Media Jumped on Pro-Obama Poll, But Was the Sample Screwy?

Wednesday morning, ABC News and the Washington Post released a new poll showing Barack Obama leading John McCain by 9 points, 52% to 43%. The next day, NBC News and the Wall Street Journal released a poll showing a much tighter race — 48% for Obama, 46% for McCain.

Any guesses as to which poll excited the press more? And which poll has come under fire for over-sampling Democrats?

ABC, naturally, reported its own poll on Wednesday’s Good Morning America, with Diane Sawyer touting at the top of the broadcast: “Breaking news this morning: Barack Obama gains ground in a new ABC News poll, a nine-point lead over John McCain.” The on-screen graphic exclaimed: "Obama Surges Ahead"

Chief Washington correspondent George Stephanopoulos was giddy: “Diane, you have to go back to 1948 for the last time when a candidate, having this kind of a lead in late September, lost.”

The ABC poll was so thrilling to journalists, it was also part of the political discussion on CBS and NBC that same morning. CBS’s Harry Smith confronted Republican Karen Hughes: “We look at brand new poll numbers this morning, nationwide poll numbers, and all of a sudden Barack Obama has jumped out to this nine-point lead.”

And NBC’s Today made room for their competitors’ poll in their news round-up. “The financial crisis is the number one issue concerning voters, according to a new poll out this morning. It shows Barack Obama gaining ground over John McCain in the race for the White House. The ABC News/Washington Post poll has Obama now leading McCain by nine points; two weeks ago in the same poll, McCain had a two-point edge over Obama,” newsreader Hoda Kotb told viewers.

As for the NBC News/Wall Street Journal poll, the head-to-head numbers didn’t even excite NBC that much. Political Director Chuck Todd went over them in a Wednesday night preview on Nightly News, which did not lead the broadcast but came about 10 minutes into the program.

After going through numbers showing how McCain “was getting clobbered on the economy over these past two weeks," Todd mentioned: "Oh, by the way, we did have a head-to-head number here, this narrow lead for Obama, 48 percent to 46. This is a slight change from our last poll, showing a little drift toward Obama.”

The next morning on Today, the discussion of the poll also downplayed the head-to-head match-up, and neither CBS nor ABC were excited enough by the narrow margin to mention any of their competitors findings.

But the ABC poll has come under fire, both from the McCain campaign’s pollster Bill McInturff and from Democratic pollster Peter Hart, who worked on the NBC News/Wall Street Journal poll, for over-sampling Democrats.

As MSNBC’s “First Read” blog reported Wednesday morning: “Those identifying as Democrats in the ABC/Washington Post poll outweighed Republican by 16 points, a wider-than-usual number, McInturff said. ‘It’s way different than what other pollsters are showing,’ McInturff said, citing many polls showing about a four-to-nine point Democratic Party ID advantage.”

After the NBC/Journal poll came out, Hart pointed out the same reason to doubt the ABC/Post poll:

The Journal/NBC News poll comes a day after a national poll conducted for the Washington Post and ABC News found Sen. Obama with a nine-point lead. Mr. Hart, the Journal's Democratic pollster, argued that these results were skewed because their sample included a disproportionate number of Democrats. Democrats had a 16-point advantage in the Post/ABC poll, which is considerably higher than most polls have found historically and this year.

The Journal/NBC poll found an eight-point advantage for Democrats.

In a posting to ABC’s Web site polling director Gary Langer defended his methodology:

Partisanship is measured by asking if people think of themselves as Democrats, Republicans or independents. That’s unleaned party ID. A next step is to ask independents which party they lean toward. That’s leaned party ID.

Unleaned party ID is more telling – it’s more rooted in actual partisan sentiment; it lets us look separately at independents, the quintessential swing voters; and it’s the number that’s comparable to what we learn from the exit polls on Election Day.

As noted, unleaned party ID among likely voters in our latest poll is 37-30 percent. Its average since we started following likely voters in June has been about the same, 36-31 percent. Among the broader population of registered voters it’s 38-28 percent, exactly matching its 2008 average in ABC/Post polls. Our leaned party ID among registered voters, also as noted, is 54-38 percent. Among likely voters it’s 52-41 percent.

The real question is whether the ABC poll that so titillated the media is the first sign of a pro-Obama trend, or merely a stray data point in a statistically-tied horse race. Both ABC/Post and NBC/Journal looked at relatively few voters (780 likely voters and 1085 registered voters, respectively) compared to the big Gallup and Rasmussen tracking polls that present a three-day rolling average of 2,700 and 3,000 voters every day.

And those polls have shown no sign of validating the big swing towards Obama that ABC/Post theorized. As of Friday morning, Rasmussen shows a five-point advantage for Obama, while Gallup has them tied at 46% each. The RealClearPolitics average shows a 3.7 percent advantage for Obama — a clear swing from the trend of two weeks ago, but hardly an affirmation of ABC’s numbers.

But as the knee-jerk reaction to the ABC News/Washington Post poll shows, the media’s interest in individual polls has less to do with an appreciation of their carefulness or exhaustiveness than with their ability to generate a sensational (pro-Obama) storyline.