Somebody, Please, Clobber the "Bradley Effect" with a Rusty Shovel Already
The Marvel No-Prize for "stupidest column of the day" goes to (drum roll) this slithering heap of offal by Susan Estrich, better known to you and me as the brains behind the Dukakis landslide. In the middle of an argument that boils down to "gee, the election is sort of close," Estrich worries that the "Bradley effect" of white voters lying about their support of black candidates will sink Obama.
The experience of the primaries, not to mention that of other African-American candidates, suggests that polls tend to overstate, not understate, support for black candidates.
For a debunkinging of the "other African-American candidates" nonsense, go here.
With the exception of Indiana, every pre-primary poll in a major state showed the race between Obama and Clinton to be closer than it turned out to be.
This is true, but not to the benefit of Hillary Clinton. Here's what the polls said, then what the voters said, in all of the big states from the end of February to the end of May.
Wisconsin. Final poll average: Obama 48.9, Clinton 41.9. Final result: Obama 58.1, Clinton 40.8. Obama beat the spread by 10 points.
Ohio. Final poll average: Clinton 49.6, Obama 43.6. Final result: Clinton 53.5, Obama 44.8. Clinton beat the spread by 3 points.
Texas. Final poll average: Clinton 47.8, Obama 45.9. Final result: Clinton 50.9, Obama 47.4. Clinton beat the spread by 1 point.
Pennsylvania. Final poll average: Clinton 49.8, Obama 42.8. Final result: Clinton 54.4, Obama 45.3. Clinton beat the spread by 2 points.
North Carolina. Final poll average: Obama 49.6, Clinton 42.3. Final result: Obama 56.1, Clinton 41.6. Obama beat the spread by 8 points.
Indiana. Final poll average: Clinton 48.6, Obama 44.4. Final result: Clinton 50.6, Obama 49.4. Obama beat the spread by 3 points.
Kentucky. Final poll average: Clinton 63.2, Obama 28. Final result: Clinton 65.5, Obama 29.9. Clinton beat the spread by less than 1 point.
Oregon. Final poll average: Obama 54.1, Clinton 41.6. Final result: Obama 58.5, Clinton 40.5. Obama beat the spread by 5.5 points.
Three conclusions: One, Obama never got a lower share of the vote than the final poll average projected, and the closest he came was winning 44 percent of the Ohio vote when polls had him at 43. Two, on average Obama outperformed Clinton in beating the spread predicted by pollsters. Her biggest surge was a 3-points-bigger-than-expected Ohio win; he won by 10 points better than expected in Wisconsin. Third, Obama beat those spreads by outperforming expectations among white voters. He won a majority of the white vote in Wisconsin, pre-Jeremiah Wright. He won whites in Oregon, post-Wright. Which segues nicely into more Estrich nonsense.
Recent polls showing America to be as racially polarized as ever don't exactly give comfort to those who would dismiss the concern that some voters may be telling pollsters one thing and then doing something very different when they actually mark their ballots.
But they're not doing that! In Indiana, for example, Clinton won 78 percent of the white voters who said "race was important" in their vote—that was 10 percent of all white voters. In West Virginia, the site of Obama's biggest primary defeat, 21 percent of whites said race was important and they broke 84-9 for Clinton. The voters who actually have a problem with Obama's race are saying so when they vote against him.
Compare this to the 1982 race for governor of California between George Deukmajian and Tom Bradley. It was a complicated election, because conservative turnout was surging ahead of what polls projected as people came out to beat the anti-handgun Proposition 15. In the final Field poll, 3-4 percent of whites actually said they were voting against Bradley because he was black. But that same poll showed Bradley winning 47-41, with Deukmajian gaining, just presumably not by enough to overtake Bradley. There is no example of a similar reversal happening to Barack Obama in any 2008 primary; indeed, there's no example of a black candidate in 2006 (there were six such statewide candidates) underperforming the polls so badly. And yet people say that there is, and get paid for it.