There is no mental gymnastics here, because it is simple mathematics. Here is the decomposition of issues #1 and #2.
Rate of police kills = Rate of stops * kills/stop
It is important that everyone understands that this mathematical relationship always holds. And it holds regardless of where your opinion on whether racial discrimination exists in policing or not. It is inviolate. We can further decompose issue #1 and #2 to add a racial bias adjustment which is subject to judgment.
Rate of stops = Crime Rate * DrivingWhileBlack
Kills/Stop = LegitimateShooting * TriggerFinger
So if we expand everything, we get:
Rate of police kills = CrimeRate * DrivingWhileBlack * LegitimateShooting * TriggerFinger
Let’s normalize everything to 1.0 for white americans, and just use a relative multiplier for blacks. That means we can drop out the LegitimateShooting since that should be the same for all racial groups. Simplifying, and grouping we get:
Relative rate of police kills = Relative crime rate * (DrivingWhileBlack * TriggerFinger)
In other words, if the relative rate of police kills is greater than the relative crime rate, the product of DrivingWhileBlack * TriggerFinger must be > 1 (indicating discrimination). If the product=1, there is no discrimination. And if <1, there is actually leniency.
We have estimates for the relative rate that police shoot blacks relative to whites (~2.4x for all blacks as a whole from the Guardian data). But we also have estimates for various crime rates. As I explained earlier, I chose murder because it is impossible to fake and difficult to ignore. The relative murder rate for blacks is 3.5x that of whites. So plugging in:
2.4 = 3.5 * ( DrivingWhileBlack * TriggerFinger )
In this case, the product of (DrivingWhileBlack * TriggerFinger) is about 0.7. Roland only measured ‘TriggerFinger’ and found it to be about 0.8, suggesting DrivingWhileBlack is close to unity.
Suppose that Roland is wrong as you suggest, and that TriggerFinger is actually 1.5. That means DrivingWhileBlack is considerably less than 1.0, which sounds wrong to me (I am hoping for 1.0).
So at the end of the day, regardless of which expert has an opinion, the mathematical expression still has to work. Feel free to plug in with your own estimates of racial crime rates and police homicide rates. Let me know how that works out.