Percentage of Voucher Students Already in Private School by State

The percentage of voucher students already in private school by state cannot be reduced to one valid national number, because the state figures in circulation do not measure the same thing. Version 1.0 of this dataset audits official materials for a defined 18-state source set and adds Wisconsin as a separate benchmark.

The direct answer is narrower than the debate around it:

• Arkansas:the revised state report records 25% as previously private and another 33% as returning 2023–24 EFA participants. Because Arkansas did not permit EFA-funded homeschool expenses in 2023–24, the best reproducible renewal-inclusive figure is approximately 58%. That is a calculation from two rounded categories, not one field printed by the state.
• Iowa: approximately 66.3%of 2023–24 participants had previously attended a nonpublic school, calculated from the department’s statement that 33.7% had not.
• Wisconsin benchmark: 20,176 of 23,417 participants, or 86.2%, were in a Wisconsin private school the prior year. Of those 20,176 students, 19,047 — 94.4% — were already in the Choice program.

That distinction changes what “already in private school” means. A stock measure across all current participants can be dominated by the program’s own returning students. It does not tell you how many private-school students received public funding for the first time that year.

1. Percentage of voucher students already in private school by state: verified data

Among the 18 states in version 1.0, Arkansas and Iowa are the only two with a direct or reproducible prior-private/nonpublic measure admitted to the dataset. Wisconsin is shown separately because its WPCP report publishes the strongest renewal split in the admitted source set. Every other admitted row either combines private school with homeschool or measures public-school history instead.

Table 1. Voucher and ESA participants by prior school status

Source: The School Choice Index, compiled and calculated from the primary materials linked in the state sections and downloadable dataset. “A” is direct private/nonpublic evidence; “B” is combined or estimated evidence; “C” is public-school-history evidence; “R” flags a renewal measure. Last verified August 1, 2026.

The first arithmetic discrepancy in the source record

Arizona’s FY2026 Quarter 3 report prints 8,894 of 15,152 as 58.6%. The displayed counts divide to 58.6985%, which rounds to 58.7%at one decimal place. The dataset retains both values: 58.6% as the agency’s printed percentage and 58.7% as the reproducible calculation.

That may look like a small difference. It is exactly the kind of difference a source-backed dataset should expose rather than silently “fix” or silently repeat.

2. What this data shows — and what it does not

This dataset shows how each admitted source classified a stated participant population. It does not produce a national average, and it does not turn every student without a public-school record into a private-school student.

The complement rule

When a source measures the full population with mutually exclusive categories, the complement of “prior public” can be calculated as not identified as prior public. It still cannot be relabeled “prior private,” because the residual may contain homeschoolers, entering kindergarteners, out-of-state students, unmatched records or other categories.

When the numerator is restricted to selected cohorts or a matching window while the denominator covers a broader population, even the residual is invalid. North Carolina is the clearest example: the report identifies 12,252 prior-public students under specified cohort rules and expresses that number as 11.5% of all 106,704 recipients. Subtracting 11.5% from 100% would not produce a clean comparison group, so this dataset leaves the residual blank.

We would rather leave a cell blank than manufacture a precise-looking answer.

3. What “already in private school” means

Four evidence types are routinely compressed into the same phrase. The table below states what each type permits a writer to say.

Table 2. Evidence classification and permitted wording

Source: The School Choice Index, Voucher Prior-School Status Evidence Index methodology, version 1.0.

A ranked bar chart mixing Classes A, B and C would be neat, easy to read and wrong. This page uses an evidence table instead.

4. How this dataset was built

Version 1.0 reviews a fixed 18-state source list: Alabama, Arizona, Arkansas, Florida, Idaho, Indiana, Iowa, Louisiana, Montana, New Hampshire, North Carolina, Ohio, Oklahoma, Tennessee, Texas, Utah, West Virginia and Wyoming. Wisconsin’s WPCP is added as a separate benchmark because its official table splits prior-year Choice from prior-year non-Choice.

This is not a census of every voucher, ESA, tax-credit scholarship, town-tuitioning arrangement or disability-specific program in the United States. It is a defined source audit.

