Since 2014, viral images of Black people being killed at the hands of the police—Michael Brown, Eric Garner, Breonna Taylor, and many, many others—have convinced much of the public that the American criminal legal system is broken. In the summer of 2020, nationwide protests against police racism and violence in the wake of George Floyd’s murder were, according to some analysts, the largest social movement in the history of the United States.2 Activists and academics have demanded defunding the police and reallocating the funds to substitutes or alternatives.3 And others have called for abolishing the police altogether.4 It has become common knowledge that the police do not solve serious crime, they focus far too much on petty offenses, and they are far too heavy-handed and brutal in their treatment of Americans—especially poor, Black people. This is the so-called paradox of under-protection and over-policing that has characterized American law enforcement since emancipation.5

The American criminal legal system is unjust and inefficient. But, as we argue in this essay, over-policing is not the problem. In fact, the American criminal legal system is characterized by an exceptional kind of under-policing, and a heavy reliance on long prison sentences, compared to other developed nations. In this country, roughly three people are incarcerated per police officer employed. The rest of the developed world strikes a diametrically opposite balance between these twin arms of the penal state, employing roughly three and a half times more police officers than the number of people they incarcerate. We argue that the United States has it backward. Justice and efficiency demand that we strike a balance between policing and incarceration more like that of the rest of the developed world. We call this the “First World Balance.”

We defend this idea in much more detail in a forthcoming book titled What’s Wrong with Mass Incarceration. This essay offers a preliminary sketch of some of the arguments in the book. In the spirit of conversation and debate, in this essay we err deliberately on the side of comprehensiveness rather than argumentative rigor. One of us is a social scientist, and the other is a philosopher and legal scholar. Our primary goal for this research project, and especially in this essay, is not to convince readers that we are correct—but rather to encourage a more explicit discussion of the empirical and normative bases of some pressing debates about the American criminal legal system. Even if our answers prove unsound, we hope that the combination of empirical social science and analytic moral and political philosophy we contribute can help illuminate what alternative answers to those questions might have to look like to be sound. In fact, because much of this essay (and the underlying book project) strikes a pessimistic tone, we would be quite happy to be wrong about much of what we argue here.

In the first part of this essay, we outline five comparative facts that contradict much of the prevailing way of thinking about what is distinctive about the American criminal legal system. In the second part, we draw out the normative implications of those facts and make the case for the First World Balance.

### A. Mass incarceration is not a world of mass policing

In one sense, prisons and police are complements. It would be impossible to have many people in prison without the police, since, to put people in prison, the police usually have to apprehend and arrest them first. It would also be difficult to have police without prisons, since the threat of imprisonment is one of the typical sanctions wielded by police around the world. Given this, and given the exceptionally high incarceration rate in the United States, many people assume that the United States must also have an exceptionally high number of police officers.

But that is not in fact the case. Figure 1 plots the police and incarceration rates of a sample of developed countries.6 The graph illustrates the chief fact that has animated the iterature on mass incarceration: America is a developed-world outlier in its use of incarceration. Yet it also illustrates the much less-well-known fact that America is not at all an outlier in its rate of policing. The United States has around 212 police officers for every 100,000 total residents, which ranks it in the forty-first percentile of today’s developed world.

Figure 1.

This figure shows the rate of incarceration and the rate of policing for a sample of countries in the developed world. The United States is a world outlier in the rate at which it incarcerates its population. But it is not a developed country outlier in the rate at which it employs police officers to patrol its streets.

Figure 1.

This figure shows the rate of incarceration and the rate of policing for a sample of countries in the developed world. The United States is a world outlier in the rate at which it incarcerates its population. But it is not a developed country outlier in the rate at which it employs police officers to patrol its streets.

Close modal

### B. Given its level of serious crime, America has ordinary levels of incarceration but extraordinary levels of under-policing

Yet this way of putting things in fact understates the magnitude of what has been misunderstood. Figure 1 denominates the scope of incarceration and policing by population. By that metric, the United States has an exceptionally high incarceration rate but a relatively normal number of police officers given the total size of its population. But we think it is more informative to denominate punishment and policing by the level of serious crime in a country. By doing so, it is possible to make inferences about cross-national differences in how countries manage serious crime.

Here one runs into some difficulties. For several reasons, it is challenging to compare levels of serious crime across countries. Some countries criminalize acts that are perfectly legal in others. Countries define many criminal acts, such as “assault,” differently from one another.7 And countries vary widely in their ability to measure the incidence of criminal acts. The result is that many international patterns in reported data are obviously misleading. Data collected by the United Nations Office on Drug and Crime, for instance, suggest that the rate of violent crime is higher in Belgium, France, and Canada than in El Salvador, Russia, or Rwanda.8 Our solution to this problem is to measure the rate of serious crime by the rate of homicides.

For the comparisons that anchor this piece—the United States to the developed world—this immediately raises a problem. Franklin Zimring and Gordon Hawkins have argued that “[r]ates of crime are not greatly different in the United States from those in other developed nations. … [O]ur extremely high rates of lethal violence are a … a distinct social problem.”9 If America has more lethal violence than Europe, but not more crime, the relatively high homicide rate in the United States would be a biased estimate of the rate of serious crime.

We have two kinds of reasons for thinking that this is wrong and that the homicide rate is the right (or best) measure. First, given the reliability issues that bedevil the police or victim survey data on which Zimring and Hawkins and others rely, this is an area in which one has to take some cues from theory and other data. Consider, then, the following trilemma.

