Timeline for Dynamic treatment timing in a panel-DiD framework
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 4, 2019 at 23:08 | comment | added | Thomas Bilach |
@Jason Goal It depends. In the 'classical' DD set-up where every treated firm receives treatment at the same time, then treated and post should be included in the estimation, along with their interaction. To be clear, post should be coded 1 for all post-treatment time points, regardless of which group a firm belongs to. However, the question notes that different firms receive treatment at different times, so the post-period is not well-defined. In this 'general' setting, you should drop post and instead use policy (i.e., treated*post) with a full set of firm and time dummies.
|
|
| Dec 3, 2019 at 20:41 | comment | added | Jason Goal |
This post and answer helped a lot about how to estimate the treatment effect in such a setting but still can not understand why the post of firm 1 is switched on from time==3, not 2, or 4. Besides, since post can not be defined for untreated firms, is that even reasonable to include them in the estimation?
|
|
| Jul 13, 2016 at 13:17 | comment | added | Andy | If you go on Google scholar and look for difference in differences and "multiple periods"/"multiple treatment periods", there should be a bunch of studies popping up. For instance these two sites.nicholasinstitute.duke.edu/environmentaleconomics/files/…, and faculty.bus.lsu.edu/bdepew/… | |
| Jul 13, 2016 at 12:11 | vote | accept | Billywob | ||
| Jul 13, 2016 at 12:09 | comment | added | Billywob | Thank you again. Do you happen to know any paper/book that works with data in which treatment timing differ for the treated? The Woolridge lecture is a bit difficult to follow in my opinion because it doesn't go into that much depth. Furthermore, I find it strange that I cannot find much regarding the topic when "searching" the web, however, I might not be familiar with the terminology used in these instances. | |
| Jul 13, 2016 at 11:56 | comment | added | Andy | No worries, I updated the answer. You can think of the policy dummy as the interaction of the treatment group and the post-treatment period indicators (as in the standard DiD setting). | |
| Jul 13, 2016 at 11:55 | history | edited | Andy | CC BY-SA 3.0 |
added 96 characters in body
|
| Jul 13, 2016 at 11:45 | comment | added | Billywob | Thank you very much for your response Andy, however, in your example with only two firms the post dummy is defined for firm one because there's only one other firm. But when there's a thrid firm with treatment happening in period 3 I'm not sure how you would define the post variable for firm one (not that this is needed in the setup you describe). | |
| Jul 13, 2016 at 10:30 | history | answered | Andy | CC BY-SA 3.0 |