Dear All,
My question concerns whether it is possible to combine a Cox proportional hazards model with a difference-in-differences approach and if so, how that would be implemented in Stata. I am very new to Cox proportional hazard models so please excuse my lack of knowledge. I truly appreciate any help.
My case is the following: I have a panel of firms covering a large range of years. In some year several industries were affected by an increase in tariffs (treatment group) whereas others were not (control group). Crucially, I have data for years both before and after this tariff change occurred. What I am interested in is how this policy change affected the probability of firm survival. Naturally, I would fit a DID model since I have before and after data as well as a treatment and control group. However, I am interested in whether the increase in tariffs decreased the likelihood of failure. As well, my data is right-censored. Thus, from reading the literature I got the sense that I should combine the DID approach with a Cox model. However, I am unsure how to do this.
Specifically, I am wondering
1) How should I structure my data? At this moment I have one observation for each firm and year it was in existence. However, for Cox models the data usually seems to be organized in a way so that each firm has only one line.
2) Is there a straightforward way to implement a DID analysis using the Cox model? How would that be coded?
Again, thank you very much for considering my case and I appreciate any help!
All the best
Leon
My question concerns whether it is possible to combine a Cox proportional hazards model with a difference-in-differences approach and if so, how that would be implemented in Stata. I am very new to Cox proportional hazard models so please excuse my lack of knowledge. I truly appreciate any help.
My case is the following: I have a panel of firms covering a large range of years. In some year several industries were affected by an increase in tariffs (treatment group) whereas others were not (control group). Crucially, I have data for years both before and after this tariff change occurred. What I am interested in is how this policy change affected the probability of firm survival. Naturally, I would fit a DID model since I have before and after data as well as a treatment and control group. However, I am interested in whether the increase in tariffs decreased the likelihood of failure. As well, my data is right-censored. Thus, from reading the literature I got the sense that I should combine the DID approach with a Cox model. However, I am unsure how to do this.
Specifically, I am wondering
1) How should I structure my data? At this moment I have one observation for each firm and year it was in existence. However, for Cox models the data usually seems to be organized in a way so that each firm has only one line.
2) Is there a straightforward way to implement a DID analysis using the Cox model? How would that be coded?
Again, thank you very much for considering my case and I appreciate any help!
All the best
Leon