1. Concentrated disadvantage is the root cause of most serious crime in developed societies.

2. America has significantly more concentrated disadvantage than European countries.

3. America has the same amount of serious crime as other developed countries.

One of these three statements must be false. Criminological theory and existing social science evidence strongly support (1).10 And we think there is good evidence to support (2).11 The main theoretical reason to believe (3) is that the United States has far more guns per capita than European countries. But while firearm availability no doubt has some impact on the level of violence, we think the effect is likely to be small. A large effect would be difficult to square with other patterns across place, persons, and time. Consider, for example, that while the United States has ten times as many guns as El Salvador, the homicide rate there is roughly ten times higher than it is here.12 And that white, richer households in the United States are much more likely to report owning a gun than Black, poorer households.13 Given this and given the reasons to believe (1) and (2), we think (3) is most likely to be the false leg of this trilemma.

The second reason—which does not depend on the first—is that homicide accounts for a large proportion of the total harm caused by crime. Insofar as standard measures of the crime rate give equal weight to each act criminalized by the state, they are conceptually meaningless. A society with a thousand petty larcenies and one murder has much less serious crime than a society with a thousand murders and one petty larceny, yet the raw crime rate would be the same in both. A meaningful measure thus has to account for the relative seriousness, or harmfulness, of each action.

It is difficult to measure how harmful different kinds of crime are with any precision, but the cost-of-crime literature furnishes a first approximation. Economists estimate the social costs of different kinds of crime by asking people how much they would be willing to pay to reduce their odds of being a victim of various offenses. Summarizing this literature, Aaron Chalfin and Justin McCary estimated that the cost of a murder is around $7,000,000, the cost of an assault less than$40,000, the cost of a robbery around $13,000, and the cost of motor vehicle theft around$6,000.14 Thus, even though homicide is much less frequent than other crimes, it is judged so much more severe that it accounts for about seventy percent of the total costs of crime. This means that it is a much better estimate of the rate of serious harm than unweighted measures of the rate of crime. Figure 2 shows the same data, but this time denominated by homicide rather than by population.

Figure 2.

This figure shows the same prisoner and police data as is shown in Figure 1, but this time denominated by the level of homicide rather than the population. America’s outlying incarceration rate looks normal, given the level of serious crime. And now, the level of policing in the United States appears exceptionally low compared to other countries.

Figure 2.

This figure shows the same prisoner and police data as is shown in Figure 1, but this time denominated by the level of homicide rather than the population. America’s outlying incarceration rate looks normal, given the level of serious crime. And now, the level of policing in the United States appears exceptionally low compared to other countries.

Close modal

One immediately sees something different. Now, America’s outlying level of incarceration looks relatively ordinary. Its prisoner/homicide ratio is a little higher than the developed-world median, but not by much. No less stark is the fact that its police/homicide ratio now appears exceedingly low. That is, if denominated by the level of serious crime, America is not normally policed but rather under-policed. America has about one-ninth the number of police officers, per homicide, than does the median developed country.15

### C. Low clearance rates in America are not driven by lack of police focus

One of the refrains of police reformers has been that American police are uniquely inefficient. Typically, when people argue that American people—and Black people, especially—have been under-protected and over-policed, they mean by this that the priorities of American police are skewed. Police focus too much on petty offenses and too little on serious crimes. This is the purpose of dwelling, for example, on the fact that only four percent of a typical police department’s time is devoted to handling violent crime.16

And indeed, it is true that in comparative context the police in the United States do not solve many serious crimes. America’s clearance rate is the lowest of all comparable countries, as Figure 3 shows.17 The median developed country records around one homicide-related arrest per homicide that occurs. In the United States, the figure is 0.56.

Figure 3.

This graph plots one measure of the clearance rate (homicide arrests/homicides) and two quantities into which it can be decomposed: the footprint of the police (police/homicides) and the rate at which any given police officer makes homicide arrests (homicide arrests/police). It shows that the clearance rate in the United States is substantially lower than the clearance rate in the median developed country and that this is a product of its exceptionally low police footprint, not the rate at which American police make homicide arrests (which is much higher than the rate in the median developed country).

Figure 3.

This graph plots one measure of the clearance rate (homicide arrests/homicides) and two quantities into which it can be decomposed: the footprint of the police (police/homicides) and the rate at which any given police officer makes homicide arrests (homicide arrests/police). It shows that the clearance rate in the United States is substantially lower than the clearance rate in the median developed country and that this is a product of its exceptionally low police footprint, not the rate at which American police make homicide arrests (which is much higher than the rate in the median developed country).

Close modal

Yet this does not seem to be, as reformers imagine, because police in the United States are exceptionally focused on nonserious offenses. Consider one measure of police focus: the number of homicide arrests made per police officer. The clearance rate (homicide arrests/homicide) is the product of police focus (homicide arrests/police) and the police footprint (police/homicide). The conventional view of policing in the United States suggests that the problem with America’s clearance rate is that footprint is high, but focus is low. In fact, as Figure 3 suggests, the converse is true: footprint is low, but focus is high.

### D. America combines low levels of certainty with high levels of severity, especially in its most disadvantaged communities

One way of summarizing much of what we have shown so far is to observe that the United States seems to emphasize the severity of punishment over the certainty of sanction. The exceedingly high prison/police ratio and the low level of police per homicide together suggest that the United States relies on long sentences rather than the sanction of arrest to control crime. One way to estimate certainty and severity more directly is to decompose the prisoner/homicide ratio into the ratio of arrests to homicide (estimating certainty) and the ratio of prisoners to arrests (estimating severity). Figure 4 plots these two ratios across the developed world. The result supports our judgment: the United States has relatively low levels of certainty but relatively high levels of severity.

