The Tightening Cometh
(updated below with some better news)
(updated again with Nate Silver's wisdom)
As I mentioned yesterday, the long awaited tightening in the national polls seems to have arrived, at least in the tracking polls. Obama is now up by only 3 points in the Rasmussen tracking poll (down from +8 last week). He's now up only 6 in the DKos tracker (down from +12 last week). The Gallup traditional model has him only up 2 (down from +6 last week). The WaPo tracker has Obama up by 7 (down from +11 last week). The Zogby Tracker has Obama up by 5, (down from +12 last week). And the GWU poll has Obama by 3 (it's remained fairly steady). Given that these polls are based on three day averages, several of them (specifically the Gallup, Rasmussen, and DKos polls) seem likely to tighten further when more favorable data rolls off the average.
[I should add that the results of the Hotline/Diageo tracking poll have stayed relatively steady over the last week. Obama is up 7, compared to a high of +8 last week]
At this point, it's pretty hard to dismiss all this as statistical noise. It's seems likely that, at least in the national horse race numbers, McCain has closed the gap somewhat over the last week. What's interesting is that this tightening is not yet appearing in the state polls. There are two possible explanations for that. The first is that the state polls are lagging indicators, that we'll start to see some tightening in the state numbers over the next few days. The second possibility is that Obama's advertising and campaigning in the battleground states has counteracted whatever phenomenon is causing the national numbers to tighten. I suspect the answer is somewhere in between.
There are two things to watch for in the coming days. First, it will interesting to see if Obama's infomerical tonight has any effect on the polls. If it does, we probably won't see it until Friday. In the meantime, given the trendlines, there's a real chance that over the next few days (maybe even today) a major polling organization will release a poll showing McCain either tied or slightly ahead. We might see that in the Gallup traditional model this afternoon. When that happens, you can rest assured that the Drudge Report will hype it with a siren and a banner headline and everyone will be talking about it. In that sense, Obama's half-hour commercial is particularly well timed. It will give the media something else to talk about tonight and a reason to wait a few days before reading too much into today and tomorrow's polls.
I don't want to sound all doom and gloom. I'm still very optimistic about Obama's chances. Indeed, I've long thought that this sort of tightening was likely in the last week of the race. I do, however, think that many in the media and on the left have gotten carried away in declaring Obama's inevitable victory. This is a race where the fundamentals favor Obama and the odds are that he will win by a comfortable margin. But it has never been a 1996 or 1984 style blowout. McCain's victory scenario is narrow but not implausible. His deficit in the polls is significant but not insurmountable. This race is not over.
UPDATE: Gallup has just released its numbers for today and they are a little better than I anticipated. Obama is now at +3 in the traditional model (up from +2) yesterday. He also gained a point in there registered voters numbers (now at +9); he stayed the same in their expanded likely voter model (+7).
As I mentioned above, the Diageo/Hotline numbers are also holding steady with Obama at +7 (only one off of his recent high water mark in that poll).
UPDATE II: Above, I referred to the possibility that state polls are lagging behind the national tracking polls. Based on my own poll watching, that seems to have happened several times in this election cycle already. Movement was evident in the tracking polls first and only later showed up in the state numbers. So I was surprised to see Nate Silver, who knows a lot about polling, say that this is a myth. He says that most pollsters release their state numbers within 24 hours, so there is no appreciable lag.
While I see his point, I think he's overstating things. First, while individual state polls may be just as fresh as trackers, state polling averages surely are lagging indicators because they necessarily include older state polls (and the oldest of those polls include even older data). Second, while most of the tracking polls include only the last three days worth of interviews, many state polls include more. For instance, the F&M poll of PA today (showing Obama up by 13) was conducted from Oct. 21-26. So a lot of the data in that poll is pretty stale. Even the most recent interviews are over two days old. The same is true of the AP polls released today, which were conducted from Oct. 22-26. And the latest batch of CNN polls, which were just released moments ago, were conducted from the Oct. 23-28. So while that data was released within 24, it stretches back five days. That means these polls are less likely to capture any movement that occurred within the last 24-72 hours. So I still think a bit of a lag effect is possible, and I would not be surprised to see some closer state polls released over the next few days.
(updated again with Nate Silver's wisdom)
As I mentioned yesterday, the long awaited tightening in the national polls seems to have arrived, at least in the tracking polls. Obama is now up by only 3 points in the Rasmussen tracking poll (down from +8 last week). He's now up only 6 in the DKos tracker (down from +12 last week). The Gallup traditional model has him only up 2 (down from +6 last week). The WaPo tracker has Obama up by 7 (down from +11 last week). The Zogby Tracker has Obama up by 5, (down from +12 last week). And the GWU poll has Obama by 3 (it's remained fairly steady). Given that these polls are based on three day averages, several of them (specifically the Gallup, Rasmussen, and DKos polls) seem likely to tighten further when more favorable data rolls off the average.
