Abstract
Analyses of the problem of social desirability (SD) in tests for personnel selection deal primarily with simple item formats with which rating scales for self-description are implemented. However, the range of possible item formats is much greater than this. In the current study we examine five different item formats in terms of their susceptibility to the tendency towards socially desired responses: problem-solving items, knowledge items, forced-choice items, self-description with rating scales, and situational judgement items. Up to now, an SD effect has been demonstrated in particular for rating scales. All items are components of a computer-based personnel selection test battery. To examine the different item formats two methods are used. First, real applicants were compared with subjects instructed to present themselves in a positive way (fake good) and with subjects instructed to give true answers. Second, the results of the five scales are correlated with an SD scale. Both methods show that the participants can distort only the results of the rating scales to their own advantage. This problem can be combated by excluding applicants with very high SD values.
Notes
1The predictive validity is a special form of criterion-related validity. The criterion (e.g., job success) in this process is measured weeks, months, or years later than the predictors. Only in this way is a prediction of the criterion possible. The simple criterion-related validity only allows statements to be made about correlative relationships, as predictor and criterion are measured at the same time.