Poll Call: University of Iowa's Over-Inclusiveness

Some folks are very excited about the new University of Iowa poll that shows, among the Dems, Clinton pulling away in a state that's been hard for her, and among the Republicans, shows Huckabee finally cracking the top tier. Woo-hooo!

But maybe we should calm down. The U of I poll suffers from the Olestra of opinion surveys: The loose screen, which means they count the responses of people of who are not necessarily going to participate in the actual election being polled.

According the methodology on the U of I site, a random list of residential phone numbers was used to generate a pool of self-identified registered voters and likely caucus-goers -- I couldn't figure out exactly how many of the 6000 or so called identified as registered voters, but U of I does report that of self-identified registered voters, 36.2 percent were eliminated for saying they were "not at all likely" to attend the caucuses. U of I kept in the sample the over fifty percent who said they were either "very likely" or just "somewhat likely" to attend. They are a very trusting group. Actual caucus-goers represent a single-digit portion (around five percent) of all registered voters in Iowa.

As I've written before, Hillary benefits from such methodology and Edwards suffers. One theory is that loose screens measure general name recognition rather than the kind of strength of support that will turn out voters on a cold January night. Tighter screens — those that do a better job of finding people who may actually caucus (such as Time's!) — tend to show the race (on the Democratic side) as a dead heat. [More on finding genuine caucus-goers here.]

One thing we can probably learn from this poll, since it does arguably do a better job of measuring name recognition over support: Huckabee is, in fact, in the top tier. Indeed, his level of support among likely caucus-goers might be even stronger than shown here, as they're more likely to have been paying attention to him all along.

Pollster.com's Mark Blumenthal has some more thoughts on U of I's methodology here, and an example of how to get a tighter screen here. I'm hoping that Mark will provide some more thoughts on this poll in particular later; will link if he does.

UPDATE: Mark has more, and adds in an email that U of I "included respondents who were 'not very likely' to attend -- that's an, uh, unusual choice."