[I should add that the results of the Hotline/Diageo tracking poll have stayed relatively steady over the last week. Obama is up 7, compared to a high of +8 last week]
At this point, it's pretty hard to dismiss all this as statistical noise. It's seems likely that, at least in the national horse race numbers, McCain has closed the gap somewhat over the last week. What's interesting is that this tightening is not yet appearing in the state polls. There are two possible explanations for that. The first is that the state polls are lagging indicators, that we'll start to see some tightening in the state numbers over the next few days. The second possibility is that Obama's advertising and campaigning in the battleground states has counteracted whatever phenomenon is causing the national numbers to tighten. I suspect the answer is somewhere in between.
There are two things to watch for in the coming days. First, it will interesting to see if Obama's infomerical tonight has any effect on the polls. If it does, we probably won't see it until Friday. In the meantime, given the trendlines, there's a real chance that over the next few days (maybe even today) a major polling organization will release a poll showing McCain either tied or slightly ahead. We might see that in the Gallup traditional model this afternoon. When that happens, you can rest assured that the Drudge Report will hype it with a siren and a banner headline and everyone will be talking about it. In that sense, Obama's half-hour commercial is particularly well timed. It will give the media something else to talk about tonight and a reason to wait a few days before reading too much into today and tomorrow's polls.
I don't want to sound all doom and gloom. I'm still very optimistic about Obama's chances. Indeed, I've long thought that this sort of tightening was likely in the last week of the race. I do, however, think that many in the media and on the left have gotten carried away in declaring Obama's inevitable victory. This is a race where the fundamentals favor Obama and the odds are that he will win by a comfortable margin. But it has never been a 1996 or 1984 style blowout. McCain's victory scenario is narrow but not implausible. His deficit in the polls is significant but not insurmountable. This race is not over.
UPDATE: Gallup has just released its numbers for today and they are a little better than I anticipated. Obama is now at +3 in the traditional model (up from +2) yesterday. He also gained a point in there registered voters numbers (now at +9); he stayed the same in their expanded likely voter model (+7).
As I mentioned above, the Diageo/Hotline numbers are also holding steady with Obama at +7 (only one off of his recent high water mark in that poll).
UPDATE II: Above, I referred to the possibility that state polls are lagging behind the national tracking polls. Based on my own poll watching, that seems to have happened several times in this election cycle already. Movement was evident in the tracking polls first and only later showed up in the state numbers. So I was surprised to see Nate Silver, who knows a lot about polling, say that this is a myth. He says that most pollsters release their state numbers within 24 hours, so there is no appreciable lag.
While I see his point, I think he's overstating things. First, while individual state polls may be just as fresh as trackers, state polling averages surely are lagging indicators because they necessarily include older state polls (and the oldest of those polls include even older data). Second, while most of the tracking polls include only the last three days worth of interviews, many state polls include more. For instance, the F&M poll of PA today (showing Obama up by 13) was conducted from Oct. 21-26. So a lot of the data in that poll is pretty stale. Even the most recent interviews are over two days old. The same is true of the AP polls released today, which were conducted from Oct. 22-26. And the latest batch of CNN polls, which were just released moments ago, were conducted from the Oct. 23-28. So while that data was released within 24, it stretches back five days. That means these polls are less likely to capture any movement that occurred within the last 24-72 hours. So I still think a bit of a lag effect is possible, and I would not be surprised to see some closer state polls released over the next few days.
posted by A.L. at 9:01 AM
13 Comments:
I suspect conservative-leaning undecideds are moving to McCain and/or there is some buyer's remorse going on. More worrisome is that Obama numbers have slipped a point or two.
One huge bright spot for Obama in Rasmussen: "Among those who have already voted, it’s Obama 54% McCain 45% with other candidates picking up a single percentage point." (Pew puts the margin at +19 Obama among those who have already voted and the WA Po has a similar number. The state numbers for already voted in Ohio and VA are +17 and +37 (!) respectively according to SurveyUSA.) Even if McCain pulls ahead by 1 or 2, it seems he's going to have a lot of ground to make up on election day.
I wonder how many Hillary supporters who (a month ago?) grudgingly supported Obama are now bailing out of that position. I don't think it's many, but would be interested to learn after the election what happened with that group.
