11/17/2011 01:13 pm ET

Iowa Caucus Poll Showing Cain Leading Based On Stale Data

WASHINGTON -- A new poll of likely Iowa caucus-goers released Thursday purports to show Herman Cain still "leading," but the poll is already dated. It was in the field for two full weeks from Nov. 1-14. Sharon Bialek, the fourth woman to make sexual allegations against Cain, held her televised press conference on Nov. 7. The pollsters confirmed to The Huffington Post that just 96 of the 377 interviews with likely caucus voters were conducted after that date.

The poll, which was conducted by Iowa State University (ISU), the Cedar Rapids Gazette and television station KCRG, shows Cain with 25 percent, followed by Ron Paul with 20 percent and Mitt Romney with 16 percent. All other candidates received less support than "can't decide," which got 8 percent. Since the survey contacted just 377 likely voters, with a margin of error of +/- 5 percent for each candidate's support, Cain's margin over Paul in this survey is not large enough to be statistically significant.

James McCormick, the Iowa State political science professor who coordinated the poll, says in the ISU release that Cain's continuing support "seems remarkable" given that "phone interviewing was done at the time of the sexual harassment allegations." Perhaps, but the national polls showing declines for Cain were fielded mostly after the Bialek news conferences on November 7.

David Petersen, another ISU political scientist who helped coordinate the poll, explains they examined the timing of the interviews before releasing the data. "There is no significant (in the statistical or substantive sense) difference in support for Cain over time in our data," he tells the Huffington Post via email. "[Cain] gets the support of about 25% of the people interviewed before the 8th and about 23% of the people on the 8th or later. "

The margin of error on 96 interviews is roughly +/- 10 percent for each candidate. So while Cain's 23 percent support on the poll after November 7 is not significantly different than the 25 percent he received before, that result would also be statistically consistent with the sort of 5 to 10 point decline in Cain's support found on recent national polls. That's the problem with a very small sample size.

Fielding a survey of likely caucus-goers is a big challenge for all pollsters, mostly because actual caucus-goers are a very small percentage of the likely electorate. In this case, the ISU pollsters had to interview a sample of 1,256 registered Iowa voters in order to identify 377 who said they would definitely or probably attend the Republican caucuses on January 3.

ISU is not the first pollster to require a long field period to complete their interviews, but the delay demonstrates the risks involved in using this sort of survey to track vote preference. In a high stakes presidential primary race, two weeks can be a very long time.