In 2025-26 the city had more high schools under 250 students than in any year since 2005-06, and nearly all of them were established schools that had shrunk, not new ones still filling in.
In the 2025-26 school year, — of the — regular public high schools in this analysis had fewer than 250 students, — had fewer than 300 and — had fewer than 350. The count under 250 has more than doubled since the 2019-20 low of —. The median high school had — students in 2025-26, down from — in 1994-95. Small schools also spend more. Compared with schools of 500 to 999 students that have similar shares of students with disabilities, English learners and low-income students, a school under 250 spent about — percent more per student in 2024-25.
Prompted by Chalkbeat New York's Oct. 28, 2026, story on small high schools in shared buildings. Read it.
Each line counts schools below that size, so the 350 line includes the schools under 300 and 250.
| 2025-26 | Schools | Students in them | Share of high school students |
|---|
The Bloomberg administration closed dozens of large high schools and opened scores of small ones in their buildings (Chalkbeat). The new schools typically opened with a ninth-grade class and added a grade a year, so for their first few years they were small by design; the state enrollment files show them as schools with no seniors yet. In 2004-05, — of the — schools under 250 were still phasing in. In 2025-26, — of — were established schools with a senior class.
"Still phasing in" means no 12th graders yet and in the state files less than four years. It is an estimate from enrollment, not a list of opening dates.
Average spending per student generally falls as schools get bigger. Part of the gap is the students: schools under 250 enroll larger shares of students with disabilities and English learners than schools with 500 to 999 students, and those students carry extra weight in the city's Fair Student Funding formula (slide 10, which lists the weights proposed for 2026-27). Small schools also have minimum staffing and operating costs spread across fewer students. Those costs may help explain the rest of the gap, alongside the city's funding protections and differences the model does not capture. The Education Department's FY2027 Fair Student Funding presentation lists a flat — foundation per school on top of per-student money (slide 9).
Enrollment on a log scale. Dashed lines at 250 and 350 students.
Percent above a 500-999-student school with the same shares of students with disabilities, English learners and low-income students, and the same middle-grade status. For schools under 250, the 95 percent confidence range is — to — percent; that reflects sampling error under this model, not factors it leaves out.
| 2024-25 | Schools | Students | Modeled difference | Per student |
|---|
These are model estimates: how much more these schools spent than the model predicts for schools of 500 to 999 students with the same measured student mix. They are not audited costs of small size alone, and they are not savings the city would book by merging schools, which would depend on which staff and buildings are kept. Results depend on the form of the model and the benchmark. A smooth model that compares each school with one of exactly 500 students estimates — for schools under 250. This analysis does not measure class size, adult attention or course breadth.
Murry Bergtraum, the school at the center of Chalkbeat's series, had — students and spent — per student in 2024-25 by the state's count (its state spending report). In October 2026, Chalkbeat reported that the school had one math teacher and two math courses last year, which would be 2025-26, the year after that spending figure.
Dollars per student, 2024-25, weighted by enrollment. The functions grouped here as shared costs (buildings, food, transportation, central offices) are nearly flat per student across size groups. The state's operating spending figures leave out some items, including debt service, so they are not an all-in cost.
| Size | Schools | Total | Instruction | Pupil support | School administration |
|---|
Chalkbeat's chart shows 113 high schools under 250 students in 2025, and its note says it excludes transfer schools, special education districts and charter schools (chart and data, retrieved Nov. 6, 2026). This page counts fewer. Under this page's school definition, — schools in the 2025-26 state files appear on the Education Department's transfer high school lists, from its School Quality Reports and earlier progress reports, and most of them are small. Keeping them in produces — schools under 250 in 2025-26; excluding them produces —. The two school lists have not been reconciled, and the reason for the difference has not been confirmed.
Sorted smallest first. Click a column to sort. "Extra vs. 500-999" is the need-adjusted premium the model gives the school's size group.
