Many thanks for creating this code in Stata!
I have a query regarding the balance of predictors when employing SDID with projected covariates. In the traditional Synthetic Control Method (SCM), it is a common practice to calculate and report the balance of all covariates used in the estimation process. This is to ensure that the algorithm achieves an approximate balance across these covariates. For example, the Stata command “synth” provides a report on the predictor balance alongside the weights and estimates.
Considering the inherent differences between SCM and SDID — particularly the focus of SDID on trend balance rather than exact covariate balance — I am interested in understanding the relevance of performing a similar balance check in the context of SDID. In your view, how relevant is it to assess predictor balance in SDID applications? Is it more relevant to check whether they are approximately parallel to the treated ones after the reweighting?
Is there an existing code or procedure that can be implemented to directly compute the predictor balance after employing “sdid”?
Dear authors,
Many thanks for creating this code in Stata! I have a query regarding the balance of predictors when employing SDID with projected covariates. In the traditional Synthetic Control Method (SCM), it is a common practice to calculate and report the balance of all covariates used in the estimation process. This is to ensure that the algorithm achieves an approximate balance across these covariates. For example, the Stata command “synth” provides a report on the predictor balance alongside the weights and estimates. Considering the inherent differences between SCM and SDID — particularly the focus of SDID on trend balance rather than exact covariate balance — I am interested in understanding the relevance of performing a similar balance check in the context of SDID. In your view, how relevant is it to assess predictor balance in SDID applications? Is it more relevant to check whether they are approximately parallel to the treated ones after the reweighting? Is there an existing code or procedure that can be implemented to directly compute the predictor balance after employing “sdid”?
Many thanks, Augusto Cerqua