I'm anything but a stat-cruncher, but don't the Democratic Primary results make the best common-sense, real-life refutation of the so-called "Bradley effect?" I'm sure there are some logical holes in my argument--I never took a stats class in college--but here goes...
- Since most electoral vote projections have Obama pretty much at the cusp of or beyond the magic number of 270, it seems that the very least Obama has to do is win ONE of the remaining EIGHT swing states: Florida, Virginia, North Carolina, Ohio, Indiana, Missouri, Colorado, Nevada. Would the "Bradley Effect" really come into in ALL of the these states, which are demographically and socially different?
- Comparing the final results in some relevant tossup states to the final polling averages (via Real Clear Politics), here's my stab at a statistics ignorant statistical analysis:
Note: I left Florida off because of the "uncontested" primary and Nevada and Colorado because they were caucuses.
Missouri
FINAL RESULT: Obama 1.3
FINAL POLL AVG: Clinton 5.7
DIFFERNCE: OBAMA +7.0
Virginia
FINAL RESULT: Obama 28.2
FINAL POLL AVG: Obama 17.7
DIFFERENCE: OBAMA +10.5
Ohio
FINAL RESULT: Clinton 10.1
FINAL POLL AVG: Clinton 7.1
DIFFERENCE: OBAMA -3.0
North Carolina
FINAL RESULT: Obama 15.7
FINAL POLL AVG: Obama 8.0
DIFFERENCE: OBAMA +7.7
Indiana
FINAL RESULT: Clinton 1.4
FINAL POLL AVG: Clinton 5.0
DIFFERENCE: OBAMA +3.6
Not only do the numbers from these swing states show no consistent pattern that confirms a "Bradley Effect," the difference between polling and final results trended in favor of Obama. Except for Ohio, which seems to perhaps be the most dubious opportunity to pick up votes. And remember the narrative of the NC and IN: it was supposed to be Clinton's chance to pick up momentum, with even a shot at keeping things close in NC. Instead, Obama closed strong in IN and closed out the primaries that night.
Some other interesting results include:
California
FINAL RESULT: Clinton 9.6
FINAL POLL AVG: Obama 1.2
DIFFERENCE: OBAMA -10.8
Massachusetts
FINAL RESULT: Clinton 15.4
FINAL POLL AVG: Clinton 7.0
DIFFERENCE: OBAMA -8.4
So if there was a "Bradley Effect" in the primaries, it was in the bluest of blue states. Living in CA, I don't think race was the matter, at least in a conventional black-white sense: the talk was that Clinton drove up the Hispanic numbers, and Asian American numbers, with a smaller effect. Plus, I think that early voting banked Hillary a lot of votes before the Obama momentum was picking up. In any case, it felt like a pro-Hillary vote and not an anti-Obama vote here.
My favorite example of the "Reverse Bradley Effect" in terms of narrative wasn't the obvious case of South Carolina, but Wisconsin.
Wisconsin
FINAL RESULT: Obama 17.4
FINAL POLL AVG: Obama 4.3
DIFFERENCE: OBAMA +13.1
Why is this example my favorite? It was the post-Super Tuesday state that was crucial to Obama's push into the lead during the primaries. Remember how Clinton was supposed to close in on Obama in Wisconsin after the Devall Patrick/Xerox-gate speeech? Didn't happen, and the big rout helped Obama keep the wind at his back.
So my basic, completely unscientific, but common-sensical conclusion is that it would be hard to believe that there could be any kind of uniform dampening across all these states. Perhaps it might happen in one or a few states, but it's not looking like it could happen in all the battlegrounds, right?
Please feel free to correct or confirm these observations, if you're killing time on the west coast before Palin on SNL!
Note: If my analysis is too simplistic (which it probably is), go to Nate Silver's "Bradley Effect" thread here:
http://www.fivethirtyeight.com/...