalty relief in the program’s history. The change was phased in over several years, a common technique for fitting tax cuts inside reconciliation’s budget windows, and it established a principle that later expansions would revisit: the tax code should not punish low-income workers for marrying.

The 2009 expansions were enacted as temporary provisions, and their afterlife illustrates the extenders treadmill on which much of the tax code rides. The December 2010 tax compromise extended the higher three-child rate and the marriage-penalty relief through 2012, as the Treasury Department noted in its fact sheet on the fiscal year 2012 budget. Each extension required a fresh legislative vehicle and a fresh round of scoring, which is why permanence became the rallying cry of the 2013 proposals discussed below. The pattern also confirms the procedural lesson: the credit grows in the slipstream of larger bargains, not through standalone bills.

Employment and health evidence

Expansions of the credit did more than move money; they gave economists a sequence of policy shocks clean enough to study. Because each enlargement changed incentives for eligible households while leaving similar ineligible households untouched, researchers could compare the two groups before and after a change and attribute the difference to the policy. That comparison design, often called difference-in-differences, underpins the most cited findings on the provision’s effects. The literature that resulted is unusually consistent: the benefit draws people into paid work, and there is credible evidence that it improves infant health as well.

The founding study of the employment literature is Nada Eissa and Jeffrey B. Liebman’s “Labor Supply Response to the Earned Income Tax Credit,” published in the Quarterly Journal of Economics, volume 111, number 2, in May 1996. Eissa and Liebman used the 1986 expansion as their experiment, comparing single women with children, who gained eligibility for a larger benefit, against single women without children, who did not. They found that the 1986 change raised the relative labor force participation of single women with children by up to 2.8 percentage points. Just as telling was what they did not find: among women already working, hours barely moved. The credit appeared to operate on the extensive margin, pulling people into jobs, rather than on the intensive margin of lengthening the workweeks of those already employed. That distinction shaped two decades of later modeling, because it implied that the subsidy’s main labor-market channel was entry into employment rather than longer hours.

Bruce D. Meyer and Dan T. Rosenbaum extended the inquiry across a longer window in “Welfare, the Earned Income Tax Credit, and the Labor Supply of Single Mothers,” published in the Quarterly Journal of Economics, volume 116, number 3, in August 2001. Looking at the years from 1984 to 1996, a period that spans the 1986, 1990, and 1993 expansions as well as welfare reform and a strong economy, they decomposed the sources of the large rise in work among single mothers. Their conclusion attributed a large share of that increase to the EITC and other tax changes, even after accounting for welfare policy shifts and labor-market conditions. The paper mattered because it separated the credit’s contribution from the confounding effects of a booming 1990s economy and the 1996 welfare overhaul, both of which also pushed single mothers toward employment. By isolating the tax channel, Meyer and Rosenbaum gave legislators a quantified basis for treating the provision as an employment policy rather than merely a transfer.

A third line of research connected the benefit to health outcomes. Hilary Hoynes, Doug Miller, and David Simon examined infant health in “Income, the Earned Income Tax Credit, and Infant Health,” circulated as NBER Working Paper 18206 in July 2012. Using variation from the credit’s expansions, they found that the EITC reduces the incidence of low birth weight. The mechanism runs through household resources: additional income during pregnancy improves maternal nutrition, reduces stress, and increases use of prenatal care, and low birth weight is among the most consequential early-life health markers because of its links to later development. Because the journal version of the paper appeared after this article’s 2013 publication date, in the American Economic Journal: Economic Policy in February 2015, the 2012 working paper is the citable source for a 2013-dated account. The finding widened the provision’s documented effects beyond earnings and employment into the next generation’s health.

Analyses by the Center on Budget and Policy Priorities have translated these mechanisms into poverty figures, finding that the credit lifts millions of people out of poverty each year. The antipoverty claim rests on the same structure that produces the employment effects: because the benefit phases in with earnings, it raises the incomes of working households without requiring them to stop working to qualify, unlike traditional cash assistance. The combination of the participation evidence, the decomposition evidence, the infant-health evidence, and the poverty analyses forms the empirical core of the case for the provision as designed. Each strand carries named authors, dated publication, and a specified method, which is why the findings have survived repeated reexamination across different expansion episodes and different decades.

Why the employment effect concentrates on labor-force entry rather than hours is worth spelling out, because it explains the policy’s design logic. Going to work involves fixed costs: childcare, transportation, work clothes, and the loss of time for household production. For a single mother weighing a low-wage job, those fixed costs can exceed the first weeks of pay, which is why many eligible nonworkers sat out of the labor force before the expansions. The phase-in rate attacks exactly that margin by raising the effective wage on the first dollars of earnings, tipping the calculation for people on the fence about working at all. Once someone is employed, the same rate matters less, because the fixed costs are already sunk and the decision shifts to how many hours to supply, where income effects and the plateau’s flat benefit blunt the incentive. Eissa and Liebman’s finding of a participation response with no hours response fits that account precisely, and later researchers have treated the extensive margin as the provision’s primary labor-market channel.

Meyer and Rosenbaum’s decomposition deserves emphasis for its care with confounding explanations. The 1984 to 1996 window they studied contains three credit expansions, a major welfare overhaul, state welfare waivers, minimum-wage increases, and one of the strongest labor markets of the postwar era. Any of those could have pulled single mothers into work. Their method controlled for state-level unemployment rates, welfare policy variation, and demographic shifts, and still left a large role for the credit and other tax changes. The phrase “and other tax changes” matters: the authors credited the tax system broadly, not the EITC alone, which is the careful reading their evidence supports. What the paper ruled out was the simpler story that the economy or welfare reform did all the work.

The infant-health finding extends the evaluation into the next generation. Hoynes, Miller, and Simon traced additional household income during pregnancy to measurable improvements in birth weight, working through channels that poverty researchers have long documented: better maternal nutrition, earlier and more consistent prenatal care, and reduced physiological stress. Low birth weight predicts a range of later difficulties, from infant mortality risk to developmental delays, so even modest reductions carry outsized significance. That a tax provision aimed at work incentives shows up in delivery-room outcomes illustrates how broadly income effects propagate, and it gave the credit’s defenders an argument that reached beyond employment statistics.

