The Minnesota School Districts With the Most Money Per Pupil Have the Lowest Reading Scores. That Is the Funding Formula Working — and It Is Why the Public Data Cannot Tell You Whether the Formula Works.

Same rules as the rest of this section: a named public dataset for every claim, every disparity stated with its denominator, the innocent explanation weighed first, and a fix a real office could actually carry out.

I set out to put per-pupil revenue beside reading outcomes and see whether Minnesota's school funding formula is buying anything. What the public data gives you is a number that looks damning, means nothing, and cannot be corrected with the files the state and the federal government currently publish. So the question this piece answers is not "does money work." It is why nobody working from Minnesota's public files can answer that, and what would have to be published before they could.

An earlier piece here covered where Minnesota's reading scores stand and what the state has done about them: Half of Minnesota's Kids Can't Read at Grade Level. This is a different question — not how the state is doing, but whether the machine that moves the money has anything to do with it.

The machine, from the statute

Minn. Stat. § 126C.10, subd. 1 defines general education revenue as the sum of fourteen things: "basic revenue, extended time revenue, gifted and talented revenue, declining enrollment revenue, local optional revenue, small schools revenue, basic skills revenue, secondary sparsity revenue, elementary sparsity revenue, transportation sparsity revenue, total operating capital revenue, equity revenue, pension adjustment revenue, and transition revenue."

Read that list with one question in mind: which of those components rise because a district has more need? Basic revenue does not — it is flat per pupil, the formula allowance times adjusted pupil units, $7,281 for fiscal year 2025 under subd. 2(a), rising after that by the greater of two percent or inflation capped at three percent under subd. 2(b). Neither do local optional revenue, a flat $724 per adjusted pupil unit under subd. 2e in two tiers of $300 and $424, or gifted and talented revenue at $13 under subd. 2b. But the components that carry the real variation between districts are all need adjustments: sparsity revenue for districts too spread out to be efficient, small schools revenue, declining enrollment revenue, and basic skills revenue, which subd. 4 defines as compensatory revenue plus English learner revenue.

Compensatory revenue is the big one, and it is targeted with real precision. Under subd. 3a, a "building concentration factor" is the ratio of a building's compensatory pupils to the pupils enrolled there on October 1 of the previous year, and "compensatory pupils" means the pupils eligible for free meals plus half of those eligible for reduced-price meals. Subdivision 3c sets the statewide compensatory allowance at $6,936 for fiscal year 2027, and subd. 3(g) puts a floor under the program: if the building-by-building sum comes in below $857,152,000 for fiscal 2027 or later, the commissioner must scale every building up until it reaches that number.

A statutory floor of roughly $857 million a year, distributed in proportion to concentrated poverty. That single design choice is why the rest of this article comes out the way it does.

The number that looks like an answer

The U.S. Census Bureau's Annual Survey of School System Finances — the F-33 — publishes revenue, spending, and fall enrollment for every school system in the country. The U.S. Department of Education's EDFacts collection publishes reading proficiency by school district and by student group, reported by each state from its own tests, which means Minnesota's rows come from Minnesota's own statewide reading assessment.

Join them for Minnesota. The most recent school year for which I could retrieve both halves is 2017–18 (more on that below, because it matters). Restricting to systems where the federal file reports a point estimate rather than a privacy band, with at least 200 valid reading scores and complete finance data, gives 273 of the 330 Minnesota school systems reporting a fall enrollment in the F-33 file — 416,148 tested students and 807,682 fall-enrolled pupils. Median total revenue per pupil: $14,093. Median reading proficiency: 61 percent.

Sort those districts into quartiles by revenue per pupil:

Monotone, in the direction that would make a certain kind of pundit very happy. The correlation between revenue per pupil and reading proficiency across those 273 districts is −0.449. Current spending per pupil: −0.492. State revenue per pupil: −0.671.

Anybody who stops there and publishes "Minnesota's most-funded districts read worst" has written something numerically accurate and substantively worthless. Here is why.

