Iowa Poll-apalooza! (Updated)
9:35 AM, Dec 28, 2007 • By DEAN BARNETT
The L.A. Times conducted a widely discussed poll of Iowa and New Hampshire voters. In Iowa, the Times had Huckabee leading Romney 37%-23%. In her write-up of the poll results, Times staff writer Janet Hook oddly neglected to mention that the results among likely voters are much closer. Counting only the likely voters, Romney trailed 36%-28%.
Ms. Hook's story on the poll reported Huckabee's fourteen point lead as definitive. Why, even Hugh Hewitt didn't notice the discrepancy between the "likely voters" result and the Times' coverage of its own poll. To find the difference, you had to crack the poll's internals and scan all the way down to Question 34. Was it merely a sloppy write-up on Ms. Hook's part or agenda journalism at work? I report, you decide.
There's another interesting aspect of the poll. It appears the LAT labeled 90% of the polled Republicans as likely voters. Given that adding the 10% of unlikely voters to the pool swelled Huckabee's lead from 8% to 14%, it looks like the non-voting public absolutely adores Huckabee. One could infer from this result that Huckabee's support is strongest with the members of the public least engaged in the process and, more importantly, least likely to vote. This strange result may mean that as voting time nears and the "polling sample" narrows from 90% of Republicans with telephones to the relative sliver of the Iowa population that actually caucuses, Huckabee's support will further erode.
Of course, it's impossible (as the above analysis implies) that an unlikely voter pool of only 10% of the total could so dramatically impact the ultimate result. In defense of my obviously faulty analysis, the poll's internals regarding these figures are utterly inscrutable. If anyone can make sense of Questions 33 and 34 and figure out exactly what they're saying, please email me (firstname.lastname@example.org) or do a post on it. While you're at it, try to make sense of the LAT coverage of its own poll. You'll have mine and the blogosphere's eternal gratitude. I'm not promising anything, but if you do a real good job of it, I may even get you a Weekly Standard coffee mug as a belated Christmas gift.
Regardless of what's going on in Questions 33 and 34, the larger point seems to hold: Huckabee does significantly better among unlikely voters than likely voters. That could be portentous. Or maybe I'm a Romney supporter grasping at straws. You'll have to wait until Jay Cost comments to find out what the discrepancy really means.
In other Iowa polling news, Strategic Vision shows Huckabee leading Romney by two points, 29% - 27%. I regret to report, but feel obliged to do so nonetheless, that Strategic Vision's polling history does not jibe with my theory of the Iowa race being extremely fluid. SV has conducted four polls since December 6. Huckabee has led in all of them. His largest lead was six points in their December 18 poll. His smallest is the two point lead registered yesterday. One could view this as a trend, but given the small movements, statistical noise is a distinct (and more likely) possibility.
Nevertheless, I'm sticking to my guns - I still sense Huckabee's support sliding away.