ESA Spending by Provider Category and State 2026 [Data]

By The School Choice Index Editorial Team · Published August 3, 2026 · Updated and last verified August 13, 2026 · Dataset version 1.1

---

In the 21-program NCSL inventory used for this audit, seven states have public sources with enough data to answer the question ESA spending by provider category and state 2026. Across those seven, the closest thing to a "school payments" line runs from 32.87% in South Carolina to 96.34% in Mississippi. Six of the seven put schools or tuition first.

An education savings account, or ESA, is a state-funded account a family can use for approved K–12 education costs such as private school tuition, tutoring, curriculum and therapy. The rules differ by state.

Now the part that will save you from a bad citation. Those seven percentages are not measuring the same thing. Arizona counts what families bought. West Virginia counts who families paid. South Carolina counts four broad buckets. Utah counts a computer-scored sample. Those are four different questions wearing the same word.

So here is our damaging admission, right at the top: there is no defensible national ESA spending figure in the public data, and we are not going to invent one for you.

If you need a single national number for a headline, this page can't give you one, and neither can anyone else honestly. What we can give you is better: seven exact source numbers, each with its period, its denominator and its limitation attached, plus the arithmetic to reproduce every one. That's the version that survives a fact-check.

We opened every report. We transcribed the tables. We reconciled the totals to the dollar. Then we did the part this release adds — we sorted the money by provider category across states, so you can see how much was reported for tutoring, therapy, curriculum, technology and enrichment, and where each figure stops being comparable.

One preview of what that turns up. In West Virginia, Amazon received $4,194,425.65 in a single school year. That is more than the four largest private schools in the program received combined. It is also more than six times what the four biggest homeschool-curriculum sellers received combined. Full arithmetic below.

A quick terminology note. This page covers state-funded K–12 education savings accounts. It does not cover Coverdell education savings accounts or 529 plans, which are federal tax-advantaged college-savings vehicles that happen to share the name. Tell them apart here.

---

Jump to what you need

• The seven-report comparison
• What "provider category" actually means
• Schools · Tutoring · Therapy · Curriculum · Technology · Enrichment
• Who actually gets paid
• Which states publish this
• How we built it · Download the data

---

ESA spending by provider category and state 2026: the seven-report comparison

Answer capsule: Seven public sources report education savings account spending in dollars by category or payment channel: Arizona, Arkansas, Florida's FES-UA program through Step Up For Students, Mississippi, New Hampshire, South Carolina and West Virginia. The closest identifiable school-payment share ranges from 32.87% to 96.34%. The figures use different periods, populations, administrators and accounting structures, so they are source-specific and cannot be pooled into a national average.

Table 1. Latest public ESA spending by provider category and state

Sources: Arizona FY2026 Q3 ESA Report, Tables 9–10; Arkansas 2024–25 EFA Annual Report, Table 3; Step Up For Students FES-UA fact sheet; Mississippi PEER Report 710, Exhibit 11; Children's Scholarship Fund New Hampshire 2024–25 EFA Expenditure Overview; South Carolina ESTF Year-One Report, Table 9; West Virginia Hope Scholarship 2024–25 Annual Report. Percentages calculated from unrounded source amounts. Verified August 13, 2026.

Freshness check: As of August 13, 2026, Arizona's official quarterly-report page still listed FY2026 Q3 as the latest FY2026 report. FY2026 Q4 had not posted.

These seven snapshots contain $1,166,988,907.62 in reported spending. That's what this release covers. It is not a national ESA total, and please don't use it as one — Arizona is one quarter, Florida is one administrator, and the years don't line up.

What this table can prove: each source's total, its biggest line, the exact share inside that source, the period, the counting unit, and the caveat that belongs beside it. Plus one narrow cross-source finding — schools or tuition lead six of seven reports.

What it can't prove: a national average, a ranking of what families prefer, a trend line, or a "typical" ESA family's budget. It can't turn a quarter into a year, one administrator into a state, a payment channel into a purchase category, or a draft into a final report.

Those aren't nitpicks. Each one changes the answer.

---

What does "provider category" actually mean in an ESA spending report?

Answer capsule: "Provider category" has no single meaning across state ESA reports. A row may identify who received the payment, what was purchased, which payment system processed it, a broad bucket, a percentage with no dollar table, or a computer-scored sample. This page sorts every source into one of those six reporting lenses before comparing anything.

This is the part almost everyone gets wrong, so it's worth two minutes.

When Arizona says "Tutoring and teaching services," that's a purchase. When West Virginia says "Payments to Education Service Providers," that's a recipient. Both look like categories. Stack them in the same column and you've quietly built a chart that means nothing.

Here are the six lenses, and which sources use each.

Table 2. The six reporting lenses in public ESA spending data

Source: The School Choice Index classification of each source's tables, footnotes and definitions. Verified August 13, 2026.

Two terms worth defining, because they show up constantly:

• A microschool is a tuition-charging school initiated by one or more teachers or by an entity created to operate a school. West Virginia defines it in state code and reports its payments separately.
• A payment channel is the plumbing — a marketplace, a direct payment to a school, or a reimbursement to the parent. It tells you the route, not the purchase.

The rule we follow, and you should too: a normalized label is for finding things. It is never permission to add the dollars together.

---

How much ESA spending goes to private schools and microschools?

