Measuring Markups with Production Data

We show standard methods to estimate production functions do not identify markups.This nonidentification creates spurious skewness in estimated markup distributions.We also show that ex-ante structure on the returns to scale solves the identificationproblem. In US public firm data and in a Monte Carlo experiment, we find thatapplying constant returns to scale performs remarkably well and reduces the skewnessin the markup distribution among public-firm by as much as half in comparison tononidentified estimates. This results in half the efficiency losses in output and laborshares when calibrated to a recent macroeconomic model