www.lesswrong.com/posts/QTLTic5nZ2DaBtoCv/birth-order-effect-found-in-nobel-laur...

The article’s own data sheet contains multiple non-firstborn laureates coded as firstborn.

In the public spreadsheet linked from the post:

• Peter Grünberg is recorded as 1,2 (first of two).
• Toshihide Maskawa is recorded as 1,3 (first of three).
• Charles K. Kao is recorded as 1,4 (first of four).

But the official Nobel autobiographies say otherwise:

• Peter Grünberg wrote that after the war his mother was left with "her two children, namely my two-year older sister and myself" — i.e. he was second of two, not first.
• Toshihide Maskawa wrote "I was born in 1940 as the second child" and then explains that he had an older sister and a younger sister — i.e. second of three, not first.
• Charles K. Kao wrote that "The first child ... was a daughter, followed two years later by a son ... Finally in 1933, I was born ... My younger brother Timothy was added to the family four years later" — i.e. third of four, not first. Those older siblings were aged 10 and 8 when they died, so they are not infant deaths that could be excluded under the sheet’s own "not counting infant deaths" convention.

So the posted dataset overstates the number of firstborn laureates by at least three cases.

That matters because the claimed effect is very small and only barely significant by the article’s own numbers: 100 laureates, +10 percentage points, 95% CI 0.2 to 19.8, p=0.044. Correcting just these three errors lowers the observed firstborn excess by about 3 percentage points (from about +10 to about +7). Given that the published 95% interval only barely excludes zero, that correction is enough to move the result back into non-significant territory.

So the claim that the analysis found a statistically significant 10-point firstborn effect is not supported by the linked data; the underlying sheet contains clear birth-order miscoding in the direction needed to create or inflate the effect.

The significance claim depends on a dataset that contains clear birth-order miscoding.

In the linked spreadsheet used for the analysis:

• Peter Grünberg is entered as 1,2 (first of two), but his Nobel autobiography says his mother was left with "her two children, namely my two-year older sister and myself" — so he was second of two.
• Toshihide Maskawa is entered as 1,3, but his Nobel autobiography says "I was born in 1940 as the second child" — so he was second of three.
• Charles K. Kao is entered as 1,4, but his Nobel autobiography says the couple’s first child was a daughter, the second a son, then "Finally in 1933, I was born" and later a younger brother was born — so he was third of four.

Those are three direct firstborn/non-firstborn errors in a sample the post says contains only 100 usable laureates. Because the reported effect is only 10 percentage points and the post’s own reported 95% confidence interval is 0.2 to 19.8 percentage points, the result is only barely above zero. Correcting three wrongly-added firstborns lowers the estimated excess by about 3 percentage points, which is enough to push the result out of statistical significance. So the statement that the Nobel-laureate result is significant at p=0.044 is not correct.