Hello everyone,
I am working with panel data to analyze the impact of a policy intervention implemented simultaneously across all treated municipalities within a specific region. My control group consists of municipalities in the rest of the country that were never treated.
I first estimated a basic xtreg model with a treatment dummy and found a significant treatment effect. However, I would like to explore the dynamic effects of the intervention. I'm considering using the Callaway & Sant’Anna (2021) Difference-in-Differences (DiD) estimator. Although it is specifically designed for staggered adoption designs, I understand that it can also be applied in cases with a single treatment period, using never-treated units as the control group.
When I run csdid, the estimated ATT is similar to the effect found in the xtreg model. It also provides a dynamic treatment effects graph that makes intuitive sense, and the parallel trends assumption appears to hold.
My question is: is it methodologically valid to use the Callaway & Sant’Anna (2021) DiD estimator in this simpler case, as long as its assumptions are satisfied?
Thanks in advance!
I am working with panel data to analyze the impact of a policy intervention implemented simultaneously across all treated municipalities within a specific region. My control group consists of municipalities in the rest of the country that were never treated.
I first estimated a basic xtreg model with a treatment dummy and found a significant treatment effect. However, I would like to explore the dynamic effects of the intervention. I'm considering using the Callaway & Sant’Anna (2021) Difference-in-Differences (DiD) estimator. Although it is specifically designed for staggered adoption designs, I understand that it can also be applied in cases with a single treatment period, using never-treated units as the control group.
When I run csdid, the estimated ATT is similar to the effect found in the xtreg model. It also provides a dynamic treatment effects graph that makes intuitive sense, and the parallel trends assumption appears to hold.
My question is: is it methodologically valid to use the Callaway & Sant’Anna (2021) DiD estimator in this simpler case, as long as its assumptions are satisfied?
Thanks in advance!
Comment