Rating Errors and Perceptual Biases

Distributional Errors

It is not uncommon to find that 80% to 90% of all employees rated by a single rater receive an aboveaverage rating. This often indicates a distributional error, wherein the rater misrepresents the distribution of performance across persons they are evaluating. In other words, these errors occur when the distribution of assigned ratings differs from the (assumed) distribution of actual job performance of the group of employees being rated. Such misrepresentations can occur both in terms of the mean level and the variability of ratings provided. The three most common types of distributional errors are leniency/severity, range restriction, and central tendency errors.

Leniency/severity errors occur when the mean of the ratings of all employees rated by a particular supervisor differs substantially from the midpoint of the rating scale. For example, if the mean ratings for all employees rated by a supervisor are very low, then the rater is thought to be overly severe; when the mean ratings are very high, he or she is thought to be overly lenient. This error can be caused by (a) raters having inaccurate or unreasonable frames of reference or expectations for performance; (b) the rater's desire to be liked, hence his or her unwillingness to give negative feedback; or even (c) expectations that other raters will also inflate their ratings. Recent research has shown that rating severity/leniency is a relatively stable characteristic of the rater and can be related to his or her personality. Specifically, individuals who score higher on agreeableness tend to provide more elevated ratings, whereas individuals who score higher on conscientiousness tend to provide lower ratings.

It is also possible for a rater to fail to make adequate distinctions among multiple ratees when rating their performance, an error referred to as range restriction. For example, consider a group of employees who vary widely in their levels of performance on one dimension, quality of work. If all of the supervisor's ratings on this dimension are clustered within a small range of scores, the variance of the supervisor's ratings will be lower than the variance of the actual performance levels of the ratees, and hence, range [Page 661]restriction is said to occur. Raters who commit this error fail to distinguish among ratees on individual performance dimensions, either because of a lack of opportunity to observe the employees or a conscious desire to avoid differentiating among ratees.

...