The Presidential Polls

Obama leads in a number of polls; McCain leads is a couple of polls. But Obama's campaign can't take much solace from that fact. All of the recent polls I've looked at that had Obama in the lead (maybe six or so of the most recent polls) are polls of registered voters. That may sound sensible enough, except that elections aren't determined by registered voters, but by voters who turn up and the polling place on election day.

Historically, Democrats have a lead among registered voters, but Republicans tend to turn out in greater proportions at the voting booth. So surveys of registered voters will tend to overreport support for Democratic candidates relative to who shows up to vote on election day.

So that fact that Obama leads in polls of registered voters, while McCain leads in two recent polls of likely voters (although one of those "leads" is within the margin of statistical error, si technically should be counted as a "tie" with Obama), has to be pretty sobering news for Obama's campaign.

Still, I'm unsure that even the polls of likely voters provide much solace to the McCain campaign: Surveys of likely voters depend critically on models of who the pollster predicts is in fact a "likely" voter. Those models seemed to track pretty well historically, but haven't seemed to be doing as well in recent presidential elections.

And there's reason to think that models based on past behavior won't predict turnout for this election either: those groups of Democrats who historically had tended to vote the least -- the young and minorities -- are those groups most activated by Obama's campaign. So I don't put must trust in "likely voter" models based on turnout in previous campaigns.

A couple of final thoughts. First, as an historical matter, that McCain is even as close as he is at this point in the campiagn is pretty surprising. GOP candidates often poll far behind their Democratic rivals during the summer, and then catch up either to win (Bush v. Dukakis) or lose very closely (Ford v. Carter). But, again, the lengthy Democratic campaign may suggest that this campaign should not be analogized to previous campaigns.

Secondly, that a Republican candidate is this close to the Democratic candidate, given eight years of a controversial Republican presidency, and continuing animus against GOP congressional candidates, also says a lot about what a strange election year this appears to be.