Evidence-Based Policing and Fatal Police Shootings: Promise,Problems,and Prospects

Perceived necessity, not legality, of fatal police shootings

Evidence matters. But so do emotions. The taking of any human life is, and should be, a matter of great emotional significance. Even when police lives are at risk when they shoot, the public can and will ask why that risk to police was not managed better in advance.

If we analyze the November 11, 2019, shooting by the Hong Kong traffic officer, our emotions could point us to a conclusion that the shooting was “lawful but awful,” if lawful it was. In the largely gun-free population of Hong Kong (whose murder rate is a tiny fraction of New York’s), the necessity of using a pistol to disperse a crowd blocking traffic is highly questionable. Many observers, and perhaps most Hong Kongers, would have preferred that the police not use live ammunition at all when the underlying law to be enforced is a traffic violation. They might have preferred that two lone officers not have been allowed by superiors to go up against a large crowd, preferring to delay any enforcement action until a large number of police arrived with less-than-lethal weapons to disperse the crowd—perhaps even the much-maligned tear gas that the U.S. plans to ban for export to Hong Kong (U.S. Congress 2019, section 9). Critics might suggest that allowing such an asymmetrical engagement between two officers and hundreds of protesters would create a condition in which the officers would almost certainly perceive a threat of injury—one that would justify a shooting. These thoughts, and many others, underlie a widespread perception of police brutality in Hong Kong, despite its fatality-free record in 2019.

Emotions are just as important in this case to the local police employee federation. Had the top commanders publicly questioned the traffic officer’s decision to take out his gun and shoot, there would likely have been fury at a failure to “back up” a courageous officer who put himself in harm’s way to enforce the law—and who probably perceived the shooting as absolutely necessary. So, too, might any challenge to the shooting cause negative emotions toward the national leadership of this one country with “two systems”: the Beijing leadership that repeatedly called on Hong Kong police to take “forceful action” against protesters. By November 11, 2019, protesters had injured many police officers, hitting some with metal rods, throwing flaming gasoline bombs in plastic bottles at others. Even a gun-free population can inflict a lot of damage on police officers, especially when police lack the legitimacy conferred by universal suffrage to elect the officials who oversee police actions.

Whatever the limitations of universal suffrage and the Electoral College in the United States may be, they are not a major issue in U.S. fatal police shootings. The United States has a police legitimacy problem, but it is not from having “one country with two systems.” It is, rather, a problem that stems from one country with two cultures. When the President of the United States makes a public statement that encourages police to use excessive force with gang members (Sherman 2018), he was supported by members of one culture. When police chiefs across the nation issued statements criticizing the president’s message (Rosenthal 2017), they were supported by members of another culture. Divisions of Americans along other lines, such as race and age, can also be identified in U.S. emotions about police use of force. Citizens and police on both sides of the cultural divide can become highly emotional on the question of when a shooting is necessary, versus when a confrontation can be avoided.

Reducing split-second decisions

Both Americans and Hong Kongers share the tendency to view police shootings in the framework of a zero-sum choice: if one wins, the other must lose. If officers do not shoot, they risk being injured or killed themselves. Stated this narrowly, harm to some seems to be inevitable. Even the U.S. Supreme Court, in decisions in both 1989 (Graham v. Connor, 490 US 386) and 2018 (Kisela v. Hughes, 138 S. Ct. 1148), has declined to regulate a police officer’s perception that they are in imminent danger of death or serious injury. Under these rulings, a U.S. jury would have been able to uphold the legality of the November 11, 2019, shooting in Hong Kong.

The “inevitability” logic of these rulings concerns what Chief Justice Rehnquist called “split-second decisions,” or what the authors of an ANNALS article a half century ago (Binder and Scharf 1980) called the “final frame” of a movie reel (see also Fyfe 1986). This conventional viewpoint places the policy for when to shoot in the context of the split second when an officer pulls the trigger—thereby ignoring all the contextual factors that shape (and limit) the choices of any officer who arrives at that split-second, final frame. Viewing a police encounter as a movie, we can rewind the movie to identify many previous “frames” in the reel of film, in which a different choice may have saved everyone from harm. In the case of the Hong Kong traffic officer, a decision to wait for further backup officers might have done just that. Split-second decisions are not inevitable. The need to make them can be reduced.

