Internal Validity: Identification in 
Economic Experiments

Chapter 3 — Internal Validity: Identification in Economic Experiments At the heart of experimental economics is the potential outcomes framework and the challenge of recovering causal effects from data where only one state of the world is ever observed. This chapter develops the assignment mechanism, introduces the four key exclusion restrictions—SUTVA, observability, complete compliance, and statistical independence—and shows how randomization solves the selection problem. It also covers heterogeneity of treatment effects, the EPATE parameter, and what can still be learned when the exclusion restrictions are relaxed.

1. The potential outcomes approach is a useful framework to understand causal inference.
2. An assignment mechanism known and controlled by the researcher allows recovery of causal effects that are internally valid.
3. Four key exclusion restrictions allow randomized experiments to recover causal effects.
4. Relaxing these assumptions might prevent us from recovering the average treatment effect (ATE), but we can still learn certain features about causal relationships of interest.