Campaigns: Polls inaccurate, irresponsible
Friday, October 29, 2010
WASHINGTON -- When a widely publicized poll showed Republican John Kasich with a commanding, 10-point advantage in Ohio's governor's race, aides to Democratic Gov. Ted Strickland fought back hard.
Against the poll.
"With just two weeks until Election Day, it is our opinion that the Quinnipiac polls are irresponsible, inaccurate and completely removed from the reality of the Ohio governor's race," the campaign said in a statement that noted other private and public surveys were showing a much closer contest.
The Quinnipiac University Polling Institute, an organization with an unchallenged reputation for nonpartisanship, responded mildly. "We stand by our numbers and our overall record for reliability," said Doug Schwartz, the organization's polling director.
The flare-up underscored a widely held view among both politicians and pollsters that polls, once used largely to help a candidate shape strategy, increasingly can affect the outcome of political campaigns in the Internet age. Candidates and their allies instantly disseminate bare-bones results, seizing on those that reflect well on their own prospects, ignoring the rest and generally skipping over details that might caution people about reading too much into them.
"They can affect contributions. They do affect news coverage in a substantial way. They can affect volunteers. They can affect [voter] interest, and through all those things can affect the outcome" of a race, said Mark Mellman, a Democratic pollster not involved in the Ohio governor's contest.
Democratic complaints this year are sometimes dismissed as sour grapes in a campaign trending against them. But Republicans, too, express unease about the proliferation of polls.
"There's a great deal of frustration with media polls, which I don't think spend the kind of money to do this the proper way," said Rob Jesmer, the executive director of the National Republican Senatorial Committee.
According to HuffPost Pollster, 26 polls have been released on the Strickland-Kasich race since Labor Day by 13 organizations. An additional 22 surveys cover the Illinois Senate race, 21 a three-way Florida Senate contest and 20 the contest in Nevada.
As in Ohio, many produce different results within the space of a few days for reasons that go unexplained in the daily communications battle of modern-day campaigns.
"The public has an absolute right to be skeptical about any polling information" that doesn't include detailed material, said Richard Czuba, whose Detroit-based firm, Glengariff Group, Inc., does survey work for The Detroit News and WDIV Local 4.
Jesmer's statement, suggesting that not all polls are equal, hints at the complexities involved.
Demographics -- making sure a survey reflects the views of a proper mix of men, women, older and younger voters, Republicans and Democrats -- are critical to producing a poll that is reliable. A pollster's decisions on which respondents are likely to vote is key.
Professional pollsters also differ on another big issue.
Most if not all firms that work for candidates and the major political parties, as well as Quinnipiac and some other organizations, use live phone operators to ask questions.
Other well-known pollsters such as Rasmussen and SurveyUSA Research rely on automated calls, in which an individual who answers the phone responds to a series of recorded questions by touching the appropriate number on the keypad.
Automated calls are cheaper, but a debate flourishes about their relative reliability.
"I am a firm advocate of live operator telephone calls," said Czuba. "For one thing, you know who is on the other end of the call. If you are doing operator calls, you can screen out the 13-year-old who thinks it would be fun to go along and say, sure, they are eligible to vote."
"We were formed to give people a lower-cost polling option," said Tom Jensen, the head of Public Policy Polling, and he and others defend surveys done by recording.
"Generally speaking, the automated survey process is identical to that of traditional, operator-assisted research firms such as Gallup, Harris, and Roper," Rasmussen says on its website.
The Associated Press has an editorial policy not to report polls that rely on automated calls.
Advocates of automated calling also point to examples in which automated polls appeared to detect the mood of an electorate sooner than operator-assisted calls. They cite last spring's Republican Senate primary in Kentucky, where automated polls showed tea party-backed Rand Paul well ahead of Trey Grayson at a time private surveys by veteran Republican pollsters showed the race tight.
Either way, the difficulties in producing a reliable survey are considerable.
"Women answer the phone more than men, older people are home more and answer more than younger people, and rural residents typically answer the phone more frequently than urban residents," the Rasmussen website adds.
Pollsters must make decisions about which people are likely to vote.
Jensen said that for this fall's campaigns, Public Policy Polling makes calls from lists of voters known to have cast ballots in at least one of the 2004, 2006 and 2008 elections.
Many firms have their own techniques for separating the "likely voters" from the rest of the electorate, often ranking individuals on a scale of 1-10 to gauge the probability they will cast ballots. Some pollsters also adjust the demographic blend that results from phone calls to make it more reflective of the known makeup of the voting age population, a process known as weighting.
The differing approaches account for at least some of the variances in poll results, but some users of surveys consider the potential political bias of the source, as well.
Among Democrats, Rasmussen is widely viewed as partial to Republicans. PPP calls itself a Democratic polling firm. Both organizations say their polling is statistically sound.
Whatever the process, the results vary widely.
In recent months, Democrats attacked a SurveyUSA poll in a House race in Virginia and sought to cushion the results of a Quinnipiac survey in Connecticut's Senate race by issuing one of their own in advance.
"The frustrating thing is that various levels of credibility are given to various polls, and you just don't know the demographics of some of these things," said Aaron Pickrell, Strickland's campaign manager.
A mid-September Quinnipiac poll showed Kasich with a 17-point lead that was larger than any other survey before or since.
The group of likely voters was 54 percent male and 46 percent female, even though women have outnumbered men in nearly all statewide elections in the United States over the past few decades. Independents accounted for 34 percent of all likely voters, Republicans for 32 percent and Democrats 29.
A few days earlier, PPP had released an Ohio survey in which women outnumbered men among likely voters, 53-47, a breakdown more in line with private surveys. Republicans and Democrats each accounted for 40 percent of likely voters, and independents for 20. That poll showed Strickland trailing by 10 points.
A string of surveys followed, most showing a single-digit race favoring Kasich.
Strickland's campaign, evidently concerned that voter attitudes were being shaped by the drumbeat of polls, released a survey of it's own in early October that gave him a four-point edge.
On Oct. 17, Quinnipiac showed Kasich ahead by 10 points, and the Strickland campaign attacked.