Why that number is an artifact and not a finding

Look at the last column of that table. It climbs as revenue climbs, because it is the thing the formula is reading when it decides how much to send. Run the correlations against it — same federal file, same 273 districts:

That last row is the whole case in one line. Local revenue per pupil — the tax base plus local referendum decisions — correlates positively with reading. State revenue per pupil correlates strongly negatively. Neither is a causal claim about anything. They are two readings of the same underlying fact about where the children with the hardest starts live: local dollars track property wealth, state dollars track measured need, and reading scores track need. Any two of those three will produce a correlation, and none of them tells you whether a dollar taught a child to read.

Economic disadvantage alone explains 62.2 percent of the district-level variation in reading proficiency. Add total revenue per pupil to that model and the explained share moves from 0.622 to 0.633. Widen the need measures to include English learner share, homeless share, and children-with-disabilities share — all from the same federal file — and they explain 64.5 percent on their own. Add revenue per pupil and you get 64.9 percent: four-tenths of one percentage point of additional explanatory power, carrying a coefficient of −0.40 reading points per additional $1,000 per pupil. Once poverty is held constant, the partial correlation between revenue and reading falls from −0.449 to −0.168, and the partial correlation between local revenue and reading falls to +0.005 — dead zero.

Split by need and the artifact shows its face. In the lowest-poverty third of these districts the correlation between revenue per pupil and reading is −0.082; in the middle third, −0.097; in the highest-poverty third, −0.541. The apparent relationship between money and reading exists only where the need-driven money is concentrated. It is the targeting, measured.

One more test, because a correlation coefficient is easy to wave away. Take every pair of districts in the set whose economically disadvantaged shares are within two percentage points of each other, where both tested at least 400 students, and whose revenue per pupil differs by at least $3,000. There are 141 such pairs. Median revenue gap: $3,897 per pupil. Median absolute difference in reading proficiency: 5.0 percentage points — against a standard deviation of 9.0 points across the full set, and an interquartile range of 55 to 67 percent. Narrow to the 19 pairs with a revenue gap of $6,000 per pupil or more and the median reading gap falls to 3.0 points. Bigger money differences among similarly situated districts did not produce bigger outcome differences.

The four closest-matched districts, on eight-year-old figures, for concreteness — and no criticism of any of them is intended or supported here, which is exactly the point:

A $4,100 gap and a $7,700 gap, both invisible in the outcome. That is not evidence that money does not matter. It is evidence that this comparison cannot see whether it does — the districts differ in a hundred things the public files do not record, from student mobility to teacher experience to the cost of delivering a service in Carlton County rather than Hennepin, and a proficiency rate mostly measures who enrolled.

I ran the whole thing again on a different year as a check. School year 2020–21 against fiscal 2021 finance, 264 districts, 319,474 tested students: revenue −0.448, current spending −0.452, state revenue −0.609, local revenue +0.241, economic disadvantage −0.756. Same signs, same magnitudes, same artifact. I am not using 2020–21 as the headline year because across the 507 Minnesota districts reporting in both years, total valid reading scores fell from 452,952 to 351,485 — 78 percent of the earlier count — and a proficiency rate computed on a fifth fewer test-takers is not comparable to one that is not.

What cuts against my own framing

The people who built this formula were solving a real problem, and they largely solved it. Minnesota equalizes across school districts whose property-tax bases differ enormously, and the equalization is not rhetoric — it is in the statute with dollar figures attached. Minnesota's statewide current spending was $17,098 per pupil in fiscal 2024 against a national $17,619, per the Census Bureau's own Table 8 — which is to say this is a state spending a little under the national average and distributing it deliberately. Section 126C.10, subd. 2e equalizes the first tier of local optional levy against a referendum market value of $880,000 per resident pupil unit and the second tier, for fiscal 2027 and later, against $671,345. Section 126C.17, subd. 6 equalizes the referendum levy at $567,000 and $290,000 per resident pupil unit. A district with a thin tax base levies less and the state makes up the difference.

