RESEARCH · COUNTY DATASET VERSION 1.0
School Voucher Access by County: What 3,144 Counties Actually Show
3,144
U.S. county equivalents in the dataset (50 states + D.C.)
950
counties in 2027 participating states with zero NCES geocode rows (39.9%)
30
states on the IRS 2027 advance-election list as of July 24, 2026
$265,800
median provisional minimum estimate for a family of four across 3,144 counties
School voucher access by county is not one national yes-or-no number. Among the 2,378 counties in the 30 states on the IRS 2027 advance-election list, 950 — 39.9% — contain zero school-location records in the NCES 2023–24 private-school geocode file. Across all 3,144 U.S. counties and county equivalents, the count is 1,064 — 33.8%.
That does notmean 1,064 counties have no private school. It means the federal geocode file contains no completed 2023–24 Private School Survey respondent location assigned to those county GEOIDs. It also says nothing by itself about voucher participation, grade level, capacity, admissions, available seats, scholarship funding, or whether another qualified education expense can be used.
The supply is thin even before program rules are applied. In the 30 participating states, 1,421 counties — 59.8% — have zero or one NCES-listed location, and 1,943 — 81.7% — have four or fewer. Those are original calculations from the row-level release published with this page.
Section 25F is a federal tax credit for qualifying donations to scholarship-granting organizations, not a direct federal voucher paid to a household. A state must elect to participate and provide the IRS a qualifying SGO list before a donor can claim the credit, while student, household, SGO, award, expense, school, and provider rules remain separate questions. The IRS describes the 2027 framework here.
The numbers on this page
The headline finding is the 950-county overlap: counties in participating states that have zero respondent locations in the NCES geocode file. It is easy to quote, but it stays accurate only when the NCES-file definition remains attached to it.
| Measure | Finding | What the number means |
|---|---|---|
| Current U.S. counties and county equivalents | 3,144 | The 50 states and District of Columbia; territories excluded |
| States on the IRS 2027 advance-election list | 30 | A dated state action, not a family scholarship or award |
| Counties in those 30 states | 2,378 (75.6%) | The county geography covered by the state elections |
| Participating-state counties with zero NCES geocode rows | 950 (39.9%) | No 2023–24 PSS respondent location assigned to the county GEOID |
| All counties with zero NCES geocode rows | 1,064 (33.8%) | A federal source-file count, not proof that no private school exists |
| Participating-state counties with zero or one NCES row | 1,421 (59.8%) | A measure of thin recorded location supply |
| NCES geocoded private-school rows | 22,510 | Completed respondent locations, not seats or program-approved providers |
| County minimum family-of-four estimate above $200,000 | 2,832 counties (90.1%) | A provisional derived threshold comparison, not a finding that a family qualifies |
Source: The School Choice Index analysis of the IRS 2027 advance-election list, NCES 2023–24 Private School Locations layer, HUD FY2026 Income Limits data, and current Census county-equivalent geography. Dataset version 1.0; last verified August 1, 2026.
Search the school voucher access by county dataset
The county lookup separates state status, provisional income, and recorded private-school supply instead of turning them into one false eligibility label. Search by county, state, or five-digit GEOID, then read the field definitions before treating any result as an answer about a family or school.
The compact lookup displays county, state, IRS advance-election status, the selected household-size estimate, NCES row count, supply band, and whether the county contains more than one HUD median. The full downloads preserve all eight household sizes, HUD area names and codes, source IDs, source URLs, and methodology notes.
| Field | What it answers | What it does not answer |
|---|---|---|
| 2027 IRS advance election | Was the state on the dated federal advance-election list? | Whether an SGO is operating, funded, or awarding scholarships |
| Provisional 300%-of-AMGI estimate | What does the disclosed HUD-based calculation produce for household sizes 1–8? | An official IRS eligibility determination |
| NCES private-school count | How many 2023–24 geocoded PSS respondent locations were assigned to the county? | How many schools accept a voucher, have seats, or serve the required grade |
| HUD area names and codes | Does one or more HUD income area apply inside the county? | Which area applies without the residence location |
| Source IDs and URLs | Which primary source supports each row? | A substitute for checking later program updates |
Source: The School Choice Index data dictionary for dataset version 1.0.
Downloadable files
| File | Use |
|---|---|
| CSV | Flat county table for analysis and filtering |
| JSON | County records, state summaries, metadata, and source information |
| Excel workbook | Readable workbook with findings, definitions, methodology, claim ledger, and source ledger |
| State summary CSV | One row for each state and the District of Columbia |
| Reproducibility manifest | Release counts, QA results, and SHA-256 source-file hashes |
Source: The School Choice Index, dataset version 1.0. These files are available without a form, registration, or gated interaction.
