Research · Dataset v1.0
School Voucher Projected Cost vs Actual Spending by State (2026 Data Audit)
Our school voucher projected cost vs actual spending by stateaudit found only four state-years that currently support a Grade A or Grade B comparison between an official cost projection and a later official outcome. Arkansas spent 20.8% less than projected in FY2024 and 4.5% less in FY2025. Iowa’s FY2025 state transfer and expenditure was $218,048,012 against a fiscal-note projection of $156.3 million — 39.5% above forecast. Iowa’s FY2024 official estimate was 19.7% above forecast, but it is not a closed-year actual.
The other high-profile comparisons reviewed here do not survive the same accounting test. Arizona’s frequently repeated four-digit “overrun” divides an incremental General Fund estimate for newly eligible students by gross scholarship awards for the entire program. Florida’s commonly compared numbers cover different populations. New Hampshire’s popular comparison crosses years and eligibility rules. North Carolina’s comparison mixes an initial allocation with scholarship funding processed after later budget action. West Virginia’s first year was interrupted by a court injunction.
The numbers are often real. The comparison is the problem.
We built this dataset to separate the comparisons that hold up from the ones that do not — and to publish both, including the rows we had to reject.
School voucher projected cost vs actual spending: the four comparisons that hold up
Four state-year comparisons in version 1.0 pass a test requiring the projection and outcome to cover the same program, fiscal year, participant population, and accounting measure. Arkansas came in below its enactment-era projection in both closed years reviewed. Iowa came in above its projection in both years, although its FY2024 outcome is an official estimate rather than a closed-year actual.
Table 1. Official school voucher and ESA cost projections matched to official outcomes
| State and fiscal year | Official projection | Official outcome | Difference | Variance | Grade |
|---|---|---|---|---|---|
| Arkansas FY2024 | $46,700,000 | $37,002,178 | −$9,697,822 | −20.8% | A |
| Arkansas FY2025 | $97,500,000 | $93,125,499.48 | −$4,374,500.52 | −4.5% | A |
| Iowa FY2024 | $106,900,000 | $127,939,695 | +$21,039,695 | +19.7% | B |
| Iowa FY2025 | $156,300,000 | $218,048,012 | +$61,748,012 | +39.5% | A |
Source: The School Choice Index analysis of the Arkansas LEARNS Act fiscal impact statement; Arkansas DFA FY2024 Budget Compliance Report; Arkansas DFA FY2025 Budget Compliance Report; Iowa LSA fiscal note for House File 68; Iowa FY2024 ESA budget schedule; and the Iowa LSA FY2025 Year-End Report. Grade definitions and row-level source links appear in the Methodology section. Last verified .
Grade definitions
- A
- Same program, year, population scope, and materially comparable accounting measure; official closed-year actual or state transfer/expenditure.
- B
- Comparable scope, but the later figure is an official estimate rather than a closed-year actual.
- C
- Appropriation or statutory cap compared with approvals, claims, transactions, or spending (utilization only; no forecast error published).
- D
- Scope mismatch: gross vs. net, all participants vs. an expansion cohort, or another materially different fiscal concept (no percentage published).
- E
- Missing source, legal disruption, unresolved authorization history, or no completed outcome (status only).
↓ Download the version 1.0 CSV
Do not average these four percentages into a national rate.Four state-years across two states are not a national denominator, and one row is an estimate rather than an actual. Version 1.0 supports no nationwide voucher “overrun rate.”
Which state’s numbers provide the clearest full projection-to-outcome chain?
Iowa FY2025 provides the clearest full chain in this audit because official state documents identify the enactment-era fiscal-note projection, the original appropriation, the year-end adjustment, and the final total for the same ESA budget unit. Those four figures answer two different questions: how far the original bill estimate missed and how much the state had to add to its initial annual appropriation.
