Education: evidence

Education evidence in the United States shows reading scores falling to their weakest levels in decades, a widening gap between higher- and lower-performing students since about 2010, and record nominal spending per pupil. Research finds that more money, smaller early-grade classes and some preschool programs can improve outcomes, but also that a large statewide pre-K program produced negative results, so the evidence supports careful design more than any single slogan. Many of the questions the public cares about most still lack settled answers.

In plain words: tests show many students, especially those who were already behind, are doing worse than before the pandemic. Schools are spending more dollars than ever, though not adjusted for inflation. Some studies say extra money helps a lot; others say it depends on how it is spent. Some preschool programs helped children for years; one big state program seemed to hurt. This page lays out the numbers, the disputes, and the gaps. Why people rank education as they do is on Education; who decides is on Education: who decides.

Defined terms

Education evidence relies on a few technical terms.

NAEP. The National Assessment of Educational Progress, often called the Nation’s Report Card, a federal assessment given to samples of students. NAEP achievement levels are not the same as state “grade level” definitions; see What NAEP civics measures.

Below NAEP Basic. Scoring under the lowest NAEP achievement level.

Randomized controlled trial (RCT). A study in which participants are assigned to a program or a comparison group by chance, so that differences in outcomes can be attributed to the program.

Lottery study. A natural experiment that uses an admissions lottery to create comparable groups.

Standard deviation (σ). A measure of spread used to express effect sizes; 0.1σ is a small but often meaningful change in education research.

Nominal dollars. Dollars not adjusted for inflation.

Evidence

Education data and research fall into a few areas: achievement, money, class size, early learning, choice, and public confidence.

Achievement is down, and gaps are wider

The 2024 NAEP results, released January 29, 2025, showed grade 4 and grade 8 reading scores each down 2 points from 2022, after a 3-point decline from 2019 to 2022.⁠¹ About 40% of fourth graders scored below NAEP Basic in reading, the largest share since 2002, and about a third of eighth graders did, the largest share on record.⁠¹ Math was mixed: grade 4 rose 2 points from 2022 and grade 8 was flat, with the nation still below 2019 levels.⁠¹

The National Assessment Governing Board reported that the gap between higher- and lower-performing students has widened for about a decade, with lower performers now about 100 points below the highest performers. Board member Martin West said, “We made progress in closing this gap until around 2010, but it’s been steadily widening since”.⁠¹ Because the widening began before the pandemic, pandemic disruption cannot be the whole explanation.

Money: spending is at a record in nominal terms

Public school systems spent $17,619 per pupil in fiscal year 2024, up 6.6% from $16,526 in fiscal year 2023.⁠² The Census Bureau states that these “Statistics are not adjusted for inflation or cost-of-living differences between geographic areas”.⁠² Federal revenue fell 4.4% to $115.1 billion that year.⁠² Public schools enrolled 49.6 million students in fall 2022.⁠³

Internationally, the OECD reports U.S. spending per student is high, “especially at tertiary level,” and that U.S. teacher salaries are low compared with similarly educated workers.⁠⁴

Does more money improve results?

Two major bodies of research reach different emphases.

Jackson, Johnson and Persico used court-ordered school finance reforms as a natural experiment. They estimated that a 10% increase in per-pupil spending each year for all twelve years of school “leads to 0.27 more completed years” of education, “7.25 percent higher wages,” and “a 3.67 percentage-point reduction in the annual incidence of adult poverty,” with effects “much more pronounced for children from low-income families”.⁠⁵

Jackson and Mackevicius later pooled many studies. They estimated that, on average, “a $1000 increase in per-pupil public school spending (for four years) increases test scores by 0.0352σ, high school graduation by 1.92 percentage points, and college-going by 2.65 percentage points”.⁠⁶

Eric Hanushek argues that “how money is spent is more important than how much is spent,” and that the long rise in real spending since 1970 did not close achievement and attainment gaps.⁠⁷ Jackson, Johnson and Persico respond that the time-trend argument “will hold only if nothing else has changed between 1970 and present day,” pointing to changes in family structure and neighborhood conditions.⁠⁸ The two sides agree that spending can be wasted and that how it is spent matters; they disagree on how reliably extra money, as actually spent by districts, produces gains.

Class size

The Tennessee STAR experiment randomly assigned about 11,600 students and their teachers to small or regular classes from kindergarten. Alan Krueger found that small classes raised performance “by about 4 percentile points the first year,” with an additional gain of “about one percentile point per year” after that, and that class size had “a larger effect on test scores for minority students and for those on free lunch”.⁠⁹ STAR is one of the strongest designs in education research. Its limit is scale: reducing class size statewide requires many more teachers, and whether new hires would be as effective as those in the experiment is a separate question.

