09/08/2008 05:43 pm ET Updated May 25, 2011

Bouncing Polls

As is obvious by our traffic, the news over the last 24 hours and the pace of new national polls dropping this afternoon, interest in political poll numbers is right now as high as it gets. Today's "cable catnip" from USA Today/Gallup (as our friends at First Read called it) have everyone pondering all of the usual polling controversies, including likely voter models, party weighting, weekend interviewing and the like. So our cup runneth over.

Here are a few quick thoughts:

  • The post from Charles Franklin this afternoon, if you haven't read it yet, is as good a review anywhere about what we can say about the national trend on the basis of all of the data available. Charles promises updates as more polls become available, so stay tuned.
  • Although most of the new polls out today are based on interviews conducted entirely after "the completion of the GOP convention," (as Gallup put it), we should remember that the Republican convention dominated the news on Friday and, to a lesser extent, continued to do so for much of the weekend. I was as guilty as anyone last week in leaping to analyze polls conducted over the weekend between the two conventions, but I know a few pollsters that would urge us to avoid characterizing public opinion at all in the period when conventions dominate news coverage. We may well see some of the bounce fade as we get new polls with data collected from Sunday night forward.
  • Today's polls clearly show a "bounce" favoring the McCain-Palin ticket, but what does it mean and more important, will it persist? Looking back at the historical data, we see that some convention bounces persist and some do not. The sample size of polling on past conventions seasons is just too small to be able to predict how much of what we are seeing today will persist and for how long.
  • Of the polls out today, one of the more interesting is the just released CBS News survey re-interviewed with 655 registered voters first polled before the conventions in mid-August. This "panel-back" design allows the pollsters to determine which respondents changed their vote preference:

McCain's move ahead of Obama can be traced in part to movement among previously undecided voters. In this survey, CBS News re-interviewed respondents to a CBS News/New York Times poll taken in mid-August. While many previously undecided voters remain undecided, more of those re-interviewed have moved towards McCain than Obama.

Again, all caveats above apply. Like the other surveys, the new CBS poll called from Friday through Sunday evening. Still, if I am reading this paragraph correctly, the CBS pollsters are telling us that most of the "bounce" they measured occurred among undecided voters. Undecided voters tend to be less attentive to news -- as should be obvious -- are more apt to move around in response to events in the news.