Table 2. Iowa education savings accounts, FY2025: four official figures for one fiscal year
| Stage | Amount | What it measures | Official source |
|---|---|---|---|
| Fiscal-note projection | $156,300,000 | General Fund ESA appropriation cost estimated when House File 68 was enacted | Iowa LSA fiscal note, Table 4, January 23, 2023 |
| Original appropriation | $179,191,922 | FY2025 amount before the year-end standing-appropriation adjustment | Iowa LSA FY2025 Year-End Report |
| Adjustment to standings | +$38,856,090 | Amount added when the standing appropriation was reconciled | Iowa LSA FY2025 Year-End Report |
| Total appropriation and total expended | $218,048,012 | Final state-reported total for the ESA appropriation | Iowa LSA FY2025 Year-End Report |
Source: The School Choice Index analysis of the Iowa LSA fiscal note for House File 68 and the FY2025 Year-End Report on General Fund Revenues and Appropriations. The original appropriation plus the adjustment reconciles exactly: $179,191,922 + $38,856,090 = $218,048,012. Last verified .
There are two different gaps in that table.
The enactment-era gapis the fiscal note against the final total: $156.3 million versus $218,048,012, or +39.5%. That answers: “How did the cost estimated when lawmakers enacted the program compare with the FY2025 outcome?”
The annual-budget gap is the original FY2025 appropriation against the final total: $179,191,922 versus $218,048,012, or +21.7%. That answers: “How much did the standing appropriation need to be increased during year-end reconciliation?”
Both calculations are real. They answer different questions.
Iowa’s ESA adjustment was 73.7% of the statewide net standings adjustment
Iowa reports $52,696,505in net FY2025 Adjustments to Standings across the General Fund. The ESA line’s $38,856,090 increase was 73.7%of that net statewide total. That percentage uses a net denominator: positive and negative standing-appropriation adjustments are combined in the state’s statewide total. It is not a share of all gross positive adjustments, and it is not a share of the entire General Fund budget.
How the 2023 fiscal-note figures compare with later official budget figures
The same 2023 fiscal note projected ESA appropriation costs through FY2027. Later Iowa documents now provide one actual, one official FY2024 estimate, one FY2026 estimate, and one FY2027 Governor’s recommendation.
Table 3. Iowa ESA appropriation figures: 2023 fiscal note vs. later official figures
| Fiscal year | 2023 fiscal-note projection | Later official figure | Difference | Status of later figure |
|---|---|---|---|---|
| FY2024 | $106,900,000 | $127,939,695 | +19.7% | Official estimate; Grade B |
| FY2025 | $156,300,000 | $218,048,012 | +39.5% | Closed-year actual; Grade A |
| FY2026 | $313,900,000 | $327,859,472 | +4.4% | Current official estimate; not an outturn |
| FY2027 | $341,100,000 | $349,646,976 | +2.5% | Governor’s recommendation; not an outturn |
Source: The School Choice Index analysis of the Iowa LSA fiscal note for House File 68, the FY2025 Year-End Report, and the January 2026 Analysis of the Governor’s Budget and Recommendations. FY2026 and FY2027 are forward-looking figures and do not receive comparability grades. Last verified .
The table does not establish that Iowa’s forecasting has permanently improved. FY2026 is still an estimate, FY2027 is a recommendation, and neither is an actual result. The largest closed-year gap currently visible in this series is FY2025.
That year also needs to be described accurately. Students who had attended public school in the prior year were already eligible in FY2024. The FY2025 phase expanded the income threshold for certain existing nonpublic-school families from 300% to 400% of the federal poverty level; it did not newly open the program to all prior-public-school students.
Why can’t Arizona’s voucher cost be compared directly with its original estimate?
Arizona’s $64.5 million figure was an estimate of the incremental FY2024 General Fund effect of HB 2853’s eligibility expansion. The later $738,002,994 figure was the total annual scholarship award value for every student enrolled in the ESA program in Quarter 4. Different populations and different accounting concepts make a direct cost-overrun percentage invalid.
The Arizona Joint Legislative Budget Committee fiscal note for HB 2853 projected that the expansion would increase state General Fund K–12 costs by $33.4 million in FY2023, $64.5 million in FY2024, and $125.4 million in FY2025. The note called those estimates “highly speculative” because participation among newly eligible students was difficult to predict.
The Arizona Department of Education’s FY2024 Quarter 4 report later recorded 74,578 students and $738,002,994 in total annual scholarship awards. That award total covers the entire program, not only the cohort created by HB 2853.
