Our main purpose here was to demonstrate that issues associated with some degree of social desirability or undesirability lead to differences in responses to survey questions about these issues, depending on which of two interviewing methods is used. Our findings seem to indicate strongly that telephone surveys in which respondents talk to human interviewers suffer from a much higher level of social desirability bias than do online surveys where respondents have already agreed to be surveyed and give their replies to a computer.

We do not claim that this is the last word on social desirability bias for these methods. We do not know the truth, so we can only guess at the magnitude of the bias in both samples. We think it likely that both surveys we used to test the concept underestimated the "socially undesirable" behaviors and beliefs, but we have no evidence to support that belief. Given that we asked these questions in only one telephone and one online survey, it is possible that there are other errors in one or both surveys that we cannot quantify.

In addition, in spite of our best attempts at trying to match samples with regard to demographic qualities and to weight for these factors, as well as attitudinal differences that might occur, differences between our online sample and the telephone sample could explain a portion of the differences we observed. However, based on our fairly extensive knowledge of such differences, we do not believe that this would be sufficient to explain the size of the ones we found. As with much research, the old adage "more research is needed" surely applies here.

Humphrey Taylor is chairman of The Harris Poll, David Krane is senior vice president of public policy, and Randall K. Thomas is director of internet research and senior research scientist, Harris Interactive.

Additional reading