Does my county have a private school that takes a voucher?
The federal location data cannot answer that question by itself.It can show whether the 2023–24 NCES geocode file assigns a respondent location to the county; the current state or program provider directory is the source for whether a particular school or provider is approved and participating.
A county result of zero should be read as “zero NCES geocode rows in this source vintage,” not “no school exists” and not “no scholarship can be used.” A result of one or more still requires separate checks for program approval, grade level, qualified expense, admissions, capacity, and award availability.
| Question | Best source | Why |
|---|---|---|
| Is the state participating in the 2027 federal framework? | IRS Section 25F state list | The IRS maintains the dated state-election record |
| Does the federal geocode file assign a private-school respondent location to the county? | NCES EDGE / PSS location file | It contains the county-coded location rows used in this dataset |
| Does a school accept a state scholarship or voucher? | Current state program provider directory | Approval and annual participation are program-specific |
| Does the school serve the student’s grade or qualified expense? | Provider profile and program rules | A provider can be approved without matching every student or expense |
| Will the student receive and be able to use an award? | Program administrator, SGO, and provider | Eligibility, funding, acceptance, and capacity are separate operational steps |
Source: The School Choice Index access-measure framework, built from the primary-source distinctions used in this release.
Why does Holmes County, Ohio show 111 private-school locations?
The NCES geocode file assigns 111 respondent locations to Holmes County, but that number is not an Ohio scholarship-provider count. Ohio maintains separate universes for chartered nonpublic schools, nonchartered nonpublic schools, and schools registered to participate in EdChoice.
Ohio’s official term is Nonchartered Nonpublic school, or NCNP. The Ohio Department of Education and Workforce says NCNP schools choose not to be chartered because of truly held religious beliefs, file an annual certification, and are outside the Department’s legal authority to regulate curriculum. Its separate private-school page explains that the state’s chartered-school directory does not include schools that are not chartered. Ohio’s definitions are here.
For EdChoice, the state says a participating provider must hold a current state charter, register for the program, and renew participation each year. A family must find a participating private school, be accepted for enrollment, and have the school submit the scholarship application. Ohio’s provider requirements are here and the family application steps are here.
| Holmes County measure | Verified result | Correct interpretation |
|---|---|---|
| NCES 2023–24 geocode rows | 111 | Geocoded locations from the federal PSS respondent file |
| Ohio charter status | Separate state universe | Ohio distinguishes chartered nonpublic schools from nonchartered nonpublic schools |
| EdChoice participation | Separate live provider universe | A provider must hold a current state charter, register for EdChoice, and renew participation |
| Participating-provider count in this release | Not published | Version 1.0 does not contain a preserved county extract from Ohio’s live provider directory |
Source: The School Choice Index count of NCES 2023–24 geocode rows for GEOID 39075; Ohio Department of Education and Workforce private-school and EdChoice provider materials, checked August 1, 2026.
The accurate sentence is plain: Holmes County has 111 rows in the NCES file. That is neither zero private schools nor 111 approved scholarship providers. This release does not publish an exact Holmes County EdChoice-provider count because the live state provider directory was not captured as a versioned county extract for this dataset.
What do county-level voucher figures actually measure?
County figures fall into different measurement layers, and the layers are not interchangeable. A location count, an approved-provider count, a grade match, a travel-distance estimate, and an actual scholarship use count answer different questions even when every number is correct.
| Code | Measure | Question answered | Unit |
|---|---|---|---|
| A1 | NCES-listed location presence | Does the dated NCES geocode file assign at least one private-school respondent location to the county? | Counties / locations |
| A2 | Approved provider presence | Does a school or provider on the dated program roster operate in the county? | Counties / providers |
| A3 | Grade-and-expense match | Does an approved provider serve the relevant grade or qualified expense? | Students / providers |
| A4 | Proximity | How far is the nearest potentially usable school or service from the household? | Families / distance / travel time |
| A5 | Actual use and capacity | Was an award available, was the student accepted, and was a seat or service actually used? | Students / awards / seats |
| A6 | Proxy | What county characteristic is being used as a stand-in for access? | Varies |
Source: The School Choice Index. Framework developed for this page from the source and denominator distinctions required to reproduce the county dataset.
The narrow ordering rule
When the same dated school universe is filtered through the same program roster and grade rules, the counts are weakly ordered: A1 ≥ A2 ≥ A3. A school cannot pass the roster and grade filters if it is absent from the starting universe.
