A social relations analysis of multi-rater performance appraisal systems

This study investigated performance appraisal in a multi-rater setting. The Social Relations Model was used to separate variance stemming from three sources: target, perceiver, and relationship. A goal of performance appraisal is to increase target variance and reduce other sources of variance that are considered error. A sample of working groups was split randomly into two conditions: treatment and control. Nineteen of the groups were assigned to the treatment condition and 20 were assigned to the control condition. There were 214 participants who interacted in a total of 1040 dyads. All participants rated the other group members on a series of work-related attributes. In the treatment condition participants were given instructions to reduce relationship variance. In the control condition participants were given only generic instructions regarding the task. Results of the study indicated that the manipulation significantly reduced relationship variance and increased target variance, as hypothesized. Implications for rater training in multi-rater performance appraisal are discussed.