Guys, national polls are not meaningless.

[Zynx]

Nate Silver would tell you almost precisely that at fivethirtyeight.com as would any professional. States follow national trends to a large extent. What do I mean by this? What it means is that a Democratic leaning state like Pennsylvania will not vote Republican if Obama wins nationwide by 6 points. It also means we will not flip Georgia in a close election nationally. States do not vote in a vacuum.

Most importantly, for the swing states, it means that pretty much all of them will flip simultaneously if Obama pulls off a multiple-point win. If he wins nationwide by 5-6 points, that means that all of the states like Colorado, Nevada, New Mexico, Virginia, Ohio, Florida, probably Indiana, and Missouri will flip our way. No state is completely alien to the rest of the country. They vote in patterns and thus the national popular vote is pretty indicative of how states are going to vote, plus or minus their partisan leanings which range from virtually even in Nevada to horrifically lopsided in Vermont and Utah.

It really isn't 50 separate little races. That is naive to think so. There isn't a chance in hell one of the candidates can win the electoral college if they lose the popular vote badly. The only time there was a significant disparity between the two was in 1876 when Tilden won by 3 points nationally, but was robbed of South Carolina, Florida, and Louisiana's electoral votes by crooked Reconstruction governments in those states. All of the other instances that weren't decided in the House, 1888 and 2000, were extremely close and we reliant on one or two states at the margins.

[bluestateguy]

1. Not meaningless, but less important than state polls

And if it is true that Obama is ahead by 6-8 as Gallup and Research 2000 say, then he is likely ahead in Florida.

[Zynx]

3. And some polls suggest that may be the case.

[MidwestTransplant]

2. I agree. Also state polls tend to lag the national polls in my observation.

[grantcart]

4. The question is when a change agent comes into play and registers

10 million new voters and they go out and vote what happens to all of the traditional models of analysis and their weighting.


The polls are fine for showing movement from point A to point B (Obama going up or going down) but they do not know what turnout model to work on.

They know it and the controversy in the polling community is huge. If you google "poll weighting" you will see thousands of recent articles trying to wrestle with it.

The pollsters are starting to put in their asterisks.

Last week Chuck Todd said "Well when it comes to states like Virginia and Florida the Obama has a huge GOTV movement and we could be significantly underpolling new voters".

They know their numbers are shaky when it comes to new voters.

[Zynx]

7. I don't dispute that, but state polls would be no better than national ones at picking that up.

Most national polls tend to be around 1000 respondents. Most state polls are 500-600.

[grantcart]

11. well the ratios are better to begin with but more to the point state polls can adjust

their weighting based on the 2006 voter turn out or are even able to make an adjustment based on documentable numbers of registered voters as confirmed by the state's Secretary of State - national polls are using the current numbers based on the demographic turnout 4 years ago - which is understandable.

But your larger point that the poll is still important I agree with. Even if it is underreporting Obama by 3-4% it is still showing the overall direction of voter opinion and an accurate record of who has the momentum.

[Zynx]

15. In terms of statistics, there is not much difference between 500 out of 3 million and 1000 out of

300 million. It's an exponential decay in meaning.

[Hippo_Tron]

12. That's why we're seeing a huge variation in the state polls

One had Obama up in Colorado by 9 one had him up by 1. That means one model is taking into account new voters and one isn't.

[elocs]

5. This is a 50 state election and not a national election.

National polls are not meaningless, but their value and meaning needs to be put into perspective. The accuracy and validity of all polls are determined by their sample size and their sampling methods. It is not difficult to manipulate a poll to obtain the desired results and many people will accept it as gospel, and then wring their hands.

[Zynx]

6. Do you dispute my basic assertion states are not in vacuums and instead follow

national trends? The evidence is very strong that they do.

[Sensitivity]

8. Are you talking about BLUE AMERICA or RED AMERICA?

:rofl:

I agree with the general principle. There are blocks of states that will be affected differently by events, but the swing states probably move the same.

[Balbus]

9. Well, that depends on who the polls show as winning...

[Zynx]

10. Lol. So true.

[LiberadorHugo]

13. Tilden was not robbed!

It's not robbery when a White supremacist sympathizer is defrauded out of electoral victory; it's justice! (That being said, the corrupt, bourgeois-cappie factions of Reconstructionism sold out African-Americans to get Hayes elected.)

[Zynx]

16. The Republicans were no more enlightened than Democrats on the subject.

They just saw black voters as useful.

[Drunken Irishman]

14. Pretty much.

Obviously state polls are important, but you're right, if Obama wins the popular vote by 6 points, those swing states aren't going to McCain. It's nearly mathematically impossible, unless Obama somehow manages to win the Democratic states by 80-20 and loses by a very slim margin in all the swing states. Ain't going to happen.

Now that isn't to say the reverse can't happen. Obama could win the popular vote by 2 points and win rather comfortably in the electoral college. As Nixon did in 1968, when he beat Hubert Humphrey by a little over 511,000 votes, or less than a percent nationally, but won the electoral college by 110 electoral votes.