The First Iowa Caucus Poll

Yesterday, the Des Moines Register released the first public poll asking likely participants about how they might vote in the 2008 Iowa Democratic Caucuses.  Thirty percent (30%) of their sample expressed an early preference for John Edwards, 26% for Hillary Clinton, 12% for John Kerry and 10% for Iowa Governor Tom Vilsack.  The other five potential candidates received support in the low single digits.  This first appearance of a pre-Caucus poll allows MP to make a point he will repeat again and again:  Identifying "likely caucus goers" is no easy task, and the methods used by the various pollsters will differ significantly.**

The Register article explained the basics of their methodology (see the "About the Poll" sidebar on the left side of the page), but left out a few useful details. So I contacted J.Ann Selzer, whose firm conducts the Iowa poll, and she kindly provided some missing details. 

Two weeks ago, the Iowa poll interviewed 600 Iowa registered voters who were likely to vote in the June 6 Democratic primary.  They started with a list of Iowa’s registered voters, selected a random sample of those who had actually cast a ballot in the 2004 general election for President, and interviewed 600 by telephone who said they "would definitely vote in the upcoming Democratic primary."   The results of that survey showed gubernatorial candidate Chet Culver leading with 36% of the vote to 28% for Mike Blouin and 21% for Ed Fallon.  Culver won 39% of the vote to 34% for Blouin and 26% for Fallon. 

Within the same survey, the pollsters identified a subgroup (n=399) that said they "would definitely or probably participate in the 2008 Democratic caucuses."  The results from that sub-sample were the ones reported on Sunday and summarized above. 

The big challenge for polling this contest, of course, is that turnout for the Democratic caucuses is typically a small percentage of eligible voters.  Iowa had roughly 2.2 million voting eligible adults in 2004, of whom (as of last month) approximately 1.9 million are considered "active" registered voters by the Iowa Secretary of State.  But only 124,331 participated in the 2004 Democratic Caucuses for President (according to the subscription only Hotline).  That number amounts to roughly 6% of all registered voters, so selecting "likely caucus goers" is no easy task.

We know that the Register asked roughly two-thirds of the voters they considered likely to vote in the Democratic primary last Tuesday about their caucus preferences (399 divided by 600 = 66.5%).  This percentage is more than reasonable if we believe the Register’s selection of likely voters for last week’s primary was on target.  The 2004 caucus turnout (124K) amounts to 84% of the turnout in last week’s Democratic primary (148,000 voters).

What an educated poll consumer would really like to know, however, is the fraction of the population of adults or registered voters represented by the poll.  In this case, it would be helpful to know how many respondents were screened out of the first poll because they said they were unlikely to vote in the gubernatorial primary.  Unfortunately, Selzer politely declined to provide that information.   [Update:  Selzer wrote back today with the information I was seeking — see update below]

Now, I do not mean to pick on the Iowa Poll.  In this instance, they have provided more details about their methodology than most pollsters, and let’s not forget, their final 2004 poll nailed the initial vote preference in the 2004 Caucuses. 

However, this example does raise a personal pet peeve:  The degree of screening in polls concerning low turnout primaries and caucuses can make a huge difference in the results, yet with a few exceptions, most public pollsters prefer to withhold the details.  As we get closer and closer to campaign 2008, let’s keep asking the pollsters:  How tight is your screen?  What percentage of registered voters or eligible adults does your sample represent?  A little more transparency in this regard would be a very good thing. 

**And yes, it is very, very early.  The Iowa Caucuses are still nineteen months away.  These initial soundings are interesting, but they tell us mostly about how which prospective candidates the respondents know and what they think of the candidates they know.

UPDATE & CLARIFICATION (6/13):

Ann Selzer’s initial relucatance to share information about the incidence of their study may have been because my query was unclear, and — as it turns out — my attempt to clarify via email managed to bounce back last night.  That’s life on the Internet (damn you "MAILER DAEMON!").

At any rate, she sent at message this morning that their initial survey of likely voters for the June primary had an incidence of 20%.  Thus, to get their sample of 600 likely primary voters, they screened from the initial sample of those who had cast a ballot in the general election for President in 2004 all but the 20% who said they were very or somewhat likely to vote in last week’s primary.   And only two-thirds of the respondents sample identified themselves as very or somewhat likely to vote in the 2008 caucuses.  Do the math (1.5 million voters in 2004 x 20% x 66%), and the sample of 399 represented a universe of roughly 200,000 likely caucus goers. 

That seems a reasonably tight screen to me, especially since the Caucuses are still 19 months away and trying to precisely identify the real Caucus-goer universe is a long way from an exact science. 

The good news is that the most respected Iowa Caucus poll has disclosed their incidence.  Thank you Ann Selzer and thank you Des Moines Register!  Let’s hope this is a sign of things to come from the others who poll in Iowa between now and January 2008.