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 that cannot be explained by rounding of survey respondents. We consider two explanations--worker left-digit bias and employer optimization frictions. We experimentally rule out left-bunching by randomizing wages for an identical task on Amazon Mechanical Turk, and fail to find evidence of any discontinuity in the labor supply function as predicted by workers' left-digit bias despite a considerable degree of monopsony. We replicate the absence of round number discontinuities in firm labor supply in matched worker-firm hourly wage data from Oregon as well as in an online stated preference experiment conducted with Wal-Mart workers. Further, the shape of the missing mass that accounts for the bunching at a round number exhibits none of the asymmetry predicted by worker left-digit bias. Symmetry of the missing mass distribution around the round number suggests that employer optimization frictions are more important. We show that a more monopsonistic market requires less employer mis-optimization to rationalize the bunching in the data. The extent of monopsony power implied by our estimated labor supply elasticities, which are in line with other recent studies, are consistent with a sizable amount of non-optimal bunching, with only modest losses in profits. Overall, the extent and form of round-number bunching suggests that "behavioral firms" can systematically misprice labor without being driven out of the market in the presence of monopsony power