Generalized DiD: How will the regression equation look like?

[Abiodun Olatunji]

Hi guys,

I´m setting up an generalized DiD model. The policy is adopted at different times (two different years). Due to the help of the forum I understood to some extent how to set up this model in Stata, but honestly I don´t fully understand how does the regression look like in general?

The code for my regression looks like this:

xtreg var1 in_effect i.year control1 control2 control3, fe

In this case in_effect is my DiD variable.
In this paper (https://www.annualreviews.org/doi/pd...-040617-013507 ; p.5) the equation is stated as follows:
Ygt = a_g + b_t + δ*D_gt + e_gt

D_gt would be in my case be in_effect, correct? What would be δ in my case?

Thanks for your help.

[Clyde Schechter]

Yes, in that equation D_gt corresponds to your variable in_effect (meaning that it is 1 in those observations where the firm/country/whatever it is has adopted the policy, and 0 elsewhere. δ is the regression coefficient of in_effect--it is the DID estimate you are trying to get. It is a result you will see in the output of -xtreg-; it is not something you have to create beforehand.

[Abiodun Olatunji]

Thank you, Clyde!

I have one further question related to the same topic.
In the literature a and b are referred as group-fixed and time-fixed effects. So, the variable year would be the time-fixed effect and correspondingly represent the b in regression equation, is this correct? But what about the group-fixed effects, in my code there isn´t any variable which would cover the a.

[Clyde Schechter]

The -fe- covers the group fixed effects.