Unintended carbon cost of automation technology in transforming economies: The role of capital dependence and structure change

Automation technologies enhance production efficiency, offering potential for sustainable development. Yet, by deepening economic dependence on inherently emission-intensive capital investment, automation may inadvertently increase carbon emissions in economies with transforming industrial and production structure. We investigate this hypothesis first through a conceptual framework incorporating heterogeneous production and industrial components. We then test this using panel data from China's second-level administrative (prefecture) units from 2007 to 2021, estimating both short-run and medium-run effects using fixed-effects and long-difference estimators. Results suggest that automation technology, measured by per capita patents, is associated with increased carbon intensity. This positive effect persists over the short- to medium-term, indicating limited mitigation through structural adaptation. This persistence highlights the inertia of emission structures even amid technological progress. The impact is larger in western and industrial regions, reflecting regional disparities in industrial composition and adaptive capacity. We further examine four mediating channels that reflect capital dependence—relative marginal productivity of capital, capital misallocation, per capita capital stock, and temporary migration—and further conduct sectoral regressions to capture heterogeneity across industrial structures. Together, these analyses reveal that capital dependence and structural change to be important channels through which the automation‑carbon intensity relationship manifests, which reveals complex environmental trade-offs from automation. Overall, the findings provide new evidence that automation may generate unintended environmental costs, calling for policy coordination between industrial upgrading and decarbonization strategies.