That rule does not license comparisons between unrelated files, different years, or different provider universes. A live state directory can include providers absent from an older NCES response file, while an NCES row can be absent from a state roster. A4 proximity and A5 actual use are not simple subsets of the county-presence fields.
A6 is not direct access. It is a proxy such as county population, public enrollment, rurality, or district density and must be labeled as a proxy every time it appears.
School voucher access by county statistics
The national distribution is more concentrated at the low end than the headline zero count alone shows. More than half of all counties have zero or one NCES-listed location, and nearly three-fifths of counties in the 30 participating states do.
| NCES location band | All counties | Counties in 2027 participating states |
|---|---|---|
| 0 NCES geocode rows | 1,064 | 950 |
| 1 NCES geocode row | 572 | 471 |
| 2–4 NCES geocode rows | 702 | 522 |
| 5–9 NCES geocode rows | 338 | 203 |
| 10+ NCES geocode rows | 468 | 232 |
Source: The School Choice Index aggregation of all 22,510 rows in the NCES 2023–24 Private School Locations file. Counts are county-location records, not program participants.
Among participating states, 11 have zero NCES rows in more than half of their counties. Texas contributes the largest number — 132 — while Alaska has the largest share at 80.0%.
| State | Zero-location counties | All counties | Share with zero NCES rows |
|---|---|---|---|
| Alaska | 24 | 30 | 80.0% |
| North Dakota | 38 | 53 | 71.7% |
| South Dakota | 42 | 66 | 63.6% |
| Oklahoma | 48 | 77 | 62.3% |
| Kansas | 65 | 105 | 61.9% |
| Nevada | 10 | 17 | 58.8% |
| Arkansas | 42 | 75 | 56.0% |
| Montana | 30 | 56 | 53.6% |
| Wyoming | 12 | 23 | 52.2% |
| Texas | 132 | 254 | 52.0% |
| Nebraska | 48 | 93 | 51.6% |
Source: The School Choice Index county aggregation joined to the IRS 2027 advance-election list, status date July 24, 2026.
The finding is geographic, not a verdict on policy effectiveness. A county can have no NCES row and still have a cross-county option or another qualified service; a county can have many NCES rows and still have no provider that matches a particular program, grade, student, or expense.
Which states are participating in the 2027 federal scholarship tax credit?
As of July 24, 2026, 30 states had made an advance election under Section 25F for calendar year 2027. The IRS page was last updated July 27, 2026, and the dataset preserves the July 24 status date carried by the list.
| Participating states | Participating states | Participating states |
|---|---|---|
| Alabama | Kentucky | Ohio |
| Alaska | Louisiana | Oklahoma |
| Arkansas | Mississippi | South Carolina |
| Colorado | Missouri | South Dakota |
| Florida | Montana | Tennessee |
| Georgia | Nebraska | Texas |
| Idaho | Nevada | Utah |
| Indiana | New Hampshire | Virginia |
| Iowa | North Carolina | West Virginia |
| Kansas | North Dakota | Wyoming |
Source: Internal Revenue Service, Federal Scholarship Tax Credit: Participating states for 2027, status date July 24, 2026.
Beginning January 1, 2027, an individual taxpayer may be able to claim a nonrefundable federal credit of up to $1,700 for qualifying cash contributions to an SGO. The IRS says excess credit generally may be carried forward for up to five years. The IRS’s current overview is here.
| The state election establishes | The state election does not establish |
|---|---|
| The state made an advance election for calendar year 2027 | That a particular SGO has completed all operating steps |
| The state intends to participate in the federal donor-credit framework | That donations or scholarship money are already available |
| The state must provide the IRS a qualifying SGO list before donors can claim the credit | That a particular student meets the statutory and SGO requirements |
| A dated state-status field for this dataset | That a private school or provider will accept a scholarship |
Source: IRS Section 25F program page, Working Families Tax Cuts overview, Revenue Procedure 2026-6, and Notice 2025-70.
How high is the provisional 300%-of-area-median-income figure by county?
The dataset reports a transparent county-area estimate, not an official IRS cutoff.It applies one disclosed, reproducible implementation of the statutory cross-references: FY2026 HUD median family income, HUD’s standard household-size adjustments, and the statutory 300% multiplier. Treasury and the IRS are still developing Section 25F regulations.
Section 25F defines an eligible student’s prior-calendar-year household income as no greater than 300% of area median gross income “as such term is used in section 42.” Section 42(g)(4) applies Section 142(d)(2), whose subparagraph (B) requires income and area median gross income to be determined consistently with Section 8 and expressly includes adjustments for family size. Read Section 25F, Section 42, and Section 142.
