I am writing a paper whose main research design is a Regression Discontinuity (RD), which I ran both parametrically using the reg command and non-parametrically using the renown rdrobust package by Calonico et al.
Now as a robustness check I’m asked to implement a Difference-in-Discontinuity approach as in Grembi et al (2016), which basically consists in taking the difference between two RD one before and one in the year of policy implementation. I found this specification as an example
Y_dt = a(Post_t * RD_d)+ e_dt for d in a bandwidth {-D, D}
where Post_t is a dummy for dates after the policy implementation and RD is the RD estimate.
I found a package that might do the job but it’s user-written and I’m not completely sure of what is the econometrics behind it. Thus, I was thinking about a running a parametric specification using more standard commands. Does anyone have a suggestion on how to proceed?
Now as a robustness check I’m asked to implement a Difference-in-Discontinuity approach as in Grembi et al (2016), which basically consists in taking the difference between two RD one before and one in the year of policy implementation. I found this specification as an example
Y_dt = a(Post_t * RD_d)+ e_dt for d in a bandwidth {-D, D}
where Post_t is a dummy for dates after the policy implementation and RD is the RD estimate.
I found a package that might do the job but it’s user-written and I’m not completely sure of what is the econometrics behind it. Thus, I was thinking about a running a parametric specification using more standard commands. Does anyone have a suggestion on how to proceed?
Comment