Don't dismiss polls you don't like (a.k.a. sticking your head in the sand)

Every day, at least a half-dozen threads are started here to discuss a newly released poll. More often than not (in fact, much more often than not), there will be a handful of posts that claim that the poll is bad, or that the pollster is generically unreliable or in the pocket of the Republican Party. In fact, if we were to believe this charge against every pollster that has earned it, then the following polling agencies would not be trusted:

-Rasmussen Reports
-Zogby International
-Mason Dixon Research and Polling
-Strategic Vision
-USA Today/Gallup
-Survey USA
-Quinnipiac University

Leaving us with virtually no nationally recognized pollster to "trust" on key races that we all are or should be paying attention to.

The symptoms are far worse than this, however. Whenever one of these pollsters (or a different one, it doesn't matter) publishes a poll that shows a Democrat up, its universally accepted and celebrating. Hell, people are popping open champagne bottles at the mention of a three-point lead in the Missouri Senate race. Two weeks later, the same exact pollster will conduct the same exact poll in the same exact state, and the Republican will be up by four points. Suddenly, the poll is unreliable and is being paid to shill for the GOP.

Consider the outrage when Quinnipiac University released a poll giving Joe Lieberman (DU's public enemy #1) a 17-point lead on Ned Lamont:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=102x2571775

Yet, today Quinnipiac released a poll showing a virtual tie between Jim Davis and Charlie Crist in the Florida Governor's race:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x2902791

Strangely, the people who insisted that Quinnipiac could not be trusted were no where to be found.

Recently, a spate of Mason Dixon polls were released that showed Democrats winning all key Senate races except for Tennessee and Virginia. Yet, Mason Dixon is being accused of being skewed to the right, even though they project the same 50-50 makeup in the Senate as everybody else. Consider the Mason Dixon poll released in Virginia:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x2902849

With the Mason Dixon poll released in Missouri:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x2902706

Oddly enough, you don't find any people complaining about Claire McCaskill having a three point lead over Jim Talent.

And I can't even count on two hands the number of times I was called a traitorous Lieberman supporter for believing every single poll that's come out that shows Lamont without a lead.

Only believing the polls that show Democrats ahead is as useful as not even looking at the polling numbers and assuming that the Democrats are winning. Polling numbers are an extremely valuable tool in determining the course of a campaign, and properly assessing the political landscape. Its a judgement on the past and an indicator of the future, and thus is invaluable in the present. You can't cherry-pick them, however, or its all meaningless. The people that stick their head in the sands and decry any poll that shows Lamont trailing are the same people who are the first to post conspiracy theories about the Lieberman victory "they never saw coming".

The greater folly of this process is that its a never-ending cycle. You don't believe poll numbers when they come out, and then you claim fraud when your candidate loses. And since you believe you lost the election because of fraud--and not the real reason of the voters supporting a different platform than the one you had to offer--then you are never challenged to move your platform to appeal to a larger base of people. And the more stagnant your positions come, the more you underperform at the polls, and the louder your calls of fraud become. And suddenly you're yelling about a Diebold conspiracy in states that don't even offer electronic voting as an option. This is the quickest route to the fringe (also known as political irrelevency) that I know of.

When using polls to look at a race, its important to look at all of the availible polling, even if you have doubts about its methodology. The difference between pollsters is often the samples they use; some oversample Republicans, while others undersample Independents, while still others use different criteria for determining who's a "likely voter". The most important thing to remember is that predictions cannot be proven wrong until Election Day. If a poll sample has 40% of Republicans, that poll is not false until less than 40% of Republicans show up at the poll on November 7th. Until then, dismissing the poll because of a percieved error in its sampling is inane.

Beyond that, even if sampling is consistently unsound, that consistency can be an important tool. For instance, Mason Dixon is accused of over-sampling Republicans. Which means that, in every poll they conduct, the amount of oversampling is roughly the same. So we can draw conclusions out of that. If they conducted a poll in late September that show Allen and Webb tied at 43%, and they released a poll today showing Allen with a four-point lead, even if those numbers are off, the trend is not. If Mason Dixon used the same methodology for each poll, then the poll still shows a four-point increase for George Allen, regardless of whether the numbers are perfectly accurate. That trending data is often more important than any given poll's numbers.

Finally, if you are going to criticize a poll, please make sure you know what you're talking about before you do it. I would bet a substantial amount of money that the credible news organizations that regularly cite polls know more about polling than you do. I would bet double the amount of money that pollsters know even more. Accusing them of a faulty methodology is like telling Albert Pujols that the mechanics of his swing are wrong.

