Quasi-Experimental Designs for Causal Inference.

Area of Science:

• Social Sciences
• Epidemiology
• Econometrics

Background:

• Randomized experiments are the gold standard for causal inference but often infeasible.
• Quasi-experimental designs offer alternatives for evaluating treatment effects in such situations.

Purpose of the Study:

• To introduce and compare the strongest quasi-experimental designs for causal inference.
• To outline the assumptions, estimation methods, and validity threats for each design.

Main Methods:

• Regression discontinuity designs
• Instrumental variable designs
• Matching and propensity score designs
• Comparative interrupted time series designs
• Formalization using Rubin causal model's potential outcomes notation.

Main Results:

• Each design has specific assumptions crucial for valid causal effect identification.
• Methods for estimating causal effects are outlined for each quasi-experimental approach.
• Potential threats to internal and external validity are discussed with mitigation strategies.

Conclusions:

• Quasi-experimental designs provide valuable tools for causal inference when randomization is not possible.
• Understanding the assumptions and limitations of each design is critical for reliable causal effect estimation.
• The Rubin causal model framework aids in formalizing causal estimands and identification strategies.