Legitimacy depends on more than number of fatal shootings

The United States and Hong Kong share a recognition that the police may shoot people sometimes without criminal prosecution, but never without risking a loss of legitimacy in the eyes of the public. Every time someone is killed by police, the basis for doing so can be contested. Legal principles notwithstanding, the argument that a shooting was “unnecessary” threatens the central claim of the state to a monopoly on the legitimate use of force (Weber 1946). If the police do not earn the moral right to use that force wisely, day in and day out (Bottoms and Tankebe 2012), the entire basis of the state may be undermined. It follows that even legal shootings pose a risk to both police and state legitimacy.

Yet legitimacy is hardly determined by the number of shootings alone. Even without killing anyone, the Hong Kong police suffered a rise from 5 percent to more than 50 percent disapproval in a few short months of 2019 (Lee 2019). This astonishing loss of legitimacy is widely blamed on injuries to protesters caused by police use of rubber bullets and tear gas, defined as excessive force by 69 percent of the population.

For reasons still unexplained, the number of fatal shootings by police in the United States declined by about 6 percent in 2019 compared to 2018 and compared to most preceding years (Washington Post 2020). This fact has attracted little attention in the United States. Nor has it changed the level of public concern when people are killed by police, or police killed by citizens. Just as evidence on crime trends takes a backseat to emotion-driving headlines about shocking individual cases of violence, much the same can be said about fatal police shootings. The goal of minimizing injuries caused by police may help to support police legitimacy. But achieving that goal is not enough.

In summary, the general goal of police killing fewer people is one that is widely shared across the various divisions and cultures within the United States. It is the specific circumstances in each case, tied to an individual officer who can be blamed or scapegoated for a shooting, that evoke the strongest emotions—and provide the greatest pressure for change. One way to avoid those emotions is to ask how shootings may be prevented well before the “final frame.”

Answering that question requires a far broader understanding of the systems that determine the risk of fatal shootings than just the facts of each case. Identifying the full range of those systems, from police training to the emergency medical transport that takes wounded persons to hospitals, is hard enough (Sherman 2018); but understanding how they work, and how they can be improved to save lives, is even harder. The analysis needed to do that cannot be done by busy police executives who must manage many issues at all times. The reflective inquiry that is required must be undertaken by a range of talents, disciplines, and professions, working together to ask not just how each death might be prevented, but how patterns of death might be predicted and interrupted. This is exactly the kind of broad inquiry to which evidence-based policing and this volume are committed.

Patterns of fatal police shootings

David Klinger’s article is the first and most comprehensive in the volume. Its aim is to map out the complex systems and contexts that shape each police encounter, long before and far beyond the “final frame.” First and foremost is the organizational adaptation to a very high risk of the citizens killing police in any encounter. The uniquely American problem (among advanced economies) is that U.S. police officers are killed in the line of duty at a high rate. As Zimring (2017, 79) points out, from 2008 to 2012, U.S. police were thirty-five times more likely than German police to be killed in the line of duty. The difference in rates of killings by police between Germany and the United States was almost identical to the difference in killings of police. It could be proposed, then, that the internationally high rate at which police kill civilians in the United States is a rational adaptation to the internationally high rates of police dying in the line of duty. But the test of that proposition is whether there are alternative policies that the United States could adopt to reduce that massive difference.

Klinger goes far beyond that starting point of elevated danger to U.S. police. In his adaptation of Charles Perrow’s Normal Accidents (1984), Klinger applies to policing a conceptual framework designed for high-risk, high-complexity organizations. Perrow’s theory of “normal” accidents was developed from his detailed analysis of a 1979 nuclear power plant meltdown, followed by further analyses of aircraft, shipping, and other crashes of highly complex systems. His key idea is that “tightly coupled” systems with little slack are unable to absorb unexpected shocks, which makes them prone to a cascading spiral of system breakdowns resulting in complete failure of the system. When a citizen encounter occurs in which a police officer is killed, does kill, or both, there is often a good answer to “why and how did this happen?” The answer lies in the “tight coupling” of police practice.