The version 1.0 calculation is:
Provisional household estimate = FY2026 HUD median family income × HUD household-size factor × 3
| Household size | HUD family-size factor used in the model |
|---|---|
| 1 | 70% |
| 2 | 80% |
| 3 | 90% |
| 4 | 100% |
| 5 | 108% |
| 6 | 116% |
| 7 | 124% |
| 8 | 132% |
Source: HUD FY2026 Section 8 Income Limits notice and methodology. Household size four is the 100% base; the other factors are HUD’s standard family-size adjustments.
| Derived finding | Result |
|---|---|
| Median of the 3,144 county-level minimum estimates, family of four | $265,800 |
| Counties whose minimum estimate exceeds exactly $200,000 | 2,832 (90.1%) |
| National minimum | Oglala Lakota County, SD: $107,100 |
| National maximum | Santa Clara County, CA: $616,500 |
| Maximum in a 2027 participating state | Summit County, UT: $503,100 |
| County equivalents with more than one distinct FY2026 HUD median | 15 |
Source: The School Choice Index calculation from the FY2026 HUD workbook and the disclosed formula. “Median” means the median of the 3,144 county-level minimum estimates; it is not a national household-income median.
The $200,000 result has a precise meaning: a household income of exactly $200,000 is below the provisional county minimum in 2,832 county equivalents. It does not mean every such household satisfies the final Section 25F rules, receives a scholarship, or has a usable provider.
Why 15 counties show a range
HUD income areas do not always align to one whole county. Sixty-eight county equivalents have multiple source records, but only 15 contain more than one distinct FY2026 median family income. Those 15 require a minimum and maximum; averaging them would invent a county-wide value that HUD did not publish.
| County or equivalent | Provisional family-of-four range | Difference |
|---|---|---|
| Bristol County, MA | $282,900–$496,800 | $213,900 |
| Worcester County, MA | $306,000–$448,200 | $142,200 |
| Crawford County, MO | $224,100–$340,500 | $116,400 |
| Norfolk County, MA | $383,400–$493,800 | $110,400 |
| Plymouth County, MA | $383,400–$493,800 | $110,400 |
| York County, ME | $320,400–$417,300 | $96,900 |
| Cumberland County, ME | $332,400–$417,300 | $84,900 |
| Penobscot County, ME | $237,000–$319,500 | $82,500 |
| Essex County, MA | $411,900–$493,800 | $81,900 |
| Rockingham County, NH | $411,900–$493,800 | $81,900 |
| Hillsborough County, NH | $365,100–$442,800 | $77,700 |
| Newport County, RI | $336,300–$413,100 | $76,800 |
| Middlesex County, MA | $429,000–$493,800 | $64,800 |
| Washington County, RI | $336,300–$387,000 | $50,700 |
| Berkshire County, MA | $328,200–$338,700 | $10,500 |
Source: The School Choice Index grouping of the HUD FY2026 source workbook and HUD area definitions.
Crawford County, Missouri includes HUD’s separately coded “Sullivan city part” area. Version 1.0 folds that pseudo-area into current Crawford County GEOID 29055 while preserving both HUD area names and both values.
Treasury said on June 10, 2026 that it expected to issue proposed regulations by the end of September 2026. That Treasury notice is here. The income table must be rebuilt or relabeled if those regulations change the operational treatment of area, family size, income, or source vintage.
Where do national private-school proximity statistics come from?
The national distance figures audited here come from a Brookings study published in March 2017, using 2011–12 private elementary-school locations. They are historical straight-line proximity estimates under an assumed universal program in which every private elementary school participates, not current voucher-provider or travel-time statistics.
Brookings measured straight-line distance from the population mean center of each Census block group to nearby elementary schools. The authors explicitly described the private-school result as an upper bound and excluded capacity, existing program rules, admissions, school participation, and changes in supply.
| Straight-line radius | All families | Below poverty | At or above poverty | Rural | Urban |
|---|---|---|---|---|---|
| 1 mile | 35% | 38% | 34% | 2% | 41% |
| 2 miles | 60% | 62% | 59% | 8% | 70% |
| 5 miles | 82% | 80% | 82% | 34% | 92% |
| 10 miles | 92% | 88% | 92% | 69% | 96% |
Source: Kristin Blagg and Matthew M. Chingos, “Who could benefit from school choice? Mapping access to public and private schools,” Brookings Institution, March 31, 2017. Private-school locations are from the 2011–12 NCES PSS; distances are straight-line.
