Are the presidential race polls accurate?

[Almighty Colin]

REFERENCE: http://www.ncpp.org/1936-2000.htm

For the popular vote:

Average error of major polls for US presidential elections post 1980.

2000 1.1%
1996 2.1%
1992 1.1%
1988 1.6%
1984 2.2%
1980 3.1%

As is typical of such things, averaging the results gives quite accurate results.

And for the coming election if you're interested;

Zogby Even
CBS Bush +2
Harris Kerry +1
Gallup Bush +4
Pew Even
IBD/TIPP Even
ICR Bush +3
NBC Bush +2
ABC Bush +1
Battleground Bush +2

Average forecast is Bush +1.3.

Now if the average margin of victory error is 3.8% (double the error per candidate) for the above 6 elections holds then the range of the average error would be Kerry +2.5% to Bush 5.1% giving Bush about a 67% probability of winning the popular vote.

But of course it's really about the electorals and that's damned close too.

[Rolo]

I think they are somewhat accurate - not perfect, but something like a weather forecast, which also builds on a not perfect science.

The thing which I find a bit troubling is how much attention and influence the polls get the closer it is to the election day - people deciding not to vote, because they think their candidate is either a sure winner or a sure loser, so why bother... this can make the final result inaccurate.

Now ofcurse this is also part of the game, and have been for decades, so its already a part of "the system".

[Mike AI]

I do not put much faith in polls.....

[RawAlex]

Colin, because of the electoral college system, overall voting trends are not as meaningful in a tight race. As was shown last time out, they guy with the most votes didn't end up as president.

The question will be how those swing states vote. Where does Florida end up, that sort of thing. That is where the numbers really count, and that requires 50+ seperate polls, not just one.

I haven't had to figure it out before, but how many people have to be contacted to get a 1% error rate (19 times out 20!)?

Alex

[Evil Chris]

I'm sure that the margin of error for predicting how many will actually come out and vote is a lot higher and tougher to predict as well. This makes the polls even harder to read.

I have a hunch more people will be out voting this year in the US.

only the vote matters, the real poll of significance

[Almighty Colin]

Alex,

Yes, what you said it is true. It's all about the electoral votes, but I already said that.

[Almighty Colin]

Originally posted by Evil Chris@Oct 30 2004, 09:36 AM
I'm sure that the margin of error for predicting how many will actually come out and vote is a lot higher and tougher to predict as well. This makes the polls even harder to read.

I have a hunch more people will be out voting this year in the US.

That's why Gallup started including the whole "likely voter" scenario. I believe that came about after the failed Dewey/Truman poll. There have been a series of improvements at Gallup over the years that have lead to better sampling and smaller errors.

[Almighty Colin]

Originally posted by RawAlex@Oct 30 2004, 09:34 AM
I haven't had to figure it out before, but how many people have to be contacted to get a 1% error rate (19 times out 20!)?

95% is two standard deviations so you need one standard deviation to be 1/2%.

sqrt(x)/x = .005
x^-.5 = 5/1000
1/x = (5/1000)^2
x = 40,000

[Peaches]

Shoot, we have problems figuring out who won after the actual election - I don't put much trust in polls.

[Almighty Colin]

Originally posted by Peaches@Oct 30 2004, 05:12 PM
Shoot, we have problems figuring out who won after the actual election - I don't put much trust in polls.

The polls can't be wrong. They just define the probability of certain things happening. In the long run, those probabilities have been accurate.

[OldJeff]

There is a major flaw in the sampling they are taking this year.

There is a huge increase in young voter registration (with good reason)

All you need to do is think about who is being polled to see the error.

College kids are not being polled - they are busy with classrooms or parties, your average blue collar joe is not being polled, he/she is too busy putting in the OT to make ends meet.

They are polling retirees and soccor moms with nothing better to do, your typical republican base, many who are in that 18% of Americans believing they are in the top 1% of income earners that the Bush Tax cut actually had any meaning for.

I feel the race comes down to this.

If there is large voter apathy (like there is so oftn) then Bush will win - your typical Senior Citizen vote that remembers the good old days.

A large voter turnout will give the election to Kerry, especially if there is a large percent of 18-25 year olds. Even though I don't think there will be a Draft, the "belief" of one happening should be enough to push those that are draft eligible to the polls to vote against Bush.

[Almighty Colin]

Originally posted by OldJeff@Oct 31 2004, 06:42 AM
There is a major flaw in the sampling they are taking this year.

There is a huge increase in young voter registration (with good reason)

All you need to do is think about who is being polled to see the error.

College kids are not being polled - they are busy with classrooms or parties, your average blue collar joe is not being polled, he/she is too busy putting in the OT to make ends meet.

They are polling retirees and soccor moms with nothing better to do, your typical republican base, many who are in that 18% of Americans believing they are in the top 1% of income earners that the Bush Tax cut actually had any meaning for.

I feel the race comes down to this.

I've seen this argument largely on sites in support of Democrats. The counterargument being used by Republicans is that the methodology in some polls assumes that there will be 60% voter turnout and since this hasn't happened since the 1960s Bush's votes are likely being undercounted in the current polls. I think both arguments are ad hominem. Everyone sees what they want to see. There are few Republicans using the young voters underrepresented argument and few Democrats using the "60% voter turnout" argument.

I agree with you that more voters means better odds for Kerry. Republicans vote in proportionally higher numbers and the strategy for Democrats is always motivating their majority to vote and doing it in the right states.

Modern polling methodology is pretty good at handling the changes you mention though. The Harris poll weights for age, income, and race for example. (They have Kerry 48%, Bush 47%). Many of the other polls use similar methodology.

I don't think there is any reason at all to assume this election won't be extremely close determined more by random factors such as whether people felt like standing in line that day, whether they had a good or bad day at work, whether traffic is bad, whether they had headaches, whether they got in car accidents, whether their football teams won this weekend, whether they got laid and so on.

[Peaches]

Originally posted by OldJeff@Oct 31 2004, 07:42 AM
Even though I don't think there will be a Draft, the "belief" of one happening should be enough to push those that are draft eligible to the polls to vote against Bush.

That's pretty sad that they'd vote so completely and totally uninformed Maybe there should be a test before you vote that sees if you understand at one or two of the issues.

[Almighty Colin]

Originally posted by Peaches+Oct 31 2004, 07:58 AM-->