The poverty figures most often cited for the provision come from analyses by the Center on Budget and Policy Priorities using poverty measures that count tax credits as household income. Traditional poverty statistics ignore the EITC entirely, because they tally only cash income before taxes, which means the official rate understates the benefit’s effect by construction. Measures that add the credit back in show millions of people, disproportionately children, lifted above the poverty line each year. The choice of measure is not a technical footnote; it determines whether the largest cash transfer for working families registers in the nation’s headline poverty statistics at all.

One caveat rounds out the picture. The strongest employment evidence concerns single mothers, the group whose incentives changed most dramatically across the expansions. For married couples, the theory is more ambiguous: in the phase-out range, the benefit shrinks as household earnings rise, which can discourage work by a second earner whose wages push the family further into the clawback zone. Empirical work has found suggestive evidence of small negative effects on married women’s participation, though the estimates are less precise than the single-mother results. The consensus, then, is specific rather than blanket: the credit robustly increases work among single mothers, with more uncertain effects at other margins.

Improper payments

The same program that the employment literature praises carries the highest improper-payment rate among large federal programs, and any honest account must give that record equal weight. For fiscal year 2023, the improper-payment rate for the credit stood at 33.5 percent, representing 21.9 billion dollars of the 65.4 billion dollars claimed, according to the fiscal year 2023 Treasury Agency Financial Report as reviewed by the Treasury Inspector General for Tax Administration. That figure is the most recent authoritative published estimate. Later reporting has shown movement in the rate: for fiscal year 2024 the estimate was 27.3 percent, per the Government Accountability Office report GAO-26-108044, and figures near 24 percent that sometimes circulate correspond to older years, fiscal 2016 through 2019, per paymentaccuracy.gov. The numbers describe overpayments, meaning dollars paid in excess of what the statute allowed, and they are measured through the federal government’s standard payment-integrity framework rather than through criminal investigation.

The drivers of those overpayments point to a design problem more than to a fraud problem. The paymentaccuracy.gov scorecard for the credit attributes roughly 20.57 billion dollars of overpayments to the government’s inability to authenticate qualifying-child eligibility, specifically the relationship and residency tests, and to misreported income. The Internal Revenue Service administers the benefit largely on the basis of taxpayer self-reporting, with few internal or external databases available for up-front verification of who lived with whom and for how long. Consider what the statute asks the agency to confirm: whether a child lived with the claimant for more than half the year, in households where custody is shared, grandparents and parents trade caregiving across months, and living arrangements shift midyear. No national registry records those facts in real time, and the agency’s matching programs can confirm wages from employer reports far more easily than they can confirm a child’s bedroom. Complexity of the rules in complex households, not simple fraud, generates most of the erroneous dollars.

That distinction shapes how policymakers have responded. If the errors were mostly deliberate deception, the natural answer would be enforcement: more audits, stiffer penalties, narrower gates. If the errors flow mostly from rules that ordinary filers misunderstand and that the agency cannot verify before paying, the natural answers look different: simpler eligibility definitions, better data matching, and payment timing that allows verification before money moves. Congress chose the timing tool in the Protecting Americans from Tax Hikes Act of 2015, Division Q of Public Law 114-113, enacted December 18, 2015. Two years after this article’s 2013 publication date, Congress required the Internal Revenue Service to hold refunds on returns claiming the EITC or the Additional Child Tax Credit until February 15, effective for the 2017 filing season, so that the agency could match return information against employer wage reports before releasing money. The hold does not change who qualifies; it changes when the government pays, buying verification time at the cost of delaying refunds that many low-income households count on early in the year.

The error-rate evidence and the employment evidence therefore describe the same provision from opposite sides. The expansions that researchers use as clean experiments are also the rule changes that filers must track; the family-size tiers that target poverty more precisely are also the distinctions that generate residency disputes; the refundability that makes the benefit reach workers who owe no income tax is also what puts tens of billions of outlay dollars beyond the reach of prepayment verification. A serious evaluation holds both pictures at once: a transfer with unusually strong evidence of raising employment and reducing poverty, delivered through rules unusually prone to payment error, with the error concentrated in the hardest-to-verify corners of family life rather than in fabricated claims.

Understanding how the government measures these errors clarifies what the headline rate does and does not claim. The Treasury’s Agency Financial Report estimates improper payments through statistical sampling of tax returns, and the Treasury Inspector General for Tax Administration reviews the methodology each year. “Improper” in this framework is a broad category: it includes payments made in the wrong amount, payments to ineligible recipients, and payments with insufficient documentation to confirm eligibility. A return that claims a qualifying child without adequate supporting records can count as improper even when no one can prove the claim was wrong. The 33.5 percent figure for fiscal 2023 is therefore best read as an upper-bound estimate of dollars that left the Treasury without full verification, not as a measured fraud rate. Criminal tax fraud, which requires proving intent to deceive, accounts for a small fraction of the total.

Read together, the published figures trace a rate that has moved within a band rather than trending decisively in either direction: roughly 24 percent in fiscal 2016 through 2019 per paymentaccuracy.gov, 33.5 percent in fiscal 2023 per the Treasury Agency Financial Report as reviewed by the Inspector General, and 27.3 percent in fiscal 2024 per the Government Accountability Office. Methodological refinements in sampling and estimation complicate year-to-year comparisons, so analysts caution against reading short-term movements as proof that the underlying error rate is rising or falling. What has not changed across the series is the ranking: in every recent year, the credit’s improper-payment rate has stood above that of every other large federal program, a persistent outlier that keeps the design debate alive regardless of the latest point estimate.

The qualifying-child rules themselves generate much of the ambiguity. When two adults can plausibly claim the same child, as happens routinely with grandparents who share caregiving or with parents who split custody across the year, the statute supplies tie-breaker rules that assign the child to one claimant through a hierarchy generally preferring parents over nonparents and longer residency over shorter. Applying those rules requires facts the tax return does not collect: who slept where, in which months, under whose roof. Paid preparers, who handle a large share of EITC returns, must complete a due-diligence checklist, Form 8867, documenting the eligibility questions they asked, and face penalties for failing to do so. But a checklist cannot conjure records that do not exist, and correspondence audits months after filing rarely reconstruct a household’s living arrangements with confidence.

