If more students are graduating than ever before, but fewer are demonstrating proficiency than ever before, what exactly does graduation mean?
Minnesota officials celebrated recently when the state announced its highest-ever high school graduation rate. In 2025, 84.9 percent of Minnesota students finished high school in four years, the second consecutive record year after 84.2 percent graduated in 2024. Governor Tim Walz called it “a moment worth celebrating,” and on its face, it is. More young people walking across a stage and collecting a diploma is unambiguously good news.
But there is a number that keeps getting buried beneath the confetti, and Minnesotans deserve to hear it plainly: fewer than half of the state’s public school students can read or do math at grade level. According to the Minnesota Department of Education’s own assessments administered in May 2025, just 49.5 percent of students met grade-level expectations in reading, and only 45 percent did so in math. We are graduating more students than ever while teaching fewer of them to proficiency. Those two facts cannot both be called success.
The Test Score Record Tells a Different Story
The National Assessment of Educational Progress, known as the Nation’s Report Card, is the only standardized measure that allows apples-to-apples comparisons across states and over time. It is administered every two years to representative samples of fourth and eighth graders in every state, and it cannot be gamed by local grading policies or graduation pathway decisions. What it shows for Minnesota is sobering.
Minnesota’s eighth-grade math scores peaked at 294 (on a 500-point scale) in both 2015 and 2017, then fell to 280 in 2022 before recovering only slightly to 282 in 2024, a net loss of 12 points over the decade. Eighth-grade reading scores followed the same arc: 271 in both 2015 and 2017, dropping to 264 in 2022 and 263 in 2024. Fourth-grade reading is the most alarming of all: scores dropped from 227 in 2015 to 213 in 2024, and 39 percent of Minnesota fourth-graders cannot reach even the basic achievement threshold on the NAEP reading test, the worst figure recorded since the assessment was first given in 1992.
Table 1. Minnesota NAEP Average Scores by Subject, 2015 to 2024
Score changes from 2015 peak to 2024. Red = decline.
Figure 1. Minnesota NAEP Scores by Subject, 2015 to 2024
Source: NAEP / Nation’s Report Card. All four grade-subject combinations show net declines over the decade.
Source: NAEP / Nation’s Report Card. All four grade-subject combinations show net declines over the decade.
The Correlation Is Striking and Statistically Significant
Now consider the graduation rate over the same period. In 2015, roughly 81.9 percent of Minnesota seniors graduated on time. That number rose gradually each year, reaching 83.7 percent in 2019, dipping slightly during the pandemic years, then climbing to 84.2 percent in 2024 and 84.9 percent in 2025.
Table 2. Minnesota 4-Year High School Graduation Rates, 2015 to 2025
Highlighted rows (2024-2025) represent record-high years.
When you plot Minnesota’s graduation rate against NAEP test scores for the five years in which both data sets are available (2015, 2017, 2019, 2022, and 2024), the statistical relationship is unmistakable.
Each point represents one NAEP year. The negative trend line confirms a very strong inverse relationship.
The Pearson correlation coefficient between Minnesota’s rising graduation rates and its falling NAEP test scores is approximately -0.90: a very strong negative relationship that holds across every tested subject and grade level.
In statistics, a Pearson r of -1.0 represents a perfect inverse relationship. A value of -0.90 falls in the “very strong” category. The probability of observing r = -0.90 or stronger by chance alone, with five data points, is less than 4 percent, meeting the conventional threshold for statistical significance. More importantly, the relationship holds consistently across all four tested subjects and both grade levels, which makes a coincidental explanation even less plausible. Something structural is happening in Minnesota’s education system, and the data demands an explanation.
The Counterargument Deserves a Fair Hearing
To be fair, correlation is not causation, and there are real explanations for the graduation rate increase that have nothing to do with lowering standards. The 2023 DFL education budget brought significant investments in English learner support, special education, mental health resources, and free school meals. Credit recovery programs expanded across the state. Schools developed alternative pathways to diplomas through career and technical education.
The gains among historically underserved students are real and represent years of deliberate effort. The table below shows the demographic breakdown.
