Friday, Jan. 11, 2008 | 2 a.m.
National public opinion pollsters, fresh off a glaring failure to pick the winner in New Hampshire’s presidential primary, are now violently queasy about trying to predict a winner in Nevada.
In fact, for a variety of reasons, major news organizations are taking a pass on polling before Nevada’s Jan. 19 caucus.
The concerns stem from the New Hampshire mistake and from knowledge that Nevada has a large transient population not familiar with the workings of a big-time caucus.
By comparison, New Hampshire is a state with a long history of political participation, a stable population and a history of respected surveys that have accurately gauged the preferences of the electorate. Nevertheless, the polls had Illinois Sen. Barack Obama leading New York Sen. Hillary Clinton on election day by an average margin of more than 8 percentage points.
Clinton won by 2 points.
In the world of pollsters, that’s outside the statistical margin of error a train wreck.
The disparity in New Hampshire added to the anxiety of pollsters already uneasy about Nevada. Even NBC, whose cable news network, MSNBC, will broadcast Tuesday’s Democratic debate in Las Vegas, has no plans to poll Nevada, said Peter Hart, the Washington, D.C.-based pollster whose firm typically conducts the NBC-Wall Street Journal poll.
Pollsters who do dare to try to gauge not only the preferences of Nevadans but also the likelihood they will turn out for a first-time prime-time caucus face a set of challenges.
Voters must show up at particular location at 11:30 a.m. on caucus day and publicly declare their support for a candidate in a process that could take more than an hour. In New Hampshire, voters had all day to cast a secret ballot in a voting booth a process familiar to citizens everywhere.
In polling language, a caucus is a “low incidence” event, meaning a relatively small percentage of eligible voters will show up to caucus.
Pollsters have to work harder to find a representative sample of likely caucusgoers, sometimes burning through thousands of households to find an acceptable pool of voters, at significant cost.
“It’s very tricky,” said J. Ann Selzer, whose firm contracts with the Des Moines Register newspaper to run the highly respected Iowa Poll every four years. “It should be impossible to ever get it right, even in Iowa.”
For the record, Selzer’s firm was spot on about Obama’s victory, just one point off. But that accuracy came after decades of experience. Iowa has held a high-profile caucus since 1972. Selzer knows a likely caucusgoer when she sees one.
Larry Harris, a principal with Mason-Dixon Polling & Research, one of two national firms that conduct polling for Nevada media on occasion, acknowledged the difficulty. “It requires a lot more energy than somebody pulling a lever in a secret ballot,” he said.
Nevada’s highly transient and increasingly diverse population is another complication for pollsters. New Hampshire and Iowa are on the opposite end of that spectrum, with nearly all-white populations and relatively static population numbers.
“It’s very difficult to know who to sample, in terms of likely voters,” said Glen Bolger, a respected Republican pollster who has conducted surveys in Nevada.
The Pew Research Center, a respected public policy foundation, regards the circumstances here as so foreboding that it opted to leave Nevada out of its pre-election poll in December.
“It’s an ascending chain of difficulty: general elections, primaries and caucuses,” said Scott Keeter, Pew’s director of survey research. “Caucuses burn up a lot of resources (for pollsters), and we thought our resources could be put to better use elsewhere.”
Nevada’s last poll was taken a month ago by Mason-Dixon, which says it will survey voters again before the caucus. (In the most recent poll, Clinton led in the single digits.) Infrequent polling is a problem, said Susan Pinkus, director of the Los Angeles Times Poll.
“My concern is that you really don’t know what’s going on in your state,” she said. “Everybody’s flying under the radar there. So much is predicated on speculation.”
Even so, Pinkus returned to the problem of identifying likely caucusgoers in a state that has never caucused in a significant way. “You’d have to poll up to wazoo to find these caucusgoers.”
Then there is the issue of race, which Andrew Kohut, president of the Pew Research Center, raised in The New York Times on Thursday.
The New Hampshire results were thrown, in part, by race. Polls frequently overstate support for black candidates among white voters, particularly white voters who are poor, he wrote. Poorer, less educated whites refuse to respond to surveys more often than affluent, better-educated whites and those who won’t answer tend to have more unfavorable views of blacks than respondents who do the interviews.
The upshot: Obama’s support was overstated.
Even before the New Hampshire primary, questions arose about the accuracy of polls anywhere this election season.
In past election cycles, presidential primary voters tended to be older. This year has ushered in a wave new voters, mostly younger, Harris noted.
Those young voters are more likely to have cell phones but not land lines and pollsters prefer land lines because they connect the voter to a specific location. So gauging the sentiment of younger voters is difficult.