Using British linked employer-employee data, we show that the establishment size effect for supervisors is approximately twice that for non-supervisors. This difference is routinely statistically significant, not explained by other controls and is an important determinant of the difference in...