Answer capsule: School and tuition payments are the largest identified line in six of seven public ESA spending reports, with source-specific shares of 32.87% in South Carolina, 52.46% in Arizona, 56.88% in New Hampshire, 66.11% in Florida's FES-UA program, 72.94% in West Virginia, 80.88% in Arkansas and 96.34% in Mississippi. The spread reflects program design, population and reporting method rather than a simple ranking of family preference.

The definitions do more work here than the percentages do.

Arizona's 52.46% is the narrow measure. It counts one source category: tuition, textbooks or fees at a qualified school. Arizona separately reports online-learning tuition and eligible-postsecondary tuition. Add all three and the broader tuition-related measure is 59.00%. That broader number is our calculation, not an Arizona category — we say so every time it appears.

West Virginia's 72.94% leaves out microschools. Add the $2,805,903.26 that went to microschools and the direct-to-school share is 79.35%.

South Carolina's 32.87% is not normal. On September 11, 2024, the state Supreme Court halted ESTF tuition and fee payments to private schools and nonpublic online providers mid-year. The report says so itself, and says independent-school tuition would have been a larger share without the decision. The number is correct. It's just correct about an interrupted year.

Mississippi's 96.34% comes from a $2.56 million program. It's a targeted special-needs program with 397 participating students in FY2024. Valid number, terrible national proxy.

The single clearest look at who spends what

Among the public reports in this audit, Arkansas is the only source that splits its spending table by educational pathway. That makes it the most useful row in the entire dataset, because it shows two completely different spending patterns living inside one program.

Source: Arkansas 2024–25 EFA Annual Report, Tables 4 and 5. Shares calculated from source dollar amounts; both totals reconcile exactly. Verified August 13, 2026.

Read those two columns side by side and the "where does ESA money go?" question basically answers itself. Private-school families spend on tuition, full stop — 97 cents of every dollar. Homeschool families spend on supplies, curriculum and program fees, and their single largest line is 39% supplies.

A statewide table that doesn't split these pathways averages two different populations into one number that describes neither group well.

One precision note. The Arkansas report rounds both Curriculum and Homeschool Tuition & Fees to 21% for homeschool families. Computed from the dollars, Homeschool Tuition & Fees is 21.49% and Curriculum is 20.83%. If you need them in order, use the dollars.

The two pathway tables sum to $84,524,869 — $18,643 more than the all-student table. The report explains it: 29 refund transactions belonged to students who withdrew mid-year and couldn't be assigned to a pathway. The reconciliation is exact.

---

How much ESA spending goes to tutoring and teaching?

Answer capsule: Tutoring is separately reported in five sources but on very different denominators: 15.23% of Arizona's quarterly spending, 5.01% among Arkansas homeschool families, 2.39% of Utah's audited subset, 0.77% in Mississippi, and 13.72% of West Virginia's education-service-provider channel — which is only 0.65% of that program's total spending. Comparing the headline percentages without the denominators produces a gap that isn't real.

Table 3. Tutoring and teaching spending, by source

Sources: the Arizona, Arkansas, Mississippi, Utah and West Virginia reports cited on this page. Shares calculated by The School Choice Index from source amounts. Verified August 13, 2026.

Look at the two West Virginia rows. Same $283,412.05. One reads as 13.72%, the other as 0.65%. Both are correct. That's what a nested denominator does to a statistic, and it's the single easiest way to accidentally publish something wrong.

The Arizona–Arkansas gap, and what actually explains it

Arizona's tutoring share is 15.23%. Arkansas's is 1.07%. That's a 14.2× difference between two states in the same year, and it looks impossible until you read the category definitions.

Arizona's source label is "Tutoring and teaching services," which is broader than Arkansas's "Tutoring" row. Arkansas separates paid instruction at academies, microschools, co-ops and learning centers into "Homeschool Tuition & Fees," which the report says explicitly covers those settings.

So we ran the adjustment. Arkansas tutoring ($904,650) plus homeschool tuition and fees ($3,151,409) equals $4,056,059, or 4.80% of Arkansas spending.

The gap narrows from 14.2× to about 3.2×. It doesn't close. Label scope explains part of it; program rules, populations and use still differ.

That's the kind of check that separates a usable comparison from a viral chart.

---

How much ESA spending goes to therapy and specialized services?

Answer capsule: Therapy spending is visible in five sources and ranges from 4.49% of Florida's FES-UA program to 0.13% of West Virginia's Hope Scholarship. The highest shares appear in programs built around students with disabilities, but program size and design matter more than that pattern alone suggests.

Table 4. Therapy and specialized-service spending, by source

Sources as cited on this page. Florida figure is a School Choice Index calculation summing four separately published rows: Speech Language Therapist $13,706,826.13, ABA Therapist Services $8,846,893.40, OT Therapist Services $6,970,875.57, and Other Therapist Services $5,941,963.52. Verified August 13, 2026.

The rough pattern holds — programs designed around disability fund more therapy. Florida's FES-UA serves students with unique abilities and shows the highest share. Arizona has a disability tier with much larger awards and comes second.

But we're not going to pretend it's clean. Mississippi's program is special-needs-only and its therapy share is 0.71%, below Arizona's. That's because Mississippi families overwhelmingly spend on tuition — 96.34% of a very small pot — which leaves almost nothing for anything else. Program size and spending rules bend the pattern.

