Posted by Jed Lewison on Fri Jul 11, 2008 at 9:00 PM Pacific

Polling the tracking polls

I haven't focused much on polling, in part because others are far more able at analyzing polls than I am, and in part because I'm not fully convinced looking at polls is a useful exercise at this point in the campaign.

Case in point -- the new Newsweek poll which is being sold as a huge 12 point swing in McCain's favor. When it showed him leading by 15, I sounded a note of caution, so I can feel good about saying that anyone who thinks that Barack has suddenly dropped 12 points should stop worrying. It's just random -- he was almost certainly never leading by 15.

Even though I think it's silly to look at individual polls in isolation, it is useful to take stock of things using a broad basket of polls. To that end, here is a chart of the daily tracking poll averages going back to early March.

It's interesting to note that Barack is now doing better than McCain's best average, and McCain is doing worse than Barack 's worst.

Barack has also had a slightly wider range -- 4.1%, from 43.5% to 47.6% -- than McCain -- 3.5%, from 43.3% to 46.0%.

So what do these numbers mean? Other than that Barack is now leading, I wish I could tell you with any degree of certainty. As folks like Nate Silver have pointed out, June leads don't mean November victories.

One thing that I have seen from other looks at the data is that this election year has had far more variability than 2004. In other words, Barack and McCain have had bigger swings in their numbers than Bush and Kerry. One theory is that the primary is the reason for that; I tend to believe that theory is correct. But it could just as easily be a fickle public, or an aggressive media, or a combination.

I've seen some indication that the numbers are settling down into a more typical distribution, but I don't think we'll know for a few more weeks, perhaps not until November.

Whatever the case may be, I'm certainly happy that we're starting the final four months with these kinds of numbers.

On the bigger picture, my basic view is that if this campaign is a referendum on the last 8 years (as it should be) Barack Obama is guaranteed victory. If it isn't, then it'll be a much closer campaign. The question is which of those two scenarios it'll end up being, and the answer won't have anything to do with polls -- it'll be determined by a combination of campaign strategy, the media, and grassroots organizers.

Note on methodology: As you can see, to build this chart, I took the Gallup and Rasmussen daily tracking polls from March 6 to July 10 and divided them into 7 different non-overlapping groups, making sure that no days were double counted (i.e., I'd use a poll from 8/8-8/11 and 8/12-8/16, but not 8/8-8/11 and 8/10-8/14). I then calculated an average for each group (the number of days in each vary slightly thanks to differences in the way Gallup and Rasmussen put their polls out), making sure that the each firm's polls were weighted equally. The result is not surprising: Barack Obama is leading, and the growth in his lead has leveled off -- but he's holding it, at least for now.