Just how accurate are the polls?

[techs]

http://www.fivethirtyeight.com...bel/pollster%20ratings

I never saw anything like this. A list of polling companies and how accurate they have been.
Pretty interesting.
CBS/New York Times and Zogby Interactive (regular Zogby in the middle) near the bottom.
SurveyUSA, Rassmussen and Mason Dixon near the top.

 

[AstroManLuca]

I saw an interview with the guy who runs that site on the Colbert Report. You can watch it online; I think it's the October 2nd episode.

He's a baseball statistician. His prediction: the Cubs won't ever win the World Series.

 

[TheSlamma]

Well, being the latest elections have not been accurate how can these be?

 

[techs]

Originally posted by: AstroManLuca
I saw an interview with the guy who runs that site on the Colbert Report. You can watch it online; I think it's the October 2nd episode.

He's a baseball statistician. His prediction: the Cubs won't ever win the World Series.

Heck, I coulda told ya that!

 

[techs]

Originally posted by: TheSlamma
Well, being the latest elections have not been accurate how can these be?

Actually not bad in 2004 Presidential


http://www.realclearpolitics.com/bush_vs_kerry.html
http://www.realclearpolitics.c...toral_Count_Final.html

 

[ProfJohn]

They tend to be correct on election day.

But this far out it is hard to tell.
Something could change the direction of the election and McCain could rebound or Obama could solidify his lead.

I guess you could say that they are a good predictor of who is winning today, but not a good predictor of who will win on election day.

 

[AstroManLuca]

Originally posted by: ProfJohn
They tend to be correct on election day.

But this far out it is hard to tell.
Something could change the direction of the election and McCain could rebound or Obama could solidify his lead.

I guess you could say that they are a good predictor of who is winning today, but not a good predictor of who will win on election day.

Exactly. What people forget is that polls don't tell you who is going to win. They tell you who would (probably) win if the election were held today. If the election were held today, Obama would almost certainly win. But like you said, a lot of things can change in the next few weeks.

The polls have never been really that far off. There are just people who jump the gun and declare victory the moment their guy takes a large lead. I have to admit it's tempting to say that an Obama win is a foregone conclusion, but it's not true. All the polls tell us is that there has to be a big shift in the next few weeks for McCain to win.

 

[heyheybooboo]

The issue with exit polling on election day in 2008 is that as much as one-third of the electorate will have already voted early.

 

[Eeezee]

This is pure counting statistics. If they polled 10,000 people, their error would be plus or minus 100, or 1%.

Reducing that error from 1% to 0.1% costs about a factor of 10. This is why we sit with errors ranging from 0.5% to 8%. It all depends on how much money the polling company is willing to dump into reducing their error bars. In most cases an error of 2% is fine because the race isn't very tight. In the 2004 and 2000 elections, however, even a 1% error was bad.

And that's only one standard deviation, meaning the uncertainty has uncertainty, too

 

[Craig234]

Originally posted by: Eeezee
This is pure counting statistics. If they polled 10,000 people, their error would be plus or minus 100, or 1%.

Reducing that error from 1% to 0.1% costs about a factor of 10. This is why we sit with errors ranging from 0.5% to 8%. It all depends on how much money the polling company is willing to dump into reducing their error bars. In most cases an error of 2% is fine because the race isn't very tight. In the 2004 and 2000 elections, however, even a 1% error was bad.

And that's only one standard deviation, meaning the uncertainty has uncertainty, too

The polls are accurate at measuring public opinion. They're not accurate at predicting at how opinion will change by the election, and not too great at who will actually vote.

 

[chess9]

Rasmussen is very solid.

Realclearpolitics is very good because they average the latest, what, 10 polls? That may give a modest statistical boost to the polls' validity.

Right now, Obama is winning in all the polls and has been for about 10 straight days. Which is why McCain and Palin have turned into raging beasts. They intend to lose ugly.

-Robert

 

[AstroManLuca]

Originally posted by: chess9
Rasmussen is very solid.

Realclearpolitics is very good because they average the latest, what, 10 polls? That may give a modest statistical boost to the polls' validity.

Right now, Obama is winning in all the polls and has been for about 10 straight days. Which is why McCain and Palin have turned into raging beasts. They intend to lose ugly.

-Robert

Actually, the FiveThirtyEight guy sort of ripped into RCP for not disclosing their methodology. But that's his word against theirs. I am sure RCP's methodology is at least fairly sound.

 

[Don Vito Corleone]

We're definitely still a ways out, but unless the economy somehow magically self-corrects in the next 3 weeks, which is obviously impossible, or something strange and catastrophic happens, I don't see how McCain can bridge the gap. I think this one is all but over. We shall see . . .

 

[chess9]

Originally posted by: AstroManLuca

Originally posted by: chess9
Rasmussen is very solid.

Realclearpolitics is very good because they average the latest, what, 10 polls? That may give a modest statistical boost to the polls' validity.

Right now, Obama is winning in all the polls and has been for about 10 straight days. Which is why McCain and Palin have turned into raging beasts. They intend to lose ugly.