Figure 4.

This graph plots measures of certainty (arrests/homicides), severity (prisoners/arrests), and their product (prisoners/homicides, or punitiveness) in the United States and the developed world, as well as for Black Americans specifically. It shows that the United States’ relatively ordinary levels of punitiveness reflect low levels of certainty and high levels of severity and that this combination is especially true for Black Americans.

Figure 4.

This graph plots measures of certainty (arrests/homicides), severity (prisoners/arrests), and their product (prisoners/homicides, or punitiveness) in the United States and the developed world, as well as for Black Americans specifically. It shows that the United States’ relatively ordinary levels of punitiveness reflect low levels of certainty and high levels of severity and that this combination is especially true for Black Americans.

Close modal

One advantage of using homicides, arrests, and prisoners to measure these two concepts is that we can say something about how certainty and severity are distributed within the United States. As Figure 4 also shows, while all Americans suffer from an exceptional balance of certainty and severity, it is Black people in the United States who are especially subject to it.

### E. Police violence may be a symptom of under-policing rather than over-policing

American police killed around 1,800 people in 2019. In the rest of the developed world, the average number of police killings is around 5 per year; the median is just 2. It seems intuitive that to reduce the level of police violence, we must reduce the footprint of the police. Yet cross-country comparisons suggest the opposite conclusion. As Figure 5 shows, there is a striking and negative cross-national correlation between the rate at which police kill civilians and the number of police officers per homicide.

Figure 5.

This graph shows a negative correlation between the log of the number of police per homicide in a country and the log of the number of civilians killed by police. Most developed countries cluster in the bottom right of the graph. The United States, however, clusters with developing countries: it is much closer to Pakistan, Iran, and Burundi than Canada, Sweden, and Germany. We take this negative correlation as supporting our expectation that a United States with more policing would be a United States with fewer rather than more police killings.

Figure 5.

This graph shows a negative correlation between the log of the number of police per homicide in a country and the log of the number of civilians killed by police. Most developed countries cluster in the bottom right of the graph. The United States, however, clusters with developing countries: it is much closer to Pakistan, Iran, and Burundi than Canada, Sweden, and Germany. We take this negative correlation as supporting our expectation that a United States with more policing would be a United States with fewer rather than more police killings.

Close modal

Countries with large numbers of police per homicide are countries in which police are much less likely to kill civilians, as compared to countries with fewer police per homicide. The countries of the developed world cluster on the bottom right of this graph (high police/homicide, low levels of police violence), while the countries of the developing world cluster toward the top left. The exception is the United States.

To be clear, a negative correlation is not proof that lower levels of police/homicide cause the police to be more violent. Several possible confounders might explain the coincidence of high levels of police killing and homicide (e.g., inequality). It is even conceivable that the relationship could run in the reverse direction (high levels of police killing cause low public demand for policing). Because police killing is rare and our data are poor, causal inference is challenging.

But there are some theoretical reasons to believe that this correlation is in fact causal. When violence overwhelms police resources, police make contact with only a small fraction of those who commit it. Under these circumstances, the civil treatment of a small fraction of offenders will have a negligible deterrent effect. Indeed, the American combination of small police footprint and brutality is reminiscent of the early modern state. As the opening pages of Foucault’s Discipline and Punish describe, when the infrastructural capacity of the state is low, exemplary but rare shows of spectacular force can be the most effective way to induce compliance with the law.18

In addition, in societies where police resources struggle to keep pace with rates of interpersonal violence and private citizens are therefore more likely to arm themselves, police officers are more likely to behave brutally out of regard for their own interests. A severe sentencing regime such as ours could well exacerbate this dynamic since, under such a regime, there are stronger incentives for suspects to take extreme measures to evade apprehension and arrest. Thus, relying on severe and lengthy sentencing, rather than policing, to deter crime could make the job of policing more dangerous. And this might in turn make interactions with the police even more dangerous—for civilians.

Of course, these inferences are speculative. Further empirical research is needed to test them. But we think the negative cross-national correlation between police killings and the number of police officers per homicide, along with the theoretical reasons to believe those correlations might be causal, should serve as a warning sign. It is not at all clear that reducing the number of police officers on the street would reduce the pervasiveness of police violence and abuse. And it suggests to us, again, that the obstacles to police reform run deep. Reformers often argue that police officers should be trained as “guardians” rather than “warriors.”19 But the training protocols that instill and entrench the warrior mindset may just be symptoms of under-policing.

The United States is ridden with much more serious crime than other comparably wealthy societies. It responds to this exceptionally high level of serious crime with an exceptional combination of relatively small police forces and comparatively long sentences. And, tellingly, this regime reaches its apogee in the way it treats most disadvantaged people. What is to be done?

The comparative observations we have made above suggest an obvious hypothesis. Perhaps the United States, like the rest of the developed world, ought to emphasize policing and penal certainty rather than incarceration and penal severity.20 Perhaps the United States ought to shift resources from incarceration to policing until the balance between the two looks more like the balance in the rest of the developed world. The implications of such a move—which we call the First World Balance—would be dramatic. The United States today has almost three times as many prisoners as police officers. If it raised no revenue but simply used the money saved by cutting prison populations to hire police officers until the ratio was the same as the ratio in the developed world (about 3.4 times as many police officers as prisoners), the new United States would have about 370,000 prisoners and 1.1 million police officers. That is, the First World Balance, if implemented in the United States, would be a society with about 1.9 million fewer prisoners and almost half a million more police officers.21 As we note later, this new United States would not be a dystopian police state. Moving to the First World Balance would in fact align the rate of policing in the United States with the rest of the developed world.

