Friday, January 11, 2008

Why Polls Are Always Wrong

On the Jan 9, 2008 edition of "A Daily Show with Jon Stewart", guest John Zogby, of the famed Zogby polling, was asked if pre-election polling actually affected the outcome of the election and he answered,"I don't know." Is it unreasonable to think that people will or will not support a particular candidate because polls show the person they were thinking of throwing support to was going to win anyway? In 2000, I voted for Ralph Nader because even without the benefit of polls, I knew that my home state, New York, was going to go to Al Gore. So I decided to try to help a third party get on the ballots by getting at least five percent of the national vote. We came up short, and Al Gore, "as predicted", won New York anyway. I am hoping that more people understand the benefit of voting this way. If your candidate wins with 70% of the vote, all of those people who were above 51% could have done the same thing I did and their candidate would still have won. If more parties can get to field candidates without having to get signatures every year (the benefit of getting at least five percent of the national vote), then our choices won't continue to be limited to the goddamn Democratic Party and the goddamn Republican Party. (Yes, I'm a liberal libertarian, but I despise both of those national parties. Local politics are different.)

As everyone who follows the polls knows by now, none of the major pollsters got it right in New Hampshire, certainly as far as Democrats are concerned. And there's a good reason for that. Political polling as a means of predicting the eventual winner of an election does not work! The theory is flawed. Stewart asked Zogby how many phone calls he would have to make to get the 850 to 900 people he uses for the poll, and Zogby answered, "About six to seven thousand." Suppose it was 6,300 calls to get 900 responses. That means that six out of seven people were not asked about their opinion on that particular question or issue. So how do they make the leap that the one out of seven that they do question is representative of the rest of us? How do they know, for example, how people who screen their calls with an answering machine will vote, or how people who don't like being interrupted to take a poll will vote, or how people who don't believe that polls are accurate will vote? They don't because they can't. How could they prove that they are "just like the people" who do give their time to a pollster (on the assumed pretense that they will give truthful and non-evasive answers)? How much do they weigh the respondent's qualifications to answer a question when taking results? Do they ask every person how politically aware they are? Or do they just ask for whom they'd like to vote (and if they plan to vote)?

Another reason for flawed results is that there are, apparently, a lot of people who don't make up their minds until the day they vote. Zogby said that about 18% of the voters made up their minds on Election Day. That's a lot of people (about 2 out of 11). Is it any wonder that they couldn't accurately predict the outcome? And since this was being heralded as a "Three Horse Race" by the supposedly unbiased media (both left and right), it was unclear to the pundits who would get the "anti-vote" of any of the candidates. Would anti-Clintons throw their support for Obama or Edwards? That would depend, of course, on why they didn't like Clinton. Would anti-Obama voters support Clinton or Edwards? It is hard to imagine that anyone could be so against Sen Obama as people are quite clearly so against Sen Clinton. Personally, I feel that a lot of the hate directed at her is unwarranted and unfounded. Many of the talking heads I hear flinging poo in her direction are using debunked and irrelevant talking points in an often inarticulate effort to do no more than continue to foment hatred toward her. While there may certainly be good reason to dislike Sen Clinton, it's probably not for any of the reasons you've heard. Unfortunately, because little is explored in depth on these types of shows very often (Keith Olbermann, one of my favorites, being an obvious exception), people too lazy to cross verify what they hear on TV continue to believe the lies. And because they believe the lies, they believe the innuendo about her and what she might do once inaugurated as our 44th President of the United States.

Oh, wait, sorry. You mean she hasn't actually won? You could have fooled me. It sure sounds like it on TV. The entire Primary Process is completely different from any other election year because of all the changes to the voting dates. Much of their theory about who will do well after each primary is based on the past performance of previous winners of those primaries. But those primaries often had days or weeks more time between them, when a lot of things can happen to a candidate, than they do this year. So how can they be certain that the "winner" in New Hampshire is going to perform a certain way in upcoming primaries that now come sooner than before? Seriously, there is absolutely no reason whatsoever to presume that any historical pattern will repeat in this election cycle. With the way these primaries are coming, the nominees of each party will be decided in plenty of time for them to screw up badly and an independent candidate gets on the ballots in all fifty states. It's not too late for that to happen. Don't say I didn't warn you.

And the analysis you hear all the time from the political talking heads is really just a bunch of nonsense. Look at what happened when Sen Clinton beat Sen Obama in the New Hampshire primary. Suddenly it was a "stunning upset" because all the polls showed that Sen Obama would win, some by double-digit figures. Could it be that their interpretation of the polling results was in error because their theory is flawed? I mean, do they ever "predict" the accurate results ahead of time? No. I know what you're going to say. "Wayne, you ignorant slut. That's what the Margin of Error is for." Oh, really? Then how come election results often fall outside the so-called "margin of error"? (It happens.) And do know why that is? The answer is mind-bendingly simple. The difference between what the polls predicted and how the election actually case out is not what the "Margin of Error" you hear so much is measuring. It has nothing whatsoever to do with that. It is not the percent the results may vary, it's the probability that your sample group (those 900 voters who answered the poll)is not representative of the actual group as a whole (them plus the 5400 people who did not take part in the poll). And even that is based on something that doesn't apply to actual political pollees: It is presumed that the group being sampled falls into a kind of general, bell-shaped graph of results (whatever they may be) where most things are more-or-less alike and there's a small number of things on either extreme. You've all seen them. It's fine for analyzing things that actually did or did not happen. They are facts. And they do generally fit the bell curve graph (for the most part). But the opinions of ill-informed people is no sound way to predict how they'll vote. People often base their opinion on things that are not true. And there's no telling when they will learn the truth and change their mind about the people they always supported before.

People, it's time to put an end to the belief that political polls can tell you anything about the outcome of an election. They can't. Not with any degree of certainty that is worth investing in, because they are flawed from the get-go. The best thing you can do is trust your own feelings about a candidate and what he or she stands for, and vote for the candidate who YOU think best represents your views. That is what Democracy is all about.


Anonymous said...

Good points, but you neglect one obvious possibility. That the polling techniques aren't wrong or inaccurate... but the vote talleys.

That might be one reason that Kucinich has spent $27,000 of his own money to pay for the recount.

With estimates already out there that the paper ballots don't match the e-voting tabulations, my guess would be it's money well spent.

But don't expect to see it on the nightly news, or your buddy Keith-O.

Of course,the pollsters from NH are already positioning themselves for the explination saying some data cards are "missing" and others may have already been erased... and that would, you see, explain the discrepancy in the numbers; can't really check if they don't have all of the evidence to add up, now can they?

Oh, and if anybody here is on the bubble about what you think Hillary and Bill are capable of in the White House again, you just come by and see me. I don't deal in exagerations or hype.


Anonymous said...

I hate to double dip but... is a comment and a link to a recent study done on the Ohio 2004 fraud. You know, one of the ones that also didn't match the polls, and, was the last state to report in (even after Alaska) and gave the race to Bush?... You know, that Ohio...

"Ohio Secretary of State Jennifer Brunner’s muchanticipated
study found
Ohio’s election system riddled with “critical security failures.” Released in
December 2007, after attempts by the Republicandominated
legislature to
block the funding, the review showed massive problems involved with
electronic voting systems."