💬 Discussion
Two Greek thinkers examine a family purse, child's toy and tax ledger with coins paid from a public fund.

A Fully Refundable Credit

The child tax credit pays up to 2,200 dollars a child, but families who earn too little to owe income tax receive only part of it or nothing [1]. Democrats have long proposed paying it in full to every eligible child, as Congress did for one year in 2021, while Republicans want to keep it tied to work. The budget law of July 2025 raised the credit and left the earnings test in place.

  • 19 million children live in households that receive less than the full credit [1]
  • 3.7 million children were kept out of poverty in December 2021, when the credit was paid in full each month [8]
  • 13.6 percent of the refundable part of the credit was paid improperly on returns for tax year 2023 [13]

The proposal

Pay the full child tax credit to every eligible child, including those whose parents earn too little to owe income tax.

The debate

Democrats argue that the poorest children are the ones left out and that the full credit cut child poverty and hunger when it was tried. Republicans answer that a credit paid without earnings becomes welfare, weakens the reason to work and should stay linked to paying tax. On 1 August 2024 a House bill that would have widened the refundable part fell short of the votes needed to proceed in the Senate. On 30 June 2025 a Democratic motion to restore the expanded credit failed along party lines.

  • “America’s poorest children, whose families are struggling to make ends meet, receive nothing from this bill.”

    argues for: Michael Bennet, Democrat, Colorado Senate floor, 30 June 2025

  • “But provisions in this bill would depart from fundamental principles that have always guided child tax credit expansions. This includes that the credit be tied to work and linked to the payment of tax, whether that is income or payroll taxes.”

    argues against: Chuck Grassley, Republican, Iowa Senate floor, 1 August 2024

The analysis

A Fully Refundable Credit

72% of the weight is for, 28% against.

An argument's weight combines three questions: how big the effect is, how certain it is and how much the good at stake matters. The larger the slice, the more the argument weighs.

For

Against

A fully refundable child tax credit is, on the current analysis, a good proposal, and the reason is almost entirely where the money goes.

Very likely better

Very likely better: Under the assumptions used, the proposal brings more than it costs in more than 99 of 100 model runs.

 Overall effect: very large

How strongly the proposal works overall, in five steps from very small to very large, comparable across all debates. The ring is always full, even when the overall effect is small; this step says how large it is.

Like the Democrats, the analysis favors the full credit, and for their main reason: the money reaches the poorest families with children. Almost all of the weight is money changing hands from the federal budget to those households, and at the bottom of the income scale a dollar does far more good. On top come gains the debate barely mentions: children who do better in school, health and earnings two decades later. The Republican objection that parents would stop working is the one claim tested in practice, and when the full credit ran for six months the drop did not appear. The largest item against is the cost to a budget in deficit, which is the same money counted at a lower weight.

In detail

All arguments

Arguments for3 arguments

Money where there is none

15points

Nineteen million children live in families that receive less than the full credit because their parents earn too little. Full refundability sends about 25.5 billion dollars a year to precisely those households. This is the lowest-income group any American tax provision reaches.

Value 5 · Household budgetsImpact 5.1Plausibility 6
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is money in household budgets, priced at the middle of the scale like every other dollar. What it buys in these households is not a marginal comfort: shoes that fit, a bus fare to a job interview, a month's rent that does not become an eviction. That the recipients are the poorest families in the country does not raise the value here; it raises the weight the dollar carries, which is counted in the Impact. Booking their poverty in both places would price the same fact two ways. What the money does to the children over their lives is a separate stream and is counted separately. The value is the middle of the scale, because the stream is a transfer of money and the budgets it lands in are priced in the Impact.

