Non binary treatment for DiD analysis using did_multiplegt_dyn

[Thibaud Marcesse]

Hello,

I have panel data for the implementation of a policy in India (2012-2021) and wanted to test whether the implementation of quotas of descriptive representation at the village level impacted policy implementation. The treatment here is the quota which reflects reservation assignment, but it is not binary (since villages can be reserved for different social and ethnic groups, in different villages) So far, I have been using different packages (xthdidregress, csdid and did_multiplegt_dyn) and comparing individual treatments (corresponding to different quotas) with the group of villages that were not treated. Given that I have 4 different treatments (that can be assigned at the same time, BUT across different villages - in other words, a village is assigned one treatment and one only), the individual analysis has been clunky (and time consuming) since every time I compare one treatment assignment to the control group (and have to repeat this 4 times). Is there a package out there that would allow me to compare these different treatment assignments simultaneously to the control group? I thought did_multiplegt_dyn might be an option but it does not seem to work.

Thanks for letting me know!
Thibaud.

[George Ford]

are the treatments assigned at the same time?

[Thibaud Marcesse]

Hi George,

Yes they are. Imagine you have 6 villages (A, B, C, and D). Village A is not reserved for any group (anyone can run for office), while Village B is reserved for women, Village C is reserved for members of an ethnic group and village D reserved for another ethnic group. The different treatments are mutually exclusive but assigned at the same time.

[George Ford]

I'd think you could do straight OLS did with 4 treatment vars. Not sure you have enough villages for clustered SE.

Were they all untreated before some period? If not, you don't have DID, just a means difference.

[Thibaud Marcesse]

Not all of them are effectively treated during the period, so there are both ‘never treated’ and ‘not yet treated villages, which I believe is enough to use the DiD packages