Source hierarchy

1. Issuing state agency or official program administrator.
2. Required annual, quarterly or legislative report.
3. State audit or official fiscal analysis.
4. Original research that identifies and explains its use of state administrative data.
5. Secondary reporting as a discovery lead only, never as the final production source for a statistic.

Every production row in Table 1 clears that gate. The Arizona 71% row comes from RAND’s original analysis of identified ADE data; the other rows come from official state or program materials.

Calculation rules

• Show the numerator and denominator whenever they are available.
• Preserve the source’s population label: applicant, approved student, awardee, recipient, entrant or active participant.
• Label every School Choice Index calculation.
• Preserve the source’s rounding, but expose an arithmetic discrepancy when the printed percentage does not match the displayed counts.
• Compute a complement only when the category and denominator support it.
• Never infer private-school history from a public-school residual.
• Never merge an all-participant stock with a first-time-entrant flow as though they were a time series.
• Never impute a missing field.
• Keep renewal participation separate from prior-school sector whenever the source permits it.

Dataset fields

The CSV records the state, program, scope role, source period, population type, denominator, numerator, direct prior-private field, prior-nonpublic field, homeschool field, combined private/homeschool field, prior-public field, renewal field, evidence class, calculation, permitted wording, prohibited wording, source location and verification date.

The “prohibited wording” column is deliberate. It records the most tempting misuse of each row so the limitation travels with the number.

5. Why renewals change the headline number

A prior-private stock is not the same as a new private-school-to-voucher flow. Wisconsin proves how large the difference can be, Arkansas shows how a revised renewal category changes the interpretation, and Indiana shows why a renewal rate cannot be used without a cross-tab.

Wisconsin: 94.4% of the prior-private group was already in Choice

The Wisconsin Department of Public Instruction’s WPCP table separates students who were in a Wisconsin private school and already in Choice from students who were in a Wisconsin private school without Choice.

Table 3. Wisconsin Parental Choice Program prior-year enrollment, 2025–26

Source: Wisconsin Department of Public Instruction, WPCP Facts and Figures for 2025-26, Table 2. Enrollment was measured as of the third Friday in September 2025; DPI labels the figures unaudited as of October 2025 and subject to change.

Three calculations follow from the table:

• Prior private: 19,047 + 1,129 = 20,176, or 86.2% of all participants.
• Renewals within the prior-private group: 19,047 ÷ 20,176 = 94.4%.
• Prior-year non-Choice: 1,129 ÷ 23,417 = 4.82% of all participants.

DPI’s footnote adds another layer. Of the 1,129 prior-year non-Choice students, 340 applied to kindergarten, 191 to first grade and 317 to ninth grade. That is 848 of 1,129, or 75.1%, in the three application grades identified in the footnote.

Among the reports admitted to this source set, WPCP is the only one that publishes this exact split. This dataset makes no claim that the split appears nowhere else in the country.

Arkansas: the revised report turns “unknown” into renewal

An earlier version of the Arkansas report said prior school type was not listed for one-third of participants. The revised official report explains why and no longer supports treating that group as ordinary missing data.

The updated narrative says one-third of participants were retained from the 2023–24 EFA year. Its footnote explains that the earlier report correctly observed that no prior school type appeared on those applications, but the reason was that the students were returning EFA participants. Arkansas’s 2023–24 family handbook states that homeschool was not an eligible EFA expense that year; the 2024–25 report calls 2024–25 the first year homeschool became an eligible pathway.

The approximation sign matters because the report publishes rounded shares, not the underlying headcounts for every prior-school category.

Indiana: renewal is visible, but the cross-tab is not

Indiana reports two separate facts for the same 76,067 participants:

• 53,285, or 70.05%, had no record of attending an Indiana public school at any point in their educational history.
• 58,433, or 76.82%, participated in the Choice Scholarship Program in the previous year.

Those figures clearly show that renewal is large. They do not show how many of the 53,285 no-public-record students were among the 58,433 renewals, because the report does not cross-tab the two fields. This dataset keeps both numbers and refuses to invent the overlap.

