Staggered Difference-in-Differences using csdid

[Darius Markovskij]

Dear all,
I'm trying to run the csdid command in Stata to estimate a difference-in-differences model with staggered treatment adoption. However, when I run the command: csdid cons_per_capita, ivar(country_id) time(year) gvar(group_year)
I get empty output tables — no estimates, coefficients, or standard errors appear.

Here’s a bit more about my setup:

• cons_per_capita: my outcome variable (float).
• country_id: unique identifier for each country (long).
• year: time variable (2000, 2001, ... , 2024).
• group_year: the year each country adopts the treatment (varies across countries). For countries that were never treated, group_year is 0. For countries that have not yet been treated, group_year is also 0, and once the treatment occurs, group_year takes the value of the treatment year.

I've double-checked the variables, and the dataset looks properly formatted. Each country has multiple years of data, and treatment starts in different years for different countries.

Things I've tried:

• Ensured no missing values in key variables.
• Verified group_year is coded correctly (some units stay untreated throughout).

Has anyone run into this issue before? Any ideas what might cause empty tables in csdid?

Thanks in advance!

[FernandoRios]

Show the cross tab between gvar and year

[Darius Markovskij]

Here it is!

[FernandoRios]

That’s the rerror
gvar has to be there through all periods

[FernandoRios]

That’s the rerror
gvar has to be there through all periods
also with so few treated variables do not add any controls

[Darius Markovskij]

Thanks, that worked. So, due to low number of treated observations I won't be able to add not-yet treated units as controls?

[FernandoRios]

You can
that isn’t a problem
madding covariants is
yoi are basically making regressions with n=1 in those cases