Early learning: strong positive and strong negative results

Early learning shows how much design matters.

In Boston, Gray-Lobe, Pathak and Walters studied more than 4,000 children who entered a preschool lottery between 1997 and 2003. Attending preschool raised on-time college enrollment “by 8.3 percentage points, an 18 percent increase,” though the study found “no discernible change on the MCAS,” the state test.⁠¹⁰ Effects showed up in later life outcomes even without test score gains.

In Tennessee, a randomized study of 2,990 children from low-income families who applied to oversubscribed state pre-K sites found that children assigned to pre-K “had lower state achievement test scores in third through 6th grades than control children, with the strongest negative effects in 6th grade,” along with negative effects on disciplinary infractions, attendance and receipt of special education services.⁠¹¹ The authors frame the study with the observation that “As state-funded prekindergarten programs expand, it is critical to investigate their short- and long-term effects”.⁠¹¹

Both are serious studies. Supporters of public pre-K point to Boston and argue that quality and alignment with later grades explain the difference. Skeptics point to Tennessee and argue that a scaled-up state program is the realistic case. The evidence does not settle which result is more typical.

School choice

Charter schools enrolled 3.7 million students in fall 2021, 7% of public school students, up from 1.8 million and 4% in fall 2010.⁠¹² Growth on that scale shows demand from some families. It does not, by itself, show whether charter students learn more than they would have elsewhere; that question requires lottery or comparison studies, and results differ by state, sector and period. This page has not reviewed that literature in enough depth to summarize it, and states it as an open question rather than a finding.

Higher education confidence

Confidence in higher education fell to 38% of adults in 2026 from 42% in 2025 and 57% in 2015, according to the Gallup and Lumina Foundation survey as reported by The 74.⁠¹³ Confidence is not evidence of quality, but it affects enrollment and political support.

Where the research disagrees

Education research disagreements fall into recognizable patterns.

Money. Natural-experiment studies.⁠^5,6 find meaningful returns; critics emphasize the national time trend and the variance in how districts spend⁠⁷ Both camps cite real data.

Early learning. Positive lottery evidence from Boston,⁠¹⁰ and negative randomized evidence from Tennessee⁠¹¹ cannot both be “the” effect of pre-K. Program quality, the alternative care children would have had, and what happens in later grades are all candidate explanations.

Test scores versus later life. Boston preschool raised college going without moving state test scores⁠¹⁰ while Tennessee lowered test scores.⁠¹¹ Which outcome a voter treats as the goal changes which evidence looks strongest.

Scale. Small, well-run programs may not keep their effects when expanded. STAR⁠⁹ was large for an experiment but small next to a statewide policy.

Competing interpretations

Education evidence is read differently by people with different values.

One reading holds that the decline in reading and the widening gap call for more resources aimed at struggling students, citing the finance studies. A second reading holds that the same decline, during a period of record nominal spending, shows that current systems convert money poorly and need structural change, such as different curricula, accountability or choice. A third reading emphasizes early childhood and points to Boston; a fourth points to Tennessee and urges caution before expanding pre-K. A fifth reading notes that post-pandemic attendance, family stress and technology may matter as much as anything schools control. Each reading is consistent with some of the evidence above, and none is consistent with all of it.

Application to this framework

Education evidence matters to My Vote Forward because a priority is not a judgment about effectiveness. A participant can rank Education first and still think a particular program fails, or rank it lower and still support a program the evidence favors. The framework keeps three judgments apart: how important a subject is, how a defined pool of money might be divided, and whether a specific program works. See Tradeoffs and Priorities.

When a proposal is shown, the framework should link to the strongest evidence on each side, label research results as research, and say so when the evidence is mixed. Hypothetical example: a district proposes universal pre-K. The proposal page would show the Boston and Tennessee results together, rather than only the one that supports or opposes the proposal.

Unresolved questions

Education evidence leaves important questions open:

• Why have gaps between higher- and lower-performing students widened since about 2010, before the pandemic?
• What makes some preschool programs produce lasting gains and others negative results?
• How much of recent spending growth was absorbed by inflation and by costs, such as benefits, that do not reach classrooms?
• How do charter, voucher and education savings account effects compare across states, once lottery and matched studies are pooled carefully?
• Can effects found in experiments survive statewide expansion with ordinary staffing?