Voucher-Program Effects on Public-School Per-Pupil Funding: Gross Deduction Is Real, but the Net Loss Depends on Whether Local Revenue Backfills It

Summary

There is no single dollar figure that applies across states, because the answer hinges on a mechanism most coverage of this question skips: whether local school revenue backfills the state aid that follows a voucher student out the door. The best available causal evidence — a quasi-experimental study of Ohio's EdChoice program (Lavertu, Ohio State University, published by the Thomas B. Fordham Institute, 2022) — found districts exposed to the voucher program lost 10-15% of enrollment relative to their counterfactual, but total per-pupil spending was statistically indistinguishable from what it would have been without the program, because local property-tax revenue (which does not follow the student and does not shrink when enrollment falls) rose 10-15% per pupil, roughly offsetting the state-aid deduction. Gross, un-offset estimates run much higher: a hypothetical 5% enrollment decline in Cleveland was estimated (EPI) to cost remaining students $364-$927 per pupil in gross state aid, and in Arizona, whose 2022 universal Empowerment Scholarship Account (ESA) program is younger, more expansive, and mostly serves students who were never in the public system (52-71% came from private/home school), public schools received on average about $300,000 less in state funding per school in FY2024, while ESA funding consumed 53% of all new K-12 spending growth for a program serving under 10% of students. In short: the state-aid deduction is real and often large in gross terms, but whether it becomes a genuine net per-pupil funding loss for the students who remain depends on the program's design and on local districts' ability to raise offsetting local revenue — a distinction most advocacy figures on both sides collapse.

1. The mechanism: gross deduction vs. net loss

Nearly every state funds K-12 education (in part) on a per-student formula. When a student leaves for a voucher, the state deducts that student's formula amount from the sending district's state aid (Ohio) or simply never allocates it to the district in the first place (many ESA states). That deduction is the gross effect, and it is what most headline figures report.

But state aid is only one of a district's two main revenue streams. The other — local property-tax revenue — is typically levied at a fixed dollar amount or fixed millage rate that does not automatically fall when enrollment falls. When a district's enrollment shrinks (from vouchers, or any other cause) while its local levy dollars stay flat, local revenue per remaining pupil mechanically rises. This is exactly the pattern the Ohio study documents, and it is the reason "how much do schools lose" cannot be answered from the state-aid deduction alone.

2. Ohio EdChoice: the strongest causal evidence available, and it found no net loss

Stephane Lavertu (Ohio State University), writing for the Thomas B. Fordham Institute (Dec. 2022), used a district-exposure difference-in-differences design over 2006-07 to 2018-19 for Ohio's performance-based EdChoice program (the state's largest voucher program, ~35,000 students as of 2021-22). Key findings for the average exposed district, 13 years after program start:

• District enrollment ran 10-15% lower than the counterfactual.
• Total district spending per pupil (capital + operating) was not statistically different from the counterfactual.
• Local revenue per pupil rose 10-15%, because EdChoice reduced state aid but had no effect on districts' ability to raise local levies, and that levy revenue was now spread over fewer students.
• No significant effects on property values, local tax rates, levy passage rates, or state fiscal-stress designations.

This is the single most rigorous state-level causal estimate available for a mature voucher program, and its headline finding directly contradicts the intuitive "vouchers drain district funding" claim — not because the state-aid deduction is fictional, but because it was offset. Critics (Policy Matters Ohio, the "Vouchers Hurt Ohio" coalition of districts) dispute this by pointing to the raw growth in statewide voucher deductions (up $47 million in one year) and by arguing that the state, not districts, should absorb the deduction — a valid state-budget critique, but a different question from whether remaining district students' per-pupil funding actually fell.

