Causal Inference Methods for Estimating Long-Term Health Effects of Air Quality Regulations.

Area of Science:

• Environmental Epidemiology
• Biostatistics
• Public Health Policy

Background:

• Air quality management requires new epidemiological evidence to assess regulatory impacts.
• Traditional methods focus on exposure-response, but direct evaluation of regulatory effectiveness is needed.
• This report introduces methods for direct-accountability assessment of air quality regulations.

Purpose of the Study:

• To provide new analytic perspectives and statistical methods for direct-accountability assessment of air quality regulatory interventions.
• To enhance policy debates with direct evidence on the causal consequences of regulatory actions.
• To anchor accountability assessment to the estimation of causal effects of well-defined interventions.

Main Methods:

• Employed established and newly developed statistical methods for causal inference from observational data.
• Utilized a potential-outcomes paradigm to frame observational studies as approximate randomized experiments.
• Applied methods including propensity scores, principal stratification, and causal mediation analysis to two case studies using national linked data.

Main Results:

• Designating areas as nonattainment for PM10 causally reduced all-cause mortality and respiratory hospitalizations.
• Sulfur dioxide scrubbers on power plants causally reduced ambient PM2.5, primarily through SO2 emission reductions.
• The study utilized a comprehensive national database linking air quality, health, and regulatory data.

Conclusions:

• Grounded accountability research in a potential-outcomes framework, providing robust evidence of health effects from air quality regulations.
• Demonstrated the causal impact of specific regulatory interventions on pollution and health outcomes.
• Contributed rigorous evidence to support the U.S. Environmental Protection Agency and stakeholders in policy development.