pudge's Journal: Electoral Polls

I was gone much of last week, will be gone some of next week. It's that summer thang. I may post things as they occur to me, if they are interesting enough, but very little is interesting to me right now.

However, Electoral Vote Predictor 2004 is quite interesting to me. I was looking for this, and Murr provided it. It's by a Bush-hater apparently, but his methodology seems sound. He collects latest polls from across the country and takes the most recent major poll from each state and puts the results up, categorizing them as strongly for a candidate (>= 10%), weakly for a candidate (5-9%), or barely for a candidate (5%).

These numbers are the most important, not the big numbers at the top, which proclaim Kerry is ahead 328-210, because most of the barelys are well within the margin of error, and are essentially statistical ties. Throwing those out, you get Kerry 231 - 199, with 108 tossups. Of course, the weak ones are still subject to change too, which is why this is updated regularly.

electionprediction.org

[turg]

Random tangent -- electionprediction.org [electionprediction.org] has a very good record but they haven't tried an American-style election yet (just parliamentary elections). In the recent Canadian election they were much much closer than any other prediction I heard (all of which were pretty much completely wrong) -- 268 out of 306 seats predicted correctly.

Re:electionprediction.org

[pudge]

Neat. I just want to clarify that the one I linked to is only listing actual major poll data, not doing subjective analysis (as best I can tell).

Re:electionprediction.org

[demachina]

I can see someone projecting Canadian elections but how would you go about projecting the likely results of a U.S. elections. Ya see Canadians use pretty reliable paper ballots so their elections probably track polls pretty well.

In the U.S. on the other hand we are using a dizzying array of voting machines many of which are child's play to tamper with. With so many close states a nefarious party wouldn't have to flip more than maybe 100,000 evotes to swing the election.

At this point someone will lob the

Re:electionprediction.org

[pudge]

I can see someone projecting Canadian elections but how would you go about projecting the likely results of a U.S. elections. Ya see Canadians use pretty reliable paper ballots so their elections probably track polls pretty well.

Then how come the liberals picked up more wins than was projected a few months ago?

If so I just have to point you to the Georgia elections of 2002 where Republicans won the governship and the senate seat despite the fact they trailed in the polls right before the election.

That

Fun link

[gmhowell]

Fun link. I read it before reading your comments and came to much the same conclusion: tons of wiggle room, and lots of time left before election day.

heh

[jamie]

I guess "Bush-hater" is a term of art now. Kinda like "backdoor draft"? :)

Re:heh

[pudge]

I've used it for quite awhile, primarily to mean someone who really does hate Bush, rather than someone who merely disagrees with his policies. That he links to a site which claims to prove Bush went AWOL clearly puts him in that category, just as I call someone a Clinton-hater if they dwell on whether or not Clinton was involved in various alleged crimes.

As to the draft, we've been over it: "draft" implies forcing someone into military service against their will, and every one of these soldiers volunteer

electionprojection.com

[Twirlip of the Mists]

I prefer Electionprojection [electionprojection.com], but only because I've been using it longer.

Electionprojection has the count at 211-327 right now, but his last blog entry makes the point that that's not accurate because he didn't get the 8/1 Gallup numbers in time to roll into this week's update. Applying those numbers puts Ohio back in the Bush column, but by a margin so narrow it's meaningless. According to those numbers, rush-hour traffic in Dayton could change the outcome of the election.