Here's the finding worth holding onto: in Arizona, every disability-specific category combined — therapies, associated goods, life-skills tuition, paraprofessionals, assistive technology and evaluations — comes to $7,986,346, or 3.62% of quarterly spending. That's against a program where 21,264 of 102,891 students, or 20.7%, are in the disability tier.

One in five students. Under four cents of every program dollar sits in categories reserved for disability-related spending. The tables do not show where disability-tier students spent the rest.

A caution before anyone runs with that. A low category share does not mean low need. It can mean the therapy is delivered by the child's school and billed as tuition. It can mean private insurance or a public program paid first. It can mean the paperwork is hard. Spending data cannot tell you which, and neither can we.

---

How much ESA spending goes to curriculum, books and supplies?

Answer capsule: Curriculum and materials are the second-largest spending area in most public ESA reports, ranging from 22.85% in New Hampshire's draft report and 19.15% in Florida's FES-UA program down to 1.99% in Mississippi. The category is split across different labels in every state, so combining rows requires checking that the source definitions don't overlap.

Table 5. Curriculum, books, materials and supplies, by source

Sources as cited on this page. Combinations are School Choice Index calculations from non-overlapping source rows; each is listed so you can rebuild it. Verified August 13, 2026.

Arkansas again shows the split that matters. Curriculum is 20.83% of homeschool spending and 0.22% of private-school spending. Supplies is 38.97% versus 2.04%. Averaging those into one state number tells you almost nothing about either group.

Arizona shows the opposite problem — a category that changes size depending on which table you open. Supplemental materials is 34.90% of Arizona's marketplace dollars and 7.62% of all Arizona dollars, because the marketplace is only 11.90% of the money. Same category, same quarter, two very different-looking numbers. More on why in the Arizona section below.

---

How much ESA spending goes to technology?

Answer capsule: Where technology is reported separately, program-wide shares run from 0.02% in Mississippi to 5.20% in West Virginia; Utah reports 6.46% inside its scored subset, not the full program. In every state that publishes a program-wide figure, technology spending sits below 10% — but these are program averages, and a per-child spending cap can still bind on individual families.

Table 6. Technology and device spending, by source

The West Virginia total sums these TheoPay rows: Computers $887,984.04, Tablets/iPads/E-Readers $396,642.08, Printers and Printer Supplies $337,129.54, Computer Supplies and Accessories $325,788.70, Audio & Visual $119,834.25, Headphones and Accessories $116,516.65, Keyboards and Accessories $43,380.41, Cameras and Accessories $19,821.93, Software $16,799.66, Calculators $7,337.89, Computer Monitors $4,817.53. School Choice Index calculation. Verified August 13, 2026.

Would a 10% technology cap actually bind?

This is the policy question behind the Texas cap and Utah recommendation, so let's answer it carefully.

Texas caps technology at 10% of a child's account for its new Education Savings Account Program. Utah's auditor recommended lawmakers consider caps on items like computers. So the natural question is whether a 10% ceiling would change much.

In every state that reports technology separately, program-wide technology spending is below 10% — 3.78% in Arizona, 4.78% in New Hampshire's draft, 5.20% in West Virginia, 0.02% in Mississippi.

Now the caveat that has to travel with that, and we mean it. A cap applies to one child's account. These are program-wide averages across tens of thousands of accounts. An average of 5% is completely compatible with thousands of individual families hitting a 10% ceiling. Utah's auditor reported 337 electronics transactions over $1,000 and 87 over $2,000. The report says those groups together accounted for $600,847, or about 54% of the scored subset's electronics and hardware dollars. The report separately noted that about 28.75% of devices in Utah public K–12 schools were non-Chromebooks. Those are different measures, but both show what an average can hide.

If you need to know whether a cap binds, you need per-account data. None of these reports publish it. Anyone telling you otherwise is reading an average as a distribution.

---

How much ESA spending goes to enrichment, sports and the arts?

Answer capsule: Enrichment spending is reported inconsistently and is usually small as a share of total program dollars — 1.50% in Arkansas and about 1.31% of West Virginia's program — but it reaches 15.36% inside Utah's scored subset, which is not a random sample. Arizona does not publish a general enrichment category at all.

This is the section people arrive at after reading a news story about a ski pass or a mountain bike. Fair enough. Here's the data.

Table 7. Enrichment, athletics and arts spending, by source

School Choice Index calculations from the named source rows. Verified August 13, 2026.

So which number is true? Both, for different questions.

Across whole programs, enrichment is small — one to two cents on the dollar in the states that separate it. But Utah's 15.36% is real too, and higher, and here's why: the auditor's subset was selected, not sampled at random. It's the transactions the model could label with at least 90% confidence. That makes 15.36% a real subset result, not a full-program share.

We report both and label both. What we won't do is pick the scarier number and call it the finding.

One thing worth noticing quietly: in Arkansas, enrichment is 7.99% of homeschool spending and 0.15% of private-school spending. It's a homeschool line, not a program-wide one. That's a 53× difference between two groups inside one state.

---

Who actually receives ESA money?

Answer capsule: In West Virginia, the source with the fullest combination of channel, category, vendor and provider-location detail, Amazon received $4,194,425.65 in 2024–25 — 64.07% of goods-marketplace spending and 9.58% of all program spending. That is more than the four largest participating private schools received combined, and more than six times what the four largest homeschool-curriculum sellers received combined.

