Already six weeks ago, I lamented that watching the Iowa polls could make your head spin.
If you’ve continued watching, you’re probably suffering whiplash by now.
I can’t offer a definitive cure, but perhaps a teaspoon of insight might help.
A USA Today-Suffolk poll had former Vice President Joe Biden leading Sen. Bernie Sanders (I-Vt.) by 6 points, while a CBS News-YouGov survey put Sanders 8 points ahead of Biden, who occupied third place. Monmouth also gave Biden a 6-point lead, but Emerson differed, reporting that Sanders led the Iowa field by 9 points.
The much-ballyhooed Des Moines Register-CNN poll had Sanders 5 points ahead of Biden, who only mustered a fourth place showing in that survey.
All the usual poll-watching caveats are in order here. In most surveys, the candidates are closely bunched, and the polls have relatively large margins of error, especially when comparing them. Emerson collects part of its interviews through a panel operated by Amazon Mechanical Turk, which we have little reason to believe is representative of anyone, let alone of Iowa caucus attendees. Averaging is better than individual polls.
All true, but none answers the question, why are these results so different?
Alas, as I noted, you won’t find a definitive answer here, but I’ll offer a theory, based on experience and data, grounded in understanding the differences between first-time caucus participants and those who have attended previously.
In 2008, then-Sen. Barack Obama (D-Ill.) won 41 percent of the vote from first-timers, but just 26 percent from veterans. Those who had caucused before went for John Edwards over Obama by 30 percent to 26 percent.
There are strong indications that the votes of first-timers and veterans will diverge substantially again this year. I’ve seen data suggesting 7-point to 9-point differences for several candidates.
Historically, about half of Democratic caucusgoers are first-timers, though there’s meaningful variation. In 2008, with the highest turnout on record, 57 percent of Democratic caucusgoers told entrance pollsters they were first-timers. In the lower-turnout year of 2004, it was 45 percent, nearly the same as in 2016 (44 percent).
Nothing keeps growing forever, and certainly not at the same rate. So, even if turnout sets another new record, there may be fewer first-timers than in the past, precisely because of the huge participation in previous cycles. The fact that so many have caucused leaves fewer first-timers on the table.
But given the differences between them, the mix of first-timers and veterans each pollster assumes could have a real impact on the poll results.
The variation in the percent of first-timers and veterans is substantial in the polls that report it.
Recent Iowa polls all had fewer first-timers than the 57 percent in 2008, or even the 45 percent in 2016, which, again, may be a reasonable assumption.
The CBS-YouGov survey was closest to these historic benchmarks, with 41 percent of its respondents claiming they would be new caucus participants, while 27 percent of those represented in the most recent Des Moines Register Iowa poll were first-timers. Quinnipiac and Monmouth had just 21 percent and 17 percent, respectively.
While several recent polls didn’t ask about, or release, information about prior caucus attendance, there seems to be some relationship between the number of first-time caucus attendees counted in a poll and how Biden and Sanders do, relative to each other.
We’ll need more data and results from caucus night to know for sure, but I’d bet that this proves to be an important reason for the disparate poll results.
Mellman is president of The Mellman Group and has helped elect 30 U.S. senators, 12 governors and dozens of House members. Mellman served as pollster to Senate Democratic leaders for over 20 years and as president of the American Association of Political Consultants.