cityparity

Childcare subsidies and US income limits: where the help stops

By Skyler Bissell · August 2, 2026 · 7 min read

San Francisco pays a full childcare tuition credit to a family of four earning up to $233,800, half a credit up to $311,700, and nothing at all above that. Those two numbers are worth more attention than almost anything in an offer letter, because between them sits a credit measured against a $2,115 monthly preschool rate, and above them sits a family paying the full San Francisco market price of $2,400 a month per child.

American childcare support is a stack of three unrelated instruments with three unrelated rules, and the one people search for is the one that vanishes first. This walks all three, says what each is worth in dollars at a professional income, and shows where the ladder ends.

TL;DR

The three layers, and what each is worth

Start with the one that never goes away. The Dependent Care FSA lets an employee set aside roughly $5,000 of pre-tax pay against care costs, which is worth about $2,000 a year in saved tax at a high marginal rate. It is not income-tested, it is offered through an employer plan, and it is the figure our engine nets from every US city's childcare line. Its weakness is that it is flat: it is capped per household, so a family with two children in full-time care gets the same $2,000 as a family with one. Against a New York bill of $57,600 that covers about eleven days.

Next, the Child Tax Credit. The federal credit is $2,000 per child under 17, and our New York model carries $2,200 per child once the Empire State Child Credit is blended in, so $4,400 for two children. It is not a childcare benefit and pays the same whether or not you use childcare, but it is real cash and it survives at these incomes: the federal phase-out does not begin until $400,000 of adjusted gross income for a married couple filing jointly.

Third, the local income-tested tuition credit. This is the layer worth the most and the only one with a hard ceiling. San Francisco's Early Learning For All programme is the one our data models in detail, and it is generous in a way that surprises people who assume the US has nothing: a qualifying preschool place is credited against a reimbursement rate of $2,115 a month, which is $25,380 a year per child. That is most of a San Francisco preschool bill.

The ladder, in dollars

San Francisco's bands are set as multiples of area median income and published as dollar figures for a family of four. Three rungs:

Look at the step between the second rung and the third. A household one dollar over $311,700 loses up to $25,380 of support for a single preschool-age child. On that dollar of income the effective marginal rate is not 37% or 50%; it is several thousand percent. With two children in the preschool band the same dollar can cost twice as much again. This is what our glossary calls a means-tested cliff, and the American childcare system is full of them at every income level.

Two warnings about those dollar figures. They are AMI multiples, so they are re-set annually and they shift with family size: a family of three faces different numbers from a family of four. And they are San Francisco's. New York, Boston, Seattle and Austin all run their own schedules on their own bands, and our engine carries an explicit income-tested schedule only where we have sourced one.

Why our headline number ignores the credit

Here is a modelling decision worth being upfront about, because it changes the biggest number on the page.

Our engine bills San Francisco childcare at the market rate of $2,400 a month per preschooler and never nets the ELFA credit off it. The schedule drives a separate informational note telling a qualifying household what it might claim; it does not move the cost. So the San Francisco figure we publish is $57,600 gross and $55,600 after the FSA, which is 40.2% of this household's take-home of $138,315, and it is identical to the New York figure even though New York has no comparable schedule at all.

The reason is availability. The credit needs city residency and a place at a contracted provider, and those places are supply-limited and prioritised toward lower incomes. A household at the top of the eligibility band is the least likely to secure one, and a comparison that hands it a $25,380 discount it probably cannot claim would flatter the US side of every calculation on this site. Where a value has judgement in it, our rule is to take the reading that makes the destination look worse. The upside is surfaced where you can act on it, and the headline stays the number you can plan against.

It is the same problem Berlin has from the opposite direction. A Berlin Kita place costs nothing and there are not enough of them, so the constraint is the queue. The difference is that a German family who waits eventually gets a free place, while an American family above the band never becomes eligible however long it waits.

What the US does not have

Read the other posts in this series and one absence stands out. None of the US metros we price has a universal fee cap.

