What Can Be Learned from Research? A Lot!

Reducing Fatal Police Shootings as System Crashes: Research, Theory, and Practice

By Lawrence W. Sherman

Annual Review of Criminology. 2018. 1:421–49

[Ed. Note: Excerpts follow from a very fine and necessary article by Larry Sherman in the Annual Review of Criminology. I have to admit that much of what I have argued since Ferguson is included in this paper. What Sherman is proposing (after his many years as a police researcher in our country and in the U.K.) needs discussion (and eventual implementation). I have highlighted some sections that I thought extremely important to consider.]


Reducing Fatal Police Shootings as System Crashes:

Research, Theory, and Practice

lawrence_shermanBy Lawrence W. Sherman

“America has reduced fatal police shootings once before, with criminology playing a key role in what can be called the First Great Awakening1of both public and scholarly sentiment against avoidable police shootings. From 1970 through 1985, 50 cities of more than 250,000 residents each took actions that cut in half the annual total count of citizens killed by police in those cities from 353to 172 per year.

“As big-city homicide rates spiked, criminologists and national news media seemed unaware that police killings of citizens went back up in the early 1990s, when many police agencies shifted from revolvers to semiautomatic pistols with large ammunition clips. The influence of the US Supreme Court decision In Graham v. Connor (1989) is also blamed for reversing the benefits garner-cause in Graham the Court ruled that police could justify killing people if they reasonably believed that the person shot was putting a life in danger.

“Yet Fyfe and Sherman both concluded, independently, that merely changing [deadly force] policies was not enough. Other elements of organizational change seemed to be essential, including external demands and internal leadership, both ensuring that policies are implemented and enforced.

“The evidence suggested that although policy restrictions made some difference, the rhetorical messages from the Mayor and Police Commissioner produced effects that overrode the policy content.

“After a series of murders of [Victoria] city police in the 1980s, a new training emphasis on danger to police apparently encouraged more preemptive use of force, with a sharp increase in fatal shootings of citizens. That led, in turn, to a public outcry—including the State Coroner’s criticism of a culture of bravery leading to rapid confrontations rather than delay and negotiation. The new response in the early 1990s was to launch Project Beacon: A new 5-day training project for all operational officers, underpinned by a ‘safety-first’ philosophy: safety for officers, safety for the community and safety for the suspect…. The immediate impact of this training was remarkable, and a culture of safety first had effectively permeated through all levels of Victoria Police.

“Fatal police shootings statewide by police in Victoria dropped by 50% in the short run after Project Beacon. By a few years later, however, the numbers began to climb again, with mentally ill suspects comprising almost half of persons killed. This echoes the conclusion from US case studies: that there is no single intervention that is likely to have a lasting effect on fatal shootings. Nonetheless, the first wave of US research featured substantial attention to policy interventions. That is more than can be said of most recent research

“[Research by Frank Zimring (When Police Kill. Cambridge, MA: Harvard Univ. Press, 2017) found that:

  1. Proportionate to the national population, African Americans are 2.3 times more likely to be killed than whites, and Native Americans are twice as likely; there is no elevated rate of death for Hispanic over non-Hispanic whites.
  2. The most common situations in which police killed people were disturbances (23%), of which approximately half were domestic.
  3. There was no firearm present in 44% of fatal police shootings.
  4. A firearm or knife was in the decedent’s possession in 72% of deaths.
  5. More than one officer is present in most police killings (65%), but 37% of killings by officers who are alone kill suspects who have no guns, compared to 11% of deaths occurring with multiple officers present.
  6. In the limited portion of cases reporting the number of bullets fired, police fired four or more bullets at the suspect in 32% of cases, thus increasing the risk of death.

“The Philadelphia study found significantly, but not substantially, higher levels of likelihood to shoot among officers with less self-control, as measured by a range of noncompliant behavior such as traffic violations.