One key feature of tight coupling that Klinger discusses is the sense of urgency police officers feel when they engage in citizen encounters. Often prompted from their radio dispatchers to go on to the next request for police service, officers may feel they do not have time for lengthy negotiations with, or even listening to, some noncompliant citizens they encounter. This need to meet “production demands” may compel them to push sensitive discussions with emotionally upset people faster than is wise—with a high risk of that quest for speed backfiring, and guns firing, in any struggle for immediate compliance with police orders. That is exactly what happened in Hong Kong on November 11, 2019: unable to wait for backup, nor expected to do so, the “systems” forced the officer into a situation where he decided to pull out his gun and shoot to protect himself. A system with less intensive “production demands”—one that would have required that he await a larger, team response—could have prevented that situation from arising, and the shooting that ended it.

As Klinger notes, the emotional resistance to delayed policing is very high in the United States. The delay-until-backup is an alternative system that might be justified in the United States with a large crowd breaking traffic laws—but not with a “shots fired” incident. When violence is in progress, delay looks like cowardice, as indicated by the disgrace heaped on the police officer assigned to the Stoneman Douglas High School in Parkland Florida in 2018 when it suffered a mass murder of seventeen people (Blinder and Mazzei 2018). While the officer delayed, people died. Yet mass murder is not a typical police situation. The culture of urgency in all tasks of American policing impedes strategies for delayed response. Urgency is overestimated, “just in case” the situation may be very dangerous. The paradox of this culture is that the failure to wait for a sufficient number of police may contribute to raising the danger level of the encounter—for the officer as well as the citizens.

Klinger’s remedies are eminently sensible, but they face substantial challenges. His own training with the LAPD SWAT team suggests many alternative systems that could save lives. Selling, or even designing, those systems may only be possible if they are clothed in the rhetoric of courage, with the bravery and skill of a SWAT team as a model. Klinger’s ideas are unlikely to sell themselves. But with a separate research agenda on how to persuade police agencies to design new systems, the fundamental rationality of Klinger’s theory could still win out.

Daniel Nagin’s article provides an essential analysis of the U.S. context for the redesign of policing systems. In sharp contrast to Hong Kong police killing no protesters in 2019, a gun-free city, U.S. police face a far more dangerous population. One indicator of that is how few suicides in Hong Kong are committed with guns, which are not even reported among all methods of suicides (Yip, Law, and Law 2003). Daniel Nagin’s article in this volume uses the high, but varying, percentage of suicides committed with guns to indicate “gun density” across the fifty U.S. states, which has been validated as a predictor of the proportion of households with guns. He finds a substantial relationship between gun density and the rate per million people at which police shoot people fatally in each state. The greater the proportion of homes with guns, the greater the proportion of people who police shoot to death.

Nagin shows that this finding is consistent with what is called “statistical prediction” theory. That theory claims, when applied to this problem, that with more guns in a state, there is a greater chance that each person the police encounter may be carrying a gun. In the application of a zero-sum choice, Nagin suggests, the police may shoot more readily to preempt being shot themselves. Whether police should do that is an emotionally contestable question. Whether there is any way to prevent it, short of reducing the prevalence of gun possession, is an empirical question. Here again, there may be alternative systems of procedures, or even training and review, that could change the link Nagin finds between gun density and fatal police shooting rates. But a long program of trial and error testing would be needed to discover what works.

Nagin’s article is also the first to report on the effects of something that can be changed by simply spending more money—and recruiting more doctors. That is, he finds that the more access to “trauma centers” there is, on average, within in a state, the lower the statewide death rate is from fatal police shootings. Trauma centers are enhanced versions of Emergency Medicine departments in major hospitals. They employ medical and support staff with much greater experience in treating gunshot wounds and with better equipment and special training. As Susan Jacoby and her colleagues report in their article in this volume, statewide average distances of homes from trauma centers has a strong correlation with higher statewide survival rates among people wounded by gunshots.