And it shows in the money. In the Census F-33 for fiscal 2024, across the 245 Minnesota school systems with at least 500 pupils and complete finance data (768,451 pupils):

Local capacity varies by a factor of two and a half. What the state sends varies by 1.28. That is a formula doing its job, and no correlation coefficient in this article erases it.

It also does not finish the job. The state contribution is nearly flat rather than compensating: the correlation between local revenue per pupil and state revenue per pupil across those 245 systems is −0.087, essentially zero. Sort them into quartiles by local revenue per pupil and median total revenue per pupil runs $17,448, $18,434, $19,288, $21,439 — a $3,991 per-pupil spread between the districts that can raise the least locally and the districts that can raise the most. Some of that is voter choice rather than wealth, and some of it is debt service and capital levies rather than classroom money, and both of those caveats are real.

Why I could not do better, and exactly what would fix it

Minnesota law already requires the report that would begin to answer this. Minn. Stat. § 123B.76, subd. 3(c): "The department must annually report information showing school district general fund expenditures per pupil by program category for each building and estimated school district general fund revenue generated by pupils attending each building on its website." Per building. Expenditures and revenue generated. And the Minnesota Report Card's own report manifest confirms the spending half is published — its "Fiscal Transparency" report states that it shows per-student spending "at the selected level (school/district/state)", excluding capital, construction, debt service, food service, and community service, divided by average daily membership. Reading proficiency is published in a different report in the same application, and a growth model — "Minnesota Growth", which compares a student's score change against students who started in the same place — is published in a third.

Note the denominator in that description, because the denominator is not pedantry. The Report Card divides by average daily membership and strips the capital and non-instructional funds out first. The Census F-33 file I used above divides by a fall headcount, and its revenue measure includes capital. Two per-pupil figures for the same Minnesota district can differ by thousands of dollars and both be correct, which is why anyone who quotes one without saying which it is has told you less than they think.

So the ingredients exist. Four things keep them from answering the question, and all four are fixable by publication rather than legislation.

1. The revenue side of the building report is partly constructed by division. Section 123B.76, subd. 3(c) tells the department how to build the per-building revenue estimate, and three of its seven clauses are uniform allocations: expenditures not reported by building "shall be allocated among buildings on a uniform per pupil basis" (clause 1), other general education revenue "on a uniform per pupil unit basis" (clause 5), and other general fund revenues "on a uniform per pupil basis, except that the department may allocate other revenues attributable to specific buildings directly to those buildings" (clause 7). A number that is partly measured and partly divided will produce exactly the artifact this article is about, because the only parts that vary are the need-targeted parts. A user of that report cannot tell measured dollars from divided dollars, and that distinction is the whole ballgame.

2. Revenue generated is not revenue spent. Section 126C.15, subd. 2(a) requires a district to allocate at least 80 percent of compensatory revenue to the buildings where the children who generated it are served. Paragraph (f) — as amended this year by 2026 Minn. Laws ch. 117, § 5, which extended the window to fiscal 2028 — lets a district allocate up to 40 percent by a plan the school board adopts. Paragraph (e) lets districts reallocate among sites for enrollment and boundary changes and requires them to report those adjustments to the department, which "must use the adjusted compensatory revenue allocations" in the § 123B.76 report. So the statutory phrase "revenue generated by pupils attending each building" is not the same quantity as the money spent on those pupils — a district may route up to two-fifths of the dollars to sites other than the ones that generated them — and the department is the party that receives both figures.

3. The Legislature just created a natural experiment and nobody outside the department will be able to use it. This is the part I did not expect to find. 2026 Minn. Laws ch. 117, § 4 — signed May 27, 2026, at 12:23 p.m. — amends § 126C.10, subd. 3 to add paragraphs (h) and (i). For fiscal year 2027 only, a building's compensatory revenue equals the greater of the amount the new building-allowance formula produces or a "building minimum amount", and the minimum is the building's fiscal 2026 compensatory revenue, times the lesser of one or the ratio of its October 1, 2025 enrollment to its October 1, 2024 enrollment, times 0.659771.