“Zimring reports (but does not relate to the organizational hypothesis of a leadership change): The 62 shootings in the year before Charles Ramsey became police commissioner (in January 2008) were reduced to 42 in his first year and (with fluctuations) down to 23 in 2015, his last year as head of Philadelphia police.

“As a matter of context, the continuing effects of change in leadership, such as the appointment of Commissioner Ramsey, are important to consider understanding how to implement any systemic changes in the causal pathways to police shooting practices… more restrictive policies appeared to reduce the use of force.

“Perhaps the greatest gap in post-Ferguson criminology is what might be called preventative imagination. It is one thing to say that policies should be more restrictive. It is quite another to say exactly what should be restricted and how compliance with those restrictions should be achieved… broad visions for policing in general, not ideas aimed at police shootings in particular.

“[Zimring’s] four ways to save lives is a clear starting point:

  1. Fewer shooting incidents,
  2. Fewer bullets fired,
  3. Immediate medical attention,
  4. Immediate transportation to a trauma center.

Just having more rules and threats of punishment may not be enough to reduce deaths. Understanding how to change police agencies can only come from studying when they do or don’t change, and the many factors affecting such changes.

“Perhaps the best theory that criminology can apply to reducing police shootings is not inherently criminological but organizational: one that helps to provide change in operational systems rather than hold more individuals blameworthy.

“System accidents: The broader term of art [Perrow] chose for these events is system accidents, emphasizing complexity rather than individual fault or blame as the root cause of the catastrophes

“Hence, this review proposes to adapt, modify, and rename the original concept of system accidents as system crashes. This evolution in both content and label is consistent with recent histories of the word accident as one developed by factory owners to avoid blame for injuries to workers caused by unsafe factory systems.

“So, what does it mean to call police shootings system crashes and why does it matter? It matters because it signals a shift from a blame culture to a learning culture. It means that we can study not only who, if anyone, was at fault but also what processes went wrong and how we can fix them.

“As Braithwaite suggests, it is important to confront individuals and organizations with the harm that their conduct causes. But it is equally important to do it in a way that allows them to express remorse and seek redemption by launching a new course of action.

Braithwaite’s graphic model is: Confrontation ⇒ Truth ⇒ Prevention.

“But the success of such confrontations depends on the emotional content of the language the confronters employ.

“Any confrontation with a person being held accountable for harm can easily come across as a statement that ‘you are a bad person’ rather than that ‘we know you are a good person, but your actions are harmful, and we must learn together how to stop them.’  It is worth trying to shift national and local dialog on police shootings from blame to learning.

“The central point Perrow made in defining the concept of system accidents is that the urge to blame individuals often obstructs the search for organizational solutions.

If a system-crash perspective can help build a consensus that many dimensions of police systems need to be changed to reduce unnecessary deaths… police and their constituencies might start a dialog over the details of which system changes to make.

“Perrow shows how the post-incident reviews rarely identify the true culprit: It is the complexity of the high-risk systems that causes extreme harm.

“Similarly, fatal police shootings shine the spotlight on the shooter rather than on the complex organizational processes that recruited, hired, trained, supervised, disciplined, assigned, and dispatched the shooter before anyone faced a split-second decision to shoot

“The good news in Perrow’s analysis is that some kinds of systems (like air travel) begin with high rates of accidents, but accident rates are lowered by redesigning the systems to reduce their interactive complexity. Airway safety is a particular success story, in which ‘experience, better designs, equipment, and procedures appeared…the unsuspected interactions were avoided, and the tight coupling reduced.’

Which type of system is policing? Reformers quoted by Kennedy suggest that policing has failed to build a culture of safety comparable to that found in commercial passenger airlines, from which not one passenger died in a US plane crash in either 2015 or 2016, despite almost 912 million passenger journeys in 2016. Yet in the long run, whether policing can become as safe as air travel may depend on how well police innovators can apply Perrow’s concepts. What follows is a first attempt to suggest detailed systemic applications of Perrow’s framework.