Impact

About 19 million children receive less than the full credit because the household's earnings are too low to claim it [1]. The credit is worth up to 2,200 dollars a child and the shortfall varies from a few hundred dollars to the whole amount; an average of 1,340 dollars is used, in a range from 800 to 2,000. Across 19 million children that is 25.5 billion dollars a year. That is close to a published ten-year estimate of the same change, which averages 27.2 billion dollars a year [10]. Where it lands is what makes the argument. Most of these families sit in the bottom fifth of American incomes, where this site counts a dollar at two and a half times its worth at median income. Not all of them do: a single parent with two children reaches the full credit only at about 34,150 dollars of earnings and a married couple at about 42,200 dollars, which lies in the second fifth, where a dollar counts one and a half times [1]. How the money divides between the two fifths is not published; four fifths in the bottom fifth and one fifth in the second is set here, because the shortfall per child is largest where earnings are lowest. That gives a weight of 2.3 and a Normalised Impact of 5.06 on the scale of this evaluation, against 2.2 for the budget that pays. No part of it reaches households above the phase-in, because the measure changes nothing for anyone already receiving the full amount. The Impact is the largest in this debate by a wide margin, and it is large because of where the money goes rather than how much of it there is.

▸ Show calculation ▾ Hide calculation
Children receiving less than the full credit [1] 19 million children
× Average shortfall per child Setting, range 800 to 2,000 dollars: the credit is worth up to 2,200 a child and the shortfall runs from a few hundred to the whole amount; a published ten-year estimate of the same change averages 27.2 billion dollars a year; take-up among families that do not otherwise file pulls the figure down [1][3][10] 1,340 dollars a year 25.5 billion dollars a year
÷ In euro-equivalent at 1.16 dollars to the euro, 22 billion euro-equivalent (range 13 to 33 billion euro-equivalent, 15 to 38 billion dollars) 1.16 dollars to the euro 22 billion euro-equivalent
× Four fifths of it in the bottom fifth of incomes, where a dollar counts 2.5 Setting: how the money divides between the fifths is not published; the shortfall per child is largest where earnings are lowest, so most of the money is set in the bottom fifth [1] 22 × 80 % × 2.5 44 billion euro-equivalent
+ One fifth of it in the second fifth, where a dollar counts 1.5 a single parent with two children reaches the full credit only at about 34,150 dollars of earnings, a married couple at about 42,200 dollars, which lies in the second fifth [1] 22 × 20 % × 1.5 6.6 billion euro-equivalent
= Weighted sum, an average weight of 2.3 44 + 6.6; 80 % × 2.5 + 20 % × 1.5 = 2.3 50.6 billion euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 5.06
Score 5.06 Impact × 5 Value × 6 Plausibility ÷ 10 = 15 of 100

Plausibility

The mechanism is a formula: removing an earnings test pays the credit to households that fail it, and who those households are is known from tax and survey data rather than estimated [1]. The counterfactual is current law, in which the refundable portion phases in at fifteen cents per dollar of earnings above 2,500 dollars, up to 1,700 dollars a child. What is estimated is the average shortfall, and the 2021 expansion supplies a check on it: that year the credit was made fully refundable and the payments that went out are on record, so the order of magnitude is observed rather than modelled [2]. Where this measure differs from 2021 is that it does not raise the credit amount, only its reach, so it costs a fraction of what that expansion did. The confounder that matters is take-up: a family that owes no tax may not file, and the 2021 experience showed that reaching the very poorest requires outreach rather than only eligibility. That is unresolved and it works against the figure, which is why the range runs down to 800 dollars. Reverse causation does not arise. The Plausibility is at the top of what a projection can carry: the rule is arithmetic, the 2021 payments are on record, and only take-up is open.

Evidence: Projection · plausibility at most 6 Evidence of effect: Definitional

Children who do better later

2.2points

Money reaching a household in a child's early years shows up two decades later in school results, health and earnings. This is among the better-established findings in economics, measured across several American programs with clean comparisons. It is a separate gain from the money itself.

Value 8 · Life chancesImpact 0.6Plausibility 5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is what a child becomes: whether they finish school, whether they are healthy at thirty, what they can earn. This site places that in the class it uses for subsistence and life chances, above economic output and below life and health. It is genuinely separate from the cash in the argument above, which is what the household gains this year; this is what the child gains over a lifetime, and the two are different goods reaching different people. What is priced is the child's own later position, not the tax the adult they become will pay, because counting both would be the same person twice. The value sits in the upper part of the scale, at the level this site uses for subsistence and life chances.

