I know this subject is slightly off-topic for many readers of our blog, but I thought it might be of interest to some, especially those interested in polling and the reliability of polling. A colleague sent along a link to this article in the Arizona Republic, “Election Pollsters Face New Challenges.”
Here’s a bit from the story:
But the polls may not be as reliable this year as they have been in previous presidential elections. Polling is getting more difficult for a host of reasons, from the increasing use of cellphones and declining polling-participation rates to the potential for racial bias among participants who are asked about the contest between Sen. John McCain and Sen. Barack Obama, the first Black person to top a major-party ticket.
The story talks about both technological challenges (especially those posed by cell phones) as well as traditional problems in survey research (insuring an accurate response, especially when the survey is focused on racial issues, or in cases where white and non-white candidates are competing for office). In such cases, “social desireability” biases can creep into survey responses; VTP colleague Adam Berinsky looked at this issue in the 1989 Dinkins-Giuliani race and found that many pre-election survey respondents might have expressed no preference to survey interviewers, instead of answering in a way that might have been socially undesireable (See Berinksy, “The Two Faces of Public Opinion”, American Journal of Political Science, Volume 43, pages 1209-1230.)