The PATH Act’s refund hold emerged from this diagnostic picture. Rather than narrowing eligibility or raising penalties, Congress bought the agency time: by holding EITC and Additional Child Tax Credit refunds until February 15, the 2015 law let the Internal Revenue Service match returns against employer wage reports before disbursing funds, catching misreported income that self-reported returns had previously carried past the payment date. The provision rode inside a year-end omnibus package, the Protecting Americans from Tax Hikes Act, with bipartisan support, and it reflected a legislative judgment that verification timing was the most cost-effective lever available. The trade-off was explicit in the debate: slower refunds for millions of compliant households in exchange for fewer erroneous payments. Whether that exchange was worth it depends on how one weighs payment accuracy against the liquidity needs of low-income filers, a value judgment the statute leaves to the reader.

Childless workers

Workers without qualifying children entered the credit’s universe with the 1993 expansion, and the terms of their admission have remained strikingly ungenerous ever since. The phase-in rate for childless claimants is 7.65 percent, meaning each dollar of earnings adds just under eight cents of benefit until the maximum is reached. Filers with one child phase in at 34 percent, those with two children at 40 percent, and those with three or more at 45 percent. The childless schedule also phases out at a much lower income level, so the benefit disappears quickly as earnings rise. The structural message is unmistakable: the provision treats childless earners as a minor add-on to a family-centered program, not as full participants.

The eligibility conditions for the childless benefit add further gates. Claimants must be at least 25 years old and under 65, must not be claimable as a dependent on anyone else’s return, and must have lived in the United States for more than half the year. The age band excludes younger workers, including many teenagers and early-twenty-somethings in low-wage jobs, and the dependent rule excludes students and others supported by parents. Together with the low phase-in rate and the early phase-out, these conditions produce a benefit that functions mainly as a small offset to payroll taxes rather than as a meaningful income supplement. An American Enterprise Institute analysis of 2012 data reported that of roughly 64 billion dollars in federal EITC benefits that year, less than 3 percent flowed to childless workers, a distribution that follows directly from the parameters Congress chose in 1993 and left largely untouched afterward.

Defenders of the existing structure have pointed to the 7.65 percent rate as calibrated to the combined employee share of payroll taxes, so that the small benefit roughly cancels the payroll-tax burden on the first dollars of earnings. Critics, in debates running through the article’s 2013 publication date, answered that payroll-tax offset was a thin rationale for excluding young adults and for phasing the benefit out so early that full-time minimum-wage workers could age out of meaningful help. Both positions accept the same structural facts; they differ on whether a work-conditioned transfer should concentrate its generosity on households with children or spread it across the low-wage labor force. The 1993 compromise chose concentration, and every Congress through 2013 left that choice in place.

The age band for childless claimants reflects a judgment about who the small benefit is for. Setting the floor at 25 excludes younger workers, many of whom are students or dependents in larger households, on the theory that the credit should support independent adults rather than subsidize youth employment broadly. The ceiling at 64 moves older workers toward retirement programs instead. Whatever one thinks of the line-drawing, the band’s narrowness compounds the schedule’s stinginess: a 23-year-old working full time at low wages gets nothing, and a 66-year-old in the same job gets nothing, while a 40-year-old gets a benefit that phases out quickly.

The phase-out mechanics deserve attention because they make the childless schedule unusually steep. The benefit reaches its modest maximum over a short phase-in range and then disappears across a compressed income band, which means the effective clawback rate in the phase-out range is high relative to the small maximum at stake. A worker whose earnings rise by a few thousand dollars can lose the entire benefit, a sharper cliff than anything on the family schedules. Defenders describe this as the price of targeting: with limited dollars, Congress concentrated the phase-out where the benefit was thinnest. Critics answer that steep phase-outs punish exactly the earnings growth the program is supposed to encourage.

By this article’s 2013 date, dissatisfaction with the childless parameters had spread across party lines. Contemporary analyses noted expressions of support for enlarging the childless benefit from figures as different as Representative Paul Ryan and President Obama, a convergence that foreshadowed the more ambitious proposals of later years. The shared premise was that a work-conditioned transfer which leaves young childless adults behind misses a population with high poverty rates and low employment. The 1993 compromise had been built for a different debate, and by 2013 its defenders were fewer.

Reform proposals

As of this article’s July 2013 publication date, two pending proposals aimed to reshape the credit, and both illustrate how the provision’s politics had settled into defense of past expansions rather than invention of new ones. In April 2013, Senators Sherrod Brown and Dick Durbin introduced the Working Families Tax Relief Act of 2013, with Senate Finance Committee chairman Max Baucus among the cosponsors. The bill would have made permanent the 2009 expansions: the higher credit rate for families with three or more children and the marriage-penalty relief that raised joint-filer phase-out thresholds. Those 2009 changes had been enacted as temporary provisions inside the Recovery Act, and the Brown-Durbin bill treated their permanence as unfinished business. The cosponsorship of the Finance chairman signaled that the proposal carried weight inside the committee that would have to move it, though introduction is not enactment and the bill faced the ordinary gauntlet of scoring, offsets, and floor scheduling.

The second proposal came from the executive branch. President Obama’s fiscal year 2014 budget, released April 10, 2013, proposed permanently extending the 2009 expansions and simplifying the rules for claiming the credit for workers without qualifying children. The simplification plank mattered because the childless rules, with their age band and residency tests layered atop the standard earned-income definitions, generated disproportionate confusion relative to the small dollars at stake. The budget’s approach was incremental: lock in the temporary gains, reduce friction for the smallest claimant group, and leave the core family-size structure alone. One clarification is necessary because later commentary sometimes blurs it. The fiscal 2014 budget did not propose doubling the childless credit; that larger proposal appeared in the fiscal 2015 budget, after this article’s date, and belongs to a later chapter of the debate.

Neither proposal had become law by July 2013, and their fates illustrate a durable feature of the provision’s legislative life. Expansions move when they can ride a larger vehicle, and standalone credit bills more often serve as markers of intent than as enacted statutes. The 2009 expansions themselves survived only because they were attached to stimulus legislation; their temporary status then generated the permanence campaigns of 2013. Anyone reading the credit’s future in that year would have been wise to watch the next budget reconciliation or tax package rather than the standalone bills, since history showed the provision growing in the slipstream of larger bargains.

