Issues with csdid

[Kapoor Manisha]

I am looking at the effect of a policy on female employment. The policy was implemented in different states at different points in time. The data I have is at the state level for time-period 2000-2020. There are some states that were never treated.
I am trying to use the csdid command but there is an issue with interpretation of the results. The policy is such that it can have an impact overtime. So, if a state received treatment in 2005, I am interested in effects one year after policy, two years after policy etc.

The two questions I have are:
First, for relative time 1: what is the treatment and control group? How is the aggregation of effects happening? As per my understanding, the treatment group should average “female employment” in one year after the policy implementation in every state that was ever treated. But what is the control group averaging?

Second, I have detailed data at state*time*industry level. Currently I am just averaging that data to state*time to run csdid. Is there any way to use industry level data and have industry fixed effects?

[FernandoRios]

1. Control groups depend on the treatment
What is used as control for cohort 10 will be different from cohort 12 or 20
things are more complex when using not yet treated
2. I’m surprised that what you did worked
you have to use time data as is
for the panel data you could interact state and industry

if you want something similar to adding fixed effects, something onhave explored is to start by demeaning your data by state&industry
this will work only if treatment is not assigned bu state nor industry
f

[Kapoor Manisha]

Thank you for getting back to me. I think I was unclear when I talked about averaging. I have employment for all states and industries overtime. Ideally, I would like some industry effects. But I didn't know how that would work with csdid, so I have summed up employment for every industry to get employment in each state overtime. The treatment has happened at the level of states and not industry. Some states were treated in 2010, some others in 2012 and some still haven't been treated.