F. Scott Fitzgerald

“The test of a first-rate intelligence is the ability to hold two opposing ideas in mind at the same time and still retain the ability to function.”
— F. Scott Fitzgerald

It’s time for what seems to be our annual cautionary post about polls.

In the past few days, we’ve seen national polls that show both Romney ahead (Gallup, by 7) and Obama ahead (Hartford Courant, by 3). Clearly at least one of those has to be wrong, right? Or maybe not. Perhaps the polls have different screens to figure out who they think is likely to vote, different ways to weight certain demographics (How big will the youth vote be? How big will the African-American vote be? How big will the Hispanic vote be?) So F. Scott Fitzgerald was right: They both could be right in their own way, even though they say different things because they’re based on different assumptions.

My first piece of advice: Ignore the national polls, no matter what they show (well, up to a point. A landslide is still a landslide, but we’re not at landslide territory yet for either guy). Some point out that Romney’s numbers in the national polls could be inflated because of his big advantages in the Southern states (see this Slate piece.) But the real reason to ignore the national polls is we don’t elect presidents in a nationwide popular vote. We elect them state-by-state through the Electoral College. So that’s where to look.

The best collection of state-by-state poll results is at Real Clear Politics. There are lots of ways to read them, of course. (My guy’s ahead! Yes! This poll is right on the money!  Or: The other guy’s ahead. This poll is clearly wrong and biased!)

Here are some things I look for. Let’s take two recent polls out of Ohio — one by Rasmussen, one by Survey USA. Both show Obama ahead, by 1% in Rasmussen, by 3% in Survey USA. Good news for Obama, right? Well, being ahead is certainly better than being behind, of course. But let’s look at how those two polls produced those leads:

* Rasmussen: Obama 49, Romney 48
* Survey USA: Obama 45, Romney 42.

The first thing I notice is — Obama’s ahead but he’s not yet at 50 percent. Yeah, yeah, there might be some third-party candidates that siphon off the votes so someone can win with less than 50 percent. Still, it you get 50 percent plus 1, you win, no matter what. And Obama’s not there yet, and that’s always a danger sign for an incumbent. Voters have had four years to size up Obama; they’ve had a lot less time to imagine Romney as a potential commander-in-chief. If you’re a Romney supporter, you have the hope that since those undecided voters haven’t yet gone with Obama, they might break your way.

And that brings us to the question of just how many undecided voters there are (and who — but that’s a more complicated question). Notice that Rasmussen has a lot fewer undecideds than Survey USA. I’d guess that’s because of how each pollster asked the question. Rasmsussen probably phrased the question in a way to force people to commit in a harder fashion than Survey USA did. (Even if they didn’t, the larger point is still true — pollsters ask questions in different ways, and gauge the likelihood of someone to vote in different ways.)

Personally, I’d think Survey USA is closer to the truth — not because it shows Obama with the bigger lead, but because it shows a larger undecided pool. Indeed, I’d think Republicans might prefer that poll, too, for the same reason. Even though it shows Obama with a bigger lead, the bigger pool of undecided voters means Romney still has a lot more opportunity there.

We see this contrast even more starkly in two polls out of Michigan. Both give Obama the lead, but in very different ways:

* EPIC-MRA: Obama 52, Romney 46
* LE&A/Denno Research: Obama 44, Romney 41

Obviously the one poll puts Obama over 50 percent. Game over. But the other poll shows Obama a long way from 50 percent — with a much higher pool of undecideds. Game still very much on.

So to glibly say, “polls in Michigan show Obama leading by 3 to 6 percentage points” is both true and quite misleading.

As get closer to Election Day, that pool of undecided voters ought to shrink. So if this were the day before the election, I’d be more inclined to believe the poll showing the smallest number of undecided voters.

Ultimately, it’s best not to hang on any specific number, but to pay attention to the direction the polls are moving: Is Romney still moving up or has his support levelled off? Is Obama dropping, or holding steady?

And, of course, there’s the best advice of all: The poll that really matters is on November 6.

Unless you’re one of those early voters, in which case you’ve already moved on.

For more consumer advisories on polls, we refer you to this piece from 2008.

– Dwayne Yancey