Difference-in-Difference with multiple periods and different treatment periods

[Kagiso Matswalela]

My name is Kagiso, currently a PhD student working with the Difference-In-Difference (DID) method to determine the impact of a programme. I was trying to follow the paper “Difference-in-Differences with multiple periods” by Callaway & Sant'Anna 2021. I could not understand how they set up their data to execute DID with multiple treatment periods. I'm using STATA 17 package.

My request is as follows. If there is anyone with knowledge of how to set up data to execute DID on STATA.

Here is the structure of my data

• I have data from 2009-2020, including firms in various manufacturing subsectors
• It includes various 8 Provinces, 4 subsectors
• Firms are treated in different periods and some are treated in all the periods.
• The programme was introduced in 2011, and few firms were treated and the majority of the firms were treated in the later years, like from 2015-2018
• Very few firms remained untreated throughout the lifespan of the programme. This would be the never treated group
• It would be fair to say that my data contains more of the “not yet treated” group than “never treated”.
• Also, this is a panel identified by firm_id

in summary, my variables include dvar(x1) ivar(x5), and treatment/programme. the year 2009-2020 (some firms were first treated in 2011, meaning we have some pre-treated data for 2009/10). group variable firm_id to identify about 330 firms.

I would like to set up my data to read the “not yet treated” as a control group. And also, still read the “never treated” as a control group in another equation.
With the STATA application in mind, how do I begin setting up my data to use in STATA to determine the treatment effect bearing in mind that treatment occurred in different multiple times?

I hope I'm making sense to anyone who would like to help me. I'm willing to engage further if my explanation of what I need help with is insufficient or not clear.

Thanking you in advance
kagiso

[Andrew Musau]

ssc install csdid, replace

Then look at the data and examples in

help csdid

[FernandoRios]

Actually csdid and drdid
both from ssc

[Kagiso Matswalela]

Hi Andrew & Fernando
thank you for taking the time to read my request. Also, thank you for your feedback. I will explore the options and revert.