Here’s the new Newsweek poll; it shows Obama ahead 44-41. Their last poll, somewhat famously, had shown Obama ahead by 15.
Here’s what I wrote three weeks ago. Just plug in ‘Newsweek’ where you see ‘LA Times / Bloomberg’ and we should be good to go.
I don’t have a big problem with the details of the McCain campaign’s pushback on yesterday’s LA Times / Bloomberg poll, which had shown Barack Obama ahead by 12 points. In contrast to some of the memos that the press was treated to from the desk of Mark Penn, the McCain team’s argument is relatively even-tempered and even-handed, fully acknowledging that their candidate does have a deficit to make up in the polls, if not the double digit margin implied by the LAT.
At the end of the day, what this really boils down to is an academic argument about whether one should weight polls by party identification, which is perhaps the most controversial subject in polling and one without any wrong answers. My take, for what it’s worth, is that weighting by party ID may increase the precision of any one individual poll, but reduce the accuracy, particularly if you are able to look at several different polls at once.
My question, however, is just who is the audience is for this critique? It’s June, and the election is in November. While the overall volume of polling data does give us some hints about what is more or less likely to occur in November, who the hell cares if McCain is down in any one particular poll? If he’s down by 3, it’s because he’s down by 3, and if he’s down by 11, it’s because he’s down by 11. The polling is simply a manifestation of that reality and not its cause.
Moreover, it is not clear to me that this is a spin war worth winning. If the media believes that Barack Obama is ahead by 5 points, then if a new poll comes out next week showing him ahead by 3, it will not get any attention. But if the media believes that Obama is ahead by 12 points, that same poll would create the perception of McCain momentum, and perhaps trigger a couple of days’ worth of bad press for Obama as whatever had been going on over the past couple of days of the campaign would be taken as the cause for his polling decline. It might lead to harsher treatment of Obama’s decision (flip-flop?) on campaign finance, for instance, or if Iran had been the subject of the week, as evidence that Obama wasn’t resonating with voters on foreign policy.
McCain’s campaign is absolutely right that the media ought not to focus too much on any one particular poll. But there are times later on when it’s going to want them to do just that.
EDIT: That said, the differences aren’t completely about party ID, since Obama lost significant ground in this poll among independents.
Here’s something interesting from the cross-tabs, though: 61 percent of Obama’s support is ‘hard’ and 39 percent is ‘soft’. McCain’s numbers are the precise opposite — 39 percent of his support is hard and 61 percent is soft. So we could describe the electorate like this:
27 Hard Obama
17 Soft Obama
15 Totally Undecided
25 Soft McCain
16 Hard McCain
So Obama has roughly an 11-point advantage among hard supporters, which corresponds to the Democrats’ edge in party ID.
About 43 percent if the country are hard supporters for one or the other candidate, while the other 57 percent are up for grabs to some extent or another.
McCain ties Obama when he wins 60 percent of those up-for-grabs voters. This is not quite the same thing as winning 60 percent of undecideds, since some of those people are decided (and decided for Obama) — but their minds aren’t completely made up. So it’s going to be difficult for McCain to get much more than 60 percent of this group. On the other hand, this universe of ‘soft’ supporters probably intrinsically tilts Republican, since across a large number of indicators, a higher percentage of Republicans are dissatisfied with their nominee and their party. So it’s going to be hard for Obama to get much more than 50 percent of this group. When the soft support is split 50/50, Obama leads by 11.
What I think you’re going to see is the national numbers continue to swing around between those two poles — as they have pretty much all year with the exception of Jeremiah Wright v1.0.