Can polling REALLY predict the election?

I keep reading all these articles about the many different polls that predict Obama to destroy McCain. The thing is, I don’t think they are really a good indicator of what will happen on election day. The reason is that it’s extremely easy for me to answer the phone while sitting in my recliner, answer a yes or no question and hang up. It’s an entirely different animal getting me to drive to a polling location to actually vote (if I’m even registered).

I don’t see how the polls can really mean anything other than popular opinion. I suspect that Obama is showing so strongly in polls because of the wow factor of voting for a black guy or people wanting change or whatever and it’s way easy to do that over the phone. I’ll bet though that there are alot more old timer republicans that will actually go out and vote than there are young new democrats that will do the same.

How reliable do you think the polls are? Weren’t both Kerry and Gore ahead in many polls before the previous elections?

I know www.fivethirtyeight.com weights the polls based on how accurate they have been in the past. They do not simply take the polls at fact value. Many TV shows simply show the poll or take a poll of polls and leave it at that. Not all polls are created equal.

Didn’t work in 1948.

Or 2000…

Electoral-vote.com has a button that lets you compare today’s projections with the projections from this day four years ago.

Obama is definitely looking very different from where Kerry was at this point in the election in 2004.

Are you saying Bush was leading in the polls because there is little dispute Gore won the popular vote, even if he did lose the electoral vote. I honestly don’t remember what the polls said back then. If the polls had Gore leading nationally, then they “worked.”

What he said.

There was a recent GD thread which seemed to be nothing but falling over itself in support for fivethirtyeight.com among statisticians.

Ann Coulter says they can’t

There are corrections that can be made based on the assumption that not everyone is equally likely to vote, and good pollsters will apply them and take similar measures to deal with other issues. But that doesn’t change the fact that even the best designed poll out there has a margin of error, and that that margin of error represents a 95% confidence interval. It’s our best attempt to predict what will happen, but not one that’s guaranteed to be exactly right all the time. And I trust Ann Coulter’s statistical skills about as much as I trust…well, something I don’t trust at all.

I used to conduct research for a univeristy and the thing with questions is in order to get the correct answers you have to ask SPECIFIC EXACT questions.

You may ask one poll “Of those people LIKELY to vote.” The same poll should be repeated with a similar question “of people who DID vote in the last two elections.”

You need to ask RANDOM people, which is the biggest issue. If you’re at a commuter depot, more people are headed from the suburbs which skew conservative.

So if you get a truly random poll, with specific question and repeated you can get results.

In 1948 the biggest issue was Truman was so far behind, the pollsters stopped taking accurate polls too early. Truman’s real surge came in the last month when people had written him off.

Finally race can play an issue. No one is gonna admit they are a racist, but when it comes time to vote, and no one can see, it’s a different story.

One last thing is the way the ballots presented. Illinois used to have a provision where you could vote a straight ticket. That helped a lot of people. Illinois doesn’t have it, but some states still might.

One poll cannot, but thousands of polls analyzed correctly can. While 2000 and 2004 were close down the stretch and could have gone either way, Obama’s lead is well outside the margin of error. All the objective evidence says Obama is going to win, and going to win handily.

There is no evidence that the “Bradley effect” still exists. People who wouldn’t vote for a black man will simply tell the pollster that they will vote for McCain.

The key is to look for the small print. A poll should show both it’s margin of error (+/- in percent) and it’s confidence factor (the percentage of times the same polling methods come up with the same result.)

A poll with a 3% margin of error and a 96% confidence factor means that if they run the numbers an inifinite number of times, they’ll come up with a result no more than 3% different.

In other words, the better the methodology, the more likely the results will be accurate, but nothing is ever 100% accurate 100% of the time.

The other thing to watch out for is that polls can only reflect public opinion at the time they are taken. If a poll is taken on a Friday and the numbers crunched over the weekend, while on Saturday candidate X is caught in bed with a live goat, there’s no way the poll is going to catch the resulting change in opinion even if it’s not actually released until Monday.

I don’t know about Gore–I wasn’t paying attention in 2000–but Kerry wasn’t. As already mentioned, electoral-vote.com allows you to compare today’s results to results posted 4 years ago. I’m not sure why so many people have convinced themselves that Kerry was ahead. Wishful thinking? They weren’t paying attention? It’s just the sort of evidence that feeds into the “Democrats can’t win elections” myth and so is bought wholesale? I mean, on Oct 22, 2004, Bush had a 271-257 lead over Kerry, and while it fluctuated after that, he never really lost his edge.

And Gore won the popular vote and damned near won the electoral vote.

So the two cases you cite in support of your doubt actually prove the exact opposite.

Polls usually try to limit there sample to “likely voters”, they don’t just assume that everybody that says they like Obama/McCain will actually show up and cast their vote.