Race of Interviewer

[Lunar]

Found this gem on the backslide of a blah Politico article:
Democratic pollster Celinda Lake said her polling found that “whites interviewed by blacks are 3 to 5 percent more supportive of Obama than whites interviewed by whites.” She said the effect was “most concentrated” among older whites who did not attend college.

But officials at other polling companies said they had not detected major problems with whites lying to black questioners.

In a review of 26,000 interviews conducted in September, the Gallup Poll “found no difference in the presidential vote choice of either black or white respondents based on the race of the interviewer,” said Gallup chief Frank Newport.

The NBC News/Wall Street Journal pollsters studied thousands of their interviews and came to the same conclusion.

“We took a hard look ourselves — it’s too important,” Newhouse said. “I was surprised to see that there was literally no difference. It was within a point or two,” and therefore not statistically significant.

Obama’s own internal polling similarly did not detect a significant Bradley effect, a campaign source said in an interview in August.

“Race of interviewer may make some difference, but I’m not a great believer that pollsters are being lied to because of the Bradley effect,” said Andy Kohut, a former director of Gallup who now heads the Pew Research Center for the People and the Press. Sampling error is a bigger problem, he said. “My continuous concern are the people we are not getting, because I know the people who are more likely not to do surveys have more negative views about African-Americans.”

Bositis agreed, in part because “the novelty of black candidates has worn off.”

“This election is, for the most part, not about Barack Obama. This election is about throwing the in team out,” he said. “And Barack Obama is the captain of our team.”

Wilder, now the mayor of Richmond, said that pre-election polls in his 1989 gubernatorial race undersampled GOP voters in Virginia, although there was a Bradley effect, as well, he said.

“I didn’t believe double-digit figures,” Wilder recalled. “Our internal polling showed us plus or minus 2 — a dead heat.”

But, he added, “I think quite honestly there was more of the Bradley effect than bad polling, in terms of exit polls.” And Wilder said he remains concerned that “raw racism, ... coupled with fears about who is going to be a leader, coupled with who has the right experience,” could alter the November election.

“Racism has not gone away,” Wilder said, “nor will it ever.”

http://www.politico.com/news/stories/1008/14343_Page2.html

Lunar's thoughts, if Gallup/NBC are wrong, and they probably aren't, the question is who is doing the lying.  Are whites lying to white interviewers or lying to the black ones?  Are blacks lying?

[Lunar]

How exactly do you know that you are interviewed by an African-American ?

Is he/she going to say: "Hi, I´m conducting a survey on behalf of Mason-Dixon. I´m African-American, just for your information."

? ? ?

Black people have different voices, it's pretty obvious most of the time over the phone.

[Lunar]

Well, the article doesn't provide internals, but Gallup (26k people) and NBC found no statistically significant difference. 

I wonder if one confined oneself to racially heterogeneous areas, if that would increase the difference between a black interviewer and a white one.  That might be what the Democratic pollster is talking about.

And again, one can only establish the fact that the gap exists (which Gallup and NBC say do not, some Dem. pollser says 3-5%), not where it comes from, which is, again, likely some combination of white people lying to white pollsters, white people lying to black pollsters, black people lying to black pollsters, and black people lying to white pollsters. 


I do not know what most pollsters do for their interviewer hiring practices, maybe I'll ask Vorlon after this thread drifts off the front page (to give him a chance).  Do you make your interviewers a demographic model of the area you're polling, make them all white, randomly select them?  What about white people that sound black or vice versa, haha.  Maybe one could select interviewers that to not have an obvious race to their voice?  hmmm