www.electoral-vote.com

[zotz]

It's Andrew Tanenbaum. This is probably obvious from the DNS registration details, but I hadn't bothered to wonder who was behind the site. It seems a bit loaded at the moment, unsurprisingly, but the estimate based on most recent polls is 298:231 in Kerry's favour . . . but so many of the states are on the edge that it could easily still go either way. The prediction with the usual split of the (now quite rare) undecided voters is 306:218, obviously with the same caveat attached. It's all still to play for.

Comments

[dennyd.livejournal.com]

Hah. I hadn't thought to check who was behind it either, despite checking the site itself most weekdays...

It swung hard to Kerry once last week, but swung back to a narrower Bush victory again the day after... the number of polls going on now is making the daily predictions a little unstable, I suppose.

[blufive.livejournal.com]

It's been back-and-forth like a wimbledon spectator for the last several weeks. Have a look at the graph...

[zotz]

It's amazing. Enough to make you wonder whether the confidence intervals on those polls are really as tight as they estimate. A lot of them just aren't compatible otherwise.

[zotz]

As the man himself points out, there are probably enough in a couple of states to plot a distribution and work out a probable figure from that.

Mmm. Meta-analysis.

[blufive.livejournal.com]

Most of them are a 2-4% margin of error, and they keep giving 1-2% leads to the candidates - hence the wild oscillations.

There's also been some kerfuffle over exactly how some of the pollsters are weighting their surveys. In particular, some of Gallup's weightings look kinda dodgy.

There are also signs that turnout is going to be way up this time around - which leaves the polls with some real problems. I suspect that US polling organisations are about to suffer the kind of whoopsie that stung their British colleagues in 1992.