DID two subsequent tretments for the same group

[Juan Price Elton]

Dear collegues. I hope someone out there is conversant with this issue. I am assessing the effect of the introduction of a tax on an environmental indicator. There is one group that was exposed to this treatment and another which was not (and let´s say that this treatment was assigned randomly). So I could in principle measure this effect. But the tax was increased (let´s call it a second treatment) after the first two years since it was introduced. I would also like to test the effect of this second treatment on the same outcome variable but there is not (I would say) a valid control group (a group that was exposed to the introduction but not to the increase). Is there any methodology that somehow allows to cope with this problem? I thank you very much and send you my best regards. Juan José Price. PS: I know this forum is intended to be a space to discuss problems users may have with Stata but I thought (and still think) that maybe there is someone amongst you who is conversant with this econometric issue and can lend me a hand.

[Jared Greathouse]

If the treatment was assigned randomly, whence cometh the need for a DD design? You can simply calculate the mean difference for each time point after the intervention... I think, unless there's a good reason why the second increase was so much different from the first one.

[Juan Price Elton]

Oh you are right. No, sorry, wasn´t random. I dont know what I was thinking about. The treatmente was assigned to two groups based on a legal status. So, yes, DID is warranted. My problem is that the tax was increased to all the treated observations two years after they were subject to the first treatment, so there is not a group that received the first treatment but not the second. Sorry again for the mistake and thank you for pointing it out. Juan

[Jared Greathouse]

Uhhhhhhhhhhhh....... you may be able to get away with ignoring the second treatment and keeping the effects you get as heterogeneous treatment effects. If not, there are Stata estimators that do DD when a unit is treated more than once, didmultiple_gt (from ssc I think).

[Juan Price Elton]

Thank Jared. I just checked this out (the command did_multiplegt). I will start "studying it and the related literature". Thank you again. Cheers, Juan

[Juan Price Elton]

Thanks Jared. I just checked this out (the command did_multiplegt). I will start "studying it and the related literature". Thank you again. Cheers, Juan