Exit poll madness
Posted: Thursday, April 24, 2008 3:05 PM by Mark Murray
From NBC's Mark Murray
By now, you've probably heard it a 100 times. Obama has a problem with working-class voters. Clinton fares poorly among African Americans -- younger Democrats, too. And these disadvantages will be big problems for either candidate, if they become the nominee.
Indeed, after 45 Democratic contests, a familiar pattern has emerged from exit polls: Clinton performs well among women, seniors, and low-income whites; Obama wins blacks, higher-income folks, young people, and independents. And pundits, campaign operatives, and superdelegates are now poring over these numbers as if they are a political crystal ball that will tell us the future about the general election.
But has this analysis gotten a bit out of hand? Consider what the exit polls told us about McCain. In South Carolina, Huckabee beat him by wide margins among weekly church-goers and born-again Christians, and Huck even beat him among those without college degrees. In Virginia -- even after McCain became the presumptive GOP nominee -- the Arizona senator once again lost decisively among weekly church-goers, evangelicals, and those without college degrees, and he also lost among those making less than $100,000. What's more, in this week's Pennsylvania GOP contest, more than a quarter of the vote went to Huckabee and Paul, not McCain.
While McCain is unlikely to do as well among evangelicals as Bush did, does anyone think that his performance with this group is a big general-election problem for him? Probably not. Is he doomed among voters without college degrees? Unlikely. Will he be unable to get a quarter of the GOP vote in November? Forget about it.
No doubt that exit polls are useful at analyzing particular races. Without them, we wouldn't have known how well Clinton performed in Pennsylvania among white women and the suburbs, which were important keys to her win.
But extrapolating their findings to tell us something about another election -- with different candidates and different voters -- is a dubious exercise.