Re: It's the HIP-Hop

[Monty Neill <Mneillft@AOL.COM>]

Regarding the predictive validity across the country of the SAT, it typically
a) is of substantially less value than Grade Point Average (GPA); b)
"explains" less than 20 percent of the variance in first year college grades.
The predictive value varies college to college, approaching zero in some
cases. For four-year GPA or for graduation, the predictive capacity is
substantially less, although both of these are far less studied, since ETS and
ACT say the point of their tests is to predict first year college grades.
(BTW, first year courses are the most likely to focus on memorization and
regurgitation of masses of data, hence are closer to what these tests
measure.)The data is also very clear that the SAT underpredicts for women and
overpredicts for men, which is to say it is biased.

What is not explained by this is how little the SAT often adds to
universities' decisions as to whom they should accept, as Crouse and Trusheim
point out (I think it was them). Not using the test might mean inclusion of
more people from groups which average lower scores (e..g, blacks, latinos,
low-income) are likely to be excluded on the basis of the test, but who will
do well. FairTest's report "Test Scores Do Not Equal Merit" examines 5
colleges and university systems which are wholly or in part "test score
optional." They report more diversity in their student body, more applicants
of color or from low-income families, and the quality of academic work has not
changed. That is, the SAT and ACT keep out students who could succeed,
discourage others from applying.

On test prep: we think it is clear that it works (though again there is
variance: it works a lot for some, not at all for others). Since it is
expensive, once again the well-to-do have an advantage. Also, to the extent it
works, it should lower the predictive value of the SAT.

Monty Neill
FairTest

--------------------------------------------------------------------------
To unsubscribe from the ARN-L list, send command SIGNOFF ARN-L
to LISTSERV@LISTS.CUA.EDU.