Dean's World

Despite some people's efforts to sneer at them, polls are scientifically valid and highly accurate, when in the hands of trained professionals using solid methodologies. Most polling does not meet those rigorous standards, but there are a dozen or so nationally recognized polling outfits who consistently do good, solid, reliable work--although even then, you should only trust their polls if they also show you the underlying data behind them. Transparency is the key to understanding whether a poll is worthwhile or just stupid fluff. (So is the willingness to put aside what you want to believe versus whatever the poll actually shows.)

The polling company I look to most often is Gallup. They're the oldest and best known. I remember growing up as a kid hearing about this or that "gallop poll" in the news, and I always wondered what kind of poll galloped. I remember when it finally clicked in my brain that a guy named George Gallup had founded the company that did these polls. Then I finally got it.

Other polling companies whose work I generally follow and find reliable are Harris, Zogby, Pew Research, Mason-Dixon, Rasmussen, and Marist. There are a few others but I mostly don't watch them. But even for those that I find most accurate, I repeat what I've said above: you should always look for the underlying data before you simply accept what the top line of the poll tells you, because how a question is asked and what their methods are always matter. (For example, polls taken over the weekend are always more suspect than polls taken on weekdays, at least in political matters.)

In recent years one particular polling company has garnered a reputation for being more accurate than any of the others: Zogby. John Zogby runs that polling outfit, and in 1998 he turned in some impressively accurate polls on the congressional elections. He also seemed closer than anybody in predicting the outcome of the 2000 Presidential election. Since then he's ridden a tide of almost mystical worship among some poll watchers.

However, the folks at Kerry Spot have a comparison of how the various polling houses did in 2000, and note that Zogby's performance was excellent but not really heads and shoulders above anyone else in 2000. And Daly Thoughts notes who the most and least accurate pollsters were in 2002. The most accurate was Mason-Dixon. Three guesses as to who the least accurate was?

None of this of course means that any of these polling houses are bad by any means. Zogby's had some terrific years and some not-so-terrific years, but always does solid, professional work. Still, anyone who tries to tell you that Zogby, or anyone else, is the "best in the business" should be automatically suspect. A scientific poll done by a reputable professional organization is usually valid, period, no matter whose name is attached to it. But they're not omniscient, they're not perfect, and they all have a margin of error and a confidence interval, neither of which is ever perfect.

Me? If I want to watch polls, I usually look at Gallup. I peek in on the others. Gallup's always got the most interesting variety of stuff up, and usually has lots of details. They also have terrific videos that look at historical data and trends that's always fun to look through.

It also tends to be sobering stuff, by the way: no matter how convinced you are that you're right about an issue, it's always humbling to see that 60 or 70% or more of the population just plain doesn't see it your way, no matter how much that bugs you. But polls are scientifically valid, at least when done right, and by learning about them and following them, you can learn a lot about how to accept when you're in the minority on something--and to think about what the best way to persuade large numbers of people to change their minds might really be.

Now I'll reverse myself on one point: Oddly enough, although I utterly stand by the general validity of scientific polling, I'll actually say that I think this year is probably going to be a bad year for the pollsters. Why? Too many unusual trends. Both major parties have embarked upon unprecedented efforts to register new voters, and it's not clear who's registering more, nor is it clear that the polling companies have a handle on that either. The Democrats are also spending more money than they ever have on getting their voters to the polls, but Republicans are doing things they've never done in that regard too. Plus, a growing number of voters are voting absentee now, and so there are millions of people who have already voted and can't change their minds.

In other words, while the polling companies are usually pretty good, none of them are entirely sure what they're measuring at the moment or exactly how representative it is of who'll show up on election day. It'll be interesting to see which polling company and (more importantly) which methodologies produce the best results this year.

But in any case, the polls released on Halloween this year will probably give us the most accurate numbers possible: they'll be the ones taken during the last full work week before election day. But then, of course, there may be a surprise over the weekend, like the Bush DUI story that Fox News broke four years ago on that same fateful weekend before the election.

Man November 2 can't get here fast enough for me. I still just want it to be over. But like a tongue poking incessantly at an aching tooth, I just can't stop constantly looking to see what's happening. It's a sickness, I tell ya.