For President Barack Obama's opponents, word of his latest downfall was stunning and welcome.
Headlines and tweets all said just about the same eyebrow-raising thing: "Poll: 71% of Obama supporters 'regret' voting for his reelection."
The purported statistic from a YouGov.com poll was the focus of a Washington Examiner post last week that spread quickly through conservative blogs, social media, radio and online forums. Tweets about the news by big guns like Fox & Friends and various accounts with smaller reach were hard to miss.
But the news is wrong. PunditFact, a Tampa Bay Times project to fact-check the talking heads, is here to explain. Let's start at the beginning.
YouGov.com did not even mention the news about regretful Obama voters in its Feb. 17 report of a new poll.
The story that accompanied the poll, headlined "Mitt Romney's phantom campaign," explored whether Romney is viewed more favorably ahead of the 2016 presidential election given more positive media buzz about a Netflix documentary about him. (The answer, based on their polling data, is not really.)
YouGov polled about 1,000 voters over two days in early February, asking voters for Romney and Obama if they would make the same decision today that they did in November 2012. YouGov found 90 percent of Romney voters would vote for Romney again, while 79 percent of Obama voters said they would stick with Obama.
Of the remaining 2012 Obama voters, 10 percent said they would not vote for him again, and 11 percent said they were unsure. The 10 percent who would not vote for him again got a follow-up question: Do you regret voting for Obama?
Of 35 people in that group, 25 said yes. That's 71 percent.
In the eyes of YouGov assistant editor William Jordan, this was not worth mentioning in the original report.
"We did not think it surprising that the small number of voters who said they would no longer vote for Obama now regretted voting for him," he said.
The next day, Washington Examiner columnist Paul Bedard published a story (since updated) that picked up on the people who said they regret voting for Obama.
"Over seven in 10 Obama voters, and 55 percent of Democrats, regret voting for President Obama's re-election in 2012, according to a new Economist/YouGov.com poll," he wrote. (The Economist has collaborated with YouGov in past polls, but not this one.)
Those words sprung to life, appearing on the Drudge Report, spreading across Twitter, and landing on sites like TeaParty.org and World Net Daily. It seeped into partisan radio broadcasts, including on Rush Limbaugh's show, and even a San Diego morning TV show.
Conservative blog Hot Air was the exception. It urged its readers to ignore stories about the 71 percent statistic.
Why? YouGov, while correctly reporting its sample sizes and percentages in its tables, left off an important disclaimer in describing who got the question. No one realized it until someone asked about it the day after posting the results.
The description about the "regrets" question should have said, "Asked of those who reported voting for Barack Obama in 2012 but would not vote for him if the election was held again."
Instead it said, "Asked of those who voted for Barack Obama in 2012." That language suggests a sample size of 396 voters, not 35.
"The 71 percent who regretted voting for Obama, which was picked up by various media outlets, is of the 3.5 percent (35 of 999 respondents) of the sample who said they voted for Obama in 2012 and would not vote for him today," Jordan said.
YouGov fixed its truncated filter. Bedard, at the Examiner, rewrote parts of his story and changed the headline to "Poll: Only 79% of Obama voters would vote for him again."
But there's no recall button for bad information on the Internet.
Surveys require a great deal of attention from pollsters and reporters alike, especially ones with political overtones, said Karlyn Bowman, a polling analyst at the conservative American Enterprise Institute. As the sample size shrinks, the margin of error rate expands.
Point being, while the overall YouGov poll had a margin of error of plus or minus 3.8 percentage points, the question about regret would have a much higher margin of error.
"You've got to be really careful when you're looking at subsamples of subsamples of subsamples," Bowman said.
In this case, that didn't happen. We rate this claim Pants on Fire.
Read more rulings at PunditFact.com.