Okay, this post is for those of you out there who thought they had this Bradley Effect figured out, only to see or read something each day that throws you back into uncertainty about what it really means and what, pray tell, is really going on as this long, strange election finally winds down.
You probably already know from Wikipedia that the Bradley-Wilder effect is named for a former Los Angeles Mayor, Tom Bradley, who lost in 1982 in his bid to become Governor of California after all the polls had him so far ahead that his election was seen as a certainty. When he lost, the conventional wisdom was that he'd lost because enough white voters couldn't bring themselves to vote for a black candidate that they lied to pollsters about their feelings and intentions.
Go below the fold for what dummies like you and I should know about this effect and how it plays (or doesn't play) for the Obama-McCain election.
For excellent, professional answers about the Bradley Effect go to pollster.com and read the archived essays there by Mark Blumenthal. He has answered this question ad nauseam, but his discussions are so precise and technical that the dummies among us can easily get lost or make mistakes.
We 'dummies' include: many, if not most, Main Stream Media anchors; professional pundits like Mickey Kaus at the New York Times Bloggingheads website; some professional political scientists (no names out of professional courtesy); Jesse Jackson, as caricatured on SNL; most internet bloggers and most of us in the general public.
I heard Mark on a Philadelphia area NPR show called Radio Times last week and his discussion with the host of that show finally cleared away the cobwebs. Go to the WHYY website and download that episode and you will never again be confused about the Bradley Effect.
This essay is meant to save you the time and effort if you can't hear Mark's discussion for yourself.
When race comes up in conversations about this election most of us have two big questions. How big a factor will race be in this fall’s election? And can the polls be trusted when one candidate is black and the other one is white?
Although, the Bradley Effect applies to only one of those questions, many pundits use it to answer both. Lets talk about the first question - the one where the Bradley Effect does not apply.
A month ago the Associated Press and Yahoo reported on a survey they commissioned at Stanford University that found that "Obama's support would be as much as 6 percentage points higher if there were no white racial prejudice."
The study went viral and led many pundits to conclude (erroneously) that unless Obama was ahead of McCain by six points or more, he'd lose because of - wait for it - the Bradley Effect.
Get it? The Bradley Effect has become a stand-in comment for a generalized concern, fear, or hope that hidden white resistance to minority candidates creates an invisible hurdle that observers have to discount when thinking about November 4.
If pundits just said "I fear there is hidden (or not so hidden) white resistance to minority candidates that creates a hurdle that observers have to discount when thinking about November 4" they'd be fine and possibly accurate, depending upon how a big a hurdle they imagined blacks would have to supplant. If they go for the short-run, "I'm worried about the Bradley Effect," that's when they get in trouble.
So now that we know what the Bradley Effect is not, what is it?
Let's assume that nationally at least 6 percent of white voters oppose Obama on racial grounds. Does this mean Obama will lose the election if polls only have him up by 51 to 48 with a margin of error lower than +/- 2 percent)? Or does this mean Obama will win the election if polls have him up by 51 to 48 (same margin of error) because the voters who oppose him on racial grounds have already been accounted for in the survey?
Depending upon how big you think the Bradley Effect really is (2-3 percent used to be the going rate, but let's stick with 6 percent), Obama will lose the election if the Bradley Effect is real. He will win if it is not.
So what is the answer Blumenthal gives? Well, he cites the work of a Harvard University graduate student named Daniel Hopkins who did a study looking at elections with black candidates pitted against white candidates since the 1980s. Blumenthal reports that the further back one goes the more real the effect, but in most recent elections after the mid-1990s the effect disappears.
Hopkins and Blumenthal do not argue that some whites have deep-rooted feelings against black candidates and won't vote for them. They report evidence that for most elections, pollsters in the aggregate capture those feelings in their surveys.
What about the New Hampshire democratic primary, when Hillary pulled off a 'surprise' win against Obama? Wasn't that the Bradley Effect? Well, no. The polls got Obama's results right, 39 estimated versus 38 actual. They got Hillary's wrong, 30 estimated, 40 actual. (Go here for a excellent chart and discussion http://matthewyglesias.theatlantic.c...
If anything in the democratic primaries the pollsters sometimes underestimated Obama's victory margins for several reasons, including underestimating black turnout and support.
So if the Bradley Effect is now a shibboleth, why should we still be afraid of the polls and their accuracy? Well there are still problems figuring out the cell phone vote (people who only have a cell phone); the newly registered vote; how many democrats, republicans and independents to put in a sample - and that's only for a preliminary list.
There still may be many surprises on November 4, but the Bradley Effect will not be one of them.