I'm a master's student working on a project for my econometrics class and I could use some advice. I'm looking at the effects of a Supreme Court decision, Tapia v. U.S, on the length of sentences handed out for drug crimes.

Statute has long been clear that a convict's need for rehab can't be used as a reason to imprison him (as opposed to probation), but before Tapia, circuits were split on whether or not the need for rehabilitation can be used to lengthen a sentence (after the prison/probation decision has been made). In Tapia, the SC ruled that a sentence cannot be lengthened due to a convict's need for rehab. After the decision, the circuits quickly split again. Some circuits allow a sentencing judge to consider rehab during sentencing, as long as it isn't the dominant factor or doesn't explicitly lead to a longer sentence (judges often list a litany of reasons as to why they are passing a particular sentence, and removing/adding a reason would not necessarily change the sentence). Others say that any consideration of rehab during sentencing violates Tapia. The before-and-after splits look like this:
Post-Tapia
Not Permitted Silent Permitted
Pre-Tapia Not Permitted 7th, 10th, 11th DC 2nd, 3rd
Silent 1st, 4th
Permitted 9th 5th, 6th, 8th
Sentencing data is readily available, so I have data from both before and after Tapia for all circuits (I'm using one year's dataset from each period). Right now, I'm just doing a difference-in-differences estimations for various drug crimes between 7th/10th/11th and 2nd/3rd, as those circuits were on the same side of the split pre-Tapia. That should show me the effects of the new split, I believe. I think I could also do an estimation between 2nd/3rd and 5th/6th/8th to see the effects of Tapia's elimination of the original split, but I haven't done this yet.

Does this sound like the best plan to you guys? I'll obviously use as many controlling variables as I can, and I'm not worried about the law side of things. Is there a different tool I could use that could use the reshuffle to my advantage and show me the separate effects of both the end of the first split and the onset of another? Any advice specific to Stata? Thanks.

(To avoid duplicated efforts, please know that I've cross-posted with StackExchange here: https://stats.stackexchange.com/ques...ervening-event.)