Interactive complexity in policing arises in many kinds of interactions, such as those between police and citizens face to face, among officers dealing with the same citizen, among officers dealing with each other, and between organizational practices and individual officers. Crucially, complexity arises whenever police face citizens who do not comply with police orders.

“There are myriad ways in which people can fail to obey police orders, which makes the next steps of those people unexpected—or at least difficult to predict with certainty. The diverse ways people can defy police authority vary on a continuum of conduct that may appear more or less threatening to the safety of police and bystanders. What also varies, as generations of systematic observation have shown, is the conduct of police officers in response to noncompliance: their discretion to use force, make arrests, call for backup, and let suspects runaway to catch them on another day.

“A graphic model might look like this:

Citizen Noncompliance ⇒ Police Response ⇒ Citizen Threatens Police ⇒ Force?

“The two variables of citizen noncompliance and officer discretion are central to the (still untested) idea of de-escalation training, giving police skills in how to calm people down when they are upset. Using such skills may be a kind of emergency brake against police anger (or fear) escalating in response to citizen anger—a dynamic many observers see in events leading up to police shootings that may have been avoidable

“What may be far more important are interactions among police officers and their formal organizations, especially those that create production pressures that make tight coupling even tighter when police encounter noncompliant individuals.

“Two kinds of production pressures are familiar to policing, both of which cause tight coupling. One is the pressure to move on: any production pressure to finish the present task to attend to the next task. The other is the pressure to contain risk: to ensure that the present situation is limited to the initial participants and does not spill over to affect more people or escalate to an injury to police themselves. Both forms of pressure create a sense of urgency in police work and work against a strategy of patient de-escalation

“Police dealing with any kind of incident are highly conscious of the need to complete their task so that another task can be addressed very soon thereafter.

The pressure to keep up production, if only to be available for a new dispatch, can create an organizational contradiction between doing the present task well and getting it over with as soon as possible.

“They hear their own counterfactual question of what else should I be doing now that could be more important than this petty job?

“The result is a dominant occupational culture theme of police work best described as urgency: a strong sense of duty to (a) finish each task in order to (b) resume readiness to provide immediate assistance elsewhere to those who need it most, wherever they may be.

“[Camden Police]… formed a cordon around a knife-wielding man: their tactic uncoupled the behavior of the man from its potential for hurting bystanders as well as from hurting police… imposing themselves as a human shield that bought time for further negotiation.

“When citizens are noncompliant, police may rightly say they do not have all day. Shooting people certainly takes less time than arresting them without injury. Yet given the potentially catastrophic costs of speed, a safer system would make it possible for police to do just that: wait all day, if that is what it takes to avoid a lethal confrontation.

“A US Park Police officer emerged, ran over to [five] other officers already dealing with the man and immediately shot the homeless man twice… This difference suggests not only production pressures but also a further system problem of excessive decentralization, in which no one is in command at the scene of a life-or-death standoff.

“Some readers, however, may challenge the claim that police shootings are rare, so let us consider whether they are rare enough for system-accident theory.

“The 2008 PPCS [Police-Public Contacts Survey of the National Criminal Victimization Survey of US residents] is the latest detailed report to provide information on the number of contacts each respondent has. Using the percent of respondents who had one versus two or more contacts leading to a mean of 1.7 contacts per person who had a police contact, we can conservatively estimate the number of contacts by multiplying 1.7 times the 40 million who had any contact at all, which results in 68 million contacts. On that basis, we can estimate the risk of any resident being fatally shot in a police encounter is 1 in 68,000. By contrast to air passenger safety, where 912 million people flew every day without any fatalities in 2015 or 2016, encountering a police officer is at least 26,000 times more dangerous than boarding an aircraft.

“From the standpoint of 18,000 police agencies, the mean would be one fatal shooting every 18 years, with many if not most agencies having never killed anyone in living memory.

