Two For the Price of One

On June 10, the Los Angeles Times published poll findings that purported to show John Kerry with a seven-point lead over President Bush. The Times headlined its poll “Voters Shift in Favor of Kerry.” The poll results were reported world-wide, generally as fact, notwithstanding that other polls, less well publicized, have shown very different results.
The administration’s chief pollster, Matthew Dowd, called the Times poll a “mess,” noting that it had obviously over-sampled Democrats.
The next day, June 11, the L.A. Times published another poll, this one headlined: “Going to War Not Worth It, More Voters Say.” In this poll, respondents said by a 53% to 43% margin that the Iraq war was “not worth it.” This poll, too, was reported world-wide, and hundreds of newspapers reported as fact that a majority of Americans had turned against the war.
But, as the eagle-eyed Dafydd ab Hugh points out, it doesn’t appear that the Times ran two polls at all:

Here is an interesting point about a “pair” of recent LA Times polls:
They published one poll result that said that Kerry was seven points ahead of Bush. This is the poll that Matthew Dowd showed was patently fraudulent. Mathematically, it’s very simple to prove: Bush is ahead of Kerry among Republicans more than Kerry is ahead of Bush among Democrats; and Bush is ahead of Kerry among Independents. The only way to combine those three results and get Kerry strongly ahead of Bush is if the Times’s poll wildly oversamples Democrats (if the two groups were roughly equal, then Bush’s lead among Independents and the fact that he picks up more Democrats than Kerry picks up Republicans would put Bush ahead, not Kerry).
Then yesterday, they published was seems to be a second, completely different poll showing that a majority of Americans think the Iraq war was unwarranted.
But there’s one amazing oddity about these two polls: they were conducted on the same day — and they interviewed exactly the same number of registered voters.
The Bush vs. Kerry poll: “The Times Poll contacted 1,477 adults nationwide, including 1,230 registered voters by telephone June 5 through 8, 2004.” The Iraq war poll: “The Times Poll, supervised by polling director Susan Pinkus, surveyed 1,230 registered voters from Saturday through Tuesday.”
A quick glance at a calendar shows that Saturday was June 5th, while Tuesday was June 8th. Same days… exact same number of registered voters contacted… say, do you think it just barely possible that this was — the same poll? Of course, that can’t be true; the Times would never take the same poll, split it over two days, and make it look as if it were two different polls, a drumbeat of polls, a murder of polls, showing a continuing erosion of support for Bush’s war and for Bush himself. That would be unethical “pseudo-journalism,” and I’m morally certain that the editor of the LA Times, John S. Carroll, would never tolerate such a thing.

Is this important? Sure. Polls shape perceptions, and perceptions are critical to election campaigns. The Times, in its article about the “second” poll on Iraq, noted that the 10% “not worth it” margin was the opposite of the results of its own November 2003 and March 2004 polls, when respondents said the war was “worth it” by the same margin. The Times announced that this showed a critical slippage in Americans’ support for the war. More likely, what it showed was a slippage in the integrity with which the Times conducts its polls. But hardly anyone understands the intricacies of polling; all that most people see is the headline that the poll generates. So the Times’ stratagem probably worked.