Why Criminal Justice Reform Needs to Think Big
There are too many police officers in the US. We need patient, tolerant, unbiased, hardheaded, softhearted cops. But we only need a few.
In the last several weeks, justice policy reform has rocketed onto the nation’s policy docket. This is a very welcome development, as anyone who has studied crime and justice can tell you that the first principle of justice reform is that the juvenile and criminal justice systems are rife with racial disparities. There is no need to reinvent that research here. At the Washington Post, Radley Balko has detailed the racial disparities that exist at every juncture of the criminal justice system. Mike Shor, an economics professor at the University of Connecticut, has a twitter thread that currently has over 700 academic papers detailing the various aspects of racial disparities in the system. The evidence for racial disparities in the system is, I think, incontrovertible.
But though welcome, I have two worries about the current calls for policing reform, especially the legislative fixes that are tickling Congress and emerging in state capitols. First, the window for big reforms is short and closing. The best example I can think of is the Equal Rights Amendment which 33 states ratified in less than two years in the early 1970s, but which ultimately floundered as opposition ramped up and swamped the early momentum. On the flip side are the absolutely terrible Stand your Ground laws which blazed through almost two dozen state legislatures in just two consecutive sessions in 2005 and 2006, before most legislators even bothered to read these bad laws. Big policy changes happen in a fever, and fevers break.
Also, the policy itch Congress seems set to scratch misses the forest for the trees. Today’s outrages are just symptoms of a larger syndrome. And big policy syndromes require big policy reforms. Drunk driving reforms went nowhere until the feds pulled out the big guns and threatened to withhold all highway funds, potentially crippling the nation’s infrastructure. Drinking ages went up, speed limits came down, and BAC levels were uniformly set. The reality is that you can’t plant a policy seed and expect a thousand flowers to bloom, or a thousand points of light to blink or whatever. You have to start big to be big. Little tweaks do not grow up to be big reforms.
And big reforms in policing are needed. Just like smoking at your desk, our policing system is a relic of a time that has passed. I can no longer avoid the broader conclusion: there are way too many police officers in the US. It's not just a bad apples problem, it's a too many apples problem. In a strict utilitarian framework, once you add into the cost-benefit calculus the harms that are done to entire communities of people from systematically racist policing, the benefits are more than offset. More importantly, from a deontological perspective, our justice system is not just. How Americans police themselves is a critical expression our shared values.
Any rigorous discussion about policing reform should begin by first asking the question: how many police should we have? There is certainly precedent for this question. When the Berlin Wall came down and the Soviet empire crumbled, progressive policy reformers, led by Margaret Thatcher and George HW Bush, chanted the mantra of the Peace Dividend and loudly asked the question, ‘how many troops should we have?” There is also empirical support for asking this question about the police today as there is no more than half as much violence in the US now as there was in the early 1990s when the peace dividend was in vogue. The parallels seem obvious.
But we need to ask this question from the right frame. The typical math for figuring if more cops should be hired is 'how much crime does one officer prevent?' This seemingly innocuous question is loaded for bear, armed to the teeth if you will, and sets off on a path that leads to only one plausible conclusion. This is not a strawman argument, by the way, as the statistical models of most serious economic scholarship and most serious empirical legal scholarship seek to parameterize the marginal deterrent effect of adding another police officer. And, by the way, these papers always study adding an officer, never subtracting one, and when they do consider what would happen if a department shrinks, they assume perfect symmetry—whatever gains were made from adding an officer (which is studied) are assumed to be lost by removing one (which is not studied). Three big ideas are missed in this framing.
First, framing the question around the marginal benefit of additional police officers assumes policing is a public good. But policing satisfies neither condition of a public good. To be a public good, policing must be nonrival, where my use of policing does not affect your use of policing. And, policing must be nonexcludable, where everyone has equal access to the benefits of policing. In the wake of unrelenting disclosures of egregious use of force, public goods claims appear farcical, prima facie.
