IDEA writes 40 percent of average per-pupil expenditure as the maximum amount a State may receive rather than a floor Congress must reach — the level was not met in any year from 1975 through FY2025

What is happening?

Federal law describes a full funding level for special education as 40 percent of the national average per-pupil expenditure. In FY2025 the appropriation for the Individuals with Disabilities Education Act Part B grants-to-states program equaled about 10.2 percent of that national average, which is just over a quarter of the full funding level, as reported by the Congressional Research Service in a primer carrying an update date of 2026-02-13. In the five decades since the Education for All Handicapped Children Act was signed in 1975, the level has never been met in any single year.

The reason sits in the statute rather than in the conduct of any agency. The subsection of 20 U.S.C. 1411 that carries the formula is headed Maximum amount, and its operative sentence describes the maximum amount of the grant a State may receive. That is a ceiling on what Congress may appropriate, not a floor Congress must reach. Nothing in the section attaches a deadline, a trigger, a report to Congress, an escalator or a sunset to appropriating less than the maximum.

FY2025 shows what such an instrument produces. The year was funded by a full-year continuing resolution approved 2025-03-15 that carried FY2024 purposes and levels forward, so the section 611 appropriation was unchanged in nominal dollars between the two years at USD 14,213,704 thousand, about USD 14.2 billion. The denominator of the share does not hold still. A constant numerator over a growing denominator is a falling share by arithmetic alone, so the federal share of the statutory reference fell in FY2025 without any appropriations decision about IDEA being taken.

Whose problem is this?

The party that would have to raise the federal share is the same party that wrote the 40 percent figure into law as a permission rather than an obligation. No other party can compel it, and the difference lands on whoever is holding the child.

Where does this problem end?

The boundary matters here because the program already exists and the entitlement is already law. What is missing is not a beneficiary and not an appropriation but an instrument that could be violated.

What is the state now, and what should it be?

Now

The target state is written in the statute itself and has not moved since 1975. It is an appropriation equal to 40 percent of the national average per-pupil expenditure times the applicable child count.

The level has been in law for fifty years and the amount appropriated has never come close to it in any year.

Needs a new measurementthe target date: no enacted instrument names a year by which the level would be reached. The single exception was the seven-year authorization schedule written into 20 U.S.C. 1411(i) in 2004, which rose to USD 26,100,000,000 for FY2011 and was estimated to deliver each state its maximum grant in that year. It expired into such sums as may be necessary and was never renewed. Neither the FY2026 budget request nor the enacted FY2026 appropriation names a date, and the federal reference primer names no proposal to set one.

How big is it?

Between 7.5 million and 7.9 million children. The low bound is the NCES count of children ages 3 through 21 served under IDEA in SY2022-23, about 15 percent of all public school students, on a page last updated 2024-05. The high bound is the count CRS cites for SY2023-24, about 7.9 million, roughly 15.9 percent of public PK-12 enrollment, as of the report update date 2026-02-13.

The band is vintage rather than uncertainty. The two counts agree about who is covered and differ by one school year of growth.

Every one of those children is served inside a system whose federal grant equaled about 10.2 percent of the statutory reference in FY2025 rather than the 40 percent ceiling. The shortfall does not select among them, because the entitlement to a free appropriate public education is individual and admits no cost defense and no queue. A district facing a funding gap cannot lawfully answer it by serving fewer eligible children, so the deficit surfaces as budget substitution rather than as denial of eligibility, and the affected count is the served count.

What the count cannot reach: children never evaluated, since identification is the one lever a district holds that this figure cannot see · general-education students in the same buildings, where a fixed local budget is reallocated · local taxpayers, who finance the residual and appear nowhere in a headcount · the distribution across districts of differing wealth, which no opened source publishes · children under 3 served under Part C · duration, because the count is one school year while the shortfall has run for fifty consecutive years and no source has summed it.

Under what conditions does it arise?

1. The obligation is mandatory and the money is discretionary. IDEA entitles every eligible child to a free appropriate public education, individually and without a cost defense. The federal contribution is set by an annual appropriation against a subsection headed Maximum amount. The two instruments are of different kinds, and only one of them binds.

