Phoebe has a post commenting on Flavia of Ferule and Fescule's reaction to a New York Times letter to the editor regarding socioeconomic affirmative action in college admissions.
A few thoughts:
Yes, SAT scores do rise with family income. That said, even for students coming from relatively well-off families, average SAT scores are not high enough to put one in the running at most very selective schools. The letter writer's scores (1300 out of a possible 1600) do put her solidly above average for test takers from any family income category. According to the data shown at the link, test takers coming from families earning over $200,000 averaged 563 verbal and 579 math. So our writer's score looks about 150 points higher than average for the very wealthiest students about whom the College Board has data. She may well be slightly less well-off than that -- it's entirely conceivable that a family with an income in, say, the $120,000 to $140,000 range could pay full freight at Bryn Mawr and spare a daughter from having to work at an after-school job -- in which case her numbers look slightly better compared to her socioeconomic peer group. Though I am not really in a position to tell people how they should feel about standardized test scores that are a few years old, on balance, I'd say there's no reason why she should be ashamed of her scores.
Second, the idea that a plucky kid from a poor background with lackluster test scores will blossom academically once given elite educational opportunities seems intuitively appealing. She had to work much harder than her privileged peer, the argument goes, and that work ethic will serve her doubly well once she actually gets to an elite institution. Maybe -- I can certainly cite anecdotes about college and law school classmates who fit this profile -- but the hard data is far from clear that these case are typical. Economist Greg Mankiw has a post here indicating that wealthier students actually do better grade-wise than their poorer counterparts, controlling for standardized test scores and other credentials. Though I have not read the Kahlenberg or some of the other studies cited at p. 8 (fns 14-16) of this paper, it appears that they also support the claim that the SAT doesn't systematically underpredict poor students's future performance.
I suppose that one might still claim that students from poor families are still likely to overperform their scores at some point after college. They'll become more successful alumni, even if they weren't more successful students. Maybe, but there are several problems here. One, it's difficult to measure what success beyond college looks like. It involves making a host of complicated value judgments that I'm not sure it's appropriate for admissions officers to try to make. I mean, does an elite school make a mistake by admitting a kid with a 1600 SAT who gets a 3.9 GPA in physics and then decides that he really wants to drop out of the rat race to start an organic farm? What about a woman with the same stats who decides that she wants to be a stay-at-home mom? Isn't it simpler and far more attractive to just have admissions officers focus on the narrower goal of picking people who will be good college students? Two, I'm also not sure that basing admissions on likelihood of success as an adult actually helps the socioeconomically disadvantaged. The George W. Bushes of the world do fantastically well for themselves because of their money and connections. An admission policy that sought to maximize the number of successful alumni would probably wind up bringing more of these types into elite universities, not fewer.
Finally, focusing this debate on preferences worth about fifty points may be a red herring. As the Schwarzschild paper I linked to above suggests, many champions of socioeconomic preferences favor them at least in part as an alternative to racial preferences. (Indeed, many of the commenters at the Ferule and Fescue post seem to muddle the two issues, though there are distinctive arguments for and against each.) Others perhaps see class preferences as a way of making sure that admissions officers treat poor students from all racial and ethnic groups fairly. It's thus worth noting that racial preferences are much larger than that at many universities;
this study cites a 190-point gap between the median SAT scores for black and white students at the University of Michigan undergrad and a 240 point gap between median black and Asian scores. I can't find a version of this chart online, but numbers presented in Stephan and Abigail Thernstrom's America in Black and White look similar. So if a university starts giving out 50-point socioeconomic preferences in lieu of the current scheme of race preferences, it will have to turn away many racial and ethnic minority students who would currently be accepted. On the other hand, if a university used both racial and socioeconomic affirmative action, many would find themselves in the strange position of giving a greater degree of preference to wealthy racial minorities than to poor students of other racial backgrounds. So I suspect that universities are unlikely to adopt a system of class preferences as limited as the one described in Leonhardt's article. And with larger preferences come larger trade-offs.