Impact

About 25.5 billion dollars a year reaches households with young children who currently receive little or nothing. What that does over a lifetime has been measured for American income-support programs [6][7], and the returns are consistently large: children in families that received more early income complete more school, report better health as adults and earn more. A return of twenty-five percent of the transfer, expressed as the annual value of the later gains, is used here, in a range from ten to fifty percent: a Normalised Impact of 0.55 on the scale of this evaluation, in a range from 0.22 to 1.1. The figure is booked against the year the money is paid, as the value of what that year's transfer does over the children's lives; the gains arrive two decades later, and nothing is deducted for that wait because nothing in this evaluation is discounted. The width of that range is not evasion: the estimates differ by a factor of five depending on the program, the age at which the money arrives and how long it lasts, and a credit paid every year to every poor family sits at neither end of that literature. What the figure does not include is the effect on the parents, which is counted nowhere in this evaluation. The Impact is about a tenth of the cash it accompanies, which is the ordinary proportion once a lifetime of gains is expressed as an annual figure.

▸ Show calculation ▾ Hide calculation
Money reaching households with young children [1] from the argument above: 25.5 billion dollars a year at 1.16 dollars to the euro 22 billion euro-equivalent
× Later gains in schooling, health and earnings, as an annual value Setting, range 10 to 50 percent: estimates for American income-support programs differ by a factor of five depending on the program, the child's age and how long the support lasts [6][7] 25 % 5.5 billion euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 0.55
Score 0.55 Impact × 8 Value × 5 Plausibility ÷ 10 = 2.2 of 100

Plausibility

The direction here is as well established as anything in this field and the size is not. The counterfactual in the studies that carry it is children in comparable families who received less, separated by something other than the families themselves: the county-by-county introduction of food stamps between 1961 and 1975 [6], or whether a mother's application to one of the first cash-welfare programs was accepted or turned down [7]. The Urban Institute's simulation of this credit [4] shows how many children it lifts above the poverty line in the year it is paid, not what that does to them later, and is not used for the size here. The confounder that would otherwise dominate — that families receiving more support differ in ways that also affect their children — is what the rollout timing and the acceptance decision remove, and reverse causation cannot arise, since a child's later earnings do not cause a county's program start date or an administrator's decision. What none of that supplies is the size for this measure. The programs measured were food assistance, health coverage and cash pensions, each delivering something different from an annual tax credit, and the transfer here is smaller per family than most of them. Carrying a return across that gap is the step this evaluation takes on its own. The Plausibility is at the middle: the finding rests on clean comparisons and the number applied here is carried across from programs that are not this one.

Evidence: Converging studies · plausibility at most 8 Evidence of effect: Quasi-experimental · at most 8

Enough food in the house

0.74points

When the credit was made fully refundable in 2021, food insufficiency among families with children fell within weeks of the first payment. By December the monthly payment was keeping 3.7 million children out of poverty. Those are health outcomes, not accounting ones.

Value 9 · HealthImpact 0.1Plausibility 5.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is the health of children and their parents in the months when money is short: meals skipped so a child can eat, a prescription not filled, the sleeplessness of an eviction notice. This site places that in the class it uses for life and health, one step below the top of it because what is lost here can be regained. It is separate from the later life chances counted above, which are about what the child becomes; this is about what happens to them now. It is also separate from the cash, because a household that goes without food is losing something the money is only the means to. The value sits one step below the maximum: the stream is health, and health that can be regained.

Impact

The 2021 expansion made the credit fully refundable and paid it monthly, and the effects appeared within weeks: food insufficiency among households with children fell by about a quarter after the first payment [9], and the sixth payment kept 3.7 million children out of poverty in December of that year [8]. This measure is a fraction of that one — it removes the earnings test without raising the credit or paying it monthly — so a fraction of the effect is what is claimed. About 1.5 million children lifted above the poverty line is used, in a range from 0.8 to 2.5 million. What that is worth in health is set at 0.025 quality-adjusted years a year per child and parent affected, covering roughly 1.5 million children and their households: 37,500 quality-adjusted years a year, valued at the figure this site uses for a healthy life year. The result is a Normalised Impact of 0.15 on the scale of this evaluation. The Impact is small against the cash, which is the honest proportion: most of what money does for a poor household is not measurable as health.

