« Public Says Iraq War a Mistake | Main | Why Isn't Kerry Farther Ahead? »

A Poll Matthew Dowd Can Be Happy About

At last! A poll that Matthew Dowd can be happy about. While the Gallup folks were busy collecting their data (June 21-23), the fair and balanced folks over at Fox News were busy collecting theirs (June 22-23). To say the Fox News results differ somewhat from the Gallup results would be to considerably understate the case.

The Fox poll has Bush up by 6 points (48-42) among registered voters (RVs). As I discussed in my previous post, Gallup has Kerry up by 4 (49-45) in the identical Kerry-Bush RV matchup. Kind of different!

And check this out. Fox has Bush ahead by 20 points in the solid red states (Gallup had Bush ahead by 8), Kerry ahead by only 3 (!) in the solid blue states (Gallup: Kerry by 14) and Kerry ahead by an identical 3 point margin in purple states (Gallup: Kerry by 9). Huh?!? Kerry ahead by only 3 in the solid blue states--and up by no more there than in the battleground states?

Was Fox really polling the same country? You've gotta wonder. The survey dates of the two polls are virtually identical, we're talking about the same universe (registered voters) and the same matchup--and yet the results are starkly different.

So who do we believe? Well, if it's a choice between Fox News or the Gallup Organization, I don't find this a particularly difficult choice to make. To tell the truth, I rarely pay much attention to Fox News polls, which are invariably and significantly pro-Bush relative to other public polls, but this one seemed so egregiously off and so directly contradicted by the Gallup data that I just had to comment on it.

You might well ask: how on earth do the Fox News folks get such weird results? I don't really know, but one possibility is that they weight their data by party ID. Under this procedure, if you've got, in your view, too many Democrats (like in that pesky Los Angeles Times poll), you simply weight them down and weight the Republicans up so you get to the presumed proper distribution of party ID (those respondents can't really be serious about indentifying with the Democrats!).

I don't know that this is true of the Fox News poll. But it certainly would help explain why their horse race results differ so much from Gallup's (who, like good girls and boys, never weight by party ID--see this good article in the Los Angeles Times explaining why public polling organizations worth their salt eschew this practice). Or how Fox News--again, polling on essentially identical days--could find John Kerry's favorability rating at 42 percent favorable/43 percent unfavorable, while Gallup has it 58 percent favorable/35 percent unfavorable.

While we're on the subject of Fox News polls, it's worth mentioning their new polls of Ohio, Florida and Pennsylvania, which also seem--well, a little surprising. Yesterday, I discussed the June 21-23 ARG poll of Ohio, where Kerry led by 6. Fox, polling on June 22-23, has Bush ahead by 4 in Ohio. Hmmm. Yesterday I also mentioned the June 21-23 ARG poll of Florida, where Kerry was ahead by 2. Fox's June 22-23 poll of Florida has Bush ahead by 9!

Curiouser and curiouser said Alice. Now, if it was anyone but Fox, I might be tempted to ascribe these differences to ARG's use of likely voters (LVs), rather than RVs, as Fox uses. But these are mighty big differences and this is Fox, so I don't buy it.

I especially don't buy it when we look at Fox's Pennsylvania results. Fox, polling on June 22-23, has Bush ahead by 3 among RVs in Pennsylvania. But, the highly reputable Quinnipiac University poll, polling on June 21-22 and also polling RVs, has Kerry ahead by 6. What a difference a day makes--or, considerably more likely, what a difference the Fox News treatment makes.

I have a new slogan for Fox News: "Pro-Bush Results Guaranteed". Unlike "Fair and Balanced", this would allow them to stick closely to their empirical record.

Comments

My theory on Fox polls is that Opinion Dynamics may be disclosing their affiliation with Fox "News" when they do their surveys, whereupon many, if not most thinking people would hang up the phone immediately. If that were the case, the resulting under-representation of thinking people and corresponding over-representation of less careful thinkers -- e.g., Republicans -- in their sampling would naturally lead to the kind of skewed results we've become accustomed to seeing in Fox polls.