The Amazing Cognitive Dissonance in Presidential Polling

There's an amazing case of cognitive dissonance among American pollsters right now. According to their predictive models, registered voters lean Obama, likely voters lean relatively more towards Romney, and votes in the bank lean heavily Obama. Wait, what? Yes. Check out this teaser from today's NBC News/Wall Street Journal/Marist poll in Ohio:
One-in-five respondents (18 percent) said they have already voted, and, of those, almost two-thirds (63 percent) said they voted for Obama.
For anyone counting, that is already a 5-point margin of the entire Ohio electorate for Obama. Ohio might just be over for Romney. Mind you, this is a poll which NBC admits has a more conservative-skewed sample than the actual voting populous in 2008 (from exit polling) and still has Obama ahead 51-45, which means Romney is still underperforming McCain in the state. Not only that, that Romney voters aren't voting yet may well mean that Romney hasn't closed the deal for his voters the way Obama has for his.

Of course, voters can't vote twice, and the more early votes Obama gets, the more late votes, or election-day votes can be banked for Romney. So overall, it's a wash. Right? Riiiiight? Right, if you assume that the universe of who votes stays constant from the day early voting opens all the way to election day. It doesn't. Consider this: the more votes the Obama campaign can get banked early and checked off the list, the more undecideds, and even Romney-leaners the campaign's ground operations can go after in the days to come. The campaign no longer has to work making sure those certain-Obama votes are cast, and thus get to use that time and resources to making more certain-Obama voters and turning out newly registered but less-motivated voters.

Yesterday, I wrote about why the Obama ground game makes "registered voters" numbers in the polls, rather than just "likely voters" numbers important when one is trying to determine the positions of the candidates. Pollsters do now, and will continue to skew the likely voter universe towards Republicans, and it's not necessarily their fault. Republicans and conservatives vote more regularly than do liberals and Democrats, so any model charged to take history into account to predict the likelihood of voting of a given voter will invariably skew the model towards conservatives.

That, however, should not stop pollsters' human instincts from kicking in and figuring out when an election is different from all the others. That shouldn't stop them from understanding, in context, why Republicans in state after state tried to eliminate or shorten early voting periods. Numbers like those coming out of Ohio should give pollsters a pause about who they think the likely voters are. At the very least, it should trigger some readjustments in their "likely voter" models, if they insist on using it.

I doubt very much that Ohio is the only place this is happening. Early voting is a major reason Obama cleaned McCain's clock in 2008. And it is again going to be his secret weapon against Romney. Pollsters and the media can do whatever they want to keep the suspense going, but votes in the bank are votes in the bank. While the media is busy putting on their freak show, Barack Obama is slowly but steadily putting the votes in the bank.