Difference in Differences when treatment ends before the end of the sample

Asked Sep 15 '16 at 21:41

Active Sep 15 '16 at 21:46

Viewed 49 times

Say, if the $T_t$ (time) dummy begins as $T_t=0$ then turns into $T_t=1$ when we have the treatment period and then goes again into $T_t=0$, how does the interpretation of the interaction term change ($D_iT_t$, when $D_i$ is the dummy for the units that received the treatment)?

In such situation, what would be (in general) the more adequate model in place of the DD?

edited Sep 15 '16 at 21:46

asked Sep 15 '16 at 21:41

John Doe

1,275
2
15
24

Related question [here](http://stats.stackexchange.com/questions/225145/difference-in-difference-with-multiple-period-pre-during-and-post-treatment). – dimitriy Sep 15 '16 at 22:09
So I would just add an interaction between the post treatment period and the treated units dummy, $D_i$? – John Doe Sep 15 '16 at 22:28
Yes, and also a post-treatment $t_{3}$ dummy. – dimitriy Sep 15 '16 at 22:49

Difference in Differences when treatment ends before the end of the sample

0 Answers0