« Time Vs. Newsweek | Main | Bush's Battleground Blues (Continued) »

New Voters for Kerry

A new Ipsos-AP analysis of their poll data shows new voters leaning very heavily toward Kerry.

Among LVs who are new voters, Kerry is favored over Bush by a smashing 25 points, 60-35. Moreover, these new voters were twice as likely to say they'd been contacted by the Kerry campaign (38 percent) than by the Bush campaign (16 percent).

The Ipsos-AP analysis provides this sketch of new voters' demographics and political attitudes:

New voters tend to be young (64% are under 35), unmarried (54%), with some college experience (36%) and holding down a full-time job (63%), often in the service sector or skilled trades. They say the country is heading in the wrong direction (68%) and disapprove of Bush’s performance as President (63%) and his handling of Iraq (65%), in particular.

Obviously, the more of these voters that show up at the polls on November 2, the better for John Kerry.


IPSOS-AP New Voters vs. Newsweek's New Voters

One of the most interesting-and frustrating-things to have observed in my poll obsession this election is the difference in poll results. Here is a good example. When IPSOS-AP polled new voters, they found Kerry with a 25% lead, but when Newsweek polled this same group, it found a 3% lead. Since I think the "new voters" will decide the election, Id like to know which is right. Heck, Id settle for splitting the difference.

Of course, there might be some explanation for these differing results. For example, Ruy's post referring to the IPSOS-AP results said that the polled group were "likely voters that were new voters" while Newsweek's group was just "new voters." Still, hard to imagine that this difference would explain the wide gap in the two results.

At any rate, I think IPSOS-AP is closer to the mark. I think Dems have done a better job registering new voters. Also, I think it is noteworthy that new voters tend to be younger, a demographic that is strongly pro-Kerry now.

Ultimately, I think Kerry will do about 5 percentage points better then the final poll results because of the undecideds breaking for him and a huge (and pro-Kerry) turnout of new voters, which have been under counted in the polls.

In my mind the best news in this entire entry is that they were more likely to have been contacted by the Kerry campaign than the Bush campaign, by over two to one ((38% to 16%)!!! This indicates that we are kicking the vaunted Rove GOTV Machine's rear!!! Lets keep up the good work people!

The variability in polls of independents is far outside the margin of error. For instance, Time, Zogby (according to this morning's AP article) and WashPost (in an article I recall seeing a couple of days ago) have Bush a dozen points ahead, while Pew has them even in its most recent poll and Newsweek has Kerry 14 points ahead. Greenberg does not give a result for independents, but given their overall result and the party breakdown for respondents, the numbers for independents must look something like Newsweek. The MOE for a sample of 300 independents that split evenly is less than 6%, so Pew is already outside the 95% confidence limit (well outside when you consider that several different polls have large Bush leads) and with Newsweek there is over a 4-sigma divergence. Clearly, we are dealing with a systematic error of some kind and not a sampling error.

I can offer a hypothesis about the cause of this. There seems to be a general tendency for tracking polls, which do not make callbacks and therefore will have to make the largest demographic weighting corrections to make Bush look stronger among independents. The Pew poll, which seems to make extraordinary efforts to call back nonresponders and therefore will require lesser demographic weighting, has Kerry looking fairly good. I don't have information about the Time, Newsweek, and Greenberg -- would appreciate any other posters' comments on that.

As I have pointed out in other postings, demographic weighting undercorrects in at least two respects: (1) "category weighting" causes undersampled low-education Hispanic (and other) immigrants to be replaced by more educated members of the same ethnic groups, and by low-education whites, both groups that trend much less Democratic than the missing low-education immigrants. (2) Education and income are highly correlated, yet in this election they show opposite voting trends. Nearly all polls weight for education and not income. This will cause undersampled low-income, low-education voters who trend Kerry to be replaced by high-income, low-education voters who are extremely pro-Bush.

The greater the degree of bias in the sampling process that needs to be corrected by weighting procedures, the greater the effect of these two mechanisms on the final results. (The magnitude of these effects will also depend on details of the weighting method.) Thus polls that call once would have a greater pro-Bush bias than polls that call back.

This hypothesis is also supported by the detailed Pew results which show a great variability among low-education respondents from poll to poll. This suggests small sample size and large (possibly fluctuating) weighting coefficients.

> In my mind the best news in this entire entry is that
> they were more likely to have been contacted by the
> Kerry campaign than the Bush campaign, by over two
> to one ((38% to 16%)!!! This indicates that we are
> kicking the vaunted Rove GOTV Machine's rear!!!

There seems to be a huge philosophical difference between Republicans and Democrats here. The former is basically focusing their efforts on the already converted, e.g. "Shrub's" campaign events are exclusively attended by adoring GOP supporters -- in most cases voluntary workers for the campaign. Nobody else is let in, so is it any wonder the "Shrub" groupies are far more enthusiastic about their Man than the Kerry supporters are! We are basically comparing apples and oranges here. I also saw an interesting comment on the NRO Corner weblog, where one conservative encouraged his fellow conservatives to contact other like-minded people to vote and/or do local voluntary work for the GOP campaign. I get the impression they are not interested in persuading undecideds or Demos at all, except DISCOURAGING people from voting in general and for Kerry in particular.
In contrast, the entire Kerry/Edwards campaign has been a concentrated effort to woo centrists. Even GOP hecklers are allowed to attend the K/E rallies. As Ruy and others have commented, this _ought_ to be a much more sensible strategy than merely trying to maintain a fiercely partisan 50.0001% majority. Next week, we will know.