Read what that floor implies. The Legislature wrote a hold-harmless guaranteeing each building at most roughly 66 percent of last year's compensatory dollars — less than that wherever enrollment fell, because of the ratio term. A floor set there does no work unless the formula the Legislature had just adopted would have cut some buildings by more than a third in a single year. It also means that in fiscal 2027, two buildings with the same children and the same needs can receive materially different compensatory dollars for a reason that has nothing to do with those children — one landed on the floor and one did not. That is the closest thing to a controlled comparison Minnesota school finance has produced in years, and it expires in twelve months.

4. The files a researcher can actually reach do not cover recent years. The EDFacts district-level reading files I could retrieve cover school years 2014–15 through 2017–18, plus the low-participation 2020–21 year; requests for 2018–19 through 2024–25 at the same path returned nothing while the earlier years returned full files, so the requests were working. On the state side, MDE's Data Center and the Report Card's data directory returned bot-detection interstitials to every automated request in this research pass, which is why not one number in this article comes from them. Nobody is hiding data — this is ordinary infrastructure defending itself against scrapers. But whatever blocks an automated request from a Minneapolis law office blocks the same request from a legislative analyst, a reporter, and a graduate student.

What we can do

Every fix below is a publication, not a program. No new agency, no new appropriation.

1. Publish the two fiscal 2027 compensatory columns before the window closes. The Department of Education has to compute both the paragraph (c) formula amount and the paragraph (i) floor amount for every building in the state in order to administer paragraph (h). Publishing them is two columns in a file it is already building. Not publishing them throws away the only clean variation in compensatory funding Minnesota has produced this decade, and it is gone after one fiscal year.

2. Make the § 123B.76, subd. 3(c) report self-describing. Every component flagged as reported-by-building or uniformly allocated, every denominator named on its own column, generated and allocated compensatory revenue as two separate columns per building. The statute already requires the report. This is a schema change.

3. Pair the dollars with growth, not just proficiency. A proficiency rate is largely a measure of who enrolled. Minnesota already runs a growth model that compares a student's score change against students who started in the same place. Put growth in the same file as the money and the artifact in this article shrinks a great deal, because growth is far less determined by the composition of a building than proficiency is.

4. One machine-readable file per year, keyed on the education site code, carrying the fiscal, achievement and growth figures for the same building side by side — and reachable by an ordinary automated request, because the people who most need to join these files are not clicking through a web form 2,000 times.

Until that exists, the defensible statement is narrow, and it belongs in every budget hearing where somebody waves a scatterplot: Minnesota's school funding formula sends more money where measured need is greater, which is what it was built to do, and for that reason the publicly available files cannot separate the effect of the money from the effect of the need. That is not a reason to stop asking. It is a specification for the file that would let us answer.

I want to emphasize the part that is nobody's fault. Not a teacher's, not a district's, not a parent's. The formula is doing what the Legislature told it to do, the department is publishing what the statute told it to publish, and the result is a public record that produces a negative correlation between money and reading which a careless reader will mistake for a verdict. Fixing that takes two columns and a schema. It is the cheapest useful thing anyone in state government could do about reading this year.

First the facts. Then the fix.

