ANALYSISApril 13, 2026
Should standardized testing be the basis of school accountability?
The United States is experiencing a renewed national debate over whether standardized testing should serve as the primary basis for school accountability, intensified by declining post-pandemic student performance and new Trump administration policies inviting states to request flexibility from federal testing mandates. Oklahoma and Texas are moving toward replacing annual state exams with tests administered multiple times throughout the school year, while states like New Jersey and Massachusetts have introduced AI scoring of standardized tests, raising new concerns about accuracy and data privacy. Meanwhile, NAEP data released in late 2025 shows historically low scores in 12th-grade math, reading, and 8th-grade science, with only 22% of 12th-graders scoring proficient in math.
If a school lifts kids out of poverty but scores stay low, has it failed — or has the test? And if we stop measuring, how do we know which schools to fix and which to fund?
Not familiar with this story? Get context →
Comparability as equity infrastructure
C
The civil rights case for standardized testing is not conservative spin — a coalition of twelve organizations including the NAACP has explicitly endorsed standardized data as their primary tool for documenting inequity and demanding resources for students of color, low-income students, and students with disabilities. That endorsement is built on a hard lesson: school systems have every incentive to pretend disadvantaged students are fine, and comparable data makes that pretense unsustainable. When Oklahoma and Texas shift to multiple-occasion local testing, they are not modernizing — they are dismantling the cross-district comparability that allows anyone to say 'this district is failing its Black and brown students relative to every district around it.'
L
We agree on the NAACP coalition — I cited them too, and their position is serious. But endorsing standardized data as an equity tool is not the same as endorsing the current accountability architecture built on top of it. The civil rights case is for comparability and disaggregation, not for high-stakes consequences that fall on schools serving the most disadvantaged students for outcomes those schools structurally cannot control.
C
If the consequences are the problem, fix the consequences — but you cannot decouple 'comparable data' from 'accountability' in practice, because the political pressure to act on inequity disappears the moment the data stops being required. ESSA proved that: more flexibility, preserved testing floor, and the trend line still went down.
L
ESSA preserved the annual testing floor precisely because that floor is where equity advocates drew the line — the problem is not that ESSA gave states flexibility, it is that flexibility without enforcement has no teeth, which is an argument for stronger federal accountability, not weaker testing.
Schools' limited control over scores
C
Critics argue that schools account for less than 10% of test score variance, meaning most of what standardized tests measure is poverty, not pedagogy — and using those scores as accountability levers punishes teachers in hard places for their students' circumstances. That objection is real, and the conservative position is not to dismiss it. But the policy implication does not follow: accountability systems can be designed around growth rather than absolute performance, measuring what schools actually move, and that is precisely what value-added models attempt to do.
L
Value-added models sound like the answer until you look at their track record — they carry error rates wide enough to misclassify individual teachers and schools at rates that undermine any claim to precision, and they have faced legal challenges in multiple states for exactly that reason. Conceding that 90% of variance is outside school control and then proposing to measure the 10% with an instrument that cannot reliably isolate it is not a solution, it is a restatement of the problem.
C
The imprecision of value-added models is a real critique of a specific methodology, not of growth-based accountability as a concept — and the alternative you are implicitly defending, letting districts self-report progress on locally designed measures, has a worse track record of precision and a far worse track record of honesty.
L
The question is not VAM versus self-reporting — it is whether we are willing to invest in assessment systems capable of actually distinguishing school quality from neighborhood disadvantage, which requires more rigor than current standardized tests provide, not less.
NAEP decline and accountability loosening
C
The 2024-2025 NAEP data shows only 22% of 12th-graders scoring proficient in math and 35% in reading, with 12th-grade reading at its lowest point since 1992. This is happening in the same decade that states have been granted maximum flexibility to move away from comparable annual assessments. That is not coincidence — ESSA's flexibility expansion preceded continued decline, and the pattern is consistent with what happens when accountability pressure is relieved without anything comparably rigorous replacing it.
L
Attributing post-2019 score declines primarily to ESSA flexibility requires ignoring a pandemic that closed schools for over a year and disrupted learning for every student in the country. The decline is real and alarming, but using it to argue against assessment flexibility conflates two separate variables — COVID's effects on learning and states' accountability structures — in a way that the data cannot actually support.
C
The pandemic explains the sharpest drops, but 12th-grade reading was already declining before 2020, and the states that had moved furthest from comparable accountability showed the weakest recovery — which suggests the structural loosening made the system less capable of responding, not just less capable of measuring.
L
If the argument is that weak accountability systems recover more slowly, that is an argument for better-designed accountability, not for preserving an annual testing regime that even its defenders acknowledge needs reform — and 'reform' cannot mean 'keep everything the same.'
