How Can AI Policy Reduce Pollution Emissions? Insights from a Pilot Project in China

• In Event: Environmental Policies at Different Scales, With Overlaps and Confounds: Analytic Solutions?

Abstract

Purpose - The rapid advancement of artificial intelligence (AI) has emerged as a catalyst for transformative shifts toward green growth, yet empirical evidence on its role in reducing pollution emissions remains limited. This study investigates the impact of the AI policy on air pollution emissions by utilizing a quasi-natural experiment of China's Artificial Intelligence Innovation and Development Pilot Zones (AIPZs), with a view to providing empirical insights into the policy framework for sustainable development.

Design/methodology/approach - To analyze the air pollution reduction effects of the AI policy, this study constructed a multi-temporal difference-in-differences (DID) model with panel data from 279 Chinese cities between 2012 and 2023. Mechanism tests examined whether environmental regulation, energy efficiency, technological innovation and consumer behavior played a mediating role in reducing air pollution emissions. Heterogeneity analyses further explored how regional disparities, digital development levels and urban scale differentials influenced the effectiveness of the AI policy. Additionally, spatial econometric models were employed to suggest AI policy's broader influence on reducing air pollution emissions beyond treated areas.

Findings - As verified by multiple robustness tests, the findings demonstrated that the establishment of AIPZs significantly reduced air pollution emissions. Mechanism tests revealed that the AI policy mitigated air pollution emissions by strengthening environmental supervision, enhancing energy efficiency, accelerating technological innovation, and stimulating the public's green consumption behavior. Heterogeneity analyses emphasized substantial differences in the role of the AI policy in mitigating air pollution, influenced by regional disparities, varying levels of digital development, and urban scale differentials. Further analyses showed that while the AI policy did not eliminate the "local-neighborhood" pollution spillover effect, it mitigated the "boundary effect" by reducing the emission intensity of cross-provincial cities.

Implications - This study provides empirical evidence for the significant impact of the AI policy on air pollution emissions. On the basis of deepening the theoretical understanding of the logical relationship between the AI policy and air pollution emissions, this study provides insights for policymakers to use AI to design targeted pollution control strategies. The findings also emphasized the need for collaborative governance approaches to address the spillover effects of air pollution across administrative boundaries.

Originality/value - This study contributes to the existing literature in several significant ways. Firstly, although many studies have analyzed the relationship between AI and environmental governance using proxy variables, studies evaluating the air pollution reduction effects of the AI policy are lacking. This study innovatively explored the impact of the AI policy on air pollution emissions, filling a research gap in the intersection of AI and pollution. Secondly, it revealed multi-channel mechanisms through which AI policy reduced air pollution, providing a detailed understanding beyond aggregate effects. Moreover, this study further examined the spatial spillover effects of the AI policy, confirming the persistence of the "pollution haven hypothesis" while demonstrating the AI policy's efficacy in mitigating environmental "boundary effects". Although this study is grounded in China's city-level data, the AI policy experience can also be extended to other countries, offering insights into reducing pollution emissions and enhancing environmental governance.

Author

• Suqi Zuo, Southeast University
  Presenting Author