The Bradley Effect:

There should be a support group for all those beleaguered progressives who over the years anxiously awaited elections in the futile hope that the polls showing their candidate behind would turn out to be wrong -- but who this year are fretting just as much that the polls showing their candidate ahead are wrong.

--David Kurtz, TPM

In 1982, Tom Bradley led in the polls from the start of the California governor's race up to election day. The exit polls showed him winning a clear victory over the GOP nominee, George Deukmejian, and seemed to be on the verge of becoming the nation's first black governor since Reconstruction.

And in the end, he lost. Since then, every time an African-American politician underperforms his poll numbers, the phenomenum known as the "Bradley Effect." It happened when Douglas Wilder actually became the nation's first black governor in 1989 by a margin much smaller than his projected total from exit polls, and then when Harvey Gantt was beaten by Jesse Helms in 1990 for the US Senate, and even more recently, when Barack Obama was unexpectedly defeated in the New Hampshire primary by Hillary Clinton at the beginning of the year.

The "Bradley Effect," the notion that there is a hidden racist vote that doesn't appear in the polls, is an anchor that every African-American politician has to carry when seeking office before a predominantly white electorate. And it is the chief reason why even four days before the election, with a larger lead in the polls than anything Bradley or Gantt had at this time in their losing efforts, there is still some skepticism among liberals that Senator Obama really has this one in the bag.

Clearly, the Bradley Effect is something that has diminshed over time; as political historian Sherry Bebitch Jeffe points out, Tom Bradley was really screwed by a hidden racist vote in his first, unsuccessful campaign to be Mayor of Los Angeles, when Sam Yorty, one of the last vestiges of the pre-Depression Democratic Party in California that was more allied with William McAdoo and the South, painted the former cop as a secret Black Panther, and turned a sixteen-point Bradley lead into a six-point deficit on election day. Four years later, Bradley decisively defeated Yorty in the rematch, and went on to win a record five terms as Mayor.

Still, there is some reason for discomfort. Obama consistently underperformed his polling numbers when it actually came time to vote during the primaries; first in New Hampshire, then on Super Tuesday in states like Massachusetts and California, then later in Ohio, Pennsylvania and Texas. He ultimately won his party's nomination because he was better organized, state-by-state, than his opponent, allowing him to survive some early defeats and take charge of the race after Super Tuesday, when Hillary Clinton's campaign ran out of money.

And there's the original Bradley Effect, in 1982. There has been quite a bit of revisionist claims that even in 1982, there was no hidden racist vote. Both Lance Tarrance, who polled for the winner, and William Bradley (no relation), who worked for the loser, as well as Prof. Jeffe, point to other factors in Deukmejian's win, including a gun control measure on the ballot that pulled in a lot of conservative voters, as well as a first-rate absentee voter drive that year by the GOP. In particular, William Bradley focuses on the flawed sample of voters polled by Mervyn Field on Election Day, a sample that also caused his outfit to project a victory for Jerry Brown in his losing Senate race against Pete Wilson; anyone who remembers the 2004 Presidential Election knows to take exit polling with a grain of salt when it comes to projecting elections.

But there was a Bradley Effect in 1982, and it can be found not in the exit polls, but in the final polling Mervyn Field did before election day. The link is to every major statewide race Field polled since 1948, when he, like the rest of his brethren, blew the Truman-Dewey election. Since then, Field's final polls have correctly predicted the winner in all but three races, and two of those races involved picks that held tiny pre-election leads and lost by slim margins. The outlier was the 1982 governor's race: Field's last poll showed Bradley with a comfortable eight-point lead, whereas the actual vote showed a two-point margin, a 10-point switch.

OK, so maybe his sample wasn't an accurate cross-section of California voters, as his exit-polling on Election Day would indicate, and as Tarrance, Jeffe and the other would concur. But the real problem with that argument is Field's final poll of the aforementioned US Senate race, showing then-Governor Brown losing by six points to Pete Wilson. As the chart shows, that was the exact margin Brown lost to Wilson. And it was a concurrent poll, taken in the final week of the campaign.

So the same polling sample, conducted concurrently, correctly predicted not only the win by the Republican Senate candidate, but the ultimate margin as well, while blowing the governor's race by ten points. I find it hard to believe that Field nailed the race that involved two white candidates, but somehow didn't pick up some trend having nothing to do with racism that would have skewed the same polling sample when it came to the governor's race.