Wednesday, September 17, 2008

Making a Mistake and Not Admitting it

"The mismatch theory may be mistaken. But suppose it were found to be valid? That wouldn't necessarily lead to the abolition of racial preferences. Another result might be the strengthening of mentorship and other programs to help less-well-prepared students achieve at higher levels."

The quote is from a recent LA Times editorial attacking the refusal of the state bar of California to allow Richard Sander, a UCLA professor, access to data that he wants to use to test his thesis that affirmative action actually hurts minority law students.

Sanders' argument, which he has supported with such data as he could get, is that affirmative actions puts minority students into the wrong schools. A student who would have done fine at a second tier school, grouped with other students of his own ability, is accepted by Stanford instead, takes classes aimed at and populated by students abler and/or better prepared than he is, fails to learn, fails the bar exam, and ends up having wasted three years and a good deal of money. It is the same argument that Thomas Sowell made long ago in the context of college admissions, pointing out that black students at MIT were much better at the relevant subjects than the average student, black or white, but much worse than the average MIT student.

What struck me about the passage I quoted above was the implicit assumption that if the result of learning that Sanders was right was the abolition of affirmative action in law schools, that would be a bad thing--that in that case the refusal to provide Sanders the data he needs might be justified. If Sanders is right, affirmative action in law schools is a mistake, hurting the very people it is supposed to help, and abolishing it would be a net gain.

Perhaps there are better solutions, as the editorial suggests. But whether or not there are better solutions, it is worth discovering the truth and acting on it. The only serious argument against doing so is that discovering the truth might mean discovering that a lot of people have done a lot of damage while claiming to do good. That would make those people, some of whom no doubt have influence in the California bar, unhappy.

Which seems the most likely explanation of the refusal to release the information.

When I was little, one of my father's pieces of advice was that making a mistake and not admitting it is only hurting yourself twice.

8 comments:

Anonymous said...

One of the main arguments given for affirmative action is essentially peer effects: Seeing women and blacks attending and teaching at MIT or law school is intended to change attitudes (of both minorities and nonminorities) making their presence feel normal and thereby (it is hoped) reducing future discrimination and inspiring the next generation of minorities.

In judging the success or failure of affirmative action, it would be interesting to try and measure these peer effects on attitudes, and the extent to which a higher failure rate impacts these peer effects.

I believe that some years ago, MIT engaged in affirmative action to increase female enrollment, altering the student body from about 90% men to nearly 50-50 in relatively few years. I don't know of any evidence of a corresponding increase in failures or drop outs among women. Is there some essential difference in how affirmative action has been implemented for race versus gender?

Mike Huben said...

Implicit in David's argument is a baseline assumption that before affirmative action, results were optimal. He then shows how affirmative action could reduce from optimality. Much more likely though (IMHO) is the possibility that results were extremely suboptimal because of discrimination. Then his argument becomes that affirmative action overcorrects.

Does it overcorrect? Does it overcorrect worse than the original discrimination? Is there a moving target, ie. are the minority candidates changing in quality? You'd have to answer all these questions to decide that the current situation could be improved.

William Newman said...

"Implicit in David's argument is a baseline assumption that before affirmative action, results were optimal."

I'm not getting it. Let's say that a reader took away the conclusion conclusion "going to race-blind admissions standards instead of affirmative action would have led to more minorities passing the bar [in the regions and times covered by Sanders' data, all other things held constant, and assuming Sanders' analysis is valid]." Would the reader be wrong to draw that conclusion? Does that conclusion depend on an assumption that before affirmative action, the policies were optimal?

William Newman said...

(Oops, sorry. Actually, my "led to more minorities passing the bar" doesn't seem to follow from the post here. I should have reread the post before commenting. David Friedman's report of the work doesn't seem to make such a strong claim. I think I was remembering that claim from an older discussion, elsewhere, of a similar study.)

John Fast said...

Implicit in Mike Huben's argument is the assumption that the only alternative to affirmative action is racial discrimination against African-Americans. He is ignoring the alternative of "color-blind admissions."

Or else he's claiming that a "color-blind admissions" policy is itself racially discriminatory.

Am I misunderstanding something about your argument here, Mike?

Mike Huben said...

Yes, John, you are missing something.

Color-blind can mean selecting close to 100% whites. Just choose admissions criteria that correlate highly with whites, but don't actually ask for color. There are lots of them: high income, certain ranges of standardized test results, etc.

I have no problem with using a proxy method such as "color blind admissions", providing that it is validated as giving equal opportunity to all groups, so that no one group is grossly over or under represented.

David Friedman said...

"I have no problem with using a proxy method such as "color blind admissions", providing that it is validated as giving equal opportunity to all groups, so that no one group is grossly over or under represented."

That assumes that there are no substantial differences among groups that are relevant. If that isn't obvious, consider that one group consists of people fifteen and under.

John Fast said...

"Color-blind can mean selecting close to 100% whites. Just choose admissions criteria that correlate highly with whites, but don't actually ask for color."

You mean the way sports scholarships discriminate by race because they are based on criteria that correlate highly with African-Americans? Because I am strongly in favor of eliminating sports/athletic scholarships completely (or at least demanding that they have to be paid entirely by specific donations from sports teams and/or fans). :-]

"There are lots of them: high income, certain ranges of standardized test results, etc."

Then you are claiming that "color-blind admissions" -- more specifically, admissions based entirely on academic performance (grades and test scores), and whatever else is a good indicator of success in college (like overcoming adversity, such as getting test scores that are average for the entire state but doing so at a very low-performing inner-city school) -- are (according to your definition) racially discriminatory, right?

And that's one of the two alternatives I accused you of, right?

(In case you don't remember, or have difficulty with reading comprehension, it was when I said, "Or else he's claiming that a 'color-blind admissions' policy is itself racially discriminatory.")

"I have no problem with using a proxy method such as "color blind admissions", providing that it is validated as giving equal opportunity to all groups, so that no one group is grossly over or under represented."

I'm not sure exactly what you mean by "giving equal opportunity to all groups" since the phrase "equal opportunity" has many different (and often contradictory) meanings.

If I understand you correctly, you're saying that we shouldn't base college admissions on academic performance because, according to you, some races -- because of their heredity or environment, I'm not sure which you believe it is -- have a harder time achieving academic performance.

Your empirical claim sounds pretty bigoted to me, but I won't bother argue with you about it. I'll let you go hang out with your racist friends at GNXP, VDare, and Half Sigma.

It sounds like you're saying that whatever system we use for admissions should be weighted to make sure that groups are admitted to college roughly proportionally to their share of the total population, rather than purely on merit. Is that what you meant?

(Personally, I don't have any objection to simply setting quotas and admitting the top 20% (or whatever) of students of each race, which would satisfy your criterion if I understand you correctly.)

However, if that's your opinion, then I'll also ask if you believe that grades should also be weighted so that each ethnic group gets a roughly proportional share of each grade, again overriding actual performance.

And, finally, I'll also ask if bar exams, medical school admissions, the Academy Awards, and other (supposedly) performance-based systems, should also be weighted to make sure each race and ethnic group gets a proprotional share, even if it means overriding actual performance.