Table 4. Arizona’s commonly compared figures measure different things
| Figure | Amount | Population | Accounting concept | Can it be divided by the other? |
|---|---|---|---|---|
| HB 2853 FY2024 fiscal-note estimate | $64,500,000 | Newly eligible students affected by the expansion | Incremental General Fund K–12 cost, including offset assumptions for public-school switchers | No |
| FY2024 Quarter 4 annual scholarship awards | $738,002,994 | All 74,578 ESA students in 11 eligibility categories | Gross annual award value for the full program | No |
Source: Arizona JLBC fiscal note for HB 2853, June 16, 2022; Arizona Department of Education FY2024 Quarter 4 ESA Report, August 30, 2024. Last verified .
Three features of the fiscal note matter:
- It estimated newly eligible participation, not the full program. Existing ESA participants remained outside the expansion-only calculation.
- It used a net differential for public-school switchers. The note estimated that a non-disabled switcher would add $529 on average above what the state otherwise would have paid through Basic State Aid, rather than charging the full award as a new net cost.
- It separately modeled currently private-school and homeschool students. Those cohorts generated new cost because the state generally was not already paying their public-school formula funding.
The Arizona enrollment comparison that does hold up
Arizona did publish two explicit total-enrollment forecasts for FY2024:
- The enacted FY2024 budget used a JLBC baseline assumption of 68,400 total ESA students.
- The Arizona Department of Education separately estimated 100,000 students by the end of FY2024, but JLBC said the submission did not explain its methodology.
- The final Quarter 4 report counted 74,578 students receiving an ESA scholarship.
Table 5. Arizona FY2024 total ESA enrollment: two forecasts and the Quarter 4 count
| Benchmark | Students | Difference from Quarter 4 count | Interpretation |
|---|---|---|---|
| Enacted-budget assumption | 68,400 | Quarter 4 was +6,178, or 9.0% higher | Valid same-year enrollment comparison |
| ADE end-of-year estimate | 100,000 | Quarter 4 was −25,422, or 25.4% lower | Valid same-year enrollment comparison; ADE methodology was not supplied in the submission |
| FY2024 Quarter 4 count | 74,578 | — | Students receiving an ESA scholarship in Quarter 4 |
Source: Arizona JLBC, “Empowerment Scholarship Account (ESA) Funding Report Background,” May 31, 2023; Arizona Department of Education FY2024 Quarter 4 ESA Report. This is an enrollment comparison, not a dollar-overrun calculation. Last verified .
The defensible Arizona finding: Quarter 4 enrollment finished 9.0% above the assumption built into the enacted budget and 25.4% below ADE’s separate 100,000-student estimate.
What participation assumptions do state fiscal notes use?
Selected state fiscal notes use participation assumptions ranging from 5% to 100%, but the percentages do not describe one standardized population. Some apply to newly eligible private-school students, some to homeschool students, some to public-school switchers, and some to a later universal-expansion phase.
Every row below preserves the population and phase attached to the percentage. Stripping those labels away would create the same comparability problem this page is designed to prevent.
Table 6. Participation assumptions stated in selected official fiscal notes
| State | Assumption stated by the fiscal analyst | Population and phase |
|---|---|---|
| Arizona | 5% in year one, 10% in year two, 20% in year three | Newly eligible existing private-school students; the same rates were applied to newly eligible homeschool students |
| Iowa | 20% at the first income threshold; 10% at the second threshold; 1.0% initial public-to-nonpublic transfer and 0.02% thereafter | Different nonpublic-school eligibility phases plus modeled public-school switchers |
| Florida | 50% participation | Identified private-school students newly eligible for FES-EO and not already receiving a scholarship |
| West Virginia | 100% participation | Existing private and parochial students in the later expansion phase; the note also described current homeschool students |
| Texas — contextual case outside version 1.0 | 50% participation | Current private-school and homeschool populations in the modeled FY2027 expansion, subject to available funding |
Source: Arizona JLBC fiscal note for HB 2853; Iowa LSA fiscal note for House File 68; Florida House Final Bill Analysis for HB 1; West Virginia Department of Education fiscal note for HB 2013; Texas LBB fiscal note for SB 2. Last verified .
The spread is real. It is not proof that one percentage is universally right or wrong. A fiscal-note percentage should never be quoted without its noun. “Five percent” is not an assumption until the reader knows five percent of whom, in which year, under which eligibility rule.
