School funding outcomes research: does money matter?

Whether money improves schools sounds like a question a spreadsheet could settle in an afternoon. It took American social science fifty years, several court systems and a new research design. The story is worth telling properly. The answer now shapes billion-dollar budget fights from state legislatures to national ministries. And the way the question got stuck is one of the clearest lessons in education research about what correlations can and cannot say.

This article traces the arc: the 1966 report that framed the doubt, the skeptic’s ledger that kept it alive, and the design flaw both sides shared. Then the quasi-experimental studies since 2016 that finally produced an answer most of the field accepts. Then the boundaries, because the new answer is routinely overclaimed in both directions.

The Coleman Report’s long shadow

In 1966, the U.S. government published one of the largest social surveys ever attempted: 600,000-plus students, thousands of schools, commissioned to document educational inequality (Coleman et al., 1966). Its most famous reading was a shock. Measured school resources — spending, facilities, libraries — explained little of the variation in student achievement once family background was accounted for. Families and peers appeared to dominate; the school’s inputs looked almost incidental.

That reading was always cruder than the report itself. But it hardened into a proverb — “schools don’t matter, families do” — and the proverb did policy work for decades. If money did not move outcomes, spending more of it on poor districts was charity, not strategy. Every subsequent funding fight, in courtrooms and legislatures, was conducted in this shadow.

What the Coleman framing could not do — through no fault of its era — was distinguish correlation from cause. It compared schools as they happened to be, with all the sorting of families, teachers and money that produced them. Holding that thought, the next three decades of research mostly repeated the same design at smaller scale.

The skeptic’s tally

The skeptical position found its most rigorous voice in economist Eric Hanushek. Across a series of reviews, he tallied the published estimates linking school resources to student performance. The 1997 update is the canonical version (Hanushek, 1997). Counting 163 estimates of the effect of per-pupil spending, he found only about 27% positive and statistically significant. Roughly 7% were negative and significant, and the remaining two-thirds were statistically indistinguishable from nothing.

His conclusion was carefully worded and endlessly quoted: there was no strong or consistent relationship between school resources and student performance. Note what the claim was. Not “money never matters” — Hanushek himself argued that how money is spent must matter — but that simply adding resources to the existing system showed no reliable payoff (Hanushek, 1997).

For twenty years this tally was the strongest card in the fiscal-restraint hand. It deserved respect: it summarized the literature that existed. The problem, it turned out, was the literature.

Why the correlations couldn’t answer

Think about how a district ends up spending more than its neighbour. Sometimes because wealthy parents demand it. Sometimes because a state sends compensatory money precisely where students struggle — poverty aid, special-education funding, remediation grants. The first pattern makes spending look artificially good; the second makes it look artificially useless, because the extra dollars flow toward the hardest problems.

Correlational studies inherit both distortions at once, in unknown proportions. That is why the old literature could stay ambiguous forever: it was not measuring the effect of money, but the net of two opposing selection stories (Jackson, 2018). No amount of statistical control fully breaks the loop, because the things that drive both spending and achievement — parental wealth, student need, local politics — are never fully observed.

What breaks the loop is money that moves for reasons unconnected to the students. American school finance, unusually, produces exactly that: court orders striking down funding formulas, and legislated reforms that redirect dollars across districts on schedules set by litigation rather than by need or wealth. From the 1970s onward, dozens of states went through such events. Each one is a natural experiment with a before, an after, and districts differently exposed. The research design was lying in the statute books, waiting.

The quasi-experimental turn

The paper that turned the field arrived in 2016, in the Quarterly Journal of Economics. Kirabo Jackson, Rucker Johnson and Claudia Persico linked the timing of court-ordered finance reforms to the life outcomes of thousands of Americans followed from childhood (Jackson, Johnson & Persico, 2016). The question, in effect: what happened to children who spent more of their school years under reform-driven spending increases?

The identification is the point. Whether a child was ten or seventeen when their state’s court struck down its funding formula has nothing to do with their family, ability or neighbourhood. It is timing — and timing determined how many school years the extra dollars covered. Comparing children on either side of these cutoffs isolates the money itself.

The results ended the stalemate. A 10% increase in per-pupil spending, sustained across all twelve school years, raised completed education by roughly 0.31 years (Jackson, Johnson & Persico, 2016). Adult wages rose by about 7%. The annual incidence of adult poverty fell by roughly three percentage points. For children from low-income families the effects were substantially larger — close to half a year of additional attainment, with the poverty reduction roughly doubling. And the mechanisms were visible: reform dollars flowed into smaller classes, higher teacher salaries and longer school years.

Two features made the study hard to dismiss. It measured life outcomes — attainment, earnings, poverty — rather than only test scores. And its effects sat exactly where the compensatory-funding critique predicted they had been hiding: among the poorest students, whose districts’ extra dollars had always looked useless in the correlations.

The replication wave

One landmark study reverses a presumption; it takes a wave to settle a question. The wave came fast, from independent teams using different reforms, different data and different outcomes.

