July 2018
Examining the market power of on-demand labor platforms in the gig economy
The authors collected data from Amazon Mechanical Turk (MTurk), a popular online platform that allows employers to post short-term tasks (some that require less than a minute to complete) that potential workers can accept for pay (as low as $0.01). Their research offers empirical estimates of recruitment elasticity and retention elasticity and a double-machine-learning estimator is used to examine the variations in these rewards. Recruitment elasticity refers to the propensity for the reward (compensation) and associated task to attract workers. Retention elasticity refers to how well bonus payments attract workers who have already accepted a job from a given employer to accept additional assignments from the same employer.
While the results displayed some heterogeneity, the implied elasticities were consistently small. The research reported that observational evidence shows low recruitment and retention elasticities for employers. These findings have important implications for online labor markets. Over time, the importance of these implications may increase as online platforms become the ideal environment to reduce costs. The study finds that the devaluation of productivity in wages is substantial if employers are exploiting the capacity to set wages. Ultimately, the study concludes that MTurk workers might be paid less than 20 percent of the value of their productivity. This “highly robust and surprisingly high degree of market power” for MTurk employers suggests an inefficient market structure that benefits employers to the detriment of workers.
“Why does this inefficient market structure persist?” The authors suggest that it is caused, in part, by limited information given to workers, the difficulty of finding alternative employment, and Amazon’s role in crafting the platform to benefit employers over workers. Using online labor markets, employers capture the surplus created by lower wages. The authors note that the platform itself exhibits market power for crowdsourcing, which allows inefficient market structures to continue. However, this lingering question is one for future study, as the authors deemed it out-of-scope for this study.