Note how systematic and inexorable the graphs are; it's not any individual college or policy or region. I suspect that this is a mechanical effect of top colleges keeping enrollment essentially static even as the population continues expansion; this means that getting in becomes ever harder (see for example average SAT scores at MIT or Harvard), which in turn changes group admission ratios by the usual tail effects - when you have two normal distributions with different means, the further out you go, the more lopsided the fraction becomes. If you have a mean difference of, say, 100 SAT points between two groups, that leads to a relatively small difference in the number from each group passing a low cutoff like 900, but it means that there will be hardly any members of the lower group reaching 1600 compared to several from the higher group. The tails thin out too fast. (The tails effect is likely also related to the Asian overrepresentation.) Presumably the larger effects in elite colleges are why the NYT investigator focused on them rather than national aggregates, which will show smaller shifts as the lower-scoring students can still go to the unselective colleges.

So as long as Harvard et al keep their incoming classes small and the number of prospective college students remains high or keeps increasing, the imbalance will continue. A few elite colleges may be able to poach from each other to improve their numbers (like Harvey Mudd) but obviously they can't all do that.