Second, this ‘how much crime does a police officer prevent?’ thinking ignores the costs of policing or presumes there aren’t any. Intrinsically and unequivocally, police engage in an activity with real social costs--they restrict liberty through the threat of detention. The way this gets ignored by this question is through the sneaky use of ‘prevent’. Police do not prevent crime, at least not as the word is prevention is applied to literally every other science. We prevent disease by mitigating the risks that cause it—we do not deter disease. We build in technology in cars and trucks to prevent crashes (think antilock brakes) and to mitigate the harm of crashes that happen anyway (think airbags)—we do not deter car crashes. The principle way police benefit the public is by making the people too afraid of the repercussions from committing a crime to actual break the law. We call this trade-off ‘deterrence’, and it is the primary mechanism of modern policing. Deterrence is about instilling fear in the population who is policed. Fear is a cost, not a benefit. So, perhaps that is two ideas in one—deterrence is both costly and wholly different in its costs from prevention.
Third, we should ask rather than assume, that additional police officers have a benefit. It is startling how rarely we apply basic performance measurement to policing, even just a few reasonable outcomes measures. One obvious outcome measure is to ask whether police, with all their new technology, training, and equipment, have gotten better at solving crimes. Police evaluate themselves using the clearance rate. By this measure, the police are not getting better. Per capita, there are half as many crimes to investigate today as there were in 1990, and there are about the same number of sworn law enforcement officers per capita (more if you throw in the rapidly growing federal forces, but let’s not). But even with fewer crimes to solve, the clearance rate hasn't gone up at all in 30 years and most crimes remain unsolved.
A much more reasonable approach to thinking about how many police there should be in America, is to ask, “what activities do police engage in that they are uniquely qualified to perform?” Put another way, we might ask, “what should police do?” If the answer is much narrower than what police do today, then the path to smaller police forces is clear. One way to think about this is to look at how police spend their time today and to ask whether having a sworn police officer perform that task is a good use of scarce public resources. I covered this to some extent last week, in writing about how traffic duties would be more effective and just if these tasks were outsourced. But there are many other activities, particularly those that are explicitly about physical and social disorder, that need to be entirely rethought, but more on that another time.
It would also be wise to think hard about whether we default to thinking about police as having unique capabilities to prevent violence. We should start by asking whether there is anything communities can do to slow crime and violence before imposing police—and their costs—locally. There is substantial evidence that community anti-violence programs are effective, perhaps even more so than the police and proportionate investments should be made there. We should invest in preventing social ills that raise the risk of violence and victimization, from job training to behavioral to health to affordable housing and better education. And, as Yale professor Tracey Meares says about police officers, we should acknowledge that, “one person couldn’t possibly be trained really well to respond to all of the kinds of things that we expect police to respond to” and instead hire experts in mental health crisis’s and domestic violence incidents and empower them as first responders.
As I sat down to write this today, I see that NYU professor Anna Harvey wrote a terrific twitter thread detailing more than a dozen studies that show that violence reduction from programs and policies that have nothing to do with policing or criminal justice. These programs, for the most part, alleviate poverty, improve education and create real, sustainable opportunity. This should challenge another one of our priors, that the crime decline was the direct result of better policing and more effective criminal justice systems. I don’t think this is true, and believe the crime decline had more to do with America becoming older and richer and having cell phones and breathing less leaded gas and a bunch of other things that are best left for another time. But the point is that there is no reason to think that history shows that the police possess a unique crime fighting ability compared to all of the alternatives.
And, regardless of how many police we choose to have, we must change police behavior. We must demand that police become more socially and economically diverse, that they hire more women, and that they prioritize trust building over inspiring fear. Policing must be transparent, and use of force must be rare, diligently recorded and carefully investigated in every instance, preferably by someone outside the jurisdiction.
We need to lighten the touch of the criminal justice system. We need fewer officers, and more officers who are better trained. We need a system that balances the harms that are inherent in deterrence, suppression and surveillance with the good police could do.
