Horse racery

[mmcirvin]

Ruy Texeira puts the best face on Kerry's poll slippage over the past month. He's an eternal optimist, but I do think that this is something Kerry's campaign must have known would happen, since they decided (probably for good political reasons) to take the hit of not delaying Kerry's nomination acceptance, and therefore laboring under much tighter spending restrictions than the Bush campaign has had over the past month. (I suspect that this has actually had a bigger effect than the Swift Boat Veterans business.) After the Republican convention, Bush will be subject to the same restrictions, and the playing field will be more level. I actually thought it would be much worse, since I thought the Boston convention would be a PR disaster (it was much less disruptive to the city than I expected) and I didn't expect Kerry to achieve the bounce immediately post-DNC that he got.

Electoral Expectations still shows Kerry ahead, though fading; electoral-vote.com, which uses a different aggregation method and is more sensitive to recent fluctuations, shows Bush ahead. In national opinion terms this is a very small shift, but, as always in US presidential elections, small shifts near 50% translate to large electoral gains and losses. Meanwhile, Bush has actually experienced his first-ever sustained up-trend in job approval ratings. It ain't much—it's barely visible against the noise and poll spread—but this is a game of inches.

I would request of fellow liberals inclined to despairing jeremiads to consider that this small and reversible gain is all that the incumbent president has to show for a month of lopsided campaign spending restrictions working in his favor, combined with a smear campaign of exquisite nastiness aimed at his opponent, at the opening of his party's national convention. It will get a bit worse before it can get better; but there is also abundant evidence that Kerry knows how to run hard in the home stretch.

Comments

[ronebofh.livejournal.com]

Note that the gains Bush has at electoral-vote.com are the result of "seven new presidential polls from Strategic Vision today... However, it is worth noting that pollsters aren't always neutral and within the law always do what the client asks for. Strategic Vision normally works only for Republicans and results like these that contradict results from other pollsters in so many states do raise some eyebrows."

That caused me to look at the poll results in the close states. It's a mess. I'm probably going to stop looking at electoral-vote.com as a result, because i don't like the way they display their results... it's too inconsistent.

Also, Kerry is still a douchebag. He started off OK on Daily Show but then slowly slipped back into dumbass campaign mode. Feh.

[ronebofh.livejournal.com]

Also, Dave Leip's predictions.

[mmcirvin.livejournal.com]

Leip's predictions are just the aggregate of the predictions everyone enters into his site. The default for his Web thingy is the state-by-state 2000 results, so, since a lot of people are more or less blundering around randomly (or posting all-D or all-R landslide predictions, absurd in today's political climate), what comes out looks a lot like the state-by-state 2000 results. Except that New Hampshire has swung. And, in an interesting mathematical phenomenon, the average electoral balance and the electoral balance of the averages come out different.

[ronebofh.livejournal.com]

Yeah, i noticed that a little while before you sent this. D'oh.

[pootrootbeer.livejournal.com]

I would agree that electoral-vote.com isn't as solidly methodical as I'd like it to be... but that may be due to the fact that there's really no reliable methodology available.

I mean, four states plus D.C. haven't even been polled ONCE this election cycle. It's probably safe to assume that Montana will go to the GOP in 2004 like they did in 2000, but shouldn't someone at least try to get confirmation of that?

[ronebofh.livejournal.com]

... but that may be due to the fact that there's really no reliable methodology available.

Well, there IS that.

shouldn't someone at least try to get confirmation of that?

I'm guessing none of the pollsters have bothered. It might not be worth their while. Then again, i wonder what pollsters get for their effort.

[iayork.livejournal.com]

Almost all of the interesting states are already within the margin of error, either directly or in aggregate (if each poll has a 5% margin of error, and there are 10 states of particular interest, what are the odds that (at least) one of the polls is more than 5% off?) (I know how to work out the answer, but not before my morning coffee and shower.) And that ignores the other difficulties with the polls: How many of these people are going to vote at all, for example? Even if the poll claims to be of people who plan to vote, probably not all do -- does that aplit evenly?

What we have is a dead heat, really. But the promising thing (to me, anyway) is that Bush's numbers are consistently going down, with occasional blips from outside events (the Iraw war, Saddam's capture, the Swift Boat liars, the convention). The blips so far have all regressed to the trend, with each new blip regressing more rapidly.

That's what America really thinks of Bush, and America just has to remember that.

Frustrating as it is to see Kerry lying low right now, it's probably pointless for him to try to overwhelm the convention (he doesn't have Tom Ridge to announce a new terror threat). Wait until the platform is set in stone, and then take shots at it.

I'm still optimistic.

[mmcirvin.livejournal.com]

Yeah, that's the interesting thing about the Electoral Expectations site; it actually tries to be reasonable about states within the margin of error by parceling out the votes probabilistically, assuming (I think) a normal distribution with 95% level (just under two standard deviations) at the MOE, instead of just awarding all the votes to the guy on top. Their assumption is that some of those polls will be off, but we don't know which ones. The flaw, which they acknowledge, being that this method doesn't take into account the (usually considerably larger) systematic errors, vote turnout, etc.

The most polls usually do to try to take turnout into account is to get separate numbers for registered voters and for self-described "likely voters". As, I think, electoral-vote.com pointed out, one thing that very partisan poll-aggregation web sites sometimes do (OK, I'll be more specific: something they've seen Republican partisan poll-aggregation sites do) is pick whether to use the RV or LV number for every poll based on which one sounds better.