But, of course, to note that moving to the First World Balance would align the United States with other countries is not to have shown that this would be a good thing. One cannot reason to normative policy conclusions from comparative empirical observations alone. To make these arguments, one has to connect fact to value. Would the First World Balance be justified?

First, it was due to our shared interest in answering questions like this one that the two of us began to work together. Sociologists write about normatively laden questions but are taught to refrain from considering the normative implications of their arguments. It is no surprise that many of them do anyway, since it is those implications that give their vocation meaning. But social scientists’ lack of training in moral and political philosophy means that these conclusions are too often founded on ideology or intuition rather than rigorous normative argument. Some philosophers are interested in applying moral and political theory to puzzles that bear on real-world problems. But a lack of social scientific training leads many of them to seek answers to these questions (or versions of those questions) that do not require empirical inputs. Thus, our aim is to combine empirical evidence and social theory with explicit normative argument.

Second, where possible, we strive to be as ecumenical as possible. In general, we do not attempt to identify the correct theory of interpersonal morality or political justice (or the correct theory of “what causes crime”) before drawing out the specific implications of just that theory. Instead, we consider the implications of a wide range of frameworks. In general, we find that first-order disagreements do not have dramatic implications for policy, which suggests that wringing our hands about these first-order issues may not always be worth the energy it consumes. We think that the combination of empirical and normative claims one would have to endorse to reject our proposal are incongruent—they represent various mixtures of ideas that effectively nobody accepts.

Finally, our case for the First World Balance should be understood as our answer to a narrowly specified question about how the United States ought to apportion a fixed pool of penal spending. Many readers will wonder, understandably, whether we stack the deck in our favor by posing the question this narrowly. Why a fixed pool of resources? And why force a choice between prisons and police, when various kinds of social or non-penal alternatives are superior to both? We say more about why we specify the question in this way in our forthcoming book What’s Wrong with Mass Incarceration, but some explanation is in order here.

We think that in the long run, a significant expansion of social policy would reduce crime by addressing its root causes and in turn reduce the need and demand for both policing and imprisonment. In other work, we argue that any coherent conception of distributive justice or economic efficiency entails that the United States should expand social policy. But a significant expansion of social policy requires significant redistribution from rich to poor. Redistribution of this magnitude would require the poor to wield some kind of leverage over the rich.22 Given the collapse of the American labor movement and the electoral fracturing of the American working class, we doubt we will see anything like this soon. Our aim in this essay is to say something useful about what should be done in the nonideal world in which we live—not just in the ideal world in which we would like to live. To say something about that question, we limit ourselves to options that are revenue-neutral.

But why consider only prisons and police? Why couldn’t the government redistribute the existing pool of money from prisons and police to social policy, just as many reformers have demanded?23 As we argue in What’s Wrong with Mass Incarceration, this is because social policy is bedeviled by what we call the efficiency-feasibility paradox. To address the root causes of crime would be to meaningfully change the opportunity structure for the most disadvantaged people in America. To do this by expanding untargeted, universal social programs would require significant resources, since the vast majority of beneficiaries are not America’s most disadvantaged people. Because penal spending is hyper-targeted in a way that social spending is not, it costs about $300 billion a year to run the developed world’s most extensive penal state but something like$3 trillion to run its most anemic welfare state.24

Now, there is good evidence that social programs that are hyper-targeted at the truly disadvantaged—particularly early childhood interventions and high-quality preschool programs—can be efficient at reducing crime.25 But the same thing that makes these hyper-targeted social programs efficient also makes it politically infeasible for governments to fund them at scale. The more targeted the beneficiaries, the more certain we can be that introducing these programs will provoke the resentment of the near-poor and middle-class. Hence, the efficiency-feasibility paradox: untargeted social policy is politically feasible but inefficient for crime control, while hyper-targeted social policy can be efficient but is infeasible.

So, it is not feasible to address the root causes of crime with revenue-neutral changes to public policy. Yet some might argue that we should divest from policing and incarceration even if this would result in more crime.26 Harsh policing and concentrated imprisonment impose deep psychological, symbolic, political, and economic burdens on disadvantaged communities, rather than merely imposing physical injuries on individuals. Concentrated incarceration in disadvantaged neighborhoods can weaken informal social controls, sow distrust in communities, undermine civic organization, and depress political participation.27 Abusive and omnipresent policing, similarly, can entrench racial residential segregation and sow “legal cynicism.”28 And arrest records can sometimes be a debilitating disadvantage in the labor and rental housing markets, despite the fact that many of those who are arrested ultimately have their charges dismissed.29 These burdens are all disproportionately borne by the disadvantaged.

