Hi there, I am currently completing my honours thesis and have been confronted with quite an obstacle due to the nature of my data. I am working with longitudinal LSAC data which has a K-cohort (aged 4-5 in 2004) and a B-cohort (aged 0-1) for the study children in the sample. I have cleaned the data for the respective cohorts, and now am using the command append to import each of the dta files for the cohorts.
The variable “cohort” is coded “B” or “K” for each child, and I have also generated binary Bcohort and Kcohort terms too.

I have then merged the test-score outcomes (Australian NAPLAN scores) on these files using merge m:1 with my “many” variable being the identification number of the child. This all works fine in my code. Whether this is where I have slipped up??

My DID is looking at those B-cohort study children in the state of NSW compared to that of the K-cohort, as in the time period between the cohorts a $77 million NSW policy injection happened only in NSW.

My issue arises when attempting to identify the treatment variable (NSW B cohort) against those who were not in NSW, and also the K cohort who never experienced any form of policy injection.

My question is: is it possible to be able to create these respective control and treatment groups and run a DID between them?
I am under the impression that the “diff” command does not allow me to specify a control group as such. Is there away around this? Is there a different more efficient way of computing a DID with two separate cohorts?


Any suggestions or recommendations would be really appreciated.

Thanks,
A stressed honours student