6. Which states directly measure prior private or nonpublic enrollment?

Among the 18 states in version 1.0, Arkansas and Iowa are the only two with a direct or reproducible prior-private/nonpublic measure admitted to the dataset.

Arkansas — approximately 58% renewal-inclusive

Table 4. Arkansas EFA previous school status, 2024–25

Source: 2024-25 Arkansas Education Freedom Accounts Program Annual Report, Figure 4 and updated footnote 3. The report covers 14,256 active participants.

The categories sum to 100%. The 33% is not an ordinary unknown field after the report’s revision. It is a program-retention category. The explicit prior-private category remains 25%; the renewal-inclusive program-history measure is approximately 58%.

A separate Arkansas figure answers a different question. In 2024–25, 10,834 active participants, or 76%, attended participating private schools and 3,422, or 24%, used EFAs for homeschooling. That is current sector, not previous school.

Iowa — approximately 66.3% previously nonpublic in 2023–24

The Iowa Department of Education reported that approximately 33.7% of 16,757 ESA participants at accredited nonpublic schools had not previously attended a nonpublic school. It separately reported 2,135 prior Iowa public-school students, about 12.7%, and 3,513 entering kindergarteners, close to 21%.

Source: Iowa Department of Education, certified enrollment release for 2023–24, January 26, 2024.

For 2024–25, the department reported 27,866 participants, 1,905 prior Iowa public-school students (6.8%) and 3,960 entering kindergarteners (14.2%). This dataset does notcalculate a 2024–25 prior-nonpublic share because the release does not state that those two categories exhaust everyone who was not previously nonpublic.

Participation reached 41,044 at accredited nonpublic schools in the October 1, 2025 certified count. That later participation total is useful context, but it does not replace the 2023–24 prior-school measure.

7. Which states combine private school with homeschool?

Texas publishes a combined private-school-or-homeschool chart. RAND’s Arizona estimate also combines the two. Neither source supports a private-school-only percentage.

Texas — 57% prior private school or homeschool

The Texas Education Freedom Accounts fact sheet reports:

• 43% prior public school
• 57% prior private school or homeschool

Source: Texas Education Freedom Accounts, 100K Awarded Applications Fact Sheet, June 10, 2026.

Three limits travel with that figure:

1. Private school and homeschool are not separable.
2. The chart says it reflects the 2024–25 school year and omits pre-K, kindergarten and first grade.
3. The fact sheet gives 102,037 total active awards, but it does not publish the denominator used for the prior-school chart.

The row describes active awards before a completed program year. It is not a full-year participation or retention measure.

Arizona RAND estimate — 71% prior private school or homeschool

RAND’s 2025 case study estimated that, among 64,133 universal-eligibility ESA users as of April 2025:

• 45,274, or 71%, were previously in private school or homeschool
• 11,813, or 18%, came immediately from public school
• 7,046, or 11%, entered as pre-K or kindergarten students

Source: RAND, Case Study of Arizona’s K–12 Education Savings Account Program: What State Policymakers Can Learn, Chapter 4, Figure 4.1 and footnote 6.

RAND says directly that the available data cannot disentangle the private-school group from the homeschool group. The estimate is derived from identified ADE administrative categories; it is not a private-school field published by ADE.

It is also not a time-series point comparable to Arizona’s Table 5 entrant percentages below. The populations and construction differ.

8. Which states report only public-school-history proxies?

Arizona, Indiana, North Carolina, Oklahoma and New Hampshire publish useful public-school-history measures. None of these rows directly measures the share previously in private school.

Arizona — one official entrant series, plus a printed arithmetic mismatch

Arizona’s FY2026 Quarter 3 report provides a four-year series for new universal-eligibility entrants in grades 1–12 who attended public school immediately before ESA enrollment.

Table 5. Arizona new universal entrants from public school

Source: Arizona Department of Education, ESA Program FY2026 Quarter 3 Report, Table 2. FY2026 covers July 1, 2025 through March 31, 2026. ⚠ denotes arithmetic discrepancy between agency printed value and recomputed value.

