Another good battleground poll for Obama today. PPP has him +4 over Romney in Nevada. 51-47.
Works for me. Setting up all the early/remote voting sites has been a major pain in the ass in the last few weeks…wish it was all over with right now.
You guys do the exact thing Nate Silver encourages you not to: cherry-pick the very best polls for your side.
There’s a difference between thinking these cherry-picked polls represent the state of the race, and believing that when added to the aggregate they will move things toward Obama. The first is irrational, the second is not.
I just go directly to 538 and look at the trendline every day after he updates it. I honestly don’t know how the rest of you manage to keep up with all the other polls every day. I find it exhausting, even just by proxy.
I was not cherrypicking the best polls for Obama, I was cherrypicking the most newsworthy polls for either side. A high-quality poll from Ohio is always like to be high on that list and that PPP Nevada poll is up there as well. If you think there are other polls as newsworthy please post them. Anyway as I suspected, 538’s forecast rose to 71%.
Ah, still talking about Gallup I see. This is simply a case of cherry picking polls to focus on. The comments on TPM simply reiterate this point. As I said about a week ago or so, Gallup was probably overstating Romney’s performance by 2 - 3 percentage points while other polls were understating his performance by about 2 - 3 points. Fast forward to now and-- sure enough-- Gallup has come down for Romney while other polls have… gone up for Romney. y no talk about the latter? 
(Remember, three or so days ago Obama actually overtook Romney in RCP’s daily tracker by, iirc, 0.1%.)
ETA> I didn’t look at the Nevada poll (simply because I never expected Nevada to go R, anyway) but the Ohio poll is laughable, especially the county-by-county crosstabs where they literally polled two to three people in a ton of counties.
They surveyed about 750 people in the state. That seems fairly standard; I checked the most recent Rasmussen Ohio poll, and they also surveyed about 750 people. The most recent SurveyUSA Ohio poll surveyed about 650 people. (Those are just the most recent Ohio polls in the RealClearPolitics list of recent polls.)
There are 88 counties in Ohio. Split perfectly evenly, that’s about 8.5 people per county. If you figure that they’re going to get a lot more respondents in counties like Franklin and Cuyahoga, which have major metropolitan areas, and fewer respondents in counties like Jackson where nobody lives (slight overstatement, but only slight), it shouldn’t be surprising that some counties in the poll only had 2-3 respondents.
There have been 10 polls in Ohio since the 18th, from every major pollster. 6 of those polls show an Obama lead, four show a tie. The simple average of those 10 polls is 1.9%. However there is also a clear pattern of the higher-rated polls ( as per 538) showing a bigger Obama lead. If you look at the five polls with the highest rating, Obama’s average lead is 2.8%.
The bottom line: Obama is 2-3 points ahead in Ohio. Anyone who believes that Romney is tied as of now is cherrypicking the best polls.
Me too. And a waste of time. Nate Silver made a decent point today: … efforts to look at the polls in a systematic way are likely to be superior to much of what passes for polling analysis, which looks at them on an ad-hoc basis. Trying to decode the meaning of an individual poll is not a terribly useful exercise when 20 or 30 of them are released every day. Silver also notes that FWIW other models that incorporate state data tend to lean more heavily towards Obama than his does. He didn’t mention names, but Princeton Election Consortium is one example.
538 gives Obama a 71% chance today. Romney’s current odds at the Presidency is nothing to sneer at. 3/10 isn’t bad and unmodeled factors could easily push the figure much higher or lower.
Cherry picking or not, PPP now shows O+5 in Virginia, up from O+2.
Yes, I know that if you’re going to look at individual polls, you’re gonna have a bad time, but the trend lines are still worth looking at regardless.
Ha. Believe it or not I was planning to write a post earlier arguing that individual polls can sometimes be informative and the hypothetical example I was thinking of was a high-quality VA poll showing Obama up by 4-5. I kid you not.
Anyway yes this poll is informative, more so than 90% of the political news and commentary written today about the race just as the Ohio poll yesterday was informative. There are some signs of Obama gaining a bit in the national polls and one might wonder whether he might gain in the swing states also. Virginia is arguably the second-most important of those states and one which tightened significantly after the first debate. Is it moving back to Obama? This poll is a useful data point which suggests that the answer may be yes.
A couple more states looking good for Obama, according to PPP:
Iowa - Obama now up 49-47, up a point from last week.
Wisconsin - Obama now up 51-45. His 6-point lead there is up from his narrower 2-point lead right after the Denver debate.
If Obama wins every state that 538 places at 70% or above he is at 271. It is very tough sledding for Romney to win.
And a one percentage point shift in the popular vote would be sufficient to put some of those 70 percenters back down to 50 percent.
Don’t kid yourselves. Nate Silver’s model does not allow for a popular vote victory for Romney and an electoral vote victory for Obama as very likely. Currently it’s rated at 7.4% chance of a popular/electoral discrepancy.
Silver’s model predicts that Obama will win the popular vote. If he does not, then he will also likely lose many of those swing states Silver currently ranks at 70%. That 70% number is completely dependent on Obama winning the popular vote nationally.
It's important to note that Nate estimates the popular vote using the state polls as well as the national polls. IIRC he has said that the former are more favorable to Obama and therefore his estimate of the popular vote is higher than if he just used the national polls. So while his model doesn't believe Obama is likely to win the election while losing the popular vote, it also doesn't believe Obama is likely to lose the popular vote in the first place.
Exactly. But historically, national polls are more accurate than swing state polls because they use a larger sample size. I really think Silver’s going to have some egg on his face this cycle. If Romney takes the lead in Ohio at some point between now and Nov. 6, and Silver’s model still shows a 60% chance or greater of Obama winning, his credibility will take a hit if it doesn’t happen.
A single national poll may be more accurate than a single state poll but since there are a lot more of the latter they provide a lot of relevant information about the national race. Nate’s basic point seems to be that combining national and state polls will give you a better of idea of the popular vote than just looking at national polls and that seems quite sound.
If someone told me there was a 60% probability of rain tomorrow, and it didn’t rain, I would not say his credibility took much of a hit. I’d think he just came up short on the very close odds of 40:60 Just saying.