Sunday, January 11, 2015

An Empirical "Justification" for Affirmative Action

Discrimination is ok as long as the results are "good", is that what I'm reading here?
In Texas, students who graduate in the top 10 percent of their high school class automatically have the right to attend any public Texas university, more or less. They still have to take the SAT or ACT, but the score doesn’t matter. This unusual situation meant that by studying this group of students, the researchers could measure how effective those tests are at forecasting future college performance without the muddling effects introduced by a typical college admissions system.

The most surprising result is that for black students, the SAT is a far more important predictor of college GPA than for white or Latino students. (This is after controlling for a host of factors including the choice of college major.)

For white students, scoring one-standard deviation higher on the SAT or ACT (that’s roughly a couple hundred points on the SAT) is associated with a college GPA that’s 0.347 points higher. For black students, higher standardized test performance is associated with an additional 0.152 point boost in college GPA. For Latino students, there was no additional boost.

Even though black students do worse on standardized tests on average, each additional point seems to be more valuable for them. “This speaks to the idea that affirmative action in admissions might empirically be a good idea,” Lincove said.

3 comments:

David-2 said...

Even though black students do worse on standardized tests on average, each additional point seems to be more valuable for them. “This speaks to the idea that affirmative action in admissions might empirically be a good idea,” Lincove said.

Let's see if we can think of an alternative explanation.

According to the paper itself, pg 21, average SATS for their cohort are: White 1218, Hispanic 1061, Black 1029 (the range is 800-1600, of course). (This is the class of 2008-2009, which was studied in the paper.) The College Board has the following statistics (2008, here: 1218 is at the 82nd percentile, and 1029 is at the 52nd percentile.

Let's leave aside the argument (which is perfectly valid and completely obvious) that someone who is accepted to college with poor qualifications (SAT at 52nd percentile) has much less of a chance of succeeding as it is, so any marginal improvement above bare minimum might have a larger impact in proportion.

Instead, let's keep looking at that College Board table.

We see there that bumping a 1030 score up 30 points (SAT scores are in increments of 10) takes you to the 58th percentile, an improvement of 6. Bumping a 1220 score up 30 takes you to the 85th percentile, an improvement of 3.

Hey, guess what?! The SAT scores aren't linearly related to your percentile rank! It is just a basic fact of SAT scores that a 30 point difference at the 1020 level is more significant than a 30 point difference at the 1220 level.

So the conclusion that for black students "each additional point seems to be more valuable for them" is not only trivial but a misinterpretation of what is going on: At the low end of the SAT scoring range, whre blacks that they studied sit, each additional point is more valuable for everyone because SAT scores are nonlinear with respect to rank.

Ellen K said...

I would rather look at the NCAA numbers on student athletes who despite full rides, academic support and no worries about the financial side of college still fail to graduate. Now THAT is something we should be talking about.

maxutils said...

EllenK--the NCAA already is doing this.