“It is clear from Perrow that different systems have different rates of failure, from missile-launching (high) to plane-landing (low). The estimates we have for fatal police shootings are certainly within those ranges.

“The consequences for public trust in the police agency can also be devastating; Skogan demonstrates with survey data that one bad event can outweigh many good deeds. Even bad publicity about police conduct in another city may depress crime reporting to police in African-American neighborhoods in cities around the country.

“Seen from a broad perspective on police organizations in relation to their environments, fatal police shootings seem appropriate for analysis with a theory of rare organizational catastrophes. But other considerations remain: whether it is ever correct to call them accidents (probably not) and whether they can be better prevented by focusing on failures of complex systems rather than just on individual actions (probably).

“One illustration of how much difference police system characteristics can make is found in a comparison of two police agencies, each of them dispatching officers to deal with a reportedly armed suspect. In the following comparison, police in Cleveland killed an unarmed boy, suffering international notoriety and criticism. Police in Camden, however, averted killing a man with a knife. One of these police agencies, Camden, had adopted a well-developed system-crash prevention strategy. The other, Cleveland, had not.

“As the county prosecutor described this case, it was a perfect storm of ‘human error, mistakes and communications by all involved that day.’ In other words, it was a system crash.

“More than 15 officers did exactly what Chief J. Scott Thomson had trained them to do under what the New York Times called his ‘Hippocratic ethos of policing: minimize harm and try to save lives:’ Officers are trained to hold their fire when possible, especially when confronting people wielding knives and showing signs of mental illness, and to engage them in conversation when commands of ‘drop the knife’ don’t work. This sometimes requires backing up to a safer distance or relying on patience rather than anything on an officer’s gun belt.

“For several minutes, the officers formed a cordon around the man and walked with him for a few blocks, trying to clear traffic ahead and periodically instructing him to drop the knife. The crisis ended when the man did just that.

“The Times also reported that had the episode taken place a year before, ‘we would more than likely have deployed deadly force and moved on,’ Chief Thomson said. The chief said he had stressed to his officers that the department ‘does not treat repositioning as retreating,’ and that backing up to put a car between a suspect and an officer ‘is not an act of cowardice.’

Note that this case does not require the reengineering of an entire police agency. All it required was a focus on patience. What Camden did can arguably be attempted in any police agency any size in the United States, until research falsifies that hypothesis.

Any research agenda for saving lives must start with the question of whether police systems were even designed to save the lives of people who police shoot. The evidence suggests the systems were not so designed. Unlike commercial airline passengers, whose safety has been steadily increased by better systems designs, people who are noncompliant with police authority have not generally been seen as customers to be protected—at least in the prevailing view of police and many elected officials.

“Policing so that all lives actually matter therefore requires that police, scholars, and the public go back to the drawing board to design a system specifically aimed at placing preservation of life on an equal (and often higher) level with swift enforcement of the law.

“As Zimring points out, the number of people who die from police shootings is approximately determined by at least four successive decision points, all occurring closely in time:

  1. Whether police will shoot at all.
  2. When to stop shooting (and after how many bullets shot).
  3. What medical care police themselves will render immediately to all persons they shoot.
  4. Whether police cars will immediately transport to a hospital persons shot by police rather than waiting for an ambulance or emergency medical technician to arrive.

“The fact that most of the 18,000 US police agencies appear to have no clear policies on decision points 2, 3, and 4 speaks volumes about the lack of design for the organizational behavior affecting the fatality rate outcomes of police shootings. At minimum, it suggests that system designs for saving lives are incomplete. Most shootings do not result in death. The fastest way to reduce fatal shootings may be to increase life-saving first aid after each shooting occurs… [battlefield-grade hemostatic bandages… but officers cannot be expected to apply those bandages if the department does not issue them along with police weapons.

