What to Make of the Polls
This has been an especially odd election when it comes to polls. We're seeing an exceptionally large divergence in both state and national poll numbers, not only in absolute numbers but in trendlines. Both of these phenomena are worth pondering.
Divergence in Absolute Numbers
Today the Pew Research Center, a respected polling organization, released a national poll with a large sample size showing Obama running away with the election. According to Pew, Obama is up by a staggering 52-36 margin among registered voters and 53-38 among likely voters. Last week there were several national polls, from Newsweek and CBS, showing similarly large margins.
On the other hand, Gallup--another venerable polling organization--now shows Obama ahead by just 2 points, 49-47 in their daily tracking poll (under their traditional likely voter model). There have been several other polls this week showing a similarly narrow lead, in the 1-5 point range.
This kind of divergence cannot be the result of simple statistical noise. It can only be explained by the use of differing assumptions about who will actually show up on election day. The Gallup poll makes this point clear, showing Obama with a larger 51-44 lead when using an expanded likely voter model. The bottom line, though, is that no one will really know which model is right until next week.
The other thing worth noting is that--in all the polls--Obama's support exists in a much narrower band than McCain's support. Obama's support ranges from roughly 49% to 53% in virtually every poll. But McCain's support ranges from as high as 47% in some polls to as low as 36% in others. I'm not sure what to make of that, but my hunch is that Obama is much closer to his ceiling of support than McCain is, and much of the fluctuation in McCain's numbers may be the result of "soft" supporters who only claim to support him when really pressed by the interviewer to make a choice (some pollsters lean on undecideds harder than others). If that's true, we shouldn't be surprised to see most of the undecideds ultimately break for McCain, in which case the final result will look closer than the polls are indicating.
Divergence in Trend Lines
The other interesting phenomenon we've been seeing is a divergence in trendlines among the many tracking polls. Some always seem to be zigging while the others are zagging. A while back, I made the following prediction:
That said, we may actually be reaching a point of convergence now. In the last few days, most of the tracking polls show movement toward McCain. All of the various Gallup models show a 3 point tightening today. The Rasmussen daily tracking poll also has Obama's lead shrinking by 3 points over the last few days. And the Daily Kos/Research 2000, ABC/Washington Post, and Reuters/Zogby polls show even steeper drops over the last few days (Obama's lead has dropped 5 points in DKos, 5 in the WaPo, and 8 in Zobgy).
The only poll that seems not to be showing this trend is the Diageo/Hotline tracking poll, which has shown a steady Obama lead of 50-42 for several days now.
Is this an actual tightening of the polls? Possibly. Will the McCain campaign be able to use this trend to sell a comeback narrative to the press? Maybe, but it will probably take another day of movement toward McCain before the media treats this as more than statistical noise (especially in light of today's Pew poll).
Divergence in Absolute Numbers
Today the Pew Research Center, a respected polling organization, released a national poll with a large sample size showing Obama running away with the election. According to Pew, Obama is up by a staggering 52-36 margin among registered voters and 53-38 among likely voters. Last week there were several national polls, from Newsweek and CBS, showing similarly large margins.
On the other hand, Gallup--another venerable polling organization--now shows Obama ahead by just 2 points, 49-47 in their daily tracking poll (under their traditional likely voter model). There have been several other polls this week showing a similarly narrow lead, in the 1-5 point range.
This kind of divergence cannot be the result of simple statistical noise. It can only be explained by the use of differing assumptions about who will actually show up on election day. The Gallup poll makes this point clear, showing Obama with a larger 51-44 lead when using an expanded likely voter model. The bottom line, though, is that no one will really know which model is right until next week.
The other thing worth noting is that--in all the polls--Obama's support exists in a much narrower band than McCain's support. Obama's support ranges from roughly 49% to 53% in virtually every poll. But McCain's support ranges from as high as 47% in some polls to as low as 36% in others. I'm not sure what to make of that, but my hunch is that Obama is much closer to his ceiling of support than McCain is, and much of the fluctuation in McCain's numbers may be the result of "soft" supporters who only claim to support him when really pressed by the interviewer to make a choice (some pollsters lean on undecideds harder than others). If that's true, we shouldn't be surprised to see most of the undecideds ultimately break for McCain, in which case the final result will look closer than the polls are indicating.
Divergence in Trend Lines
The other interesting phenomenon we've been seeing is a divergence in trendlines among the many tracking polls. Some always seem to be zigging while the others are zagging. A while back, I made the following prediction:
[A]t some point in the next 21 days, there will be a day or two where most of the major tracking polls show movement toward McCain. Statistically, that's just bound to happen. And when it does, Drudge will hype McCain's "comeback" and various GOP surrogates will attribute the rise to some external event (a "gaffe" by Obama or a new argument by McCain, etc.) and the press will run with it--if for no other reason than they are tired of the Obama running away with the election narrative.That prediction has not (yet) come to pass, and I think that's likely due to the proliferation of daily tracking polls. There are just so many of them now that it's hard to make statistical noise look like an actual trend. Tightening in some polls has consistently been balanced out by expansion in others. Interestingly, this cancellation effect has likely affected coverage of the campaign. If, like 4 years ago, we only had the Gallup and Rasmussen trackers to look at, there would have been a few instances over the last month where both polls seemed to show movement toward McCain and the McCain campaign would have had a much easier time selling the media a comeback narrative.
