Most online commentary I’ve seen about this paper is incorrect. I’ve seen this paper used as evidence of police malfeasance because the amount of cocaine seized jumped to 280g. This is the opposite of what’s described in the paper, where the author notes that, based on drug seizure records, amounts seized do not appear to be the cause of this change. After noting that drug seizures are not the cause, the author notes:
I do find bunching at 280g after 2010 in case management data from the Executive Office of the US Attorney (EOUSA). I also find that approximately 30% of prosecutors are responsible for the rise in cases with 280g after 2010, and that there is variation in prosecutor-level bunching both within and between districts. Prosecutors who bunch cases at 280g also have a high share of cases right above 28g after 2010 (the 5-year threshold post-2010) and a high share of cases above 50g prior to 2010 (the 10-year threshold pre-2010). Also, bunching above a mandatory minimum threshold persists across districts for prosecutors who switch districts. Moreover, when a bunching prosecutor switches into a new district, all other attorneys in that district increase their own bunching at mandatory minimums. These results suggest that the observed bunching at sentencing is specifically due to prosecutorial discretion
This is mentioned in the abstract and then expounded on in the introduction (the quoted passage is from the introduction), so I think that most people commenting on this paper can’t have read it. I’ve done a few surveys of comments on papers on blog posts and I generally find that, in cases where it’s possible to identify this (e.g., when the post is mistitled), the vast majority of commenters can’t have read the paper or post they’re commenting on, but that’s a topic for another post.
There is some evidence that something fishy may be going on in seizures (e.g., see Fig. A8.(c)), but if the analysis in the paper is correct, that impact of that is much smaller than the impact of prosecutorial discretion.
[return]