Yet the issue with citing the many negative consequences of incarceration and policing to sanction a rise in crime is simple: serious crime has those very same consequences. Rampant crime can suppress social mobility and exacerbate poverty and concentrated disadvantage at the community level.30 Living in a violent neighborhood reduces people’s incentives to do things that are important for upward mobility, social cohesion, civic organization, and physical health.31 Neighborhood violence can be traumatizing and stressful, causing lack of sleep and impeding children’s cognitive development, attention, and impulse control.32 Businesses are wary of investing in neighborhoods where crime is rife, leaving residents without access to basic retail services, including groceries and healthy food options. Property values drop, reducing funding for local public schools and other municipal services. Upwardly mobile residents of high-crime neighborhoods often move out, which exacerbates the concentration and clustering of disadvantage in those areas.33 And just as a criminal conviction or an arrest record can be a stigma on the job market, victims of crime can be stigmatized in extremely disadvantaged communities, where mutual respect can become a zero-sum game.34

Given that serious crime causes the same kinds of problems as abusive policing and concentrated incarceration, we argue that state and local governments face an inevitable and tragic trade-off between the harms of crime, punishment, and policing. The least well-off have to bear the brunt of these trade-offs, no matter how state and local governments choose to make them. The fact that trade-offs of this nature are inevitable is itself an injustice. It is an indictment of our society. But we think some ways of balancing these harms are more just and more efficient than others.

Thus, we think that the question before us is how to strike the right penal balance. What ought to be done about the level of incarceration and the level of policing in today’s United States? Where, in the 2 × 2 space plotted above in Figure 1, should the United States lie?

Consider, first, how a consequentialist might approach this question. On that view, the state should decide how to strike the balance between incarceration and policing by choosing that point in this 2 × 2 space that maximizes aggregate welfare. How should we expect different ways of striking the balance between policing and incarceration to affect people’s well-being in the aggregate?

Consider what we take to be the first-order welfare consequences of moving to any point in this space: what it would imply for the level of homicide and crime, the number of people in prison, and the number of people killed and arrested by the police. What would happen to each of these if the United States were to implement the First World Balance?

First, homicide and other kinds of serious crime would decline. The empirical literature on deterrence is unequivocal that increasing the size of police forces is a much more efficient way to prevent crime than increasing the length of prison sentences for those who are apprehended and convicted.35 The explanation for this asymmetry is well established in behavioral psychology and economics. People do not make decisions the way rational choice models imply. Rather than calculating the expected utility of the options they are choosing between, people tend to weigh the immediate consequences of their actions much more heavily than consequences in the distant future.36 It is by definition possible to add to the length of a prison sentence only on the back end, in the future. So it is to be expected that increasing the probability of arrest and conviction would do more to deter crime than increasing sentence lengths. Today in the United States, a single dollar spent on policing is almost sixteen times more effective at deterring crime than a dollar spent on incarcerating additional prisoners. Our best guess is that the First World Balance would be a world of a little more than four thousand fewer homicides (and substantially less crime more generally).37

Second, the costs to aggregate well-being imposed by mass incarceration would be substantially smaller. Whatever its general consequences, prison is extremely detrimental to the well-being of prisoners. It is difficult to put a precise number on this suffering. But suppose one thinks that a year in prison is even one-half as good as a year spent outside. On this view, a reduction of two million in the prison population would be the equivalent of saving one million years of life (roughly, fifteen thousand lives, assuming a life expectancy of about sixty-five). The smaller the exchange rate (e.g., if living for a year in prison generates only one-tenth of the utility of living for a year outside prison, rather than one-half), the larger the welfare consequences of decarceration.

Finally, consider the costs of policing. On the one hand, a world of more policing would, perhaps unsurprisingly, be a world of more arrests. Based on recent work by Chalfin, our best guess is that the First World Balance would be a world of almost 7.8 million more arrests.38 On the other hand, for the somewhat speculative reasons we gave earlier, we guess that a world of more policing would be one of less police violence (about nine hundred fewer people killed by the police).39

Consequentialists have to propose some way to weigh these consequences against each other. This is not straightforward, but however one chooses to do it, the deck seems stacked against the status quo. For the additional arrests to be reason to rule against our proposal on consequentialist grounds, the costs of these arrests to aggregate well-being must outweigh the sum of the benefits of less crime and less incarceration (and possibly fewer police killings). We think this is implausible on almost any accounting.

In fact, the first-order welfare costs of increased arrest are probably smaller (in absolute value) than any one of the benefits of the First World Balance considered in isolation. Suppose, for instance, that the average arrest is about as bad as three days in prison. (Most arrests do not last nearly this long, so this stacks the argumentative deck against our proposal.) On this assumption, 7.8 million arrests in a year is the equivalent of roughly sixty-five thousand prisoner-years (65,000 prisoner-years = 7,800,000 arrests * 3/365). This is not even five percent of the benefits of decarceration. Further, if we convert prisoner-years to lives at the rate given above, these 7.8 million arrests would be the equivalent of about five hundred lives (65,000 * 0.5 * 1/65)—substantially less than the four thousand lives we estimate would be saved by the reduction in homicide.

One might object that the indirect psychic and social costs of these almost eight million additional arrests would outweigh the benefits of the substantial reduction in serious crime we think the First World Balance would yield. Yet those who make this argument would have to give compelling reasons to think that these nth-order costs outweigh the nth-order psychic and social benefits of less crime. As we have argued above, it seems strange to single out the nth-order costs of one of these to the exclusion of the others. Serious crime, abusive policing, and concentrated incarceration seem to all have the same kinds of harmful consequences.

Thus, from a consequentialist perspective, we think that the First World Balance is justified. In What’s Wrong with Mass Incarceration, we show the robustness of our proposal to a range of alternative assumptions. The only assumption that yields a substantially different verdict requires an extreme pessimism about the effect of expanding police on crime that is out of sync with the empirical literature.

