[FoRK] Irregularities, or the lack thereof

J. Andrew Rogers andrew at ceruleansystems.com
Wed Nov 3 12:06:31 PST 2004


>  From 3% to 99% of precincts reporting, I never saw CNN waver from 51% 
> Bush and 48% Kerry.  This seems extremely unlikely.
> 
> Virginia was for a while called too close to call, a complete surprise 
> as noted, then 30 minutes later it was magically in the proper range.


This mostly has to do with the order in which precincts are tallied.  I
didn't see any real surprises.


 
> There are a lot of comments about exit polls being so far off.  What if 
> they weren't?


The question isn't why the exit polls were so far off, but why such
badly broken samples were leaked to the world at large.  The internals
of the exit polls in question were significantly non-representative of
the population of voters, and the results reflected that without being
corrected for this fact.

That caught my attention as well, but other folks (on the Internet, not
in the media) almost immediately pointed out that the internals were
ridiculous, and when normalized to a more representative model, showed
Bush in the +1-3 range which turned out to be the actual case.

So the exit polls were correct, sort of.  I saw normalized versions of
those exit polls in less than an hour of them being released that showed
Bush with a small but comfortable lead.

Again, the more interesting question is why clearly broken exit poll
numbers were released, since that isn't the kind of error made by accident.

j. andrew rogers



More information about the FoRK mailing list