The mechanics of permanence explain why the Brown-Durbin bill mattered beyond its policy content. Temporary tax provisions live under a scoring rule that makes extension look costly: the Joint Committee on Taxation measures a permanence bill against a baseline in which the provision expires, so making the 2009 expansions permanent carried a ten-year price tag even though the policy merely continued existing law. That scoring convention is why temporary provisions so often get extended rather than made permanent; extension bills can be tucked into larger packages where their cost dissolves into the total, while permanence bills must carry their full score in daylight. With Finance Chairman Baucus as a cosponsor, the Working Families Tax Relief Act had a plausible path through committee, but the score, the search for offsets, and the crowded 2013 calendar stood in its way.

The fiscal 2014 budget’s simplification plank addressed a quieter problem. The childless rules layer an age test, a dependent test, and a residency test on top of the standard earned-income definitions, generating confusion out of proportion to the dollars involved. Simplification could mean fewer tests, clearer definitions, or better coordination with information the agency already holds, though the budget left the specifics to Congress. The proposal’s modesty was the point: rather than redesigning the benefit, the administration asked to reduce the friction that kept some eligible childless workers from claiming it at all. As of July 2013, both the bill and the budget proposal remained pending, awaiting a legislative vehicle large enough to carry them.

The politics of 2013 also featured a quieter contest over which committee would own the credit’s future. With the Finance Committee under Baucus focused on comprehensive tax reform, some advocates hoped the EITC expansions would be folded into a broader rewrite, while others preferred a standalone extenders vehicle where the provisions faced less competition for offsets. The choice mattered because tax reform’s revenue-neutrality constraint would have forced the expansions to be paid for with other tax increases, while an extenders package could carry them with less visible cost. That jurisdictional jockeying helps explain why the 2013 proposals stayed in the introduction-and-endorsement phase through the article’s July publication date: no consensus had formed on the vehicle, and without a vehicle the provisions could not move.

Where the credit sits among antipoverty programs

The EITC does not operate alone. It works alongside the Temporary Assistance for Needy Families program, created by the welfare reform statute of 1996, and the Supplemental Nutrition Assistance Program, and the three form a rough division of labor. TANF provides time-limited cash assistance oriented toward moving recipients into work, SNAP addresses food needs regardless of employment, and the credit supplements the earnings of people already working. The design logic is complementary: the benefit rewards labor-market participation while the other programs address nonworkers and needs that wages alone do not cover, such as nutrition and the transition costs of entering employment. A household can receive all three at once, and the programs’ different eligibility rules mean they reach overlapping but distinct populations.

That complementarity also defines the credit’s limits. Because the benefit phases in with earnings, it offers nothing to people with no earnings at all: the nonworking poor, including many elderly and disabled people, fall outside its reach by construction. TANF and SNAP, whatever their own restrictions, do not share that particular boundary. The provision is therefore best understood as an antipoverty tool for the working poor specifically, not as a general guarantee against destitution. Its refundability extends its reach down the earnings scale to workers whose incomes are too low to owe income tax, but it cannot reach below zero earnings.

In the hierarchy of American antipoverty efforts, the credit occupies a distinctive middle position. Social Security remains the largest overall antipoverty program in the United States, a standing documented in Social Security’s poverty record, and its poverty-reducing power dwarfs every means-tested program because of its scale and its near-universal coverage of the elderly. Among cash means-tested transfers, however, the EITC stands as the largest operating through the tax code for working families, moving tens of billions of dollars a year as a mix of forgone revenue and direct outlays. The tax-code delivery is the point of the comparison: where Social Security and TANF pay benefits through agencies, the credit pays through the annual tax return, reaching eligible workers without a separate application to a welfare office. That delivery channel explains both its high participation among eligible filers and its high improper-payment rate, since the tax system was built to collect revenue, not to verify where children sleep.

The complementarity with the 1996 welfare law runs deeper than shared clientele. The Personal Responsibility and Work Opportunity Reconciliation Act imposed work requirements and lifetime time limits on cash assistance, deliberately making nonwork less sustainable for recipients. That policy only makes moral and practical sense if work itself pays enough to live on, which is the gap the credit fills. Analysts across the ideological spectrum have described the two as a matched pair: the welfare law supplies the push toward employment, and the EITC supplies the pull by raising the return to low-wage work. Neither works as intended without the other, which is why retrenchment debates tend to implicate both at once.

The comparison with Social Security sharpens what “largest” means in each claim. Social Security’s antipoverty power comes from scale and universality: it reaches nearly all elderly Americans with benefits large enough to lift most above the poverty line, and its effect shows up overwhelmingly among retirees. The EITC’s distinction is narrower but real: among cash transfers that phase out with income and flow through the tax system to working households, nothing else moves comparable dollars. The delivery channel is the substantive difference. Social Security and TANF pay through benefit agencies with applications, interviews, and caseworkers; the credit pays through the annual tax return, reaching eligible filers who may never set foot in a welfare office. That accessibility helps explain the provision’s high take-up among eligible workers, and it also explains the verification problem, since the tax system was engineered to collect revenue rather than to confirm household composition.

Program rules are written to keep the layers from working at cross-purposes. Federal provisions generally prevent a one-time EITC refund from immediately disqualifying a household from nutrition assistance, so that the credit supplements rather than supplants other aid in the months after filing. The coordination is imperfect, and caseworkers and claimants do not always understand the disregards, but the intent is clear: the safety net is supposed to function as a stack, with the earnings supplement sitting atop nutrition and cash assistance rather than knocking households off them. Evaluating the credit in isolation therefore understates its role, since part of its value lies in making the rest of the system cohere around work.

The boundary of the design is as important as its reach. By conditioning aid on earnings, the credit writes out of the program everyone with no labor income: the long-term unemployed, many people with disabilities, and elderly people outside the workforce. Those populations depend on the programs the credit complements, which is why no serious analyst treats the EITC as a substitute for the safety net as a whole. It is one layer, designed for one population, and its celebrated employment effects are inseparable from that limitation.