Three figures that are often repeated as though they disagree — 34% of rural families within five miles, 69% of rural families within ten miles, and 92% of all families within ten miles— are cells in the same figure. The denominator and distance have to travel with the number.
The Brookings values remain useful as a clearly defined historical benchmark. They are not 2026 or 2027 access figures, current drive time, current provider participation, or current school capacity.
Full state table
The state table combines the current county aggregation with the dated federal participation field and one clearly labeled historical proximity benchmark.The NCES columns use the 2023–24 geocode release; the Brookings column uses 2011–12 private elementary-school locations and must not be read as current program access.
| State | Counties | Zero NCES | Zero share | NCES rows | 2027 IRS | Brookings 5 mi. |
|---|---|---|---|---|---|---|
| Alabama | 67 | 19 | 28.4% | 266 | Yes | 63% |
| Alaska | 30 | 24 | 80.0% | 25 | Yes | 73% |
| Arizona | 15 | 2 | 13.3% | 307 | No | 83% |
| Arkansas | 75 | 42 | 56.0% | 120 | Yes | 58% |
| California | 58 | 5 | 8.6% | 2,335 | No | 95% |
| Colorado | 64 | 31 | 48.4% | 272 | Yes | 84% |
| Connecticut | 9 | 0 | 0.0% | 232 | No | 92% |
| Delaware | 3 | 0 | 0.0% | 74 | No | 85% |
| District of Columbia | 1 | 0 | 0.0% | 50 | No | 100% |
| Florida | 67 | 3 | 4.5% | 2,004 | Yes | 90% |
| Georgia | 159 | 67 | 42.1% | 620 | Yes | 78% |
| Hawaii | 5 | 1 | 20.0% | 100 | No | 93% |
| Idaho | 44 | 20 | 45.5% | 119 | Yes | 71% |
| Illinois | 102 | 31 | 30.4% | 971 | No | 89% |
| Indiana | 92 | 22 | 23.9% | 590 | Yes | 79% |
| Iowa | 99 | 42 | 42.4% | 186 | Yes | 65% |
| Kansas | 105 | 65 | 61.9% | 165 | Yes | 73% |
| Kentucky | 120 | 55 | 45.8% | 270 | Yes | 62% |
| Louisiana | 64 | 14 | 21.9% | 282 | Yes | 75% |
| Maine | 16 | 0 | 0.0% | 104 | No | 58% |
| Maryland | 24 | 1 | 4.2% | 543 | No | 92% |
| Massachusetts | 14 | 0 | 0.0% | 504 | No | 95% |
| Michigan | 83 | 13 | 15.7% | 564 | No | 83% |
| Minnesota | 87 | 18 | 20.7% | 406 | No | 79% |
| Mississippi | 82 | 25 | 30.5% | 174 | Yes | 57% |
| Missouri | 115 | 35 | 30.4% | 501 | Yes | 73% |
| Montana | 56 | 30 | 53.6% | 109 | Yes | 58% |
| Nebraska | 93 | 48 | 51.6% | 174 | Yes | 76% |
| Nevada | 17 | 10 | 58.8% | 95 | Yes | 90% |
| New Hampshire | 10 | 0 | 0.0% | 134 | Yes | 70% |
| New Jersey | 21 | 0 | 0.0% | 784 | No | 96% |
| New Mexico | 33 | 14 | 42.4% | 124 | No | 71% |
| New York | 62 | 3 | 4.8% | 1,231 | No | 90% |
| North Carolina | 100 | 15 | 15.0% | 594 | Yes | 73% |
| North Dakota | 53 | 38 | 71.7% | 48 | Yes | 60% |
| Ohio | 88 | 9 | 10.2% | 1,006 | Yes | 81% |
| Oklahoma | 77 | 48 | 62.3% | 140 | Yes | 61% |
| Oregon | 36 | 8 | 22.2% | 290 | No | 86% |
| Pennsylvania | 67 | 3 | 4.5% | 1,567 | No | 90% |
| Rhode Island | 5 | 0 | 0.0% | 87 | No | 96% |
| South Carolina | 46 | 4 | 8.7% | 293 | Yes | 73% |
| South Dakota | 66 | 42 | 63.6% | 69 | Yes | 55% |
| Tennessee | 95 | 34 | 35.8% | 391 | Yes | 70% |
| Texas | 254 | 132 | 52.0% | 1,330 | Yes | 82% |
| Utah | 29 | 12 | 41.4% | 140 | Yes | 68% |
| Vermont | 14 | 2 | 14.3% | 93 | No | 48% |
| Virginia | 133 | 35 | 26.3% | 651 | Yes | 81% |
| Washington | 39 | 6 | 15.4% | 517 | No | 85% |
| West Virginia | 55 | 17 | 30.9% | 100 | Yes | 55% |
| Wisconsin | 72 | 7 | 9.7% | 730 | No | 83% |
| Wyoming | 23 | 12 | 52.2% | 29 | Yes | 59% |
Source: The School Choice Index county dataset version 1.0; IRS 2027 advance-election list; NCES 2023–24 Private School Locations; Census county-equivalent geography; and Brookings 2017 proximity analysis. The Brookings column is historical and uses 2011–12 private elementary-school locations.
