I think it's worse than that. I heard a talk given by a Black ETS
researcher about this once a number of years ago. Here's what I was told
at the time. The situation may have changed and I'd be interested to hear
if anyone knows that it has.
There are gender identifiable questions and racially identifyable
questions. Those are questions that a given gender or race, as a group,
answer in a particular way (e.g. right or wrong). Remember that all the
questions asked on the SAT are not used for scoring, there's a percentage
(I don't remember how much, and it's probably different now.) of the
questions every time that are part of the ETS research effort.
My understanding of the process is that the test designers use some of
those gender and race identifyable questions to make up the actual SAT
question pool, based on what they believe the population distribution on
the SAT should be. So if they put in a number of questions that females
don't do well on in math -- if they assume that's what the population is
like -- they'll get those results. If there are language questions that
males do better on than females using those would help bring up the male
It would be nice to assume that race and gender identifyable questions were
not used, but I don't believe that's the case.
This archive was generated by hypermail 2.1.2 : Fri Apr 12 2002 - 15:15:35 EDT