Which school voucher cost comparisons are not valid?
A comparison fails when the projection and outcome cover different years, populations, programs, or accounting measures. Version 1.0 rejects or holds prominent comparisons in Arizona, Florida, New Hampshire, North Carolina, and West Virginia instead of publishing precise-looking percentages that the source definitions do not support.
Table 7. Rejected and held projection-to-outcome comparisons
| State | Tempting comparison | Why it fails | Grade |
|---|---|---|---|
| Arizona | $64.5 million vs. $738,002,994 in annual awards | Expansion-only incremental General Fund estimate vs. gross awards for the full program | D |
| Florida | $209.6 million vs. a later all-program scholarship total | Incremental impact of two HB 1 components vs. spending for populations and programs outside those components | D |
| New Hampshire | An initial six-figure budget figure vs. a later multimillion-dollar year | The circulated comparison crosses fiscal years and eligibility regimes; a same-year outcome has not been cleared | D/E |
| North Carolina | $293.54 million allocation vs. $432,238,544 in scholarship funding processed | Later funding action has not been fully reconciled, and processed or awarded scholarship funding is not automatically the same as cash expenditure against the initial ceiling | E |
| West Virginia FY2023 | $23,667,486 forecast vs. $9,188,026.43 transferred or $7,771,761.39 spent | A court injunction interrupted the launch from July through October 2022 and led to prorated awards | E |
Source: The School Choice Index analysis of official state fiscal notes, budget summaries, annual reports, and program records. Last verified .
Florida: the $209.6 million estimate was incremental
Florida’s final bill analysis estimated a $209.6 millionstate expenditure impact from two components of HB 1: $112.1 million for expanded FES-EO eligibility and $97.5 millionfor the higher FES-UA growth allowance. The same analysis treated a public-school student moving to a scholarship as having no incremental fiscal impact because the student was already funded through the Florida Education Finance Program. The $209.6 million figure is an incremental bill-impact estimate, not a forecast of all scholarship spending in Florida.
New Hampshire: the popular comparison crosses regimes
The widely circulated New Hampshire comparison pairs an early budget figure with spending in a later year after program participation and eligibility changed. Version 1.0 does not publish a percentage until the original line can be paired with an official outcome for the same fiscal year and legal scope.
North Carolina: allocation, authorization, and processed awards are different stages
North Carolina’s 2024–25 annual report says the State Education Assistance Authority processed scholarship funding for 80,472 students totaling $432,238,544. An earlier legislative budget summary identified $293.54 million for awards. Subtracting those figures before reconciling later authorization and identifying the exact cash measure would turn a funding-history question into a false forecast-error statistic.
West Virginia: an interrupted launch year is not a normal outturn
The Hope Scholarship fiscal note estimated $23,667,486 for the initial year. The first annual report recorded $9,188,026.43 transferred, $7,771,761.39 spent, and $1,416,265.04 unspent. A court injunction halted the program from July through October 2022, participation declined, and awards were prorated. We publish the source figures and do not treat the result as ordinary forecast variance.
West Virginia also demonstrates why “the fiscal note” is not always one number. The bill record contains the initial-year estimate, an estimated $102,890,453 cost for the later private- and homeschool-expansion phase, and $126,557,939 across the two phases.
Why do two official numbers for the same program sometimes disagree?
Official program reports and official finance reports can measure the same program at different levels of precision or under different accounting presentations. Arkansas’s annual program report uses rounded figures, while its state budget-compliance reports provide exact non-GAAP budgetary-basis expenditures.
Table 8. Arkansas Education Freedom Accounts: rounded program report vs. exact finance report
| Fiscal year | Fiscal-note projection | DESE annual-report figure | DESE variance | DFA budget-compliance expenditure | DFA variance |
|---|---|---|---|---|---|
| FY2024 | $46,700,000 | $37.3 million | −20.1% | $37,002,178 | −20.8% |
| FY2025 | $97,500,000 | $93.8 million | −3.8% | $93,125,499.48 | −4.5% |
Source: Arkansas LEARNS Act fiscal impact statement; Arkansas DESE 2024–25 EFA Program Annual Report; Arkansas DFA FY2024 Budget Compliance Report; Arkansas DFA FY2025 Budget Compliance Report. The state documents do not publish a reconciliation between the rounded and exact figures. Last verified .
