DID estimation; variables getting dropped due to collinearity

[Mehreen Mookerjee]

Hello,

I am trying to run a basic DID model where a policy was implemented in 1 state. I have 2 rounds of data, one pre-policy and one post policy. I have data for 3 neighbouring control states.

gen treat=0
replace treat=1 if state==101 //State 101 is treated

rename time post //this is our time dummy

gen treatpost= treat*post //Interaction term

reg Y treatpost treat post (other controls), vce(cluster district)

However, Stata just drops the post and treatpost variable as "omitted due to collinearity". I am not sure why this is happening as I have checked that they are not perfectly collinear. Please advise!

[Jared Greathouse]

Why not do

gen treat=0 replace treat=1 if state==101 //State 101 is treated
  
  rename time post //this is our time dummy
  
  reg Y i.treat##i.post (other controls), vce(cluster district)

Edit: the only other possibility is that your predictors are perfectly collinear with one of those. but since you do not show your data, as the FAQ asks, it is impossible for me to comment.