Learning to Successfully Hire in Online Labor Markets

Hiring in online labor markets involves considerable uncertainty: which hiring choices are more likely to yield successful outcomes and how do employers adjust their hiring behaviors to make such choices? We argue that employers will initially explore the value of available information. When employers observe successful outcomes, they will keep reinforcing their hiring strategies; but when the outcomes are unsuccessful, employers will adjust their hiring behaviors. To investigate these dynamics, we propose a two-component framework that links hiring choices with task outcomes. The framework’s first component, a Hidden Markov Model, captures how employers transition from unsuccessful to successful hiring decisions. The framework’s second component, a conditional logit model, estimates employer hiring choices. Analysis of 238,364 hiring decisions from a large online labor market shows that, often, employers initially explore cheaper contractors with a lower reputation. When these options result in unsuccessful outcomes, employers learn and adjust their hiring behaviors to rely more on reputable contractors who are not as cheap. Such hiring tends to be successful, guiding employers to reinforce their hiring processes. As a result, the market observes employers transition from cheaper, lower-reputation options with poorer performance to more expensive reputable options with better performance. We attribute part of this behavior to employer confidence and risk attitude, which can change over time. This work is the first to investigate how employers learn to make successful hiring choices in online labor markets. As a result, it provides platform managers with new knowledge and analytics tools to target employer interventions. This paper was accepted by Anandhi Bharadwaj, information systems.