Significant attention has been devoted to the implications of automation for decades, ranging from employment, to productivity, to worker health and safety. This work revisits the question of worker safety in the context of warehousing and logistics, with a specific focus on how the adoption of novel forms of robotics affects the mix and volume of injuries suffered by warehouse workers. The relationship between these inputs and outcomes is unclear. On the one hand, we might expect injuries to fall through the elimination of dangerous or highly repetitive tasks. On the other hand, injuries might rise among residual human workers, whose tasks remain unautomated, if the incorporation of robotics leads to a decline in task variety and a faster pace of work. To explore this tension, we draw on annual warehouse-level records of worker injuries from 141 Amazon Fulfillment Centers, between 2016 and 2020. Results indicate that the relationship between robotic automation and worker health is complex. While robotically automating a warehouse is associated with reduced rates of severe injury (injuries requiring the employee to miss days of work), it is positively associated with the rate of non-severe injury (injuries which do not require missed work). Specifically, estimates indicate robotics are associated with a 40% reduction in severe injuries, but a 77% increase in non-severe injuries. Further, evidence indicates that the rise in non-severe injury is at least partially driven by the increased intensity of work that accompanies robotics, implying that automation of some tasks (via robotics) has spillover consequences for the health of workers performing complementary, non-automated tasks. Using a difference-in-differences-in-differences model, findings further indicate a relative increase in the volume of stress-based injuries, namely sprains and strains (versus cuts, fractures, bruises, etc.) that occurs in Robotic (versus Legacy) Facilities, during periods of particularly high operational load (i.e., Black-Friday, Amazon Prime Day). Implications of our findings for regulators, workers, and adopting firms are discussed within
