Education Funding

What does education actually cost?

Less than the published price, for most families, and the gap between the two is where planning should start. Colleges publish a sticker price. What a particular household pays is the net price: the published figure minus the grants and scholarships that do not have to be repaid. The difference is routinely large, it varies enormously by institution and by family, and it is the only figure worth building a savings target around.

The number the financial aid system works from is not the tuition bill either. It is the cost of attendance, a college's official estimate of what one year there costs a particular student. Federal law defines what it must include, and the list runs well past tuition and fees to housing, food, books and course materials, transportation, personal expenses and a series of situation-specific items. That matters in two practical ways. Tuition is frequently under half of it, so a family that plans for tuition alone has planned for part of the problem. And the non-tuition components are estimates the institution sets, not charges it sends, so a student can spend more or less than the allowance without anything having gone wrong.

Because each institution sets its own allowances, two comparable colleges can publish figures thousands of dollars apart for the same underlying reality. Federal law requires each one to publish every element of its cost of attendance rather than only the total, which is what makes an element-by-element comparison possible instead of a comparison of two bottom lines. When a figure looks out of line with a similar school, the answer is usually in the housing or food allowance rather than in the price of the education.

The most useful structural fact about these budgets is that the non-tuition part barely varies with the price of the college. Housing, food, books, transportation and personal expenses come to roughly the same annual figure at an in-state public university and at an expensive private one, because a student has to live somewhere and eat either way. Only tuition varies. In the College Board's 2025-26 figures that non-tuition block is about 61 percent of the whole annual budget at an in-state public four-year college and about 81 percent of it at a community college. That has two consequences. The entire sticker-price gap between a public in-state option and a private one is tuition, so that is the part a scholarship can address. And "tuition is free" at a community college leaves four-fifths of the bill outstanding, which is the commonest misreading of a tuition-free program.

The gap between sticker and net is created by grant aid, and its composition differs by sector in a way worth knowing before you read an award letter. In 2025-26 average grant aid covered roughly four-fifths of published tuition and fees at in-state public four-year colleges and roughly three-fifths at private nonprofit ones. But in the most recent detailed year available, 2022-23, about 88 percent of the average grant at a private nonprofit came from the college itself, against about 52 percent at a public four-year. A private college's aid is largely its own discount off its own list price; a public college's is largely somebody else's money. That is why a large scholarship from an expensive private college can still leave a higher net price than a modest one from a public university.

Is the cost still rising faster than inflation?

Not over the last decade, which is the opposite of the story most readers carry, and getting it right changes how a family should plan. Published tuition did rise far faster than inflation for roughly three decades through the mid-2010s. Since then, in the College Board's own inflation-adjusted series, average published tuition and fees are down about 7 percent in real terms at public four-year colleges over the ten years to 2025-26 and down about 10 percent at community colleges, while private nonprofit tuition is up about 2 percent. Average net tuition after grant aid at public four-year colleges peaked in 2012-13 and, on the College Board's estimate for the current year, is roughly half that level in real terms now. Treat the current-year net figure as a projection: the underlying federal grant data runs two years behind.

Without four qualifications that becomes its own misleading claim. A flat decade sits on top of a long climb, so real published tuition at a public four-year college is still more than double its mid-1990s level. Housing and food have not fallen; they rose again in the most recent year, and since they are the majority of the budget at public institutions the total bill for a residential student has held up much better than the tuition line. These are averages, and net-price averages cover first-time full-time students only, so an individual family's price can rise while the average falls. And a good part of the real decline happened because colleges raised nominal prices by less than inflation during the 2021 to 2023 inflation spike, which is genuine money in a family's pocket but is not a trend to project forward.

The planning consequence is worth stating plainly, because the false version pushes families in two unhelpful directions at once. It encourages over-saving into an account with a penalty for non-education use, and it encourages the fatalism that says saving is pointless because the target is running away. Neither follows from the data. A savings target built on a projection of costs compounding well above inflation for eighteen years is building in an assumption that has not held for a decade.

Two structural features change what "the cost" even means. The aid calculation runs on household income from two years before the year of enrollment, so the financial picture being assessed is not the current one. And a college's published cost is for one year, while the decision is a four-year one, during which price generally rises and some scholarships do not repeat.

Is it worth the money?

On average, clearly, and the average is a poor guide to an individual decision. In the Bureau of Labor Statistics' 2024 data, full-time workers with a bachelor's degree had median weekly earnings about two-thirds higher than those with a high school diploma, and an unemployment rate roughly 1.7 percentage points lower. Treating the degree as an investment, the New York Fed puts the internal rate of return for the median bachelor's holder at about 12.5 percent, a figure it describes as having been consistent at 12 to 13 percent for several decades.

The decision-relevant half is the spread around that median, and it is wide. The same New York Fed work estimates a return of about 2.6 percent at the 25th percentile of college graduates, which by construction describes at least a quarter of everyone who finishes. Three things move an individual outcome more than institutional prestige does.

