Self-assessment in a selection situation: An evaluation of different measurement approaches

The effect of measurement conditions on the psychometric properties of self-assessment was investigated in a selection context. Two experimental studies are reported which investigated the effect of format and instructions on leniency and halo. In Study I it was hypothesized that questionnaires based on overall descriptions of competencies and asking for examples of past behaviour would display less leniency and halo than when raters did not provide such evidence, or when the competency was broken down into its behavioural elements. The effect of social comparison instructions (SCI) was also explored. Participants were 738 candidates who applied to join a public sector organization. Contrary to predictions, it was found that questionnaires based on 'behavioural elements' displayed less leniency. Furthermore, an interaction between format and instructions was found; questionnaires based on behavioural elements/no SCI showed less leniency than 'competencies/no SCI' and 'competencies/evidence/SCI'. No significant findings were obtained for halo. Study 2 used a subset of 224 candidates from Study I who attended the next stage of the selection process. It was hypothesized that unbalanced, positively toned scales would show less leniency and halo than conventional scales, as would instructions requesting a realistic self-assessment of the individuals' strengths and weaknesses. The first hypothesis was supported. A trend towards lower leniency was observed for 'honesty' instructions, which failed to reach significance. Gender differences were observed such that mates were significantly less lenient when dimensional questionnaires were employed (Study 1), and unbalanced questionnaires and instructions pertaining to realism (Study 2). Females' self-assessment displayed less leniency and halo when unbalanced questionnaires were employed. Findings are discussed in terms of implications for practice.