Measures of desirability beliefs and their validity as indicators for socially desirable responding

Social desirability (SD)-bias is a serious threat for survey-data quality, and the respondents’ desirability beliefs have proven in many studies to predict the incentives for the strength as well as for the direction of this bias. However, the issue of the relative validity of different ways to measure these incentives has hardly received any attention. We introduced three such measures and discussed the respective tradeoffs between their parsimony on the one hand and the implied assumptions which have to be fulfilled on the other. In the empirical part of our paper, we tested with four questionnaire topics whether and how strong these assumptions are violated and thus how much measurement effort is necessary to obtain a sufficiently good indicator for the perceived incentives for SD-bias. These topics were: (a) having prejudice against the elderly, (b) making charitable donations, (c) participating in political elections and (d) being environmentally conscious. We found the most parsimonious one-point measure to be strongly affected by measurement error in the case of all questionnaire topics. Depending on how strong the assumption of monotony was violated, for some topics the medium elaborated simple difference scores were sufficiently valid, but partly only the most elaborated domain-specific difference scores were a valid predictor for the potential strength and direction of incentives for SD-bias