I suspect that data on so-called random stops would be next to impossible to produce short of actually getting the specific police logs or radio logs.

Quote Originally Posted by SadisticNature View Post
While you're at it you could try and find data on random stops. Police tend to be more suspicious of black people, so there are a far higher rate of random stops, and a far higher rate of searches at the border etc.

I have several acquaintances who smoke marijuana, 4 of whom took their personal amounts across the border, the three white guys weren't searched, the one minority was. Admittedly this is only an anecdotal case, but if stuff like this plays out in the larger data, then its quite likely blacks are not necessarily committing more crimes but rather are being treated with suspicion and hence are caught more frequently.

So looking it how arrests compare to convictions wouldn't show you the larger picture of what level of crimes are being committed. It would only show you what level of crimes are being caught. Assuming a random sampling is certainly problematic as there is strong evidence of bias. Take for instance racial profiling:

The idea behind it was that blacks committed a higher percentage of crimes, so if a police officer has two suspicious people (one white, one black) fleeing the scene of a crime and can only chase one of them they go after the black guy. There are several possibilities for what actually happened here:

Case (i): The black guy did it. They likely catch him and prosecute.

Case (ii): The white guy did it. He escapes the initial scene, and chances are somewhat poor that they track him down to catch him and prosecute.

Case (iii): They were accomplices. The black guy likely gets caught and is prosecuted. He may or may not turn over his accomplices.

So if you have a police force that responds to a chase scene in this way, you would have bias in your data. The white guy is far more likely to not be caught for this crime than the black guy.