• Completion, which is the dominant variable. The financial case rests almost entirely on finishing, and the worst outcome available here is borrowing for a program that is not completed: the debt behaves exactly as it would have and the earnings increase never arrives. In the 2024 BLS figures, workers with some college and no degree earned about 10 percent more than high school graduates, which is roughly a seventh of the bachelor's premium. And a Federal Reserve analysis of borrowers surveyed between 2017 and 2019 found those who borrowed and did not finish far less likely to report doing at least okay financially, a gap its authors put at roughly two and a half times the size of the improvement associated with completing.
• Time to the degree, which families control and ignore. The New York Fed estimates that taking five years rather than four pushes the median return down to about 9 percent, and six years to about 7 percent. Roughly a quarter of the return goes with each extra year, through both the added cost and the delayed earnings. This is worth holding against the common advice to spread a degree over more years to ease the annual bill: it does ease the bill, and it is not free.
• The price actually paid. This is a household variable rather than a school variable. The New York Fed's own scenarios put the return nearer 10 percent for a student receiving no financial aid, and lower still where costs run above average. The same degree at the same college is a materially different investment for a family paying list price and one paying a heavily discounted net price.

Put the first two together and you get the framing that matters most, and the reason it is easy to miss is that the two halves are published separately. The return to a degree is conditional on getting one, and the probability of getting one is a published, institution-specific number. In federal data for students who entered in 2017, the six-year graduation rate at open-admissions institutions was around 30 percent, against roughly 92 percent at institutions accepting fewer than a quarter of applicants, with a steady gradient between. Across all four-year institutions about 65 percent finished within six years and about 49 percent within four, so only around three-quarters of those who finish do so on time. The honest expected value multiplies a completion probability by a return that is conditional on completing, and both inputs are available per institution and per program on the Department of Education's College Scorecard.

Two things keep this honest. Those graduation rates count only students who started full time and finished at the same institution, so a student who transfers and graduates elsewhere is recorded as not completing, which means the published rate understates true completion. And all of these figures describe what people with different levels of education actually earn, not what a particular person would have earned had they chosen differently. Who goes to college, and who finishes, is not random, so none of this is a causal estimate of what the degree itself does.

The reasoning applies with more force to graduate study, where costs are higher and grant aid is scarcer. Work published by the Urban Institute in 2022, using federal program-level data, found that a rule sanctioning programs whose graduates owe more than they earn early in their careers would reach close to one in four master's recipients who borrow, concentrated in a small number of fields. The undergraduate conclusion does not carry over program by program, and for some programs there is no positive financial return at all. It cuts the other way too: a shorter credential, a public institution, or two years at a community college before transferring are financially serious options, not consolation prizes.

Where does education saving sit among other goals?

After an emergency reserve, after genuinely expensive debt, and usually after retirement funding. That ordering upsets people, so it is worth stating the reason plainly, because the reason is mechanical, not a judgment about family priorities: there is borrowing available for education and none for retirement. A student can borrow, work, choose a cheaper school, start somewhere less expensive and transfer, or take longer. Nobody can borrow to fund a retirement.

The second half of the argument is the part that resolves the guilt. A parent who arrives at retirement underfunded does not absorb that shortfall privately. It reappears as support the adult child may be asked to provide, at a point in that child's life when they are funding their own household. Funding retirement first is not a choice of self over child; it is a choice about which of two costs the child eventually carries, and the retirement shortfall is generally the larger and the less avoidable of the two.

The general sequence most planners work from is the financial order of operations: capture any employer retirement match, which is the one return no other use of the money can match, clear high-rate debt, build an emergency fund, fund tax-advantaged retirement accounts against a target you have actually checked, then fund education. Our retirement planning guide covers what that target involves.

What matters more than the sequence is this. Partial funding is a legitimate plan: a family that intends to cover half of a public in-state cost and says so out loud has planned better than one that intends to cover everything and has not started. And the ordering is not absolute. A household with a high income, a fully funded retirement plan and a long horizon is in a different position from one choosing between a retirement contribution and a 529 contribution this month, and time in the market is worth enough that a very young child argues for starting something small early rather than waiting for perfect sequencing.

How much should we save?

Enough to cover a share of a named cost that you have chosen deliberately, which is a very different exercise from working toward a number you read somewhere. A usable target needs four inputs, and the value of writing them down is that each one is arguable and none of them is a forecast: which kind of institution you are planning for, what proportion of the cost you intend to cover, how many years away it is, and how the money will be invested in the meantime.

The share-of-cost framing survives contact with reality better than the full-cost one. Aiming at the entire projected cost of an expensive private college produces a number large enough to cause paralysis, and it also over-saves for a likely outcome, because most students do not attend the most expensive option and most families do not pay the published price. Choosing a fraction, half of a public in-state cost, or a fixed monthly amount you will actually sustain, converts the goal into something you can start this month and revise later. Under-saving is recoverable from income, current earnings and modest borrowing; a target nobody starts is not.

The horizon then decides the investment mix, and education money differs from retirement money in one way that matters more than any other. It has a hard deadline. A retirement portfolio that falls in value can be left alone for a few years; a college bill arriving in September cannot wait for a recovery. That is why education accounts are usually built to reduce risk as the deadline approaches rather than holding a constant mix, and why most 529 plans offer an age-based or enrollment-date option that does this automatically, much like a target-date fund does for retirement. Two consequences follow: money for a child a decade or more away can reasonably be invested for growth, and money needed within the next two or three years generally should not be. Our investing guide covers the underlying asset allocation question of how the money is split between stocks and bonds, and compound growth is the reason an early start does disproportionate work.

One habit is worth more than precision in the target: revisit it when something changes rather than annually out of duty. A new child, a move to a state with a different public system, a change in income, or a teenager whose plans have come into focus all change the inputs enough to be worth an hour.

