I apologize for being pedantic but you gave this guy too much of the
argument.....
> Not too long ago I sat at a table in a faculty lunch room while another
> faculty member explained that in a search with 100 applicants (of which
> 20 are women) if you hire a woman, it is unlikely that you hired the
> best candidate. His argument washat if only 20% of the sample is
> female the probability is much smaller that the best candidate will be
> female than that the best candidate will be male.
This argument is correct as far as it goes...
> He was trying to
> explain why so many searches in math and science hire men. (Nearly 100%
> of the searches I have seen in the last 5 years.)
And this is where it stops being correct. If "bestness" is randomly
distributed among applicants then roughly 20% of the hires should have
been women.
So what is "nearly 100%" and what is "roughly 20%"? If you've seen 10
searches in the last 5 years, 1 hired woman is lower than the 2 you
expect but not too low to reject the idea that hiring is random with
respect to gender (if the pool is 20% women). If you've seen 100
searches, 10 or fewer will be the result under the null hypothesis less
than 2% of the time.
But few universities have hired 100 people into any department in the
last 5 years. What one runs into all the time in academia are small
samples so that a closet (or lunchroom) MCP can always make the argument
that the women in this applicant pool weren't the best. At some point,
the argument that you are trying to hire the best and take no notice of
gender is not credible. That is when the words *discrimination* and
*quota* start to be used.
> Needless to say, I pointed out the ramifications of his argument if
> taken to the next logical point. He had no problem with the implication
> that most women hired in science and math are not the best ones for the
> jobs. He didn't say they weren't good, mind you, but they were probably
> not the best.
I don't see this as the next logical point, unless the number of women
hired is significantly above the expectation. If you accept the
assumption that bestness is not gender linked, your faculty of 20% women
could certainly contain only the best candidates from several
independent searches.
There is also an important and questionable assumption here, that there
is a single definition of best. Some think finding "best" is like using
GRE exams (or priority scores) with higher score as being better. But
NIH panels don't always get it right and neither do search committees.
A better scholar, a better teacher, a better colleague, a better grant
writer? Lots of dimensions here.
Mike Kahn
