My objective is to run stacked regression (DiD) with entropy balancing in the setting of multiple timing for treatments.
Initially, my data consisted of panel data for firm-years. I then stacked this data by forming cohorts based on the treatment year, resulting in a data structure where the unit of analysis is cohort-firm-years.
My primary concern is the imbalance in covariates between treated firms and control firms within each cohort.
To address this, I aimed to use entropy balancing.
My approach involved utilizing the "ebalance" command and a loop in Stata to generate weights, with the number of weights corresponding to the number of cohorts. Subsequently, I attempted to run a regression using the reghdfe package with these weights.
However, I am encountering difficulties with the multiple weights. How can I proceed with running a regression that incorporates these multiple weights?
Is there a flaw in my approach? Has anyone successfully applied entropy balancing to form cohorts for stacked regression?
I know there are published papers out there based on the same method here. But they did not disclose their codes..