Veteran San Francisco Chronicle reporter Carla Marinucci provided a rundown today of how the New Hampshire polls could have been so wrong, quoting numerous experts, each offering their explanations. But the article and the experts fail to fully analyze how the pollsters got Tuesday's Democratic primary so wrong both before and after the election.
Last night, Chris Matthews, anchor of MSNBC's Hardball, revealed that initial exit polling -- polls taken of voters as they're walking out of polling stations -- showed that Barack Obama won by five percentage points on Tuesday. Fox News, which employs the same exit pollsters, had it 39 percent for Obama versus 34 percent for Hillary Clinton. Were those exit polls also wrong, just like the pre-election polls?
Despite all of the reasons proffered in Marinucci's story and elsewhere about Hillary Clinton's victory -- from her teary-eyed moment revealing her human side to women angry about how she was treated in the media coming out in huge numbers for her -- there appears to be only one plausible explanation for why both the pre-election and post-election polls were so different from the actual results: the so-called Bradley Effect.
First, let's dispense with three major theories so far for Clinton's surprise victory. The teary-eyed moment, the huge women's turnout, and a third theory that independents who told pollsters before the election they planned to vote for Obama but then voted for Republican John McCain because Obama was so far ahead they believed he would win easily -- some pre-election polls had him up by 13 percent. Those theories are all attractive and can explain the difference between the pre-election polls and the election-night results, but none of them explain the exit polling. After all, if voters switched to Clinton because of the teary-eyed moment or decided last-minute to vote for her in large numbers, or they suddenly turned from Obama to McCain, then why did more voters tell pollsters that they actually voted for Obama?
This is where the Bradley Effect comes in. It's named for former Los Angeles Mayor Tom Bradley, an African American who was leading in 1982 pre-election polls for California governor, but then lost on Election Day to white Republican candidate George Deukmejian. Post-election research showed that white voters were afraid to tell pollsters prior to the election that they planned to not vote for Bradley for fear of being labeled racists, but then went ahead and voted for Deukmejian in their private polling booths.
Marinucci's piece touches on the Bradley Effect, and whether it was in play in New Hampshire, but fails to note that it explains both the pre- and post-election polling showing Obama had won -- while none of the other theories do. If Democratic voters were afraid to tell pollsters they weren't going to vote for Obama before the election, then they would be afraid after the election, too. (Later, so-called "final" exit polls showed Clinton winning, but those polls were readjusted to "match" the actual election results.)
The debate over the Bradley Effect has touched off some heated arguments in the liberal blogosphere with some skeptics arguing that the pre-election polls in Iowa showed Obama ahead, and he ended up winning the election. So why wasn't the Bradley Effect in play there? That argument, however, overlooks how the Iowa caucuses are conducted. They're not done in secret, but in community settings in which voters must publically declare who they support. Anyone afraid of telling a pollster that they're not voting for the black candidate would be afraid of telling hundreds of others standing around them as well.
So what if the exit polls were just wrong? Well, if they were, then that also knocks out two of the three theories mentioned above. Because remember, the women coming out in huge numbers and the independents making an eleventh-hour switch to McCain theories are based on the very same exit polls. It was those polls (after they were adjusted to match the final results) that revealed large numbers of women voted for Clinton and fewer than expected independents voted for Obama.
So what about the teary-eyed moment? To believe that it dramatically changed the race, then you would have to believe that both the pre- and post-election polling was wildly wrong. It's possible, but is it likely? Polls have been proven accurate time and again. Look at Iowa, for example, or the Republican race in New Hampshire. In both cases, and thousands of others, the pre-election polls were dead on. But in New Hampshire, the post-election polls would have to have been off by a large margin -- 8 percent. Clinton eventually won by three percent.
If you're skeptical about the Bradley Effect, and think the polls were probably right, then check out liberal blogger Brad Friedman at BradBlog.com. Friedman, who claims to neither support Obama nor Clinton, is raising questions about the highly vulnerable Diebold voting machines used to count most of the New Hampshire ballots.
He and some supporters of GOP candidate Ron Paul are reporting interesting data comparing how Clinton and Obama fared in counties that used the Diebold machines versus those that did not. If their research is true, then Obama won the non-Diebold precincts while Clinton won the Diebold ones. Friedman is not alleging that the election was rigged; he just maintains that this type of data demands a hand recount of all the ballots. -- Robert Gammon