Home > Political Ranting > Or Maybe They’ve Just Got It Wrong

Or Maybe They’ve Just Got It Wrong

September 18th, 2004

In my immediately previous post, I opined about the unreliability of polls, but now with more information in hand, the issue may be more clear–it may be the “very different models” answer. As I had mentioned quickly in a post a week and a half ago, Rasmussen had explained the discrepancy between their polls and TIME and Newsweek polls by pointing out that those two news magazines weighted Republican voters far more greatly than Democrats. And now it looks like Gallup has been doing the same.

This post by Steve Soto at The Left Coaster (via DailyKos) points to specific data about Gallup’s polling model: they predict that Republicans will make up 40% of all voters, and Democrats 33%, and that’s reflected in their polling data. So no wonder Bush is ahead.

The next obvious question is, are those numbers accurate? Steve points to this quote from Zogby:

If we look at the three last Presidential elections, the spread was 34% Democrats, 34% Republicans and 33% Independents (in 1992 with Ross Perot in the race); 39% Democrats, 34% Republicans, and 27% Independents in 1996; and 39% Democrats, 35% Republicans and 26% Independents in 2000.

Read the whole Zogby article if you can, it discusses TIME and Newsweek in detail–just like the Rasmussen article.

But Zogby makes it pretty clear: in the last three elections, Democrats voted either just as much as Republicans, or more–D:34-R:34, D:39-R:34, D:39-R:35. So why are organizations like Newsweek (D:31-R:38) and Gallup (D:33-R:40) weighting so heavily for Republicans? If anything, recent elections suggest it should be the other way around, and in this election, Democrats are galvanized like crazy–I personally (and very unscientifically) am expecting a heavy Democratic turnout.

So a key question in evaluating a poll is, how do they weight by party affiliation? A new CBS poll just came out a few hours ago(source PDF file), and they have Bush ahead by 9 points (50% – 41%), and though their poll is not so outrageously weighted towards Republicans, it is still weighted in their direction (D: 31-R: 33), when, by the last two election figures, it should probably be more like 39-35 or 38-36, weighted towards Democrats–at least.

And where we find ourselves winding up is, essentially the same as in the last post–in the land of unpredictability–albeit with a crude compass for which data to dismiss with extreme prejudice. But there are still too many factors out there even after party affiliation weighting is factored out, I believe, to make any of the polls believable to a point where it would make too much of a difference. In other words, mind those margins of error, boys and girls.

Categories: Political Ranting Tags: by
Comments are closed.