Ok, let's go with the stop and frisk example. Suppose that a white person has a 1 in 1000 chance of carrying an unlicensed firearm in public. Now suppose a black person has a 1 in 500 chance. If you frisk the population RANDOMLY, you will see that the black people are packing heat twice as much as a white person. Great! Now, what do you think will happen to the data if you sample blacks at twice the rate as whites?

If your sample is large enough the rates shouldn't change whether you're frisking 10,000 people or 1 million people. I mean the argument could be that whites are sampled in such small amounts (even though they really aren't purely from the fact that demographically they are a much higher % of absolute population) resulting in a inaccurate representation of population mean of white crime, but it's a weak one.

The reason why there's relatively higher sampling for blacks is because they live in hot spot areas which result in higher interactions with police compared to whites. There's no great conspiracy here. Inner city violent crime rates dwarf suburb or rural. Inner cities have high % black meanwhile the other two don't.

Very good, the rates SHOULDN'T change but the raw data does (because you obviously have more samples). You should still have a solid estimator if you know what you're doing.

Now, suppose your goal is to maximize criminals caught via stop and frisk. What is the optimal allocation of resources assuming that blacks are outnumbered 10 to 1?

I also disagreed with that sentiment, but that's what they argued back then. I dont have that discussion bookmarked though, not sure if it really were skinned and thor, might have been other Pard lefties.

But they don't have equal crime rates and thats the point. Take two neighborhoods, one with 10x the murder rate and 10x the policing. Is that policing just producing a statistical artifact? Do the dead spring back to life if no cops are around to investigate their murders?

My neighborhood averages about 0.1-0.2 per 100,000 homicide rate (only if you count enough decades into the past to measure a single homicide and divide it out- its been zero for many years). Minneapolis has already had 50 homicides in 8 months so lets say 75 for 17.6 per 100,000.
Should police be patrolling my mean streets as vigilantly as lake street?

The point is that "per encounter" is not sufficient. You have to add in other factors to get even a basic analysis. Just taking the per encounter rate results in masking of obvious abuses where they stopped Jamal 10x more so their killing rate doesnt look bad, even though they should have only stopped Jamal once

Murder is an exception in this regard, since murders are unlikely to go unnoticed. For most other types of crimes, the dark figure is far higher, so that the intensity of policing has a higher chance of perpetuating itself.

There's also this. Most crimes are unreported and unnoticed. I've known dozens of white drug dealers but they will never get stopped for a "broken tail light" and searched, so they're never going to be in the official statistics

Straight from the mouth of your favorite painting leftist

The optimal resource allocation isn't dependent on race but where crime happens. You're going and deploying stop and frisk where most of the crime happens. Again, there's no conspiracy here to maximize heads on spikes to show off. Reactive policing like stop and frisk happened because the crime stats based on data directed the cops there. They are doing stop and frisk in this neighborhood because 80% of all crack related violent crime happened in this area not because some other reason.

I don't have the models for optimal because these aren't linear and there are asymptomatic qualities to modeling this stuff but the way you're framing the question misses the point of why it got instituted.

This post was edited by ofthevoid on Aug 25 2020 02:18pm