“Officers can save lives of severely injured gunshot-wound victims if they place wounded persons in their police cars for immediate transport to the hospital, but they may be barred from doing so unless they have been trained and authorized by a local system designed to save lives

“Officers are also unlikely to know how to decide when enough bullets have been fired unless they are trained to make that decision on clear principles.

“Those policies, in turn, must be designed not in isolation from other dimensions of a policing system but rather in full light of the coupling and interactions with all relevant dimensions—from training and supervision to dispatchers and health-care systems.

“[Some pertinent questions the author suggests that must be asked:]

Redesigning Core Functions to Generate Fewer Confrontations

 “Starting with the core-functions issues, criminologists could address the following questions:

  1. Can production pressures on police to act quickly be reduced to slow down decision-making in every citizen encounter—thus leaving time to avoid split-second decisions?
  2. What do many officers or police chiefs do to avoid shooting people despite legally sufficient provocation and justification (averted shootings), and how can we find links to other parts of the system that may inhibit their success?
  3. Do police records show some police officers to be predictably more at risk than others to shoot illegally or unnecessarily, such that evidence-based decisions could be made to remove them from street encounters with civilians?
  4. Do certain kinds of training raise the risk of avoidable shootings so that such training can be discontinued in favor of other training tools that save more lives?
  5. Can on-the-scene protocols safely divert authority to shoot civilians from officers to super-visors (as in high-speed chases)?
  6. Can organizational incentives offered to encourage delay and de-escalation (bring-them-back-alive medals) help to reduce avoidable shootings?

Designing Innovations to Reduce Deaths when Confrontations Occur

“Once police engage in a lethal confrontation, Zimring’s decision points show that there are still many opportunities to save lives. Criminologists could therefore work with police agencies to answer such questions as these:

  1. Can cease-fire protocols limit the number of bullets fired by police under clear circumstances?
  2. How many lives can be saved by police applying hemostatic bandages to civilians immediately after a shooting by police?
  3. How many lives can be saved by requiring police to drive wounded civilians in police cars to hospitals as soon as they have received first aid?
  4. Should every accidental or intentional gunshot wound caused by a police officer be subject to a peer review process located outside of the shooter’s organizational unit?
  5. Can police leaders build more community trust after shootings by various statements of regret or efforts at reconciliation?

“The central obstacle to research and development for the majority of fatal police shootings is a key state-level policy: the vast decentralization of US police agencies. Any attempt to reduce fatal police shootings would be better targeted on the agencies in smaller communities, where the majority of all fatal shootings occur and where the rates of shootings are highest. Any attempt to introduce innovations in those communities will likely require the support of state legislatures and the police training boards that most states have created.

“FN: The idea of RJCs [Restorative Justice Conferences] about fatal police shootings has a relevant precedent in Montgomery, AL, where police failed to protect a Freedom Rider civil rights group from beatings by a white mob in 1961. Five decades later, Montgomery Police Department Chief Kevin Murphy took an opportunity to apologize for this harm on behalf of his police agency. The apology was something Chief Murphy had planned to do since his first days as some Montgomery police officer decades after the mob attack. The apology was delivered on camera to Congressman John Lewis of Atlanta, who had marched with the Rev. Dr. Martin Luther King in Montgomery.

“Foremost among the ideas that could be developed is interagency cooperation within states, i.e, ways in which smaller agencies can share in specialist expertise for life-saving tactics and strategies—with or without cooperation from state police or even nearby cities over 100,000.These strategies could include first-aid and psychological training for police in treating gunshot wounds they have effected themselves, protocols for equipping police cars to transport shot persons to hospitals, and in hostage negotiations or siege situations—all of them introduced with careful impact evaluation research.

I encourage you to read Sherman’s full article and think about ways in which your police agency (either as a police or community leader) can take these recommended steps to ensure everyone, police and citizens, gets home safely at the end of the day or shift. The current problem of deadly force can be addressed and solved.

1 Comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.