Which account should the money go in?

For most families saving specifically for education, a 529 plan, and the reason to know the alternatives is that the exceptions are real. Every vehicle in this section differs on the same four axes, and comparing them on anything else wastes time: how the growth is taxed, who controls the money, what happens if it is not spent on education, and how the financial aid formula sees it.

529 plans

A 529 plan is a state-sponsored investment account for education savings. You contribute after-tax money, it grows untaxed, and withdrawals are free of federal tax when spent on qualified education expenses. There is no federal deduction for contributions, but many states offer their own income tax deduction or credit, usually limited to their own plan, which is the main reason to look at your home state's plan before shopping on fees. You can otherwise use almost any state's plan regardless of where you live or where the student eventually studies.

Three features make it the default rather than merely the most familiar. Capacity is high enough that it rarely binds, and there is no federal annual contribution limit at all, only the gift tax rules and a lifetime cap set by each plan. The owner keeps control, so unlike a custodial account the money does not become the student's to spend at the age of majority. And the range of qualified expenses has widened considerably: beyond college tuition, fees, books, required equipment, a computer and internet access, and housing and food for a student enrolled at least half time, it reaches K-12 costs up to $20,000 a year, plus apprenticeship costs, which have qualified since 2019, and, since the 2025 tax law, both a much broader list of K-12 items than tuition alone and certain postsecondary credentialing and licensing expenses, including continuing education required to keep a credential current.

Two details about that K-12 allowance are easy to miss and both change the answer. The figure is fixed in statute with no inflation adjustment, and it applies per beneficiary across every plan and every account, not per account, so opening a second account does not create a second allowance. And state conformity is where this expansion breaks down: some states have not adopted every federal change, and at least one large state treats a withdrawal that is federally qualified for K-12 as nonqualified for its own purposes. The same hazard applies to moving an account to another state's plan, which some states treat as a nonqualified withdrawal and use to reclaim a deduction they gave you earlier.

A related but distinct product is a prepaid tuition plan, offered by a handful of states, which lets you buy future tuition credits at something closer to today's price. It trades the investment risk of a savings plan for a different set of risks: the range of schools the credits cover is narrower, the guarantee behind them varies by state, and the flexibility if the student goes elsewhere is more limited.

Coverdell education savings accounts

A Coverdell education savings account works like a 529 in its tax treatment, with tax-free growth for qualified education expenses, and differs in three ways that make it a niche choice and not a competitor. Total contributions for a beneficiary are capped at $2,000 a year from all sources combined, which is small enough to make it impractical as a primary college vehicle. Contributions phase out above an income level for the contributor, which 529 plans do not impose. And the account carries age limits at both ends: contributions generally stop when the beneficiary turns eighteen, and the balance must generally be used by thirty, with exceptions for a beneficiary with special needs. Neither figure is indexed: the $2,000 limit dates from 2001 and the contributor income range from 1997, so both have eroded substantially in real terms and will keep doing so.

What it retains is breadth of expense. Its definition of qualified elementary and secondary school costs is wider than the 529 version and carries no dollar cap, reaching items such as room and board, uniforms, transportation and extended-day programs. A family paying private school costs during the years before college may therefore find it useful alongside a 529 and not instead of one.

ABLE accounts, where the beneficiary has a disability

An ABLE account is not an education account, and it belongs here because for a family with a disabled beneficiary it often outranks one. It is a tax-advantaged account for qualified disability expenses, which include education among a much broader list, and its central feature is that a substantial balance does not disqualify the beneficiary from means-tested benefits the way ordinary savings in their name would. Eligibility depends on the disability having begun before a statutory age, which recent law widened. Money can be moved from a 529 to an ABLE account for the same beneficiary or a family member.

Below the headline, three details do the work. The annual contribution limit is set by reference to the gift-tax exclusion but, since a 2025 change to how it is computed, is no longer the same figure, so the two have to be looked up separately. Where the balance passes the resource threshold, Supplemental Security Income is suspended rather than terminated and Medicaid continues, which is a materially better outcome than the usual description implies. And the account carries a provision allowing the state to be repaid for certain Medicaid costs out of what remains at death, which is a real cost to weigh and not a footnote. A household in this position is making a benefits-eligibility decision as much as a savings decision, and it is one of the clearest cases on this page for professional help.

Custodial accounts, taxable accounts, and savings bonds

A custodial UTMA or UGMA account is a plain investment account holding an irrevocable gift to a minor. It has no education restriction at all, which is its appeal and its problem: the money legally becomes the child's to use for anything at the age of majority, and because it is the student's own property the aid formula assesses it far harder than a parental asset. Its investment income is taxed under rules designed to prevent income shifting, so the tax advantage is smaller than families expect.

A taxable brokerage account in a parent's name buys complete flexibility at the cost of tax drag on dividends and realized gains. It is the right answer where the purpose is genuinely uncertain, where the money may be needed for something other than education, or as a place for savings beyond what a 529 can sensibly absorb.

US savings bonds carry an education provision of their own: interest on certain Series EE and Series I bonds can be excluded from income when the proceeds go to tuition and required fees, subject to an income limit. Whether it works turns on two conditions, and one of them is set years before anybody thinks about it. The bond must be issued in the name of the person claiming the exclusion, not the child's, and that person must have been at least 24 when it was issued. A bond bought in a child's name, which is the obvious thing to do and a common gift, permanently defeats the exclusion for everybody. Note also that the qualifying expenses here are narrower than elsewhere, tuition and fees only, though a contribution of the proceeds to a 529 or Coverdell counts.