Nagin’s finding, as he notes, is a correlation without proof of causation. States with more trauma centers may, for example, have more cautious police officers, or better police systems to minimize the perceived need to shoot to protect police. Yet the trauma center access correlation is powerful, and the theory is consistent with other data. If even one state that is a “trauma desert”—with no access to a trauma center anywhere in the state—were to create one, it would constitute an initial experiment in reducing lives lost to police shootings.

The problems of funding and staffing such centers in rural areas are of course substantial. Yet the idea would fall on common ground between propolice and antipolice constituencies. Everyone, including police officers shot in the line of duty, would stand to gain from “watering the trauma deserts.” For people shot hundreds of miles from a trauma center, the impact may be limited, but helicopter ambulances and first aid using hemostatic bandages could enhance survival even across large distances from a trauma center. Nagin’s data are clear about the correlation. It is now up to public health advocates to use the analysis to strengthen the case for better treatment of bullet wounds.

Despite this volume’s emphasis on systems and context, there is a widespread concern about the kinds of people who wear badges and carry a gun. The attribution of blame, incompetence, racism, and other moral defects is more often focused on persons than on systems. But fatal police shootings are rare events, with little known about how different officers would deal with such situations—the same, or differently? Faced with the same facts, what percentage of officers would shoot, and what percentage would not? Most important, who are the people who can avoid shooting without getting shot, or without other harm resulting?

The research methods needed to answer these crucial questions require holding everything constant except the officers. Any study that compares officers across different situations is confounded by those differences. To study officers faced with exactly the same facts, Greg Ridgeway identified a unique dataset in the New York City Police. It offered detailed information on 291 officers present at a total of 106 shooting incidents, each incident involving multiple officers, in 2004 through 2006. In each of these incidents, at least one, but not all, of the officers at the scene discharged their firearms. Analysis of these incidents can identify the correlates of officers who shoot—or do not—in exactly the same situations as officers who make the opposite choice. The differences in characteristics of those officers provide the best possible research design for predicting which officers will be less likely to shoot—for better or worse.

The most striking finding in his analysis is the role of prior arrests made by the officers. Officers who had made ten or more misdemeanor arrests were 80 percent less likely to shoot than officers who had not made ten such arrests. But for felony arrests, the opposite was true: officers with 9.3 felony arrests were twice as likely to shoot as officers who had not made as many as 9.3 felony arrests. While the latter finding was a less certain (nonsignificant) finding, it suggests a very different orientation or work history between the two groups. Those with frequent experience in making “quality of life” arrests, possibly with persons suffering from mental illness, may be used to noncompliance and managing people through words and physical restraint. Those who make many felony arrests may expect to find weapons on the arrestees, giving them greater tendencies to make “statistical predictions” in Nagin’s terms. The latter may shoot preemptively, for safety, in situations in which the misdemeanor arresters do not shoot.

Two other powerful and certain predictors in Ridgeway’s New York study were (1) the officers who had repeated misconduct reports: these officers were three times more likely to shoot than those who did not; and (2) African American officers, who were also three times more likely to shoot than white officers. The prediction of more shootings by officers with more misconduct is one that could be dealt with by removing such officers from front-line assignments and removing their firearms. That policy was widely practiced in the New York City Police Department (NYPD) in the 1970s for alcoholic police officers, who performed such tasks as courier service or reception desk duty. But the prediction of more shootings by black officers is one that has no obvious application, except perhaps more research to understand it.

In this volume, Ridgeway shows that the findings for how many bullets are fired follow similar patterns to those who shoot or do not shoot, at least in the NYPD. But in a larger national sample of (only) officers who shoot, comparing how many bullets they fired, Ridgeway could not find the same predictors as he found in the NYPD. What that may mean is a need for this kind of study to be done in each large police agency and that all such studies should include data on officers present who did not shoot.

The article by Phillip Atiba Goff and Hilary Rau provides a more theoretically intensive approach to studying police use of force. Framing the question of individual officer differences in relation to “bad” policing, Goff and Rau include racism and excessive force in the target for prediction and explanation. Reviewing a wide range of laboratory experiments and field studies, Goff and Rau weave together various strains of situationist and identity psychology. Their review suggests that there may be much more powerful selection tools for appointing police officers than multiple choice tests and background investigations.