There is only one poll that should be categorically dismissed, at least for now. John Zogby, one of the premiere pollsters in the United States, oversees two polling agencies: Zogby International and a new, innovate form of polling called Zogby Interactive. Zogby International remains one of the most trusted and reliable polling groups in the world. Zogby Interactive, however, is an experimental approach to polling that involves contacting users through email, and allowing them to be polled online. This method is not only untested, but it also has very serious flaws its surface when it comes to verification and sampling. Zogby is offering this service as a test of its effectiveness, rather than as a legitimate poll--at least for now. If Zogby can work out the kinks, interactive polling may be the way its done in the future. In the meantime, however, its numbers should not be counted at all.

It's all automated, which would seem to bias out a large number of people who speak tenuous English, and feel uncomfortable with the impersonal approach of a machine's commands in English. I think that filters out a lot of Democratic voters from the Rasmussen samples.

Especially when dealing with questions that people may want to impress the interviewer.

You are more likely to say you are voting on election day to a live person than to a automated call.

and it gives you a trend. You can spot the outliers easily. Only problem is, it ain't real time.

And my experience tells me that FauxSnooze polls are ALWAYS off by three to five percent, at a bare minimum, from every other poll that asks about the Monkey's job approval.

You're never going to get everyone to "agree" to vouch for or dismiss polls. Politics is a team sport, and people want to root for their team. Favorable polls are like a group of friendly fans, and people will cite them, because it strengthens their beliefs.

And. like it or not, some polls ARE bullshit, because someone's getting a little dough on the side, quite possibly, to make sure the results look as good as plausibly possible. If Frank Luntz is doing the polling, it's ALWAYS skewed to the GOP, because that's who pays him. If seventy one percent of the country wants the Monkey's head on a pike, he'll make it sixty-eight, just to cushion the blow.

When the GOP wins, it's because the country is fairly evenly divided, and there's room to cheat the system within the margin of error or maybe a point or two beyond it.

We'll have to see how they'll be able to swap votes out when the margins are well outside the MOE.

THEY change minute to minute and depend on WHO is doing the polling. 'WE' should have ALL learned that polls; NOT even exit polls; have ANY validity. Just GO VOTE and make sure you take a (preferably a Democrat) friend along so THEY can vote too.

the bottom line is people want change -- badly . . . Democrats want change, Independents want change, and even many Republicans want change . . . the vote this year will be against the party in power, and no matter what the polls say, it will be a Democratic landslide . . .

now, whether the votes are actually counted to reflect that is another story altogether . . .

There is nothing to indicate that there will be an overwhelming amount of turnout this year. That's not to say that there won't be, but there's nothing in the way of proof you can offer, except personal anecdotes. Then if things don't turn out the way you think--"whether the votes are counted to reflect that"--you blame it on election fraud, instead of recognizing the possibility that maybe people weren't motivated to vote this year.

tested under laboratory conditions or conform to legal standards of proof to be true and utterly compelling.

Given a choice between counting yard signs and trusting a poll, I'm going to trust a poll. Given a choice between a random sample of 600 people in my state and me asking five or six of my closest friends, I'm going to trust the random sample to provide a more accurate analysis of voter's behavior.

criterion.

However, election polls generally don't change "minute to minute". Even the close elections have followed a general pattern than different methodologies have found. You can track these patterns, no matter which pollster you use. Some examples:

In the Michigan Senate race, Stabenow started off with a large sized lead over Mike Bouchard. After Bouchard won the primary, and DeVos was at his peak, Bouchard closed to within five points. Now Bouchard has moved past his peak, and Stabenow is back in double-digits. All of the polls conducted in this race have followed this general trend.

Others:

-In Minnesota, Mark Kennedy and Amy Klobluchar started off neck-and-neck, and Klobluchar slowly pulled away, and now has a double-digit lead. Her lead started to generate in late July, and all availible polling reflects that.

-In Washington, Mike McGavick slowly closed in over the course of six months on Maria Cantwell, going from 20 points down to 4 points down. Then he hit a ceiling, and now Cantwell is back in a double-digit lead.

-In Virginia, George Allen was 8-10 points up on Jim Webb until Macacagate, then Webb pulled into a statistical tie with him. In the last six weeks, Allen has eeked out a slim 2-6 point lead over Webb.

-In Tennessee, Corker saturated the airwaves with his ads in the primary, and after winning the primary enjoyed a large lead over Harold Ford. However, in the next month, Ford surged and outcampaigned Corker for the lead. Now the momentum is slowly changing back into Corker's favor, but the two are statistically tied.

