Monday, April 28, 2008

Obama exit polls well, finishes lousy

One of the problems in the 24-hour news cycle is how quickly information is thrown out for consumption before it can be put into proper context. Mickey Kaus has been on the question of the reliability of exit polls and it's worth noting:

How Crappy Were the Exit Polls? Pretty crappy! They certainly didn't capture the 10 point Clinton win. [Looks like 9.2 to me--ed Print the legend!] According to National Review, the early exits even had Obama up by five. Maybe nobody believed that--but the exits still distorted the narrative of election night. The theme of the first New York Times story seemed to be

Mrs. Clinton faces major challenges going forward ...

The cable nets, primed by the exits, also spent the initial evening hours asking whether Hillary could or would go on--as opposed to why Obama had suffered an embarrassing drubbing that revealed real weaknesses. .... P.S.: Always trust content from kausfiles! ... P.P.S.: Brendan Loy has more on the crappy exits. ... P.P.P.S.: If the exit polls are this unreliable for press' result-predicting purposes, why aren't they also unreliable for all the scholarly purposes they are supposedly put to? Garbage is garbage, no? ...

The exit polls run into three major issues. One issue is the oversampling of urban voters. The second issue is the bias of the poll takers in selecting their sample face-to-face (what Mickey Kaus called the "kid in the hoodie phenomena"). Both of those factors led some to believe John kerry was going to win in 2004, for example. Both of those factors can be corrected by weighing them against the actual results. The problem, of course, is that the exit polls go from data for immediate commentary to data for later analysis. Not much good on live TV.

The third factor is the truthfulness of the voters' answers. In the Obama case, this has possibly meant an inflation of Obama's electoral strength, possibly even after the results are weighed. It'll be interesting to see the studies that come out examining whether some white liberals would rather lie than admit they voted for the white candidate.

Fox News last Tuesday actually did a good job of explaining why they weren't making any projections on the margin of Clinton's victory because of the pro-Obama bias in the exit polls. It still didn't stop them from using the exit poll data for discussion of voters' motivations, etc., when perhaps it should have. But at least they didn't go into immediate "story" mode telling viewers how to interpret the results. There was even a humorous moment where former Presidential Advisor Karl Rove was on camera trying to peek at the Pennsylvania Secretary of State's website, only to announce in frustration that the site "must've crashed."

Almost a week later, Brian Fraley has a look at the numbers from the Chicago Tribune that confirms a bit of what was suspected, that Obama is losing ground amoung "values" voters, specifically white Catholic Church-goers.

In each case, exit polls found regular Mass-goers registering within a few percentage points of Catholics who do not attend church regularly. In most of the primaries, Mass-goers were slightly more likely to vote for Obama–though not by a statistically significant margin. The same was true in the March primary in Ohio, a neighboring industrial Midwestern state that is much like Western Pennsylvania.

But that was not the case in Pennsylvania last week. Catholics who attend church weekly turned against Obama in much greater numbers than less observant Catholics. Catholics who attend church weekly voted 74 percent for Clinton and 26 percent for Obama; Catholic who do not attend church weekly voted 65 percent for Clinton and 35 percent for Obama. That’s a 48-point margin for Clinton among observant Catholics versus a 30-point margin among Cahtolics who are not regular churchgoers.

And that despite the anti-abortion, socially conservative Catholic Sen. Bob Casey Jr (D-Pa.) at Obama’s side throughout most of the Pennsylvania campaign.

Keep in mind that Casey is one of the most popular political figures in Pennsylvania (an advantage Obama did not have in Ohio) and outspent Clinton nearly 2:1 (ditto) and still suffered a rather convincing loss. Even if the weighted survey results come in pretty close to Ohio (with the church-going exception above) we're still looking at a significant decline in Obama's political strength. (Just think if Clinton had done a better job campaigning here what kind of race we would be talking about.)