Dynamic synthetic control method for evaluating treatment effects in auto-regressive processes

Abstract

Abstract Motivated by evaluating the effects of air pollution alerts on air quality, we propose the dynamic synthetic control method for micro-level data with time-varying confounders and spatial dependence under an auto-regressive model setting. We employ the empirical likelihood to define the synthetic control weights, which ensures a unique solution and permits theoretical analysis. The dynamic matching increases the feasibility of matching and enables us to assess the unconfoundedness assumption using pre-treatment data. For statistical inference, we develop a normalised placebo test to address the asymmetry issue. The method is illustrated and evaluated on numerical simulations and a case study on air pollution alerts.