Supervisor Accuracy in Subjective Evaluations and Employee Careers

We investigate how the accuracy of supervisors’ performance evaluations affects employee careers. We develop a simple model in which a supervisor receives a noisy signal of an employee’s performance and submits a subjective rating based on that signal. The model predicts that more accurate supervisors generate greater dispersion in their rating scores across subordinates, leading to more promotions. Using personnel records from a large manufacturing firm, we identify supervisors with higher rating dispersion as more accurate raters and estimate the effect of being assigned to them. Consistent with the model’s prediction, we find that employees assigned to accurate raters are promoted at higher rates. We also show that supervisors who drink alcohol and those hired more recently tend to be more accurate raters.