The retirement accounts people reach for

A Roth IRA is a retirement account, and using it for education is a decision with a cost that is easy to miss. Contributions can be withdrawn at any time without tax or penalty, which is why the account gets suggested; and the higher education exception means that even earnings withdrawn for qualified education expenses escape the early-withdrawal penalty, though those earnings remain taxable income unless the withdrawal is otherwise qualified. The cost is that retirement contribution room is annual and cannot be recovered. A dollar spent on tuition from a Roth IRA is a dollar of tax-free retirement growth given up permanently, and it may also appear as income on a future aid application. Borrowing from a workplace retirement plan for the same purpose carries a further hazard covered in our employee benefits guide.

Who should own the account, and will it hurt financial aid?

Ownership answers two separate questions, control and aid treatment, and families systematically overweight the second. The aid effect of a parent-owned education account is small, and the federal formula is specific enough that the size of the effect can be stated. Parental assets above an allowance are assessed on a sliding scale, and the statutory rates work out so that at most about 5.6 cents of each dollar of a parental asset ends up in the index. A dependent student's own assets, by contrast, are assessed at a flat 20 cents on the dollar with no allowance at all. Parental income drives the formula far more powerfully than either. A family reorganizing its savings to protect aid eligibility is usually optimizing the smallest of the three variables.

Three ownership arrangements come up, and the differences between them are worth knowing before money moves rather than afterward.

• Parent-owned is the default. The account is reported as a parental asset, the parent controls investment choices and withdrawals, and the beneficiary can be changed later. This is the arrangement most families should want, and it is the one that preserves the option to redirect the money if plans change.
• Owned by a dependent student. Federal law is explicit that an education savings account designated for a dependent student is the parent's asset "regardless of whether the owner of the account is the student or the parent", so it draws the low parental rate rather than the flat student rate. That removed what used to be a real penalty. Note the contrast with a custodial UTMA or UGMA account, which is not an education savings account, is genuinely the student's property, and is assessed at the student rate.
• Owned by a grandparent or other third party. The account is not reported on the student's aid application at all, and neither, now, is a distribution from it. Older guidance warned that money paid out of a grandparent's 529 counted against the student as untaxed income on a later application, which was a real and awkward trap. The FAFSA rewrite removed the questions that produced it, and the Department of Education has confirmed the categories were eliminated. A college using its own methodology may still ask, so the caution survives in a narrower form.

One consequence of those rules is easy to miss: an account held for one child is not reported on a sibling's aid application, since it is neither designated for that student nor an asset of theirs. Families with children close in age sometimes assume the reverse.

Gifting into someone else's account has its own dimension. A contribution to a 529 is a completed gift for tax purposes, which for most families is unremarkable because ordinary annual gifts fall under the amount anyone can give any number of people each year without a filing. For larger gifts a special election allows a contributor to treat a lump sum of up to five times the annual exclusion as if it were spread over five years, which is how a grandparent funds an account substantially in one step; it requires a gift tax return to elect, and further gifts to that beneficiary during the five years can have gift tax consequences. Paying tuition directly to the college is a different and in some ways better route, for reasons set out under tax benefits below.

One asymmetry is worth holding onto: control and tax efficiency generally point the same way, toward a parent-owned 529, while the aid question is small enough that it should rarely be the deciding factor. Where it does decide something, it is usually because a college uses its own formula rather than the federal one, and the way to find out is to ask that college.

How does financial aid actually work?

One subtraction sits at the center of the whole system: cost of attendance, minus the index the aid application produces about your family, minus other assistance already committed, equals need. Everything else is either an input to that line or a decision about what to do with its output.

Three things follow from it immediately, and each corrects a common misreading. Need is a calculated measure, not an offer. Nothing obliges a college to meet it, and many colleges meet less than the full amount, so the difference between calculated need and the aid actually offered is where the real price of a place sits and a family should not assume the calculated figure will be covered. And the index that gets subtracted is not a bill. The Student Aid Index, which replaced the older Expected Family Contribution, is exactly what its name says: a number that gets subtracted. It can be negative, which increases measured need rather than reducing it, and it is not an amount anyone is asked to pay.

The form that produces it is the FAFSA, the Free Application for Federal Student Aid, filed once for each academic year at no cost. Federal law requires the Department of Education to open it by October 1 before the year of enrollment, which is the earliest a family can file rather than a deadline to meet, and the deadlines that actually bind are set by states and by individual colleges, some of them early and some of them first-come while funds last. The form runs on income from two years back and requires consent to retrieve tax data directly from the IRS, which is a condition of eligibility rather than an option. Many private colleges also require a second form, the CSS Profile, which asks more and feeds an institutional formula that can treat assets, home equity and a non-custodial parent quite differently from the federal one.

What comes back divides into two kinds of money with different logic. Need-based aid follows from the subtraction above and includes federal grants, state grants, institutional grants and work-study. The largest federal grant for undergraduates with need is the Pell Grant, which does not have to be repaid, is capped over a student's lifetime and, since July 2026, carries two new disqualifications worth knowing: a student aid index at or above twice the maximum grant removes eligibility outright, and so does receiving non-federal grant aid equal to or above the full cost of attendance. That second rule runs opposite to every intuition about aid, because it lands on the student holding a full scholarship. Merit aid, by contrast, does not come from this formula at all. It is a discount a college chooses to offer to attract a student, or a private scholarship awarded on its own criteria, and it is competed for rather than qualified for.

