US national test scores show the largest maths decline on record

The signal

The Nation's Report Card, the main US education-sector achievement measure, released two sets of results in autumn 2022, pointing the same direction. The 2022 NAEP mathematics highlights show the national average score for fourth-graders fell 5 points and for eighth-graders fell 8 points, each measured on NAEP's 0–500 scale against 2019. The report states these are the largest maths declines at both grades since the assessment began in 1990, with grade 4 below every prior year to 2005 and grade 8 below every prior year to 2003. A second, separate release covering only age 9 students, the 2022 long-term trend highlights, reports a 5-point reading decline since 2020, the largest average reading decline in that series since 1990, and a 7-point mathematics decline, the first fall mathematics has recorded in the long-term trend measure.

The evidence

Both documents describe assessments given to nationally representative samples: roughly 116,200 fourth-graders in 5,780 schools and 111,000 eighth-graders in 5,190 schools for the main mathematics assessment, and about 7,400 age 9 students in 410 schools for each long-term trend subject. The two series measure different things: the main assessment tracks grades 4 and 8 since 1990 on subject-specific scales, while the long-term trend tracks fixed ages on a separate scale designed for consistency across decades. That the two independent series moved together, one showing the steepest maths fall in its history and the other recording maths' first-ever decline, is what makes this a signal rather than a single anomalous score. Both reports also note that lower-performing students fell further than higher-performing students, changing the shape of the distribution and not only its average.

Timeframe and confidence

The assessments were administered between January and March 2022, comparing against a 2019 (main NAEP) or 2020 (long-term trend age 9) baseline. This is editorial reading: two nationally representative federal assessments, run independently on different scales, agreeing on direction and magnitude is comparatively strong evidence of a real shift in measured performance over that window, not a sampling artefact. Neither document identifies a cause; both are score reports, not explanations.

What would change the reading

The long-term trend release for ages 13 and 17, which the report says was scheduled for spring 2023, would show whether the decline was confined to younger students or ran across ages. A subsequent NAEP main assessment cycle showing scores stabilising or recovering toward 2019 levels would weaken the reading that this was a durable break rather than a one-off measurement.

• Did the next main NAEP cycle show recovery, stagnation, or further decline at grades 4 and 8?
• Did the age 13 and 17 long-term trend results released in 2023 show the same pattern as age 9?
• Did the score distribution narrow or keep widening between lower- and higher-performing students?

Two independently designed federal assessments recording the same direction in the same testing window is a stronger signal than either alone, but neither report says why scores moved, and the age 9 result is only part of a series still being published.

Source trail

1. Mathematics: NAEP Highlights 2022www.nationsreportcard.gov · Source publication: not established · Retrieved 2026-09-16

   Reports the 2022 national mathematics assessment score declines for grades 4 and 8 versus 2019, and states these are the largest since 1990.

2. NAEP Long-Term Trend 2022: Age 9 Highlightswww.nationsreportcard.gov · Source publication: not established · Retrieved 2026-09-16

   Reports the 2022 long-term trend results for age 9 students, showing reading's largest decline since 1990 and mathematics' first-ever decline in the series.

Read across the evidence

• US undergraduate enrolment kept falling two years into the pandemic
• Coursera's IPO filings showed how few learners became payers
• Retail sales data is also a survey, just of firms not people

Platforms & distribution · All 100 historical records