Minimizing social desirability bias in measuring sensitive topics

Social science research has long been concerned with attitudes, beliefs, and behaviors that are potentially objectionable, immoral, or illegal. These types of topics include, for example, racism, ableism, cheating, and stealing, among others. Referred to as 'sensitive topics', their investigation usually involves questions that require respondents to admit to attitudes, beliefs, and behaviors that violate social norms, making their assessment susceptible to error due to social desirability bias. This article describes an empirical investigation of an approach to minimize this bias, the use of 'forgiving language' in survey item development and the effect on item variability. Using secondary data initially collected as part of a measurement development study of mental health providers’ stigmatization of service users, 15 pairs of similar, thematically targeted items, varying with respect to wording approach were tested using a purposive sample of mental health providers (N = 220). Findings indicate that items crafted in a forgiving manner were not significantly influenced by social desirability bias, in contrast to items developed in more traditional language. Additionally, forgiving language-items produced higher levels of agreement, on average, when compared to those written in more traditional language. More research is indicated, including systematic variation of wording approaches, but these results seem promising.