Sources

Statutes, all fetched as raw text from revisor.mn.gov on September 11, 2026, with the 2026-session amendment banner checked on each section. Minn. Stat. § 126C.10, subd. 1 (the fourteen components of general education revenue, quoted verbatim); subd. 2(a)–(d) (formula allowance of $7,281 for fiscal year 2025; the fiscal-2026-and-later index of the greater of 1.02 or one plus CPI change capped at 1.03; the commissioner's duty to publish the allowance by the end of February); subd. 2e(a), (c), (d) (local optional revenue of $724 per adjusted pupil unit, in a first tier of $300 and a second tier of $424, each times adjusted pupil units; first-tier levy equalized against referendum market value per resident pupil unit of $880,000; second-tier equalizing factor of $671,345 for fiscal year 2027 and later); subd. 3(b), (c), (f), (g) (the fiscal 2024–2026 "formula allowance minus $839" calculation; the fiscal-2027 building compensatory allowance calculation; the statewide floors of $838,947,000 for fiscal 2026 and $857,152,000 for fiscal 2027 and later); subd. 3a(b), (c) (building concentration factor; compensatory pupils as free-meal-eligible plus one-half of reduced-price-eligible on October 1 of the previous year); subd. 3c (statewide compensatory allowance of $6,936 for fiscal year 2027); subd. 4 (basic skills revenue as compensatory revenue plus English learner revenue under § 124D.65, subds. 5 and 5a). Minn. Stat. § 126C.15, subds. 1 and 2 (permitted uses; the 80 percent building-allocation requirement and the 20 percent board-plan allowance; the "building means education site" definition; the subd. 2(e) reallocation reporting duty). Minn. Stat. § 126C.17, subd. 6(b), (c) (referendum equalization levy equalizing factors of $567,000 and $290,000 per resident pupil unit). Minn. Stat. § 123B.76, subd. 3 (the per-building accounting duty; subd. 3(c) quoted verbatim, and clauses (1), (5) and (7) on uniform per-pupil allocation) — no 2026-session amendment banner appears on this section.

Session law. 2026 Minn. Laws ch. 117 (H.F. No. 2433), read in full as enacted text at revisor.mn.gov rather than from the Revisor's posted statutory text, because the Revisor flags § 126C.10, subd. 3 and § 126C.15, subd. 2 as amended by this chapter. Section 4 adds § 126C.10, subd. 3, paragraphs (h) and (i): the fiscal-2027-only greater-of rule and the building minimum amount equal to fiscal 2026 building compensatory revenue times the lesser of one or the ratio of October 1, 2025 to October 1, 2024 enrollment times 0.659771; it also corrects the paragraph (f) cross-reference from (c) to (b). Section 5 amends § 126C.15, subd. 2(f) to extend the 40 percent board-plan allocation allowance to fiscal year 2028. Section 6 amends the general education aid appropriation in Laws 2025, First Special Session ch. 10, art. 1, § 28, subd. 2, from $8,509,608,000 to $8,550,641,000 for fiscal 2026 and from $8,765,730,000 to $8,783,520,000 for fiscal 2027. The chapter was presented to the governor May 20, 2026 and signed May 27, 2026 at 12:23 p.m.

Finance data. U.S. Census Bureau, Annual Survey of School System Finances (F-33), individual unit files elsec18t.txt (fiscal 2018), elsec21t.txt (fiscal 2021), elsec23t.txt (fiscal 2023) and elsec24t.txt (fiscal 2024), plus the fiscal 2023 and fiscal 2024 summary-table workbooks elsec23_sumtables.xlsx and elsec24_sumtables.xlsx, whose Table 8 ("Per Pupil Amounts for Current Spending of Public Elementary-Secondary School Systems by State") supplies the published statewide figures of $16,116.71 for Minnesota and $16,525.88 for the United States in fiscal 2023 and $17,097.98 for Minnesota and $17,619.39 for the United States in fiscal 2024 — all downloaded from www2.census.gov on September 11, 2026, with the accompanying school24doc.docx file documentation for field definitions (ENROLL is a fall headcount; STRFORM is general formula assistance; TCURSPND is total current spending). Minnesota units identified by FIPST 27 in the fiscal 2023 and 2024 files and by an NCESID beginning "27" in the fiscal 2018 and 2021 files, which carry no state column. All per-pupil figures in this article are computed by the author as the unit's reported dollars divided by the same unit's reported ENROLL.

