Sorry to be joining this conversation late. My impression is that Lani Guinier is an advocate much more than a scholar.
There are lots of ways that Harvard and Yale et al. advantage the already advantaged. [If I’m a class warrior, I’m doing it pretty badly as I’d be attacking my class here]. The Z-list, admission preference for legacies, the preferences for offspring of celebrities, hedgies, and politicians as well for lacrosse and squash players, etc. Giving weight to interesting things a kid has done that required the bank of Mom and Dad. Guinier’s assertion that AA has led to an upsurge in Harvard or elite admission of upper middle class African-Americans and immigrants rather than native US-born inner city kids is supported by a couple studies (one may be from her with colleagues).
Incidentally, as someone pointed out, the claim that SAT scores don’t predict grades may reflect a methodological problem. The low correlation between SAT scores and grades may well have something to do with schools’ own selection mechanisms. Not only is there likely to be a restricted range, as pointed out earlier, but kids with lower scores who are admitted may well have that special je ne sais quoi that got them selected in the first place. It would be interesting to look at schools in a country like Canada where a much broader range of students go to the same schools (alas, they don’t use SATs except for Americans) or maybe even a state where almost everyone goes to the same school and college admissions themselves don’t bias the results.
Whether SATs advantage the rich seems more complex. These are complex issues that can’t be dealt with well, except in something much longer than a long post, so I apologize and hope I don’t say something offensive because I am not putting in all the qualifications and nuances.
The tests measure a certain set of problem-solving skills. Those skills probably are helpful in certain kinds of tasks and classes but not others. Posters who are corporate recruiters (Blossom perhaps) say that doing well on the math SATs is a prerequisite for being interviewed for many finance jobs and likely management consulting jobs. Those skills probably aren’t useful in some other adult tasks (e.g., I doubt SAT scores predict who will be a great painter and maybe not even who would be a good psychotherapist).
In general, kids with higher SES get higher SAT scores, though as @JHS points out, not necessarily at the very top of the income distribution. Higher SES families contribute to this by reading to their kids early on (even in the womb), living in towns with good school districts, sending their kids to private schools, hiring test prep firms, etc. These environmental contributions clearly make a difference. It would be a mistake to assume that just because Daddy and Mommy have a higher income, Kiddo will have higher SAT scores without these environmental inputs, as some people seem to suggest. The higher SAT scores arise in part because of the enriched environment, which helps with test scores but other aspects of academic performance as well (and perhaps also to life outcomes).
There may aslo be another direction to the causation. Higher SAT scores seem to be correlated with life outcomes. In a study of kids who placed in the top 1% (of IQ or SAT tests) at age 13, those with higher scores did better at age 33 on a number of metrics (see http://www.businessinsider.com/this-chart-proves-just-how-much-sat-scores-predict-future-success-2012-5) including income. On average, if you looked at the population as a whole, controlling for parental income, I don’t think it would be surprising to find that on average, higher SATs would be correlated with higher income or other life outcomes, even controlling for parental income. That is, how surprising would it be to find out that, controlling for parental income, kids with 1500 SATs in general had higher incomes than kids with 800 SATs?
Just as there are genetic and environmental contributions to IQ, there are likely to be genetic and environmental contributions to SAT scores. If so, having higher SAT parents is likely to help kids both genetically and, if the parents perform better in terms of income as suggested by the study quoted in the link above), with environmental inputs as well.
The racial / ethnic issues are complex. “Scientists” used to say that immigrants had low IQs because when they took them IQ tests, they performed poorly, ignoring the fact that many immigrants didn’t speak much English and the tests were in English. Beyond the language divide, some groups are much more likely to come from highly non-enriched environments. Comparisons across groups may be hard. Just giving free test prep (as the College Board has vowed to do) will not equalize the environmental inputs. This is consistent with Blossom’s age 5 hypothesis. That is why, I suspect, Harvard ends up with immigrant and upper middle-class African Americans making up something like 80% of their African American student body – their parents have been giving the environmental inputs. If AdComs have a lower SAT threshold for URMs as the Espenshade (?) paper suggests that they do, the current use of the SATs may not discriminate against African Americans (unless the threshold is not set well). If so, getting rid of SATs may make it harder to find URMs who, despite weaker educational background, will be able to perform in the elite academic environments. But this is complex stuff.