Yesterday, I wrote about a few of the variables that can affect the accuracy of polls. Polls conducted of just adults, for example, deliver a huge seven-point bias in favor of Democrats. Notwithstanding this bias, it is simply silly to report political polling results of a sample of adults, since a large part of that sample won't actually be voting. It can, however, be useful if you want to make a partisan political point. This is exactly what the Tennessean newspaper and other media outlets have done.
Yesterday, the state's biggest paper splashed across its front page the breathless headline that "Obama closed the gap" and was essentially tied with Romney in the Volunteer State.
The poll of 1,002 Tennessee residents who are 18 and older found 42 percent would vote for Romney and 41 percent for Obama if the election were held now. The survey, conducted May 2-9 by Princeton Survey Research Associates International for Vanderbilt, had a margin of error of 4 percentage points.
Do you see that? "residents...18 and older," meaning adults. Sure enough, a couple more paragraphs into the story, we learn that the Vanderbuilt University poll did add in a registered voter screen. Those results? Romney is beating Obama by seven points.
That's a much different headline isn't it? To its credit, Politico immediately threw cold water on the headline results of the poll, describing the campaign in Tennessee as a "blowout wrapped in a nail-biter."
But, Brett LoGiurato, the resident Obama palace guard at BusinessInsider, went with the shocking headline that Obama was "only down one" in Tennessee. He eventually got around to noting the registered voter number, but still tried to argue that the poll was bad news for the GOP. (Warning: for LoGiurata, the sun coming up is pretty much bad news for the GOP.)
The thing is, we can't really tell if the poll is bad news for the GOP, because Vanderbuilt University won't release the poll data until Wednesday. We don't know if they over sampled Democrats to get these results or played around with the sample in other ways. It seems Vanderbuilt and the Tennessean paper will be releasing bits of the poll over the next couple of days. The needed context to understand the poll won't come until long after they "report" the findings.
(Note: I'm assuming the May 2-9 date for the poll is a typo by the Tennessean. One, seven days is a very long time to have a poll "open". More importantly, why report on a poll taken two to three weeks before reporting it? A lot has happened in the campaign in those weeks, right? Of course, most of what's happened has been negative for the Obama campaign. The newspaper wasn't just sitting on this poll until they needed to change the narrative of the campaign, were they?)