10.3 I Just Want an ATE

Which treatment variation do we allow to count

Rules of Thumb (do not apply 100 percent of the time)

1. True randomization \(\rightarrow\) ATE
2. Randomization in a group \(\rightarrow\) CATE
3. Closed back doors \(\rightarrow\) weighted average treatment effect
4. Assuming treatment on untreated \(\rightarrow\) ATT
5. Isolate exogenous variable \(\rightarrow\) LATE

(Słoczyński, 2022)

• Comparing treated to counterfactual outcomes

  • More treated people, more represents ATUT