Saturday, October 09, 2004

More About Polls

Many of my readers know that I am infatuated with polls. A number of readers have sent me mail, asking how to tell what a poll is worth, and which polls I like the best. Those who read my stuff over on Polipundit, also know that I ripped into Ipsos this week. So, in this article I want to talk about polls in general, and revisit the question of where I see the race.

First off, I can’t say I really have a favorite poll, or a most-accurate. The reason I say that, is that every poll should be able to stand up on its own information, so that no one gets a pass, and every poll gets a chance for consideration. What I look for, is details, especially demographics and party support, consistency from report to report, and a sound methodology. I don’t insist on a specific weighting or position, because if the data is complete, I can test the results for myself. I find Gallup and CBS/NYT win my respect regularly on those aspects, but I still check out their claims to be sure.

Now, before I go on, I want to balance out my comments on the Ipsos poll a little bit. Over on Polipundit, I ripped them down pretty harshly, because they totally ignored basic points of poll methodology in the polls put out for Newsweek and the AP. But, even so, their poll has its value, because the information is valuable for noting certain additional facts. You see, their respondent poll had only 49 self-described Independents, a little over 5%. That means that the people watching the debate last Thursday, were overwhelmingly partisan going in. It means that almost nobody was going to change their mind from what they saw or heard in that debate; some people might have slipped in their support for Bush a little, or some Kerry supporters might have gained confidence. But there is no substantive reason to expect major opinion shift from that debate, or the other debates. Also, the race has basically come down to getting out the vote; if Kerry can match Bush in getting his supporters to vote, he can win. But if Bush shows the kind of confidence and optimism we saw in early September, Kerry’s soft followers will fall off again.

Back to polls in general. One of the most irritating thing in trying to see who’s best, is that there really is no good standard for the measure right now. In the first place, a lot of the polls out there are pretty new, so they don’t have a track record for Presidential Elections. But even the ones which do, are judged by one element only: their final poll versus the actual Popular Vote in the election. And that can be deceiving. For instance, the Zogby poll, whose owner thinks far more highly of himself than he has any right to claim, got within 2 points of the PV in 2000. Sounds good, but most people don’t know that Zogby moved 8 points in the final two days of the campaign, which is not supported by any evidence in the public mood at the time, and which makes no sense as methodology. In other words, Zogby was faking out his readers on purpose for most of 2000, or else he was simply careless for most of the run, and he made a quick change at the very end to be able to claim he was accurate. Don’t be fooled. There is no scoreboard to prove someone is better than everyone else; you just have to do your homework.

There are basically three components to a poll. The numbers which say someone is ahead and job approval, the questions related to the key issues, and the breakdown of demographics. Not every poll asks the same questions, and not every poll presents its details. I don’t want to be harsh, but while there are some interesting polls which don’t show their work, I strongly recommend insisting a poll show the details before you buy into it. Otherwise, a poll is just ego candy.

Also, when you read a poll, that you consider who is putting the thing out. There are partisan outfits who have a slant to their position, and there are companies which put out a poll, but really don’t work at it. Make sure you are comfortable with how they work their poll, before you include it in your calculations. For those who want my opinion in general, please review my earlier article on classes of opinion polls.


No comments: