Thanks for pointing this out and citing the paper. When I saw the Andersson graph I thought, "oh hey, those two roughly Gaussian distributions are indeed a wee bit distinguishable", then I read
> were drastically more likely to be in the Swedish financial elite than people with the surname Andersson.
and I was like, "Drastically? Really?"
Another day another shoddily reported science article.
Two distinguishable Gaussian distributions will have very different behavior in the tails. So if you knew two groups had slightly different means in wealth, then it would not be surprising to discover that at a particular cutoff in the tail, one group will be much overrepresented. Thus if you look at Clark's paper http://faculty.econ.ucdavis.edu/faculty/gclark/papers/Sweden... , you see that while incomes may only be 21% higher for descendants of noble surnames vs a normal surname, "we see that Noble Surnames are about 6 times more likely to be doctors or attorneys, and about 4 times more likely to be graduating an elite university in 1998-2012"; likewise, the mean increase in education is not that big, but the Royal Academy overrepresentation is something like 5x.
> were drastically more likely to be in the Swedish financial elite than people with the surname Andersson.
and I was like, "Drastically? Really?"
Another day another shoddily reported science article.