Pollster.com

Articles and Analysis

 

Iowa's Entrance Poll: A Correction

Topics: 2008 , Exit Polls , The 2008 Race

Earlier in the week, I made the mistake of being a bit less careful in an email to Mickey Kaus than I might have been writing for this blog. In an item on Wednesday about the "entrance poll" that the networks have planned for the Iowa Caucuses, Kaus accurately quoted me saying that the Iowa entrance poll has "zero check against younger interviewer bias." I subsequently had a chance to interview Joe Lenski, the consultant who will run the National Election Poll (NEP) exit poll operation for the networks in 2008, and learned that I was not quite right. While past Iowa entrance polls lacked a mechanism to correct for demographic non-response for past Iowa caucuses, the entrance pollsters will attempt such a procedure in Iowa on January 3.

The issue involves a procedure that is standard in all other network exit polling. NEP instructs their interviewers keep a hand tallies of the gender, race and approximate age of every voter they are unable to interview. The exit pollsters use these tallies to weight their data. If they find, for example, that the respondents are 65% female, but that tally of interviews, misses and refusals adds up to 55% female, they will weight the completed interviews to the latter (more accurate) statistic. The age correction has always been especially important because, I as I have written previously, the predominantly younger interviewers typically have more trouble completing interviews with older voters.

The network consortium has not asked entrance poll interviewers to keep such tallies for past caucus entrance polls, because the time pressure is greater: The interviews must be conducted as voters stream in for the 6:30 p.m. caucus start, rather than gradually over the course of a typical election day.

Lenski tells me, however, that given concerns about the big demographic variation in the Democratic race, the networks have insisted on implementing a non-response tally and correction this time around. So my dire warnings to Kaus about the potential for error on the entrance poll results (that they might "show Obama doing better than he'll really do even among those entering") were a bit premature.

Having said that, I have wondered about the accuracy of the correction given what we learned about the shortcomings of the exit polls in 2004. If interviewers have trouble keeping to "select every nth voter" selection procedure, they may also have trouble keeping accurate tallies of the voters that refused or avoided participating. Still, that is my speculation. The bottom line is that I was wrong to assert that the entrance poll would make "zero effort" to correct age bias. Apologies to all for the error.

We will post a full transcript of my interview with Lenski, which focuses on the mechanics of the Iowa entrance poll, within the next few days . Lenski also did a long interview with Andrew Kohut of the Pew Research Center that looks more broadly at how NEP will handle the "super crowded election season." Exit poll junkies will definitely want to read it all.

 

Comments

Interesting to hear about how they correct for sample bias. I would have thought it would be even better to correct using responses obtained by interviewers who either had no response bias or the opposite response bias. (Eg older male interviewer who speaks mainly to older males.) Then one could see what the actual, rather than assumed, trends were in the under-responding populations and correct for response bias accordingly.

Have the two different ways of correcting for bias been studied? Is there a difference? I mean, do we actually *know* that the current method is accurate?

A bone to pick: it's customary to talk about how badly exit polls failed in 2004. That avoids the odious alternative: the exit polls were fine, but the election was not. There's some solid scientific work suggesting the latter, and an interesting refusal on the part of Mitofsky and Co. to release the raw data that could have settled the issue. So, in the interests of scientific objectivity, you can't really say the exit polls missed badly. We don't know that.

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR