[WizzleKake]

The problem, however, is that most standardized tests of this type discriminate against non-white, non-male test takers (at least in the United States).

Can anyone explain this? I don't understand the explanation that follows.

[gdubs]

The famous example of this is the so-called "Regatta Question" in which an SAT question assumed a knowledge of Crew -- an elite sport that would have been a rather obscure reference for minorities from the inner-city. [1]

1: http://articles.latimes.com/2003/jul/27/local/me-sat27

[justathrow2k]

Isn't this more an issue of class more than race, doe? Even with the argument that members of a certain race tend to belong to a certain class it doesn't change the fact that its still inherently a class and environment issue.

Like, from your post, I would say that Crew is an obscure reference for anyone living in the inner-city, not just minorities.

[rory]

In this specific instance, it's really excluding all but the elite. I remember learning what crew was for the first time while visiting colleges, and I grew up in middle-class suburbs.

[usea]

The idea is that if something negatively affects protected group(s) disproportionately, then it doesn't matter whether the mechanism by which it does so is via class -> strong race/class correlation. It's still discriminatory.

[justathrow2k]

I don't think anything was arguing that they weren't discriminatory, or in the very least I wasn't trying to argue that. The claim was that it was discriminatory towards non-whites. I'm claiming its discriminatory to anyone regardless of race and instead its more dependent on their income level and environment they grow up in.

I'm a white male but I know that personally I had no clue what Crew was during my high school years. Anecdotally, I went to school in an area where the public school population majority was, well, what we consider the minority when discussing race relations.

[whybroke]

Also, although racism is quite possible, the lack of social mobility may explain why race and social class are closely related (i.e. if you inherit both race and social position we'd expect a society to look like this)

[untog]

That would assume that race and class can be entirely separated. I'd say that they cannot.

[justathrow2k]

But even if they cannot be separated (a conversation for another time) these questions are still difficult for anyone living in the inner city, especially those in the lower income class, not just those who are minorities, wouldn't you agree?

[shamney]

one very old question does not prove anything about the actual cultural bias of the SAT now or then! research on this topic mostly shows that cultural bias is not a serious problem: http://lp.wileypub.com/HandbookPsychology/SampleChapters/Vol...

[gaius]

Only if you stretch the definition of "white man" to include Asians who actually tend to out-perform whites on standardized tests. But of course it is actually nothing to do with race and everything to do with having parents who value education above all else.

[Retric]

Consider, some groups are more willing to guess when given incomplete information and the SAT promotes guessing. If there was a larger penalty for wrong answers other groups would do better. Another issue is retaking the SAT tends to increase scores as you keep your best individual score.

Individually adding ~30 points might mean little, but when you look at large groups of people such small changes become important.

[codexon]

I don't see what's wrong with the guessing mechanism.

If you cannot eliminate any answers, then your average expected score gain will be the same as not answering.

If you can eliminate choices, then your average expected score gain will be proportional to how many answers you've eliminated.

If a group fails to realize this I would say the SAT was successfully in measuring their cognitive abilities in this regard.

[gohrt]

is the SAT about [whatever it's testing], or is it about how good you are at estimating your confidence in your partial-guesses?

[codexon]

The penalty is designed so that you will not gain more points than you deserve if you guess and had no idea which answers were wrong.

If there are 5 choices and you randomly guess, you have 4/5 chances to get -1/4 and 1/5 chance to get 1 point. That means if you had no idea which answer is correct, you will net 0 points over the long run, and this will be the same as leaving the question blank. 4/5 * -1/4 + 1/5 = 0.

If you are able to eliminate 1 choice, you will on average get 1/4th more points by picking a random answer among the remaining choices. If you are able to do that, then you deserve the extra points, because it took more knowledge for you to eliminate it.

Anyone that doesn't realize this deserves to have a lower score, and if you can't eliminate answers then you also deserve a lower score. Bringing in racial and gender biases into this is ridiculous.

[Retric]

It's a completly arbitrary choice that happens to benifit white males. Of course people are going to bring up racial and gender bias issues.

[codexon]

I just showed you it wasn't an arbitrary choice and that it has a mathematical basis.

Someone who has the ability to eliminate at least 1 choice understands the question better than someone who doesn't know what the question is asking at all, and is appropriately rewarded for it.

[shamney]

is there any evidence for this theory or did you just make it up?

[Retric]

From a quick google:

"We find that when no penalty is assessed for a wrong answer, all test-takers answer every question. But, when there is a small penalty for wrong answers and the task is explicitly framed as an SAT, women answer significantly fewer questions than men."

http://freakonomics.com/2011/11/18/sat-strategy-by-gender-me...

Retaking the test: http://philvol.sanford.duke.edu/documents/SAN01-20.pdf

PS: Making an unbiased test is really hard, the SAT comes reasonably it’s not there.

[untog]

The core of it (as I understand it) is that these standardized tests are written by groups of white men. Thus, the "standard" they are working to is, in fact, not all that standard, but is actually a self-replicating formula that makes sure that the next series of people to create the standardized tests are also white men.

[URSpider94]

Pragmatically, this conclusion comes from looking at how various demographics do on the SAT, versus how they perform in college after being accepted. Once you establish that this is true (since performance in college is the "ground truth"), then you can go back and look at reasons WHY this happens.

Put another way, the discrimination is the fact, demonstrated by a disparity in prediction vs. outcome. The reasons why this happens are hypotheses.

[gmarx]

I don't think anyone is going to touch that; I will point out though, one wouldn't expect the SAT to predict college performance because it is used to sort people into less and more competitive schools and within those schools people select their majors (and some majors are more cognitively demanding than others)

[mathattack]

Some of it is the guessing that other people mentioned. Some of it is vocabulary. Since there are correlations between race and wealth, certain words (say Commodore) may be more quickly understood by some than others.