The tools that could be used to screen applicants for positions would be the kind of simulator technologies that are widely available for police training. These technologies generally create some kind of virtual reality experience for the officer, who is armed with a fake gun that can register when the officer would shoot (or not) in relation to a screening of a situation in which officers are treated as if they are present. These tools allow the trainee to be tested on whether they shoot too much, too little, or mostly right, in relation to risk. But in the research Goff and Rau review, the officers are provoked in various ways, to test how well they can control their temper.

The important point about this method is that training itself may do little to change (or improve) the personal conduct of any one officer. What could be more useful is to use such scenarios to screen out applicants whose reactions constitute “bad policing.” As in many employment situations, it is much more effective to screen out people who are unsuited to the job than it is to try to increase their suitability once they are hired.

Perhaps the most surprising findings of Goff and Rau’s review concern the “identity” theories being tested. One theory is that men react badly when their masculine identity is threatened and are more likely to pull the trigger than officers who are less threatened by aspersions to their masculinity. Predicting which men are more insecure about that identity may be something that the tests Goff and Rau review can accomplish. With long-term research to support the predictive value of the tests, they might withstand the inevitable court challenges brought by people excluded from police jobs on the basis of such evidence.

The more surprising finding is that “officers who are concerned about appearing racist are more likely to use force in racially disparate ways,” as Goff and Rau report. Their evidence comes from surveying officers before they are tested with the simulators, so that their identity as people who are not racist is clearly established. If they both identify with being unbiased, and are insecure about whether they will be perceived that way, that condition seems to make them more prone to bad behavior. “In one 2012 study of officers from a large urban police department,” Goff and Rau write, “the Center for Policing Equity found a correlation between stereotype threat and greater use of force against black civilians.”

Here again, the problem of using this approach in training is that it may be too late after they have already been hired. Using it in hiring would pose legal challenges. But given the core importance of the personality traits needed to use deadly force with restraint, Goff and Rau’s approach seems to be well worth exploring as a hiring technique. As long as there is a sufficient number of applicants for police jobs, and adequate pay to attract them, this approach may prove less problematic than the current hiring practices at reducing unnecessary shootings.

Once an officer accumulates a higher-than-usual number of complaints from the public, they may become prone to being transferred, or “shuffled,” as Linda Zhao and Andrew Papachristos put it. Yet such “shuffling” of high-complaint officers may just make matters worse, at least in terms of their risk of shooting people—and of shootings by other officers in their social networks. That conclusion emerges from the Zhao and Papachristos article, which examines shootings of citizens between 2004 and 2016 by all 9,210 of the unique Chicago Police Department (CPD) police officers subject to one or more complaints, who collectively were the subjects of 38,442 complaints from 2000 and 2003.

By examining the links among officers who were identified in relation to the same incidents, the article counts the number of other officers to whom each “deviant” (complained-about) officer is linked, and by how many degrees of separation. The more links, and the fewer the average degrees of separation, the more likely the officer is to become part of a small “power few” of highly “central” people in these networks. Using standard terminology in social network analysis, Zhao and Papachristos call these power few officers “brokers”—because they are potential conduits for information, values, and reputations between different social networks.

As the authors put it, “Our findings show that although only a small percentage of all CPD officers shoot—those officers who do shoot appear to occupy a unique structural position, that of a network broker. This finding appears to hold even when considering individual-level factors such as age, race, gender, and activity. Our estimates suggest that each standard deviation increase in betweenness centrality predicts a roughly 1.17 higher odds of shooting net individual factors. The remaining association is strongly related to organizational-level factors such as shuffling.”

The movement of high-complaint officers to other units may have two effects on the risk of their shooting. One is that these “brokers” themselves become more likely to shoot civilians. The other is that the brokers may “spread” their propensity to shoot people, so that transfers give them an opportunity to spread proshooting views among other officers who had not met them before. As the authors note, the correlational nature of the study limits their ability to conclude that transfers actually cause more shootings. But that is not, as they suggest, any reason to ignore social network analysis in making transfers, especially from one district to another. Social network analysis offers great prospects for improving police management, but only if the insights of the results can be specified and clearly linked to choices of optimal units to which high-complaint officers could be transferred.

