Is Clark the anti-Dean?

Polls show it - sort of. That is, if the pollsters know about the new voters Dean and Clark are bringing into the system.

The American Research Group is running a tracking poll on the New Hampshire Democratic primary. What ARG call a tracking poll is a system where they poll a third of their sample every day, and every day's poll results include the last three days' worth of responses. So if something happens on day 4, you'll see some reactions on day 5, but you'll only see the full effect on day 7. (Except that stuff may have happened on days 5 and 6, so you're not out of the woods yet.)

Note that they ask different people every time - which are presumed to be representative - so it's not a proper panel poll, so you can't see which people are changing their mind from day to day, just that the overall results are changing.

In 5 days, the following evolution has happened:

 Dec 26-28Dec 31-Jan 2

Why is Kerry dropping? Well, the ARG said at this time that it was because of an anti-war Kerry ad, featuring children playing happily and promising never to go war over oil. This unfortunately reminded women (predominently) that Kerry had voted for the war. Hence the meltdown. I must say, it's pretty tragic that a few days after a candidate announces he's borrowing money and mortgaging his house so he can keep his ads on the air, he goes into freefall because one of his own ads.

Happily, according to today's version of that page, he doesn't have much further to slide. If you compare the total number of people who would vote for a candidate, and the number of people who are "strongly committed" to them, you get the following:

 SupportersStrongly committed% strongly committed

Clark has the most undecided supporters. (Gephardt, Edwards and Kucinich have figures that are too low, and therefore too imprecise, to be included in this study.) This can mean one of two things.

First of all, Clark could be a less compelling candidate, who is nonetheless better than the others. His supporters aren't particularly convinced, and he could lose them as easily as he's gained them. That's one argument.

The other argument is that Clark's supporters are less entrenched because they've recently joined him from other campaigns, or from uncertainty. (There are 18% undecided in the latest poll, and 1% voting for "Other", whoever that is - Lyndon Larouche?)

This matters, because those 18% undecided have to go somewhere. If the ratio of committed to recent converts holds on election day, and the 18% undecided have to choose a candidate (or stay at home), then given the slimness of Kerry's lead, on current standing I think this poll predicts Clark to take 2nd place in New Hampshire.

And that does two things at a stroke: 1) It knocks out John Kerry, and 2) It shows that Wes Clark is the only one of Dean's rivals that can successfully get the votes of people who don't want Dean. If Clark can do it in New Hampshire, and Gephardt, Lieberman and Edwards can't, that gives him immense momentum for the next round of primaries.

Of course, there's a lot of problems with such an analysis. First of all, it assumes that a poll - a snapshot of public opinion now - represents underlying tendencies that will persist until the election. No serious pollster would tell you that; all a poll can tell you is that at a certain time people told someone that they would vote in a certain way.

The other reason is more interesting. Both Dean and Clark have been running highly innovative Internet-based campaigns, and they're attracting a different kind of voter as a result. Young people in particular, and non-voters in general, are being energised to turn out and vote - and it's anyone's guess who they are. Do the pollsters know about this new phenomenon (which hasn't actually been tested in a ballot box yet, of course)? Are the polls we see accurate predictors of the coming elections? It will be instructive to compare the last of the New Hampshire phone polls with the exit polls, especially if the pollsters get things wrong. Because if Dean and Clark are bringing new voters into the system, many of the polls everyone has been relying on may have to be rethought.