An Introduction to Causal Inference for Observational Studies

Speaker: Alexander Stubblefield, Ph.D. candidate Department of Economics, CSTAT consultant

Causal inference combines statistical methods with institutional knowledge to estimate the impacts of events, interventions, and decisions. It is particularly valuable when working with observational or pre-existing data, where natural experiments can provide insights in place of randomized controlled trials. By moving beyond correlational studies, causal inference enables an examination of the cause and effect relationship. Causal inference allows us to address questions such as: What is the impact of receiving a college education on future earnings? What is the effect of maternal smoking on low birth weight? What are the effects of Medicaid expansion on mortality?

This seminar will provide a brief introduction to common techniques used to test causal hypotheses with observational data, such as:

• Difference-in-Differences
• Instrumental Variables
• Regression Discontinuity

We will discuss when each method is appropriate, the assumptions they require, and the types of data needed for implementation.