The direction is the same in both official series: Arkansas came in below the fiscal-note projection in both years. Table 1 uses the exact DFA expenditure figures because they provide the clearest budget-to-actual measure. This is the whole thesis in miniature: two official numbers are not automatically interchangeable merely because they concern the same program.
Does the headline cost of a school-choice law equal the voucher line inside it?
Not always. An omnibus education act may fund teacher pay, literacy programs, administration, and scholarships in one fiscal package. Comparing the full act’s headline cost with actual scholarship spending creates an apparent underrun that says nothing about the scholarship projection.
Arkansas provides a clean example. The LEARNS Act fiscal impact statement estimated approximately $297.5 millionacross the act’s first-year provisions. The Education Freedom Account line inside that total was $46.7 million, or 15.7% of the full first-year fiscal impact.
The $46.7 million line is the projection compared with EFA expenditures in Table 1. The $297.5 million omnibus figure is not.
Can spending exceed a fixed statutory ceiling?
A program cannot lawfully approve more than a fixed cap without a legal funding change, but cap utilization does not measure forecast accuracy. In a capped program, unmet demand may appear as denied applications or a waitlist instead of an expenditure overrun.
Table 9. Funding-ceiling utilization in version 1.0
| State and program | Funding ceiling | Approved or represented amount | Utilization | What the figure shows | Grade |
|---|---|---|---|---|---|
| Oklahoma Parental Choice Tax Credit, tax year 2024 | $150,000,000 | $150,000,000 approved | 100.0% | The statutory cap was fully allocated | C |
| Utah Fits All, 2024–25 | $82,573,400 in base and supplemental appropriations | $67,733,684 in the program-manager transaction population | 82.0% | The identified transaction population represented 82.0% of appropriated funding | C |
Source: Oklahoma Tax Commission 2024 Parental Choice Tax Credit Report; Utah enacted base and supplemental budgets; Office of the Utah State Auditor Report No. 25-49. The Utah review was limited and did not audit every program financial activity. Last verified .
Oklahoma’s report recorded 38,756 applications, 27,099 approvals, 11,657 denials, $150.0 million approved, and $62.2 million associated with denied applications. Those figures demonstrate demand against a cap. They do not demonstrate that a projection was right or wrong.
Texas shows the difference between a cap and a demand estimate
Texas is a current contextual case, not one of the 11 states in dataset version 1.0. The Texas Legislative Budget Board said the Education Freedom Account program’s ultimate cost in the 2026–27 biennium would be limited to the amount appropriated. The same fiscal note estimated potential costs of approximately $3.3 billion in FY2028, $3.7 billion in FY2029, and $4.8 billion in FY2030 under the modeled assumptions. The Comptroller later reported more than 274,000 applications; by June 10, 2026, more than 102,000 students had received awards and nearly 145,000 eligible students remained on the waitlist.
That is why “stayed within the appropriation” and “demand exceeded funded capacity” can both be true. One describes the legal ceiling. The other describes applications.
What counts as actual school-voucher spending?
There is no universal state field called “actual spending.” An official outcome may be an audited expenditure, a budgetary-basis expenditure, a state transfer, an approved tax credit, a scholarship award, a reimbursed transaction, a final estimate, or a funding ceiling. This dataset names the measure instead of silently treating those terms as synonyms.
Table 10. Data dictionary
| Term | What it means in this audit | Common misuse |
|---|---|---|
| Fiscal-note projection | Prospective estimate prepared before or during enactment | Treated as a forecast of total program spending when it estimates only an incremental bill effect |
| Appropriation | Legal authorization or funding provision | Treated as exact expected demand |
| Standing unlimited appropriation | Statutory appropriation reconciled to the amount required under the formula | Treated as a fixed cap |
| Original appropriation | Amount available before later adjustments | Treated as the final annual total |
| Final budget | Available amount after transfers, supplements, and revisions | Treated as the original projection |
| Expenditure | Amount recorded as spent under the source’s accounting basis | Treated as identical across state accounting systems |
| Transfer | Amount moved into or through a state fund or budget unit | Treated as proof that families spent every dollar in the same year |
| Award | Scholarship or credit approved or assigned | Treated as cash disbursed |
| Recipient spending | Amount paid or reimbursed for eligible expenses | Treated as equal to an account’s full award value |
| Statutory cap | Maximum amount that can be approved without a legal change | Treated as a demand forecast |
| Gross program cost | Total awards, transfers, or expenditures before defined offsets | Compared with a net fiscal-impact estimate |
| Net fiscal impact | Cost after the source’s defined savings or offsets | Compared with a gross total |
Source: Definitions derived from the accounting treatments documented in the primary sources cited throughout this page.
