Many firms use forced rating systems in which supervisors must evaluate employees according to a predefined distribution. We develop new theory suggesting that forced ratings are less likely to enhance performance when supervisors assess subjective dimensions of employee performance (e.g., creativity), but can have some harmful side effects. In a laboratory experiment, employees work on a creative task, and supervisors rate their performance. We do not find any difference in the employees’ performance or effort in a creative task setting between forced and free ratings. We do, however, find that forced ratings create higher stress for employees (ex post stress scales and biomarkers). Higher stress in turn mitigates the positive effect of effort on creativity. Furthermore, we find that actual creativity explains less of supervisors’ ratings of employees’ performance under forced ratings. Instead, factors that are unrelated to actual creativity, such as eloquent writing and strategic gaming behavior, matter more. Results of an additional online experiment confirm that forced ratings work differently in tasks where performance needs to be evaluated subjectively compared to tasks where objective measures are available.