Why I Don't Think It's Over For McCain, Even Absent A Game-Changer

And this chart for early August of 2000:

And this chart showing the final polls from October and November of 2000:

Tell me, looking at those polls, what would your prediction for the 2000 election have been? What in these polls would lead you to believe anything other than that Bush would win, probably by a substantial margin (Answer: Following the tracking polls closely, and noticing the Gore surge at the end).

How about these taken from 1996 (taken from here):

Wouldn't you have guessed at least a double-digit win for Clinton/Gore?

Yes, polls have error margins. But the whole purpose of aggregating polls is that the error margins should cancel each other out. When all of the polls are on one side or another of the actual result, something else is probably going on other than sampling error.

Now what that something else is can be a source of considerable debate. Some will attribute it in these cases simply to bad luck: a campaign finance scandal breaking late against Clinton and a DUI arrest breaking against Bush. But I've always been skeptical that this was what really caused the late polls to break so heavily against Bush and Clinton.  After all, the final tracking polls still showed Bush ahead consistently (save Zogby, who you will note pretty well nailed both elections), even after the story broke.

If you look at the polls, they pretty much got Clinton's vote share in 1996 and Bush's in 2000.  What happened is that they underestimated Gore's and Dole's vote share.  My interpretation has always been that in both elections, the voters were confronted with a likeable agent of change against a dull, dour candidate who could largely be affiliated with the status quo (remember Dole's "bridge to the past" theme?).  They liked the change agent, but for different reasons had real doubts about him.  Sound familiar to anyone?

Would a similar pattern occur in 2008?  Who knows.  This election might be more like 1980, where last minute voters became comfortable with the change agent, and broke heavily toward him.  That is certainly the goal of Obama's massive ad buy right before the election.  It's probably where I'd put my money if I had to, quite frankly.

And to be certain, Obama is in better shape than Bush was in 2000, though he never quite hit the peaks Bush hit in 2000 either.  He isn't in as good of shape as Clinton in 1996 (at least not yet).  But if the economy stops looking like Armageddon is near and the Ayers/ACORN angle gets traction, Obama's edge might fade by a point or two.  Then, the 65+, uneducated, white independent voters who make up the bulk of the undecideds might hold the balance of the election in their hands.  Not certain that's great news for Obama.

Like I said, I expect Obama to win.  I think his odds are actually probably better than what he's presently trading at on InTrade.  But if you've been following politics as long as I have, you know that the polls are sometimes off in a manner that can't just be explained by error margins.  In 2000, Republicans were supposed to do quite a bit better in the Presidential and Senate races than they ended up doing.  In 2002, Democrats were supposed to do better in Senate races than they ended up doing.  In 2006, Democrats won the House with 7% of the nationwide popular vote, when of the seventy-five or so polls taken from July of 2006 through election day, seven (7) showed Republicans trailing by seven points or less (while 23 polls in that time period showed Democratic leads of 15 points or more).  The Democrats' lead that year from October 15 through election day was 14.5% -- nearly double what they actually got (It's also interesting to note that only 6 generic polls all year so far have shown leads for the Dems of 15 points or more).

So while I do think McCain is unlikely to win, I don't think his victory is contingent upon some Hail Mary pass or game-changer.  While something like that certainly could help, all that needs to be true is that the undecideds break against the charming change agent.  It's happened before. 

And that's without questioning whether the electorate will really be 40% Dem, 27% Rep, and 30% Indep, as some polls are assuming.  I don't know if there's a "shy tory" effect at work or not, but I'll be shocked if the electorate flips from D+2 in 2006 to D+13 today.  But hey, I've been shocked before . . .