Causal Inference Methods for Assessing the Public Health Benefits of Air Pollution Regulations in the United States

The incomplete nature of epidemiological evidence on the public health benefits of large scale regulatory actions is due in part to the lack of sound statistical methods that can isolate the causal pathways leading from a regulation to both air quality and health. First, we are linking national data sets that will provide information on regulatory actions, ambient levels of criteria pollutants, health outcomes, and confounders for the entire U.S. Second, we are planning to use a potential outcomes framework to define causal effects of interest for single and multiple pollutant accountability assessment and develop methods for estimation. We will develop a causal inference method for accountability research that uses principal stratification to isolate the causal pathways leading from regulation to changes in air quality and health. The proposed method will allow us to quantify and disentangle causal effects of the regulation on health that are 1) associated with causal effects of the regulation on air quality and 2) associated with causal pathways capturing other factors that do not involve changes in air quality. We will apply our proposed methods to our national data sets to estimate the impact on health indicators of separate regulations that target different pollutants.