Many argue that public policy should not aim simply to maximize aggregate well-being; it must also be sensitive to how the benefits and burdens of society in general, and the criminal legal system in particular, are distributed. Consider, for example, what Parfit calls the “Priority View.”40

The Priority View: Benefiting people matters more the worse off these people are.41

One corollary of the Priority View (or “prioritarianism”) is that burdening people matters less the better off those people are.

The burdens of how governments choose to strike the balance between policing and incarceration—no matter how they choose to do so—will be disproportionately borne by the disadvantaged. Victims of crime, victims of police abuse and brutality, and those behind bars all tend to be drawn disproportionately from the ranks of the least well-off. They are disproportionately Black and disproportionately poor. Those who are most likely to be victims of the worst kinds of crime—in particular, homicide—are the same people who are most likely to be abused, brutalized, or killed by the police. And they are also those who are at the highest risk of serving time in prison.

But these groups are not identical. As Figure 6 shows, Black people are more disproportionately overrepresented among murder victims, the incarcerated, and those arrested for serious offenses than they are in the ranks of those who have been arrested for petty offenses or killed by the police in any given year. In fact, Black people seem to be underrepresented among those who report ever having been arrested in their lifetimes.42 High school dropouts are far more disproportionately overrepresented in the incarcerated population than they are among those who have been arrested in their lifetimes. They make up fifty-four percent of the former group but only fourteen percent of the latter (and ten percent of the total adult population).

Figure 6.

This figure shows the race, class, and race × class composition of different groups in the United States based on data from the Survey of Prison Inmates, the American National Election Studies, Fatal Encounters, the FBI and the Supplementary Homicide Report. It shows that the “ever arrested” are the least disadvantaged population of the different subgroups that figure in the First World Balance.

Figure 6.

This figure shows the race, class, and race × class composition of different groups in the United States based on data from the Survey of Prison Inmates, the American National Election Studies, Fatal Encounters, the FBI and the Supplementary Homicide Report. It shows that the “ever arrested” are the least disadvantaged population of the different subgroups that figure in the First World Balance.

Close modal

The Priority View thus lends further support to the case for the First World Balance. The burdens of the status quo—under which the United States leans so heavily on long prison sentences relative to policing—fall more disproportionately on Black people and the poor, and especially the Black poor, than do the benefits. Shifting to the First World Balance would seem not only to reduce the burdens associated with the penal system overall but also to increase the benefits. Because it would reduce the number of prisoners, victims of crime and homicide, those arrested for serious offenses, and those killed by the police, it would also shift the burdens from a more disproportionately disadvantaged population to the somewhat better-off.

We can thus summarize the prioritarian argument for the First World Balance as follows.

### Prioritarian Argument

AP. Burdening people matters less the better off those people are.

1. The burdens of incarceration and homicide are more disproportionately borne by the disadvantaged than the burdens of arrest and police violence.

2. Therefore, we should weigh the burdens of incarceration and homicide more heavily than the burdens of arrest and police violence.

3. The First World Balance lightens the burdens of incarceration and homicide (and possibly police violence) while increasing the burdens of arrest.

∴ The Priority View strengthens the case for the First World Balance.43

Consider what we have argued in this essay. We have noted that the United States, uniquely among developed countries, leans especially heavily on penal severity to the neglect of penal certainty. This is reflected in the fact that it has roughly three prisoners for every police officer, while every other developed country has about 3.5 police officers for every prisoner. The United States would have to reduce the incarcerated population by around 2 million people and hire half a million more police officers to bring its prisoner-to–police officer ratio in line with the rest of the world—what we called the First World Balance.

We first defended the First World Balance on consequentialist grounds. We suggested that it would be a substantially more efficient way to use the pool of resources America currently devotes to penal spending. The human, social, and economic costs of incarceration would be substantially reduced; homicide and other serious crime would decline; and police violence might also drop. The main downside of our proposal would be the costs associated with a significantly greater number of arrests. But these costs pale in comparison to the benefits.

We then showed that arrests affect a more advantaged group of people than crime or incarceration. (Victims of police violence tend to be somewhat less advantaged than the average person who is arrested but somewhat more advantaged than those who are incarcerated or victims of homicide.) From a prioritarian perspective, according to which we have stronger reasons to confer benefits on people the worse off they are, this finding simply strengthens the case for the First World Balance.

In What’s Wrong with Mass Incarceration, we offer a more wide-ranging normative case for the First World Balance than we can in this essay. We think that the First World Balance is preferable on a variety of grounds. But we do think that some forms of civil libertarianism probably rule out our proposal.

Some civil libertarians might favor radical decarceration, without any increase or perhaps even some reduction in police force size, on the grounds that state-imposed violence or harm is morally different from—and worse than—interpersonal violence committed by private individuals.44 An extreme version of this position would hold that no amount of interpersonal violence could ever justify the use of coercive force by the state. But a state completely lacking in coercive power would be unable to enforce tax law and policy and thus unable to collect revenue. Without revenue, governments could not provide public goods or a social safety net. So this extreme form of civil libertarianism is essentially a kind of political anarchism. And we doubt many are in fact committed to this brand of anarchism.

