IANAL, but I would say it's because these are recruitment letters, not acceptance policies. As in: it alters where you look, not what you're looking for. And that can alter the balance in the representation that you end up with, even if your selection criteria are exactly the same across all groups.
It's the same as hiring policies. Imagine you have an objective test that you can give candidates, and you are willing to accept anyone who gets a 5 on a 5-point scale. The test is completely independent of race. You have two schools to draw candidates from. One is 90% white, the other 90% Black. You have resources to recruit at only one school. You want to increase the representation of Black employees, without lowering your standards.
Which school do you recruit at? The 90% Black one. You end up with employees who are exactly as qualified as if you had recruited at the other school, but a higher percentage of them are Black. If you're white and go to the second school, your chances of getting an offer are no different. If you're white (or Black) and go to the first school, you get no recruiter and unless you find out about the position on your own, you're screwed. If the schools are the same size, then you could say that being white lowered your chances of getting hired, as a direct result of the choice that the recruiter made.
Does the recruiter have a responsibility to make your odds of being admitted independent of your race? What would that mean? If there are 100 schools to choose from, it would mean that the recruiter would have to go to every one of them, and spend an amount of time at each inversely proportional to the school percentage of some race, which doesn't take into account the overall size of each school... it doesn't even work mathematically.
Back to the original example, imagine if Sparse Country states had twice as many good Asian male candidates as Asian female, and twice as many of those as good white candidates. (The percentages of not good candidates could be completely different, even reversed, and it wouldn't change anything.) Nothing stops any of those candidates from applying. But you can adjust your recruiting policies to get roughly the same number of good applications from each group.
This reduces the odds that a given Asian male (sampled evenly from the population) will end up admitted. But it does not affect the odds for an Asian male who chooses to apply.
It's the same as hiring policies. Imagine you have an objective test that you can give candidates, and you are willing to accept anyone who gets a 5 on a 5-point scale. The test is completely independent of race. You have two schools to draw candidates from. One is 90% white, the other 90% Black. You have resources to recruit at only one school. You want to increase the representation of Black employees, without lowering your standards.
Which school do you recruit at? The 90% Black one. You end up with employees who are exactly as qualified as if you had recruited at the other school, but a higher percentage of them are Black. If you're white and go to the second school, your chances of getting an offer are no different. If you're white (or Black) and go to the first school, you get no recruiter and unless you find out about the position on your own, you're screwed. If the schools are the same size, then you could say that being white lowered your chances of getting hired, as a direct result of the choice that the recruiter made.
Does the recruiter have a responsibility to make your odds of being admitted independent of your race? What would that mean? If there are 100 schools to choose from, it would mean that the recruiter would have to go to every one of them, and spend an amount of time at each inversely proportional to the school percentage of some race, which doesn't take into account the overall size of each school... it doesn't even work mathematically.
Back to the original example, imagine if Sparse Country states had twice as many good Asian male candidates as Asian female, and twice as many of those as good white candidates. (The percentages of not good candidates could be completely different, even reversed, and it wouldn't change anything.) Nothing stops any of those candidates from applying. But you can adjust your recruiting policies to get roughly the same number of good applications from each group.
This reduces the odds that a given Asian male (sampled evenly from the population) will end up admitted. But it does not affect the odds for an Asian male who chooses to apply.