Franklin Zimring on When Police Kill
What this covers
Criminologist Franklin Zimring discusses his 2017 book on the roughly 1,000 Americans killed annually by police, speaking with Russ Roberts on this EconTalk episode. Zimring's central claim is that approximately half of these deaths are unnecessary—occurring in situations where officers' lives are not meaningfully threatened—and that this specific fraction can be reduced by 50–75% through departmental rule changes and financial incentives rather than individual criminal prosecution. The conversation moves through the anatomy of police killings, the mechanics of what makes shootings lethal, and the policy levers that have already proven effective in reducing deaths without making policing more dangerous.
Zimring develops several empirical anchors. He notes that only firearms pose real lethal risk to police officers, yet hundreds of annual killings involve knives, bats, or no weapon at all—establishing a factual floor for unnecessary force. He cites New York City's reduction from roughly 70 police shooting deaths per year in the 1970s to seven or eight today as evidence that intensity vs lethality of policing are distinct; aggressive policing can coexist with fewer killings. He argues the core problem is not individual rogue officers but system vs individual responsibility: ambiguous departmental rules that justify lethal force in situations that do not require it. This leads to his policy prescription: clear don't shoot rules and stop shooting rules enforced through financial liability on police departments rather than prosecutions of individual officers. The discussion also addresses barriers—police unions, the absence of reliable national data, and the 100%-fault standard that defeats criminal prosecutions when blame is shared between officer and system.
Zimring argues that roughly half of the ~1,000 annual police killings in the U.S. are unnecessary—occurring when officers' lives are not actually at risk—and that this specific, soluble problem can be cut by 50-75% through clear 'don't shoot/stop shooting' rules enforced primarily through financial liability on police systems rather than individual criminal prosecution.
- Only firearms pose a real lethal threat to police, yet hundreds of killings involve knives, bats, or no weapon at all
- Reducing police shootings in cities like New York did not increase danger to officers, showing civilian lives can be saved without endangering police
- Individual criminal prosecution fails because fault is shared between officers and ambiguous departmental rules, so system-level financial sanctions are the effective lever
About half of police killings are unnecessary because they involve no lethal firearm threat, particularly non-lethal weapons and situations where danger has passed.
- Police shooting deaths divide roughly 50/50 between necessary and unnecessary killings, where 'necessary' means situations posing a meaningful risk to officer life; since only firearms create real lethal risk to officers and 97.5% of police fatalities from assault are by gun, the more than 400 killings a year provoked by non-lethal weapons (blades, bats, personal force) are unnecessary lethal force.
“The best way to divide that thousand shooting deaths is into the necessary--and then define what I mean by necessary--and the unnecessary. And, that division is almost 50/50.”
- Knives and bladed weapons were never a significant life-endangering threat to police; the infrequency of fatal knife assaults long predates Kevlar body armor, and the relevant events are typically displays ('he had a blade') rather than actual assaults.
“the infrequency of fatal knife assaults antedates the Kevlar circumstances, very substantially.”
- The central finding is a good-news/bad-news structure: only firearms are real risks to police lives (good, because it narrows what justifies lethal force), but there are vast numbers of guns—especially concealed ones officers cannot see—in the United States; therefore reducing killings to the realistically necessary level could cut civilian deaths by half to 75%, yet would still leave hundreds of deaths where a real gun threat exists.
“The good news is that only firearms are real risks to police officer lives. The bad news is there are a lot of firearms in the United States. And, it is particularly the concealed ones, the ones that police can't see which are specific and unique threats.”
- The lethality of police gunfire rises sharply with the number of wounds: a single wound from a police gun carries a 20.8% death rate, but three, four, five wounds, or fifteen shots push death into the majority of cases—motivating 'stop shooting' rules for situations where the danger has clearly passed (suspect on the ground, fleeing, or already wounded).
“one wound inflicted from a police gun, 20.8% death rate. Three wounds, four wounds, five wounds, 15 shots--and all of a sudden death occurs in the majority of cases.”
- Given 300 million people and 650,000 police, the death risk to officers from non-firearm weapon assaults is extremely tiny—it is much more dangerous for an officer to drive around the city than to face such an attack.
