I turned on NPR yesterday to hear that “Colorado is now a statistical dead heat” in the presidential race. I don’t know why journalists love that phrase so much—I suppose it gratifies their desire for a close, dramatic contest. In fact, the last five polls of Colorado all show Obama leading; I suppose the reporter was ignoring four of these, picking one with Obama up by 3 points, and basing her statement on the idea that a 3-point lead in the poll is consistent with the assumption that the race is really tied. Of course, it is equally consistent with the assumption that Obama has a six-point lead, so to call it a “statistical dead heat” makes exactly as much sense as saying “Obama now has a statistical 6-point lead.” That is, no sense at all.
If you wanted to say something true, you could say that Obama’s lead of 3 or 4 points is a bit disappointing in a state he won easily in 2008, and thus that CO is one of Romney’s better-looking targets among the swing states.
Also yesterday, CNN released a poll showing Obama with a 3-point lead, down from 6 points in their previous survey. According to pollster Keating Holland: “That’s a strong suggestion that whatever bounce President Obama received from his convention has, as expected, faded away. That’s why they call them ‘bounces’.”
This is emblematic of a phenomenon even more depressing that the “statistical dead heat.” When someone publishes a poll, they often act as though their poll were the only one in existence; this leads them to say things on the basis of woefully inadequate information, things any well-informed reader can identify as bullshit. In this case, the idea that the race has reverted to where it was before the Democratic convention, when it was actually tied, or even to where it was before the Republican convention, when Obama had a 2-point lead…well, that’s fool talk.
For example, Monday was not a very good polling day for Obama, but even so, the other 8 (!) national surveys showed, on average no drop-off for Obama since the previous version of the same surveys (all conducted after the Dem convention). (See Nate Silver’s piece on 538.com for details.) There have also been a ton of recent state surveys showing Obama ahead in every battleground state except possibly North Carolina. Yesterday’s haul included a poll giving Obama a 15-point lead in New Hampshire, almost certainly an outlier but it also isn’t a ‘statistical dead heat.’
And today’s state polls are pretty awful for Romney:
The (bama +7 in NH is only a little bad for Romney, since PPP tends to lean Democratic by a point or two. The fact that a deficit of 5 or 6 points in a state he has to win is an OK result for Romney should tell you something about his situation. Being down 11 in Nevada is nowhere near OK for Romney, and We Ask America has given him good results in the past, so don’t blame the pollster.
But my favorite is Obama +8 in Virginia. This is the fourth VA poll published by Roanoke this year, and the first three all had Romney leading, most recently by 6 points. Given that background, and the fact that Obama has only been up 2-4 points in most recent VA surveys, this has to be one of the worst state polls Romney has received all year.
The election is not being held today, except for some early-voting states, so I’m not counting any chickens. But what instruments we have agree that Obama’s bounce is doing just fine (it’s actually not the convention bounce any more, it’s the extra bounce from the 47% tape, but whatevs).