On New Year's Eve exactly four years ago, the Des Moines Register released its final poll of Iowa caucus-goers and turned the political world upside down.
While the newspaper's final Iowa Caucus poll of 2011, set to be published Saturday night at 7 p.m. Central Time (8 p.m. Eastern Time), may not confound the conventional wisdom this time, it is among the most eagerly anticipated political polls of the season for good reason. The Register has a hard-earned reputation for accuracy grounded in the fundamentals of survey research: Assume as little as possible about the likely caucus-goers, and let the voters speak for themselves.
Four years ago, most polls taken in late December showed a close three-way race for Democratic caucus votes among Barack Obama, Hillary Clinton and John Edwards. Earlier that fall, one-third or fewer of the respondents on most surveys reported that they would be attending their first caucuses in 2008, below the 45 percent of Democratic caucus-goers who had participated for the first time in 2004.
The last Register poll, however, looked very different. Sixty percent of the respondents said 2008 would be their first caucus, and more than usual were younger and politically independent. The survey gave Obama 32 percent of the vote and a big lead over both Clinton (25 percent) and Edwards (24 percent). The potential first-time caucus-goers gave Obama more than two-thirds of his support (72 percent).
The results were met with considerable skepticism. The Clinton and Edwards campaigns immediately attacked the poll's "unprecedented turnout model," saying it was "at odds with history." Joe Biden's pollster described the results as "impossible." Even David Yepsen, then the Register's renowned senior political columnist, telegraphed disbelief, noting that the results would "raise some eyebrows among party pros."
But the paper's pollster, Ann Selzer, had not modified her "turnout model." She had assumed nothing about the demographics or party allegiance of the likely caucus-goers interviewed by her company. Her method involved sampling all potentially eligible voters and letting the likely caucus-goers self-identify through a series of questions. As she explained to the PBS "NewsHour" four years ago, "we put our method in place, and we let the voters speak to us."
The final caucus results vindicated Selzer's approach. Obama won by an even larger margin than the poll predicted. The network entrance poll showed that 57 percent of the voters were attending their first caucus. Forty-one percent of those first-timers supported Obama, providing two-thirds (68 percent) of his support.
The 2008 outcome cemented a reputation that had been growing for decades. The Register's Iowa Poll is the nation's longest continuously running newspaper poll, and it has asked Iowans about their caucus preferences since the 1980s. In 2004, its final survey was the only one to correctly forecast the order of finish for the top four Democratic candidates. An informal survey of pollsters in late December 2008 showed the Register/Selzer poll to be the clear leader in perceived reliability.
All that said, the Register's past success is neither magic nor a guarantee of future performance. Purely random error lurks as the enemy of every poll, and Iowa remains what famed Republican pollster Richard Wirthlin once described as "the ultimate sand trap for pollsters." As in 2008, Selzer is fielding her final poll in the week between Christmas and New Year's Day, a time when many Americans are traveling and away from home.
It helps that Selzer is based in Iowa and knows her state and its voters. It helps that her reputation rests on the accuracy of this particular poll, while Iowa is just another stop on the campaign trail for other pollsters. It helps that her interviewers introduce themselves as calling on behalf of the well-known "Iowa Poll" (they do not volunteer the Des Moines Register affiliation unless pressed), a practice that likely increases cooperation from voters, especially when they're besieged by calls from campaigns close to caucus day.
But the real "secret sauce" of the Iowa Poll isn't much of a secret. Selzer and the Register have gained their reputation by sticking to the sound fundamentals of survey research and trusting their results even in the face of ferocious disbelief. Saturday night's poll results will probably not shatter the conventional wisdom as they did four years ago -- but if they do, pay close attention.
CORRECTION:The original version of this article had incorrect percentages for Hilary Clinton (28 percent) and John Edwards (26 percent) in the final Des Moines Register poll in 2008. The correct numbers were 25 percent for Clinton and 24 percent for Edwards.