Intuitively I agree with this, and think militarization of law enforcement is terrifying, but this report at a glance has some issues. I want to understand the

Intuitively I agree with this, and think militarization of law enforcement is terrifying, but this report at a glance has some issues. I want to understand the trade off here, but they fail to pose SWAT as a real trade off for officer safety in many cases. For example, they cite that of the raids where police predicted weapons, 35% of the raids turned out to have weapons, 32% did not, and for the remaining 33% it is unknown if they did or did not. The way they interpret this data is: "We found in the course of our investigation that the SWAT team found weapons (the overwhelming majority of which were firearms such as handguns, but rarely assault rifles) in just over one-third of the incidents in which they predicted finding them, which suggests the police are not particularly good at accurately forecasting the presence of weapons." This is transparent bullshit, because in their very table they state that a third of the raids it was unknown. If we assume that the 'unknown' state is random (an assumption), ie ~half of those unknown accounts had weapons and the rest didn't based upon the known positive and negative rates being close to even, this means that no, the police didn't guess only one third of the time correctly, they guessed correctly approx 53% of the time, more than half of the time. The next step is to ask if a 50/50 chance of being right that you might be under threat of a weapon means you should err on the side of SWAT, but they don't ask that, they just say that we should "we would expect them to find weapons in nearly all of the incidents studied." Says who? Risk is a game of probabilities, and obviously there needs to be some threshold for which cops deploy SWAT with regards to the likelihood of weapons being present. Regardless of if the randomness of the unknown assumption is true, they are transparently mis-characterizing the data here. Additionally, their section on race seems transparently broken. They are measuring absolute measures, not relative measures. To understand race bias in SWAT, you need to understand the rate at which SWAT is deployed for warrants for a given demographic. Instead, they spent most of their analysis (if this rate is in there, I missed it) on absolute metrics; but we all know already that on an absolute basis, law enforcement spends a disproportionate amount of time in minority communities. You can argue if this is due to systemic bias or not, but this is obviously a dependent variable if we're trying to answer the question "Do cops deploy SWAT more if they are dealing with minorities?", and it doesn't seem to be treated as such. edit: Classy downvotes, HN thought police!