As you know, Gallup has been running three separate numbers in its daily tracking poll -- registered voters and then both a 'traditional' and an 'expanded' likely voter numbers. Much has been made, especially by McCain supporters, of the fact that the 'traditional' number has shown Obama ahead by as little as 2 points -- well within the margin of error. However, in the last few days, the three numbers have been converging. As of yesterday's reading
(the most recent), the registered number had Obama 11 points ahead, the 'expanded' number 9 points ahead and the 'traditional' number had him 8 points ahead.
Beside the numbers being generally good for Obama, one question I have is why the three numbers are converging, as opposed to moving in Obama's direction with the same spread between them.
One possibility is the number of voters who've already voted. Gallup is also monitoring how many voters say they plan to vote early and how many say they have already voted
early. I'm not certain how Gallup is computing their numbers. But the key in these different likely voter 'screens' is what you'll take as evidence from a voter to believe they're likely to vote. One of the key differences in a lot of these 'expanded' models, though I'm not certain about Gallup specifically, is whether the pollster is willing to believe you're a 'likely' voter even if you have not voted before. Now, once someone has already voted, they're not only a 'likely' voter but a certain voter. So I wonder whether some of the convergence may due to uncertain 'likely' voters, who made the cut in the expanded but not the 'traditional' screen, now reporting that they've actually voted and are thus now being counted in both tallies.
To be clear, I'm not certain about Gallup's methodology in this respect. So count this as informed speculation. But these are the approaches being used by a few other polling outfits. So I think this is a good supposition. We'll look into it today and report back to you.