Performance Index

Its arrival changes what the corpus can be asked. adequacy records an open contest over “whether outcome measures or input measures should define the requirement.” Until now the corpus could only take the input side of that question, because it held no outcome variable at all. The Performance Index is the outcome side, and it makes the outcome formulation of adequacy testable rather than merely contested. inference

It does not make it answered. A score is not a requirement. Adequacy in the outcome formulation needs a threshold — a level of measured achievement the state declares sufficient — and Ohio publishes an index, not a floor. The Index tells you where a district sits; it does not tell you where it was supposed to be. inference

Use the Index, not the star rating. The overall rating groups many districts into the same category and cannot distinguish among them; the Index is continuous and can be ranked, correlated, and quintiled. Any Ohio claim about relative district performance that quotes stars is answering a coarser question than it appears to. verified

What this repository computed # ↑ Contents

Not what Ohio publishes. Figures derived here from committed fixtures, each one citing the test that reproduces it.

The dominant fact about this measure is poverty # ↑ Contents

Across the 606 districts where both are held, the Performance Index correlates with the economically disadvantaged share at −0.846, which is 71.6% of the cross-district variance in the Index. verified computed; see crates/xcheck/tests/report_card_2425.rs

Nothing else in this corpus comes close to that. It is the first thing to say about any district-level correlate of the Index, because almost every such correlate is partly this variable wearing a different label — including OCG White Paper 013’s strongest reported finding, that federal expenditure per pupil tracks the Index at −0.558. Holding disadvantage constant, that −0.558 falls to −0.158. verified The paper hypothesized that its federal result was a poverty signal and said it could not test the hypothesis directly. It can be tested directly, it now has been, and the hypothesis is largely right: about seven-tenths of the association was poverty.

The measure of poverty matters as much as the fact of it # ↑ Contents

Ohio publishes two economic-disadvantage shares and they are not the same variable. The report card’s own 2024-25 share is top-coded by community eligibility — where a district qualifies for universal free meals every student is counted disadvantaged — so 87 districts sit at exactly 100% against 37 on the Cupp Report’s FY2024 measure. They correlate with each other at only +0.823, and against this Index they give −0.734 and −0.846. verified The censored measure gives the weaker association, as censoring predicts. The −0.846 above is the Cupp measure and is the better estimate; both are committed, and neither substitutes for the other. See Ohio School Report Cards — District Details.

The measured relationships, all computed rather than quoted verified:

economically disadvantaged share, Cupp FY24     -0.846    rho -0.847
economically disadvantaged share, report card    -0.734    (top-coded; see above)
ADM weight ratio, weighted/unweighted            -0.745    rho -0.734
state revenue per pupil (FY24)                   -0.549    rho -0.521
federal expenditure per equivalent pupil (FY25)  -0.558    rho -0.694
operating expenditure per UNWEIGHTED pupil FY25  -0.337    rho -0.334
operating expenditure per pupil, Cupp FY24       -0.388    rho -0.350
assessed valuation per pupil (FY23)              +0.321    rho +0.372
total expenditure per EQUIVALENT pupil (FY25)    -0.016    rho +0.048
Progress (value-added) effect size 2024-25       +0.375

The equivalent-pupil row is the one that gets quoted, and what separates it from the unweighted row above it is a denominator rather than a fact about Ohio. See expenditure per equivalent pupil. The Progress row is the other outcome measure, not a correlate of spending — see Progress (Value-Added) for what it does to every sign here.

A distributional asymmetry to state before any of those coefficients. Performance has a coefficient of variation of about 0.124; per-pupil spending, on either denominator, about 0.202. Ohio districts differ from one another about 60% more in what they spend than in what they score. verified A near-zero coefficient between a wide distribution and a narrow one is a weaker statement than it sounds.

It barely moves, and that cuts both ways # ↑ Contents

Three school years are held. Adjacent years correlate at +0.988 and +0.989, 2022-23 against 2024-25 at +0.981, and the median district’s three-year range is 2.1 points against a cross-district standard deviation of 10.9. verified So a single-year cross-section is a fair stand-in for a district’s Index, and the “single year, single measure” limitation is weaker than it looks. But a measure this static cannot detect anything that changes annually — including what a district spends. Over this window the Index is closer to a fixed district attribute than to an annual result. inference

Lagging the spending does not help, and the direction is the point # ↑ Contents

FY2024 per-pupil spending covers the 2023-24 school year. Against the Index for the year before it, the same year, and the year after: verified

spending one year AFTER the outcome    -0.426
spending contemporaneous               -0.412
spending one year BEFORE the outcome   -0.388

The backwards lag is the strongest of the three. A causal reading needs the opposite ordering, and this is the remedy OCG White Paper 013 proposed for itself — “test lagged and cumulative spending measures rather than same-year expenditure.” It does not rescue the analysis; it closes it off. Nothing propagates forward here. inference

A second outcome measure sits beside this one, disagreeing # ↑ Contents

Against Progress (value-added), the same spending variable carries the opposite sign. In a need-adjusted model controlling for economic disadvantage, English-learner and disability shares, log enrollment, and valuation per pupil, the standardized spending coefficient is −0.073 here (t = −2.84) and +0.209 on growth (t = +4.60). verified The negative association recorded on this metric is a property of measuring level. It does not survive the switch to a measure of gain, and neither metric should be quoted on outcomes without the other.

That model also explains most of this metric and little of the other: R² 0.752 against 0.227, with economic disadvantage carrying a standardized −0.659 here. Three-quarters of what this measure records is demographics. verified

The average student’s district scores below the average district. Unweighted mean 87.7; weighted by enrollment, 85.1. verified Ohio’s larger districts score lower, which is why the district-weighted and student-weighted views of every relationship here diverge, and why neither can be quoted alone.

    not tested    0.0        accomplished  1.1
    limited       0.3        advanced      1.2
    basic         0.6        advanced plus 1.3
    proficient    1.0

  minimum   53.1
  p25       81.5
  median    88.2
  mean      87.7
  p75       95.1
  maximum  112.8
  sd        10.9      coefficient of variation ~0.124

  92.5   89.6   88.4   86.7   81.4      unweighted district means
  92.6   88.4   88.6   87.4   72.2      enrollment-weighted