The Bradley Effect?

We hear often these days of the 'Bradley Effect', the theory that polled individuals tell pollsters they'll vote for a black candidate, then choose not to when actually in the booth.  This theory is named after the former Los Angeles mayor, ahead in the polls when running for governor of California in 1982, only to ultimately lose to George Deukmejian.  Might this theory be a bit overblown?

I recall the 1982 election somewhat vividly.  Yes, Bradley was ahead of Deukmejian in the polls prior to the election.  At the same time, in the race for US Senate, Jerry Brown (D) was running slightly behind Pete Wilson.  

On election nite, I watched coverage on San Francisco's then-NBC affiliate, KRON-TV.  When the polls closed, their analyst, Rollin Post, claimed that based on exit polling, Bradley would win, and Brown vs. Wilson was too close to call, with perhaps Brown having a narrow advantage.

As it turns out, Bradley lost a close race, and Brown was destroyed.  So, while the polling was wrong on Bradley, it was equally wrong on Brown.  This suggests race may not have been a factor in the erred polling.  

Why, then, were the polls off by so much?  In that election, there was a gun control proposition on the California ballot.  Every redneck from the boonies (conservative heartland) made it to the polls.  While voting against the proposition, they also voted Republican, adding huge numbers for Deukmejian and Wilson.

I can't speak to the Doug Wilder election, and I won't argue there are no instances of those polled telling a different story than how they actually vote, but in the Bradley election, the pollsters grossly misrepresented the rural vote, which was the primary reason they were so wrong.