Getting the ATT in the form of percent increase in csdid_plot

[Luiza Lins]

Hello,

I am estimating the impact of the opening of a specific public facility on the number of people registered as homeless in cities. This public facility opens at different times for different municipalities (and never opens in some municipalities). Hence, I'm using Callaway and Sant'Anna's csdid function.

I estimate it as follows:

csdid count, ivar(city) time(year) gvar(open_year) notyet
estat event
csdid_plot

The issue is that the ATT in absolute terms doesn’t tell me much on its own. I would like to know the relative change instead. Since count = 0 for many cities, I cannot simply use the log transformation. Is there any way to obtain the ATT (preferably in the plot) in terms of percent increase or decrease?

Sorry if I made any mistakes.

[FernandoRios]

Not possible
Your alternative would be using jwdid and use predict(xb) when calling for results

[Luiza Lins]

Hi Fernando, thank you for your response. I think I’ll just include the baseline mean in the plot.

If I may ask another question: I’m concerned about time-varying confounders, such as municipal GDP, population growth etc. Would it make sense to include municipality-year fixed effects to account for these? Or does the command already handle these differences inherently?

Best regards