▸ Show calculation ▾ Hide calculation
Children lifted above the poverty line Setting, range 0.8 to 2.5 million: the 2021 expansion kept 3.7 million children out of poverty in December 2021, but it also raised the credit and paid it monthly, which this measure does not [8] 1.5 million children
× Quality-adjusted years gained, children and their households Setting, range 0.01 to 0.06: enough food in the house, a prescription filled, an eviction that does not happen 0.025 a year each 37,500 quality-adjusted years
× Value of the years the value of a healthy life year used across this site 40,000 euro-equivalent each 1,500 million euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 0.15
Score 0.15 Impact × 9 Value × 5.5 Plausibility ÷ 10 = 0.7 of 100

Plausibility

The 2021 expansion is the closest thing to a test this measure will ever have, and it is a good one for direction and a poor one for size. The counterfactual is the months immediately before the first payment, on households with children, measured by a national survey running continuously through the period, and the drop in food insufficiency followed the payment date rather than any trend [9]. The confounder that matters is that 2021 contained several other pandemic-era supports, so isolating the credit from the rest requires the timing to do the work, and it largely does — the payments started in July while the other supports did not change that month. Reverse causation does not arise. What does not transfer is the size: the 2021 credit was worth up to 3,600 dollars a child and was paid monthly, and this measure is neither, so the effect here should be considerably smaller than the fraction assumed. The valuation of a year of adequate food as health is a construction of this evaluation rather than anything measured. The Plausibility is above the middle: the effect was observed on the right population at the right moment, and the version of the policy that produced it was much larger than this one.

Evidence: Single precedent · plausibility at most 7 Evidence of effect: Quasi-experimental · at most 8

Arguments against3 arguments

Twenty-five billion dollars a year

6.6points

The money comes from the federal budget, which is already borrowing 5.8 percent of national output. Full refundability is permanent by nature: there is no year in which the earnings test comes back on its own. The cost grows with the credit as it is indexed.

Value 5 · Public financesImpact 2.2Plausibility 6
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is federal money, priced at the middle of the scale. A refundable credit is spending in the shape of a tax provision and this site treats it as such. Public money is not weighted differently from private money; the whole of what makes this measure worth considering is that a dollar is worth more where it lands than where it came from, and that difference is counted on the receiving side. The deficit is not a second harm on top of the spending, because borrowing shifts who pays and when and that shift is inside the dollar figure. The value is the middle of the scale, the level this site uses for public money whatever it is spent on.

Impact

The measure pays out about 25.5 billion dollars a year, the same figure that appears as a gain on the receiving side, weighted here at the standard one for public money rather than at the 2.3 that applies where it lands; it follows from the 19 million children [1] and the average shortfall set there. A published estimate of the same change puts it at 272.4 billion dollars over the ten years from 2027 to 2036, about 27.2 billion dollars a year, a little above the figure used here and well inside the range [10]. The range is the same 15 to 38 billion dollars as on the receiving side, because the two sides move together: if take-up is lower than assumed, both the cost and the benefit shrink at once, and if the average shortfall is larger, both grow. Over the six years counted here the total is about 150 billion dollars. Nothing offsets it inside this argument; what the money buys is counted in the three arguments above. Two things would raise it over time and are not in the figure: the credit is indexed, and the number of children in low-earning households moves with the economy. The Impact is the second largest in this debate and is the mirror of the largest: the same money, seen from the side that pays it.

▸ Show calculation ▾ Hide calculation
Credit paid to children who currently receive less than the full amount Setting, range 800 to 2,000 dollars a child, the same as on the receiving side; cost and benefit move together [1][10] 19 million children × 1,340 dollars 25.5 billion dollars a year
÷ In euro-equivalent at 1.16 dollars to the euro, 22 billion euro-equivalent (range 13 to 33 billion euro-equivalent, 15 to 38 billion dollars) 1.16 dollars to the euro 22 billion euro-equivalent
× Weight of a dollar in the federal budget the standard weight for public money on this site, against 2.3 on the receiving side 1.0 22 billion euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 2.2
Score 2.2 Impact × 5 Value × 6 Plausibility ÷ 10 = 6.6 of 100

Plausibility

That paying a credit to households currently excluded costs money is arithmetic on a formula, not a prediction, and the counterfactual is current law [1][3]. What is estimated is how much is claimed, which depends on take-up among families that may not file a return at all — the same open question that limits the benefit figure, working in the same direction. The 2021 year of full refundability supplies an observed check: the payments that went out are on record, and scaling them down for the smaller credit and the absence of monthly delivery gives an order of magnitude consistent with the figure here [2]. The confounder that could push the cost above this figure is improper payments, which have historically been higher for the refundable portion of this credit than for most federal programs; part of that money reaches households not entitled to it and is counted as a separate argument below rather than added here. Reverse causation does not arise. The Plausibility is at the top of what a budget projection can carry: the rule is certain and only the claiming rate is estimated.

