Social desirability describes the phenomenon when a respondent does not give an honest answer to an interviewer, typically on a controversial issue. The respondent wishes to provide the answer that is most "socially acceptable" rather than speak his true feelings. In other words, the respondent is actually responding with public opinion rather than his own personal views.
In general, there are two types of Social Desirability. The first is the respondent's desire to appear "politically correct"--the desire to answer a question as the power structure would have her answer. The second is a desire to please the person who is interviewing rather than give an accurate response. Social desirability theory can play havoc in polling and election predictions.
The basic cause of Social Desirability is the desire to please someone. It may take the form of giving the "right" answer as the power structure or majority opinion wold define it, or even more, a response might reflect a desired self image (one, for example, of being "enlightened" or "progressive") rather than true personal feelings. Responses given by white people to black interviewers often differ radically from responses given by white people to white interviewers.
There are numerous examples of the effects of this distortion. When David Duke, a well-known social dissident, ran for Senate in Louisiana in 1992, social desirability was widely reported as playing a part in underestimating the support for Duke. Respondents to polls did not want it publicly known they supported him. Duke regularly outperformed the polling agencies' predictions on this basis.
"The Bradley Effect" is a term that arose from the 1982 gubernatorial election in California, when white poll respondents regularly overstated their support for the black candidate, Tom Bradley. Bradley lost the race even though all polls showed him ahead. This phenomenon was frequently discussed during the Presidential campaign of Barack Obama.
In 2008, The Los Angeles Times rejected the usefulness of the "Bradley Effect." It argued that several liberal groups supporting Bradley imposed their views of gun control and many other hot issues on the Democratic platform, alienating many white voters who initially supported Bradley. Hence, the Times held that it was not "latent racism" that cost Bradley the race, but rather the ideological movement of the campaign itself. The Times suggested that the "Bradley Effect" specifically, and the Social Desirability effect more generally, have a limited use in polling predictions.