In just about every race this cycle, all the availible polling follows a singular trend for that race, no matter which poll it is. While its possibility that every single poll is off in a race, these polling institutions invest hundreds of thousands of dollars and years of practice into perfecting their system, and people wouldn't pay them and news sources wouldn't cite them if they didn't have some validity.

practicing and perfecting their system, will NEVER be able to predict the outcome of an election. While polling may enhance the drama of an election and be used heavily by the MSM just for THAT effect, polling will NEVER provide a high degree of accuracy in predicting the ACTUAL election results. Polling, with its constant changes over short periods of time, has little REAL meaning or validity.

First of all, most pollsters don't use their polling to predict the outcome of an election, they use it to get a sense of what people are responding to and how they are responding. Its a diagnosis of a campaign, not a prediction for November 7th. And that's why most polls ask "if the election were held today, would you vote for...?" instead of asking "Who are you voting for on November 7th?"

better than claimed by the pollsters.

..I know you mean well but the partisans do not like inconvenient facts.

And the way to deal with inconvenient facts is to destroy the messenger any way one can.

concerned beggars belief, doesn't it? Where's the faith, I ask? Where's the faith? Same with the voting machines. They're neutral, aren't they? Just metal and wires and plasma screeens, maybe.

And what kind of a suspicious mind would you need to think that pollsters might want to please their paymaster?

Personally, I'm most suspicious of all of good figures for the Democrats, which would be sure to be understated.

But the OP has a very salient point. Polls are ruthlessly attacked for bias etc usually with little basis in fact.

Partisans seem to live in a world of their own making.

I prefer reality.

It's obviously escaped the attention of you and your posse of kindred spirits on this thread that politics in the US is a blood sport. You try to insult our intelligence, but end up make fools of yourselves.

By starting this thread and reinforcing it in a little mutual-admiration society, you are the arch-partisans; the partisans par excellence. Name one other aspect of your elections that has not been corrupted by Republican fraud.

It is thoughtful.

However, I'd be seriously betting that in no way can a phone poll begin to approximate just who is truly going to show up to vote on November 7. Not in as polarized a place as America has become today.

We are going to show up to vote. We are mad about what has been done to our country.

And I equally believe that there is a great number of people out there who did vote for B*sh in 04 that would really love to have that vote back right now. Not that they would vote for a democrat - necessarily, but that they sure wouldn't vote republican.

For myself, I am quite certain that in the end, we will do much better than it currently appears. And it currently appears we do at least get back the House.

We'll get the senate too.

Joe

"The greater folly of this process is that its a never-ending cycle. You don't believe poll numbers when they come out, and then you claim fraud when your candidate loses. And since you believe you lost the election because of fraud--and not the real reason of the voters supporting a different platform than the one you had to offer--then you are never challenged to move your platform to appeal to a larger base of people. "

You NAILED it!

While election/electronic machine fraud is a problem that we must fight, the automatic fallback position whenever we lose seems to be "We got Diebolded" without any consideration that sometimes we just get fewer votes than the other side. But instead of figuring out why more people voted for the other side, we just continue on under the assumption that we did everything just right and but for evildoers and evildoings over which we have no control, we'd have come out on top. And then we repeat the same mistakes.

Thank you so much for explaining it so eloquently!

SurveyUSA and StarTribune. They are both far outside all other polls almost every time. SurveyUSA tends to go conservative more often than not and StarTribune tends to go left more often than not. There is a plethora of polls out there and when one is usually a complete outlier from all other polls, something isn't right. So those are two polls I disregard for favoring one side or the other.

However I still do look as to whether there is a trend within that same both because trends are the most important thing, especially in outlier-polls. If StarTribune shows Wetterling up by 3% when all other polls show the race tied, for example, then she increases her lead to 8% when all other polls show her up by 3-4%, it shows and confirms the upward trend that the other polls show.

But while I take what they have to say with a grain of salt, they've been ahead of some trends. For instance, the Star-Trib had Klobluchar up by 22 points when everyone else had the race in single-digits. It was derided from a number of sources--and deservedly so--but now that number doesn't seem so far off. In the same way, SUSA was the first polls to show a tightening in the Maryland Senate race after the primary, the Washington Senate race back in late spring, the Tennessee Senate race in late August, and so forth. While many times their numbers look out of "mainstream", other times the mainstream numbers end up reflecting what they had a month earlier.

Of course, its often times coincedence, but its enough for me to consider the polls, even if I don't believe them as much as I do other pollsters.