Studying Section 32

Legislative staff and researchers approaching Section 32 for the first time should start with the statute text itself, then work outward through the layers that give the bare language its meaning. The Internal Revenue Code section supplies the definitions: earned income, adjusted gross income thresholds, the phase-in rates and phase-out mechanics, the qualifying-child tests, and the special rules for childless claimants. The committee reports on each expansion supply the intent: the 1986 report explains indexing, the 1990 and 1993 reconciliation reports explain the family-size tiers and the childless extension, and the 2001 and 2009 materials explain the marriage-penalty adjustments. Because every enlargement rode inside a larger vehicle, the reports to read are those of the Tax Reform Act of 1986, the 1990 and 1993 reconciliation acts, the 2001 tax act, and the 2009 Recovery Act, not freestanding EITC documents that do not exist.

Quantitative work should lean on three recurring sources. The Joint Committee on Taxation’s revenue estimates score each expansion and reveal how scorekeepers split the cost between revenue loss and outlay, a split that matters for budget enforcement. The Congressional Research Service’s legislative histories collect the public law citations, effective dates, and parameter tables in one place, sparing researchers the work of reconstructing the phase-in schedules from session laws. The Internal Revenue Service’s Statistics of Income data and the Treasury’s payment-integrity reporting supply the participation and error-rate series, including the fiscal 2023 improper-payment figures discussed above. For the employment literature, the original journal articles remain the essential reads: Eissa and Liebman for the participation margin, Meyer and Rosenbaum for the decomposition across the 1984 to 1996 window, and the Hoynes, Miller, and Simon working paper for the infant-health channel.

A practical method for keeping this material straight is to build the file expansion by expansion, with each public law number anchoring its own set of notes: the parameters before, the parameters after, the effective dates, the JCT score, and the studies that later used that round as variation. Researchers who maintain that structure find that later questions answer themselves, because most disputes about the provision reduce to which expansion changed which parameter in which year. To keep your statute notes, citations, and case chronologies together free on VaultBook is one way to hold the growing file in a single searchable place as the rounds accumulate.

For the legislative history in one place, the Congressional Research Service’s report on the credit’s legislative history, catalogued as R44825, collects the public law citations, effective dates, and parameter tables across all the expansions, and it remains the fastest way to confirm which round changed which threshold. The Joint Committee on Taxation’s revenue estimates for each vehicle, published as JCX documents, show how scorekeepers divided each expansion’s cost between forgone revenue and outlay, a division that determines how the provision interacts with budget enforcement rules. Reading the conference reports on the reconciliation acts repays the effort: the 1990 and 1993 reports explain the family-size tiers in the managers’ own language, including the reasoning that larger households needed larger supplements.

For numbers, the Internal Revenue Service’s Statistics of Income program publishes annual tabulations of EITC claims by income band and family size, and the Treasury’s Agency Financial Report carries the payment-integrity estimates that underpin the error-rate discussion. Researchers studying effects before journal publication should watch the National Bureau of Economic Research working paper series, where papers like the Hoynes, Miller, and Simon study circulate months or years before print. The habit to cultivate is dating every figure: because the parameters move with almost every tax bill, an undated EITC statistic is a trap, and the careful researcher attaches a tax year to every number.

A final research tip concerns the reconciliation process itself. Because the 1990, 1993, and 2001 expansions moved as reconciliation bills, their legislative histories live in budget committee reports and conference agreements rather than in the hearings of a single authorizing committee. The budget resolution’s reconciliation instructions set the dollar targets, the tax-writing committees filled in the policy, and the conference report reconciled the chambers’ versions. Tracing a single parameter change therefore means following it through three documents: the instruction, the committee markup, and the conference agreement. Staff who master that paper trail can answer almost any question about why a given threshold sits where it does.

What does the EITC cost the Treasury each year?

The Treasury bears the cost in two forms: forgone revenue when the benefit offsets income-tax liability, and outlays when the refundable portion exceeds liability. For tax year 2012, the credit totaled about 64.1 billion dollars, of which about 56.2 billion was paid as refunds, according to the Congressional Research Service.

Those are dated, attributed figures from the era of this article’s 2013 publication, and they illustrate the provision’s fiscal scale after the 1990s and 2009 expansions. The trajectory behind those figures runs through every expansion described above. In 1975 the first-year cost ran about 1.2 billion dollars; a decade later it stood near 2.5 billion; and by 1995, after the 1986, 1990, and 1993 enlargements, Representative Dave Camp told the House the annual cost had reached almost 25 billion dollars, as recorded in the Congressional Record of October 25, 1995. Each enlargement raised the phase-in rates, widened the income bands, or added family-size tiers, and each of those design choices mechanically increased the dollars flowing through the provision. Scorekeepers treat the refundable share as spending and the nonrefundable share as a tax expenditure, which is why budget documents show the program in two places at once. The practical consequence is that debates about the credit’s cost are always debates about its parameters: to spend less, Congress would have to lower rates, narrow bands, or remove tiers, and every expansion since 1986 moved in the opposite direction.

The Congressional Research Service’s tabulations add texture to the totals. For tax year 2012, about 27.8 million returns claimed the credit, for an average benefit of roughly 2,300 dollars per return, with the refundable share accounting for nearly seven-eighths of the total. Averages conceal the distribution: most claimants cluster in the phase-in and plateau ranges with benefits well above the mean, while a long tail of filers in the phase-out range receive small amounts that pull the average down. Budget analysts therefore track the two components separately, since the refundable portion behaves like spending, growing automatically with eligibility, while the nonrefundable portion behaves like a tax cut, worth only as much as the liability it offsets. That dual character is also why the credit appears in debates under two guises: as welfare spending to its critics and as tax relief to its defenders, with the accounting supporting either framing.

One more feature of the cost deserves notice. Because the benefit is indexed for inflation but the underlying wage distribution shifts with the economy, the credit’s cost moves with the business cycle even when Congress does nothing. Recessions push more workers into the eligible income ranges, raising outlays automatically, which makes the provision a modest automatic stabilizer: it spends more precisely when low-wage households need it most. Expansions layer policy-driven growth on top of that cyclical movement, which is why the cost curve ratchets upward in steps rather than climbing smoothly.

