Mark Blumenthal | October 15, 2008
Topics: Adam Berinsky , Bradley/Wilder , Gary Langer , Kathy Frankovic
Just yesterday, I expressed a bit of frustration that much of the recent media focus on the "Bradley-Wilder Effect." Much of it, and much of the debate in the blogosphere, inevitably turns on looking backward at what surveys did or did not miss 20 or 30 years ago. Instead, my hope is that pollsters will share with use new measurements based on their current survey data that can help us determine whether any such "effect" is in store for 2008.
Today, the ABC Polling Unit and its director Gary Langer posted data indicating that on at least one measure, the 2008 surveys are not yielding evidence of the "effect."
First, a bit of background for those who are just tuning in: The pattern seen in surveys in the 1980s and early 1990s was that in races pitting black Democrats against white Republicans, the final public opinion polls would accurately forecast the percentage of the vote for the African American candidate but would understate support for the white candidate. One piece of evidence that the effect was real was the apparent presence of "race of interviewer effects." Here is the way I described the issue in a column earlier this year:
Pollsters concluded [20 years ago] that racial attitudes were at work in the Bradley-Wilder effect when they observed that white respondents tended to report different vote preferences depending on whether the interviewer was black or white.
The best known study of this phenomenon -- a 1989 Public Opinion Quarterly article [PDF] based on surveys of the Virginia governor's race -- closed with an "unambiguous recommendation: survey organizations must record the race of interviewers and check for these effects whenever they conduct polls in black-white electoral contests."
That is exactly what Langer and ABC have done with 7,281 white respondents interviewed so far this year, and they see no evidence of race of interviewer effects in their recent surveys:
Among registered voters in a dozen national ABC/Post polls this year, 53 percent of white respondents told white interviewers they supported McCain - as did 52 percent of white respondents speaking with black interviewers. Forty-one percent of whites told white interviewers they supported Obama; an identical 41 percent said the same to black interviewers. And 93 or 94 percent of back respondents backed Obama, regardless of the interviewer's race. (It's unclear, moreover, how well respondents can even tell the interviewer's race in a telephone survey. In one of our recent polls 55 percent identified it correctly, but 23 percent were wrong and 22 percent declined to hazard a guess.)
The pollsters at CBS News did a similar analysis earlier this year and also found some "limited" evidence of race of interviewer effects in surveys conducted in 2007, but not in 2008, and then only on questions asked about the primary election contest between Clinton and Obama. Here is the way CBS polling directory Kathy Frankovic described that evidence recently (emphasis added):
This spring we aggregated several polls and looked at the interaction between the race of the interviewer and the race of the respondent and how that might impact vote choice. We did find some race of interviewer effects among white respondents, but it was relatively small, and appeared limited. White Republicans, the oldest white voters and the youngest white voters, and whites in the South and the West, showed no race of interviewer effect. There was some among white voters who were Democratic identifiers, those who were between the ages of 44 and 64, and those who lived in the Northeast or Midwest. However, more recent polls haven't found a significant race of interviewer effect. But clearly we need to look at this on a regular basis.
Again, pollsters including Langer and Frankovic continue to debate the theories and evidence behind the apparent "effect" seen 20 or so years ago. That is an interesting discussion, but not particularly relevant to the 2008 election. The bottom line: if present day surveys are yielding little or no evidence of race of interviewer effects, that is a big strike against the possibility of a pronounced "Bradley effect" this year.
Bonus: And speaking of looking backward, friend-of-Pollster Adam Berinsky reminds me of his three-part series on past evidence of the Bradley-Wilder effect. His second post in the series provides some very helpful explanation of why interviewer effects are an important clue:
They key to understanding the Bradley effect from this point of view is to think about who would be uncomfortable answering the candidate preference question. For this, we need to put each election in its proper context. The 1989 New York City mayoral election is the case I'm most familiar with, both because I've done some analysis of the pre-election polling data, and because I was a New York resident at the time. In that election the preferred candidate of older Jewish Democrats (or, as I like to call them, Mom and Dad*), Ed Koch, lost a contentious Democratic primary to David Dinkins, who is black. Considering that many older Jewish Democrats had never in their life voted for a Republican candidate, a vote for Giuliani in the general election could be seen as nothing but a vote against Dinkins. Indeed among Jews over 50, 30 percent claimed that they didn't know who they were going to vote for a week before the election, even though 93 percent said they would definitely cast a vote (among non-Jews under 50, 7 percent said they didn't know who they were going to vote for and 89 percent said they would definitely vote). These are the precise circumstances where we would expect to see the polls perform poorly - and they did.
The whole series is worth reading in full.