Political scientist named to panel studying pre-primary polling

Unraveling the mystery behind why pre-election polls in the New Hampshire presidential primary were so dramatically off-target is the mission of a select panel that includes Charles Franklin, a nationally known polling expert and political scientist at the University of Wisconsin–Madison.

"In 1948, the polls predicted Harry Truman’s defeat," says Franklin. "Pollsters responded to that with a serious study of what went wrong. That self-examination ultimately improved polling practice. I hope that something similar will come from this study."

Polls in New Hampshire wrongly predicted a double-digit Democratic victory for Sen. Barack Obama over Sen. Hillary Clinton. After the votes were counted, Clinton won, taking 39 percent of the vote to Obama’s 36 percent.

The 11-member committee established by the polling industry’s leading trade association — the American Association for Public Opinion Research — will also examine subsequent 2008 pre-election polls to see if they help explain what happened in New Hampshire.

A variety of explanations for the wayward poll numbers have emerged, ranging from faulty turnout models to last-minute changes in voter sentiment.

"Voters may change their minds between the poll and election day," says Franklin, co-developer of Pollster.com. "One important question we hope to answer is how much of the error in New Hampshire was due to voters changing their minds, and how much was due to the shortcomings of the polls themselves."

Franklin says that the circumstances surrounding this primary season also present difficulties for pollsters.

"Primary polling is a particular challenge because of low voter turnout and fast-changing support for candidates," Franklin says. "This year’s compressed primary season and a surge in turnout among young voters in the Democratic primaries have made polling particularly challenging."

Committee chair Michael W. Traugott, a University of Michigan professor of communication studies and president of the World Association for of Public Opinion Research, says the panel plans to report its findings in April.

"There are a lot of different explanations floating around about what happened with the polls, but the committee will look at the available data to see which ideas have more merit than others," Traugott says. "Pre-election polling has a long history of accuracy. This committee’s work will help to make sure that it continues to remain accurate and reliable."

Franklin says a last round of polling on the Monday prior to the Jan. 8 New Hampshire primary may have identified changing trends in voter opinion.

"But if lots of voters change their minds on election day, there is little the polls can do to predict that," Franklin says.