Category tables tell you what got bought. They don't tell you who got paid. New Hampshire's draft also names recipients. West Virginia is the source that connects named payees to payment channels, category sub-tables and provider-state geography, so this section leans on it hard and says so.

Where the goods money went

West Virginia families buy physical goods through TheoPay, an open online marketplace facilitated through Student First Technologies' TheoPay application. Total 2024–25 TheoPay spending: $6,546,970.94.

Source: West Virginia Hope Scholarship 2024–25 Annual Report, TheoPay vendor and category tables. Groupings and shares calculated by The School Choice Index. Verified August 13, 2026.

Amazon alone received 6.3 times what the four biggest homeschool-curriculum sellers received combined.

Put it against the schools and it's starker. The largest single private school in West Virginia's program, Cross Lanes Christian School, received $1,004,246.09. Amazon received 4.18 times that. Add up the top four schools — Cross Lanes Christian, Faith Christian Academy, Covenant School and Heritage Christian School — and you get $3,805,396.27. Amazon still comes out ahead, by $389,029.38.

What this does and doesn't show. Amazon is one company competing against hundreds of individual schools and small vendors, so a concentration finding is partly just arithmetic about market structure. It's also a fair reading of where much of the goods-marketplace money lands: at general retailers, not only at education publishers. Both readings are true. We'd rather give you both than let someone else supply the second one.

For context, New Hampshire's draft report shows a similar pattern — Amazon received $2,941,744.35, making it the single largest named recipient there too, ahead of any school. Two states, two different administrators, two different platforms, same answer.

One more detail worth ten seconds

West Virginia also discloses that $659,405.09 of TheoPay spending, or 10.07%, is "Unclassified." The report explains why plainly: the field used to categorize purchases didn't exist for the first two months of the school year, and staff reconstructed categories where they could, partly using manufacturer commodity codes.

That's a state telling you where its own data stops. More reports should do it.

---

How much ESA money leaves the state?

Answer capsule: West Virginia is the only source in this audit publishing where its providers are located, and the answer depends entirely on the provider type: 99.39% of microschool payments and 94.50% of nonpublic school payments went to in-state recipients, but only 67.27% of education-service-provider payments did. About one-third of West Virginia's service-provider money went to providers in other states.

Table 8. Share of West Virginia Hope Scholarship payments going to in-state providers, 2024–25

Source: West Virginia Hope Scholarship 2024–25 Annual Report, payment-distribution tables. Shares as published. Verified August 13, 2026.

There's a clean gradient in this page's calculation: the more school-like the provider, the more the money stays home. Buildings don't move. Tutors, online academies and virtual programs do.

Where the service money went instead: Nevada took 7.72% of all education-service-provider dollars — almost entirely one online academy. Ohio took 4.93%, Pennsylvania 4.27%, Texas 2.60%, Virginia 2.50%, Iowa 2.13%.

One line in that appendix is worth a second look. Steubenville City Schools, a public school district in Ohio, received $100,750 in West Virginia Hope Scholarship payments. That's public education money from one state landing in a public district in another, through a private-school-choice program. Current West Virginia code does not require participating schools or education service providers to be located in the state, and this is the kind of flow that only shows up when a state publishes provider locations.

House Bill 5249, introduced February 5, 2026, would add an in-state requirement for private schools and education service providers. As of August 13, 2026, it remained pending in the House Education Committee. This table is the evidence base for that debate, and it says the answer depends heavily on which provider type you're talking about.

---

Which ESA programs publish spending by provider category?

Answer capsule: The National Conference of State Legislatures' July 21, 2025 inventory lists 21 directly state-funded ESA programs across 18 states. Under the documented search protocol, seven publish category-dollar tables, one publishes a limited subset analysis, one publishes a percentage without a dollar table, and no current category-dollar table was located for the remaining 12.

Table 9. Public spending-disclosure status for the 21 programs in NCSL's July 21, 2025 inventory

Program universe: NCSL, Education Choice State Policy Scan: Education Savings Accounts, updated July 21, 2025. Evidence status: The School Choice Index search described in the methodology. "No current category-dollar table located" means none was found in the named public-source classes by August 13, 2026; it does not prove no responsive record exists.

No single source contains this table. NCSL supplies the program list. Agencies, evaluators, administrators and auditors supply the evidence. The status column comes from checking them one at a time and refusing to call a rulebook a spending report.

Two things stand out. We located no current category-level spending table for Florida's universal FES program — the FES-UA figures on this page come from one administrator's fact sheet for the disability-focused program. And Arizona and West Virginia publish two of the most detailed final reports; New Hampshire's similarly detailed source remains a draft. Size doesn't predict transparency here. Reporting law helps: Arizona's quarterly reporting exists because A.R.S. § 15-2406 requires it.

---

What does each state's report actually measure?

Answer capsule: Each ESA report has a different stopping point. Arizona publishes complete dollars across two payment routes but incompatible count units; Arkansas is the only source splitting spending by educational pathway; Florida publishes top categories only; New Hampshire's report is a draft; West Virginia's headline table is a payment channel rather than a purchase category.