2. The yardstick disappeared after FY2011. The 2004 authorization schedule was the one instrument naming a dollar figure per fiscal year. It ended at USD 26,100,000,000 for FY2011 and lapsed into such sums as may be necessary, leaving a ceiling as the only benchmark. A permission cannot be missed.

3. Standing still is a cut, and standing still requires no decision. FY2025 is the clean case, funded by a continuing resolution that carried FY2024 levels forward. The appropriation did not change and the share fell, with no debate and no recorded vote about IDEA.

4. The ceiling has drifted away from the children it counts. Since FY2007 the maximum grant is computed on the SY2004-05 headcount adjusted by general child population and poverty change, not by the number of children identified as having a disability. The served count grew about 17 percent between SY2012-13 and SY2023-24 by derivation, and none of that growth raises the ceiling.

5. Any increase is discounted at the point where its effect would show. Under 20 U.S.C. 1413(a)(2)(C)(i) a local agency whose Part B allocation rises may reduce its own required local effort by up to half of the increase. An added federal dollar can therefore deliver as little as fifty cents of net new special education spending.

What has been tried?

Two directions have been tried. One writes a larger number into an authorization and the other pushes money through in a single year. The first lapsed on schedule and the second expired by design.

What was found?

Why is it still unsolved?

Institutional gap — the target is drafted as the definition of a maximum, so there is no instrument that could be violated and nothing ever fires.

The first movement is the mismatch between two kinds of instrument. IDEA entitles each eligible child to a free appropriate public education, individually and open-endedly, and that entitlement is enforceable in court. The federal contribution is set annually against a subsection headed Maximum amount, which says what a state may receive and not what Congress shall provide. Appropriating a quarter of the maximum is therefore not a breach of IDEA. It is IDEA operating exactly as written. Nobody is in breach of anything, so nothing fires. The single instrument that ever named a dollar figure per fiscal year was the schedule written in 2004, and after FY2011 it became such sums as may be necessary, which leaves only a ceiling as a benchmark.

The second movement is that the share moves on its own, in the direction of the gap, and the ceiling itself has drifted from the children. Because the numerator is a nominal appropriation and the denominator is a per-pupil measure that is restated each year, holding the appropriation constant lowers the share automatically. The mechanics deserve care here, because a compressed account of them is false: the statutory definition of average per-pupil expenditure at 20 U.S.C. 7801(2) uses aggregate current expenditure from the third fiscal year preceding the determination year divided by children in average daily attendance, so what moves the denominator in a given year is a stale measurement being replaced by a slightly less stale one, not a cost increase in that year. Alongside this, the multiplicand in the ceiling has been frozen since FY2007 at the number of children served in SY2004-05, escalated only by general child population and poverty change. The served count has grown faster than the general school-age population. So even paying 100 percent of the statutory full funding level would no longer correspond to 40 percent of the average per-pupil expenditure for the children actually served, and the percentage everyone quotes carries that detachment invisibly.

The third movement is where the difference lands, and it is the part that makes the arrangement stable. The gap between the cost of the entitlement and the federal grant is absorbed by state and local revenue, and IDEA then makes that absorption a legal ratchet. Section 1413(a)(2)(A)(iii) and 34 C.F.R. 300.203 forbid a local agency from spending less local money than it spent the year before, measured by any one of four methods, with the state liable to repay federal funds from non-federal resources if it fails. A state that reduces its own support has its next grant reduced by the same amount. IDEA thus contains a maintenance-of-effort obligation enforceable against every party except the one whose effort is measured against the 40 percent formula, and it further discounts its own increases through the offset that lets a local agency convert up to half of any rise into local budget relief. That asymmetry is the mechanism, and it is not bad faith by anyone. The reference products are careful, the Department of Education supplies the data, the enforcement machinery functions as written. The failure is in what the instrument is.