“in a country of 300 million people with 650,000 police is talking about a very, very tiny death risk in any attack or cumulatively over the year. It's much more dangerous to drive around the city.”
- Of roughly 1,120 deaths in police custody or interaction each year, almost exactly 1,000 are fatal shootings, making fatal shootings about 90% of the events, so the phenomenon should be analyzed as a singular, aggregate category in which the officer intends to wound with a lethal instrument.
“There are about 1,120 deaths in police custody or police interaction each year, but almost exactly 1000 of those are fatal shootings. And, so, that cluster is 90% of the events.”
Large financial damages, clear shooting rules, and administrative priority can drive departmental reform by making unjustified killings costly and visible.
- Because you cannot jail a police department, the effective lever for preventing hundreds of unnecessary deaths is big money-damage incentives that jointly sanction the system and the officer, which would drive departments to adopt simple 'don't shoot' rules (don't shoot when the weapon can't kill police) and 'stop shooting' rules (cease fire once the danger is over).
“You can't put a police department in jail. But, what you can do, is you can create a money-damage incentive that will then produce the miraculous cure for hundreds of unnecessary deaths, which are very simple rules that police departments announce and administer.”
- National policy can still provoke local reform: qualified immunity rules must change, states and the federal government must create private damage actions and effective monetary-damage measures, and departments must be prevented from treating unjustified-killing payouts as a small budget line item—because if unjustified killings become a real administrative wound, departments will fix the problem.
“You can't put a police department in jail. But, what you can do, is you can create a money-damage incentive that will then produce the miraculous cure for hundreds of unnecessary deaths, which are very simple rules that police departments announce and administer.”
- Because police officers care about promotion, salary, and ratings, once an administrative priority is set to make sense and save civilian lives, killings can fall—as in New York City where citizen deaths from police guns dropped from about 70 a year in the 1970s to roughly seven or eight, driven not just by rules but by officers knowing the priorities.
“police officers care about promotion. They care about their salaries. They care about their ratings. So, that, once there is an administrative priority to make sense and to save civilian lives, the numbers can go down.”
- The clearer the don't-shoot and stop-shooting rules are, the more shootings will plainly appear to be the individual officer's fault, because the system can no longer share the blame; clear rules thus simultaneously prevent killings and make individual responsibility assessable.
“The more those rules are clear, the greater the number of unjustified shootings where it will look like the clear fault of the individual officer or group of officers. Then the system isn't adding to the blame and sharing the blame.”
- Disciplining a rogue or over-zealous officer is difficult but possible in a union setting; it is easier to use financial and promotional sanctions and to withhold benefits than to rely on quasi-criminal disciplinary tactics, because broader, more administrative-looking controls avoid requiring a criminal conviction or a 100%-fault standard.
“it's probably easier to create financial and promotional sanctions and to withhold particular benefits than it is to use what are essentially quasi-criminal disciplinary tactics exclusively.”
Reducing police shootings does not increase officer deaths or crime, as proven by successful urban examples and lack of statistical evidence otherwise.
- Roughly 50 police officers die per year in the line of duty in the United States, a figure that has stayed low after a large decline, but is still many more than France or Germany because of the prevalence of guns.
“50 police deaths a year is still, if you want to compare it to France or Germany, many more. And, the reason is, again, all of those guns.”
- Intensity of policing is a completely different phenomenon from the lethality of policing: New York reduced its police shootings from about 70 to seven or eight a year while still doing an enormous amount of effective policing, showing you can lower killings without reducing the level or effectiveness of policing.
“it turns out that intensity of policing is a completely different phenomenon. And, it was one that was very effective in New York.”
- There is no persuasive statistical study showing that decreasing shootings by police increases the number of attacks against police; this is the place to look for a causal link because it is the only situation where the particular risk (lethal threat to officers) directly connects to the particular counter-risk.
“there has been no persuasive indication or statistical study that decreasing shootings by police increases the number of attacks against police.”
- When departments achieved large drops in police shootings—Philadelphia's recent decline and New York's 40-year reduction—policing did not become markedly more life-threatening for officers, demonstrating that civilian deaths can be cut without raising the risk to police.
