Title: Human premium in work allocation between humans and machines

Abstract: We study whether there exists a human premium in the work allocation between humans and machines from a firm perspective. We elicit preferences for work allocation between humans and machines in an online experiment on Prolific using a convex budget sets design. Our results indicate that there is a high consistency in the preferences for human-machine work allocation. These preferences reveal a human premium: a tendency to allocate more work to the human worker than the machine at identical relative prices. These findings suggest that preferences for work allocation to humans can be an important constraint on the proliferation of automation technologies in the workplace.