Here’s the relevant Espenshade/Chung article: https://www.princeton.edu/~tje/files/webOpportunity%20Cost%20of%20Admission%20Preferences%20Espenshade%20Chung%20June%202005.pdf
Quick read: Their basic model is really simplistic: They essentially reduce all admissions variables to race, athletic something or other (they say “recruited athletes,” but their “recruited athletes” have less than a 50% admission rate, so they are not talking about anything like actual athletic recruitment), legacy status, U.S. citizenship, SATs, and randomness. The number of athletic and legacy candidates is comparatively small in their dataset (the 1996-1997 admission season), so those preferences have a fairly small effect on overall numbers. As a result, when they run their simulation without taking race into account, essentially it becomes a model where the only meaningful variable is SAT scores, and the class is shaped to conform to the SAT distribution among unhooked white students. If you do that, Asians replace most of the Black and Hispanic students.
But no one does anything like that, or would.
One of the things I really noticed, reading the article, is how wildly out of date the numbers are. The overall admission rate for U.S. citizens at the three elite colleges in the study is over 23%; legacies are a little less than 50%; anyone with SATs over 1500 (out of 1600) is 40%. As noted above, recruited athletes are just under 50%, too, which is a total puzzler, because experience tells me that number is more like 95%, if not 99%. (The number of athletes admitted for the three colleges is also significantly greater than current Ivy rules would permit for recruited athletes.) In short, the world being modeled is not the world of today, and not really the world of a generation ago, either.
Another interesting point: In the study, international applicants are considered in the race distribution numbers. You can tell that because the sum of all racial categories equals the sum of all applicants, including international applicants. But it wasn’t clear to me how international applicants were handled in the simulations, because the relevant tables don’t have those numbers. (I may not have read hard enough, but I couldn’t find any discussion, either.)