The table excludes kindergarteners and every eligibility category other than universal eligibility in grades 1–12. Its FY2026 denominator, 15,152, is 14.7%of the report’s 102,891 total ESA students. That is not a flaw; it is the defined population. It is a flaw only when the percentage is quoted as though it describes all current ESA participants.

The same report separately says 42,993 current students entered from public schools between FY2017 and FY2026 Q3. It also says that table excludes FY2011–FY2016 entrants because of a student-information-system migration. Dividing 42,993 by the full 102,891 would mix coverage periods, so this dataset does not publish that quotient.

Indiana — a lifetime Indiana public-school record

Indiana asks whether a student previously attended an Indiana public school at any point in the student’s educational history.

Table 6. Indiana Choice participants by Indiana public-school record

Source: Indiana Department of Education, 2024-2025 Annual Choice Report, Table 11.

“No record of attending an Indiana public school” is exact. “Never attended public school” is not: a student may have attended public school outside Indiana, entered kindergarten, been homeschooled or lacked a matched Indiana record.

North Carolina — 11.5% under a restricted cohort and matching method

North Carolina DPI reported 12,252 identified prior-public attendees, equal to 11.5% of all 106,704 Opportunity Scholarship recipients. The report also says only 80,879 recipients were in the three included fall/spring cohorts, and it excludes 636 spring-only recipients from the primary analysis because of timing.

Source: North Carolina DPI, Report to the North Carolina General Assembly: Opportunity Scholarships, June 4, 2026.

DPI defined prior-public attendees as recipients beginning with the 2024–25 scholarship year who were enrolled in public school for the majority of either the first or second month of the preceding school year. Because NCSEAA does not require the state’s unique student identifier, DPI used probabilistic matching on name, date of birth, address, county and available identifiers.

The state itself expresses 12,252 as 11.5% of all recipients. This dataset records that published percentage, but it does not calculate an 88.5% inverse. The numerator is produced under cohort and matching restrictions that do not define the rest of the full recipient population.

Oklahoma — 7.9% taxpayer-reported immediate prior public

The Oklahoma Tax Commission reports 3,120 prior-public students among 39,637 approved studentsfor the 2026–27 cycle.

The arithmetic complement is 92.1%, but the correct description is not reported as prior public in the immediately preceding semester. It is not a private-school percentage.

Source: Oklahoma Tax Commission, Parental Choice Tax Credit Report for School Year 2026-2027, July 2, 2026. The report says the prior-public field is based on what the taxpayer indicated on the application.

New Hampshire — the same 343 students produce 3.3% and 7.2%

New Hampshire’s November 2025 fact sheet reports:

• 10,510 total EFA recipients
• 4,745 new recipients
• 343 students who left public school the prior year

The same switcher count produces two valid percentages:

• 343 ÷ 10,510 = 3.3% of all recipients
• 343 ÷ 4,745 = 7.2% of new recipients

Source: New Hampshire Department of Education, November 2025 Education Freedom Account Data Fact Sheet, October 1 count.

The numbers do not conflict. They answer different questions. This is the cleanest denominator lesson in the dataset.

9. Which states publish no comparable distribution?

Ten states in the defined 18-state source set did not produce a prior-school distribution that cleared the version 1.0 source gate. That is a source-audit result, not proof that no administrative data exist.

Table 7. Version 1.0 source-gate status for all 18 audited states

Source: The School Choice Index source audit, last verified August 1, 2026.

Table 8. Official material reviewed for states without an admitted row

Source: Official issuing-agency and program-administrator materials linked in each row; source-gate review by The School Choice Index, last verified August 1, 2026.

A state can move into the admitted table when an official source publishes the missing denominator, category wording and observation population. The visible verification date and version history should change only after that source is actually checked.

10. Why the percentages differ

Most of the apparent spread comes from five measurement choices. Check these before quoting any state figure.

11. Why prior enrollment matters in fiscal analysis

Prior-school status changes the baseline of a fiscal calculation, but it does not settle the fiscal result by itself.

