leads appear as lags and vice versa using the eventdd command

[Niklas Klinger]

Hallo everybody,

I am using STATA 17 and I have a problem using the user-written eventdd command.
Somehow when I try to implement a panel event study in my output leads are named lags and vice versa.

This e.g. happens when I try to replicate the example provided by Clarke and Tapia-Schythe (2020)¹.
When running the following code

webuse set www.damianclarke.net/stata/

webuse bacon_example.dta, clear

gen timeToTreat = year - _nfd

eventdd asmrs pcinc asmrh cases i.year i.stfips, timevar(timeToTreat) ci(rcap)  cluster(stfips) graph_op(ytitle("Suicides per 1m Women") xlabel(-20(5)25))

accoding to the authos the coef. for lag21 should be -22.92073 . In my output this is the coef. for lead 21. Somehow all leads are called lags and all lags are called leads.

Does anyone know why this happens? I am relatively new to STATA. Any help would be greatly appreciated.

All the best,

Niklas

---

¹Clarke, Damian; Tapia-Schythe, Kathya (2020): Implementing the Panel Event Study. IZA Discussion Paper Series (13524).

[Jared Greathouse]

Unless there's something about the options I'd need to read, this problem doesn't replicate on my machine. Please, show the exact output Stata gives you here.

[Niklas Klinger]

Thank you for the quick reply.
I attached the output aswell as the graph Stata produces when I use the code cited above.
I apologize if this is not the correct way to cite Stata output in a post.

[Jared Greathouse]

Okay so in the event study world, unless I'm mistaken, we use lags and leads a little differently than is common in regression.

Here, what you're looking at is to see if the intervention had any impact before the intervention happened. I don't know why we call them lags and leads, but they're not used in the same sense that we're used to seeing them. My point here is that there's nothing objectionable here, all you're seeing here is the effect of the intervention t time periods before and after the event.

[Niklas Klinger]

Thank you very much Jared. That helped a lot.

[Aiqi Sun]

I was confused by this as well. I suspect there was a bug in an earlier version of the eventdd package, which the authors used for their working paper. The published version appears to have corrected it, and the results now align with yours.

Published verion:

https://journals.sagepub.com/doi/full/10.1177/1536867X211063144