This is not primarily a cost-structure failure. The derived gap of roughly USD 41 to 43 billion a year does compete inside discretionary caps, but that explains why any program is not enlarged, and it does not explain why this particular target has gone unmet for fifty consecutive years while remaining largely uncontroversial. Nor is enforcement absent. Enforcement is present, funded and directional. There is simply no rule pointing the other way for it to attach to.

What observation would mean it is solved?

Candidates — (a) an enacted appropriation for a fiscal year that reaches 40 percent of the national average per-pupil expenditure (b) an enacted instrument that converts the level into a floor or a dated schedule (c) the federal share measured against what special education actually costs rather than against the statutory proxy.

(a) alone is a one-year event measured against a drifting ruler. The ceiling it would satisfy is computed on a headcount frozen at SY2004-05, so reaching it would no longer mean 40 percent of the average per-pupil expenditure for the children being served today. It is also reversible by the next appropriations cycle, and the FY2025 case shows that reversal requires no decision at all.

(b) alone has already been tried in the form that failed. The 2004 schedule was exactly such an instrument, naming seven annual figures and an end point for FY2011. Appropriations tracked roughly half of it and the schedule expired without renewal. An authorization is still a permission, so a new schedule would have to differ in kind and not only in the numbers it carries.

(c) alone cannot currently be computed. The last national measurement of what educating a student with a disability costs used SY1999-2000 data and was published in 2002, and no successor collection was found. On that measurement the federal government covered 7.5 percent of total special education spending, and 40 percent of the average per-pupil expenditure amounted to roughly 44 percent of measured excess cost by derivation, so full funding was never half the bill. Without a fresh measurement this observation has no denominator.

The three have to be read together, and none of them alone would settle the question.

What is it connected to?

Fills with researchschool finance litigation over state and local shares, Medicaid reimbursement for school-based health services, the ESEA Title I formula which relies on the same statutory per-pupil definition, Part C early intervention for infants and toddlers, and comparable special education funding gaps in other countries. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• The dollar size of the gap. The federal reference primer publishes the ratio and it publishes the appropriation, and it never multiplies them out. The central number about this problem exists in public only as a percentage. The figure of roughly USD 41 to 43 billion a year given above is derived by this round and belongs to no source.
• What produced the 10.2 percent. The primer states that its calculation rests on data supplied by the Department of Education, including the annual appropriation, the national average per-pupil expenditure for the relevant year, the applicable child count and the set-asides, and then publishes none of the three inputs other than the appropriation. It also publishes the fifty-year series as a figure rather than as a table, so the year-by-year values can only be read off pixels. The headline statistic of this problem is not auditable from any public document.
• That the level is anchored to a headcount from SY2004-05. The statute is explicit and the reporting is not. One CRS product mentions the base year inside a paragraph on allocation mechanics; the product that publishes the percentage does not flag that its full funding level rests on a school year now twenty-one years past. The phrase about 40 percent of average per-pupil expenditure is doing work in public discussion that the statute stopped doing in FY2007, and no opened source says so.
• What the formula denominator means inside IDEA. The IDEA definitions section defines assistive technology service and excess costs and contains no definition of average per-pupil expenditure in public elementary and secondary schools. ESEA defines a like-named term. Nothing opened here states that IDEA uses it, or which value the Department applied in which year.
• The section 611 figure in enacted law. The FY2026 statute appropriates one sum for IDEA and the Special Olympics act together, and the grants-to-states amount that is the numerator of every percentage here appears nowhere in it. For FY2025 not even an explanatory statement exists: CRS records as of 2025-08-06 that none accompanied the full-year continuing resolution and that amounts for numerous programs remain unknown. The year of the most-quoted figure rests on a program allocation that no enacted document states.
• What the 40 percent stands in for. The last national measurement of what educating a student with a disability costs used SY1999-2000 data and was published 2002-09 under a Department of Education contract. Twenty-six years later there is no successor and no national per-student special education expenditure collection. The target has never been re-checked against the thing it targets.
• Where the shortfall lands. The federal share is reported as a single national percentage. No opened source disaggregates the residual between state and local revenue, or by district wealth, or by state. The burden is known in aggregate and unmapped in distribution, and because the district-level literature was not reachable in this round, this silence should be read as not found here as well as not published.
• Any remedy. The authoritative federal primer, updated 2026-02-13, names no bill, no sponsor and no Congress proposing to reach the level, and does not discuss mandatory funding proposals. It records the gap and stops. That same primer, updated ten days after the FY2026 appropriations act was signed, carries no FY2026 figure, so its series ends at FY2025.

