The traditional literature on method of pay distinguishes workers who are paid time rates from those who are paid piece rates. The theory predicts that the piece-rate workers will earn more, and empirically they do. A simple generalization is to divide time-rate workers into two groups: those whose wage depends on their supervisor's ratings and those whose wage does not. Theory predicts that the workers whose pay is linked to supervisor ratings will earn more than the other time-rate workers. Wage data for workers in over 3,000 manufacturing establishments show they do not, and several simple explanations fail to resolve this empirical puzzle.