Intersting fact about polling

[donneyp]

A lot of people are proclaiming this race over because the polls tell them its over.

I was checking some numbers and thought this was interesting.

In 2004, the final CNN/USAT/Gallup poll predicted

-Kerry would win Ohio by 4%
-Kerry would win Florida by 3%
-Bush would win Pennsylvania by 2%

But in actuality
- Bush won Ohio by 2%
- Bush won Florida by 5%
- Kerry won Pennsylvania by 4%

This year? CNN probably has 1 more poll to go so stay tuned.

[Phaedrus]

In conudcting a poll, the poll takers must calculate the margin of error, this is based on the assumptions they made about the polling practice, the sample space they are using, take into account the obvious biases in the sample space etc. The margin of error is usually stated at the bottom of the poll numbers to remind the people reading about what being in the weeds actually mean.

Obviously, the more samples taken means more accuracy, but is more samples are taken in a biased sample space, the statistics doesn't bear out, hence we have the Bradley effect, in reference to the Governor race in California when Mayor Tom Bradley was predicted to win the Governorship and instead was trounced. The initial explanation was that the Caucasian voters said that they would vote for Bradley but when alone in the privacy of the voting booth, they decided against voting for a black man. In actuality, the Bradley organization did a very poor job of assessing possible biases and actually neglected to sample a large swath of California which gave them the wrong results. This is why the pollsters get paid big bucks, because they understand this stuff.

Now, as for the CNN poll, most sane people will take the numbers with a grain of salt. The people on the inside have their own opinions on the veracity of the poll taking organizations. By and large they try to be as fair as possible, some organizations will skew their results to reflect their own biases, Fox News comes to mind.

New York Times has resorted to taking the top polls and averaging their results and coming up with a combined poll. It isn't any more or less accurate than the component polls but it does tend to nullify the peaks and valleys and give a more sensible set of numbers.

So the truth is somewhere in the middle and there is no sense in believing the polls as a predictor of the political races until the votes have been counted, correctly of course.

The statistics lessons on polling could also be extrapolated to volleyball statistics, there are biases and variations and randomness which does not appear on game stats. Which is pretty contrary to the founding assumption of the Cauldron crowd.

[OverAndUnder]

The fact is, rich people, old people, white people, suburbanites, and strongly religious people, show up to the polls.

College students, the poor, and urban minorities, talk talk talk talk talk talk talk but year after year they don't actually vote when the day arrives.

A picture is worth a thousand words but an uncast vote ain't worth %*$#, pardon my French.

[bunnywailer]

Fundamental flaw in any of the polls being conducted is that they can only draw a representative sample of people to poll from those who are willing to be polled. There is too much of an unknown because there are that many other people who do not wish to be polled prior to the election.

Also, I was surprised myself at how small the actual sample sizes are for these polls. Alot of them only have between 2,000 and 3,000 people responding. I don't know how they arrive at their margins for error, but I don't know how you can accurately assess a state's voting trend when there may be a couple million people voting and you only polled 2000 of them.

But then again, I did get a C- in my statistics class in college, so what do I know?

[bunnywailer]

Oct 31, 2008 19:26:53 GMT -5 OverAndUnder said:

The fact is, rich people, old people, white people, suburbanites, and strongly religious people, show up to the polls.

College students, the poor, and urban minorities, talk talk talk talk talk talk talk but year after year they don't actually vote when the day arrives.

A picture is worth a thousand words but an uncast vote ain't worth %*$#, pardon my French.

In all honesty, if there were one election where all of these demographics DO get off their asses and actually go vote, this will be the one.

That does bode favorably for Obama/Biden.

Conversely, do we know which way the silent and undecided voters will sway?

[TheSantaBarbarian]

The largest betting joint in Ireland is so sure of an Obama win that they have already paid off. Of course, they were also so sure that Ireland would vote to join the EU that they paid off early on that as well, and then had to pay off again when the EU vote failed...

CNN had been running various scenarios on the Bradley effect and where the final undecided will go. The pundents seem to think that the undecideds will probably break toward McCain, but that may be offset by increased voting by young voters and blacks. Bottom line, they have no idea. I really hope we don't have another 2000, we really can't afford that right now.

[TheSantaBarbarian]

Anyone know what the Vegas odds are right now?

[bunnywailer]

Sad thing is that if Gore hadn't listened to his party's advice in 2000 and conceded that election, he woulda been a shoo-in for 2004 and probably running for re-election this coming week.

[donneyp]

Nov 1, 2008 3:03:40 GMT -5 @bunnywailer said:

Sad thing is that if Gore hadn't listened to his party's advice in 2000 and conceded that election, he woulda been a shoo-in for 2004 and probably running for re-election this coming week.

Don't feel bad for Gore. After he unconceded he had a call with his advisors. Bill Galston told him basically what you propose now, and his advice was not solicited again.