Table 3. Minnesota Graduation Rates by Student Group, 2015 vs. 2025
All racial and ethnic groups saw improvements over the decade.
State officials also note that graduation rates measure something different from test scores. A diploma reflects four years of accumulated work, persistence, and engagement, while a standardized test is a snapshot of a single day. There is genuine value in that distinction. But the distinction only goes so far.
A Harvard and Stanford joint analysis of learning loss found that Minnesota students today are approximately three-quarters of a grade level behind their 2019 peers in math and more than half a grade level behind in reading. When NAEP scores are adjusted for student demographics, including race, income, and English learner status, Minnesota’s rankings fall sharply: from the top seven states in fourth-grade math to roughly 26th, and from third in eighth-grade math to 15th. The state has long comforted itself with high average scores that largely reflect its favorable demographics. That comfort is no longer available.
A Diploma Without Preparation Is a Deferred Cost
The consequences of this gap are not abstract. Nationally, only 33 percent of 2024 high school graduates were considered academically prepared for college-level math, down from 37 percent in 2019, even as more students reported planning to attend a four-year college. Stanford economist Eric Hanushek has calculated that Minnesota ranks eighth among states in expected lifetime income loss attributable to pandemic-era learning disruption. A diploma that does not represent genuine academic preparation is not a gift to a student. It is a postponed reckoning that tends to arrive in the form of remedial coursework, mounting debt, stalled careers, and lower lifetime earnings.
Mississippi Offers a Lesson Worth Studying
Mississippi passed a literacy-based retention law in 2013 requiring struggling third-grade readers to repeat the grade rather than advance unprepared. Critics predicted the policy would stigmatize students and widen gaps. Instead, students retained under the law scored, on average, 1.2 standard deviations above their promoted counterparts by sixth grade, a difference driven largely by gains among Black and Latino students. Mississippi climbed from near the bottom of national reading rankings, rising from 49th in 2013 to among the top tier of states in fourth-grade reading on demographically adjusted NAEP scores by 2024. Minnesota, which once occupied a comfortable spot near the top of those same rankings, now sits 39th.
Louisiana is the only state whose fourth-grade reading scores are now higher than their pre-pandemic level. Alabama is the only state to achieve that in fourth-grade math. Both pursued aggressive, structured literacy reforms that prioritize reading proficiency over age-based promotion. Minnesota’s READ Act, passed in 2023, begins to move in that direction, but experts note it lacks the accountability mechanisms that have made comparable reforms work elsewhere.
What a Diploma Should Mean
None of this means Minnesota should abandon efforts to keep students in school. Dropout prevention matters enormously, and the gains among historically underserved populations are real and worth protecting. But graduation rates and test scores should move together over time. In a healthy system, more students graduate because more students are learning.
When the two measures diverge as sharply as they have over the past decade in Minnesota, with a Pearson correlation of -0.90 between a rising graduation rate and falling test scores, that divergence is not a story of success waiting to be framed correctly. It is a warning. It tells us that something is being counted that is not being learned.
“That’s the news from Lake Wobegon, where all the women are strong, all the men are good-looking, and all the children are above average.” — Garrison Keillor
Garrison Keillor’s A Prairie Home Companion signed off for decades with that gentle joke about Lake Wobegon, a fictional Minnesota town where, by definition, every child was above average. It was a warm satire of Midwestern self-satisfaction, the comfortable belief that things here are simply better than elsewhere. For a long time, Minnesota’s education statistics seemed to confirm the joke. The state consistently ranked near the top of national comparisons, and its leaders were happy to take credit.
But the data no longer supports the punchline. When you look past the record graduation rates and examine what students are actually learning, the Lake Wobegon effect collapses. Only 45 percent of Minnesota students meet math standards. Only 49.5 percent meet reading standards. In both cases, more than half of students are not at grade level. NAEP scores have fallen to their lowest levels in decades. And the Pearson correlation of -0.90 between rising graduation rates and falling test scores raises a question that Minnesota’s leaders have not answered honestly: if more students are graduating than ever before, but fewer are demonstrating proficiency than ever before, what exactly does graduation mean?
Related: Are MN Graduation Rates Rising? (What does that really mean?)
