The effect of Dichotomous versus Likert response format on the measurement of Socially Desirable Response bias
Socially Desirable Responding (SDR) is defined as a biased tendency to endorse personality inventory items that are generally judged to be socially acceptable, and to avoid endorsing items that are judged to be socially undesirable. Rare Virtue (RV) scales, such as the MMPI-2 Lie scale, have been designed to detect SDR. The items of such scales, called RV items, typically invite respondents to claim unlikely virtues or deny common faults. The present study hypothesized that RV items perform much differently when a Dichotomous response format (e.g., true/false) is used than when a Likert response format is used. A total of 457 undergraduates in six experimental conditions made ratings on a questionnaire that consisted of 55 RV items and 59 randomly selected items from the Schedule for Nonadaptive and Adaptive Personality (SNAP). A variable called the Over/Underestimation Value (OUV) measured how much the endorsement rate for each SNAP and RV item differed from what would be predicted based on the item's social desirability. As predicted, OUV was significantly higher for RV items than for SNAP items, and this difference was significantly higher when a Dichotomous format was used than when a Likert format was used. Four RV scales, including the Lie scale and Wiggins Sd scale of the MMPI-2, were administered to respondents who were instructed to fake good on the questionnaire. Each of the four RV scales significantly discriminated fakers from a control group of non-fakers. However, contrary to prediction, the scales' success in discriminating fakers from non-fakers was not significantly greater when a Dichotomous response format was used than when a Likert format was used. Overall, the findings for the experimental hypotheses were mixed. However, follow-up analyses provided tentative new insights regarding the relationship of items' social desirability and endorsement rates with the items' sensitivity to faking good.
Medrano, Celia, "The effect of Dichotomous versus Likert response format on the measurement of Socially Desirable Response bias" (2011). ETD Collection for University of Texas, El Paso. AAI1494364.