What Does the New University of Iowa Poll Mean?

University of Iowa Professor, pollster, and caucuses guru David Redlawsk released the results of his latest Hawkeye Poll today, eliciting mixed reactions from the national media. Every blog entry I have read that covers the poll has linked to Pollster.com’s analysis of the Hawkeye Poll and their explanation of why its results are not comparable to other poll results.

The short synopsis is that the Hawkeye Poll has a problem with oversampling Iowa’s electorate. Based on its sample sizes and reliance on self-reporting for the “likely voter” question, it is accurate if over one million Iowans head to precinct caucuses in January. The record for turnout on the Democratic side is around 120,000, and the record on the Republican side is about 100,000, which makes Redlawsk’s turnout model untenable.

Also notable about the Hawkeye Poll — and although many report this as a disadvantage of using the poll’s results, I don’t see it that way — is that the “Which candidate will you support?” question is left open-ended, so voters have to name their candidate without hearing a list of options. If this were a standard election where “likely” voters are going to show up to the polls and see a ballot with lists of candidates’ names on it, this would not be a particularly useful methodology; but the caucuses are different, because (1) turnout is difficult to project based on past voting history, on a certain level, and (2) Iowans don’t show up at their precint caucuses unless they already know whom they will support in the first round and, often, in the second round, should they be forced to record a second choice.

So the Hawkeye Poll oversamples the electorate, but it asks the “Which candidate will you support?” question more effectively than other polls. Here’s what that means:

• The least likely caucus-goers included in the survey are most likely to report support for candidates with high name recognition, positive coverage in the mainstream media, and TV advertisements. That means the candidates whose support is most likely to be overstated include (in descending order, based on my guesses and very little empirical data) Sen. Hillary Clinton (28.9%), Sen. Barack Obama (26.6%), Gov. Bill Richardson (7.2%), and former Sen. John Edwards (20.0%), with the majority of the oversampled “less likely” voters breaking for Clinton and Obama. (Notably, Clinton’s Iran mailing hit Iowa mailboxes on October 20, right in the middle of the eight day window of interviews for this poll, perhaps further inflating her name recognition. Obama’s response mailing would not have hit mailboxes until the final day of the interview window or, more likely, the day after interviews ended.)
• Although Richardson’s numbers may have been slightly inflated based on name recognition and warm feelings generated by his TV ads, this is probably mitigated by the oversampling that serves to benefit primarily Obama and Clinton. To measure the influence of his TV ads (and, potentially, TV ads in general) on this poll in particular, compare his 7.2% in October to his 9.4% in August, which was closer to the time of his summer ad blitz.
• The voters who reported support for Sen. Joe Biden (5.3%) are almost certain to caucus for their candidate. In fact, given the number of unlikely voters sampled who probably inflate the numbers of the candidates ahead of the Delaware senator, Biden’s support is probably slightly greater than 5.3%.
• The candidates who did not meet the threshold of 2% in the Hawkeye Poll (Sen. Chris Dodd, Rep. Dennis Kucinich, and former Sen. Mike Gravel) have serious issues with name recognition, but some of them may well have support greater than 2% once the frontrunners’ inflated numbers are filtered out.
• The reported number of undecideds (8.9%) is virtually meaningless. Only 13.8% of individuals who were reached by interviewers and were qualified to answer the questions participated in the survey. Of the 86.2% of qualified survey participants who declined to participate, many may have refused out of annoyance, but some may also have refused because they are undecided or because they do not know much about the race yet.