Already, rumors of a contested election - no matter what the outcome - are swirling around the blogosphere. Anecdotes of touchy touch screens switching votes from Obama to McCain have materialized in West Virginia. The same thing is happening in Tennessee, but in reverse. A programming error in the touch screens used in New Mexico revealed that straight ticket votes left out the votes for President and Senate in democratic strongholds. Luckily this error was caught early and has been corrected. With stories of improperly purged voter rolls, allegations of voter registration fraud, potential polling place shenanigans, and warnings of long waits due to malfunctioning equipment - it's all enough to make you want to stay home on Election Day.
Which is exactly why you shouldn't.
Polls have been used, misused and maligned for years, and they have become an obsessive touchstone to explain or doubt all manner of uplifting or downright depressing trends about us, as a country. A depressed or unenthusiastic electorate stays home. But research is suggesting that one of the most dispiriting poll-driven stories of this election season - that racist voters would somehow mislead pollsters by saying they'll vote for Sen. Barack Obama when they'll secretly vote for Sen. John McCain - just isn't true.
Researchers are saying it is not very likely.
Researchers discovered that in telephone polling leading up to the primary, Sen. Obama's name was recited last. Listeners tend to remember the last name better and choose it more often - a well-known phenomenon that can artificially inflate a candidate's true support by a few percentage points. Conversely, voters tend to choose the first name listed on a ballot more than they should - another phenomenon called the primacy effect. In New Hampshire, Sen. Clinton's name was listed first, an advantage that could have led to a pickup of about 3 percentage points. Studies by Stanford researcher Jon Krosnick showed that 85 percent of statewide races in California from 1976 to 2004 had a primacy effect.
Incredibly, the first name you encounter on the Presidential ballot can depend on the political party of your governor (hint: Florida had a Republican governor in 2000), the party that received the most votes during the last election, or whose last name goes first alphabetically, among many, many other arbitrary reasons. Poor or unfair ballot design may be the true culprit in many close elections.
As for the "Bradley effect" itself, Dan Hopkins at Harvard University studied political races beginning in 1989 in which a black and white candidate squared off. Hopkins found that a slight Bradley effect may have existed until about 1995. After that, it actually became a slight anti-Bradley effect (black candidates did better than polls suggested) before leveling off into no discernable effect whatsoever.
That's not to say the racism has vanished from the country. Sadly, it hasn't, as amply evidenced by the frightening if ill-formed plot by two white supremacists to assassinate Sen. Obama, and by studies that routinely find unconscious bias even in people who honestly believe they are bias-free. (see Harvard's Implicit Study to try it for yourself: https://implicit.harvard.edu/implicit). In one experiment by Harvard researcher Mahzarin Banaji , Barack Obama was judged by study volunteers to be less American even than former British Prime Minister Tony Blair.
It would be truly depressing indeed if implicit bias strongly influenced voting behavior. But science suggests that we are perfectly capable of holding two contrary views at once, and carefully worded surveys this election season suggest that qualities like leadership and shared values are far more important to individual voters. Other surveys have found that some voters who might not normally vote for a Democrat support Sen. Obama solely because of the historic nature of his candidacy. Last night, Sen. McCain downplayed the potential for racism to factor into the election, telling CNN's Larry King that people will vote "for the best of reasons, not the worst of reasons." (I may not agree with many of his points, but on this one, I think he's right).
Exit polls have been rightfully scorned, but researchers have suggested abundant ways to make them better, and many of those suggestions are being implemented this year. And what of the wildly divergent pre-election polls that seem to be multiplying like rabbits? If they too are imperfect, and reflect the quirks of individual pollsters, perhaps there's safety in numbers. Few pollsters have released their methodologies, meaning that most cannot be adequately evaluated by researchers (one exception: most experts agree that the Zogby online polls just aren't scientific). But adding numerous polls together can cancel out the eccentricities of most, and the average is a more accurate snapshot of the public's true sentiments - a strategy adopted by many online polling analysts (like www.pollster.com and www.fivethirtyeight.com) this year. And despite the wildly erratic polls of the not-so-distant past, cumulative final polling data from state to state have converged to within a percentage point or so of the finally tally.
If nothing else, it's yet one more reason for the country's decision-making - and especially voting regulations - to be based on better science. The bigger picture is that neither deliberate lying nor implicit bias seems to be wreaking havoc on the polls, and that polls are less likely to mislead the public about what's really happening at the ballot box. Something that's proven to have an effect? Not voting at all.