Consider the plight of the unfortunate undecided voters -- and, let me preface what I am about to say with the statement that this applies to the class of undecided voters in general, and is not meant to demean or insult any particular individuals. I'm sure you have, individually, valid reasons for sitting on the fence.
Collectively, the undecideds at this point have to be archetypical low-information voters. That is, ignorant, barely-sentient, displaying the intellectual curiosity of a houseplant, unaware of the broad sweep of human events, except possibly as they interfere with the TV schedule.
That these folks are going to break for McCain is just so difficult to comprehend.
On the other hand, the question arises as to how pollsters classify a voter as "undecided." We have a household policy here that we politely decline to participate in polls. Therefore, we could be considered undecided. In case you were wondering, we're not. We just enjoy withholding information from pollsters to make their lives (and the lives of politicians they report the results to) more difficult.
That it seems to be raising A.L.'s blood pressure is an unfortunate side-effect, and one for which I apologize.
There may be a plausible--i.e. not fanciful--McCain victory scenario. However plausibility and probability are two different things.
There's absolutely nothing wrong with warning against complacency. But it's also not wrong to accurately describe the likelyhood of a McCain victory. It is very unlikely. Less than 1 in 20 chances, if I read the state races right at the usual sites.
It is very unlikely. Less than 1 in 20 chances, if I read the state races right at the usual sites.
The problem is that the state numbers are pretty hard to reconcile with the most recent national numbers. *If* the national polls showing Obama with a 2-3 point lead are accurate, then it simply can't be the case that Obama has as comfortable an electoral college lead as some of those sites suggest. *If* Obama is only up 2-4 points nationally, then McCain has a very realistic shot at winning. If Obama national lead is really in the 5-7 range, McCain's has a plausible but unlikely victory scenario (which would require further tightening over the next few days). If the lead is actually greater than 7, it's over.
If the polls really are tightening, I don't understand why this is not reflected on Nate Silver's site, for instance. He still has Obama as better than a 20-1 favorite to win and those odds are about as high as they ever have been. Also, his projection of the popular vote has not changed significantly in quite a while.
I'm not so sure the "tightening" isn't mostly baked in at this point. If you look at the DKos tracker, the tightening is mostly due to the +14 and +11 days last week rolling off. The last 3 days are +5,+6,+5, respectively, which seems pretty steady (each day's writeup on the site publishes the actual daily number). Additionally, it looked like Rasmussen had a big McCain day Sunday. Awaiting today's Gallup, but I think tomorrow's number will be demonstrative if there's really some additional movement, or just a convergence to a more "natural" 4-5 point Obama lead.
rashomon, I agree to an extent. What concerns me about the DKos poll is that it has been consistently showing a bigger lead than the others, until this week. It does seem to be hanging steady in the 5-6 range, but that's dramatically lower than what it was just last week. And the DKos poll makes some very favorable partisan weighting and turnout assumptions.
It'll be interesting to see what Gallup shows today.
The non-tracking national polls still mostly show large Obama leads. I'd put more stock in them because of their larger sample sizes and smaller margins of error than in the daily tracking polls.
When we are getting state polls showing Georgia, Indiana and Arizona as toss ups there is no way that the national horse race can be anywhere near tied. Obama should win a very comfortable victory on Tuesday and we're all going to look back and wonder why we wasted our time obsessing with national popular vote tracking polls.
The problem is that the state numbers are pretty hard to reconcile with the most recent national numbers.
Sure, but it's not hard to reconcile the state numbers with the AVERAGE of the national numbers. Look, the average national lead of all the polls at pollster.com and RCP is about 6 points. A 6 point win certainly nets AT LEAST a 300 electoral vote haul for Obama, probably a 325-350 vote haul. (Clinton won 379 with an 8.5 point win.) Thus, a 6 point national lead would square with Obama having decent leads in most of the battleground state polls.
Chris Bowers has a great discussion on state and national polls at:
http://www.openleft.com/showDiary.do;jsessionid=C8C15AF1C101133E89FCD0385BF660BD?diaryId=9503
Bottom Line: No need to panic.
You said:
"The first is that the state polls are lagging indicators"
Nate Silver has a post[1] up that debunks this theory.
[1] http://www.fivethirtyeight.com/2008/10/myth-of-lag.html
As usual, the Electoral Vote Predictor on the front page of https://www.intrade.com/ is interesting. For state races, there are (almost certainly) some high-information bettors, trying to take money from low-information bettors.
Post a Comment
Links to this post:
Create a Link
<< Home