Comparisons across time require one more adjustment. The dollar figures above are nominal, meaning inflation flatters the growth: 25 billion dollars in 1995 bought more than 25 billion would a decade later. Analysts who convert the series to constant dollars find the same upward ratchet, because the expansions outpaced inflation by wide margins, but the real growth is less steep than the nominal figures suggest. The Congressional Research Service’s tables present both nominal and inflation-adjusted series for exactly this reason, and careful readers should check which column a cited figure comes from before comparing across decades.

How is the credit claimed on a tax return?

Eligible workers claim the benefit on their annual federal income tax return, typically by completing the EITC worksheet and, when qualifying children are involved, attaching Schedule EIC. The Internal Revenue Service computes the amount from reported earnings and either reduces tax owed or adds the excess to the refund.

The agency applies the phase-in rate, plateau, and phase-out for the filer’s family size, and no separate application to a benefits agency is required. That simplicity of claiming is both the design’s genius and its vulnerability. Because the return relies heavily on taxpayer self-reporting, the agency pays first and verifies later through document matching, audits, and, after the 2015 PATH Act’s refund hold, a February delay that allows wage-report cross-checks before money moves. Filers must substantiate the qualifying-child tests on examination: the relationship, age, residency, and joint-return conditions described in the statute, plus the earned-income and investment-income limits. Paid preparers handle a large share of EITC returns, which concentrates both compliance assistance and, in some documented cases, error generation in the preparation industry. The claiming process therefore mirrors the program’s larger bargain: maximum accessibility through the tax system, purchased at the price of limited up-front verification.

Timing completes the picture. Before the PATH Act’s hold took effect for the 2017 filing season, the agency generally issued refunds within weeks of accepting a return, which meant EITC dollars reached households early in the calendar year, when many used them to catch up on bills or make large purchases. The February 15 hold shifted that calendar, and the delay fell hardest on the households with the least savings to bridge the gap. Preparers adapted by warning clients early, and some refund-anticipation products repriced around the new timeline. The episode illustrates a recurring tension in the program’s administration: every control that reduces erroneous payments also slows or complicates payment for the compliant majority, and the statute offers no formula for striking the balance. The claiming process, simple on its face, thus carries the full weight of the provision’s central trade-off between access and accuracy.

For filers selected for examination, the process becomes document-intensive. The agency typically asks for school or medical records showing the child’s address, childcare provider statements, or lease agreements to substantiate the residency test, along with proof of the relationship where it is not obvious from the return. Many eligible claimants struggle to assemble such records, particularly in informal caregiving arrangements, which is one reason taxpayer advocates have long argued that the audit process itself deters legitimate claims. The complexity that generates improper payments on one side generates nonparticipation on the other, and both flow from the same statutory demand: prove, on paper, the facts of family life.

Frequently Asked Questions

Q: What does the earned income tax credit actually do?

The earned income tax credit supplements the wages of low-income workers through the federal income tax system. For each dollar of earnings up to a threshold, the tax code adds a percentage as a credit; that percentage depends on how many qualifying children the filer claims. The payment grows with earnings through the phase-in range, holds steady at a maximum across a plateau, then shrinks as income rises further until it disappears. Because the credit is refundable, workers whose earnings are too low to owe income tax still receive the full amount as a refund. The design rewards work directly: only people with earned income can claim it, and within the phase-in range every additional dollar of wages increases the benefit. Policymakers created it to offset the payroll-tax burden on low-wage earners and to make employment more financially rewarding than nonemployment. Over successive expansions it grew into one of the largest antipoverty transfers for working households in the United States.

Q: Who created the earned income tax credit?

Congress created the earned income tax credit in 1975, and President Gerald Ford signed it into law as part of the Tax Reduction Act of that year. The original provision was modest: a temporary offset for the payroll taxes paid by low-income working families, conceived partly as an alternative to raising the minimum wage and partly as a way to blunt the regressivity of Social Security taxes. Lawmakers made it permanent in 1978, and every major enlargement since then has also come from Congress through tax legislation: the Tax Reform Act of 1986, the reconciliation acts of 1990 and 1993, the 2001 tax act, and the 2009 Recovery Act. No president created the credit by executive action; each expansion required a statute moving through the House Ways and Means and Senate Finance committees. The bipartisan character of that history is notable, with Democratic and Republican Congresses alike voting to enlarge the benefit across four decades.

Q: Was the earned income tax credit an alternative to welfare?

In its original conception, yes. The 1975 credit was designed as a work-based alternative to traditional cash welfare: instead of paying benefits to people outside the labor force, the government would supplement the earnings of people already working. Supporters argued that tying aid to employment avoided the work disincentives they associated with programs like Aid to Families with Dependent Children, because the benefit grew rather than shrank as earnings rose through the phase-in range. That framing persisted through the expansions. When Congress overhauled cash welfare in 1996, creating Temporary Assistance for Needy Families with work requirements and time limits, the credit’s role as the work-rewarding complement became explicit: TANF pushed recipients toward jobs, and the EITC made those jobs pay more. The two were never substitutes in a strict sense, since the credit cannot help people with no earnings, but the alternative-to-welfare idea shaped its political appeal.

Q: Who qualifies for the earned income tax credit?

Qualification turns on earnings, income, and household composition. A claimant must have earned income from wages or self-employment and must have adjusted gross income below the statutory ceiling for the filer’s family size. Investment income must fall below a threshold that Congress indexes for inflation, which keeps the benefit targeted at wage earners rather than asset holders. Filers with children must satisfy four qualifying-child tests covering relationship, age, residency, and joint-return filing. Workers without qualifying children face a separate, narrower set of rules: they must be between 25 and 64 years old, cannot be claimed as a dependent on another return, and must have lived in the United States for more than half the year. All claimants need valid Social Security numbers and, with limited exceptions, must file as something other than married filing separately. Meeting every applicable test is necessary; failing any one of them disqualifies the claim.

Q: Is the earned income tax credit refundable?

Yes. Refundability is the feature that lets the credit reach workers whose earnings are too low to generate income-tax liability. A nonrefundable credit can only reduce tax owed to zero, which would leave the poorest workers with nothing. The earned income credit instead pays out any excess as a refund: if the computed benefit is larger than the filer’s income-tax liability, the Treasury sends the difference as a payment. Budget scorekeepers reflect this by treating the nonrefundable portion as forgone revenue and the refundable portion as an outlay, which is why the program appears in both the revenue and spending sides of federal budget documents. Refundability also explains the provision’s scale; for tax year 2012, about 56.2 billion of the 64.1 billion dollars in total credits was paid as refunds, according to the Congressional Research Service. Without refundability, the EITC would be a middle-income tax cut rather than an antipoverty program.