Policy-making and fatal police shootings

While the first section of the volume reveals the complexity of the patterns of fatal police shootings, it merely hints at the difficulty of finding better policies to govern those patterns. Starting the second section of the volume at the highest level of governance, Franklin Zimring spells out the reluctance of all levels of government to take effective action that could save lives of both police and civilians. This problem is tied to the unique structure of American government, in which the federal government has scant obligation to govern policing, the states make scant use of their power to do so, and juries make scant use of often-strong evidence to convict police officers of criminal charges.

Zimring begins with the context of high murder rates of police by U.S. residents, but then shows how many more fatal police shootings are unrelated to the actual presence of guns. While Daniel Nagin’s paper provides a statistical explanation for that in the evidence for a higher probability of guns being present in some states compared to others, that fact does not justify the deaths in some 40 percent of cases in which guns are not in the possession of the people police shoot. These are the arguably unnecessary fatal shootings against such risks as knives, in the kind of encounters faced by police in other countries (like the UK) that kill almost no one in such circumstances. ²

Zimring quickly dismisses prosecution as an effective means of reducing unnecessary shootings, making the case for administrative reforms in policing. But his second theme sadly shows the limits of the first: the federal dispersion of powers makes administrative reforms unlikely to happen by systematic governance at the state level, where the power lies to create and destroy police agencies. With such rare exceptions as Camden, New Jersey—where the state abolished a failing police department and replaced it with a county-wide agency—the state legislatures have been reluctant to take decisive action to improve governance. And while some might say that federal oversight of local police reached an apex under President Obama, the rapid swings in federal governance make that level of intervention short lived.

Most readers will readily support Zimring’s proposals for a more aggressive federal program of research and statistics on police use of deadly force, with a modest cost of $10 million per annum. Perhaps some policing organizations would even campaign for such an initiative, to save police lives. If the ever-changing gridlock of political parties in Washington after 2020 would allow it, the idea could actually become reality. Yet as former Assistant Attorney General Laurie Robinson observes in her article in this volume, the chances are still very low.

On a much smaller scale than national governance, Wolfe and his colleagues offer new evidence for hope in de-escalating risks of violence in police-citizen encounters. While “de-escalation training” recommendations from the Obama Task Force (President’s Task Force on 21st Century Policing 2015) have been adopted in eighteen states (Sherman 2018), there is no consistent definition of either “de-escalation” or its training. Neither are there any impact evaluations of such training linked to police behavior in real-world settings.

Nonetheless, preliminary findings from a training experiment in Tucson (AZ) and Fayetteville (NC) suggest that a form of training called “social interaction skills” can help to reduce police use of force. Like so many policies, the devil may be in the details. Their careful documentation in this volume of how the program was implemented reveals much about the challenges of delivering any training in police agencies. Bluntly put, training is seen as a threat to operational performance of the police mission. To the extent that availability of police officers for operational work is reduced by time spent in training, abstracting officers for such courses may be seen as a threat to the safety of officers remaining on duty. Yet that is only a starting point.

What Wolfe and his colleagues show is a critical issue in all training: results may depend on the personality and skill of the trainer. In their words, “We conclude that the conceptual framework was generally well-received, but that results depended heavily on the selection and performance of each agency’s own trainers.” This conclusion provides evidence that training in policing is hardly as homogeneous a treatment as taking a pharmaceutical pill. States may require training with a certain label, but the effect of training on saving lives can be quickly undermined by a loose regulatory framework for delivering the training. Absent a state-level inspector-general of police to audit the delivery of such quality assurance strategies (Sherman 2015), the states may be unable to produce the effects they can agree should be sought by legislation.

Robin Engel and her colleagues carry us further into the organizational details of trying to reduce unnecessary use of force. Even in the wake of a nationally criticized fatal police shooting, the University of Cincinnati struggled mightily to find evidence-based guidance for reform. In her unprecedented appointment as an eminent police scholar leading a major university’s policing, Engel provides a case study of looking for both ideas and evidence about those ideas. Ideas were far more plentiful than the evidence supporting them, as she recounts. The solution of generating new evidence was simply too slow for meeting the demand. The legitimacy of the police agency depended on taking action, regardless of whether such actions would work. Indifference is a worse sin than ineffectiveness. That, in an epigram, is what holds back generating new evidence when innovations are applied to fatal police shootings.

