OCG White Paper No. 013 — Does Per-Pupil Spending Track Academic Performance?

Source. Ohio Common Ground Research Center, White Paper No. 013, “Does Per-Pupil Spending Track Academic Performance? An Unadjusted, Single-Year Cross-Sectional Analysis of Expenditure per Equivalent Pupil and Performance Index Across Ohio’s Traditional Public-School Districts, 2024–2025.” Tracking ID RL-2026-020, Version 1.0. Dublin, Ohio, August 7, 2026. Type. Secondary source — analysis over primary data, with a published replication package. Location. ohiocommonground.com.

What it contains. A cross-section of 607 rated traditional districts joining the report-card Performance Index to report-card expenditure per equivalent pupil by IRN. Enrollment comes from the FY2025 Expenditure Expanded List. Reports Pearson and Spearman coefficients, Fisher-z intervals, spending and performance quintiles both unweighted and enrollment-weighted, and a seven-row sensitivity table.

Headline figures, as published:

Quantity	Value
Total spending per equivalent pupil vs Performance Index	Pearson −0.016 (95% CI −0.095 to +0.064; p = 0.703); Spearman +0.048
Same, weighted by unweighted ADM	Pearson −0.149
Federal expenditure per equivalent pupil vs Performance Index	Pearson −0.558; Spearman −0.694; partial (holding state-and-local) −0.556
State-and-local expenditure per equivalent pupil vs Performance Index	Pearson +0.086 (p = 0.034)
Enrollment vs spending per equivalent pupil	Spearman −0.366
Spending distribution, 607 districts	median $12,856, mean $13,224, SD $2,660
Performance Index distribution	min 53.1, median 88.2, mean 87.7, max 112.8, SD 10.9

Replicated against the primary sources — every figure reproduces. The corpus now retrieves the same three DEW files and recomputes the paper’s results from them. To the precision the paper displays, all of it lands: −0.016, Spearman +0.048, federal −0.558, state-and-local +0.086, enrollment-weighted −0.149, enrollment against spending ρ −0.366, median $12,856, mean $13,224, and the Performance Index at median 88.2, mean 87.7, sd 10.9, range 53.1 to 112.8. Pinned in crates/xcheck/tests/report_card_2425.rs. verified

This is an unusually clean replication and it should be said plainly: the paper’s arithmetic is correct, its sources are cited accurately enough to find and reproduce, its sensitivity table holds, and its stated caveats are honest. The disagreement below is about measurement, not competence.

Independent corroboration of a corpus figure. Its spending dispersion, CV = $2,660/$13,224 = 0.201, matches the coefficient of variation of 0.202 that crates/dispersion computes over the FY2024 Cupp Report — different year, different source file, different denominator, same dispersion. This is the first external check on that statistic. verified

Where this corpus departs from it, now measured. The paper’s central claim is that the association is near zero. It is near zero on the denominator the paper chose. The department’s FY2025 Expanded List publishes one operating-expenditure total against both a weighted and an unweighted pupil count, and correlating the Performance Index against the same dollars gives −0.015 on the first and −0.337 on the second. verified Same districts, same year, no covariates — only the divisor changes.

“Unadjusted” is therefore true of the paper’s model and false of its measure. The stated limitation, “no need-adjusted model in this version,” understates it: there is a need adjustment, inside the independent variable, on one side only, pointing the way that flattens the result. See Expenditure Per Equivalent Pupil.

Two of its own conjectures, tested. The paper hypothesized that its federal-spending result was a poverty signal and said it could not test that directly. Holding the economically disadvantaged share constant, federal per-pupil falls from −0.558 to −0.158 — the hypothesis was largely right. And the variable it was standing in for is far stronger than anything the paper reports: disadvantage against the Performance Index is −0.846, 71.6% of the variance. verified

The measure it names in Future Research answers the question differently. The paper lists “add other outcome measures, including the Progress (value-added) component.” Done: against Progress, holding disadvantage constant, per-pupil spending correlates +0.146, where against the Performance Index it correlates −0.125. Its own proposed next step reverses the sign of its subject. Its other proposed step — lagged rather than same-year spending — does not help either: FY2024 spending tracks the prior year’s Index (−0.426) slightly better than the following year’s (−0.388), so there is no forward-directed signal to lag into. verified

Its priority next step has been taken, and it reverses the paper’s subject. The paper names “a need-adjusted, multivariable descriptive model relating Performance Index to per-pupil spending while accounting for economically disadvantaged, English-learner, and disability shares, district enrollment, typology, and local wealth” as its first Future Research item. That model now exists in this corpus, missing only typology. The standardized spending coefficient is −0.073 on the Performance Index and +0.209 on the Progress component. verified Its own proposed test returns opposite signs depending on which of Ohio’s two published outcome measures is used, and the paper analyzed only one of them.

One correction of description. The paper reads state-and-local spending across performance quintiles as “moving in the opposite direction, from $12,403 to $13,324.” That row is flat within $200 across four quintiles and steps $984 at the top. It is a step at the wealth end, not a gradient.

Caveats on using it:

• Secondary. Every figure above is as-published and reaches DEW only through this paper.
• Single year, single outcome measure, no covariates. The paper says so repeatedly and does not overclaim; it draws no causal conclusion and makes no recommendation.
• Its district count (607) is not the corpus’s (606 for FY2024 Cupp, 609 for the FY2027 model).
• Its enrollment total (~1.47M, FY2025) exceeds the corpus’s FY2024 enrolled ADM total (1,439,473) while Ohio enrollment is falling, so the two ADM definitions are not the same quantity. open