We need good cops, perhaps now more than ever. We need patient, tolerant, unbiased, hardheaded, softhearted cops. They are out there and we need a system that identifies them and rewards them and roots out evil. But most of all, we need experts trained in the nuances specific to each call for service. And we need to acknowledge that many, perhaps most of those experts do not need to wear blue. We need fewer police.
Idea (Updating Priors)
It's not just a bad apples problem, it's a too many apples problem
I used the phrase in the header in the bit I wrote above, but then came across this really impressive scholarship this week from Aaron Chalfin and Jacob Kaplan at the University of Pennsylvania. In this working paper, Chalfin and Kaplan test the idea that policing has a bad apples problem. The paper is notable for both its conclusions and its method. The conclusion is surprising, which is that they disagree with my assumption (based on an untested prior belief of mine) that there is a bad apples problem in policing. They find evidence that suggests that bad apples are not the problem in contemporary policing. I think this leads to the conclusion that the problem is actually much worse than it appears. If it is not a bad apples problem, it is a structural problem and this is much harder to solve. Although they disagree with my sophistry, I think their finding lends some credence to the idea that there should be fewer police.
I want to take a few sentences to describe their method because though not entirely novel, it is a really powerful way to challenge untested ideas. It also makes their finding more credible. This gets a bit tricky, but bear with me, it is worth the journey. The idea is to compare the number of police officers with use of force (UOF) complaints against the right baseline. The wrong baseline is no baseline at all. If you just did that you’d see that there were an average of 1.4 UOF complains with a few officers having several and most having no complaints. You would probably conclude that there were a few bad apples. However, if you thought about it for a moment, you would have trouble articulating what pattern you would expect to see if there were no bad apples. What would a no bad apples pattern look like? 40 officers have two complaints and 60 have 1? 1 has 140 and the rest have 0?
A better way to understand UOF would be to compare the observed UOF complaints against a plot of randomly generated complaints. The key here is that there are very few complaints in the data, and thus the few with more than one complaint really stands out. I’m over simplifying here, but essentially the authors randomly assigned 140 UOF complaints to 100 officers with replacement (meaning that once an officer had one random complaint there were returned to the pool with a chance to have another randomly assigned to them). Due to the vagaries of chance, many officers had no UOF complaints randomly assigned, many had a single complaint, and a few had several complaints. That plot of randomly assigned complaints turns out to look a lot like the observed data—both are Poisson distributions if you want to be technical about it. Which suggests there aren’t that many bad apples. Which is troubling.
This way of thinking about the universe is helpful in other problems were events are relatively rare. This is essentially the same kind of reasoning University of Wisconsin professor Jordan Ellenberg uses in his book How Not to be Wrong when he shows that multiple lottery winners don’t have a magical ability, they just won the luck lottery.
I study firearms and the one issue around guns that seems to cross partisan lines in the role of guns in suicides, as gun rights supporters are often surprised (and open) to the data on gun ownership and increased risk. While doing that research I came across this utterly remarkable article: I am not Always Very Attached to be Alive. And how mundane that feeling is.
When asked on twitter what I don’t like that everyone else likes, I have an all-purpose answer: plot. It’s no wonder I am drawn to travel books, where there is never a plot except that Bill Bryson wanders around and makes snarky comments, or John Steinbeck and Charley drive around and have rich reflections. The best travel book I had never heard of is, hands down, is William Least Heat-Moon’s Blue Highways. The New York Times says, “the book is wonderful.” It is that.
I am sorry that I don’t have the skills to put this graphic up here, but it shows precisely how highly correlated shootings and poverty are. Check it out, it’s just one click. https://phl.maps.arcgis.com/apps/StorytellingSwipe/index.html?appid=d3394c20f95d472b9038976c8791ecf5#
Back before Marcus Mumford kicked the drum while signing the Cave, the incomparable Levon Helm, did all that and played guitar too. There’s winners and there’s losers and I’m south of that line.