A participant identified as moving directly from public school and a participant already outside public school do not begin from the same state-funding baseline. That is why North Carolina’s statute requires a report comparing scholarship awards for identified prior-public students with the state per-pupil allocation, and why Oklahoma’s reporting statute separately requires the count of students who attended public school in the immediately preceding semester.

The North Carolina report also shows why the next step must remain separate. It calls its funding difference a modeled estimate under simplified timing assumptions and says it is not a realized current-year dollar difference. Prior-school status is one input. Award amounts, public-school funding formulas, timing, fixed costs, enrollment counts and statutory assumptions are other inputs.

This page does not turn a prior-school percentage into a savings or cost claim.

For adjacent questions, see School Voucher Tuition Gap by State, School Voucher Payments by Private School, Education Savings Account Statistics and School Choice Statistics.

12. Limitations of this dataset

The limits are part of the result, not a disclaimer pasted on afterward.

• The scope is defined, not exhaustive. Version 1.0 audits 18 named states and adds one Wisconsin benchmark.
• Source periods differ. Rows span 2023–24 through 2026–27 awards.
• Populations differ. Applicants, awards, participants, all recipients and new entrants cannot be merged.
• Several values are calculated. Arkansas’s approximately 58%, Iowa’s approximately 66.3%, Wisconsin’s 86.2% and 94.4%, and the public-history percentages calculated from counts are labeled as calculations.
• Published rounding can obscure the underlying count. Arkansas publishes rounded shares. Wisconsin’s displayed shares sum to 99.99% because of rounding.
• One official percentage does not match its displayed counts. Arizona prints 58.6%; 8,894 divided by 15,152 rounds to 58.7%.
• Some fields are self-reported. Oklahoma’s prior-public field is taxpayer-indicated. Wisconsin’s table is based on the program’s enrollment reporting and is labeled unaudited as of October 2025.
• Record systems have boundaries. Indiana measures Indiana public-school history; Arizona omits FY2011–FY2016 from one migration-affected table; North Carolina uses probabilistic matching.
• No national average is produced. Any national percentage attributed to this dataset would be a misuse.
• No causal claim is produced. The data show recorded status, not what a family would have done without the program.

Frequently asked questions

Downloads

How to cite this page

Primary sources

• Arizona Department of Education. Empowerment Scholarship Account Program — Fiscal Year 2026 Quarter 3 Report, pursuant to A.R.S. §15-2406. May 2026.
• Arkansas Department of Education and University of Arkansas Department of Education Reform. 2024-25 Arkansas Education Freedom Accounts Program Annual Report. October 2025.
• Arkansas Department of Education. Education Freedom Account Annual Report, 2023-2024. September 30, 2024.
• Arkansas Department of Education. EFA Family Handbook, 2023-2024. June 2023.
• Indiana Department of Education. 2024-2025 Annual Choice Report. 2025.
• Iowa Department of Education. Certified enrollment for 2023-24 holds steady; 16,757 ESA participants enrolled at Iowa accredited nonpublic schools. January 26, 2024.
• Iowa Department of Education. Certified enrollment for 2024-25 holds steady; 27,866 students participating in ESA program. January 17, 2025.
• Iowa Department of Education. Certified enrollment for 2025-26 slightly declines consistent with national trends. December 19, 2025.
• New Hampshire Department of Education. November 2025 Education Freedom Account Data Fact Sheet. November 2025.
• North Carolina Department of Public Instruction. Report to the North Carolina General Assembly: Opportunity Scholarships. June 4, 2026.
• Oklahoma Tax Commission. Parental Choice Tax Credit Report for School Year 2026-2027. July 2, 2026.
• RAND Corporation. Case Study of Arizona’s K–12 Education Savings Account Program: What State Policymakers Can Learn. November 13, 2025.
• Texas Comptroller of Public Accounts. Texas Education Freedom Accounts: 100K Awarded Applications Fact Sheet. June 10, 2026.
• Wisconsin Department of Public Instruction. WPCP Facts and Figures for 2025-26 As of October 2025. October 2025.