Difference in differences estimation using a regression

samr

New Member
#1
Hello,
Difference in difference estimation implies calculating (P2-P1)-(S2-S1). That would give the result of the treatment.
1533828678508.png
It seems to me that the estimation is valid for a specific time point, time 2 at this example.
I have read however (for example, at wikipedia), that DiD estimation can be estimated using a regression model
1533828903465.png
Where Y(ist) is the dependent variable , γ(s) is a dummy for the control\treatment group, λ(t) is a dummy for the before\after period. D(st), afaik, is a dummy variable that equals 1 if a person is in the treatment group, AND if the time is after the treatment time.

What I don't understand however is the equivalence between the two. While the first form is supposed to calculate difference-in-differences estimator between two specific time points, the regression formulation takes into account all the time points after the treatment. So it seems that the regression will somehow fit the estimator to all the points after the treatment period, rather take a specific point, and estimate DiD based upon it.

Am I wrong?
Thank you
 

Attachments

samr

New Member
#3
I think I understood my mistake. While I have written that
While the first form is supposed to calculate difference-in-differences estimator between two specific time points, the regression formulation takes into account all the time points after the treatment.
The first point is wrong - in standard diff-in-diff estimation, I have learned that one is supposed to calculate difference-in-difference between averages before, and averages after. I simply missed that.

Thanks.