Evidence: Projection · plausibility at most 6 Evidence of effect: Definitional

Parents who stop working

0.24points

A credit paid without an earnings test removes the reason to earn the first dollars, which is the standard objection and a serious one. It was simulated at 1.5 million workers leaving the labor force. Then the policy ran for six months in 2021 and the effect was looked for and not found.

Value 6 · OutputImpact 0.2Plausibility 2
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is output that does not happen because someone who would have worked does not. It belongs to the class this site uses for economic systems and prosperity. What is counted is the production, net of the time the parent gains back, which for a parent of young children is not worthless — time at home is not idleness, and treating a withdrawal from paid work as a total loss would overstate it. Nothing is counted here for what the parent's absence from work does to the child, in either direction, because the evidence points both ways and pricing it would be guesswork. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

Impact

The simulation that framed this objection put the number at about 1.5 million parents leaving the labor force under a fully refundable credit paid monthly at a higher amount than this measure proposes [12]. What is booked here is not that figure but the expected one, which after the 2021 evidence is much smaller: 200,000 parents, in a range of 60,000 to 700,000 if a withdrawal happens at all; whether it happens is a question of Plausibility. The 700,000 is the upper edge of what the 2021 measurements could not rule out rather than the center of what they found. Each represents about 29,000 dollars of annual earnings, and what society loses is the production behind that net of the time the parent gains back, for which forty percent is used. The result is about 2.3 billion dollars a year. This measure removes the earnings test only; it does not raise the credit or pay it monthly, both of which the 2021 version did and both of which would strengthen any withdrawal effect. The Impact is a tenth of the cash it argues against, and it is stated at the expected size rather than the simulated one.

▸ Show calculation ▾ Hide calculation
Parents leaving the labor force Setting, range 60,000 to 700,000 if a withdrawal happens at all: 700,000 is the upper edge of what the 2021 measurements could not rule out, against a simulation of 1.5 million for a larger, monthly version of the policy [2][5] 200,000 people
× Annual earnings each 25,000 euro-equivalent (about 29,000 dollars) 5 billion euro-equivalent
× Production net of the time the parent gains back time at home with young children is not idleness, so a withdrawal from paid work is not a total loss 40 % 2 billion euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 0.2
Score 0.2 Impact × 6 Value × 2 Plausibility ÷ 10 = 0.2 of 100

Plausibility

The objection was tested and the test came back empty. From July to December 2021 the credit was fully refundable and paid monthly, and researchers compared employment among families with children against families without, using the fact that only the former received the payments [2]. Employment among recipient parents did not fall relative to the comparison group, in the aggregate or at any income level examined. The confounder that matters is that 2021 was an extraordinary labor market with several other supports running, so a real effect could have been masked by conditions that no longer apply; that is a fair objection and it is why this argument is not scored at zero. Reverse causation does not arise, since eligibility followed from having a child rather than from any employment decision. A second limit is that the 2021 payments were announced as temporary, and parents rarely change jobs or hours for a benefit they expect to end within the year, so the test says less about a permanent credit than its clean design suggests. The simulation that predicted a large withdrawal was published before the measurement and has not been reconciled with it. The Plausibility is low because the effect claimed here was looked for under a proper comparison, in the one period the policy actually ran, and did not appear.

Evidence: Single precedent · plausibility at most 7 Evidence of effect: Quasi-experimental · at most 8 Placement: Effect did not materialise · plausibility 1.5–2

Open: A permanent version would allow a longer comparison in a normal labor market, which is what the 2021 test lacked. Two years of employment data for families just above and just below the current phase-in threshold would settle it and could carry P to 6.