See the evidence

Four documents were read end to end rather than summarized. The US Code text of section 1411 was downloaded and extracted locally, which is why every digit of the authorization schedule and the 2004-2005 base-year clause is quoted as read rather than as reported; the enacted FY2026 public law was extracted the same way and searched for the special education account; the FY2026 discretionary request was extracted and read at the special education line; and the 1999-2000 expenditure report was extracted from its ERIC full text. Everything else in the table is second hand, and the two most-cited figures in this document are among the second-hand items: the 10.2 percent share and the section 611 dollar amount both come from a single reference product returned by a page summarizer, and neither was read in an appropriations act. Three disagreements are left visible rather than resolved. First, the enacted FY2026 statute gives its own short title as the Consolidated Appropriations Act, 2026, while two reference products call it the Further Consolidated Appropriations Act, 2026; the statute governs and the discrepancy is in the record. Second, the FY2024 and FY2025 percentages come from two reference products carrying the same update date, and neither states whether its numerator is section 611 alone or section 611 together with section 619, so the two figures are not subtracted anywhere in this document and the direction of travel is established instead from an identical appropriation over a growing denominator. Third, the most-quoted per-pupil expenditure figures in American education use a fall enrollment denominator and a different vintage from the statutory measure, and no opened source states which value the Department applied in which year. The derived quantities — the annual gap of roughly USD 41 to 43 billion, the FY2026 change of about 0.6 percent, the growth of about 17 percent in the served count, and the relation between 40 percent of expenditure and measured excess cost — are arithmetic performed in this round and belong to no source. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 7,500,000–7,900,000 As of SY2022-23 to SY2023-24

Sensitivity The width of this interval is vintage, not uncertainty. The two counts come from different federal sources one school year apart and agree about who is covered, so there is no serious dispute about the denominator of this problem. That makes the absence of a published dollar gap harder to excuse rather than easier. The count is a coverage count and not a harm count: it says how many children are served inside a system whose federal grant equaled about 10.2 percent of the statutory reference in FY2025 against a 40 percent full funding level, and it does not assert that any individual child was denied a service. It is used as the affected population because the entitlement to a free appropriate public education is individual and admits no cost defense and no queue, so a district facing a funding gap cannot lawfully respond by serving fewer eligible children, and the deficit surfaces as budget substitution rather than as denial of eligibility. Six things the count cannot reach. Children never evaluated, because identification is the one lever a district holds that this figure cannot see, and children never evaluated never enter the count. General education students in the same buildings, to the extent a fixed local budget is reallocated; the last national measurement put special education at 21 percent of total K-12 expenditure in SY1999-2000 and no current figure exists. Local taxpayers, who finance most of the residual through property tax revenue and appear nowhere in a headcount. Distribution across districts of differing wealth, since the same federal percentage means different things in a district with a large tax base and one without, and no opened source publishes that disaggregation. Infants and toddlers under Part C, a separate program with a separate appropriation of USD 540 million in FY2025, who sit outside the Part B count entirely. Duration, because the count covers one school year while the shortfall has run for fifty consecutive years and no source has ever summed it.

Regional breakdown No source opened in this round publishes a state-by-state or district-level count of children served against the federal share they receive. The federal share is reported as a single national percentage and the residual is nowhere disaggregated between state and local revenue. Splitting the national count by state population would be proportional allocation, and it would be wrong in both directions at once, because identification rates and the local revenue base that absorbs the residual both vary by state and district. The district-level and state-level literature was additionally unreachable in this round, so this absence should be read as not found here as well as not published.