Oh, David, you seem to be suggesting that graduation requirements and NAEP scores are created by the same people for the same purpose. That’s not even close to being true and very likely will never be true.
The achievement level descriptors (Basic, Proficient, Advanced) were set through a ‘judgmental process’ called the Modified Angoff method, where panels of educators estimate how a “just proficient” student would perform on each item. This process is inherently subjective and has been criticized extensively. The National Academy of Sciences issued a report in the 1990s concluding that the achievement levels were “fundamentally flawed” and should not be used for reporting. NCES acknowledged the criticism but continued using them anyway. The cut scores have never been validated against external criteria like college readiness or workforce outcomes in a rigorous way
Many psychometricians argue NAEP Proficient is set unrealistically high, meaning the grim statistics about students “below proficient” are partly an artifact of the cut score rather than a true reflection of learning. Comparing countries on NAEP-equivalent scales often shows that even high-performing nations would have large shares of students below NAEP Proficient.
NAEP cannot validly characterize individual students. I’m not sure classroom teachers can, either. But this limitation extends to small subgroups (rural students, specific ethnic minorities in small states, students with disabilities), which can have very wide margins of error that get underreported in policy discussions. Journalists, Times opinionists, and policymakers treat NAEP point estimates as precise facts when the confidence intervals are sometimes quite wide. Why do you do that?
Because no student takes the full NAEP test, linking across booklets requires strong statistical assumptions. If those assumptions are violated, for instance, if certain booklets were administered to systematically different student populations, score comparability is compromised in ways that are difficult to detect.
Maintaining a stable scale over decades requires that the underlying construct being measured stays constant, but reading and math instruction, curriculum, and even what skills matter have shifted. And, politicians love to offer opinions about what matters.
When NAEP updates its frameworks (as it did for reading in recent years), it creates discontinuities that complicate long-term trend interpretation. The transition to digital administration also introduced a mode effect. Students who grew up testing on paper perform differently on tablets, and the adjustment for this is imperfect. And teacher and society opinions about ‘screens’ continue to be an issue.
Nonresponse bias Is a real issue with NAEP. School nonparticipation and student absenteeism are not random. The statistical adjustments for nonresponse assume a certain pattern of missingness that may not hold, particularly in urban districts or schools wit high mobility. There is no strong external validation that the adjustments fully correct for this. Do you actually know how scores are collected in MS, AL, and LA, those miracles of modern education you cited?
States and schools vary considerably in how they apply exclusion criteria for students with disabilities and English learners. This makes cross-state comparisons potentially misleading. A state that excludes more low-performing students will appear to do better, and while NAEP flags high exclusion rates, it doesn’t fully correct for the distortion. As the student population has changed demographically over decades, comparisons of aggregate national scores conflate compositional changes with actual learning changes. A state whose scores declined may have enrolled more high-need students, not experienced a true decline in effectiveness. NAEP subgroup analyses don’t resolve this. Maybe try a Pearson thingy on it.
Despite standardized protocols, the actual experience of taking NAEP varies across schools, classroom culture, how field staff explain the test, and student familiarity with the format.. All of these introduce noise and doubt about validity. The very inconsistent shift to digital has added new variability around device familiarity and typing ability, particularly for written responses.
The National Assessment Governing Board (NAGB), which sets policy for NAEP, includes political appointees and representatives of stakeholder groups who have interests in how results are framed. Hmm. Achievement level descriptions and reporting choices reflect negotiated compromises, not purely technical judgments
NAEP is maybe the ‘best large-scale measure of U.S. student achievement available’, but “best available” is not the same as “reliable and unbiased.” Its core problems, subjective and unvalidated achievement levels, nonresponse issues, and mode transition effects mean that confident policy pronouncements based on small NAEP score changes, or precise cross-state comparisons, outrun what the data can actually support. Treat NAEP as a useful but very imperfect signal, not a grounded truth. How you treat teachers and schools is another topic.
The Mn constitution requires teaching proficiency in reading and writing—— a sign of the times.
AI will make literacy more important—— and critical thinking. Learning why
-to will be more important than how-to.