Money that reaches the wrong households

0.24points

The refundable part of this credit has an improper payment rate well above the federal average, because eligibility turns on where a child lived and with whom, which no database records. Removing the earnings test removes one of the few checks that exists. What is paid wrongly is not recovered.

Value 6 · Trust in institutionsImpact 0.1Plausibility 4
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is the standing of a program with the people who fund it, which this site places at the level it uses for the working order of public institutions. What is priced is not the money itself: a payment that goes to a household of similar income is close to a wash in welfare terms, and treating it as a total loss would be wrong. What it costs is the program's defensibility — an error rate that can be quoted is what such measures are attacked with, and repeatedly cut back on. That is a real thing to lose and a modest one. The value sits in the middle of the scale, at the level this site uses for the working order of public institutions.

Impact

The Treasury's report for fiscal year 2025 puts the improper payment rate of the refundable portion at 13.6 percent, measured on tax year 2023 returns, with a sampling margin of plus or minus 2.9 points at 95 percent confidence [13]; the inspector general's review of that reporting is source [11]. The cause is that eligibility depends on which household a child lived in for more than half the year, and no administrative record establishes that. Fifteen percent is used here as a rounded scenario value for the changed program, not as the published figure, inside that margin and in a range from 5 to 25. Applied to 25.5 billion dollars a year that is about 3.8 billion dollars reaching households not entitled to it, in a range from 1.3 to 6.4 billion. Most of that is not lost in welfare terms — the recipients are usually households of similar income, so the dollar is worth about what it would have been worth — and what remains is the cost to the program's standing. Thirty percent of the wrongly paid amount is used for that, in a range from ten to sixty, giving a Normalised Impact of about 0.1 on the scale of this evaluation. Removing the earnings test removes one of the few automatic checks on a claim, which is why the rate is assumed to hold rather than fall. The Impact is the smallest in this debate, because what is priced is the program's defensibility rather than the money.

▸ Show calculation ▾ Hide calculation
Credit paid under full refundability, 25.5 billion dollars a year at 1.16 dollars to the euro [1] 22 billion euro-equivalent
× Share reaching households not entitled to it Scenario setting, range 5 to 25 percent, rounded up from the published rate of 13.6 percent (fiscal year 2025, tax year 2023 returns, plus or minus 2.9 points): eligibility turns on which household a child lived in for more than half the year, which no administrative record establishes [13][11] 15 % 3.3 billion euro-equivalent
× Cost to the program's standing Setting, range 10 to 60 percent: most of the money reaches households of similar income, so the dollar is worth about what it would have been; what is lost is the program's defensibility 30 % 0.99 billion euro-equivalent
÷ Normalised Impact scale of this evaluation 10 billion euro-equivalent a point 0.1
Score 0.1 Impact × 6 Value × 4 Plausibility ÷ 10 = 0.2 of 100

Plausibility

The improper payment rate itself is measured and published annually by the tax administration, so the first link is not in doubt: 13.6 percent for tax year 2023 returns, plus or minus 2.9 points [11][13]. The 15 percent used is a rounded setting for the changed program, not a measurement. Everything after it is this evaluation's construction. Whether removing the earnings test raises, lowers or leaves the rate unchanged has not been studied — a simpler rule can reduce errors as easily as increase them, and that counter-mechanism is real and unanswered. The counterfactual is the current phase-in, under which the same measurement is made. What a program's standing is worth, and whether an error rate actually costs it anything, is not measured anywhere: the price used here is set rather than found, and the whole quantity rests on it. Reverse causation does not arise. The Plausibility is below the middle because the step from a measured error rate to a cost is one this evaluation takes on its own, and the possibility that a simpler rule reduces errors is unanswered.

Evidence: Plausibility · plausibility at most 5 Evidence of effect: Mechanistic · at most 6 Placement: Chain closed, unevidenced · plausibility 4–5

Open: The improper payment rate is published each year. Two years of it after the earnings test is removed would show directly whether a simpler rule raises or lowers errors, and could carry P to 6.

How robust the result is

Under the assumptions used, the proposal brings more than it costs in more than 99 of 100 model runs.

gains more than it costscosts more than it gains

Each dot stands for 200 of the 20,000 model runs.