Norway caps what any provider may charge at NOK 1,200 a month, and every household pays it regardless of income. Germany's Länder fund the place directly, and five of our eight German cities bill nothing. The Netherlands charges full price and refunds a statutory minimum percentage that no income can fall below. England funds a fixed block of hours for every 3 and 4 year old with no income test at all. Four different mechanisms, and all four share one property: they have a floor.

The American design has no floor. It has a ladder that ends, and above the last rung the household absorbs the full cost base of a labour-intensive service in an expensive city. That is the single structural reason two children in daycare cost $55,600 a year in San Francisco and €0 in Berlin, and it is why the childcare line moves a relocation decision further than the salary line does. The whole spread sits on our childcare costs by country table.

If a raise is about to cross the line

Say you are at $305,000 in San Francisco with a four-year-old in a contracted place, and a $20,000 raise is on the table. Take it and you cross $311,700, lose a credit worth $12,690 this year and the chance of $25,380 next year, and pay tax on the raise as well. The raise is real; the net gain is a fraction of it, and possibly negative in year one.

Three things are worth doing before you sign anything. Check the current year's AMI bands for your actual family size, because the dollar figures move and a family of five sits on different rungs from a family of four. Ask whether pre-tax deferrals move the income the schedule looks at, since a 401(k) or an FSA contribution can change which side of a rung you land on. And price the raise against the whole household ledger rather than the paycheck alone, which is what running the numbers here does in about two minutes.

Which rung you sit on this year is a fact about an AMI table and your family size, and both of those move without consulting you. That is the real difference between American childcare support and the four European systems in this series: elsewhere the number is fixed in law and reaches everyone; here it is fixed in a table you have to go and look up, every year, and it can leave without warning. The worked San Francisco vs Berlin family comparison and our cost of raising kids breakdown carry the rest of the ledger.

FAQ

What is the income limit for childcare assistance in the US?

There is no single national limit, because the schedules are set by cities and states against local area median income. San Francisco's Early Learning For All programme is the one our engine carries in full: a family of four qualifies for a 100% tuition credit up to $233,800 of household income, a 50% credit up to $311,700, and nothing above that. Because the bands are AMI multiples, the dollar figures shift with family size and are re-set each year.

How much is the Dependent Care FSA worth?

About $2,000 a year of saved tax for a high earner, against a pre-tax contribution cap near $5,000. The important property is that it is flat per household: two children in daycare and three children in daycare both produce the same $2,000. It is available through an employer plan and is not income-tested, which makes it the only childcare support that reliably survives a two-earner professional income anywhere in the US.

Does the Child Tax Credit phase out for high earners?

Eventually, and far above the point where childcare help disappears. The federal credit is $2,000 per child under 17 and its phase-out starts at $400,000 of adjusted gross income for a married couple filing jointly, or $200,000 for a single filer. Our New York model carries $2,200 per child once the state top-up is blended in, so $4,400 for two children, and it survives intact at the incomes on this page.

Why does cityparity charge the full market rate if a subsidy exists?

Because a credit you cannot claim is not a discount. The San Francisco credit requires city residency and a place at a contracted provider, and those places are supply-limited and prioritised toward lower incomes, so the household at the top of the band is the least likely to get one. Our engine bills the unsubsidised market rate and surfaces the schedule as a separate note, which keeps the headline figure the one you can rely on and puts the upside where you can see it.

More in this series

Sources. The reimbursement rate comes from the San Francisco Department of Early Childhood's ELFA FY25-26 rate sheet, and the income bands and credit design from Children's Council San Francisco's help-paying-for-child-care guidance. Child Tax Credit amounts and phase-outs come from the IRS Child Tax Credit page, and the dependent-care rules from IRS Publication 503. Field-level provenance and dates are in data/_meta.json; the method is on the methodology page.

Figures here come from cityparity's per-city engine and were current at publication; AMI bands, reimbursement rates and credit amounts are re-set annually, so treat any single number as a strong estimate and run your own inputs. See the methodology.