That’s what I also can’t quite understand. I would have thought that a straight forward “who are you supporting” is going to be the most accurate indicator.
For anything else
Isn’t it self evident that if you are voting Obama, you will tend to consider yourself a democrat and vice versa?
If you are NOT voting according to your registration (either for president or down ticket), aren’t you more likely to call yourself independent?
What are people more likely to be confused about - who they are voting for or who they are registered with? (in relation to people being wrongly resigtered as Repub)
In short, I’d follow the polls rather than registrations.
That Karl Rove article is a masterpiece of cherrypicking. He goes to great lengths discussing the national polls which supposedly show Romney in the lead (though right now even that is debatable) while ignoring the Ohio polls which show a much clearer Obama lead. If you believe the polls are rubbish let’s at least be consistent.
I am not sure about his sources for early voting data. This is what the WaPosays about Ohio:
If the state doesn’t have traditional registration where is Rove getting his numbers from? In any case it’s not showing Romney leading in the early voting. It’s showing a smaller Obama lead than 2008 but still very healthy at 54-46. Romney will have to win big on election day to make up for that.
I know. It’s ridiculous how some of us refuse to ignore the polls, despite their historical record of accuracy.
Not that they can’t be wrong, just that it’s unlikely that they’re wrong. You haven’t convinced me yet (Party registration and self-ID are not the same!).
It’s possible you’re right. I don’t know why you’re so convinced, though, considering the polls. If you’re right, all it means is that the polls are wildly off. But they could be off in any direction- they could be undercounting the # of Republicans and independents who support Obama, for example, just as they could be overcounting the # of Democrats. It sounds like wishful thinking to me… but we’ll see on Nov 6. I’ll put my money on people with records of accuracy.
And Romney wouldn’t need to buck the trends against him in the polls? You make me laugh.
I could be wrong. Nate could be wrong, and the polls could be wrong. Intrade could be wrong. So could you, obviously. I’m not going to just take your word for it, because we’ll find out in 5 days. But considering that the pollsters and Nate have addressed this topic before, and seem to be taking it into account, I think it’s more likely that they’re right and you’re wrong.
So either the few big national trackers are producing results that are (unintentionally) biased to Romney, or, as a group, the much larger number of state polls are systematically biased to Obama.
Is there a plausible mechanism for either?
Nationally the potential source for bias is within the demographic correction being off. Do state polls “correct” for regional demography as much? If so what datasets do they use? Are they systematically doing it in the same way and is that way more or less likely accurately reflecting the likely voter pool compared to the methods of the national trackers? If not is that a potential source of error or an avoidance of one at this point?
Because if it is true that if the numbers from the states are systematically biased by some error they all share against Romney by much more than 2 points then this race is likely his. As per the recent 538 analysis the past record is that national polls are more often biased than the states polls overall, but past results are no assurance of future results …
OMG is making the same argument that some Democrats made in the days before the election in 2004- that the polls innacurately represented (oversampled/overestimated) the number of Republicans, showing a Bush advantage. But the polling averages were right in 2004, and these partisans were wrong. I think it’s more likely than not that the polling averages are right in 2012.
Yes, the argument parroted by OMG and his ilk are silly. But there was not the disconnect between the national polling (Gallup and Rasmussen) and the state polling datasets in 2008. In fact that year the national data (see the recent 538 article for the table) had no bias while the state data had a small one. In 2004 the state data was less biased than the national polls but also by a similarly small amount. You have to go back to 2000 to get a significant discrepancy between the datasets, the states data doing better by 2 points then. And yes in 1996 the national data was much more biased than the state data was, being off by over 5 points compared to the nearly dead on of the state data.
Does disparate performance in 1996 and 2000 really meaningfully predict performance in 2012? Are the sorts of potential systemic errors made then likely to be the same sorts of systemic errors that could be made now? Agreed arguing that firms should be correcting for voter ID is a silly position, but there can be other systemic approach factors that the state pollers share that create some error.
I’m not resting easy. But there’s not really a disconnect between the national and state polling- unless you disregard most of the national polling. The national polling averages show a very tight race- within 1 point or so- while the state polling shows consistent leads of various sizes for Obama in most of the swing states. These do not conflict.
