So the researchers in the Atlantic article point out that “when they controlled for things like race, gender, socioeconomic background, and standardized-test scores, the courses that students took in high school had very little impact on college grades.”.
Well, no, if you do a regression analysis and only compare kids in these extremely defined little subgroups, there will not be much difference in what they do or learn in high school, or even what type of college they go to, what type of course they take there and what type of grade they get. Doesn’t meant it would not make a difference if these kids actually did very different things in high school. Since most kids probably don’t, outliers disappear in the noise.