Iowa’s $218,048,012is a state-reported appropriation and expenditure figure for the ESA budget unit. It should not be rewritten as “families spent $218 million.” North Carolina’s $432,238,544is described as scholarship funding processed for 80,472 students. Arizona’s $738,002,994 is an annual scholarship award total attached to Quarter 4 participants.
How we produced this dataset
We located official projection documents and official later outcomes, recorded the exact terminology used by each source, tested every candidate pair for program, year, population, and accounting compatibility, and calculated a variance only when those dimensions matched. Failed comparisons remain visible in a rejected-comparison ledger.
The comparability grades
| Grade | Standard | Variance published? |
|---|---|---|
| A | Same program, year, population scope, and materially comparable accounting measure; official closed-year actual or state transfer/expenditure | Yes |
| B | Comparable scope, but the later figure is still an official estimate rather than a closed-year actual | Yes, with the estimate label attached |
| C | Appropriation or statutory cap compared with approvals, claims, transactions, or spending | No; utilization only |
| D | Scope mismatch: gross vs. net, all participants vs. an expansion cohort, or another materially different fiscal concept | No; the mismatch is explained |
| E | Missing source, legal disruption, unresolved authorization history, or no completed outcome | No; status only |
Inclusion criteria
A row qualifies for Grade A or B only when all six conditions hold:
- The same program or legally defined expansion is being measured.
- The same fiscal or school year is being measured.
- The same participant population is being measured.
- The accounting measures are materially comparable.
- The projection comes from an authoritative contemporaneous source.
- The later figure comes from an authoritative post-close or official final source.
Exclusion criteria
We do not calculate a variance when:
- An expansion-only projection is paired with all-program spending.
- A net fiscal impact is paired with gross award value.
- One year is paired with another.
- Awarded or processed funding is automatically treated as cash spent.
- A cap is treated as a demand forecast.
- Litigation materially disrupted the implementation year.
- A revised or supplemental authorization has not been reconciled.
- The numerical claim can be found only in a secondary source.
- A derived estimate requires an unstated assumption about population growth, attrition, or accounting treatment.
Calculation rules
Variance in dollars = Official outcome − Official projection Variance as a percentage = (Official outcome − Official projection) ÷ Official projection Utilization percentage = Approved, transferred, or represented amount ÷ Cap or appropriation
A utilization percentage is never labeled a forecast error.
Source hierarchy
- Audited state financial reports or official budget-to-actual reports.
- Official closed-year budget and agency expenditure reports.
- Official legislative or program annual reports.
- Official final budget estimates.
- Official fiscal notes and legislative analyses.
- Official program-manager records where the manager is legally responsible for administration.
- Secondary sources only to locate a claim or capture public wording — never as the numerical source of record.
What this school-voucher cost data shows — and what it does not show
The dataset shows whether specified official projections and specified official later figures were above, below, or incapable of comparison under one accounting test. It does not decide whether a program is good policy, measure educational outcomes, or establish a causal effect on a state’s broader budget.
What it shows
- Four Grade A or B state-year projection comparisons.
- Exact dollar and percentage differences for the rows that clear the test.
- Why prominent Arizona, Florida, New Hampshire, North Carolina, and West Virginia comparisons do not clear it.
- Whether a later figure is an actual, estimate, transfer, expenditure, award, transaction population, or cap utilization measure.
- How participation assumptions differ across selected state fiscal notes.
- Which source or reconciliation is still missing for an unfinished state.
What it does not show
- Student achievement or program quality.
- Distributional equity or who benefits.
