Basics
- Data:
- Outcome Y: …is observed twice, before and after treatment (2 timepoints)
- Treatment D: …is observed once, between the two outcome measurements (1 timepoint)
- We know that it is 0 at the first time point
- Controls X: …observed at t1 or before
- Idea: Take trend of control group as counterfactual for unobserved trend in treatment group
- Treatment Effect: See gray line in graph!
- Assumption: Treatment group’s trend would have been the same has control group’s trend in the absence of treatment (parallel trends assumption)
- …it’s all about defending this assumption, i.e., similarity between treatment and control group