The table sums to 3,144 county equivalents, 1,064 zero-row counties, and 22,510 NCES geocode rows. Those totals are release checks, not separately entered editorial figures.
How we built the school voucher access by county dataset
The release is a reproducible multi-source join, not a survey conducted by The School Choice Index. Every original statistic comes from the 3,144 county rows and can be recalculated from the public downloads and stated source files.
Source hierarchy
| Source | Version or status date | What this release uses |
|---|---|---|
| IRS Federal Scholarship Tax Credit page | Participating states as of July 24, 2026 | The 2027 state advance-election field |
| 26 U.S.C. §§ 25F, 42, and 142 | Current preliminary U.S. Code checked August 1, 2026 | The statutory income language and cross-reference chain |
| HUD FY2026 Section 8 Income Limits workbook and notice | FY2026 | Median family income, area definitions, and household-size factors |
| NCES Private School Locations / PSS | School year 2023–24; First Look published May 2026 | The 22,510 geocoded respondent-location rows |
| Census county geography | Current geography checked August 1, 2026 | The 3,144-county universe and current Connecticut regions |
| Brookings proximity report | Published March 31, 2017; 2011–12 private-school locations | Historical straight-line proximity context only |
Source: The School Choice Index source ledger, dataset version 1.0.
1. Build the current county universe
We began with Census county codes for the 50 states and District of Columbia, then used current geography for Connecticut. Census replaced Connecticut’s eight former counties with nine Councils of Governments or planning regions as county equivalents. The Census change notice is here. The final geography contains 3,144 unique five-digit GEOIDs. Territories are not included.
2. Add the dated state-status layer
Every county inherits its state’s “Yes” or “No” status from the IRS advance-election list dated July 24, 2026. “No” means the state was not on that list at that status date; it does not mean the state can never act later.
3. Aggregate the NCES geocode rows
We counted each unique PPIN row in the 2023–24 NCES private-school geocode file and grouped the rows by the NCES state and county codes. The release contains 22,510 unique rows and zero county codes outside the canonical geography.
4. Group HUD records by county equivalent
We grouped all FY2026 HUD records by current county GEOID. Where more than one record produced the same median, the county retains the source-area names without publishing a false range. Where the medians differ, it reports the minimum and maximum.
HUD pseudo-code 29056, “Sullivan city part,” was assigned to Crawford County, Missouri GEOID 29055 because that area lies inside the current county. Both source records remain named in the row.
5. Calculate household-size estimates
For each HUD median, we multiplied by the disclosed household-size factor and by three. All eight household sizes remain in the CSV, JSON, and workbook; the page uses the family-of-four result for compact comparisons.
6. Run the release tests
| Release test | Result |
|---|---|
| County rows | 3,144 |
| Unique five-digit GEOIDs | 3,144 |
| HUD source rows grouped | 4,764 |
| Canonical counties without a HUD match | 0 |
| NCES geocode rows | 22,510 |
| Unique NCES PPIN identifiers | 22,510 |
| NCES county codes outside the canonical geography | 0 |
| Participating states at the dated IRS status | 30 |
| Current Connecticut planning-region county equivalents | 9 |
Source: The School Choice Index reproducibility manifest. The manifest also publishes SHA-256 hashes for the three local source files used in the build.
The source hashes preserved in the manifest are part of the methodology. They make it possible to distinguish a calculation change from a later upstream-file revision.