Of course, even if there is no categorical moral distinction between harm that the state itself does and harm it merely allows private citizens to do to one another, one might think that the former is worse than the latter as a matter of degree. We think there are reasons to be skeptical of even this scalar distinction between the state’s “doing” and “allowing” of harm. But we cannot defend a view about these first-order moral questions here. In What’s Wrong with Mass Incarceration, we show that one would have to weigh state harm much more heavily than interpersonal harm to reach different conclusions. In our view, this fact highlights the same tension in the more moderate civil libertarian position that we see in the anarchist view. As we argued above, when serious crime runs unchecked in poor neighborhoods, it has any number of negative nth-order consequences on political, social, cultural, and economic life. So in order to sustain the moderate civil libertarian position, one must, like the anarchist, think that the “negative liberty” to be free from state-imposed violence or coercion is more important than the promotion of equality—including democratic or political equality.

The appeal of this civil libertarian position to those on the left seems especially puzzling in light of the fact that, by international standards, the United States under the First World Balance could hardly be considered a police state. As Figure 7 shows, the police-per-capita ratio in this counterfactual United States would still be roughly equivalent to the police-per-capita ratio in today’s Spain and broadly in line with the rest of the developed world. Indeed, even after the increase in policing and the decline in homicide, there would only be about eighty police officers per homicide in the United States. This would still be the lowest police/homicide ratio in the developed world (almost one-fifth the median value). Thus, even after this dramatic shift of resources from incarceration to policing, one could make the case that it would still be relatively under-policed.

Figure 7.

This figure shows the same data we showed earlier, but now with the First World Balance included. As it illustrates, the new United States could hardly be considered a police state by these standards. The number of police officers per capita would be well within the range found in the rest of the developed world.

Figure 7.

This figure shows the same data we showed earlier, but now with the First World Balance included. As it illustrates, the new United States could hardly be considered a police state by these standards. The number of police officers per capita would be well within the range found in the rest of the developed world.

Close modal

Note also that under the First World Balance, the United States would be a society of about three hundred thousand prisoners and roughly fifteen thousand homicides. This would make it the least punitive society in the developed world; the prisoner/homicide ratio in this world would be about twenty, which, as Figure 2 shows, would be the lowest among advanced capitalist countries. So even if the First World Balance is justified, one might wonder whether it is feasible and thus worthy of the kind of normative analysis and defense we have given it.

We admit that there are significant obstacles to changing the balance that state and local governments strike between the arms of law enforcement. There are, after all, reasons that the United States has evolved its present-day penal balance. But our view is that the First World Balance is nonetheless substantially more feasible than the kinds of things that reformers tend to demand today. In the highly unequal, oligarchic America in which we live at present, calls to reallocate a fixed pool of revenue will meet with less powerful opposition than calls to tax the rich. That is why we assume it is infeasible to expect the United States to build a generous welfare state in the mold of the Scandinavian social democracies. Proposals to use hyper-targeted social policy to address the root causes of crime are similarly infeasible. As we have argued, to be efficient, a social policy intervention must meaningfully transform the opportunity structure of those most likely to commit crime. Yet an intervention that transforms the opportunities of only those in this position will upend the incentive structure of unequal societies, thus gumming up the economy and eliciting political opposition.

In any case, readers should not think of the First World Balance as an alternative to social democracy. Justice and efficiency demand that state and local governments in the United States balance the allocation of resources between the two arms of law enforcement radically differently from the way they do at present. That would be true even if this society were more equal. Indeed, we think it is telling that more equal societies do precisely what we have defended in this essay.

1

Last updated February 13, 2022. Thanks to Randall Kennedy, Tara Menon, Seanna Shiffrin, and participants at the UCLA Legal Theory Workshop for incisive comments.

2

Larry Buchanan et al., Black Lives Matter May Be the Largest Movement in U.S. History, N.Y. Times (July 3, 2020), https://www.nytimes.com/interactive/2020/07/03/us/george-floyd-protests-crowd-size.html.

3

See, e.g., Jennifer Cobbina-Dungy et al., “Defund the Police”: Perceptions Among Protesters in the 2020 March on Washington, 21 Criminology & Pub. Poly 147 (2022).

4

See, e.g., Amna Akbar, An Abolitionist Horizon for (Police) Reform, 108 Cal. L. Rev. 1781 (2020).

5

See, e.g., Randall Kennedy, Race, Crime, and the Law 19 (1997); Alexandra Natapoff, Underenforcement, 75 Fordham L. Rev. 1715 (2006).

6

Unless indicated otherwise, all data come from the History of Punishment data set maintained by John Clegg and Adaner Usmani, which collects comparative and historical data on prisons, policing, the courts, and crime from hundreds of primary and secondary sources. See John Clegg & Adaner Usmani, From Prison to Plantation (forthcoming). The majority of the data used specifically in this essay are originally from the United Nations Office on Drugs and Crime, the World Prison Brief, or Eurostat. Estimates of statistics such as the police and prisoner rate are the median of estimates from all sources reporting that statistic. To maximize coverage and minimize the effect of year-to-year fluctuations, in other countries “today” refers to the period spanning 2015 and the latest data available. By the “developed world,” we mean high-income countries with large populations. We exclude countries with small populations such as Luxembourg, inflated incomes such as Bermuda, and the oil-rich monarchies of the Middle East such as Qatar.

7

Zelia Gallo et al., Comparing Serious Violent Crime in the United States and England and Wales: Why It Matters, and How It Can Be Done, in American Exceptionalismin CrimeandPunishment 332 (Kevin R. Reitz ed., 2017).

8

See Emily Widra & Tiana Herring, Prison Policy Initiative, Statesof Incarceration: The Global Context 2021 fig.3 (Sept. 2021), https://www.prisonpolicy.org/global/2021.html.

