October 16, 2008
Polls may underestimate Obama’s support by 3 to 4 percent
Current polls of the presidential election may be underestimating Barack Obama’s support by 3 to 4 percent nationally and possibly larger margins in the Southeast and some strongly Republican states, according to UW researchers.
Psychologist Anthony Greenwald and political scientist Bethany Albertson, who analyzed data from the 32 states holding Democratic primaries, said race played an unexpectedly powerful role in distorting pre-election poll findings and the same scenario could play out in the election between Obama and John McCain.
“The Clinton-Obama raced dragged on so long, but it generated a lot of data. It is the only existing basis on which to predict how a black candidate will do in a national general election,” said Greenwald, who pioneered studies how people’s unconscious bias affects their behavior. “The level of inaccuracy of the polls in the primaries was unprecedented.”
Prior to the start of the primary season, the UW researchers thought the so-called Bradley effect would play a key role in the 2008 election. Previously, this effect showed exaggerated pre-election poll support for black candidates in some prominent elections in the 1980s and 1990s.
The Bradley effect is named for former Los Angeles mayor Tom Bradley, an African American, who lost a close 1982 gubernatorial election in California after holding a solid lead in the polls. As the 2008 primaries played out, Greenwald and Albertson found that the Bradley effect only showed up in three states — California, New Hampshire and Rhode Island.
However, they found a reverse Bradley effect in 12 primary states. In these states they found actual support for Obama exceeded pre-election polls by totals of 7 percent or more, well beyond the polls’ margins of error. These errors ranged up to 18 percent in Georgia.
“The Bradley effect has mutated. We are seeing it in several states, but the reverse effect is much stronger,” said Greenwald. “We didn’t have a chance to look at these effects before on a national level. The prolonged Democratic primary process this year gave us a chance to look for this effect in 32 primaries in which the same two candidates faced each other.”
Albertson and Greenwald believe the errors in the polls are being driven by social pressures that can operate when voters are contacted by telephone prior to an election. They said that polls from states in the Southeast predicted a large black vote for Obama and a much weaker white vote. They found that, in a few Southeast states, exit polls showed that both whites and blacks gave more votes to Obama than the pre-election polls had predicted.
“Blacks understated their support for Obama and, even more surprising, whites did too. There also is some indication that this happened in such Republican states as Montana, South Dakota, Oklahoma, Missouri and Indiana,” Greenwald said.
“If you call people on the phone today and ask who they will vote for, some will give responses influenced by what may be understood, locally, as the more desirable response. It is easy to suppose that these people are lying to pollsters. I don’t believe that. What I think is they may be undecided and experiencing social pressure which could increase their likelihood of naming the white candidate if their region or state has a history of white dominance. They also might give the name of the Republican if the state is strongly Republican.
A good analogy of a desirable response and social pressure, he said, would be if you lived in Detroit and you get a call asking if you will participate in an anonymous survey about automobiles.
“You agree and are asked if you prefer American or foreign cars. Even if you own a Japanese car, you might experience some pressure to give an answer that might be more appreciated by the caller — that you prefer American cars,” said Greenwald. “When it comes to politics, although voters are presumably anonymous when speaking to pollsters, the fact that the person calling them knows their phone number may not let them feel anonymous.”
Albertson noted that the polls have systematically underestimated Obama’s support and this can have an impact on the election.
“This distortion is interesting because poll numbers are part of the story journalists tell the public and they can also affect campaign strategy, such as states in which to spend resources,” she said.
A graph showing how the Bradley and reverse Bradley effects played out in the Democratic primaries is available at http://faculty.washington.edu/agg/Bradley_&_Reverse_Bradley.Dem_Primaries.2008.pdf.