Skip to content

Latest commit

 

History

History
18 lines (15 loc) · 1.6 KB

README.md

File metadata and controls

18 lines (15 loc) · 1.6 KB
Raw

Compare-DiD-Estimators

Case: staggered treatments, homogenous (constant) TEs. (100 units, 15 time periods)

  • "Cohort 1": 35 units are treated once at t=7, constant ATT(g=1,K)=49 for K=0-8 periods after the treatment (t=7-15)
  • "Cohort 2": 35 units are treated once at t=11, constant ATT(g=2,K)=49 for K=0-4 periods after the treatment (t=11-15)
  • 30 units are never treated image image

Case: staggered treatments, heterogenous (constant) TEs. (100 units, 15 time periods)

  • "Cohort 1": 35 units are treated once at t=7, constant ATT(g=1,K)=20 for K=0-8 periods after the treatment (t=7-15)
  • "Cohort 2": 35 units are treated once at t=11, constant ATT(g=2,K)=40 for K=0-4 periods after the treatment (t=11-15)
  • 30 units are never treated image image
  • TE estimates for K>4 periods post-treatment are non-sense b/c for the 2nd cohort, we only have data up to 4-periods post treatment, so estimates of ATT_K for K>4 only average out the TE for the first cohort (which has a smaller ATT=20)
  • TWFE-OLS & CDLZ are biased.