A dollar reaching these families is valued at 2.3 times a dollar at median income. That is 2.5, the highest weight this site uses, for the four fifths of the money placed in the bottom fifth of incomes, and 1.5 for the fifth placed in the second fifth. How the money divides between the fifths is not published; the split is a setting of this analysis. That weight is the main reason the margin is wide, and with it the measure lies ahead in all runs. At 1.5 throughout, the measure would still come out ahead by a wide margin. The two sides would only weigh the same if a dollar at the bottom of the income distribution were valued at about six tenths of a dollar at median income.

Show the distribution of the model runs

20,000 simulated futures

Very likely better · 100 %

In more than 99 of 100 simulated futures the proposal brings more than it costs.

No effect is known exactly; every argument has a range. The calculation draws a possible size for each argument from its range 20,000 times and counts how often the benefits exceed the costs.

Net effect in points per future. Left of zero the drawbacks outweigh the benefits, right of zero the benefits win. Eight in ten futures fall between +7.0 and +16 points.
Scale and assessment

Scale of this evaluation: Normalised Impact, unitless and calibrated to this topic. For comparison, one point here corresponds to about 10 billion euro-equivalent a year.

Size class:large

Effect very large (+110) · 96 % of the effect is positive

Areas show individual argument weights. The positive share of the effect nets out money transfers between the two sides; the probability of a positive net effect is determined separately.

How we score →

Review and sources

Last reviewed by Claude Opus 5.5 · October 29, 2026 · 11× AI, 2× human
  1. October 29, 2026AI reviewClaude Opus 5.5record updated

    Page texts for the new page layout (draft G) with quotes from primary sources; numbers unchanged.

  2. October 21, 2026AI reviewClaude Fable 5.1record updated

    Money amounts in argument texts stated in dollars per style-en.md; scores, chains and balance unchanged

  3. October 20, 2026AI reviewClaude Fable 5.1record updated

    Summary and scale note rewritten to the style of 20 Oct 2026; numbers unchanged

  4. October 16, 2026AI reviewClaude Opus 5.5record updated

    Range of con-2 now covers only the case in which parents do withdraw (60,000 to 700,000); whether they withdraw at all stays in Plausibility.

  5. October 14, 2026AI reviewClaude Opus 5.5record updated

    Englische Debatten, Befund SO-16: ACTC-Fehlerquote mit Treasury-Bericht FY2025 belegt (13,6 % für Steuerjahr 2023, ±2,89 Pkt.), 15 % als gerundete Szenarioannahme gekennzeichnet.

  6. October 14, 2026Assessed by a human, implemented by AIClaude Opus 5.5record updated

    Bilanz 2.0 nachgetragen (Julians Freigabe 14.10.): P(D > 0) 1,00, D-Mittel +17,9; Kategorie deutlich besser statt besser nach altem Maßstab.

  7. October 12, 2026Assessed by a human, implemented by AIClaude Opus 5.5re-scored

    Entscheidung Julian 12.10.: Empfängergewicht in pro-1 als Mischgewicht 2,3 statt 2,5 (80 % erstes, 20 % zweites Einkommensfünftel), Mitte des Transferpaars pro-1 ⇄ con-1 von 19 auf 22 Mrd € (Tax Foundation 23,5 Mrd €); pro-2 und con-3 ziehen mit der Summe mit, Anzeigemaßstab k 0,2 → 0,1.

  8. October 12, 2026AI reviewClaude Opus 5.5re-scored

    con-3 nachgerechnet (19 × 15 % = 2,85 → I 0,855, score 2,05), evidence_basis-Token an pro-1, pro-3, con-2 berichtigt, Gleichstandsaussage korrigiert (Gleichstand erst bei Gewicht 0,6), fünf Belege nachgetragen, Erstattungsdeckel 1.700 $ im Maßnahmentext, US-Schreibweise.

  9. September 14, 2026AI reviewClaude Opus 5record updated

    Gesamtprüfung 08.09.: pro-2 kausal jetzt auf Hoynes/Schanzenbach/Almond 2016 und Aizer u. a. 2016 statt Urban-Simulation gestützt, Kohortenbuchung der 25 % benannt, con-2 um Befristung 2021 ergänzt; keine Zahländerung, r 0,73.