Engel’s tenure in governance of a large police agency faced calls for five prominent reforms: body-worn cameras, de-escalation training, implicit bias training, early intervention systems, and civilian oversight. As the article observes, “These heavily endorsed interventions, however, are not based on a strong body of empirical evidence and may result in unintended consequences.” Her efforts to decide whether to adopt these policies were hampered by the low statistical power that comes with rare events. Because her time was short, but time between shootings could last years, she was unable to test the ideas herself during her time in office. Many other police executives could say the same.

The time-to-evidence issue supports the conclusion that Engel and her colleagues reach: that university-based researchers should undertake long-term tests of policies to reduce fatal shootings. Even if these tests take years, the tenure of a university professor is very likely to outlast that of an agency police chief. In the end, the article suggests, only long-term scholarship can take the time to generate strong evidence on (thankfully) rare events.

One answer to the call by Engel and colleagues is the Pollack and Humphreys article. In a catalogue of policies for improving police encounters with persons with psychiatric or substance use disorders (PWPSUD), the evidence ranges from substantial to modest. Overall, their article suggests that the interventions reviewed might be the most rapid means of reducing fatalities nationwide. As Zimring (2017) points out, a high proportion of people shot fatally by police are mentally ill, if only transiently. Yet this appears to be one dimension of the problem that offers ample evidence for effective prevention of unnecessary shootings.

The Pollack and Humphreys review covers three categories of policies: (1) those intended to prevent violent confrontations between police and PWPSUD; (2) “in the-heat-of-the-moment” strategies, that is, how police can handle potentially dangerous situations involving people who are impaired in some way; and (3) long-term management strategies for PWPSUD who have been involved in the criminal justice system.

Most of these policies require coordinated action at various levels of operation. For police to take guns away from PWPSUD, for example, requires a state law to allow it, as well as local priorities to undertake it. The fact that the most aggressive such law in the United States—Indiana’s—followed the death of a police officer suggests the widespread benefit of preventing violent confrontations in this way. They cite further evidence of lives saved from other states as well.

The heat-of-the-moment strategies connect to the evaluation described by the article in this volume by Wolfe and his colleagues. Strategies for keeping police out of harm’s way while PWPSUD become less dangerous can be promoted and taught. But as the Camden, NJ, example Sherman (2018) discussed illustrates, the effective implementation of such strategies may depend on a police chief who is willing to terminate officers who reject such delay tactics.

The long-term management of PWPSUD offenders who are known to authorities may have the greatest potential of the three categories of innovations. Inexpensive tracking technologies for alcohol or drug use could trigger immediate interventions that would neutralize risks as soon as they begin to elevate. Alcohol use, in particular, seems subject to judicial oversight and mandatory monitoring for detection of violations, followed by immediate incarceration. The promising evidence for this approach, as reviewed in the Pollack and Humphreys article, points to much harm reduction beyond fatal police shootings; the policy may prevent accidents and domestic abuse at the highest proportions. Yet the publicity around any fatal police shooting may create windows of opportunity to introduce more expensive, and perhaps effective, day-to-day offender management.

Preventing avoidable fatalities

The topic of preventive policies in general leads to a highly focused, yet widely ignored, dimension of police fatal shootings. That dimension comes out of Zimring’s framework of policies, starting at the end: how can more deaths be prevented even after people have been shot? Our invitation to Jacoby and her colleagues to start that discussion leads them to a cautious, if powerful, conclusion: “Prehospital transport known as ‘scoop and run’ . . . may shorten the time between injury and hospital care when emergency medical services are delayed or unavailable. Current evidence suggests that when comparing police hospital transport and emergency medical services transport, survival rates are at least equivalent, and in some studies, police-transported survival rates are better, for victims of violence.”

