Regression Discontinuity Design

Social policy and clinical care are filled with thresholds—people receive or don’t receive interventions or treatments based on factors such as income, test results, year of birth, and others. These thresholds are a potential source of quasi-randomization for strong observational studies. Regression discontinuity designs are a quasi-experimental research design that utilize these thresholds to derive causal estimates about the impact of interventions and policies.

In this talk, Dr. Law will provide an introduction to regression discontinuity designs and share an example from his research based on thresholds in BC’s Fair Pharmacare drug coverage program. He will discuss the counterfactual assumption in these designs, their data requirements, how to model them, how to interpret the results, and what threats to validity analysts should be concerned about.