It sounds to me more like the stop to arrest ratio simply depends on something other than the race of the person being stopped.
Let's say the police notice from experience that driving a recent model mustang while speeding means you're 10x as likely to be carrying drugs than average, so they make sure they stop any speeders in recent model mustangs. The fact that black recent model mustang drivers get caught with drugs in the same ratio as white recent model mustang drivers means the signal the police are actually using works for both groups, not that the police are focused on black drivers over white drivers because someone notices blacks speed in recent model mustangs 3x as much as whites do.
With both groups being arrested with the same ratio (1/3 of stops), that's evidence racism isn't an issue with the stops. If blacks who were stopped only had drugs 1/10 of the time compared to whites 1/2 the time, then you might be able to make a case that's because cops stop too many blacks compared to the pool of actual black drug couriers. In this situation, it sounds like they stop just the right proportions using some other set of indicators (not race) to determine who to stop.
Having played around with this for a while, I agree that the deterrence-maximizing strategy is highly model-dependent and does not necessarily entail a higher search rate for the larger group. It's easy to write down models where this is the case, but equally easy to write down models where it's not.
It sounds to me more like the stop to arrest ratio simply depends on something other than the race of the person being stopped.
Let's say the police notice from experience that driving a recent model mustang while speeding means you're 10x as likely to be carrying drugs than average, so they make sure they stop any speeders in recent model mustangs. The fact that black recent model mustang drivers get caught with drugs in the same ratio as white recent model mustang drivers means the signal the police are actually using works for both groups, not that the police are focused on black drivers over white drivers because someone notices blacks speed in recent model mustangs 3x as much as whites do.
With both groups being arrested with the same ratio (1/3 of stops), that's evidence racism isn't an issue with the stops. If blacks who were stopped only had drugs 1/10 of the time compared to whites 1/2 the time, then you might be able to make a case that's because cops stop too many blacks compared to the pool of actual black drug couriers. In this situation, it sounds like they stop just the right proportions using some other set of indicators (not race) to determine who to stop.
If you write a book on economics and Robin Hanson can only find one point of disagreement with it you must have done a pretty amazing job.
You are a gentleman and a scholar. :)
In short, touche.
Having played around with this for a while, I agree that the deterrence-maximizing strategy is highly model-dependent and does not necessarily entail a higher search rate for the larger group. It's easy to write down models where this is the case, but equally easy to write down models where it's not.