Arizona: why the report doesn't contain the number people quote

Arizona's report, filed under A.R.S. § 15-2406, ends with two spending tables. Table 9 covers ClassWallet Marketplace purchases, recorded item by item. Table 10 covers Pay Vendor, reimbursement and debit-card spending, recorded order by order — and an order can hold several items.

The two tables tell opposite stories. In Table 9, supplemental materials is the biggest line at 34.90%, and qualified-school tuition is 0.09%. In Table 10, qualified-school tuition is 59.53% and supplemental materials is 3.94%.

Quote either table by itself and you're quoting a payment-route subset, not the program.

Both reconcile exactly: $26,266,939 + $194,390,410 = $220,657,349.

We tested Arizona's own headline. The department says about 60% of Q3 spending went to tuition, textbooks or fees at a qualified school, online program or eligible postsecondary institution. We ran that same three-category calculation on every possible denominator:

Only the combined-dollar basis gives 59.00%, which rounds to "about 60 percent." That makes the combined reading reproducible and well supported. It's still our inference from two tables plus a highlight, not a third table Arizona published.

The Arizona arithmetic in full, so you can rebuild it:

• Combined spending: $26,266,939 + $194,390,410 = $220,657,349
• Qualified-school tuition: $23,269 + $115,725,292 = $115,748,561 → 52.46%
• Online-learning tuition: $811,787 + $11,987,644 = $12,799,431
• Postsecondary tuition: $12,969 + $1,626,028 = $1,638,997
• Broader tuition measure: $115,748,561 + $12,799,431 + $1,638,997 = $130,186,989 → 59.00%

One more Arizona figure the source does not publish. The average marketplace item cost $50.66 ($26,266,939 ÷ 518,486 items). The average order on the other three routes cost $599.35 ($194,390,410 ÷ 324,335 orders). That's an 11.8× difference — and it's exactly why the marketplace generates most of the transactions and a small share of the money. Note the units differ; we are comparing item averages to order averages, and we don't add the counts.

Table 10. Arizona ESA spending by category, FY2026 Q3, with prior-year comparison

Sources: Arizona FY2026 Q3 and FY2025 Q3 ESA reports, Tables 9 and 10. Combined columns and percentages calculated by The School Choice Index. Verified August 13, 2026.

Two changes deserve attention. "Multiple categories selected" grew 5.4×, from $1,369,395 to $7,429,121. The label doesn't mean anything was improper — it means the order couldn't be assigned to one category. And "Fees to manage an ESA" disappeared entirely; it carried $1,112,503 a year earlier. Any multi-year series has to preserve category additions and removals rather than quietly forcing them into a stable list.

Florida: nine categories and a residual

Step Up For Students reported $789,122,967.25 in 2023–24 FES-UA spending — the Family Empowerment Scholarship for Students with Unique Abilities, Florida's disability-focused ESA. Private-school tuition was 66.11%; curriculum and instructional materials 19.15%.

The nine published categories account for 97.42% of the stated total. That leaves $20,321,118.94, or 2.58%, that the fact sheet doesn't name. We publish the residual as a calculated line rather than pretending the nine categories are the whole ledger.

Transcription note. The source prints Part-Time Tuition & Fees as "$32,485,26807," with no decimal. Reading the last two digits as cents gives $32,485,268.07, which matches the format of every surrounding figure and leaves a positive residual against the stated total. We record the normalized amount and the reason rather than silently fixing it.

Two limits stay attached: Step Up For Students is the named administrator and data producer, and the document publishes "top categories," not a complete ledger.

Mississippi: small, targeted, and almost entirely tuition

Mississippi's FY2024 reimbursement table totals $2,563,448 across 397 participating students. Tuition and academic fees are $2,469,608, or 96.34%. Every other expense type is under 1.25%.

The report also says 63% of participants used funds only for tuition, and 37 participants — about 9% — used tutoring or therapy in addition to tuition. Those are participant counts, not dollar shares. Don't swap one for the other.

New Hampshire: detailed, and a draft

New Hampshire's current expenditure overview reports $20,221,069.81 in 2024–25 spending and $11,501,321.84, or 56.88%, for private-school tuition. The document is visibly labeled DRAFT on its summary and data pages.

It's unusually useful anyway — the private-school tuition amount and percentage are both exact in the detailed appendix. But it isn't internally consistent everywhere. The summary pie labels online and virtual tuition as 3.3%; the narrative says about $637,000, or 3.1%; and the detailed appendix gives $636,612.08, or 3.15%. We use the exact appendix amount when needed and use neither rounded percentage in the headline comparison.

Every New Hampshire number on this page keeps the DRAFT label until a final report replaces it.

South Carolina: the year a court changed the answer

South Carolina's 2024–25 Year-One Report totals $6,151,754.09 across four broad groups:

This is the only source in the comparison where school tuition isn't the largest line. The report explains why: the South Carolina Supreme Court's September 11, 2024 Eidson decision halted ESTF tuition and fee payments to private schools and nonpublic online providers. The report states independent-school tuition would have been a larger portion without it.

The $2,022,102.74 is accurate. It's just court-interrupted. Anyone using South Carolina as evidence of normal first-year family preference is using a correct number for the wrong claim.

That was the 2024–25 rule set. The current program is governed by Act 11 of 2025, so this table is a historical spending record, not the current rulebook.