School enrollment comes from the New York State Education Department's public school enrollment files, 1994-95 through 2025-26 (BEDS day, the state's Basic Educational Data System count date; all students). Each year's count is taken in the fall: the 2025-26 counts are as of Oct. 1, 2025, in a file the state pulled on Oct. 8, 2026. A high school here is a New York City public school, not a charter, where grades 9 to 12 make up at least a quarter of enrollment. That keeps 6-12 schools and drops elementary and middle schools with a stray ninth grader. Excluded: District 75 special education schools, District 79 alternative programs, the pre-2004 alternative high school district (state code 77), schools through 2009-10 that share a borough and school number with one of that district's schools, and every school the Education Department has labeled a transfer high school in its progress reports (2006-07 to 2012-13) or School Quality Reports (2014-15 to 2023-24). The transfer list pools those designations and applies them to every year, including 2025-26; it was not checked against each year's designations. Size is total enrollment, including middle school grades at 6-12 schools, as in Chalkbeat's chart. School years are labeled by span; Chalkbeat appears to label them by the fall year.
A school counts as still phasing in if it has no 12th graders and first appears in the state files less than four years earlier. That is a proxy from enrollment, not a register of opening dates, and "established" is everything else. Schools are followed across the 2003-04 recode by borough and school number, and transfer labels from later years are applied to earlier ones, so a few schools' histories may be misassigned.
Spending per student is the state's 2024-25 school-level figure (federal, state and local; report card database, table "Expenditures per Pupil"). The breakdown comes from each school's 2024-25 Financial Transparency Report on data.nysed.gov (example: Murry Bergtraum): instruction is functions 1 and 3, pupil support function 2, school administration function 5, shared costs functions 4 and 6 to 12. Needs shares come from the state's 2024-25 enrollment files for students with disabilities, English language learners and economically disadvantaged students. The state suppresses some small subgroup counts, and a suppressed count is not a zero. For the 23 schools in the spending analysis whose English learner count is suppressed, the model uses total enrollment minus the published count of students who are not English learners (41 students across the 23 schools, zero to four at each). Those derived counts are used only inside the model; the school-level files leave them blank, as the state does. The $225,000 foundation and the funding weights are from the Education Department's FY2027 Fair Student Funding presentation (slides 9 and 10); slide 10 labels the weights as proposed for 2026-27, so the presentation alone does not establish the final adopted weights.
A regression of log spending per student on size group plus the three needs shares and a flag for schools with middle grades (420 schools; reference group 500-999 students; robust standard errors). It shows an association between size and spending among schools that look alike on those measures, not what small size itself causes. Each school's premium is its predicted spending at its own size minus its predicted spending at 500-999, times its enrollment. The dollar figures convert predicted log spending back to dollars, which estimates typical rather than average spending; a standard adjustment for that would raise them by about 0.3 percent. A smooth version using log enrollment instead of size groups gives a smaller estimate for schools under 250 (—). The two differ partly because the grouped model lets the extra cost climb steeply at the smallest sizes and partly because they use different benchmarks: one school of exactly 500 students, against the 500-999 group as a whole. The 2023-24 data show a somewhat steeper size gradient than 2024-25, but that check reuses 2024-25 needs shares, so it is not fully independent. The city's hold harmless policy, in place since 2023-24, has protected most of a school's prior-year funding at the initial allocation even as enrollment falls (city comptroller, FY2027 budget comments). Some of the small-school premium may reflect that policy; the model cannot say how much.
The functions grouped here as shared costs are nearly flat per student across size groups. These data cannot show whether that reflects how the state's reports allocate those costs or what buildings and central services actually cost small schools. The state's figures are operating spending and leave out debt service and some other items. The analysis cannot say what a merger would save, which depends on which staff and buildings are kept. Class sizes, adult attention and course offerings, the breadth problem in the Chalkbeat story, are not in these data.
The page's data: counts.csv, schools_2024_25.csv, size_bins_2024_25.csv, premium_2024_25.csv, fsf.csv. The scripts that build them are in the scripts folder.