Death by Robots? Automation and Working-Age Mortality in the United States
Key Takeaways
Increased automation between 1993 and 2007 lead to increases in death rates from all causes among 45-54 year old American adults.
Across all age groups, automation led to increases in death rates from drug overdoses, suicide, homicide, and cardiovascular disease, indicating broad social and health consequences of economic restructuring.
Automation need not lead to premature death. The impacts of industrial robots varied depending on factors such as the generosity of social safety nets and labor market policies.
Executive Summary
Introduction
The study investigates whether the shift toward automation in U.S. manufacturing — specifically the adoption of industrial robots — contributed to rising working-age mortality. Automation has been linked to reductions in economic opportunities and wages in manufacturing. Such declines may underlie increases in the so-called “deaths of despair” – like suicide and drug overdoses — and other causes of death. This paper uses novel data on variation in robot adoption over time and across labor markets to estimate the causal effect of automation on mortality among American adults.
Main Findings
- Localities experiencing greater increases in automation saw larger increases in mortality from all causes among adults aged 45–54. This suggests that displacement from manufacturing jobs — via automation — had measurable, deleterious effects on population health.
- When breaking down deaths by cause, we found that overdose, suicide, homicide, and cardiovascular deaths all increased and that patterns varied by age group. The cross-cutting effects of automation on diseases with different risk factors suggests that automation works through multiple mechanisms to affect health and well-being. These could include reducing financial resources and access to health care as well as raising stress.
- The effects of automation varied from place to place. In American labor markets located in states with stronger social safety nets and labor protections, we find a much more muted impact of automation on mortality. In contrast, in labor markets located within states with less generous safety nets and labor protections, the link between automation and death is even stronger. Together, these findings imply that policy context can buffer or exacerbate the health consequences of economic disruption.
Conclusion
Our findings illustrate how changes in the economy can negatively affect health and well-being. But it does not have to be this way. Stronger social safety nets and policies to support workers displaced by automation may mitigate negative effects on health. Another option is to rethink how we introduce new innovations in our economy. For example, some countries have managed to automate different sectors in ways that make workers more productive, rather than displace them.