West Virginia: complete by recipient, partial by purchase

West Virginia reported $43,766,093.47 across five payment channels:

Read that last column carefully. West Virginia publishes purchase detail for 20.65% of its money and none for the other 79.35%, because direct school payments are reported as one line item here. That's not a criticism. It's just the reason a channel table can't be relabeled a category table.

Inside the education-service-provider channel, here's the full 13-row breakdown, published with both denominators so nobody misreads it:

Source: West Virginia Hope Scholarship 2024–25 Annual Report, Education Service Provider Payments by Category. Program-share column calculated by The School Choice Index. Verified August 13, 2026.

One source typo, resolved by arithmetic. The report prints Arts Instruction as "224,752,91" — a comma where a decimal belongs. Reading it as $224,752.91 makes all 13 rows sum to exactly the stated $2,065,869.69. We publish the corrected reading and the reason.

Utah: a scored sample, not a ledger

The Utah State Auditor reviewed 212,856 Utah Fits All transactions totaling $67,733,684, then assigned categories to a high-confidence subset of 41,070 transactions totaling $17,139,736. That's 19.3% of transactions and 25.30% of dollars.

Source: Office of the Utah State Auditor, limited review of the Utah Fits All Scholarship Program, issued June 26, 2026. Shares use the $17,139,736 subset denominator. The auditor warns the results cover only part of the population and that a transaction can receive more than one category. Verified August 13, 2026.

The auditor's asterisk marks categories that were allowed under the pre-2025 guidance but became prohibited or limited for 2025–26 and later. The subset describes the year reviewed, not today's expense rules.

Tuition is 62.66% of the scored subset. It is not 62.66% of the $67,733,684 program. The subset wasn't randomly drawn, and overlapping categories mean these amounts shouldn't be added as if they were exclusive.

Utah stays out of the seven-report comparison for that reason. Including it would quietly convert a scored sample into a full category table — which is the exact error this page exists to prevent.

Tennessee and Iowa: what rules do to the numbers

Tennessee's January 2026 pilot evaluation reports that students spent 95% of ESA funds on participating-school tuition in each of the program's first three years. Useful, and narrower than it looks — it's a repeated percentage from one targeted pilot, with no dollar ledger. The same evaluation reports students spent an average of 90% of their total ESA funds across three years, which is a different denominator entirely.

Iowa's program page says funds must be used for tuition and fees first, before remaining money can go to marketplace purchases. So a tuition-heavy Iowa distribution wouldn't mean the same thing as a tuition-heavy distribution in a program with no sequencing rule.

That's the principle behind half this page: spending percentages describe the rules and the payment system at least as much as they describe the families. They are not a preference survey.

---

Why can't these ESA spending numbers be added together?

Answer capsule: The same activity can be recorded as an item, an order, a transaction, a reimbursement, a provider payment or a school payment. The programs also use different periods, category lists, refund treatments, populations and disclosure levels. A shared word like "tutoring" does not create a shared variable.

Here's the short version of what breaks:

Different periods. Arizona is one quarter. Florida is 2023–24. Arkansas, New Hampshire, South Carolina and West Virginia are 2024–25. Mississippi is FY2024. No trend line survives that.

Different populations. Universal programs, income-limited programs, special-needs programs, first-year programs, court-interrupted programs.

Different reporting entities. A state department, a university research center, a legislative evaluator, a scholarship administrator, a state auditor.

Different counting units. Arizona's marketplace counts items; its other routes count orders. Arkansas counts transactions. West Virginia counts payments. Utah counts scored transactions.

Different refund treatment. Arkansas explicitly subtracts refunds and cancellations as a negative row inside its total. The other six main tables do not display a comparable negative row.

Where each source stops

Sources as cited throughout. Verified August 13, 2026.

We deliberately do not roll these into a single "data quality score." Item-level detail, named recipients, categorized dollars and top-category publication are different properties. Mashing them into one percentage would invent a statistic with no stable meaning — the exact thing this page warns against.

Comparisons we treat as fair: each source's biggest line, exact within-source percentages from a named denominator, whether a source is final or draft, whether subgroup data exist, whether the complete table is organized by purpose or by payee, and how much each source leaves mixed or unclassified.

Comparisons we don't publish: a national category share, a ranking of state family preferences, a pooled average built from quarterly and annual and administrator-specific and court-interrupted figures, transaction counts that add items to orders, a full-program Utah share pulled from a scored subset, or any outcome claim based on what was purchased.

Refusing to calculate a national average isn't a missing feature. It's the main result.

---

How was this ESA spending dataset produced?

Answer capsule: We searched the 21 programs in NCSL's July 21, 2025 inventory for public documents reporting completed expenditures or payments, transcribed the source labels and dollar amounts, recalculated exact shares from named denominators, and reconciled every table used. Derived measures, draft evidence, percentage-only evidence and scored subsets are labeled separately rather than blended.

What gets into the main comparison

A program enters only when all three are true:

1. A public source reports completed spending or payments.
2. The source names at least one category or channel and gives a total.
3. The amounts can be transcribed or reconciled without guessing.

Lists of legally eligible expenses are not spending data. Participant counts, awards and appropriations are not spending by category. Percentage-only claims are kept as context, outside the dollar comparison.