Because the formula looks two years back, families whose circumstances have changed are the ones it serves worst. The route is not to argue with the index but to ask the college's financial aid office for a documented case-by-case adjustment, and two statutory protections make that request more than a favor: an institution may not maintain a policy of refusing all such requests, and it may not charge a fee to review one. A job loss, a death, a divorce, high unreimbursed medical costs or a one-off spike in the base year are all ordinary grounds.

How do we compare offers?

By reducing each one to a single figure that only counts money you do not have to repay: the college's cost of attendance, minus grants and scholarships. An award letter typically presents grants, work-study and loans together under a heading like "your financial aid package", which makes a larger package look like a better offer when part of it is simply permission to borrow. A loan in an award letter is not aid; it is a cost with the payment deferred.

Three adjustments turn that figure into something comparable across schools.

• Use each college's own cost of attendance, not a national average and not tuition alone, and check whether the housing and food assumptions match what the student will actually do. A commuting student and a residential student at the same college have materially different real costs.
• Run all four years, not the first. Ask whether each scholarship renews, what conditions attach to renewal such as a minimum grade average or full-time enrollment, and what the college's recent history of price increases looks like. An offer built largely on a one-year entering award is a different proposition from one built on a four-year commitment.
• Subtract what the family will actually contribute from savings and current income, and look at what is left. That remainder is the borrowing the offer requires, and it is the number the next section is about.

Ask every aid office two questions directly, because the answers are rarely in the letter. Whether the college practices scholarship displacement, meaning it reduces its own aid when a student wins an outside scholarship, which determines whether hunting for private scholarships will actually reduce the bill. And whether the office will reconsider the offer, which is a normal request rather than a confrontation and is strongest when it rests on documented changed circumstances or on a competing offer from a comparable institution.

Which tax benefits apply, and how do they interact?

Several, and the useful first question is never whether an expense counts but what it counts for. Each federal education benefit defines its own list of qualified education expenses, and the lists genuinely differ, so an expense can qualify for one benefit and not another. Room and board is the clearest case: it counts for a 529 withdrawal and not for either education credit.

The two credits are where most of the value sits for a household paying tuition. The American Opportunity Tax Credit is the larger of the two, is claimed per student for up to four years of undergraduate study, requires at least half-time enrollment in a program leading to a recognized credential, which includes a certificate program and not only a degree, and is partly refundable, meaning it can pay out even where no tax is owed. The Lifetime Learning Credit is smaller, is claimed per tax return rather than per student, and exists for everyone the first credit excludes: graduate students, part-time students, someone taking a single course to improve job skills, and anyone who has already used four years of the other one. A household with two students commonly claims one credit for each rather than choosing between them.

Both credits phase out over the same income range, and that range is fixed in statute with no inflation adjustment: it runs from $80,000 to $90,000 of modified adjusted gross income, or $160,000 to $180,000 on a joint return. The separate indexing provision that used to move the Lifetime Learning Credit's thresholds was repealed at the end of 2020. The consequence is worth naming because it works quietly: an unindexed phase-out tightens in real terms every year, so a household that qualified comfortably several years ago can find the credit shrinking on an income that has only kept pace with inflation. Neither credit is available to a married person filing separately, at any income. And one eligibility condition changed for returns covering 2026 and later: both credits now require a work-authorized Social Security number for the taxpayer and for the student, so an individual taxpayer identification number no longer works. The deadline for having the number, and what happens if you miss it, are on the credit pages.

The rule that turns all of this into a planning exercise is that no single expense may support more than one benefit. A family paying tuition, fees, housing and food in the same year is therefore allocating those costs among the benefits available instead of counting them twice. The usual shape of a good allocation is to reserve enough tuition to claim the more valuable credit in full and cover the remaining qualifying costs with a tax-free 529 withdrawal, because the credits are capped and worth more per dollar while the 529 exclusion is uncapped and worth only the tax on the growth.

One ordering rule sits above the allocation. Tax-free educational assistance comes off the pool of qualifying expenses first: a scholarship excluded from income, a Pell Grant, employer help and veterans' education benefits all reduce the expenses available to support anything else. A family whose costs are largely met by grants may have fewer qualifying dollars left than the bill implies. One consequence of that ordering is genuinely counterintuitive and worth knowing when an employer is involved: a tuition reimbursement that you take as taxable pay can leave a credit fully intact, where the same reimbursement excluded from income reduces the qualifying expenses and shrinks the credit. The tax-free version is not automatically the better one.

Scholarships and fellowships are generally excluded from income to the extent they cover tuition and fees, and also course-related expenses such as books, supplies and equipment where the course requires them. The portion covering housing and food is taxable to the student, which surprises families and occasionally produces a filing obligation for a student who has never had one. That taxable portion is also the basis of a well-known maneuver, in which a student deliberately treats part of a grant as taxable income spent on housing in order to free tuition dollars to support a credit. It can work, and it has two costs that are invisible if you reason only from the first step. A taxable scholarship counts as earned income for working out the student's standard deduction, which is the part everyone notices and the part that makes the move look free. But for the tax actually charged on it, it is treated as unearned income, so it can be taxed at the parents' rate rather than the student's under the rules for a child's investment income. And where those rules apply to the student, the refundable portion of the larger credit is denied outright. Two of the three effects run against the family, which is why this belongs in a conversation with whoever prepares the return, not in a rule of thumb.

