Mission
Our mission is to Build a More Numerate Future
Numbers are how the modern world argues. And we want to teach our students to hold their own in that argument. This page shows the evidence that brought us here, along with the cautions that come with it, so you can check our reasoning yourself.
“[T]he time may not be very remote when it will be understood that … it is as necessary to be able to compute, to think in averages and maxima and minima, as it is now to be able to read and write.”
- 34% of U.S. adults scored at Level 1 or below in numeracy in 2023
- 45% of high school seniors scored below NAEP Basic in math in 2024
- 87% of gynecologists understood a positive test after a one-hour lecture, up from 21%
How are adults doing?
In the United States, 34% of adults scored at Level 1 or below in numeracy in 2023, up from 29% in 2017 (NCES). Across OECD countries the average is 25% (OECD figures, via Ireland's Central Statistics Office). Adults at Level 1 can do basic arithmetic with whole numbers, money, and simple percentages like 25% and 50%, and can read a value from a table or graph, as long as the problem takes only one or two steps (OECD level descriptions, via England's national PIAAC report).
Source: NCES, PIAAC 2023 U.S. results · OECD average: OECD Survey of Adult Skills 2023, via Ireland's Central Statistics Office
The average score moved the same way. The U.S. average was 249 in 2023, down from 255 in 2017 and 257 in 2012, and below the international average of 263 (NCES release). A score under 226 is Level 1 or below, and a score of 276 or higher is Level 3 or above (NCES). The share at Level 3 or above, 38% in 2023, was not measurably different from 2017, so the change we can measure is at the bottom of the scale.
The vertical axis starts at 240, not zero, so the drop looks steeper than it would on the full scale. Source: NCES release, Dec 10, 2024.
Some of that rise may come from the test itself. The 2023 test moved to tablets, and NCES says comparisons with earlier years need caution because of differences in the assessments and in scoring (NCES). So we also check a second source.
How are students doing?
The National Assessment of Educational Progress (NAEP) reports 12th-grade math as one trend line back to 2005 and ran a bridge study when it changed devices in 2024 (NAEP). About 19,300 seniors at 1,500 schools took the 2024 test. There we see the same direction: in 2024, 12th-grade math scores were the lowest ever reported, and 45% of high school seniors scored below NAEP Basic, the largest share on record (NAGB).
Source: The Nation's Report Card, 2024 Grade 12 Mathematics · NAEP says its achievement levels are used on a trial basis and should be interpreted with caution.
An average can hide who moved, so NAEP also reports scores by percentile. The 10th percentile is the score that 10% of students fall below, and the 90th is the score that 90% fall below. Between 2019 and 2024, students near the top held steady, and the drop was largest in the bottom quarter.
Can these skills be taught?
We think so, and one of our favorite pieces of evidence starts with a question. Try it before you read on. In a screening program, 1% of women have breast cancer. If a woman has it, the chance that her mammogram is positive is 90%. If she doesn't, the chance is still 9%. A woman tests positive. What is the chance that she actually has breast cancer?
Gerd Gigerenzer asked 160 gynecologists this at the start of a training session. A majority, 60%, answered 80% to 90%, and 19% answered 1% (Gigerenzer, BMJ, 2011). Now look at the same facts as counts of people.
- 9 have cancer and test positive
- 89 don't have cancer but test positive
- 1 has cancer but tests negative
- 901 don't have cancer and test negative
Now keep only the 98 women who tested positive
9 out of 98 is 9.2%, or roughly one in ten.
Source: Gerd Gigerenzer, “What are natural frequencies?”, BMJ 343, d6386, 2011.
Gigerenzer calls these counts natural frequencies. After a one-hour lecture to 150 gynecologists that taught them, 87% understood the chance of cancer after a positive test, compared with 21% beforehand (Gigerenzer, Observer, 2007).
Source: Gerd Gigerenzer, Observer, 2007
One lecture on one kind of problem is not a curriculum, and we can't promise the same jump in a classroom. What this result does show us is that a person with years of training can miss a number like this, and then read it correctly once it is presented and taught differently.
What we do about it
When we fail to equip our students with data literacy, we aren't just letting their test scores slip; we are disenfranchising them from the conversations that shape their lives. LWN is our attempt to help close this gap and empower the next generation to participate in a data-driven society and have a say in shaping their own future.
So students in our projects collect real data, work with it themselves, and learn to ask how a number was collected, estimated, analyzed and presented.