Missouri data and predictive stakes
C
The Missouri longitudinal study found that 8th-grade students who scored at the top of reading assessments are 62 times more likely to earn a bachelor's degree than their lowest-scoring peers. That number is not produced by a biased instrument — it is telling you exactly where life trajectories diverge and when. If you eliminate the annual comparable assessment that surfaces that gap at 8th grade, you lose the earliest reliable signal that intervention is possible.
L
The 62-times figure actually strengthens the case for intervening on the conditions producing low scores, not just flagging them annually. If 8th-grade reading predicts college completion that powerfully, and schools control less than 10% of score variance, then the accountability system should be triggering resource interventions in the neighborhoods producing those scores — not consequences for the schools embedded in them.
C
Resource interventions require knowing which students need them, and that requires a comparable signal that exists independent of what a district chooses to report about itself — you cannot trigger the intervention without the measurement.
L
Agreed that measurement must precede intervention — the dispute is whether current testing regimes are designed to trigger the right interventions, or whether they are designed to produce annual headlines that substitute for the structural investment those students actually need.
AI scoring and modernization risks
C
The Massachusetts AI scoring debacle — errors affecting hundreds of exams in 2025 — is a concrete warning about what happens when political pressure to make testing feel less burdensome drives technological shortcuts. The impulse to modernize assessment is legitimate, but modernization that introduces systematic error at scale while reducing transparency is not an upgrade. It is accountability laundering.
L
Massachusetts is a cautionary tale about bad implementation, not about assessment innovation in general — the lesson is 'validate AI scoring before deploying it statewide,' not 'never update the technology underlying standardized tests.' Standardized bubble tests from the 1970s have their own systematic errors; we just stopped noticing them because they are familiar.
C
The familiarity of existing errors is actually the point — known error characteristics can be accounted for in policy design, while novel AI scoring errors arrive opaque and at scale before anyone understands their distribution, which is a qualitatively different accountability risk.
L
Then the standard should be transparency and validation requirements for any scoring methodology, AI or otherwise — which is an argument for rigorous oversight of testing infrastructure, not a defense of the status quo.
Conservative's hardest question
The strongest challenge to this argument is the evidence that schools account for less than 10% of student test score variance, meaning most of what standardized tests measure is outside any school's control — which makes test-based consequences for schools a potentially unjust lever. This is genuinely difficult to dismiss, because if the accountability system punishes schools for their students' circumstances more than their instructional quality, it may demoralize good teachers in hard places without improving outcomes.
Liberal's hardest question
The claim that schools account for less than 10% of test score variance is genuinely difficult to dismiss — if that figure is accurate, then high-stakes school accountability built on test scores may be holding schools responsible for outcomes they structurally cannot control, punishing poverty rather than pedagogy. A credible liberal accountability framework must grapple with whether annual testing, however equitably administered, can distinguish school quality from neighborhood disadvantage at the level of precision that funding and intervention decisions require.
Both sides agree: Both sides agree that standardized testing data is a legitimate and necessary tool for documenting inequity affecting low-income, minority, disabled, and English-learning students — citing the same NAACP coalition as evidence.
The real conflict: They disagree on a factual-interpretive question: whether the finding that schools account for less than 10% of test score variance means high-stakes school accountability is structurally unjust, or whether that variance can be isolated through growth-based models to make school-level accountability valid.
What nobody has answered: If value-added and growth-based models can isolate school-level contributions from neighborhood disadvantage, why have decades of their use failed to produce a politically durable accountability system that both civil rights advocates and teachers in high-poverty schools find credible and just?
Sources
- NAEP Nation's Report Card 2024–2025 data on 8th-grade science and 12th-grade math and reading proficiency levels
- No Child Left Behind Act (2001) statutory requirements for annual testing in grades 3–8
- Every Student Succeeds Act (2015) provisions on state accountability systems
- Education Trust public statements on standardized testing and comparable statewide data
- NAACP and coalition of 12 civil rights organizations' public statements on testing and equity
- 2024 Missouri student study on 8th-grade test scores and bachelor's degree attainment rates
- Georgetown University Edunomics Lab research pairing test results with K-12 spending data
- Massachusetts Department of Education reports on AI scoring errors in 2025 statewide assessments
- New Jersey education officials' announcement of AI scoring for spring 2025 standardized tests
- Trump administration U.S. Department of Education Dear Colleague letter authored by acting assistant secretary Craig Trainor
- FutureEd policy analysis on revamping federal testing provisions
- Massachusetts Consortium for Innovative Education Assessment (MCIEA) program description
- Dr. Robert Sternberg public statements on standardized testing and higher-order skills
- Reports on Oklahoma and Texas proposals to shift from annual to multiple-occasion testing