Two other routes matter. Employer educational assistance is a separate exclusion, and its use to repay student loans, which had been due to expire, is now permanent. Note the limit that matters on a page about paying for a child's education: it reaches only the employee's own education and the employee's own loans, so it cannot pay down a loan a parent took for a child. Our employee benefits guide covers how it fits a benefits package. And there is a deduction for interest paid on a qualified education loan, capped at $2,500 a year, available without itemizing, phased out over an income range and unavailable to a married person filing separately. It carries a trap for families who split responsibilities: the deduction requires you to be legally obligated on the loan and denies it to anyone claimed as a dependent, so where a dependent student is the only borrower and the parent makes the payments, nobody gets to deduct the interest. Our taxes guide sets credits and deductions in their broader context.

The grandparent tuition payment, which four rules happen to favor

One arrangement in this territory is unusually efficient, and it is efficient because four separate provisions line up rather than because of any single break. Where a grandparent pays tuition directly to the college for a dependent student, four things are true at once.

• The payment is not a gift at all, under the unlimited exclusion for tuition paid straight to an institution.
• It sits outside the generation-skipping transfer tax, so it does not consume that exemption either.
• It is not treated as tax-free educational assistance, so unlike a scholarship it does not reduce the pool of expenses available to support a credit.
• And it is treated as made by the student, whose payments are in turn treated as made by the parent claiming them, so it can support the parent's education credit.

That combination is worth knowing precisely because the obvious alternatives are less efficient. Handing the same money to the student or the parent to pay the school is an ordinary gift. And a contribution to a 529 is expressly not a payment of tuition for this purpose, so it does not get the unlimited treatment, which is why the five-year front-loading election exists at all. The exclusion also reaches tuition only. Housing, food, books and supplies paid directly to the college are ordinary gifts, a distinction our estate planning guide covers in more detail.

Should we borrow, and how much?

Borrow after everything cheaper is exhausted, in the student's name before anyone else's, and only up to an amount that survives the least favorable version of events. The order to work through is grants and scholarships first, then savings and current income, then federal loans in the student's own name, and only then parent borrowing or private credit. Each step down that list is more expensive than the one above it, and the last two are where families do damage that lasts. How much a family can find in the second step is a cash flow question rather than a savings one, and for many households it is the largest single lever.

The federal borrowing rules changed on 1 July 2026, so anything written before mid-2025 describes a system that no longer exists. What follows is the structure for someone starting now. A transitional rule, set out below, keeps the older rules in place for students who were already partway through, and it matters more than it sounds. The dollar amounts here are fixed in statute with no inflation adjustment, so unlike most numbers in this territory they will not drift.

• Undergraduate limits are unchanged. A dependent undergraduate can borrow $5,500 in the first year, $6,500 in the second and $7,500 a year after that, of which a portion may be a Direct Subsidized Loan where the student has demonstrated need, meaning the government pays the interest during enrollment. An independent undergraduate can borrow more. The lifetime undergraduate total is $31,000 for a dependent student and $57,500 for an independent one, of which at most $23,000 may be subsidized.
• Graduate students no longer have a separate PLUS loan, and their total capacity fell. For instruction beginning on or after 1 July 2026 a graduate or professional student cannot take a federal PLUS loan. What remains is the Direct Unsubsidized Loan, on which interest accrues from disbursement rather than being covered during enrollment, capped at $20,500 a year for most graduate study and $50,000 a year for professional study, with lifetime totals of $100,000 and $200,000 respectively, and those aggregates sit on top of whatever was borrowed as an undergraduate rather than including it. Read the annual figures carefully, because the graduate one is unchanged: only the professional limit rose. Since the PLUS loan it replaces could be taken up to the full cost of attendance and nothing took its place, a graduate student's total federal borrowing capacity is lower than it was, often by a lot. Graduate students have not been eligible for subsidized loans since 2012, and the transitional rule below does not reopen that.
• Which programs count as "professional" is not settled. The $29,500 difference between the two graduate figures turns on a definition the statute pins to a regulation as it stood in July 2025. The Department of Education's attempt to narrow it was stayed by a federal court in June 2026, and the Department is administering an interim list of qualifying programs while the case runs. Nobody should assume a program into the higher figure: ask the school which limit it will originate against.
• Parent borrowing now has hard caps. Federal parent borrowing is limited to $20,000 a year per dependent student and $65,000 in total per dependent student, and both figures are the total across all of that student's parents rather than per parent. The older shorthand that a parent could borrow up to the full cost of attendance stopped being true on the same date.
• There is a lifetime cap per student of $257,500, across all federal borrowing in the student's own name. It excludes parent loans taken on their behalf, which are the parent's debt and sit under the separate caps above.
• A transitional rule protects students who were already partway through, and it suspends all four changes above. A student who, as of 30 June 2026, was both enrolled in a program and had already received a loan for it, or had one made on their behalf, keeps the old rules for the lesser of three academic years or the time left in the program. That means a graduate student in that position can still take a PLUS loan, their old annual limit still applies, their parents are not yet subject to the new caps, and the lifetime cap does not yet bite. It is conditional on staying enrolled and lapses if the student withdraws. So the rules that applied to an older sibling may genuinely not apply to a younger one, and a continuing student should confirm their position with the aid office before concluding that federal borrowing has closed.