That said, we may actually be reaching a point of convergence now. In the last few days, most of the tracking polls show movement toward McCain. All of the various Gallup models show a 3 point tightening today. The Rasmussen daily tracking poll also has Obama's lead shrinking by 3 points over the last few days. And the Daily Kos/Research 2000, ABC/Washington Post, and Reuters/Zogby polls show even steeper drops over the last few days (Obama's lead has dropped 5 points in DKos, 5 in the WaPo, and 8 in Zobgy).
The only poll that seems not to be showing this trend is the Diageo/Hotline tracking poll, which has shown a steady Obama lead of 50-42 for several days now.
Is this an actual tightening of the polls? Possibly. Will the McCain campaign be able to use this trend to sell a comeback narrative to the press? Maybe, but it will probably take another day of movement toward McCain before the media treats this as more than statistical noise (especially in light of today's Pew poll).
posted by A.L. at 2:15 PM
10 Comments:
"McCain's numbers may be the result "soft" supporters who only claim to support him when really pressed by the interviewer."
Is this accurate? Pollsters pressing for support for one candidate?
Or do you mean pressed for an answer?
Which also would surprise me.
yeah, i should have been clearer. Some pollsters lean on undecideds harder than others, really pressing them to state a preference. That's why the undecided numbers vary from firm to firm.
Hmmmm. Interesting. I think polling is something I could get interested in doing.
Seems to me some conclusions could be drawn from how hesitant a callee was to respond.
Anyways, thanks for the response and I enjoy your blog.
Peace and good health to you and your family.
Yikes A.L, these recent posts of yours are freaking me out a bit. However, one thing gives me comfort - GOTV. I suspect that undecideds are likely to break for McCain (although at this point you have to woner what kind of cabbage could be undecided). However, I think that the enthusiasm gap in favour of Obama, combined with his awesome ground game, will get out the vote like nobody has ever seen before. Many of the pollsters may not have fully captured this effect. So, add enthusiasm to ground game and subtract racism, voter suppression and machine tampering and you have, I hope, a 7 point win in Penn, a 5 point win in Colorado and a 3 point win in Virginia and that, as they say, is the ball game.
By Juliet Eilperin and Alice Crites
STATE COLLEGE, Pa. -- Sarah Palin delivered tonight a sweeping -- and not completely accurate -- portrayal of what Barack Obama said about the Supreme Court in a radio interview seven years ago.
"He called the court's failure to take up the redistribution of wealth 'a tragedy,' " Palin said. "Is that an indication he would want to rewrite the founding document of our great nation to achieve his goals, and what does that say about what he would do about future Supreme Court justices?"
The audience, which packed Penn State University's Recreation Hall, roared at her remarks ...
The discrepancy does seem to hinge on what a "likely voter" actually is. But even then, Obama is so far ahead in the electoral college that McCain has only a very remote chance.
That being the case, Obama would do well to look at the senate races: MN. KY. GA. TX.
The senate is now the weakest link by far. He should do what he can, now, to shore that up. He should swing through those senate battlegrounds, and leave McCain to beat his brains out in PA.
Boy, wouldn't that be awesome if Mitch McConnell went down to defeat?
Heaven would be McCain losing in Arizona - It would make up for Gore's loss in Tenn.
Franken (i always found him annoying, but i want him to win) will win Minn and Bill O'reilly is going to freak out.
Georgia will be tough. IMO - Chamblis is a nasty asshole, but a good politician. It will be close. Max Cleland should
pray for a monster level turnout among blacks..
Texas is doable too But tough- Too bad there isn't a good right wing third party candidate taking away votes from Mac.
Re: Texas--not for Obama/pres. race, but to knock off Cornyn. It's a 6 point gap, +/- 3, I think, but the poll is dated, and the gap was closing. With a little outside help, Cornyn could be gone.
Hey A.L., what do you think of the chances that the polls, across the board, could be systematically underestimating McCain's support, the way they underestimated Hillary in the New Hampshire primary? Assuming there's not a big change in the state-by-states I don't see Obama losing in the Electoral College unless this is the case. Could NH happen again?
publius,
I'm not too worried about systemic polling error. I think some of the more conservative polling models could prove to be correct, but I doubt they're all wrong (i.e. I doubt McCain is actually ahead right now). I think NH was a very fluid situation and that's why the polls were off. My main concern here is that McCain manages to get the race down to 1-3 points nationally and then wins in a Bush 2000 scenario, losing the popular vote but barely scraping together enough red state electoral votes to win.
Post a Comment
Links to this post:
Create a Link
<< Home