- A causal effect on public-school budgets.
- Whether a state should expand, change, or repeal a program.
- Every long-term offset or secondary fiscal effect.
- A nationwide average, median, or voucher-overrun rate.
- Every voucher, tax-credit scholarship, and ESA program in all 50 states.
Limitations
| Limitation | Practical significance |
|---|---|
| Version 1.0’s core audit covers 11 states, not every school-choice program in the country. | This is an accounting-comparability audit, not a 50-state spending census. |
| Texas appears only as a separate current case study. | Texas is not counted in version 1.0’s 11-state dataset totals or schema coverage. |
| States use different fiscal years and accounting systems. | Cross-state totals should not be added without further normalization. |
| A transfer or appropriation total is not automatically recipient spending. | Iowa keeps its state transfer/expenditure label. |
| An award or processed-funding total is not automatically cash disbursed. | North Carolina remains unresolved. |
| A cap is not a demand forecast. | Oklahoma and Utah receive Grade C utilization treatment. |
| Some source projections are rounded. | Percentage results can differ slightly when an exact enacted amount becomes available. |
| Arkansas has a rounded program-report series and an exact finance-report series. | Table 1 uses the exact finance figures and discloses the rounded series separately. |
| Program eligibility changes between years. | Cross-year New Hampshire comparisons are not published as forecast error. |
| Litigation can interrupt implementation. | West Virginia’s first year is not treated as a normal variance result. |
| Arizona’s $738.0 million is annual award value, not net fiscal impact or paid expenses. | No dollar-overrun percentage is published from that figure. |
| Utah’s State Auditor report is currently linked through a media-hosted copy. | Replace it with a state-hosted copy when one becomes available. |
| Ohio’s enactment-era forecast has been located, but later official totals have not been reconciled to one matching measure. | No Ohio variance is published in version 1.0. |
| No homogeneous national denominator exists in this release. | We publish no national rate. |
State coverage status
| State | Best grade | Version 1.0 treatment |
|---|---|---|
| Arizona | D | Cost-scope mismatch explained; separate FY2024 enrollment comparison published |
| Arkansas | A | Two closed-year comparisons using exact state finance figures |
| Florida | D | Incremental HB 1 estimate separated from all-program totals |
| Indiana | E | Status only; no cleared enactment-era projection located |
| Iowa | A/B | One closed-year actual and one official estimate published separately |
| New Hampshire | D/E | Popular cross-year comparison rejected |
| North Carolina | E | Held pending authorization and cash-measure reconciliation |
| Ohio | E | Forecast located; matching later measure not reconciled |
| Oklahoma | C | Statutory-cap utilization only |
| Utah | C | Appropriation utilization with audit-scope caveat |
| West Virginia | E | Court-disrupted launch explained; no variance published |
Why this matters when a new fiscal note is published
The first number attached to a school-choice bill may be an incremental net effect, a gross appropriation estimate, a statutory ceiling, or one phase of a larger program. Before repeating it as “the cost,” a reader has to identify the population, year, and accounting measure attached to it.
The practical lesson is not that voucher programs always exceed projections. Arkansas did not in either closed year reviewed. Iowa did in FY2025. West Virginia’s first year came in far below its initial figure under a court-disrupted launch. Oklahoma fully used its cap, which says nothing by itself about forecast accuracy.
The useful questions are narrower:
- Does the fiscal note model the entire program or only newly eligible students?
- Does it report gross appropriation cost or net fiscal impact?
- What participation rate does it assign to private-school, homeschool, and public-school students?
- Is the later figure an expenditure, transfer, award, account value, or cap?
- Did eligibility, funding, or implementation change before the outcome year closed?
A number without those labels is easy to quote and easy to misuse. The labels are the analysis.
Frequently asked questions
Which states spent more on school vouchers than projected?
In version 1.0, Iowa is the state with publishable above-projection rows. Its FY2025 state ESA transfer and expenditure of $218,048,012 was 39.5% above the $156.3 million fiscal-note projection. Its FY2024 official estimate was 19.7% above forecast but receives Grade B because it is not a closed-year actual. Arkansas came in below projection in both closed years reviewed.
Did Arizona’s voucher program exceed its forecast by more than 1,000%?