Q: Does the earned income tax credit increase employment?

The best available evidence says yes, principally by drawing people into the labor force. Nada Eissa and Jeffrey Liebman’s 1996 Quarterly Journal of Economics study found that the 1986 expansion raised labor force participation among single women with children by up to 2.8 percentage points relative to single women without children, while leaving the hours of those already working essentially unchanged. Bruce Meyer and Dan Rosenbaum’s 2001 Quarterly Journal of Economics study, covering 1984 to 1996, attributed a large share of the rise in work among single mothers to the credit and other tax changes, after accounting for welfare reform and economic conditions. The mechanism is straightforward: by adding a percentage to each dollar of earnings in the phase-in range, the provision raises the effective wage and makes employment more attractive than nonemployment. Researchers find the effect concentrated on the decision to work at all rather than on hours worked.

Q: Why does the earned income tax credit have high improper payments?

The high rate reflects verification difficulty, not primarily fraud. For fiscal year 2023, the improper-payment rate was 33.5 percent, or 21.9 billion of 65.4 billion dollars claimed, per the Treasury Agency Financial Report as reviewed by the Treasury Inspector General for Tax Administration, the highest among large federal programs. The paymentaccuracy.gov scorecard attributes roughly 20.57 billion dollars of overpayments to the inability to authenticate qualifying-child eligibility, meaning the relationship and residency tests, plus misreported income. The Internal Revenue Service administers the benefit on taxpayer self-reporting with few databases capable of confirming, before payment, which child lived with which adult for more than half the year. In households with shared custody or multigenerational caregiving, those facts are genuinely ambiguous, and the statute’s intricate tie-breaker rules are easy to misunderstand. Complexity in complex households generates most erroneous dollars, which is why analysts describe the problem as a design flaw rather than a crime wave.

Q: Does the earned income tax credit have a marriage penalty?

It can. The credit’s phase-out thresholds for joint filers have historically been less than double the thresholds for single filers, so two earners who marry can face a combined benefit smaller than the sum of what each could claim single. The penalty bites hardest when both spouses work and their combined income pushes the household further into the phase-out range than either income would alone. Congress has twice acted to soften it. The Economic Growth and Tax Relief Reconciliation Act of 2001, Public Law 107-16, raised the income level at which married couples filing jointly begin losing the benefit, supplying the first dedicated marriage-penalty relief in the provision’s history. The American Recovery and Reinvestment Act of 2009, Public Law 111-5, lifted the joint-filer thresholds further. Neither change eliminated the penalty entirely; both narrowed it. Whether a given couple faces a penalty or a bonus depends on the spouses’ earnings split and family size.

Q: How does the earned income tax credit phase-in rate work?

The phase-in rate is the percentage the tax code adds to each dollar of earnings at the bottom of the income scale. A filer with one qualifying child faces a 34 percent rate, so each dollar earned adds thirty-four cents of credit until earnings reach the end of the phase-in range. The rates rise with family size: 40 percent for two children and 45 percent for three or more, reflecting the judgment that larger households need larger supplements. Workers without qualifying children phase in at 7.65 percent, a much smaller supplement. The phase-in design is what makes the provision a work incentive rather than a flat grant: within this range, earning more always increases the benefit, so the effective wage exceeds the market wage by the phase-in percentage. Once earnings pass the phase-in ceiling, the credit stops growing and holds at its maximum across the plateau until the phase-out begins.

Q: What is the earned income tax credit plateau?

The plateau is the flat stretch of the benefit schedule where the credit holds at its maximum value. After earnings carry a filer through the phase-in range, the computed benefit reaches its ceiling; across the plateau, additional earnings neither increase nor decrease the payment. The plateau exists because Congress set separate income points for the end of phase-in and the start of phase-out, creating a band in which the maximum benefit is constant. Its width varies by family size and filing status, and lawmakers have adjusted it in successive expansions. Conceptually, the plateau marks the range where the provision functions as a pure income supplement: the work incentive of the phase-in has done its job, and the phase-out has not yet begun to claw the benefit back. Filers whose earnings fall anywhere inside this band receive the same maximum amount, which is why the schedule resembles a trapezoid when drawn on a graph.

Q: What are the four qualifying-child tests for the earned income tax credit?

Section 32 conditions the family-size benefit on four tests that a child must satisfy. The relationship test requires the child to be the claimant’s son, daughter, stepchild, foster child, sibling, half-sibling, or a descendant of any of those. The age test requires the child to be under 19 at year’s end, under 24 if a full-time student, or any age if permanently and totally disabled. The residency test requires the child to have lived with the claimant in the United States for more than half the year, which is the test that generates most disputes in shared-custody and multigenerational households. The joint-return test bars the credit when the child files a joint return, except when the return was filed only to claim a refund of withheld tax. All four must be met for the same child, and tie-breaker rules assign the child to one claimant when more than one person qualifies.

Q: Why does the earned income tax credit limit investment income?

The investment-income limit keeps the benefit targeted at wage earners. Congress designed the credit as a supplement to labor earnings and an offset to payroll taxes, not as a general subsidy for people with substantial assets. Without a cap, a filer with large investment returns but modest wages could qualify for a benefit meant for the working poor. The statute therefore disqualifies anyone whose investment income, meaning interest, dividends, capital gains, and similar returns, exceeds a threshold that Congress indexes for inflation. The limit is deliberately set low enough to screen out asset-rich claimants while leaving ordinary savers unaffected. It works alongside the earned-income requirement to define the eligible population from two sides: the claimant must have enough labor income to enter the schedule, and must not have so much capital income as to fall outside the program’s purpose.

Q: How did the 1986 act change the earned income tax credit?

