A reader calls my attention to AP’s poll in Nevada of those who had already voted, Obama is ahead, 56-33.

Except… that poll is of an entire 318 people. Which means it’s got a margin of error at about 5.7 percent. (I suppose I should be pleased; the AP puts Obama ahead 52-34 among those who have already voted in New Hampshire, with a sample size of 59. No, I didn’t leave off a number, it’s one less than sixty.)

The DNA poll is of 7,147, which one of my readers puts at a margin of error of 1.15 percent.

But one reader writes in with some skepticism:

There are two mitigating factors you’re overlooking with your optimistic dispatch. One is that exit polls [stink], invariably, and no huge sample size can overcome that. A sample size of 1000 is just as good as a huge sample size like the one in this guy’s poll IF the respondents are scientifically (randomly) selected. The whole problem with exit polls is that the respondents are self-selecting. More enthusiastic voters want to talk. Sometimes this favors Democrats (see Kerry, John) but it doesn’t have to. For some reason, perhaps because they feel outnumbered and under-courted, Nevada Republicans REALLY want to affirm their positions.
The other problem is that, if it’s accurate, the exit poll reflects depressed Republican turnout relative to their registration numbers. This will never be good in a state where Democrats have drastically improved their registration numbers.

A few points on the “stinkiness” of exit polls… First, they’re of people who have actually voted, so that gives them one leg up on your average pre-election poll. Two, I suspect that the potential for self-selection is greater when polling is done in person than by phone. (The 2004 exit pollsters outside voting sites were disproportionately college-age women, which seems like a good formula for lots of “Of course I voted for Kerry, because I care so much about the rain forests. Say, could I buy you a cup of fair trade coffee when you’re done with your shift?” responses.)

“Nevada Republicans REALLY want to affirm their positions.” Okay, I’m open to that theory, but why? And why would Nevada Democrats who just voted early for Obama disproportionately not want to tell a pollster they voted for him? Is this some sort of reverse-Bradley effect where a voter doesn’t want to admit they voted for the popular cool guy?

Those who have not voted yet are disproportionately Republican. Yes, it is possible that enough of them stay home on Election Day to cede the state to Obama. But for the state GOP and the state McCain campaign, their job is to get them out. For the Obama camp, they have to look at the DNA poll and wonder how many Democrats have shown up at the polls and marked the box for McCain…