Outcome data. U.S. Department of Education, EDFacts LEA-level reading/language arts achievement files rla-achievement-lea-sy2017-18.csv and rla-achievement-lea-sy2020-21.csv, downloaded from www2.ed.gov on September 11, 2026. These files carry each state's own assessment results as reported to the Department by the state education agency; Minnesota's rows are Minnesota's statewide reading assessment results as reported by the Minnesota Department of Education. Proficiency values are published as point estimates for larger groups and as suppressed ranges for smaller ones; the analysis set is restricted to districts reporting a point estimate for all students with at least 200 valid scores, which is why the fiscal 2018 set is 273 of 330 Minnesota systems. Subgroup shares (economically disadvantaged, English learner, homeless, children with disabilities) are computed as that subgroup's valid-score count divided by the all-students valid-score count in the same file and year.

Minnesota Report Card. The application's own report manifest at rc.education.mn.gov/data/reportList.json, retrieved September 11, 2026, listing 24 reports including "Fiscal Transparency" ("How are school funds spent?"), "North Star Academic Achievement", "Test Results and Participation", and "Minnesota Growth". The quoted description of the Fiscal Transparency report's level of aggregation, exclusions, and average-daily-membership denominator is taken verbatim from that file.

All correlations, regressions, quartile tables, partial correlations and matched-pair counts in this article were computed by the author from the files named above. The regression models are ordinary least squares on district-level observations, unweighted except where a tested-student-weighted figure is identified as such.

What I could not verify, and am not asserting. The Minnesota Department of Education's Data Center at pub.education.mn.gov and its document server at education.mn.gov returned bot-detection interstitials to every automated request during this research pass, and the Report Card's own data directory returned an "unusual activity" block; no figure in this article comes from any of them, and nothing here should be read as a claim about what those systems do or do not contain. I obtained no per-building compensatory revenue figure for any Minnesota building in any year, and make no building-level claim. I did not retrieve the fiscal 2026 or fiscal 2027 formula allowance amounts the commissioner publishes under § 126C.10, subd. 2(d), and do not state them. EDFacts district-level reading files for school years 2018–19 through 2024–25 were not available at the path that served 2014–15 through 2017–18 and 2020–21, so the most recent joinable year in this article is 2017–18 and the relationships described may have changed since; ED Data Express returned a JavaScript challenge to automated requests and was not used. Summing the Census individual-unit file by hand does not reproduce the Bureau's own published statewide per-pupil figure: my summation of the fiscal 2023 unit file yields $16,891 in current spending per pupil for Minnesota against the $16,117 the Bureau publishes in Table 8 of the same release. Every district-level per-pupil number in this article is therefore computed on the unit file's own enrollment denominator and should not be added up to a state total; the only statewide figures stated are the Bureau's own published ones. Revenue figures include capital and debt-service revenue, which is driven by local construction decisions rather than the general education formula, and current spending excludes it; both measures are reported separately above for that reason. Economically disadvantaged share of tested students is a crude proxy for need and does not capture student mobility, teacher experience, program mix, or regional cost differences, and nothing in this article identifies a causal effect of money on reading in either direction. Minnesota's 2026 assessment results had not been released as of publication. Where the federal file reports no count for a subgroup — 60 districts for English learners, 47 for homeless students — I treated that subgroup's share as zero rather than dropping the district, so that all four need measures run on the same 273 districts. On complete cases only (190 districts) the four need measures explain 74.3 percent, revenue still adds about a tenth of a point, and the revenue coefficient is positive rather than negative. The sign of a coefficient that small is not a finding either way; that it is that small is. This piece runs well past this site's 1,100-to-1,400-word norm, at roughly 3,340 words, because the statutory reading and the two joined datasets each produced verified material I was not willing to drop. This is data journalism about public files, not legal advice and not an audit. The statutory passages are quoted to show what the formula instructs, no reader's situation and no client of mine is discussed, and reading it creates no attorney-client relationship. Corrections: campaign@madgettformn.com.