[ Home ] [ Up ] [ Perot, Bush, Clinton ] [ Withstanding the Wave ] [ Rise of Robopolling ] [ Marriage and the Courts ] [ Delegates and the Democrats ] [ How to Forecast an Election ] [ Case for Online Polls ] [ Presidency on Life Support ] [ Iowa Caucuses ] [ 2000: Clinton's Coattails ] [ NAES 2000 ] [ Support for Clinton ] [ Sampling Error ] [ Incumbent Rule ] [ Focus Groups ] [ Internet Polling ] [ Questions About Polling ]

J. Ann Selzer is president of Selzer & Company and director of The Des Moines Register’s Iowa Poll.

This article appeared in the Jan. 26, 2004, edition of The Polling Report.

The Iowa Poll and the 2004 Caucuses

The whirlwind of the final week before the Iowa caucus led to one moment of consensus: The pundits believed the race was volatile and too close to call. Tracking polls in the final days showed four candidates clustered in a pack at the top. Additionally worrisome was a high percentage of likely caucus-goers who, though they had an initial preference, said they could still change their minds. The result was very few television or print commentators who were confident enough to make predictions.

The caucuses by their nature are hazardous duty for pollsters. To participate, a person must be a registered Democrat, which sounds simple enough. However, participants can register to vote at the caucus. And, they can change their party affiliation at the caucus. In addition, 17-year-olds who will be 18 by the date of the November general election are eligible to participate.

Yet, the entrance poll numbers and the final delegate counts reflected what had been reported in The Des Moines Register’s last Iowa Poll, published the Sunday before the Monday caucuses on January 19th. This was the only poll that correctly positioned Senator John Edwards in second place and Howard Dean in third place behind the leader, Senator John Kerry, with Congressman Richard Gephardt in fourth place.

This article offers a bit of Tuesday-morning quarterbacking, describing how our analysis of Iowa Poll data fit the eventual caucus results. It’s a job easier done after the fact than before, but is perhaps instructive for future caucus and primary polls.

How Other Polls Fared
While there is an argument to be made (and Warren Mitofsky certainly makes it) that the entrance poll was not designed to predict the eventual distribution of delegates, in fact it did exactly that.

Zogby’s tracking poll in the last few days showed Dean remaining in second place, with virtually no change over the final five days. While the tracking poll showed a small, steady increase for Kerry, and a small jump for Edwards, it still looked very close, with the spread from first to fourth widening only from five points to seven points.

A local media poll also showed Dean retaining a lead, with just a four point difference between first and fourth place.

The Iowa Poll
What gave us confidence in the Iowa Poll numbers?

First, we had reasonable confidence in our sampling methodology. While a random-digit sample assures every land-line telephone-equipped household has an equal probability of being contacted, thereby catching every potential caucus-goer, the voter registration list is efficient and nearly comprehensive. We added recent registrants to our database two weeks before the caucuses, making it as fresh as possible.

The overwhelming majority of Iowans eligible to vote are registered.

• 2,192,686 Iowans counted in the Census are age 18 and over.
• 50,516 of those adults counted in the Census are not citizens.
• An estimated 42,300 of Iowa residents are convicted felons and therefore not eligible to vote.
• 1,983,628 Iowans are registered to vote (not including new registrants on caucus night).
• That’s a registration rate of 94%.

Source: U.S. Census and The Sentencing Project of Human Rights Watch.

The same calculation for New Hampshire reveals a registration rate of 77% (same sources).

This made us less nervous about talk that the Dean campaign was bringing new people into the political process. Even though the entrance poll showed that 45% of all caucus-goers were first-time attendees, that turned out to be more about bringing currently registered voters to their first caucus than about creating political interest where there had been none. Our sample was drawn from a proportionate mix of registered Democrats, Greens, and no-party registrants; 24% of our respondent pool identified themselves as independents; the entrance poll showed 19% identified as independents.

Second, our trends told a powerful story. The two-day rolling averages for the four days the Iowa Poll was in the field showed Kerry rising fast, followed by a hot Edwards. Dean was sinking fast, losing seven points from the first two days of interviewing to the last two days. He would have needed a miracle to recover a second place showing.

Third, Kerry was turnout-proof. His first-place finish with the full sample was bolstered by his strength among those who said they would definitely rather than probably attend. Respondents were asked how likely it was they would attend the January 19th caucuses -- would they definitely attend, probably attend, might or might not attend, or probably not attend. Only definite and probable attendees qualify for the poll. Kerry led both with the full sample and with just definite caucus-goers -- an indicator he would win regardless of turnout.

Fourth, the big turnout was expected by all and that would help Edwards. He was first among those who said they would probably, rather than definitely, attend. Only if bad weather convinced the less than fully committed to stay home was Edwards’s second place finish threatened.

Finally, while the race was fluid, no single candidate seemed poised to gain from late decision-making. Forty-seven percent (47%) of likely caucus-goers expressing a preference said they could still be persuaded to change their minds and support another candidate. However, the race was not substantially different between the two groups—those whose minds were made up and those who could still change. The only thing that could have made a difference in the outcome would have been for all of those who said they could still be persuaded to change their vote to indeed have moved to their second choice candidate. Kerry would still have led, but Howard Dean would have taken second place, with Edwards and Gephardt in a tie for third. Yet, the story was that Dean supporters were moving away from their candidate; something major—that miracle again—would have to happen to bring them back.

That miracle didn’t happen. Something else that was expected to influence the final outcome didn’t happen. The long-standing axiom of caucus politics in Iowa is to organize, organize, organize and then get hot at the end. Of the two elements, getting hot is far more important. John Edwards had a skeleton organization in the state, yet his heat catapulted him from the bottom of the pack (he won just 5% of likely caucus goers in a November Iowa Poll) to second place. Dean had the most money and therefore the strongest organization, yet didn’t have enough glue to keep supporters in place.

The caucuses by their nature are hazardous duty for pollsters.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

---

 

 

 

 

 

The overwhelming majority of Iowans eligible to vote are registered.

 

 

 

 

 

 

---

 

 

 

 

 

 

 

 

 

 

 

The long-standing axiom of caucus politics in Iowa is to organize, organize, organize and then get hot at the end. Of the two elements, getting hot is far more important.