Issue with csdid in Event Study Using Repeated Cross-Sectional Data

[Simon Holden]

Hi Statalist,

I'm using the csdid command in Stata to conduct an event study. My dataset is repeated cross-sectional with observations from the years 2011, 2013, 2015, and 2018. The first time to treat (time1) ranges from 2009 to 2018. The dependent variable (out) is a binary variable.

When I run csdid, the results seem odd. Despite having 7306 observations in my dataset, only 305 observations are used in the analysis. It appears that csdid is only utilizing observations from 2011 and 2018.

Here are my questions:

Why is csdid only using observations from 2011 and 2018?
Is it possible to include all the sample years (2011, 2013, 2015, and 2018) in the analysis?
Any insights or suggestions would be greatly appreciated!

Thank you!
Simon

use "https://raw.githubusercontent.com/JBZ-researcher/csdid_example/main/example1.dta", clear

csdid out, time(year) gvar(time1) notyet
g sample = 1 if e(sample)

[FernandoRios]

Two problems
1. csdid assumes data is equally spaced (you have 2011,13,15 and should have 2017 not 18)
2. Your Gvar should only contain years that are observed in your tvar. Otherwise odd things happen

[Simon Holden]

Two problems
1. csdid assumes data is equally spaced (you have 2011,13,15 and should have 2017 not 18)
2. Your Gvar should only contain years that are observed in your tvar. Otherwise odd things happen

Many thanks! Fernando.