Two front pagers talking about the polls yesterday, including Markos himself. Both were correct in their analyses. But I believe two critical aspects weren't addressed. And let me just say, I'm not one of those rose-colored-glasses, everything's-great-for-Obama spinners. Nor am I one of those gloom and doomers ready to leap from the Golden Gate Bridge just because Zogby says McCain's ahead by 5%. Rather, I try to be as objective in my analysis as I can. And the bottom line, from what I can see, is that although the poll averages are tightening some, the race is virtually unmoved from where it was when Obama's lead was proclaimed to be larger than it is now. Here's why...
...okay, the first problem is the polling models. Most professional pollsters will tell you that a "likely voter" poll is less accurate than a "registered voter" poll this early. The reason is very basic and simple... that people aren't paying too close attention yet, the conventions haven't occurred (so, enthusiasm is tough to gauge), etc. Obamaniacs might argue that the likely voter model will be worse than the registered all the way up to the end this year, due to previous non-voters who are fired up this year getting screened OUT by the likely voter model due to past voting inactivity. This argument may, in fact, be true... to a degree (pollsters don't ONLY use past voting performance to determine likely voting, and certainly we can't trust 100% of the supposedly-enthused previously non-voters to show up). And, gee, let's go take a look at likely versus registered voter polls...
Go take a look at the list of national polls.
The RCP average shows Obama up by only 1.2%. But let's do some math here. Check out which ones of those are likely voter polls... Zogby, Rasmussen, Quinnipiac and Battleground. The average of those four is a dead even tie. The remaining recent four (all registered voter polls): CBS/NYT, NBC/WSJ, LAT/Bloomberg and Gallup tracking, average out to Obama being up 2.5%.
So, Obama is likely REALLY up about 2.5%. Still, one could argue that there is tightening. Looking at the same aforementioned page and scrolling down to look at older polls, we can examine all of the registered voter polls for the 10 days prior to the above sample. That'll include IBD/TIPP, Pew, CBS, Gallup Tracking and AP-Ipsos, which, in their average, show an Obama lead of 4.6%.
Therefore, Obama's lead MAY have been cut in half. But even this is debatable. All of these pollsters have different methodologies. So, one might reasonably argue that we need to compare apples-to-apples. In doing so, the trend isn't quite so obvious over the past 20 days. Once again, consult the above link... Gallup's Tracking had Obama up 3% in the old RCP entry (from 8/3), they're now at Obama up 2%; Rasmussen's Tracking had McCain up 1% in the old RCP entry, they now have Obama up 1%; CBS News had Obama up 6% and now has him up 3% (though the new poll was done with the NYT, so it's unclear if the methodology was identical). That's an average of less than a 1% shift to McCain.
Furthermore, just look at the trackers themselves. Here's Rasmussen. Obama was up by a wider margin in June and into early July, but that's obviously merely a result of good press from clinching the nomination (with a slow fade into early and mid-July). Otherwise, with leaners, and excluding a brief uptick in late July, related to his "world tour", he has always been between down 1% and up 3%. Without leaners it's the same story except that Obama does a bit better. In short, there has been essentially zero movement in this tracker.
...and as for Gallup? Well, yes, Obama did have a bigger lead between August 8th and 12th. But other than that, since the beginning of the month it's been a steady 0-4% lead... and now stands smack dab in the middle of that, at 2%. So, there has been no movement here either.
In short, the trackers show Obama off as his peaks, but zero movement from the median over the past couple of months. The remainder of the polls show some movement (about 2% towards McCain), but this movement can be debated as we are comparing against different polls with different methodologies.
The second important point is the demographics issue. What's missing in all the analysis is a look at who's undecided. In a close race, the way the undecideds break is critical. Unfortunately, I've not found a lot of crosstabs available in these polls. So, I'm not going to bother linking to any here (if you're interseted you can follow the links from the RCP site above). But we do know this:
From the LA Times poll 3% of Republicans are undecided while 11% of Democrats are. That same poll also shows slightly more women undecided than men. However, another recent poll (unfortunately, I don't recall which... possibly the NBC/WSJ poll, since it was referenced on MSNBC by Chuck Todd) indicated twice as many women undecided as men.
So, women and democrats, two demographics significantly more likely to vote for Obama, comprise a large percentage of the undecideds. Therefore, as people settle on a choice, all else remaining equal, Obama's lead is almost certain to expand. The only "bad" demographic for Obama is that the LA Times poll also showed more whites undecided than blacks. This is contrary to what we've seen in all of the primaries, but then again, John McCain is no Hillary Clinton. So, the LA Times data may be correct. That demographic split of undecideds would favor McCain. But that racial breakdown difference (as far as the number of undecideds) is smaller than the wide gap in party-identity undecideds. So, it remains likely that Obama will capture the majority of undecideds.
And on a final note, a national popular vote lead of 3% (which is about where Obama is likely at ...assuming his 2.5% lead in registered voter polls and giving him a small advantage among undecideds) is not insignificant. I've done a study comparing popular vote margin to electoral vote margin. Fitting a curve to the data shows that a 3% popular vote lead yields, on average, a 20-25% electoral vote victory. That is, in other words, a 100+ electoral vote margin of victory for Obama. (If Obama's lead shrinks to 2% in the national popular vote he's still likely to win the electoral college by over 50 votes; and if his popular vote lead expands to 5%, he's likely to win the electoral vote by a margin of about 170.)
Bottom line: Yes, the race has probably tightened. But, it has not tightened as much as some are implying. They're looking too hard at "likely voter" polls, and they're comparing against Obama's peaks (more often than not, he's only been ahead by 1-4%, just like now ...not 5-10%, as is being implied). Furthermore, that lead, if it has dwindled, will be recovered through a capture of the majority of undecideds. And, finally, as small as that popular vote lead looks, it would likely lead to a substantial electoral vote victory for Obama.