I employ a DiD model and want to compare the effect of independent variable (post##treatment) on a dependent variable (i.e. vote). I want to compare this effect across two sub-groups:
  • Votes where the crime involved was violent (i.e. violentcrime==1)
  • Votes where the crime involved was white-collar (i.e. whitecollar==1)
There are other subgroups, which I am not interested in. I started with a subgroup analysis, where I estimated the regression vote post##treatment if violentcrime==1 and did the same for the variable whitecollar. However, these two sub-groups only form a portion of the full sample. I was wondering whether this approach is viable or creates selection bias and requires a heckman 2 stage model to correct for this.

Thank you for your help,

Karan