Two of those ceilings carry a feature that is easy to miss and changes how they should be planned against. The $65,000 parent aggregate and the $257,500 lifetime cap are computed without regard to any amount already repaid, forgiven or discharged, so paying a loan down does not restore capacity under either. That is a departure from how the undergraduate aggregates work, which are measured against what is still outstanding.

Before deciding how much of that capacity to use, it helps to know how much other families actually use, and to know how the figure in general circulation is put together. Among 2023-24 bachelor's degree recipients, roughly half graduated with no education debt at all, and the average among those who did borrow was under $30,000. Note what that figure leaves out, because all three exclusions push it down: anything their parents borrowed, students who transferred rather than finishing where they started, and for-profit colleges entirely. Across all graduates rather than only borrowers the average is closer to $14,000, and the median across every adult borrower sits in the low twenty-thousands. The much larger figures in circulation are produced by dividing the whole federal portfolio, interest included, by the number of students rather than borrowers, and by folding graduate and parent debt into a number presented as undergraduate. The distribution is the reason a mean misleads here: about a third of borrowers owe under $10,000 and hold 4 percent of the money, while about one in nine owes $80,000 or more and that group holds nearly half of it. Those very large balances sit almost entirely outside undergraduate borrowing, because the undergraduate aggregate limits are too low to reach $80,000: the tail is graduate, professional and parent debt. Those distribution figures are as of mid-2025, on a series that is moving quickly.

When a federal limit binds before the bill is covered, the instinct is to treat the gap as a financing problem. It is better read as information about the choice. The limits are fixed in statute and take no account of what any particular program costs, so a gap above them is a fact about the price of this option rather than a hole to be filled with other credit. Most of the alternatives, a less expensive institution, living at home, transferring credits in, or working part time, remain available for longer than they feel available, and each is reversible in a way that a loan is not. One of them deserves a caveat rather than a recommendation: stretching a degree over more years does lower the annual bill, and it also reduces the return, for the reasons set out under is it worth the money. Finishing late beats not finishing, and finishing on time beats both.

Where the gap is closed with credit, who signs matters as much as the amount. A federal student loan in the student's name is the debt of the person whose earnings the degree is meant to raise. Parent borrowing puts the obligation on someone with fewer earning years left and no claim on the benefit, and it does not disappear if the student leaves without a credential. That is why it competes directly with retirement funding rather than sitting alongside it.

The federal-versus-private choice is not principally about the interest rate. Private student loans are ordinary commercial credit priced on the borrower's or a cosigner's credit history, and what they lack is the set of statutory options attached to federal loans: an income-based plan that sets the payment from income rather than from the balance, and Public Service Loan Forgiveness for borrowers in government or nonprofit work. Note which income-based option a new borrower actually gets. The older income-driven repayment menu is closed to loans first made on or after 1 July 2026; those borrowers choose between a tiered standard schedule and a single new income-based plan. Both belong to the repayment stage rather than this one, and the glossary's student loans section covers them. What belongs here is the consequence: refinancing a federal loan privately converts it permanently into a contract with none of those options, so it is not a decision to take while still enrolled or before knowing whether forgiveness is in play. Our credit and debt guide compares the two kinds of loan in more detail, including the collection powers behind federal debt and the reason student loans are hard to discharge in bankruptcy.

That choice is about to be forced on more people, and graduate and professional students should plan for it rather than assume a substitute exists. In 2024-25, the last full year before the change, about 454,000 graduate students borrowed through the PLUS program that has now closed to them, at an average of roughly $34,000 each, and by design those were borrowings above the unsubsidized limits. Set that against the size of the market expected to absorb them: total nonfederal education lending, meaning every private, state and institutional loan to every undergraduate, graduate and parent borrower in the country, originated less in the same year than graduate PLUS did on its own. That is on the College Board's own estimate of a market it says it measures less precisely than the federal one, which is the right caution to attach and does not change the order of magnitude. Full substitution would require the private market to more than double, for borrowers at the point in their lives with the least income and the most debt. The practical consequences for someone choosing a program now are that private credit may be dearer or harder to obtain than recent experience suggests, that a cosigner is likely to be required, and that the amount a program costs above the federal limits is a live part of choosing it rather than a detail to sort out later.

One asymmetry between those two federal routes belongs in the borrowing decision, because it changes what the options are worth: public service forgiveness is not taxed, while a balance cancelled at the end of an income-driven plan generally is. The detail sits with the repayment stage, but the gap can be large enough to affect which program is worth planning around.

The familiar rule of thumb, that total borrowing should not exceed the graduate's expected first-year salary, is a reasonable starting constraint and worth understanding rather than obeying. Its logic is that a standard ten-year repayment on a balance around one year's income lands near a tenth of gross pay, which most budgets absorb. It breaks in both directions. It is too permissive for a field where the entry salary is a poor guide to the following decade, or where the expected salary is itself uncertain, and too restrictive for a professional degree with a reliable and much higher earnings path. It becomes more useful with two changes: apply it to the whole family's borrowing for that student and not to the student's loans alone, and take the salary figure from data on that specific program rather than from optimism.

What if plans change and the money is not needed?