How we searched

For every program in NCSL's July 21, 2025 inventory, we checked the administering agency or program site, annual-report repositories, legislative and oversight evaluations, auditor publications, and the original administrator where it produced the data. Search terms paired each program name with "annual report," "quarterly report," "expenditures," "spending," "transactions," "payments," "category" and "financial report."

"No current category-dollar table located" means that process didn't find one by August 13, 2026. It doesn't establish that no internal record exists, that a public-records request would fail, or that the NCSL inventory includes every program enacted after its July 2025 source date.

How we calculated

• Source category names are preserved exactly in the dataset.
• Shares use the source dollar amount divided by the named source denominator.
• Arizona dollars are combined across Tables 9 and 10 only where labels match. Counts stay separate.
• Arkansas percentages use the $84,506,226 transaction total, not the separate $93.8 million program-cost figure.
• Florida's residual is the source total minus the nine published categories.
• New Hampshire's exact figures come from the draft appendix; rounded summary figures stay labeled approximate.
• West Virginia payment channels are never renamed as expense categories.
• Utah's subset stays separate and is never generalized to the full population.
• Every multi-row combination on this page lists the rows we added.

Reconciliation checks

• Arizona: $26,266,939 + $194,390,410 = $220,657,349 ✓
• Arkansas: Table 3 sums to $84,506,226; Tables 4 and 5 exceed it by the footnoted $18,643 ✓
• Florida: nine categories + $20,321,118.94 residual = $789,122,967.25 ✓
• Mississippi: expense types sum to $2,563,448 ✓
• New Hampshire: detailed appendix ends at $20,221,069.81 ✓
• South Carolina: four rows sum to $6,151,754.09 ✓
• West Virginia: five channels sum to $43,766,093.47; the 13-row ESP table sums to $2,065,869.69 ✓

---

What we actually verified

Answer capsule: This release transcribes and reconciles public documents from eleven named issuers, recalculates every published percentage from source dollar amounts, and labels every derived figure. It does not include a national estimate, an outcome claim, or any figure we could not trace to a named table.

Who made this. Researched, calculated and written by The School Choice Index Editorial Team. The School Choice Index is an independent comparison and research resource for U.S. school choice programs.

No external subject-matter expert reviewed this release, and we don't claim one did.

What we checked, source by source: report issuer · source period · table or exhibit number · reported total · original category labels · exact dollar amounts · denominator · calculated share · draft or final status · scope limitation · whether the source link resolves · whether the download matches the visible table.

What we did not verify and do not claim: every responsive record that might exist outside public repositories · family intent or preference · provider revenue or profitability · national market share · fraud or misuse based on a category label alone · whether any purchase was educationally appropriate · what will happen next year.

Corrections. Found a source we missed or a total that doesn't reconcile? Send the report title, issuing entity, table or page number, and the correction through our corrections page. Material corrections go in the changelog. We don't quietly overwrite.

---

What are the limits of this data?

Answer capsule: This is a seven-source comparison and a 21-program disclosure audit, not a national transaction census. Periods, populations, administrators and accounting structures differ; one source is a draft, one major row belongs to a single administrator, South Carolina was court-interrupted, and Utah is a scored subset.

Source-dated universe. The 21-program audit uses NCSL's July 21, 2025 inventory and needs refreshing when NCSL updates it.

Period mismatch. Arizona is a quarter. Florida is two years back. The rest are 2024–25 or FY2024.

Scope mismatch. Florida's figure belongs to one administrator. Mississippi is a targeted special-needs program. Utah's analyzed subset covers 19.3% of transactions. Tennessee gives a percentage, not a ledger.

Document status. New Hampshire's report is labeled DRAFT, and its internal online-tuition discrepancy stays visible rather than buried.

Classification. Source systems assign categories through family selection, provider information, order-level labels, reconstructed fields or model scoring. Every percentage inherits the limits of whichever method produced it.

What none of this shows. These reports don't establish whether a purchase helped a student, whether a family would have bought the same thing without the program, how much eligible money went unspent, or whether one category is a better use of funds than another.

No national average. The seven rows are never pooled. The $1.167 billion figure describes this release's coverage, not national ESA spending.

---

Frequently asked questions about ESA spending by provider category

Answer capsule: The core distinction is between what an ESA program allows and what its reports show families actually paid for. Eligible-expense guides describe possible purchases; this dataset uses completed-expenditure evidence and keeps percentage-only, rules-only and subset-only evidence in separate tiers.

Is this about Coverdell education savings accounts?

No. This covers state-funded K–12 school-choice education savings accounts. Coverdell ESAs and 529 plans are privately funded, tax-advantaged accounts under federal tax rules. Full comparison here.

What percentage of ESA funds goes to private-school tuition?

There is no defensible national percentage. The source-specific school-payment shares are South Carolina 32.87%, Arizona 52.46%, New Hampshire 56.88%, Florida FES-UA 66.11%, West Virginia 72.94%, Arkansas 80.88% and Mississippi 96.34%. Definitions and periods differ across all seven.

What is the largest ESA spending category?

Tuition or direct school payments lead six of the seven category-dollar sources. South Carolina's court-interrupted first year is the exception, with parent-directed tools, supplies and services at 60.80%.

What is a "provider category" in an ESA report?

It depends on the state. It may describe who received the payment (a school, a microschool, a service provider), what was purchased (tutoring, therapy, curriculum), how the money moved (marketplace, reimbursement), or a broad bucket. This page sorts every source into one of six reporting lenses before comparing anything.

