"Armed with this learning “Rosetta Stone,” we revisit various well-known results, showing, inter alia, that learning differences between most- and least-developed countries are larger than existing estimates suggest."
So, it's another item linking study. The idea is to find items that have been reused across these tests, and thus one can link the scores with some math tricks. Coverage looks like this.
The results look pretty much like every other such ranking.
Unsurprisingly, the correlation of these new results with existing ones are very similar. There is a comparison to the Altinok 2018 scores, these are the World Bank ones, r = .90 or so.
These authors are very PC and only talk about vague "human capital", "test scores" and the like. Like the other PC researchers in the area, they are puzzled by the oil country results, and returns to schooling. The magic education pill is in another castle.
There is of course no mention of intelligence, nor any of the intelligence researchers who have been using these country comparisons for decades: Lynn, Meisenberg, Rindermann, Becker, et al, not even Garett Jones.
"Once dull, always dull". Special education classes have very good metrics for teacher education and students per teacher, but they don't seem to improve much from this extra effort. So Terman infers from this that these factors cannot be of great importance.
New study of 2700 Indian whole genomes shows that Indo-European/Yamnaya/Steppe is quite low, about 15%. It is a bit higher in the north (almost 20%) and among those who speak IE languages, but not impressively so. Too much mixing since the arrival.
Inbreeding is strong. The median person had identical blocks in their genome at the level suggesting their parents were 3rd cousins, whereas the human average is about close to 4th cousins (Africans less).
Annoyingly, the data is not public, even though American taxpayers paid for this. Can't wait for Trump admin to maybe do something about this abuse (inb4 fell for it again award).
Not everybody commit scientific fraud at the same rate. Let's look at some data. First retraction rates.
We can also look at the top list of most fraudulent researchers ever (so far those that were caught). Note: data from 2019 list. There's quite a few non-Europeans.
This is important because we want a per capita measure of sorts. For highly regarded journals, European built countries produced about 75% of science (Nature Index), but maybe 30% of top fraudsters.
You have maybe seen this plot. It shows ratings of races in USA by race. The key result is that each race favors their own, except for Whites who are apparently race-blind.
Since it only covered a single year, 2020, I wondered if this would replicate across years of data. So I downloaded the data for 1964-2024.
So we see that the result replicated (I didn't use survey weights). And indeed, in 2020 and 2024, Whites have about zero ethnocentrism.
But this pattern among Whites is more complex. Here are the ratings by political ideology (self-rated 1-7 scale). The below average attitude towards fellow Whites is concentrated on the left, just as @ZachG932 found before. Especially the far left.
Mental issues roughly follow a hierarchical pattern, like cognitive abilities, like a general factor on top. At least, statistically.
The motivating factors behind this approach compared to the categorical (diagnostic) approach are: 1) evidence of continuity between clusters, 2) binary encoding of continuous data loses information, 3) correlations among diagnoses are the norm, 4) a given person may not quality for any particular diagnosis, yet have severe symptoms.
Some aspects of mental problems haven't been integrated into the hierarchical model yet. Say, unusual sexual interests (from foot fetishes to pedophilia and rape fetishes).