Three new national polls were released late Wednesday, and while they all show President Obama and Mitt Romney locked in a tight race, they also differ on the way in which voters perceive the candidates.
A new CBS News/New York Times poll shows Romney leading Obama, 47 percent to 46 percent, well within the poll's margin of error of plus-or-minus 4 percentage points. A bipartisan poll conducted for NPR shows Obama leading Romney by 2 points, 47 percent to 45 percent, again within that poll's margin of error. Meanwhile, yet another bipartisan poll conducted for Fox News shows Obama leading Romney by 4 points, 45 percent to 41 percent, technically within that poll's margin of error as well.
But despite similarities in the topline results, there are significant differences on other measures. First, it is helpful to examine the ways in which the various methodologies for the three surveys differ.
All three polls were conducted by live interviewers calling both landline and cellular telephones. In all three surveys, respondents were selected randomly according to their phone numbers. The NPR poll (July 9-12) is slightly older than the CBS News/New York Times (July 11-16) and Fox News (July 15-17) surveys, though it is not clear that events over the course of the last week have significantly changed the trajectory of the race.
Though each poll includes a substantial number of respondents, there are some key differences here. The CBS News/New York Times poll was of all Americans aged 18 and older, though the questions pertaining specifically to the election were asked of only those who said they were registered to vote. The Fox News poll interviewed only those respondents who reported they were registered to vote, while the NPR poll added a further screen, interviewing only those who said they were registered and likely to vote in the November election.
Looking beyond the topline results, the polls do diverge on some key measures. Among all Americans, 44 percent approve of the job Obama is doing in the CBS News/New York Times poll, down from 48 percent in April. The NPR poll shows Obama's approval rating significantly higher, with 49 percent of likely voters saying they approve of the job he is doing. The Fox News poll splits the difference, showing Obama's approval rating at 47 percent among registered voters.
Considering that most likely-voter screens make a poll more Republican, it is surprising that the poll of all adults would show lower approval ratings than polls of all voters or likely voters. But only 90 percent of respondents in the CBS News/New York Times poll registered an opinion of Obama's job performance; the percentage who disapprove in the poll is only 46 percent. That means that twice as many respondents in the CBS News/New York Times poll registered no opinion of Obama's job performance than in the other two surveys. It is not uncommon for non-registered adults to be more unsure on questions about politics and elections.
The next area of divergence concerns image ratings for both candidates. The CBS News/New York Times poll asked only registered voters about their opinions of the two candidates, finding both to be upside-down. A shockingly-low 36 percent of voters view Obama favorably in this survey, compared to 48 percent who view him unfavorably. Twelve percent were undecided, and 3 percent say they haven't heard enough about Obama to have an opinion. But, in the Fox News poll, the percentage of voters who view him favorably is significantly higher, 52 percent (46 percent view him unfavorably). The NPR poll did not test the candidate's respective image ratings.
Similarly, for Romney, his favorability rating in the CBS News/New York Times poll is 20 points lower than in the Fox News poll (32 percent versus 52 percent).
Since both polls only asked these questions among registered voters, this is an apples-to-apples comparison, at least from a sampling perspective. But there are important differences in how each pollster asked the question that make comparing the two results unfair.
The CBS News/New York Times poll asked voters, "Is your opinion of [Obama/Romney] favorable, not favorable, undecided, or haven't you heard enough about [Obama/Romney] to form an opinion. Meanwhile, Fox News pollsters told voters, "I'm going to read you the names of several individuals. Please tell me whether you have a generally favorable or unfavorable opinion of each. If you've never heard of one, please just say so."
Voters interviewed by CBS News/New York Times pollsters were given the explicit option of choosing "undecided," or saying that they haven't heard enough to form an opinion. And significant percentages -- 13 percent for Obama, 37 percent for Romney -- chose those options. CBS News polls have often showed lower favorability ratings for elected officials than other polls as a result of these options.
But the Fox News poll only offered voters the option to say they have never heard of candidate, which is different from not hearing enough to form an opinion. And Fox did not offer an explicit "undecided" option, meaning that voters had to volunteer that. As a result, just 2 percent said they could not offer an opinion of Obama, while no one said they had not heard of him. For Romney, 7 percent had no opinion, and 2 percent had never heard of him.
The differences in the way the three polls were conducted are important and help explain why polls sometimes achieve diverging results. Those differences are sometimes instructive, and examining them can provide further insight than choosing one poll over another -- or simply averaging all the polls together.
The CBS News poll surveyed 1,089 adults, including 942 registered voters. The Fox News poll, conducted by the Democratic firm Anderson Robbins Research and Republican firm Shaw & Company Research, consisted of interviews with 901 registered voters, for a self-reported margin of error of plus-or-minus 3 percentage points.
The NPR poll was conducted by two advocacy groups co-led by prominent campaign pollsters: Stan Greenberg's Democracy Corps and Whit Ayres' Resurgent Republic. That poll surveyed 1,000 likely voters, for a margin of error of plus-or-minus 3.1 percentage points.