Comparison period when g>t, i.e., when looking at pre-treatment effects, is t-1, rather than g-1

[jtorcasso]

Suppose no anticipation, and we consider estimating ATT(g,t). Then when g <= t, the comparison period is g-1. But when g > t, the package will use t-1 as the comparison period. Is there a reason why g-1 is not used as the comparison period when g > t? To my knowledge, the paper only refers to g-1 as a comparison period, regardless of the value of t.

Using t-1 as the comparison period also leads to pre-trends that look closer to zero, since you compare adjacent periods, rather than with a common comparison period (e.g., g-1).

Just wondering if this is a bug, or if there is a particular reason for this choice.

Thanks for a very useful package, Jake

[jtorcasso]

I see that you've already discussed this

https://github.com/bcallaway11/did/issues/27

[bcallaway11]

Hi Jake,

Yes, a longer discussion is available in this post.

We'll have the option to use a "universal"/fixed base period in the new version of the code which should be available in the next week or two.

Brant