Abstract
Unlike organisms in test tubes and Petri dishes,humans generally know when they take part in a research study. As a result, respondents often behave in ways to please the researcher,avoid embarrassment, or “look good”. In sodoing, respondents misrepresent their true preferences, resulting in social desirability bias (SDB). SDB is a complex and multidimensional construct caused by self-deception (anoverly favorable self-image), and impression management (Fisher and Katz 2000). Social scientists are keenly aware of the SDB issue and have developed a number of methods to minimize its influence.In the case of economic experiments, SDB is often thought to be mitigated with the use of monetary payments meant to increase the saliency of the task. However, the most widely used approach to detect and control for SDB in social science research is to investigate correlations between measures of SDB using psychometric scales and people’s behaviors,or answers to survey questions. SDBscales are constructed by asking a series of questions designed to determine if respondents say they engage in an activity that is socially desirable, but that is thought to rarely be acted on. For example, one question might ask whether the respondent agrees that they“have never intensely disliked anyone.” How-ever, most people have harbored an intense dislike for someone (e.g. an anonymous journal referee), but face social pressure to refrain from admitting or acting upon such feelings.Someone who agrees with the above statement is, in all likelihood,not telling the truth; they are misrepresenting their true feelings to provide a socially-desirable answer. An SDB scale or score is obtained by aggregating an individual’s responses to several such questions.The typical view is that, “… a significant correlation between scores on the [social desirability] scale and a target construct scale suggests that the data are confounded because of the respondent’s desire to answer in socially desirable ways, thereby seriously weakening the validity of the findings. Conversely, a low correlation suggests that the measure is relatively free of SDB,” (King and Bruner 2000,p. 88). We disagree. Rather than interpreting the low correlation between an SDB scale and behavior as evidence of a lack of SDB,we argue that such a finding can result as an artifact of the hypothetical context in which behaviors and attitudes are often measured. In this paper, a conceptual model of socially desirable behavior is developed, which shows that common interpretations of SDB scales may be misleading. We show that in hypothetical situations,where there is little or no cost to displaying socially desirable behavior, that such a view need not be true. As a result, utilizing SDB scales to neutralize SDB is unlikely to be completely effective, and therefore we discuss an alternative method that can be used in this regard – inferred valuation.