Heterogeneity in the Effect of Environmental Protection Expenditure in China: Causal Inference from Machine Learning

The amount of fiscal expenditure on environmental governance in various regions of China in recent years has increased greatly, but its effect on environmental protection is often controversial in existing research. The conflicting finding is largely due to endogenous problems or selection bias. Generally, the cities with high smog pollution have been encountering environmental protection expenditure (EPE) shocks more greatly since 2011. To estimate the causal effects of EPE shocks more accurately, this research uses new machine learning to the estimate heterogeneous treatment effect of the EPE shock on urban air pollution from 216 cities in China, from 2011 to 2018. Empirical results from causal forests show the following. First, the positive EPE shock has contributed to China’s short-term improvements in air pollution. Long-term causality has not been confirmed statistically, though there is a strong correlation between an increase in EPE and a decrease in PM2.5. Second, a positive EPE shock leads to deterioration of smog in low-income and above-medium smog areas in the long term. Third, the positive EPE shock is conducive to the long-term improvement of air quality in moderately industrialized areas, but it does not lead to the long-term improvement of urban air pollution with a high or low degree of industrialization. Similarly, an EPE shock in cities with medium population density improves air quality for a long time, but runs the opposite in cities with high or low population density.