  10. September 6, 2026AI reviewClaude Opus 5record updated

    i_spanne an allen 6, normalisierung an 5 von 6 (con-3 Programm-Ansehen hat keinen Anker), Transferpaar pro-1 ⇄ con-1 verdrahtet (gleiche Menge, beide P 6, beide definitorisch — kein P-Angleich noetig), massstab_hinweis ohne r. Die beiden Paar-Beine widersprachen sich in der Spanne derselben Summe und sind jetzt exakt proportional. Kategorie steigt von Besser (r 0,73) auf Deutlich besser (P(D>0) 1,00) — allein durch die Paar-Buchung.

  11. September 6, 2026AI reviewClaude Opus 5First evaluation

    Created for the English side: the labour supply objection placed on the finding ladder against the 2021 full-refundability episode.

Evaluations are produced with AI support and reviewed on a schedule for new developments; human passes are marked separately.How we review →

Show all 13 sourcesHide sources
  1. Center on Budget and Policy Priorities: To Strengthen Economic Security, Policymakers Should Expand the Child Tax Credit to the 19 Million Children Who Receive Less Than the Full Credit Amount. cbpp.org
  2. Center on Budget and Policy Priorities: Gains From Expanded Child Tax Credit Outweigh Overstated Employment Worries. cbpp.org
  3. Congressional Research Service: Selected Issues in Tax Policy: The Child Tax Credit. congress.gov
  4. Urban Institute: How a Permanent Expansion of the Child Tax Credit Could Affect Poverty. urban.org
  5. Tax Policy Center: Let the Child Tax Credit Work. taxpolicycenter.org
  6. Hoynes, Schanzenbach and Almond, American Economic Review: Long-Run Impacts of Childhood Access to the Safety Net (2016): county rollout of the Food Stamp Program, 1961 to 1975. aeaweb.org
  7. Aizer, Eli, Ferrie and Lleras-Muney, American Economic Review: The Long-Run Impact of Cash Transfers to Poor Families (2016): accepted against rejected applicants to the Mothers' Pension program. aeaweb.org
  8. Center on Poverty and Social Policy, Columbia University: Sixth Child Tax Credit Payment Kept 3.7 Million Children Out of Poverty in December (January 2022). povertycenter.columbia.edu
  9. Shafer, Gutiérrez, Ettinger de Cuba, Bovell-Ammon and Raifman, JAMA Network Open: Advance Child Tax Credits Reduced US Food Insufficiency by 26 Percent (summary of the 2022 study, Boston University School of Public Health). bu.edu
  10. Tax Foundation: Make the Child Tax Credit Fully Refundable: 272.4 billion dollars over ten years, 2027 to 2036. taxfoundation.org
  11. Treasury Inspector General for Tax Administration: Assessment of Fiscal Year 2025 Compliance With Improper Payment Reporting Requirements (Report 2026-400-039). oversight.gov
  12. Corinth, Meyer, Stadnicki and Wu, Becker Friedman Institute: The Anti-Poverty, Targeting, and Labor Supply Effects of the Proposed Child Tax Credit Expansion (Working Paper 2021-115). bfi.uchicago.edu
  13. U.S. Department of the Treasury: Agency Financial Report, Fiscal Year 2025: Payment Integrity, refundable tax credits (Table 1, tax year 2023, page 207). home.treasury.gov

More debates in this area

Very likely worse · 0 %
Size: large
The Forced-Labour Tariffs

An additional duty of 10 or 12.5 percent on imports from sixty economies, set by whether each bans, or has committed to ban, imports made with forced labour.
Unclear · 48 %
Size: medium
School Scholarship Credit

Let every taxpayer send up to 1,700 dollars of federal income tax to scholarship organizations for school children instead of the Treasury, from January 2027.
Very likely better · 100 %
Size: very large
Lifting The Payroll Cap

Charge the Social Security payroll tax on earnings above the taxable maximum of 184,500 dollars, which today stop being taxed altogether.
Very likely worse · 0 %
Size: large
Keep the Tips Deduction

Make the deductions for tipped and overtime income permanent instead of letting them lapse after 2028.
All debates in Taxes, Tariffs and the Federal Budget →