Sunday, October 01, 2006

Follies in poll-land

So the story from a major newspaper chain's wire service starts out this way:

WASHINGTON - Democrats are slightly ahead of Republicans in three election battleground states that will help determine control of the Senate, a series of polls released Sunday showed.

Interesting stuff, on the face of it, so let's read a little farther, shall we, as Steve Thomma -- a former co-worker of mine a long time ago and far, far away (Fort Wayne, to be exact) rolls out the numbers:

In Montana, Democrat Jon Tester had the support of 47 percent of registered voters while incumbent Republican Sen. Conrad Burns had the support of 40 percent.

OK, respectable numbers, and quite possibly the basis for suggesting someone might be ahead. A 7-point margin might well fall outside one margin of error, though unless it is a big sample, it's unlikely to be outside the two-margin area that makes us pretty statistically sure (how much of an oxymoron is that) the lead is genuine.

In Ohio, Democrat Rep. Sherrod Brown had 45 percent of registered voters, while incumbent Republican Sen. Mike DeWine had 43 percent.

Now it's getting a little dicey. A 2-point margin is pretty much going to be a dead heat in anyone's poll. You have to survey a darn lot of people to get below a 2 percent margin of error.

In Tennessee, Democratic Rep. Harold Ford Jr. had 43 percent and former Chattanooga Mayor Bob Corker, the Republican Senate nominee, had 42 percent.

OK, let's cut the charade. No pollster in the world -- even the shadiest -- is going to tell you that a 1-point margin between two candidates in a poll can be anything but a statistical dead heat.

But there's that lead again, Democrats are slightly ahead ...

In fact, all these Mason-Dixon polls, each of which surveyed 625 registered voters in the noted states, had a margin of error of 4 percent (OK, 4 percentage points for FEV over at Headsuptheblog). So in not one of these races does any candidate have a lead, statistical or otherwise. Even in our widest race, it is possible the results could be Burns 44, Tester 43.

And let's not open the box of whether registered voters really vote, especially in midterm elections.

Thomma's better than that -- and at least his editors should be. But if there's any sure sign that the silly horserace season has begun, it's silly poll stories like these that provoke misleading headlines like this one in South Carolina's largest paper:

Polls find Democrats
ahead in 3 vital states

Uh, yeah. Sure. Whatever you say.


Post a Comment

Links to this post:

Create a Link

<< Home