On CNN this afternoon, Bill Schneider took the Bradley Effect theory to a whole new level.
In talking about Ohio polling, he highlighted the following numbers:
Obama 49%
McCain 46%
Unsure 5%
and said that Obama couldn't feel secure in that lead because the number of 'unsure' voters was greater than his lead.
What was his rationale for suggesting 'unsure' voters might all break for McCain? The fact that very few African-American voters are unsure.
For the few of who who don't know, the Bradley Effect was a polling phenomenon that suggested minority candidates underperformed their polling numbers on election day due to people who were reluctant to admit they wouldn't vote for a minority candidate when talking to pollsters. Nate at 538 analyzed the phenomenon and pointed to a Harvard study indicating that there was no statistical evidence of it in the last 10 years of elections.
What Schneider suggested today isn't that people who won't vote for Obama might lie to pollsters though. What he suggested is that the only sure support Obama gets comes from African-Americans, and that if anyone else at this point in the game is still not sure who they're voting for, the Obama campaign should expect them to vote for McCain.
Nate had another post recently about how he allots undecided voters in his modeling. Looking at the Democratic primaries, for instance, he found that 60-65% of undecided white voters went for Clinton (about the same number that were voting for her anyway). Using a number of other demographic factors, the ratio he arrived at for Ohio in splitting undecideds was about 53:47 McCain.
53% McCain. Not 100% McCain.
There is simply no logical, rational reason for Schneider to think 100%, or anything close to 100%, of 'unsure' Ohio voters might go for McCain -- other than, perhaps, Schneider's own prejudices.
Schneider's been flogging the theory that all undecided voters might go for McCain, solely because they aren't African-American, more than once this week:
CAMPBELL BROWN: So, OK, you mentioned '82, the Bradley race, that was what 26 years ago. The country is very different now, I think you could say. We've advanced somewhat in terms of our attitudes. Do you think it still exists, The Bradley Effect?
SCHNEIDER: Well, it appears to have gotten smaller. The research indicates that in 2006, the last midterm election, the pre-election polls were actually pretty accurate in predicting how the races would go if there was a black candidate and a white candidate. And one new report says that in a lot of southern battleground states right now, the -- that Barack Obama is doing much better than John Kerry did four years ago.
BROWN: Well, let me ask you about that. I mean, are there certain states where the Bradley Effect may be more of an issue than in other states?
SCHNEIDER: There are several battleground states right now where Obama's margin is very close, like Florida and Ohio. Now, take a look at Florida. It's very interesting. Obama leads by 3 points. 5 percent of voters say that they're unsure. Very few black voters are unsure how they're going to vote. If those unsure voters end up going for McCain, that would overtake Obama's lead in Florida, which is very narrow. So, Obama cannot feel comfortable unless his lead is larger than the number of unsure voters.
We could also see a reverse Bradley Effect. Where an African-American candidate does better than predicted. You could see that in Virginia and North Carolina, if black turnout turns out to be unexpectedly high. We might also see whites voting in unexpectedly high numbers for Barack Obama if they want to make a statement for change. What could happen is, economic anxiety could overwhelm the Bradley Effect.
BROWN: Interesting stuff, Bill. Basically, you're telling us, we don't have any idea. We're not going to know until November 4. But thank you for clarifying that.
SCHNEIDER: OK.
Schneider, and CNN, need to be called on this, well, bullshit immediately. His email address is:
bill.schneider@turner.com
If you want to be polite, ask him for something to back up his theory that 'unsure' voters will vote any differently than voters in the same demographic group who have already decided who they'll vote for (which is what I just did):
Mr. Schneider-
This week on CNN, on multiple occasions, you have put forward the idea that voters identified as 'unsure' or 'undecided' by polling in states such as Ohio and Florida might shift en masse towards John McCain. The only rationale you have offered for this shift is that those 'unsure' voters are probably not African-American.
Quite frankly, I have no idea what could have led you to the conclusion that McCain might win all 'unsure' non-AA voters. There's absolutely no reason that I've ever seen to think undecided voters of any race will vote differently than decided voters of that race. If you have some evidence to the contrary I would be very interested in hearing about it, because right now the theory you're espousing could (charitably) be described as specious.
Thank you for your time.
UPDATE: Just received a rather prompt reply to my email. Schneider pointed me to an AP article which indicates that in elections featuring an African-American candidate and a white candidate, undecideds break heavily for the white candidate.
The article, however, was written in 1989. One of the races it cites is Dinkins beating Giuliani...
I've changed the diary title as well. I still think the theory Schneider is offering has a racist element to it, but he's basing it on (dated) conventional polling wisdom rather than his own assumptions, so I'll give him some benefit of the doubt.
