Affirmative action policies at universities are being challenged more and more often nowadays. And actually, some states have already banned them. So what lessons have we learned from these bans?
The question is not only whether minority enrollment rates decrease, but also whether many minorities will actually not even bother applying after such a ban, and whether schools are able to maintain their current level diversity.
Yagan (2014) (ungated PDF) takes a look at California’s 1996 affirmative action ban and its effects on law school admissions. He uses a dataset on law school applicants from an elite college from 1990 to 2006. This includes roughly 38,000 applications of 6,000 applicants to 187 law schools.
Predictably, after California’s ban on affirmative action, black admission rates dropped as shown below. (Note: the author focuses on blacks, but a similar story applies to Hispanics as well.) The figure below is for Berkeley, but a similar trend can be observed at UCLA, home to the UC (University of California) system’s other top law school.
There are two main possible mechanisms with different implications, were a nationwide ban to take place.
First, the California ban merely decreased the size of the applicant pool because black applicants decided not to apply to “black-unfriendly” schools. If a nationwide ban were to take place, this option would not be available anymore as there would be no “black-friendly” schools left. Therefore, a nationwide ban’s consequences would probably be small: the applicant pool and hence enrollment rates would not change significantly.
Second, fewer black people applied because their chances of admission decreased after the ban. But if admissions advantages were to decrease, then there would be a reduction in black enrollment rates even without a smaller applicant pool. Furthermore, we would see the same effects in case of a nationwide ban.
So in the first scenario, black enrollment decreases only because blacks do not apply in as high numbers as previously. In the second scenario, it decreases because blacks’ admission advantages are reduced.
To investigate which of the above two options was the case in California, Yagan examines black enrollment rates while holding the applicant pool constant. This means that we assume that the applicant pool did not change after the ban. If this is the case, then clearly the first effect disappears. So if there is any drop in black enrollment rates while holding the applicant pool constant, it is entirely due to the second effect, that is reduced advantages.
But how to check this? Yagan’s idea is quite nice. He estimates a probit model separately for each UC (University of California) school in his sample (which are Berkeley and LA). This model tells us the probability of gaining admission for an applicant with a given GPA, LSAT score and other characteristics. Two other controls include: whether the applicant is black, and if yes then whether the year in question is a pre- or post-ban year.
So with this model, we can estimate the probability of admission for a given applicant both pre- and post-ban. Then Yagan just simply takes the applicant pool from pre-ban years, and estimates their admission chances assuming that we are in a post-ban year. This way, the applicant pool is kept constant at pre-ban levels. So after doing this prediction procedure for all pre-ban applicants, the results are aggregated and black admissions rates can be calculated post-ban with a constant (pre-ban) applicant pool.
What do we see then? It turns out that black admissions rates would have decreased from around 60% to 30% even if the applicant pool were constant. This lends credence to the second mechanism described above: that there are reduced advantages for blacks in admission. But are all the advantages gone?
Using a procedure very similar to the one discussed above, the author estimates the black admission rate we would have seen, had black applicants been subjected to the same requirements as white applicants. The average black admissions rate at UC schools would have been only 8% in this case, as opposed to that 30%. So clearly, schools could retain some advantages in admissions for blacks despite the affirmative action ban.
Then why did schools not keep the original 60% admissions rate? There are two possibilities. Either they could not, because the alternative measures they used to diversify their student body (like low income status) are obviously not perfectly correlated with being black. I.e. if you can only admit people because they’re poor and not because they’re black, then you will mostly admit poor people. As long as not all or most poor people are black, you will not be able to admit as many black people as before. Or, the second possibility is that the schools just simply did not want to or could not (for legal reasons) admit as many black people as before.
The data indicates that UC schools would have been able to maintain a black admission rate of around 63%, though the confidence interval is quite wide (it could be as low as 33%). Anyways, it seems the schools could have maintained a higher advantage and they did not do so because of unwillingness or legal reasons.
There could be two interesting reasons for this.
First, if schools have preferences over diversity and non-racial excellence measures (LSAT, GPA, etc.), then once they can only evaluate race imperfectly (based on low income status or diversity essays), it will be more risky to maintain a high level of diversity. This is because the risk of admitting a below average white student or rejecting an above average black student grows, when schools cannot observe race. Therefore, schools still want diversity, but the opportunity cost of maintaining pre-ban diversity levels is too high for them.
Second, if the courts cannot perceive the precise criteria on which applicants were admitted (a likely scenario), then they may base a noncompliance suit on the black-white admissions gap they observe. I.e. if they see a high enough gap, they may be able to find probable cause for violating the affirmative action ban. To avoid noncompliance therefore, schools reduced their admissions gap to a level that does not raise red flags from a legal point of view.
The three main conclusions of the paper are the following:
- An affirmative action ban will lead to a large reduction in black admissions advantages, but at the same time it will not completely eradicate them. In the current study, the advantages were roughly slashed in half.
- It appears the UC schools in this study did not find any “workable” alternative to race-based admissions that could sustain as high a diversity as affirmative action. This provides support to the constitutionality of affirmative action.
- Finally, to achieve larger black admissions advantages under a ban than what we see in California, some policy is necessary that makes it mandatory for colleges to place a higher weight on variables such as low income or diversity essays which correlate with being black.