problem including never-treated with Sun & Abraham(2021) eventstudyinteract

[Anna DSouza]

I am using the IW estimate and the eventstudyinteract command from Sun and Abraham (2021) with 'treated' (variable turns on when unit is treated) and 'never_treated' units, and with pre and post variables, as seen:

CODE:
[eventstudyinteract DEPVAR month_pre6 month_pre5 month_pre4 month_pre3 month_pre2 month_post0 month_post1 month_post2 month_post3 month_post4 month_post5 month_post6, cohort(treated) control_cohort(never_treated) absorb(i.m_y##i.gov hhid) vce(cluster Code)]

Since the estimator uses never treated units, I assumed that they would be included in the estimation sample, however, only treated units are in the estimation sample.

Is this because there are three steps in the estimation procedure and the last step is creating a weighted sum of the first step estimates - which already incorporated the never treated units? I can't think of another reason that the never treated units woutd not be included in the estimation sample. In their paper and other papers that use the estimator, authors are clear that never treated units can be part of the control group.

Any feedback would be much appreciated.

[Jeff Wooldridge]

Anna: I don't know what the problem is, but I know another way to produce the Sun-Abraham estimator that has easier syntax. You have to define a cohort variable: zero for the never treated and the period of first treatment otherwise. Then

jwdid DEPVAR i.gov, ivar(hhid) tvar(i.m_y) gvar(cohort) never