Monopsony and employer mis-optimization explain why wages bunch at round numbers

We show that administrative hourly wage data exhibits considerable bunching at round numbers. We run two experiments, randomizing wages around 10 cents and $1.00, to experimentally measure left-digit bias for identical tasks on Amazon Mechanical Turk, and fail to find any evidence of discontinuity in the labor supply function at round number, despite estimating a considerable degree of monopsony. We replicate these results in administrative worker-firm hourly wage data from Oregon. We can rule out inattention estimates found in the behavioral product market literature. We provide evidence that firms “misoptimize" wage-setting. More monopsony requires less employer misoptimization to explain bunching