« Days of Ice & Awe | Gene Expression Front Page | Concurrent evidence »
January 26, 2004

Reverse Affirmative Action

An oft-cited benefit of eliminating affirmative action is that the non-Asian/Jewish minority members who are admitted would no longer be suspected to be "affirmative action admits", of lesser expected competence.

I was shocked a few days ago when I realized that eliminating affirmative action does NOT in fact accomplish this!

To illustrate the problem, suppose all undergraduate institutions graduated all applicants with IQ greater than the mean, and no others. Surely, being easily identifiable as a member of a group X with mean IQ one standard deviation below the mean (and, as another simplifying assumption, let's say this subpopulation's IQ distribution has standard deviation identical to that of the majority, and the population as a whole as well) would no longer be disadvantageous as long as you could also demonstrate you were a college graduate?

WRONG. Members of this group X would tend to be "barely good enough", while everyone else would tend to exceed the minimum requirement by a greater margin. Specifically,

~31.7% of graduates in the general population would exceed the mean by at least one standard deviation. Only ~14.4% of group X graduates would. This is a greater than 2x difference.
~4.6% of graduates in the general population would exceed the mean by at least two standard deviations. Only ~0.8% of group X graduates would. This is almost a 6x difference.
In ANY interval above the mean, group X graduates would cluster toward the bottom of the interval more than the graduate population as a whole.

Thus, it would be rational for any employer incapable of measuring IQ perfectly, and with greater use for people with higher IQs than lower IQs, to discriminate against group X. Even if they can estimate IQ, if they can't do so perfectly, knowledge that you're a member of group X gives them additional information that would cause them to adjust the estimate downwards, if they were behaving perfectly rationally. (Griffe discussed this in the past in his post on differential cutoff.)

The corollary: to make race next to worthless with respect to making hiring decisions, colleges would need to APPLY AFFIRMATIVE ACTION IN REVERSE, not just eliminate it! To make the mean IQ of group X graduates match that of the graduate population as a whole, the bottom ~42% of qualified group X applicants would need to be rejected, while all other qualified applicants were graduated. To make the overall IQ distribution of group X graduates emulate that of the overall graduate population, even more qualified group X members would need to be rejected.

In other words, this is essentially an unattainable goal. Barring extensive genetic engineering or the like, the informational content of race and other visible attributes can only be reduced (which elimination of affirmative action would accomplish), not eliminated.

Godless comments:

As "minority" is inaccurate (since there are numerically significant minorities with high achievement statistics), I have made the minor edit of replacing "minority" with "group X" where group X has a trait one SD below that of the population. Note that this substitution is very relevant when it comes to the actual composition of elite institutions like Microsoft and MIT. Consider, for example, the demographics of MIT:

(MIT stats)
8% of MIT's undergrads are foreign
27% of MIT's undergrads are Asian.
20% are black/Hispanic
Hillel counts another 10% as Jewish.
That leaves 35% white gentiles

Subtracting the blacks/Hispanics gives about 80% of the seats assigned on a merit basis. Of the 72% that went to Americans, slightly more than half (37%) went to Jews and Asians, while the rest (35%) went to white gentiles. Again, Jews and Asians only make up 5-6% of the population, while white gentiles make up 69%. In other words, a tiny fraction of the population beat out the largest group in the population for the merit-based seats at arguably the best engineering school in the country (alongside Caltech and Stanford).

Note that this assumes there to be no negative discrimination against Asians and Jews, which is not necessarily a valid assumption. Lowell High School in San Francisco had quotas on the number of Asians for quite a while, and elite universities still intentionally set the bar higher for Asians.

Posted by Dog of Justice at 05:06 PM