Margin of error: The new L.A. Times survey on the presidential race
One key result from the L.A. Times/Bloomberg News presidential poll released this morning is not too close to call: The Times itself misinterpreted its own poll.
And because widely hyped polls from major papers are often self-fulfilling prophecies (who doesn’t want to jump on the bandwagon?) that become “news” bits played and replayed by the electronic media and are marketed back to us by the candidates and their handlers and the news readers on TV, the best course is for you to ignore them.
Based on my independent survey of my own opinions — and what I learned long ago about polling from pollsters themselves and statistics professors — here’s what you should ignore:
The headline on veteran L.A. Times reporter Doyle McManus‘s bannered story in his own paper:
Wrong. The poll’s main finding is that Barack Obama has continued his rapid momentum upwards, slicing and dicing Hillary Clinton‘s previous lead.
McManus’s lead paragraph is even worse:
The problem — see for yourself — is that the poll reveals that Clinton’s lead may be as small as 1 percentage point. “Solid”? Her edge over Obama, according to the poll’s own results, continues to erode even more rapidly than the ground under one of those cliff-top houses on the Southern California coast.
Let’s not even get into the fact that the poll itself is questionable, claiming to be nationwide but consisting of telephone interviews of “1,312 registered voters, 532 Democratic-primary likely voters and 337 Republican-primary likely voters.” The supposed 42-33 lead for Hillary is based on scripted phone conversations from a boiler room with 532 people across America.
And let’s not even quibble about the highly questionable wording of the questions, which play on the vague notions of “experience” and “change.” And let’s not even get into the fact that the pollsters reveal the statistical margin of error but don’t reveal the “confidence level.” (Read this explanation of those terms, and of political polling in general, from the Portland Oregonian.)
Leaving all that aside, let’s just say that the story blares the “news” that Clinton leads Obama 42 percent to 33 percent. Of course, the second paragraph from the end notes that the margin of error is plus-or-minus 4, meaning that Clinton could have 38 percent and Obama 37 percent. In other words, the results fall barely, just barely, within the margin of error. That’s too close to call, unless you’re Dick Cheney and you mobilize GOP officials and courts to steal the election.
The only thing these polls are good for is assessing movement, change, trends from previous polls taken the same way and from the same pollster. And in that case, Obama is actually the winner. He has the clear momentum the more exposure he gets. If the rate of his momentum had slipped, that would be news.
(Full disclosure: Obama disclosed the other day that his favorite TV show is The Wire and that he thinks that the show’s Omar is a “fascinating character.” I couldn’t agree more. And who wouldn’t support Obama after that disclosure?)
If you don’t believe me about McManus’s own fuckup on his paper’s own poll, here is the headline on the formal, final press release and detailed results that the L.A. Times issued on its poll:
Yes, the pollsters’ own story does not lead with Hillary’s holding an edge. And the press release notes up high:
McManus doesn’t even mention that earlier poll or Clinton’s earlier 24-point advantage.
The fact that such a shift has occurred in less than two months’ time is astounding, especially considering that Clinton outpolled Obama in New Hampshire.
The subhead on McManus’s story does refer to a change in Obama’s support. But it’s wrong-headed. The subhead:
The poll actually shows that there is also no longer a “clear front-runner” on the Democratic side.
Unless you see 1 percentage point — which the margin between Clinton and Obama could be, according to statistics — as a “clear” lead.