The Art And Science Of Political Polling

By .(JavaScript must be enabled to view this email address)
Wednesday - November 24, 2010

E-mail this story | Print this page | Comments (0) | Archive | RSS

Del.icio.us

Political polling affects elections. Read that your favorite candidate trails the well-financed, better-known incumbent by 21 percentage points and you may decide on election day to go to the beach rather than your polling place.

Or consider the experience of Linda Lingle in 1998. Her respected Mainland pollster told her a few days before the election that she had a comfortable lead over incumbent Gov. Ben Cayetano. So she flew home to Maui for the weekend.

Local pollster Barbara Ankersmit of Qmark Research worked for Cayetano that year, and today has an overlay of her 1998 polls framed and displayed on her office wall.

“Our polls showed us close to Lingle that final weekend,” she says. “So we told our fundraiser to go out and raise more money so we could buy all the available (advertising) space possible.”

He did, they bought the space, and Cayetano beat Lingle by a razor-thin margin.

Polling can be misused. Consider the practice of push-polling.

“A push poll creates opinion,” says Becky Ward of Ward Research. “For example, you ask the question: ‘If you knew that candidate X routinely beat his wife, would you be more likely to vote for him, less likely to vote for him, or willing to report him to the police?‘Ask that of 600-1,000 people, and you’ve started a whispering campaign that can decide a close election.”

Ward conducted this year’s Hawaii Poll for the Honolulu Star-Advertiser and Hawaii News Now. “As a media poll, we’re much more concerned about polling too close to the election and influencing the outcome,” says Ward. “In the 1998 Lingle-Cayetano election, we published the Friday before the election. This year we published nine days out.

“A media poll needs to be as transparent and objective as possible,” says Ward. “Media polling suits me. It’s an objective undertaking, rather, versus a matter of life and death.”

Ankersmit did media polls when she worked for SMS Research, once the proprietor of the Hawaii Poll. “I don’t like to do media polls,” she says. “With a media poll you have to be totally balanced and very careful. They have an impact on elections, but it’s difficult to prove how much of one. So many factors go into how people vote.

“From a professional and personal point of view, I’d much rather poll for a campaign,” Ankersmit adds. “The pollster’s part of the inner circle of the campaign. You help shape strategy and messaging.”

Pollsters face new challenges, most of which are technological, and the spreading use of cell phones is the dominant technological problem.

“Getting the proper cell phone sample is particularly important,” says Ward. “For the Hawaii Poll we called 80 percent land lines, 20 percent cell phones only. As things change, we’ll go to 70 percent land line, 30 percent cell phone.”

Exclusively, cell phone users are overwhelmingly young and Democrat. According to the Pew Research Center, a poll with a land-line bias gives Republicans 4-6 percentage points in polls.

Something like that (and more) must have occurred this year in the Rasmussen Poll of Hawaii’s U.S. Senate and gubernatorial races. Three weeks out, Rasmussen gave never-defeated U.S. Sen. Dan Inouye only a 13 point lead over often-defeated Republican Cam Cavasso. Two weeks out, Rasmussen had Neil Abercrombie-Duke Aiona at 49-47, a statistically dead heat. On election day, Inouye beat Cavasso by 51 percent, Abercrombie took Aiona by 17 percent.

Automated polls, where those polled answer questions by hitting a number on their phones, also were used in Hawaii’s 2010 elections with a high degree of accuracy. “I think they were lucky,” says Ankersmit. “I don’t trust them. They’re too simplistic. They don’t screen the person at the end of the line. There’s no live human being involved.

“Polling is somewhere between an art and a science. New technology isn’t going to change that.”