Julien Lafortune, Jesse Rothstein and Diane Whitmore Schanzenbach studied the post-1990 “adequacy” reforms — finance overhauls aimed at guaranteeing minimum funding in poor districts (Lafortune, Rothstein & Schanzenbach, 2018). Using event-study methods — tracing outcomes year by year around each reform — they found low-income districts’ achievement flat before reform, then climbing steadily after, accumulating roughly a tenth of a standard deviation within a decade. Gradual, unglamorous, and exactly the trajectory a real resource effect should trace.

Jonah Hyman followed Michigan children into adulthood: a roughly 10% spending increase in the primary and middle grades raised college enrolment by about three percentage points and degree completion by around two (Hyman, 2017). Christopher Candelaria and Kenneth Shores examined court-ordered reforms nationally and found graduation rates in the highest-poverty districts rising on the order of seven to twelve percentage points about seven years after reform (Candelaria & Shores, 2019).

Different decades, different states, different outcome ledgers — attainment, test scores, graduation, earnings — and the same sign. By the time Jackson reviewed the modern literature, the summary was blunt: policy-induced increases in school spending improve student outcomes, and the studies finding otherwise are the rare exceptions (Jackson, 2018).

What the pooled evidence now says

The settling document is a 2024 meta-analysis. Jackson and Claire Mackevicius pooled the modern quasi-experimental studies of policy-driven spending changes in the United States (Jackson & Mackevicius, 2024). Their central estimate: an extra $1,000 per pupil, sustained for four years, raises test scores by roughly 0.03 standard deviations and college-going by close to two percentage points. Over 90% of the pooled estimates are positive, and the distribution of results looks like a real effect measured with noise — not a contested literature split down the middle.

Read those magnitudes honestly in both directions. They are not miracles: 0.03 standard deviations is a small nudge on tests, which is why spending advocates who promise transformation are overclaiming. They are also not nothing: sustained across a schooling career, and carried into attainment, earnings and poverty, the same dollars clear standard cost-benefit bars in most of the pooled scenarios (Jackson & Mackevicius, 2024). Money is a slow, broad lever, priced like one.

The reconciliation with the skeptics matters too. Hanushek’s core intuition — that how money is spent conditions what it buys — survives intact inside the new consensus (Hanushek, 1997). Effects vary across places and uses; spending on instruction and people shows up more reliably than spending on some capital projects. What did not survive is the null: the claim that adding resources to schools has no reliable payoff is now the position at odds with the evidence (Jackson, 2018).

What the evidence doesn’t show

The new consensus has firm edges. Six of them.

• No blank cheque. The estimates come from specific policy instruments — court orders, formula reforms — in specific decades. They do not certify that any future increase, spent any way, will replicate them (Jackson & Mackevicius, 2024).
• Effects are averages over heterogeneity. Gains concentrate in low-income districts and in spending on instruction; some reforms moved outcomes little (Candelaria & Shores, 2019).
• Test-score effects are modest. The headline life-outcome gains dwarf the near-term score movements; leaders promising rapid test gains from budgets alone are outside the evidence (Jackson & Mackevicius, 2024).
• The evidence is overwhelmingly American. The quasi-experimental base rides on U.S. finance litigation; transfer to other systems — different baselines, different labour markets — is an assumption, not a finding.
• Cuts and increases may not be symmetric. Much of the literature studies increases; recession-era cuts show harm, but magnitudes need not mirror each other (Jackson, 2018).
• How-to-spend is the thinner literature. The causal ranking of uses — salaries vs class size vs time vs technology — remains far less settled than the does-it-matter question (Hanushek, 1997).

Where the evidence stops

1. 1No blank cheque
2. 2Effects are averages over heterogeneity
3. 3Test-score effects are modest
4. 4The evidence is overwhelmingly American
5. 5Cuts and increases may not be symmetric
6. 6How-to-spend is the thinner literature

© 2026 FUTURE PROOF™

Spending by the evidence

For ministries, states and school boards, the literature compresses into five working rules.

Retire the “money doesn’t matter” card. The strongest modern designs, pooled, put the burden of proof on the null (Jackson & Mackevicius, 2024). Budget debates should now be about targeting and delivery, not about whether resources can work at all.

Target where the effects live. Every major study finds the largest gains for low-income students and under-funded districts (Jackson, Johnson & Persico, 2016). Progressive targeting is not just equity politics; it is where the measured returns are.

Buy the mechanisms, not the metaphor. Reform dollars worked through people and time — teacher pay, class size, instructional hours (Jackson, Johnson & Persico, 2016). A funding increase should arrive with a delivery model naming what it purchases and why.

Judge on the right clock and ledger. The achievement effects accumulate over five to ten years, and the largest payoffs — attainment, earnings, poverty — arrive in adulthood (Lafortune, Rothstein & Schanzenbach, 2018). A reform judged on a two-year test window is being graded before the mechanism has run.

Instrument the spending. The gap between districts where money worked and where it did not is a measurement gap as much as a management one (Jackson, 2018). Tie each marginal program to observable outputs — instructional time delivered, mastery gained, attendance held — so the next budget fight can cite your ledger, not the literature’s averages.