Two recent national polls showed one of the presidential candidates moving ahead of the other.
They just couldn't agree on which one.
A Pew Research Center survey of likely voters conducted Oct. 4 to Oct. 7 gave Republican nominee Mitt Romney a four- percentage point lead over President Barack Obama. A Gallup poll of registered voters Oct. 1 to Oct. 7 showed Obama advancing over Romney by five points one day after, in a shorter tracking survey of three days immediately following their Oct. 3 debate, it called the race a tie.
Another post-debate poll, the ABC News/Washington Post survey released today, put Romney's favorable rating at 47 percent, up three points over last month and his highest score since the survey first asked the question in September 2011. Even so, 51 percent view him unfavorably, second only to his 52 percent negative mark in March. Obama emerged with a 55 percent favorable rating in the latest poll, three points higher than last month and his best mark since April 2010.
The released results didn't include a match-up between the two candidates. The poll of 845 registered voters was taken Oct. 4-7 and has a margin of error of plus or minus 4 percentage points.
Lack of Clarity The conflicting results underscore the lack of clarity in an environment in which, almost every day, new polls are released that vary depending on how voters are contacted, how they are counted and other variables tied to an individual polling group's approach to conducting surveys.
Both Democratic and Republican pollsters have a general sense that Romney's support has grown and Obama's slipped since the former Massachusetts governor's performance in last week's first presidential debate in Denver. There's little consensus, though, on how deep or enduring that change in the race's dynamic may be.
"Romney clearly has gotten a boost coming out of the debate and it's for obvious reasons," said Whit Ayres, a Republican pollster who's worked on several presidential and congressional campaigns and is part of National Public Radio's bipartisan polling team.
Even so, Ayres said he's looking for more of a pattern that shows Romney closing a gap with Obama before pronouncing that the gains are lasting ones.
'Too Few' Peter Hart, a longtime Democratic pollster who has run surveys for the Wall Street Journal and NBC News, agreed. "Romney has been helped" by his debate performance. Yet "a pattern is a series of results showing something in the same direction," he said. "It's just too few polls and too few days." Romney's improved standing in some polls conducted since the debate has "added an exciting twist because things were moving in one direction," toward Obama, said J. Ann Selzer, an Iowa-based pollster who conducts surveys for Bloomberg News.
Still, conflicting results also have only emerged in polling of the states both campaigns say will decide the election.
Rasmussen Reports had Obama ahead, 49 percent to 48 percent, among likely voters surveyed Oct. 7.
Rasmussen Reports, which Ayres said "tends to lean more Republican in the makeup of its sample," had Romney leading in Virginia, 49 percent to 48 percent, according to an Oct. 4 poll of likely voters. The firm uses a technique, questioned by some nonpartisan polling experts, called robo-calling, which relies on automated phone surveys that require participants to respond using dial pads.
A Democratic-aligned firm, Public Policy Polling, uses the same robo-calling technique and put Obama ahead by three points in Virginia, 50 percent to 47 percent, according to its Oct. 4-7 poll of likely voters.
Both Democratic and Republican pollsters express concern that groups using robo-calling may fall short in adequately screening voters to verify their identity, in reaching cell phone users who tend to be younger and Democratic-leaning, and in providing enough transparency to confirm their sampling populations are balanced to reflect the makeup of the general population.
Demographic Issue "The focus of a robo-poll is being very quick and short so, as a consequence, they may or may not ask enough questions to help them judge the likelihood of someone voting or enough demographic questions to weight their sample appropriately," said David Hill, a Texas-based pollster who has worked for Republican candidates since 1984.
Scott Rasmussen, founder of Rasmussen Reports, said automated polls mean that each respondent hears the same questions the same way, without any nuances from a live operator.
He said his company uses an online polling tool to reach people without conventional telephones.
While pollsters usually adjust their data to reflect the demographics of the population they're sampling -- such as age, race and gender -- Rasmussen also weighs his data to reflect political party enrollment or preference. He calls it "the best single indicator of how" someone will vote.
Other pollsters don't do that, because they say party identification can be fluid and change from survey to survey.
'It Changes' "It goes up and down; it changes," said Evans Witt, chief executive officer of the nonpartisan Princeton Survey Research Associates. "Party ID is not stable.
It varies." Voters excited about a candidate might tell a pollster they identify with that nominee's party, and then decide they lean independent in the next survey if their enthusiasm has waned, Witt said.
Another complicating factor in reading the most recent polls is the impact of a Labor Department report Oct. 5 showing the jobless rate fell to 7.8 percent, the lowest since Obama became president in January 2009.
"We have to get to the point where we have enough polls that were conducted past the debate and at least long enough past the jobless numbers," Witt said.
"I don't think we've had a huge move. We're seeing a good debate for Romney, a good jobless number for Obama and I don't think that fundamentally changed the race." Interest Groups A number of interest groups have added to the morass of presidential election polling. The Service Employees International Union sponsors polls with the Democratic-leaning Daily Kos website. Wenzel Strategies of Columbus, Ohio, partners in polls with groups including the Family Research Council, which supports a ban on abortion and calls homosexuality "unnatural." "For whatever reason, if a poll is sponsored, it can influence the results," said Frank Newport, editor-in-chief of the Gallup poll. "Be very suspicious of polls sponsored by a political party or advocacy group. When they report to you, they're spinning." Paul J. Lavrakas, president of American Association for Public Opinion Research, based in Deerfield, Illinois, said respondents may decide to participate in a sponsored poll depending on whether they agree with the aims of the group behind it.
"You're going to get biased results," Lavrakas said.