Why does Arizona have two tuition percentages?

The narrow 52.46% counts one source category — tuition, textbooks or fees at a qualified school. The broader 59.00% adds online-learning and eligible-postsecondary tuition. The broader figure is our calculation, with the formula published.

Why is West Virginia's biggest number called a payment channel?

Because the complete table groups money by recipient route — nonpublic schools, microschools, service providers, TheoPay and reimbursements. It doesn't provide one complete, mutually exclusive purchase-purpose table for all dollars.

How much ESA money goes to Amazon?

In West Virginia, Amazon received $4,194,425.65 in 2024–25 — 64.07% of goods-marketplace spending and 9.58% of all program spending. New Hampshire's draft report shows Amazon as its largest single recipient as well. Most states don't publish vendor-level data at all.

Does Utah publish spending by category?

Only partly. The State Auditor published a limited model-based analysis of a scored subset — 41,070 of 212,856 transactions. It's real evidence, but it isn't a full-program category table and its categories can overlap.

Do ESA programs spend much on therapy?

Where it's reported, therapy runs from 4.49% of Florida's disability-focused FES-UA program down to 0.13% of West Virginia's Hope Scholarship. A low share doesn't prove low need — therapy may be delivered by a child's school and billed as tuition, and insurance or another public program may pay first.

Is there a national average for ESA spending by category?

No. Pooling would combine different periods, populations, administrators, category systems and counting units into a figure that looks precise but holds no consistent meaning.

How often is this page updated?

Arizona is checked when a new quarterly report appears. Annual and administrator sources are checked when new reports are posted. The 21-program audit is refreshed when NCSL or a program source changes. The date at the top changes only after we re-verify.

---

Download the data

Free, no email required. Version 1.1 adds the provider-category crosswalk, the six reporting lenses, West Virginia's channel sub-tables and provider geography, and cross-state category ledgers with their denominators attached.

• CSV, version 1.1 — flat source-row file with normalized category, reporting lens, denominator and source link on every row
• XLSX workbook, version 1.1 — comparison, crosswalk, per-state reconciliations, 21-program audit, data dictionary and claim ledger
• JSON, version 1.1 — machine-readable, same canonical data object as the tables above

Last verified August 13, 2026. Dataset version 1.1.

---

How to cite this page

The School Choice Index Editorial Team. "ESA Spending by Provider Category and State 2026 [Data]." The School Choice Index. Published August 3, 2026; updated and last verified August 13, 2026. https\://theschoolchoiceindex.com/research/education-savings-account-spending-by-category/

For the dataset:

The School Choice Index. ESA Spending by Provider Category and State, Version 1.1. Verified August 13, 2026. https\://theschoolchoiceindex.com/research/education-savings-account-spending-by-category/

---

Primary sources

• National Conference of State Legislatures. Education Choice State Policy Scan: Education Savings Accounts, updated July 21, 2025. Source for the source-dated 18-state, 21-program inventory used in this audit.
• Arizona Department of Education. ESA Program FY2026 Quarter 3 Report, transmitted May 29, 2026. Tables 9 and 10.
• Arizona Department of Education. ESA Quarterly Reports, checked August 13, 2026. FY2026 Q3 was the latest listed FY2026 report.
• Arizona Department of Education. ESA Program FY2025 Quarter 3 Report. Prior-year comparison.
• University of Arkansas Department of Education Reform and Arkansas Department of Education. 2024–25 Arkansas Education Freedom Accounts Annual Report, October 2025. Tables 3–5.
• Step Up For Students. FES for Students with Unique Abilities fact sheet, 2023–24 ESA spending top categories.
• Mississippi PEER Committee. PEER Report 710, December 9, 2024. Exhibit 11.
• Children's Scholarship Fund New Hampshire. 24–25 EFA Expenditure Overview, December 2025 DRAFT.
• South Carolina ESTF Program. Year-One Report, 2024–25. Table 9.
• South Carolina Supreme Court. Eidson v. South Carolina Department of Education, Opinion No. 28235, filed September 11, 2024.
• South Carolina General Assembly. Current Chapter 8, Title 59, including 2025 Act No. 11 amendments.
• West Virginia State Treasurer's Office / Hope Scholarship Board. Hope Scholarship Annual Report, 2024–25. Payment channels, category sub-tables, provider geography and named-recipient appendices.
• West Virginia Legislature. Current Hope Scholarship Code, Article 31 and House Bill 5249 status, checked August 13, 2026.
• Office of the Utah State Auditor. Limited review of the Utah Fits All Scholarship Program, June 26, 2026.
• Tennessee Comptroller of the Treasury, OREA. Education Savings Accounts: Program Evaluation, January 2026.
• Iowa Department of Education. Students First Education Savings Accounts, updated April 2026. Tuition-first sequencing rule.
• Texas Legislature. Senate Bill 2, enrolled version, establishing the Education Savings Account Program and its technology limit.

---

Related research

• Education Savings Account Statistics — programs, students and funding totals by state
• School Voucher Payments by Private School — this page covers what ESA money was spent on; that one covers whether the school receiving it is publicly named
• ESA Audit Reports by State
• School Voucher Tuition Gap by State
• What ESA Funds Can and Can't Be Used For — the rules, not the spending

---