3. Arizona ESA: a newer, larger, and structurally different program

Arizona became the first state with a universal voucher program (Empowerment Scholarship Accounts) in 2022. Its cost structure differs from Ohio's in ways that matter for the funding-loss question:

• Most new ESA recipients were never public-school students. 52% (2024) to 71% (2023) of ESA users came from private school or homeschooling — not from a public-school seat whose state formula dollars simply follow the student. This is new state spending, not a reallocation, and it does not free up anything for remaining public-school students.
• The voucher amount exceeds the traditional per-pupil state allocation by $424 (elementary/middle) to $540 (high school) per student, because Arizona bases ESA awards on the (higher) charter-school funding formula.
• Scale and growth: the program cost $738 million in FY2024 (up from far less at 12,127 participants in 2021-22 to 61,689 by 2022-23, +409%) and is projected at $864 million for FY2025 — roughly 10-12% of state K-12 funding for under 10% of students.
• Effect on public schools: EdTrust estimates every Arizona public school received, on average, about $300,000 less in state funding in FY2024 because of the program's growing cost, even as public schools continued to educate 92% of the state's students. ESA funding accounted for 53% of all new K-12 spending growth that year, versus 47% for traditional public-school funding growth.
• Arizona's voucher expansion has been cited (12News, Learning Policy Institute) as the single largest driver of the state's $1.4 billion FY24-25 budget shortfall, forcing across-the-board state agency cuts.

Because Arizona's program only went universal in 2022 and this connector's own district-finance table (below) does not carry usable pre-2022 state-coded data, no rigorous causal before/after estimate for Arizona could be computed here or found in the published literature at the same design quality as the Ohio study. The EdTrust figures above are descriptive/accounting estimates, not a quasi-experimental estimate of the counterfactual.

4. Illustrative gross estimates from other sources

The Economic Policy Institute modeled a hypothetical 5% enrollment decline for Cleveland Metropolitan School District (a district affected by Ohio's older, smaller, Cleveland-specific voucher pilot) and estimated remaining students could lose $364 to $927 per pupil in gross state aid — i.e., the deduction alone, with no adjustment for local backfill. This is useful as an order-of-magnitude illustration of the gross mechanism, but per the Ohio causal study above, a gross estimate like this should not be read as the empirically observed net effect on total per-pupil spending, which the causal study found to be roughly zero for EdChoice-exposed Ohio districts overall.

5. Supplementary check against Census F-33 district finance data

To test the backfill mechanism directly (per this corpus's own guidance on this exact question), I queried edu.f33_district_finance for Arizona (state_fips 04) and Ohio (state_fips 39), aggregating state and local revenue per pupil. This table has a significant coverage gap that limited the test: only fiscal years 2016, 2023, and 2024 are loaded at all (2017-2022 are entirely absent), and the 2016 vintage has state_fips null for every row nationwide, making it unusable for a state-level comparison. That leaves only FY2023 and FY2024 — both entirely after Arizona's 2022 universal ESA expansion — so no genuine pre/post causal test could be built from this table for either state. Within that one comparable window:

Both states show nominal state and local per-pupil revenue rising year over year — consistent with "no outright collapse in funding" in the one window observed, and directionally consistent with the local-backfill mechanism (though this is descriptive, not causal, given the missing pre-period). This gap in the warehouse's F-33 coverage (missing FY2017-FY2022) is itself worth flagging: it is exactly the window that would let a state-level diff-in-differences test be built around Arizona's 2022 program start, and it currently cannot be.

Bottom line

How much per-pupil funding public schools lose after a state starts a voucher program is not a fixed number — it depends on program design, program age, and whether local revenue offsets the state-aid deduction. In Ohio's mature, performance-targeted EdChoice program, the best causal study found essentially no net per-pupil spending loss, because local levies (which don't follow the student and don't shrink with enrollment) rose enough to offset the state-aid reduction. In Arizona's much larger, newer, and mostly non-diversionary universal ESA program — where most new recipients were never public-school students to begin with — the effect looks different: no comparable causal study exists yet, but the program consumed the majority of all new K-12 spending growth and public schools received roughly $300,000 less in average state funding per school in FY2024, alongside a $1.4 billion state budget shortfall the program is cited as the largest driver of. Gross, un-offset hypothetical estimates (e.g., $364-$927 per pupil for a 5% Cleveland enrollment decline) illustrate the deduction mechanism but should not be quoted as an observed net effect.