-Robert

Actually, the FiveThirtyEight guy sort of ripped into RCP for not disclosing their methodology. But that's his word against theirs. I am sure RCP's methodology is at least fairly sound.

I've wondered about their methodology as well. One day Obama is 4.7 and the next 6.2. RCP appears to weight some polls higher than others. Also, they don't use The Daily Koz poll. EVER! Probably good....

-Robert

 

[piasabird]

I saw a report on polling and the changing demographics. More people are living in locations like myself where we do not maintain a land based phone, so we can not be polled. This report was on MSNBC. There were some reports from Stepenopholis or however his name is spelled. I cant find it now.

What they were also looking at was what happens with polling when the opponents are one black person and one white person. They found that black voters vote along the polls predictions but the White voters did not vote along the polls predictions. They also pointed out that households with no land based phones are on the rise. This means like 14% of people can't be polled. So they are waiting to see how well the polls will predict this election.

At first they were saying that 18-25 year olds tend to have only a cell phone. However, I am 50 and I only use my cell phone. And I had a coworker that was doing the same thing and he was over 50. So I was wondering what effect this has on polls.

Here is one article on Cell phone useage:
http://www.itu.int/ITU-D/ict/n...crease+In+The+USA.aspx

 

[Don Vito Corleone]

Originally posted by: piasabird
I saw a report on polling and the changing demographics. More people are living in locations like myself where we do not maintain a land based phone, so we can not be polled. This report was on MSNBC. There were some reports from Stepenopholis or however his name is spelled. I cant find it now.

What they were also looking at was what happens with polling when the opponents are one black person and one white person. They found that black voters vote along the polls predictions but the White voters did not vote along the polls predictions. They also pointed out that households with no land based phones are on the rise. This means like 14% of people can't be polled. So they are waiting to see how well the polls will predict this election.

At first they were saying that 18-25 year olds tend to have only a cell phone. However, I am 50 and I only use my cell phone. And I had a coworker that was doing the same thing and he was over 50. So I was wondering what effect this has on polls.

Here is one article on Cell phone useage:
http://www.itu.int/ITU-D/ict/n...crease+In+The+USA.aspx

In 2004 there was a lot of speculation that the number of cell-phone-only users would lead to underreporting of Kerry supporters, since cell-phone-only folks were disproportionally likely to be young and/or of lower income, making them predisposed to be Democrats (please don't take that as an insult - I have been cell-phone-only for nine years and am neither particularly young nor particularly low-income). Ultimately that didn't prove to be the case. This year MAY be different, in that black/younger voters may be underrepresented due to the cell phone factor - we shall see. As I said above, this election has all the earmarks of a landslide anyway, so it may be more an academic issue than anything to be concerned about.

 

[Drakkon]

http://www.pollster.com/blogs/...ds_in_2008_vs_04_a.php - interesting look at the trends during the 00 and 04 elections compared to the 08 ones.

 

[imported_six]

Thought the Iowa Election Markets is more accurate than any opinion poll???

 

[Jiggz]

As popularized by Mark Twain, "Lies, damned lies and statistics!" That's how accurate polls are.

 

[sportage]

Polls are pretty much on the mark.
Back in 2004 Bush was up like 1%.
Everyone thought these new "young" voters would
make the diff for Kerry. That never happened.
End results were close, with Bush winning like 1% popular.
I find it hard to believe the country is split almost exactly 50/50.

In this election, its the younger new voters again.
Its said these new registered are 60-80% democrat.
We'll see.

With the outright near violent anger at the Mccain rally?s,
if Mccain doesn?t win, they will turn on him and Palin for blowing it.
If I were Mccain, I'd have two plane tickets ready for Brazil on Nov 5th,
just in case.

They are not going to be mad at Obama for winning, they will be pissed at Mccain
for not winning.
Remember, even Rush once said he'd vote for Hillary if Mccain were the nominee.
There is this underlining republican "I told you so" anger just waiting to explode.
And explode it will?

 

[Craig234]

Originally posted by: Jiggz
As popularized by Mark Twain, "Lies, damned lies and statistics!" That's how accurate polls are.

Sigh. I thought the know-nothing party had disbanded. Welcome back.

Sorry, Palin is not available.

 

[shira]

Originally posted by: techs
http://www.fivethirtyeight.com...bel/pollster%20ratings

I never saw anything like this. A list of polling companies and how accurate they have been.
Pretty interesting.
CBS/New York Times and Zogby Interactive (regular Zogby in the middle) near the bottom.
SurveyUSA, Rassmussen and Mason Dixon near the top.

538 factors that rating (which is that pollster's average error = the average difference between the pollster's results and actual election results, after eliminating the unavoidable sampling errors associated with polling) into a weighting scheme used to incorporate a given poll result by the pollster into the overall picture for the state or nationally (538 also uses the size if the poll and how recently the poll was taken). They then use the results to run 10,000 simulations for a given contest, and indicate the percentage of the time they expect a given contest to be won by each candidate.