“when you look at these big drops and you ask, 'Well, have police been markedly less safe under those circumstances?' Evidently not.”
- There is no clear indication of any significant or measurable crime-prevention benefit from police shootings, because the situations provoking shootings (mostly disputes and assaults against police) barely overlap with cops-and-robbers crime; reducing lethal force is therefore very unlikely to increase general crime.
“there is absolutely no clear indication of any significant or measurable crime prevention. And, it would be so unlikely because of the almost complete lack of overlap.”
- The assumptions of causality embedded in statistical attempts to tease out crime-prevention effects of police shootings are very tricky, which is why a fact-based rather than regression-based analysis is more reliable for this question.
“the assumptions of causality that have to be embedded in those kinds of statistical attempts to tease out prevention are pretty tricky.”
- The big drop in the risk of death to police occurred from the mid-1970s to roughly the late 1990s or turn of the century; what has happened since is the maintenance of that low level rather than further decline, and absent a different country (with fewer guns) further drops below 50 a year are unlikely.
“The big drop in the risk of death to police goes back from the mid-1970s to probably the late 1990s or the turn of the new century. What we have done since the dramatic lowering of police risks is maintain them at that low level.”
Police killings go largely unaddressed because decentralized departments treat each death as an isolated incident rather than summing them into visible aggregate harm.
- Police killings received little policy attention before 2014 because, unlike capital punishment which is aggregated state by state, the 18,000 separate police departments mean each police killing is treated as a single isolated news event and never summed, obscuring the fact that the state uses lethal force on a steady, collective, and very large scale (1,000-1,100 deaths a year).
“the fact that we have 18,000 different police departments and that the police department is the operational unit to focus on meant that each police killing happened as a separate news event. And, we never added them up.”
- Police use of lethal force is a very specific, very soluble problem and should not be used as the general organizing principle for taking on the much more pervasive and difficult questions of policing intensity, budgeting, harassment, and organizational change; the soluble lethal-force problem should be solved first, with broader conversations following.
“I don't think that the very important but very specific and very soluble problem of police use of lethal force is a good, general organizing principle to take on these other much more pervasive kinds of policing changes”
Domestic disputes and mental health crises drive a quarter of police killings because police are first-line responders even when no crime occurs.
- The 'dispute' category—personal conflicts, often domestic, where an individual has a weapon and uses it in a way callers perceive as threatening others or themselves—accounts by itself for a quarter of all situations that provoke killings by police, a chronic condition arising because police are the first line of intervention for problematic, potentially threatening behavior even when no offense is being committed.
“The dispute category is, by itself, a quarter of all situations that provoke killings by police.”
- While police may not excel at domestic conflict resolution, they bring a capacity to use force that is very important when the weaker party in a conflict calls for help; substituting a social worker only works if that social worker is unusually strong, so simply replacing police is not a straightforward solution.
“police bring a capacity to use force, and to respond, that is very important when the weaker of two people in a conflict--we can take the domestic assault as the typical one--wants to call for help. If you call a social worker, it had better be a pretty strong social worker.”
Individual criminal prosecution fails because departmental rules and systemic ambiguity share blame; union opposition matters less than targeting police leadership.
- Many situations—brandishing a blade at an officer, fleeing, bolting from arrest—are illegal and justify the use of force, but they do not justify killing; force and lethal force are distinct authorizations.
“there are many situations which justify force. They just don't justify killing.”
- Police unions act as a barrier to reform and are unwilling to focus any disciplinary attention on use-of-force policies; however, because there is no police chiefs' union, generating systemic pressure that targets the chief reduces the importance of union opposition.
“It is still the case that there is an unwillingness by police unions to focus disciplinary attention of any kind to use-of-force policies.”
- Criminal prosecution of individual officers rarely succeeds because in few killings is all the fault on the individual officer; departmental rules justifying deadly force are ambiguous and cover many settings where officer lives aren't at risk, so the unnecessary killings are the joint responsibility of police systems—the chief, the administrative rules, and the officer.
“very few of these killings are situations in which all of the fault is on the individual officer... if you're talking about the criminal responsibility of the individual officer, it was all his fault. That is a small minority of all the cases”