Bradley Effect & Automated Polling

Great Debates
1

The Bradley Effect being, of course, the notion that a minority candidate will underperform in an election vs his polling numbers because people will give inaccurate responses to a poll in order to avoid sounding racist. So perhaps 5% of the people claiming to vote for Obama in a poll actually won’t but lied about it.

Having been polled by Rasmussen recently, the point jumped out at me that the poll was completely automated. I had no one to impress except for the computerized voice that responsed to all my opinions with an encouraging “Ok… how about this question…” There would have been no reason for me to lie unless it was to myself.

I know SurveyUSA uses automated polling but couldn’t find an answer for Gallup. I’m guessing that the smaller or local outfits use live interviewers. In any event, is there a reason to believe that the Bradley Effect would have a significant impact on automated polling?

2

I have in my hot little hands a textbook on HR practices. It quotes a few studies of varying ages (1969, 1989, and 1995) that claim that people are more likely to disclose information that might reflect negatively on them to a computer rather than to a human. The book says people reported their GPA, desire to stay with the company, etc., more honestly to a computer than to a live person. These studies looked at job interviews conducted at a PC (or probably terminal in the earlier studies), not at polls done by phone, but maybe the same effect would work in both cases. I don’t know, of course. Just offering some data.