That percentage cannot be established by dividing the commonly cited official figures. The $64.5 million estimate covered the incremental FY2024 General Fund effect of expanding eligibility, while $738,002,994 was the gross annual award value for all 74,578 Quarter 4 participants. Arizona’s valid same-year enrollment comparison shows Quarter 4 participation was 9.0% above the 68,400 assumption used in the enacted budget and 25.4% below ADE’s separate 100,000-student estimate.
Is a voucher appropriation the same as a spending forecast?
No. An appropriation authorizes or supplies funding under state law. A forecast estimates expected cost or demand. A standing appropriation can be reconciled upward or downward, while a fixed statutory cap limits approvals unless the law or funding changes.
Does actual spending always mean families spent the money?
No. A state figure may represent a budgetary-basis expenditure, transfer, approved scholarship, tax credit, reimbursement, account award, or another measure. Iowa’s $218,048,012 is a state appropriation and expenditure figure for the ESA budget unit, not proof that families spent every dollar during the same fiscal year.
Why is Iowa’s FY2024 figure Grade B rather than Grade A?
The official source labels the $127,939,695 figure as an estimate rather than a closed-year actual. Its accounting scope is comparable enough to calculate a labeled difference, but it is kept separate from Grade A outcomes.
Why isn’t West Virginia’s first year treated as a normal forecast comparison?
A court injunction stopped the Hope Scholarship from operating from July through October 2022. Participation declined and awards were prorated, so the partial launch year does not provide a normal test of the fiscal note’s participation assumptions.
Can a capped school-voucher program go over budget?
It cannot approve spending above a fixed legal ceiling without a funding or legal change. That does not mean demand was forecast accurately. Oklahoma filled its $150 million tax-credit cap, and Texas’s separately discussed program received far more eligible applications than funded awards.
Why do official sources give different totals for the same state and year?
They may use different accounting bases, levels of precision, or stages of the funding process. Arkansas’s program report uses rounded figures while its budget-compliance reports provide exact expenditures. Iowa’s fiscal note publishes both a gross ESA appropriation-cost series and a broader net-fiscal-impact series. A valid comparison has to match the measure, not just the program name.
Data downloads
The complete dataset is published in machine-readable form with row-level source URLs, projection and outcome measures, comparability grades, and verification dates.
- School Voucher Forecast-to-Outturn Audit — CSV (version 1.0, July 31, 2026)All 13 state rows with projection measures, outcome measures, grade, notes, and source URLs
Primary sources
Arizona
Arkansas
- Arkansas General Assembly, LEARNS Act fiscal impact statement, SB 294
- Arkansas Department of Finance and Administration, FY2024 Budget Compliance Report
- Arkansas Department of Finance and Administration, FY2025 Budget Compliance Report
- Arkansas Division of Elementary and Secondary Education, 2024–25 EFA Program Annual Report
Iowa
New Hampshire
North Carolina
Ohio
Texas — contextual case outside version 1.0
Utah
How to cite this page
The School Choice Index Editorial Team. “School Voucher Projected Cost vs Actual Spending by State.” The School Choice Index. Dataset version 1.0. Last verified .
https://theschoolchoiceindex.com/research/school-voucher-projected-cost-vs-actual-spending-by-state/
About this research
Who created this: The School Choice Index Editorial Team.
How it was produced:By manually matching official state cost projections with official later spending, transfer, award, budget, tax-credit, and audit records under a documented comparability test — and by preserving the comparisons that failed that test instead of discarding them.
Why it exists: To make school-voucher cost claims easier to verify and to keep figures with different populations or accounting scopes from being presented as direct comparisons.
The School Choice Index is an independent, non-partisan editorial reference for U.S. K–12 school-choice programs.
Version 1.0 — . Initial 11-state core audit: three Grade A state-years, one Grade B estimate, two Grade C utilization rows, five rejected or held comparison groups, and status-only entries for Indiana and Ohio. Texas is included only as a separately labeled current case study.
Last updated: · Last verified: · Dataset version: 1.0
The School Choice Index is an independent comparison and research resource for U.S. school choice programs. Material corrections are recorded with the date, affected records, previous and revised values, reason, and source. Read our methodology, editorial standards, and corrections policy.