Do small classes help young children learn?
✓ Numbers check out
Counts as holding up
The code and data produced the same result. That doesn't prove the study's method is right.
An independent rerun of Tennessee's STAR class-size experiment, using a public copy of the data, reproduced the reported first-grade gain from small classes, so the numbers check out.
The study asked whether putting young children in smaller classes helps them learn. Tennessee's STAR experiment assigned thousands of young children at random to small or normal-size classes, which is why the results are considered strong evidence. Krueger reported that first graders in small classes scored higher on tests than peers in regular classes.
The rerun was an independent rebuild, meaning it was not the original author's code, and used a public copy of the data with reading and maths scores only, so it could not match the original exactly. The rebuild found a first-grade gain of 8.5152 percentile points, compared with the 8.57 reported, and the verdict was that the numbers check out. This only shows the same data and method give the same number; it does not confirm the sample was fair, that the method was right for the question, or that the gain lasts or holds elsewhere.
The text was written by a private AI model on NEAR AI (z-ai/glm-5.3-flash). The verdict and numbers come straight from the blockchain, and the AI can't change them.
- What the paper said
- Small Class 8.57 (1.97). Table 6, panel B, column 1, of the working paper (NBER 6051), read from the page image
- What we got
- 8.5152
- Whose code ran
- Independent rebuild
- How well known
- Cited by 1,116 studies
- What forecasters thought
- Nobody forecast this one. We posted it with the result already known, to show how a famous paper looks.
- Proof
- Read the receipt or see claim 33
What we didn't check
- Whether the sample was fair or large enough.
- Whether test data leaked into training or tuning.
- Whether the result holds on new data.
- Whether the study's method was the right one for its question.
- Whether the gain lasts, or what happens in other places.
- The published journal table, which isn't free to read. We used the working paper's.
Community notes
Notes are people's own words, kept on the blockchain. They never change a verdict, and the Lab's AI never reads them. Posting costs a little NEAR for storage, and what isn't used comes back.