RCT

Propensity score matching

IV

RDD

Diff-in-Diff

For each method:

1. Describe the estimation strategy. Who is the treatment and control group?
2. What data would you need on the treatment and control groups?
3. What is the assumption required to interpret the estimate as causal?
4. How safe is this assumption? (almost always true, can be convinced depending on context, usually pretty hard to believe)
5. How can you test the assumption?