What this dataset shows — and what it does not show
This release measures three access layers and refuses to collapse them into a legal or practical verdict. It is strongest when its scope remains narrow: dated state status, a disclosed provisional income model, and a dated federal location-file count.
| This dataset shows | This dataset does not show |
|---|---|
| Whether a state was on the dated IRS 2027 advance-election list | Whether a scholarship is funded or available to a particular family |
| A reproducible provisional local income estimate for household sizes 1–8 | An official IRS cutoff or individualized tax determination |
| How many NCES 2023–24 geocode rows are assigned to a county | The complete number of private schools currently operating |
| Whether more than one HUD median applies inside a county equivalent | Which HUD area applies without the residence location |
| A dated historical proximity benchmark from Brookings | Current drive time, current school participation, capacity, admissions, or available seats |
| State and county aggregates derived from the row-level release | Every existing state voucher, ESA, tax-credit scholarship, town-tuitioning program, or public-school choice rule |
Source: The School Choice Index methodology and the limitations of the underlying IRS, HUD, Census, NCES, and Brookings sources.
The NCES file is not the weighted national school estimate
The NCES First Look report published in May 2026 gives the response disposition for the 2023–24 Private School Survey. The geocode release used here has 22,510 rows, matching the report’s completed-interview count.
| 2023–24 PSS disposition | Unweighted cases |
|---|---|
| Cases included in data collection | 57,265 |
| Out of scope | 27,946 |
| Completed interviews | 22,510 |
| Nonresponse | 6,809 |
| Unweighted response rate | 76.8% |
| Base-weighted response rate | 76.0% |
| Weighted national estimate of private schools | 30,553 |
Source: NCES 2026-015, “Characteristics of Private Schools in the United States: Results From the 2023–24 Private School Universe Survey”, May 2026.
The weighted national estimate of 30,553 schools and the unweighted 22,510 completed records answer different questions. Comparing them as though 22,510 were a prior-year national total would manufacture a decline that the data do not show.
Additional limits
- Version 1.0 does not inventory every state voucher, education savings account, tax-credit scholarship, town-tuitioning program, public-school choice rule, or provider roster.
- State advance election does not establish SGO funding, award availability, student eligibility, or provider acceptance.
- The income values are provisional derived estimates pending Section 25F regulations and operational guidance.
- HUD areas can be smaller than a county; a range does not identify which value applies to a residence.
- NCES locations are not seats, enrollment capacity, tuition, quality, accreditation, admissions, or program participation.
- The 2023–24 source can lag openings, closures, relocations, or reporting changes.
- The Brookings proximity figures are historical, straight-line, elementary-school estimates based on 2011–12 school locations.
- The data cover the 50 states and District of Columbia, not U.S. territories.
Why the county layer matters for the 2027 launch
The 2027 state map covers three-quarters of U.S. county equivalents, but the recorded private-school supply inside that footprint is uneven. Forty percent of participating-state counties have zero NCES rows, nearly 60% have zero or one, and 11 participating states have zero rows in more than half of their counties.
The income layer points in a different direction. The median county-level minimum estimate for a family of four is $265,800, and the derived minimum exceeds $200,000 in 90.1% of counties. A wide provisional income range therefore does not erase the supply, provider, grade, capacity, travel, or award layers.
That is why one national “access” percentage fails. The legal state footprint is broad, the recorded county supply is often thin, the provisional income ceiling is often high, and none of those facts alone tells a family whether an award can be used.
The refresh triggers are a federal rule change, an update to the IRS state or SGO lists, or a new HUD, NCES, or Census source vintage. The visible date changes only after the affected source and QA checks are rerun.
How often will this data be updated?
The visible date changes only after the affected primary sources are checked and the complete dataset is rebuilt or revalidated. State participation and federal rules need the fastest monitoring; HUD, NCES, and Census follow separate release cycles.
| Element | Refresh cadence | Verification method |
|---|---|---|
| IRS participating-state list | Weekly through the 2027 launch; monthly afterward | Recheck the official IRS status date and state list |
| Treasury and IRS implementation rules | Immediately after proposed or final guidance | Compare definitions, income treatment, SGO requirements, and effective dates |
| HUD median family income | Annual | Rebuild from the new fiscal-year workbook and notice |
| NCES private-school locations | Quarterly release check; rebuild when a new file appears | Download, hash, count unique IDs, and rerun county joins |
| Census county geography | Annual | Check Gazetteer/TIGER releases and geography notices |
| Headline statistics | Every source refresh | Recompute from row-level data; do not edit totals by hand |
| Visible update date and dataset version | Only after the complete source and QA review | Publish a version note and preserve the prior release |
Source: The School Choice Index release protocol for this dataset.
Any correction records the old value, the new value, the source that changed, and the dataset version in which the correction appeared. Silent number changes make a dated research page harder to use and audit.
Frequently asked questions
These answers keep the state, income, school-location, provider, and actual-use layers separate. A county row can narrow the question, but it cannot replace the official program administrator, SGO, school, or qualified tax professional for an individual determination.