Here’s a very specific 2004 example on party ID and polling (starting from arguments in this article)
In 2004, in the presidential election in North Carolina, pre-election polling showed George Bush leading John Kerry about 52-44 (with 3-4% undecided). These polls (see here and here for the two SurveyUSA polls closest to the election), showed an approximately even D/R split in voter self-identification (40D/38R/21I and 39D/39R/21I in the two linked polls).
However, actual voter registration in NC (summarized here; go here for the raw numbers) showed a significant Democratic edge at 47D/34R/19I.
Why the party ID difference between the polling and registration? One possible reason would be that the polls, for some unknown reason, over-sampled Republicans; thus, Kerry’s actual results in North Carolina should be much better that the polling showed, and Bush’s results worse. Alas, the actual results on election day were Bush 56/Kerry 44, meaning Bush actually did a little better than what one would predict from the aggregate polls, despite the clear Democratic advantage in party registration.
But what’s the most interesting is that CNN did some exit polling in North Carolina in 2004. The exit poll shows Bush winning 56% to 43% (very close to the actual result), but also shows party self-identification of those polled. This was 39D/40R/21I, almost exactly matching the party **self-identification **from the pre-election polls, and nowhere near the breakdown of registered voters.
The conclusion here is that there were a significant number of people who were **registered ** with one party, but **self-identified **with another.
The application of this 2004 example to the 2012 election is left as an exercise for the reader.
The best you can say about the embarrassing lies about Chrysler and GM and abortion are that they’re the final desperate act of a craven campaign that knows it is fucked.
Anyway, this morning’s aggregate of national polls on Real Clear Politics has the national race at a tie. Which is going Obama’s way, since Romney’s small lead has been steadily coming down.
The requirement to register with a party varies from state to state. Even in states that do require an affiliation with a party at registration, the only possible meaning it can have afterward is limiting primary voting to the party of registration. In New York, e.g., if you register as a Republican you can only vote in Republican primaries. The registration form requires you to pick one of the following options:
If you pick “I do not wish to enroll in a party” you are legally considered to be a Blank not an independent. Notice that there are both an Independence Party and a Conservative Party among the third party options. Registering as one of them does not necessarily mean you are conservative or independent any more than registering as a Democrat means you’re a liberal.
Once you’re registered, you’re set for life if you want. If you don’t care about voting in primaries - which is the majority of voters - then whatever you put as your party 20, 40, 60 years ago is what you stay no matter how you vote in any main election or what your beliefs are.
Other states are even less strict. They allow any registered voter to vote in any primary.
Are party registrations and voting behaviors and political affiliations positively correlated? Of course they are. But the correlation is not 1 or anywhere close to it.
That’s why Gallup and others ask the question as of today. There is no reason to assume that party registration plays any part in any individual’s current behavior.
And the same for bengangmo’s " Isn’t it self evident that if you are voting Obama, you will tend to consider yourself a democrat and vice versa?"
No, it’s the opposite of self-evident. It may not be true at all. Some people who consider themselves independents will vote for Obama. Some people who consider themselves members of third parties will vote for Obama. Some people who consider themselves Republicans will vote for Obama. (And some people who consider themselves Democrats will not vote for Obama.) Voting behavior switches from election to election as people convince themselves that - in this particular instance - a vote has to be cast that varies from what they consider to be their core beliefs.
In this particular instance… That’s what polls try to measure, that’s what forecasters try to predict. The only sure way to be wrong is to do nothing but measure past behavior or what box you happened to check 50 years ago.
Because they are taking small random samples and then attempting to draw wider conclusions from those small samples. If your sample size is 500 and 498 of them are Democrats you are going to get different results than if you have a good random mix…and you are going to be able to draw wider conclusions from your poll results than polling those 498 Democrats and asking who they are going to vote for in this election.
It’s a reasonable question to ask, as long as people don’t fall in to the trap of thinking it’s immutable or comparing it to registration numbers. I might still be a registered Republican for all I know, but haven’t voted for the GOP presidential candidate since 2000.
Getting 498 out of 500 Democrats is obviously a bad sample, but seeing the D/R/I share vary by 5% or less is perfectly typical, and not something you really want to be weighting against. If you do, you are in danger of missing reality.
Race, age, and gender are completely different, as folks don’t generally change those attributes between elections.