It is relatively well-known that blacks and Hispanics perform worse in college than whites and Asians do. This, some people propose, is an obvious consequence of affirmative action. By allowing disadvantaged groups to gain admission to college with lower test scores, lots of underqualified applicants will be admitted. This will then generate the aforementioned inequality in college performance.

But would race-blind admissions help eliminate this inequality? On the one hand, such an admission policy would evaluate students with the same standards regardless of race. On the other hand, if there’s a more fundamental inequality among students that already exists prior to college (like inequality in high school performance), then even race-blind policies can’t prevent unequal performance in college.

To see what I mean by this last point, consider the paper of Krishnamurthy and Edlin (2014) (ungated version). It’s a fact that the distribution of test scores such as the ACT or the SATs (which are used to evaluate students applying to college) differs by race. In particular, restricting our attention to blacks and whites only, the score distribution of whites is generally better than that of blacks. The figure below illustrates this for some ACT test data.

Let us summarize admission decisions in the form of a cut-off score such that if you score less than this cut-off, you won’t gain admission. If we want the same cut-off score for both blacks and whites (e.g. 20), then we’d have the situation shown below.

It can clearly be seen that even if we have identical cut-off scores, most admitted black students will be concentrated just above the cut-off, whereas white admittees will be more equally spread out above the cut-off. So even with group-blind admissions, we’d have that black students would perform worse than white students on average (to the extent that high school scores are predictive of performance in college).

In fact, if one wants to equalize the mean ACT scores of admittees, some discrimination *against* blacks would be necessary (a clearly preposterous idea).

So the first interesting and relatively straightforward result of the paper is that group-blind admissions do not eliminate the black-white gap in college GPAs. This is because the gap results from a fundamental inequality in high school performance, which cannot be eliminated by college admissions policies unless one is willing to discriminate against blacks.

This result shows that if schools merely wanted to reduce the black-white college GPA gap, then group-blind admissions wouldn’t be an optimal policy. What would be optimal is discrimination against disadvantaged groups, which is clearly not happening. Therefore, – assuming that schools follow admissions policies that are optimal given their preferences – it must be the case that schools want to achieve more than just minimizing the black-white college GPA gap when they make admissions decisions.

The authors next examine what if schools want to minimize the black-white college GPA gap *and* have the cut-off scores of various groups as close to each other as possible (i.e. equal treatment). This case yields a similar result to the previous one: that some discrimination *against* disadvantaged students (in the form of a higher cut-off score) would be optimal.

The authors thus conclude that there must be some third factor schools care about besides (1) the desire to eliminate college GPA gaps among groups and (2) the desire to treat applicants equally. They look at the situation in which schools care about equal representation. That is, schools like if the racial composition of their incoming freshman class is similar to that of the general population.

In this case, group-blind admissions can be optimal under a restriction that is unlikely to hold in real life. Namely, because this restriction requires that some schools still prefer to have higher cut-offs for disadvantaged students. Given that schools care about equal representation (in addition to the two other factors mentioned above), it seems that some amount of affirmative action is often optimal.

As far as I can tell, the whole message of the paper rests heavily on the implicit assumption that optimal policies under certain preferences are always followed. For instance, one of the main points of the paper is that group-blind admissions are not optimal if one’s goal is to reduce the college GPA gap, what’s optimal is discrimination against disadvantaged students. Clearly, in our society such discrimination is unacceptable (and rightly so). But the goal of reducing the college GPA gap is still a reasonable one. So we could think of group-blind admissions as a policy that gets us as close to eliminating this gap as possible while still obeying the institutional/legal/moral/political constraints of our society (namely, that we cannot have higher cut-offs for disadvantaged students). As far as I understand the paper, if we add this simple societal constraint to the model, we would have group-blind admissions as an optimal policy, because it’s still better than any amount of affirmative action in reducing the college GPA gap.

Perhaps if we want equal representation to some extent, we need some affirmative action. However, the point the authors are trying to make is merely that group-blind admissions advocates shouldn’t use the argument that “group-blind admissions decrease the college GPA gap” to support their preferred policy. The problem is that the paper doesn’t convince me of this. Group-blind admissions do make the gap narrower relative to affirmative action.

The only way in which affirmative action is an optimal outcome in this paper is if we consider equal representation (or call it diversity if you will) desirable and a goal in and of itself. And even in this case, depending on the exact “weights” of one’s preferences for diversity, equal treatment and reducing the college GPA gap, it may be that group-blind admissions (or discrimination against disadvantaged students) are optimal.

And further, if we consider discrimination against disadvantaged students as socially unacceptable, then any time the optimal policy involves some of this discrimination, the actual policy should be the one which gets us as close to the outcome of the optimal policy as possible without discriminating against disadvantaged students. And in all cases, this is group-blind admissions.

Now, I’m not really making judgments here as to whether group-blind admissions or affirmative action is better in general (for some research on that, see this post). But what the authors are trying to convince the reader of is that group-blind admissions advocates shouldn’t argue that group-blind admissions decreases the college GPA gap. They failed at convincing me of this. What I can see is that group-blind admissions is definitely not the optimal policy if we want to reduce or eliminate the gap, but it is still (1) better than affirmative action, and (2) the best policy if we eliminate socially unacceptable ones.