Extending synthetic control method for multiple treated units: an application to environmental intervention

Abstract

Taking the environmental interventions on air quality at G20 Hangzhou Summit as a natural experiment, this paper innovatively establishes an extended synthetic control method with multiple units to evaluate the dynamic treatment effects on air quality improvement at the Summit. The method constructs data-driven weights according to the fluctuation of urban air quality to obtain a more robust and stable estimation with smaller root mean squared prediction error (RMSPE). By minimising RMSPE for pre-intervention model fitting, the study takes nine cities under policy intervention in Zhejiang as treatment cities, and 45 key cities without policy intervention as control cities during 201501–201706 as the final improved experimental scheme. The policy effect of environmental regulations on the average monthly air quality composite index of treated cities in Zhejiang is -0.84 during 201607–201702; while no significant treatment effect is observed since 201702. The results indicate that the environmental policy for the G20 Hangzhou Summit lasted a relatively short period, and it had a significant short-term improvement effect while losing its long-term improving effect on air quality in treated cities. The identification validates the extended synthetic control method with multiple units could also be applied to the policy effect evaluation in other areas.