The idea of scoop and run was pioneered in Philadelphia, a city with multiple trauma centers, in which police have routinely driven shooting victims to emergency rooms in police cars for nearly two decades. The authors cannot find evidence of any other police agency adopting such a policy. They also point out that the life-saving benefits in Philadelphia may depend heavily on the proximity to trauma centers, or even standard emergency rooms. As Nagin showed in his article in this volume, proximity to emergency medicine is a predictor (if not definitely a cause) of higher rates of death from police shootings. They note that “Pennsylvania, for example, has approximately 2.04 Level I and II trauma centers per million people and 99 percent of its population has access to a trauma center within an hour. In comparison, Arkansas has only recently received its first trauma center—just one in the entire state.” Yet the policy has a clear theoretical basis in the biochemistry of blood loss.

Trying to persuade police officers to partake in scoop and run is a challenging task, as I personally experienced in Trinidad & Tobago. With more than three hundred murders a year, police were reluctant to risk disease, lawsuits, and bloodstains on the backseats of their cars. Despite urging by their Police Commissioner in 2014 to 2016, police in Trinidad & Tobago made very little use of either scoop and run or hemostatic bandages, which can substantially slow blood loss. When it came to people they had shot themselves, police were particularly morally reluctant to intervene. Some saw intervention in the deaths as a disruption of justice, especially where the wounded had shot at police officers. Similar dynamics may arise in U.S. cities.

The challenge of reducing the conflicts between police and community are also intensified after a fatal police shooting. Many police agencies refuse to issue any statements at all about the deaths that take place as a result of their officers. Some issue statements, but without an apology, or an apology without an acknowledgement of responsibility. The question of what kinds of statements can best promote reconciliation is an urgent one. It is especially urgent when African Americans are shot by white officers.

Testing alternative ways to address this challenge is difficult to do in field settings. O’Brien and colleagues offer the creative solution of randomly assigning written scenarios to hundreds of anonymous respondents on an internet crowd-source site. The respondents are surveyed for their reactions, and standard measures of police legitimacy are deployed. The method seems like a highly cost-effective way to test a range of statements police could make in the aftermath of fatal police shootings. It is the kind of approach that could benefit from multiple replications, as well as trial-and-error variations.

It is for the latter reason that the specific findings in this article are not nearly as important as the demonstration of the research method. If a program of research across tens or hundreds of such experiments could lead to a systematic review of what approaches work best, police executives could be much more evidence-based in how they respond to arguably unnecessary shootings.

Reflections

Robinson’s commentary addresses all the articles in the volume except the special editor’s recommendations. Her comments offer candid assessments of a wide range of problems and prospects for the best use of the evidence in this volume. They draw on a great breadth and depth of political experience, especially in bridging research and policy. She remains cautiously optimistic that progress can be made on fatal police shootings, despite the enormous barriers to change.

Robinson’s article identifies four areas of promise. One is the centrality of building trust between police and community and rebuilding it in the aftermath of fatal shootings. While evidence on how to do this best is limited, the substantial research on community policing offers some direction. A second area of promise that she sees is training, especially of the kind reported by Wolfe and colleagues. A third area is, more generally, operational police officers’ buy-in for procedures and systems that would make everyone safer. The fourth is building better bridges between policing and academic institutions, which may in some sense be the most difficult. Starting with the very language used to present the research in much of this volume, there is a problem of communication between the two sectors.

The most important message she conveys is the need for scholars of policing to make their research more accessible to the general public, including police. Meaningful and helpful illumination of the dynamics of police shootings puts scholars under substantial pressure to use complex statistics and creative, sometimes complicated, research methods. In using the tools of experts, scholars may limit their audience to people who can take no action. So the research must also speak to those who can act.

Like Robin Engel in this volume, I offer an appeal to academics and “pracademics” of police practice: write a plain language summary of all the research that has implications for policy-makers. Spell out the policy implications, rather than pointing in a general direction. The true promise of evidence-based practice is that it focuses on concrete, applied solutions. By doing so, it can make research better able to serve democracy, even in a world roiled by “irrationality.” Testing ways of saving lives can add much clarity, and rationality, to decision-making at all levels of governance. It is, after all, irrational to think that better evidence should win out, just because it is scientifically better than the alternative. Like police legitimacy, better evidence must sell itself in a public dialogue, every day, even with setbacks. The audience is ready and waiting. We need only reach out to it, with clarity and humility.