There are six exits, and knowing them in advance is what makes it rational to fund an account properly rather than under-fund it against a risk that turns out to be small. The fear of being trapped with unusable money is the most common reason families under-save for education, and it is largely misplaced.

• Change the beneficiary. A 529 beneficiary can be changed to another family member, which includes a sibling, a cousin, a parent, and the account owner themselves. This is the simplest answer and it costs nothing. Note that a change to a beneficiary in a younger generation can have gift tax consequences worth checking before making it.
• Do nothing. There is no deadline by which a 529 must be spent. An account can wait for a change of mind, a graduate program a decade later, a professional credential, or a grandchild.
• Withdraw against a scholarship. Where a scholarship covers costs the account was meant to cover, an amount up to the scholarship can come out without the additional 10 percent tax. The earnings portion is still ordinary taxable income, so this is a penalty exception, not a tax exemption.
• Roll to the beneficiary's Roth IRA. Up to a lifetime total of $35,000 for a beneficiary can move from a 529 to that beneficiary's own Roth IRA. The conditions are real and worth reading before relying on it: a fifteen-year account age, earned income in the beneficiary's hands, and the annual Roth limit, which together mean exhausting the full amount takes several years. The 529 term page sets them out.
• Roll to an ABLE account, where the beneficiary or a family member has a qualifying disability.
• Repay student loans, up to a point. A 529 can pay principal and interest on the beneficiary's student loans, or a sibling's, subject to a separate lifetime cap of $10,000 for each person whose loans are repaid. Two things to note: the 2025 tax law raised several 529 limits and left this one alone, so a reader who has absorbed the general expansion will overestimate it; and interest paid this way reduces what can be claimed under the student loan interest deduction, to the extent the money came from the account's earnings.

If none of those fits, a plain non-qualified withdrawal is available and costs less than most people assume. Your own contributions always come back untaxed, because they were made with after-tax money in the first place. Only the earnings portion is taxed as ordinary income, with an additional 10 percent on top, and it is taxed to whoever receives the money. On an account that has not grown dramatically, the real cost of the escape hatch is modest. Check two things first: a state that gave a deduction on the way in may reclaim it, and states that have not adopted every federal expansion of qualified expenses can tax a withdrawal that is federally fine.

Common mistakes

• Planning against the published price. The sticker figure is not what most families pay. Get each college's own net price for your household before drawing conclusions about affordability.
• Budgeting for tuition and forgetting the rest. Housing, food, transport and course materials are frequently more than half of what a year costs, and they are the components a family is most likely to underestimate.
• Skipping the FAFSA because income looks too high. Federal student loans are not need-based and require it, many states and colleges run their own awards off it, and it costs nothing to file.
• Reading October 1 as a deadline. It is the date by which the Department of Education must make the form available. The deadlines that matter are set by states and colleges, and some are much earlier than families expect.
• Treating the Student Aid Index as a bill. It is a number that gets subtracted, not an amount anyone will ask you to pay, and a negative one increases measured need rather than signaling a problem.
• Assuming a full scholarship is always safe. Non-federal grant aid at or above the full cost of attendance now removes Pell Grant eligibility for that period, which is the opposite of what almost everyone expects.
• Funding education ahead of retirement. Education can be borrowed for and retirement cannot, and a retirement shortfall usually returns to the same child later.
• Closing a gap with parent borrowing. A gap above the federal student limits is information about the choice and not a financing problem, and parent debt sits with the person who has the fewest earning years left to repay it.
• Refinancing federal loans privately too early. The conversion is permanent and gives up income-driven repayment and forgiveness, so it is not a decision to make while still in school or before knowing whether either applies.
• Claiming a credit and a tax-free withdrawal on the same dollar. One expense supports one benefit. Allocate deliberately, and remember that tax-free grants reduce the pool of qualifying expenses before anything else is computed.
• Putting education money in the student's own name. A custodial account becomes the student's property at majority and is assessed far harder by the aid formula than a parental asset.
• Assuming state rules follow federal ones. Some states have not adopted every federal expansion of qualified 529 expenses, and a state that gave a deduction can reclaim it on a withdrawal it does not recognize.

When is it worth getting professional help?

Much of this is manageable alone, and the parts that are not have a recognizable shape. Opening a 529 and choosing an age-based option is a half-hour task. Filing the FAFSA is free and the form walks you through it. What is worth paying for is a decision where the amounts are large, the choice is hard to reverse, and the interaction with the rest of a financial plan is where the answer actually lives.

Four moments qualify. The year before the first FAFSA, because the formula looks at income from two years earlier and there are things that can be positioned in advance that cannot be fixed afterward. An offer that requires borrowing beyond the federal student limits, because that is the point at which parent borrowing, private credit and the retirement plan all collide in the same decision. A graduate or professional program, where the sums are larger, grant aid is scarcer and the borrowing rules changed most in July 2026. And any family with a disabled beneficiary, where the education question is inseparable from benefits eligibility and a well-meant account in the wrong name can do real harm.

For any of those, our advisor directory can be filtered for planners who work on education planning, and a single piece of work is often the whole engagement, not an ongoing relationship. Where the issue is existing education debt rather than paying for a coming year, the directory also filters for student loan specialists, and there is a credential specifically for that work, the CSLP®.

Key terms in education funding

Definitions for the terms this guide uses most, each linking to a fuller entry.

Browse all 32 education funding terms in the glossary.

Frequently asked questions