The Tax Reform Act of 1986, Public Law 99-514, delivered the credit’s first major enlargement. It raised the credit rate to 14 percent, substantially increasing the supplement per dollar of earnings, and it permanently indexed the income thresholds to inflation. Indexing was the quieter but more durable change: before 1986, inflation steadily eroded the real value of the benefit by pushing workers into higher nominal income bands, and only new legislation could restore it. After 1986 the parameters adjusted automatically, preserving the benefit’s purchasing power between congressional actions. The expansion fit the broader reform bargain, in which lawmakers cutting marginal tax rates wanted to show that low-wage earners were protected. Economists later treated the 1986 change as a natural experiment: Eissa and Liebman’s landmark 1996 study measured the employment response by comparing single mothers, who gained from the expansion, with single childless women, who did not.

Q: How did the 1990 and 1993 expansions change the earned income tax credit?

The Omnibus Budget Reconciliation Act of 1990, Public Law 101-508, introduced family-size differentiation by splitting the schedule into a one-child formula and a two-or-more-child formula, so larger families received larger benefits. The change recognized that additional children raise household need without raising wages, and it phased in across 1991 through 1994. The Omnibus Budget Reconciliation Act of 1993, Public Law 103-66, then delivered the largest single expansion in the program’s history, raising benefit levels substantially across the board and, for the first time, extending eligibility to workers without qualifying children. The 1993 childless benefit was intentionally small, structured as a modest payroll-tax offset rather than a full antipoverty payment. Together the two laws transformed the credit from a flat, family-blind supplement into a tiered system scaled to household size, and the cost trajectory reflected it: by 1995 the annual price had reached almost 25 billion dollars.

Q: Why is the childless-worker earned income tax credit so much smaller?

The disparity traces to the terms Congress set when it first extended eligibility to childless workers in 1993. Lawmakers conceived that extension as a narrow payroll-tax offset, not as an antipoverty benefit comparable to the family credit. The phase-in rate of 7.65 percent mirrors the combined employee share of payroll taxes, so the small benefit roughly cancels the payroll-tax bite on the first dollars of earnings. The schedule also phases out at a much lower income level than the family schedules, and eligibility is further restricted to claimants aged 25 to 64 who are not dependents and who lived in the United States for more than half the year. The result is a benefit that reaches relatively few dollars per claimant: an analysis of 2012 data found less than 3 percent of total EITC dollars flowing to childless workers. Congress debated enlarging it before this article’s 2013 date but left the 1993 parameters essentially intact.

Q: How does the earned income tax credit differ from a tax deduction?

A deduction and a credit reduce tax liability through different arithmetic, and the difference determines who benefits. A deduction subtracts from taxable income, so its value equals the deduction amount multiplied by the filer’s marginal tax rate; a thousand-dollar deduction saves 150 dollars for a filer in the 15 percent bracket and nothing for a filer with no taxable income. A credit subtracts directly from the tax owed, dollar for dollar, so a thousand-dollar credit is worth a thousand dollars to any filer who owes at least that much. The earned income credit goes further because it is refundable: filers who owe no income tax still receive the full amount as a payment. That makes the credit far more valuable to low-income workers than any deduction could be, since deductions are worth least to the people with the lowest marginal rates. The distinction explains why antipoverty policy favors credits over deductions.

Q: What did the 2015 PATH Act do to earned income tax credit refunds?

Two years after this article’s 2013 publication date, Congress changed the timing of EITC refunds. The Protecting Americans from Tax Hikes Act of 2015, Division Q of Public Law 114-113, enacted December 18, 2015, required the Internal Revenue Service to hold refunds on returns claiming the earned income credit or the Additional Child Tax Credit until February 15, effective for the 2017 filing season. The purpose was verification: the delay gave the agency time to match return information against employer wage reports before releasing money, addressing the improper-payment problem at the point of disbursement. The law changed no eligibility rule and reduced no benefit amount; it altered only when the government pays. For households that rely on early-year refunds to cover winter expenses, the hold imposed a real cost in delayed cash. The provision illustrates the post-2013 policy response to payment error: slower refunds in exchange for better verification.

Q: How do researchers know the earned income tax credit increases employment?

Researchers exploit the credit’s expansion history as a series of natural experiments. Each enlargement changed work incentives for eligible households while leaving similar ineligible households untouched, creating a treatment group and a comparison group. Eissa and Liebman’s 1996 study, for example, compared single mothers with children against single women without children across the 1986 expansion, attributing the divergence in their employment trends to the policy because the two groups otherwise faced similar labor markets. This difference-in-differences method rules out many alternative explanations: a general economic boom would lift both groups, so only a policy affecting one group can explain their divergence. Meyer and Rosenbaum applied the same logic across the longer 1984 to 1996 window, controlling separately for welfare reform and economic conditions. The consistency of results across different expansions, different decades, and different research teams is what gives the employment finding its credibility.

Q: How does the earned income tax credit interact with TANF and SNAP?

The three programs divide the antipoverty labor. Temporary Assistance for Needy Families, created by the 1996 welfare reform law, provides time-limited cash assistance oriented toward moving recipients into employment. The Supplemental Nutrition Assistance Program addresses food needs for low-income households regardless of work status. The earned income credit supplements the earnings of people already working, raising the payoff to employment without requiring a separate welfare application. A household can receive all three simultaneously, and their eligibility rules reach overlapping but distinct populations. The key boundary is that the credit requires earnings: it offers nothing to people with no labor income, while TANF and SNAP can reach nonworkers. In policy terms the programs are complements rather than substitutes. TANF pushes recipients toward jobs, SNAP covers nutritional needs along the way, and the EITC makes the resulting wages go further, which is why analysts describe the American safety net for working families as a layered system.

Q: Can married couples claim the earned income tax credit?

Yes. Married couples are fully eligible, and the statute provides a separate schedule of phase-in ranges, plateaus, and phase-out thresholds for joint filers. In most cases the couple must file a joint return to claim the benefit; married filing separately generally disqualifies the claim, with only narrow exceptions. The complication is the marriage penalty: because joint-filer phase-out thresholds have historically been less than double the single thresholds, two earners who marry can receive less combined than they would have received as two single filers. Congress addressed this twice, in the 2001 tax act and the 2009 Recovery Act, by raising the income level at which joint filers begin losing the benefit. Whether marriage helps or hurts a particular couple’s credit depends on how earnings are split between spouses and how many qualifying children they claim. Couples with one earner typically face no penalty; two-earner couples near the phase-out range may.