Kenmlin wrote:
>
> >PS: Pardon my flippancy in response to a serious inquiry. I'm not
> >really planning to write the SOA exams, either. I just don't understand
> >why an association of supposedly numerate people would do such a
> >damnfool thing as grade on a curve. Surely they realize that the
> >candidates writing on two different dates are not random samples from
> >the same population, and have some idea of the objective criteria that
> >make somebody "actuariable"?
> >.
>
> Well, anyone could take the actuary exams so the exam takers may not be
> "numerate." Unlike federal government, SOA wants to screen out
> and keep only those people at the very top. Nothing wrong with that.
If you define "the very top" to be some fixed percentage - which is my
interpretation of what was described - then yes there is something
wrong.
Not so much in terms of breaching anybody's rights as in terms of loss
of efficiency for the society.
Suppose in one year there is high unemployment, TIME has run an article
about how great it is to be an actuary and how all you need to do is
pass a few exams that anybody's allowed to write, and some popular
sitcom has a positively-presented actuary character. Fifty percent more
people try to write the exam, most of them utterly hopeless. If indeed
it is marked on a curve, people get through who wouldn't have got
through otherwise.
Now a couple years later, the job market is booming, Silicon Valley is
tempting strong math students into CS, Newsweek ran an article on how
hard the SOA exams really are, and the actuary character on the sitcom
has been put in jail for cocaine possession (in the script) because the
vastly popular actor who played him has been hired away. Fewer people
than average write, and a score that would have got you into the top 50%
(or whatever) two years before doesn't come close this year.
Result: the pool of actuaries contains weaker people from the year in
which the median performance was lower, and does not contain some
stronger people who wrote in the year with higher median performance.
Industry, no doubt, will do its best to hire the N strongest candidates
they can, whatever number write the exam. Certification should depend
on an objective criterion, not on either ow many candidates are needed
that year or on how mnay chose to write the exam.
By the same token, grading on a curve is normally evidence that the
instructor either has no idea what constitutes competence in their
subject or is too lazy to set an exam that has been properly
calibrated. Minor adjustment based on results is one thing - for
instance, if it's discovered that an apparently easy question had a trap
in it that the instructor hadn't seen. But out-and-out deciding that "An
A means you're in the top 10% of the class, no matter how good or bad
the other 90% are" is daft.
-Robert Dawson
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
