Hello Statlist members, this is both an econometric question and a Stata implementation question. I am studying the effect of Medicaid expansion on an outcome. I hypothesize that the effect of Medicaid expansion is greater for disabled individuals compared to all other individuals. I started with the following TWFE model:
regress outcome medexp##disability disability i.year i.state control_variables, vce(cluster state)
medexp is a binary variable that takes a value of zero for state s in year t before treatment takes place and a value of one in the years after treatment for states that were treated and zero for states that are never treated. disability is a binary variable that takes a value of 1 if a person is disabled. States adopted Medicaid in different years so treatment is staggered.
I would like to use other estimation strategies that have recently been proposed in the econometrics literature to address limitations of TWFE regressions in the presence of heterogeneous treatment effects and staggered timing such as the Callaway and Sant’Anna (2021) estimator or Borusayk et al. (2021) but I run into a problem implementing the statistical packages because my treatment variable is interacted with another binary variable. For example, I cannot use the hdidregress or csdid because of this. Any suggestions how I could proceed?
regress outcome medexp##disability disability i.year i.state control_variables, vce(cluster state)
medexp is a binary variable that takes a value of zero for state s in year t before treatment takes place and a value of one in the years after treatment for states that were treated and zero for states that are never treated. disability is a binary variable that takes a value of 1 if a person is disabled. States adopted Medicaid in different years so treatment is staggered.
I would like to use other estimation strategies that have recently been proposed in the econometrics literature to address limitations of TWFE regressions in the presence of heterogeneous treatment effects and staggered timing such as the Callaway and Sant’Anna (2021) estimator or Borusayk et al. (2021) but I run into a problem implementing the statistical packages because my treatment variable is interacted with another binary variable. For example, I cannot use the hdidregress or csdid because of this. Any suggestions how I could proceed?
