Define a "good measure". Also - what's the alternative?

You'd have to show representative statistics (with variability measures) to arrive an any definite conclusions.

@ventsyv Convictions of fraud, harassment, corruption, etc. If law and order matters, then the procedures of law and justice matter.

@curiousdannii How can that metric matter, unless there's an assumption that some exact proportion of complaints (which aren't met in most departments) are valid?

@fredsbend Any number of indisputable cases of police brutality that have no consequences matter. I don't know whether this guy's previous complaints are indisputable though.

@fredsbend The standard of "beyond reasonable doubt" is specifically for criminal trials. There are arguments for why this might be the best standard in that context, where somebody's liberty or even their life may be at stack. It's not at all obvious that this is the only appropriate standard for something like NYT commentary.

@curiousdannii Well, if you can find a true scotsman, let me know. Otherwise, that's why we have the courts. And, ironically, that's where government wrongs (such as corruption that protects bad cops) are also tested and corrected. If you value the system we have then you believe the procedures are reliable.

@GeoffreyBrent When juxtaposed against a wrongful killing, I'd hope this "list of complaints" theory, which is used to argue there's some base corruption in most police departments in the country, produces comparable wrongs, as in, criminal wrongdoing. If not, then that leaves the NYT a bit dishonest here.

I would just like to point out that even if there was the strongest correlation imaginable, it wouldn’t be actionable since anybody can file a complaint against any police officer to try and get them in trouble. (Note that the question doesn’t specify whether the complaints were considered valid.)

@fredsbend "If you value the system we have then you believe the procedures are reliable." - I think we can broadly value the system and also understand that sometimes things go badly wrong: sometimes systematically, sometimes inadvertently, sometimes recklessly or negligently and sometimes deliberately. Partially because police are protected from the consequences of their actions that non-police would face. E.g "qualified immunity" cato.org/blog/why-qualified-immunity "police may [steal from,] shoot, beat, kill, and fatally neglect people in their custody, practically at will"

Policing in the us is similar to schools in that there are dramatic differences from locale to locale. You need a boatload of statistical controls before any of the numbers have any real meaning.

For example @curiousdannii suggestion is laughable for districts like ferguson -- the whole problem there is that the entire department was corrupt which is exactly why the feds had to step in. Of course bad cops there wouldn't be disciplined, that was the whole problem in the first place.

@Lag "Reliable" in my context refers to the whole of our government. Corruption rings are busted with some frequency, and ring leaders are punished. Yet articles like this NYT one push a story that nearly all police departments protect bad cops. Any claim of systemic problems inherently questions the reliability of justice procedures.

@fredsbend Counting complaints is certainly an imperfect measure of police wrongdoing, but it's not at all clear why counts of convictions would be a more reliable one.

@GeoffreyBrent Because due process is a thing. Because a police department can be held liable for wrongful termination, as much as for officer misbehavior. Because presumption of innocence is the foundation of our jurisprudence. Plainly, because complaints alone can't mean anything other than complaints, else we begin to violate the procedures for justice. It's safe to assume a bad officer will generate complaints. It's not safe to assume a good officer won't.

@fredsbend but will the "bad" officer generate more complaints, than the "bad" one in the same department? If he will, than it can still be used as some measure of "abusiveness" or whatever.

@11684 Yup, this is the problem with using complaints as a metric. In a discussion elsewhere on this a cop pointed out that he had been the subject of a complaint--for the "wrong" of wearing sunglasses while talking to someone.

This question has earned a number of (unacceptable) unreferenced theoretical answers of the form "In my view of the world, there would be noise. Noise would make the metric useless. QED." Compare that to my answer: "This peer-reviewed study of the actual data shows there is noise. This peer-reviewed study of the actual data shows that despite the noise, the metric is strongly predictive of certain outcomes." If you want to argue against that, you need to bring better empirical studies, not your assumptions and speculations about how the world might work.

My truck has burned nearly 14,000 gallons of gasoline. That's not a very useful fact without other information. Adding in that my truck is 26 years old and has well over 200,000 miles makes the figure more insightful.

@fredsbend A NYT article is not "jurisprudence" and the NYT does not have the authority to fire police, so I'm not sure how that's relevant. Even within jurisprudence, different standards of proof apply to different contexts - this is how OJ was found civilly liable for the deaths of Nicole Brown Simpson and Ron Goodman, but not criminally guilty of their murder. You still haven't made a case for why the "beyond reasonable doubt" standard is the only relevant one in this context.