Is Section 25F itself a school voucher?
Section 25F is a federal individual tax credit for qualifying cash contributions to scholarship-granting organizations. SGOs use qualifying funds for elementary and secondary education scholarships, but the federal mechanism is not a direct voucher payment from the federal government to a family.
Does a state’s advance election guarantee scholarships?
No. It records the state’s advance election for 2027. The qualifying SGO list, donations, SGO operations, scholarship funding, student verification, award decisions, qualified expenses, and provider participation are separate steps.
Does zero NCES-listed locations mean there is no private school in the county?
No. It means the 2023–24 geocode release contains no completed PSS respondent location assigned to that county GEOID. The source can miss nonrespondents and can lag later openings, closures, and moves.
Does one NCES-listed school mean a voucher can be used there?
No. The school may not participate in the relevant program, may not serve the required grade, may not have capacity, or may not accept the student. Use the current program-provider directory and the school’s own information.
Why are the income figures called provisional?
Treasury said proposed Section 25F regulations were expected by the end of September 2026. This page was verified before that date. The dataset publishes the exact statutory cross-reference and HUD-based calculation so the result is reproducible without pretending it is a final IRS table.
Which household-income year does Section 25F use?
The statute refers to household income for the calendar year before the scholarship application date. The SGO must verify annual household income and family size under the federal framework.
Why can one county show two income values?
More than one FY2026 HUD income area can apply inside a county equivalent. When the source medians differ, the dataset publishes the minimum and maximum and directs the reader to the residence-specific HUD area.
Can a scholarship be used across a county line?
The county table does not impose or test a county-line rule. The answer depends on the SGO, program, qualified expense, provider rules, admissions, and practical travel or service availability.
Are the 22,510 NCES locations confirmed scholarship providers?
No. They are geocoded respondent locations from the 2023–24 Private School Survey. Program approval is a separate state or SGO roster.
Does this dataset include every state voucher or ESA program?
No. Version 1.0 is deliberately scoped to the 2027 federal Section 25F state layer, provisional local income estimates, and the NCES location supply field. For broader state-program context, use the related research below.
How to cite this page
This block identifies the page, publisher, version, and verification date so the same release can be distinguished from later updates. It is neutral attribution information, not a request.
The School Choice Index Editorial Team. “School Voucher Access by County: What 3,144 Counties Actually Show.” The School Choice Index. Dataset version 1.0. Last verified August 1, 2026. https://theschoolchoiceindex.com/research/school-voucher-access-by-county/
For a claim derived from the downloadable data, keep the field definition and source vintage with the number. For example: “The School Choice Index found that 950 counties in states on the IRS 2027 advance-election list had zero school-location records in the NCES 2023–24 private-school geocode file.”
Published: · Last verified: · Dataset version: 1.0
Primary sources
The page relies on issuing agencies, statutory text, original federal datasets, and the original Brookings proximity study.Secondary articles are not used to establish the release’s headline counts.
- IRS: Federal Scholarship Tax Credit and participating states for 2027
- IRS: Working Families Tax Cuts — Federal Scholarship Tax Credit overview
- IRS: Notice 2025-70 in Internal Revenue Bulletin 2025-50
- U.S. Code: 26 U.S.C. §25F
- U.S. Code: 26 U.S.C. §42
- U.S. Code: 26 U.S.C. §142
- Treasury: Education Freedom Tax Credit guidance preview
- HUD: FY2026 Income Limits
- HUD: FY2026 Section 8 source workbook
- HUD: FY2026 Income Limits notice
- NCES: Private School Locations 2023–24
- NCES: 2023–24 PSS First Look, NCES 2026-015
- Census: Connecticut county-equivalent change
- Census: Gazetteer and current geography files
- Brookings: 2017 private-school proximity study
- Ohio Department of Education and Workforce: Private Schools
- Ohio Department of Education and Workforce: EdChoice Provider Information
Related research
These pages cover adjacent questions that version 1.0 intentionally leaves separate. Keeping the subjects separate prevents a county supply count from being mistaken for state-program eligibility, payment size, or affordability.
Who produced this research
The School Choice Index Editorial Team produced the release by joining and checking the listed primary sources, calculating the county statistics, documenting the transformations, and publishing the row-level files.The School Choice Index Research desk is an independent, non-partisan reference section that assembles primary-source K–12 school-choice data into transparent, reproducible datasets.
This page exists to make the county evidence easier to inspect without hiding what the sources cannot establish. It provides educational research and general reference information, not legal, tax, accounting, or individualized eligibility advice.