[dragonwriter]

> Hm... Aren't SATs widely used for college admissions

Yes.

> (not exclusively, of course, but if they were, I'd say the system is totally meritocratic)?

Wait, what? Unless you make that tautologically true by defining the merit you are trying to assess in admission as "SAT scores", I don't see why you would.

> I'm not saying it's the best indication of the kind of merit required to excel at the university, but it's probably one of the best we have

Its not. For predicting college performance, its a very weak predictor of college grades, or even first-year grades (weaker than high school grades, class rank, or even just the high school you attended), and, to the extent its useful, at least one study has indicated that its predictive power is almost entirely explained by the degree to which it serves as a proxy measure for the high school that the student attended, and that within-school variation in SAT scores lacks predictive power.

Also, there is evidence that the relationship between SAT scores (and the same is true of other measures, like GPA) and college performance is not consistent across various axes of demographic variation (race, income, etc.) -- I've cited one analysis on this elsewhere in the thread.

> remember, we're favoring the disadvantaged so that they can improve despite them being worse, not because we believe SATs wrongly assess them as being worse

One of the many reasons for admissions preference for traditionally disadvantage groups is advocated is that many of the measures used as signals to admission disadvantage those from traditionally disadvantaged groups, where the measure reflects disadvantage of circumstance rather than lack of merit.

> Also, in which way (relevant to university admissions or completion) are blacks better than whites/Asians?

I never said they were.

[yummyfajitas]

It's true - black people tend to underperform relative to their SAT.

http://www.mindingthecampus.org/2010/09/the_underperformance...

Your source agrees with this if you scroll down to the data tables. Note the sign on the black and hispanic coefficients.

So your real critique seems to be that SAT is unfair because it gives a non-predictive boost to blacks and hispanics. Consequently, if we wanted to have the most accurate predictor possible, we'd include negative weights for black people. Is this really the critique you are making of using SAT and GPA?

One of the many reasons for admissions preference for traditionally disadvantage groups is advocated is that many of the measures used as signals to admission disadvantage those from traditionally disadvantaged groups...

That's simply incorrect. If it were true, then blacks would overperform relative to SAT rather than underperform.

[dragonwriter]

> It's true - black people tend to underperform relative to their SAT.

Its actually a lot more complicated than that; the degree to which each objective factor predicts grades (and the degree of difference in expected college grades each unit change in the factor represents) differs by race, income, and other factors. (In the case of SAT scores, higher SAT scores mean more in terms of better college grades than they do for whites.)

So the average black student underperforms based on their SAT scores -- but that is less true (and eventually the opposite of the case) for blacks with higher SAT scores.

> So your real critique seems to be that SAT is unfair because it gives a non-predictive boost to blacks and hispanics.

No, my real critique is that the assertion that without AA, existing admissions would be meritocratic is unsubstantiated, and that there is plenty of reason to be skeptical of it.

And I will maintain it until someone presents a coherent definition of the merit sought and evidence that supports the claim based on that definition.

[yummyfajitas]

...that is less true (and eventually the opposite of the case) for blacks with higher SAT scores.

From the cited article: "Blacks with the highest SAT scores and high school GPAs...are those underperforming the most.

Do you have have data showing the opposite? Note that your source on the other thread did linear regression, so it can't possibly capture this effect.

No, my real critique is that the assertion that without AA, existing admissions would be meritocratic is unsubstantiated...

Your sole evidence for this is a source stating that SAT and GPA are good predictors of college performance, but SAT and GPA plus a penalty for black people is even better.

[tomp]

> For predicting college performance, its a very weak predictor of college grades

That's interesting. Naively, I would expect SAT correlated with IQ, and that correlated with success in university. Are there better predictors of college grades?

> where the measure reflects disadvantage of circumstance rather than lack of merit

Hm... I can hardly imagine how a bad score on the math SAT would not indicate that you lack merit in math.

> I never said they were.

Well, then it doesn't much matter which definition of merit (who is better) you take (as you claim in your original response to me)...

[dragonwriter]

> Well, then it doesn't much matter which definition of merit (who is better) you take

Yes, it does.

The definition of merit is necessary to assess whether the current system is, excluding AA, actually meritocratic, as claimed. While it would also be necessary in order to assess whether one race or another was better than another, if that were an interesting question to answer -- the fact that I'm not asserting that one race is better doesn't alter the fact that an actual definition of merit is necessary for the claim that the existing system (but for AA) is meritocratic to even have meaning.