How to deal with treatment assignment times vary problem in STATA17?

[Phuc Nguyen]

Dear all Statalist experts, today I read the new function of Difference-in-Difference test from STATA17 called xtdidregress

I applied this code to my data and got the results below:

xtdidregress (FAT_w1  $tatcov_) (post33treat1), group(TYPE2) time(year1)

Among them, FAT_w1 is the dependent variable, $tatcov_ is the covariates, post33treat1 is a binary variable( receiving value of 1 for the treatment group after event day and 0 other wise), TYPE2 is the firm type and year1 is the real year in me data sample.

After that, following the guide of the same link above, I conduct a parallel trend test visually by using the syntax

estat trendplots

However, the results turns out an error like that

I am wondering why they said "treatment assignment times vary" here. Because in my datasample, I only have one event date and examining 3 years before and 3 years after this event date.



Could you please kindly to help me in sorting it out?

Many thanks and warmest regards.

[Øyvind Snilsberg]

in your screenshot treatment assignment times vary from 2019 to 2021. first observed treatment time by group,

egen first = min(year1) if post33treat1, by(TYPE2)

[Phuc Nguyen]

in your screenshot treatment assignment times vary from 2019 to 2021. first observed treatment time by group,

egen first = min(year1) if post33treat1, by(TYPE2)

Dear Øyvind Snilsberg , thank you for your help, my concern is how to address the technical issue above. Hope that I did not misunderstand your point.


Much thanks and best regards.

[Marco Antonio]

Hi Phuc Nguyen, did you find an answer to your question? I am struggling with the same issue!

[Nguyen Hoa]

The solution is to make sure that (i) all units in control group start in the sample at the same time and (ii) all units in treatment group start to be treated at the same time.
The easiest way is to create a balanced sample from the current unbalanced sample. But if you want to keep dynamics of firms (firm entry and exit), then you need to create a new (unbalanced) sample such that it satisfies the two conditions above. You can easy do both approaches in a few steps by using the egen() function.

Good luck!

---------
Hannah (Hoa) Nguyen

[Phuc Nguyen]

The solution is to make sure that (i) all units in control group start in the sample at the same time and (ii) all units in treatment group start to be treated at the same time.
The easiest way is to create a balanced sample from the current unbalanced sample. But if you want to keep dynamics of firms (firm entry and exit), then you need to create a new (unbalanced) sample such that it satisfies the two conditions above. You can easy do both approaches in a few steps by using the egen() function.

Good luck!

---------
Hannah (Hoa) Nguyen

Thank you so much, I solved it with the same approach previously, this mainly caused by the missing data that makes the treatment date of unit varying