I am talking about the leading group, whatever that may be, why would you split only the United States by a certain dimension and compare it to the entirety of other populations?
Ah, I get it now. Yes, it would be more apples-to-apples to, for example, compare white Americans with ancestral Europeans, by disaggregating scores in European countries by ethnic/ancestral group of origin. However, this would not affect comparison greatly, because in European countries, the children of ancestral Europeans comprise 80-90% of the total, compared to <45% in US, so taking the whole aggregate instead of this 80-90% doesn’t change that much. Same is true for East Asian countries: immigrant population in Japan or Taiwan is pretty negligible, as these countries are fairly homogenous. Finally, for countries with highly heterogeneous populations, like eg India or Indonesia, none of the large ethnic groups is ahead of US whites.
I think I've found the answer and it's simpler than that. There is no analysis splitting whites or other ethnic groups because PISA has national variations of questionnaires (source on the US one below), and in all likelihood no other country splits by race.
It's still weird, probably lacking rigor and methodological soundness, to split American races out and compare them to the bulk of other populations.
It might not be super sound methodologically to only do it for US in PISA, but I don’t think that it affect results significantly. Other high performing countries have highly homogenous populations, so the aggregate score of the entire country is very close to the score of the top performing group. On the other hand, in countries with heterogenous populations like India, Indonesia, Afghanistan or Nigeria, even the top groups are not performing very well. If you know of a country other than US, where the top performing group does significantly better than country average, and their performance is on the level of, say, European average, I’d be very curious to learn about it!
The main methodological problem is how one can leave out like 40% of the United States, and say that the US comes on top (which it still doesn't!). Like, I'm not even questioning the racial split data, but selecting the two groups and saying that the US is the best is a weird flex that wouldn't pass muster in Stats 101.
I’m not leaving out anyone. What I’m saying is that each American group individually comes out out of top, relative to matching groups elsewhere, and the fact that overall we don’t, is just an example of Simpson’s paradox. American blacks and Hispanics are ahead of foreign blacks and Hispanics too, it’s not just whites and Asians.
There are regularly PISA reports on European countries that split the data by natives (n-generation Turks are usually counted here) and immigrants. Sometimes all the immigrants are in one single group, sometimes there is fine-grained data (Somalis do very poorly, probably to nobody's surprise).
