DU polling experts: how are state polls superior to national polls?

[CTyankee]

I understand about over or undersampling in the national polls, for instance. But how are the state polls designed to avoid that error?

Like many other DUers I am placing a great deal of faith in the state polls and ignoring the national polls (at this point). I just wonder why state polling, with a smaller "universe", is more accurate. My understanding is that the larger the poll, the smaller the margin of error and vice versa.

Please help those of us less knowledgeable in science of polls...and thanks!

[Barack_America]

1. I don't think it's the accuracy, but the significance...

Considering we don't elect our President by popular vote.

At least that's what I assume people mean when they suggest we disregard the national polls.

:shrug:

[Exilednight]

2. It's because we use the electoral college to elect the President. In theory...........

the popular vote count could be 15 total votes for Obama and 22,000,000 for McCain - and Obama could win as long as those 15 votes are in the 15 states with the most electoral votes.

[suston96]

3. Polls are not based on accuracy.......They are just guesses. and bad ones at that.

"Faith in polls" of any kind is foolish. Use them to note trends but if you have been attentive to past polls and to the past elections they cover, you should realize how "inaccurate" polls have been.

If all pollsters used the exact same techniques and parameters maybe, just maybe they could be considered reliable, up to a point. Even then, don't bet on them.

[glowing]

4. Its the way that the President wins... Electoral college map is the accurate

assessment.. so state by state polls are the polls to watch.

[NYC_SKP]

5. Sample size and the electoral college, we win state by state.

National polls of, say, 600 to 1200 respondents (our of 3,500,000) are just peachy for gathering data on how Americans feel about single issues.

If we elected presidents by nationwide popular vote, it might even be useful (but the sample size would need to be larger).

The president is elected state by state with electoral college votes, with "winner take all" states giving all electoral college votes to that candidate with the highest popular vote.

Thus, my Aunt Emma's vote for McCain may show up on a national poll, but in November in California, her vote won't matter.

We'll win or lose this election state by state with electoral votes. Please visit: http://www.fivethirtyeight.com/ Do it today.

While state polls are not conducted daily, they are much more important in understanding where the election is headed.

Here, for example, is Florida....trending nicely for Obama:



Looking at national polls is fine, but we need to pay attention to the details and to individual states.

[CTyankee]

6. Thanks. I should have stated that I understand the Electoral state vote importance but really

my question went to the issue of methodology. What I think you are telling me is that methodology per se is not that important (within certain parameters)since it is irrelevant to the imperatives of our Electoral College system of choosing a president. And also that "all thing being equal" applies: state polls use the same methodology across the board (and the same as national polls).
However, we have noted that there have been serious problems with over and undersampling in national polls and also in using historical data from 2004 that is inaccurate as applied to 2008 (youth vote, e.g.). But can these problems also exist in state polls?

[NYC_SKP]

7. I realized that after I replied.

I'd say that yes, these problems can also exist in state polls.

I don't think there is any single great answer to your question, there are too many variables among the entities which conduct polls (universities, media, pollsters, etc.).


Every four years our culture changes more than it did during the prior four years and it has to make it tougher for pollsters to be accurate. When oversampling, undersampling, and other strategies are used and based on four year old assumptions, they will be problems.

But I know of no particular differences in strategies between national and state-wide polls that would make differences.

:shrug:

[CTyankee]

8. In your opinion, which of the types of state polls do you consider better designed?

You mentioned universities, media and pollsters. So an entity such as Quinnipiac could be superior to Rasmussen, or vice versa, just to throw two examples out there?

[suston96]

9. Here.........

http://www.electoral-vote.com/

Interact with each state to check latest poll.

Since the national popular vote doesn't count, the national daily tracking polls have to be considered in that light.

It's the aggregate of the state electoral numbers that have some perspective. And even that cannot be taken seriously.

More:

http://www.pollster.com/polls/

http://www.realclearpolitics.com/epolls/maps/obama_vs_mccain/

http://www.fivethirtyeight.com/

National daily tracking polls are a peculiar feature that average several days of polling and then arrive at a "national" figure for that date. But, I repeat, this is not a national popular vote election.