Victory probability mapObama lead over time

Saturday, September 25, 2004

16%

I haven't updated because I've been in Milwaukee without Internet. Updates may become rare as I start working 100-hour (or more) weeks until November. Lots of new polling has come out in the past week.

I doubled my "daily volatility parameter" to reflect that this campaign seems more volatile than usual.

In a nutshell, without maps: We have a 16% chance of winning (1% if the election were today) with an expected 233.6 electoral votes (228.0). We have 48.9% of the two-party vote. My prediction (which I didn't get a chance to post) on the 18th was a 13% chance of winning, so this is a slight uptick over that nadir. Speaking of nadirs, my estimate of Ralph Nader's effect is higher than it's ever been this year (something like a 1.4% effect on Kerry's share of the two-party vote).

2 Comments:

Blogger david said...

Are you taking into account the unbalanced nature of most of these polls? It's been shown that most of the heavily-Bush polls have samples that are much more heavily Republican than the actual voting body.

And really, 16% seems far, far too low. Of course, your numbers probably know more than I do, but it just seems so unreal.

September 27, 2004 1:59 AM  
Blogger Benjamin Schak said...

I believe the 16% is realistic, with the following huge caveat: I depend on the reliability of pollsters and their polls. If the average pollster has even a tiny bias one way or the other from reality, that seriously throws off my results. For example, it's possible that polls are failing to factor in the massive voter Democrat-targeted registration efforts mentioned in today's NY Times; or failing to factor in the massive GOTV efforts that I'm a part of. Basically, if you believe polling, you should believe me; but if you don't believe polling, you shouldn't believe me.

A lot of people on both sides have argued for and against weighting poll results by party. I won't get into it now, but I personally am against partisan weighting, but for reweighting by respondents' 2000 votes.

September 27, 2004 4:30 PM  

Post a Comment

<< Home