[FoRK] polling meta-analysis discussion [was Re: Pitches]

Luis Villa luis at tieguy.org
Tue Oct 14 13:04:03 PDT 2008

On Tue, Oct 14, 2008 at 3:49 PM, Joe Barrera <j3 at barrera.org> wrote:
> On Tue, Oct 14, 2008 at 12:42 PM, Stephen Williams <sdw at lig.net> wrote:
>> And here ya go for balance:
>> Chairman Tom Cole wrote:
>>> The polls continue to show Republicans gaining ground all across America as voters learn the true agenda of Barack Obama and his liberal friends.
> Wow. Is that what the polls say? Talk about a reality-free zone.

Reality appears to have a well-known liberal bias.

Places I've been going to for polling meta-analysis:

http://www.pollster.com/blogs/ [best graphical display of existing
trends I can find; thoughtful discussion of the ins and outs of
polling; maddening refusal to extrapolate from the present to the

http://www.fivethirtyeight.com/ [interesting simulation-based approach
to the problem; draws on historical data to model/extrapolate what
might happen between now and election day; maddening lack of
transparency about the details of the model being used]

http://election.princeton.edu/ [overall maddening lack of detail;
occasional very sharp commentary]

I used to go to Real Clear Politics, but then I read this:

I used to read electoral-vote.com as well, but its poll-averaging
methodology seems fairly primitive compared to pollster and

Does anyone else have any suggestions?

The core problem right now for McCain/the RNC is that in all the
polls, McCain is down at least *ten* points in a near-majority of
electoral college states, and at least 7-8 points in a large number of
other states. At this point, he's going to have to fight to win North
Carolina- not a good sign.

Certainly there are some national polls indicating some tightening,
but they need a *lot* of tightening, and they need it in a lot of
places. Barring McCain very, very effectively landing the Ayers card
tomorrow night, that seems unlikely. Not impossible, but unlikely.


More information about the FoRK mailing list