I think what OP alluded too is not understanding preexisting bias.
"Where should police patrol" ->where crime is highest.
Where is crime highest ->existing reports point to X neighborhood having the most arrests.
Without asking "why" it may ignore That neighborhood is black, so historically racist decisions lead to extra patrolling (but no evidence of increased crime rates) which led to more arrests.
Your logic could be misleading if not parsed carefully. It's not that police shouldn't patrol where crime is highest--of course they should patrol there, if they know where that is. Rather, the point is that "most arrests" is not necessarily an accurate way to detect where crime is highest.
Yes, except this is trivial to normalize for - just track arrests per cop. I feel like I constantly see criticisms like this, that use extremely naive metrics as straw men to argue against the idea of measuring things. It's extremely misguided (I know you're just explaining what he meant and not necessarily promulgating that idea yourself).
Sure, but then a conservative outcry is guaranteed: "Everybody can see that blacks are more criminal, the statistics show it clearly. But now the government is trying to hide it by gaming the data!"
As the public doesn't understand statistics it would be easy to make that claim a truth by repeating it a few times.
Everyone manipulates statistics in their favor, be that to support a claim ("the numbers say that if we do X then Y", guns/less gun deaths, social healthcare/improve quality of life) or to cause outcry ("you can't say X is/are/has Y, the numbers are Z", gender A/higher standard deviation for Q/sexist, monopolies/damaging/biased) regardless of the side of politics you are on in the US.
Statistics are hard and the implications of some result are sometimes not intuitive. People don't care about what a standard deviation is, they just want more standard deviation than the other people over there because that makes them better.
More like: "In a political context gathering data is not at all a guaranteed way to get further to the truth."
I don't want to make a case for dismissing an empirical approach per se!
An current example from Germany right now: There's a part of the country that's in uproar because of new immigrants that came in 2015. Some people felt threatened and demanded more police in public spaces were immigrant youth gathers. So more police is present at those public spaces. Naturally this resulted in them being aware of more violent events (mostly groups of immigrant kids scuffling), which made the immigrant share of violent crimes double, and people feel "rightfully" afraid, as the stats now proof their suspicions! But the reality of course is that German kids have fights in public, too; police just didn't take note and so those fights don't end up in the data gathered.
> More like: "In a political context gathering data is not at all a guaranteed way to get further to the truth."
It's not guaranteed, no. But it's the only way we have any hope of it.
> An current example from Germany right now: There's a part of the country that's in uproar because of new immigrants that came in 2015. Some people felt threatened and demanded more police in public spaces were immigrant youth gathers. So more police is present at those public spaces. Naturally this resulted in them being aware of more violent events (mostly groups of immigrant kids scuffling), which made the immigrant share of violent crimes double, and people feel "rightfully" afraid, as the stats now proof their suspicions! But the reality of course is that German kids have fights in public, too; police just didn't take note and so those fights don't end up in the data gathered.
Is that actually true, though? I'm not familiar with the figures, but it's not unreasonable to imagine there may genuinely be more violent incidents among migrants from a war torn region like that. There were certainly serious cultural assimilation problems:
Well, this isn't about murder. Murder rates didn't raise, but statistics for scuffles did. In both cases this is about crimes among immigrants, though, not about crimes of immigrants against Germans.
Yes, as the entire point of this thread is problematic application of data in a political context. The data is violent crime stats. The false interpretation is fear of the native polulation. The native population is not affected by the violence, ergo the fear is misplaced.
That you brought up the murder in the first place is exactly the misuse of statistics I spoke of. About 400 people get murdered every year in Germany, and you bring up exactly the one recent case were the suspects are immigrants.
This is easy. What are we optimizing for? Crime fighting effectiveness of police officers. Can we measure that? Sure, it's number of convictions per officer-hour of patrol.
You could have pointed out that number of arrests per area is the wrong measure and why. That would have been rational. Instead, you chose to drag identity politics into it. Brilliant.
I'm not saying "crime arrests per area is wrong".
@xamuel explained it better than I could:
>It's not that police shouldn't patrol where crime is highest--of course they should patrol there, if they know where that is. Rather, the point is that "most arrests" is not necessarily an accurate way to detect where crime is highest.
Here are the murder offender data from the FBI [1]. Your job is to bring down murder rates. You have limited resources. How would you now allocate them using "evidence based policies"?
I think the comment above yours may be misleading. Perhaps lawmakers are following a different set of metrics (contributions from private prison industry) when considering their rulemaking.
One look at crime stats and this will not be a favorable idea.