9

Franklin E. Zimring & Gordon Hawkins, Crime Is Notthe Problem: Lethal Violencein America 3 (1999).

10

For a comprehensive survey, see Robert Lilly et al., Criminological Theory: Context and Consequences (7th ed. 2019); see also Robert J. Sampson & William Julius Wilson, Toward a Theory of Race, Crime, and Urban Inequality, in Crime and Inequality 37 (John Hagan & Ruth D. Peterson eds., 1995).

11

On America’s higher levels of disadvantage, see Lane Kenworthy, Social Democratic America (2014); Alberto Alesina & Edward Glaeser, Fighting Poverty in the US and Europe: A World of Difference (2006). On America’s spatial concentration of poverty and wealth, see Yonah Freemark et al., Varieties of Urbanism: A Comparative View of Inequality and the Dual Dimensions of Metropolitan Fragmentation, 48 Pol. & Socy 235 (2020), https://doi.org/10.1177/0032329220908966; Margaret Weir & Desmond King, Redistribution and the Politics of Spatial Inequality in America, in Who Gets What? The New Politicsof Insecurity 188 (Frances Rosenbluth & Margaret Weir eds., 2020); see also Nicola Lacey & David Soskice, Crime, Punishment and Segregation in the United States: The Paradox of Local Democracy, 17 Punishment & Socy 454 (2015), https://doi.org/10.1177/1462474515604042.

12

For data on firearm possession, see Aaron Karp, Small Arms Survey, Estimating Global Civilian-Held Firearms Numbers (June 2018), https://www.smallarmssurvey.org/resource/estimating-global-civilian-held-firearms-numbers.

13

See Tom Smith & Jaesok Son, NORC, General Social Survey Final Report: Trends in Gun Ownership in the United States, 1972–2014 (2015), https://www.norc.org/PDFs/GSS%20Reports/GSS_Trends%20in%20Gun%20Ownership_US_1972-2014.pdf.

14

Aaron Chalfin & Justin McCrary, Are U.S. Cities Underpoliced? Theory and Evidence, 100 Rev. Econ. & Stat. 167 (2017). Note that the \$7 million figure is drawn from estimates of the value of a statistical life; see also Mark A. Cohen et al., Willingness-to-Pay for Crime Control Programs, 42 Criminology 89 (2004).

15

The police/homicide ratio in the United States is about 43. The median police/homicide ratio in the developed world is 381.

16

Jeff Asher & Ben Horwitz, How Do the Police Actually Spend Their Time?, N.Y. Times (June 19, 2020), https://www.nytimes.com/2020/06/19/upshot/unrest-police-time-violent-crime.html.

17

Here we measure the clearance rate by the proportion of homicides that result in an arrest, but one could also measure this by the proportion that result in a conviction. We do not do this here because the probability of a conviction is a function not just of the nature of policing but also of the court system. But even if measured by convictions per homicide, the United States has a lower clearance rate than the median developed country.

18

Michel Foucault, Discipline and Punish: The Birth of the Prison (Alan Sheridan trans., Vintage Books 2012) (1975); see also Peter Baldwin, Command and Persuade: Crime, Law, and the State Across History (2021).

19

“Law enforcement culture should embrace a guardian—rather than a warrior—mindset to build trust and legitimacy both within agencies and with the public.” Presidents Task Force on 21st Century Policing, Final Report of the Presidents Task Force on 21st Century Policing (2015).

20

Others have proposed something similar. See, e.g., William J. Stuntz, Law and Disorder, 14 Wkly. Standard, Feb. 23, 2009, reprinted in Stuntz: Use Federal Dollars to Put More Cops on Streets, Harv. L. Today (Feb. 24, 2009), https://today.law.harvard.edu/stuntz-use-federal-dollars-to-put-more-cops-on-streets/ (last visited Mar. 14, 2022); Mark A.R. Kleiman, Toward Fewer Prisoners & Less Crime, 139 Daedalus 115 (2010); Philip J. Cook & Jens Ludwig, More Prisoners Versus More Crime Is the Wrong Question 8 (2011). We have learned from these and related arguments, but we believe no one has yet thought through all the normative and empirical issues that the proposal raises.

21

This calculation is based on the number of prisoners and police officers in the United States in 2019, which is the latest date for which we have data on all the relevant variables in this paper (arrests, homicides, prisoners, police). It accounts for the fact that the addition of police officers will increase the number of petty arrests, some fraction of which will result in additional incarceration. Note that since the number of prisoners has declined considerably between 2019 and the date of writing (as of March 2022, the prison population in the United States was around 1.7 million), the specifics are a little different today, though the principle is the same.

22

“In a capitalist democracy the probability of a development in the direction of economic democracy depends primarily on changes in distribution of power resources.” Walter Korpi, The Democratic Class Struggle 4 (1983); see also Adaner Usmani, Democracy and the Class Struggle, 124 Am. J. Socio. 664 (2018).

23

See, e.g., Keeanga-Yamahtta Taylor, We Should Still Defund the Police, New Yorker (Aug. 14, 2020), https://www.newyorker.com/news/our-columnists/defund-the-police; Michelle Alexander, America, This Is Your Chance, N.Y. Times (June 8, 2020), https://www.nytimes.com/2020/06/08/opinion/george-floyd-protests-race.html; Felicia Gomez et al., A Roadmap to Community Safety: A Guide for Local Lawmakers (2021), https://civilrightscorps.org/wp-content/uploads/2021/10/Local-Policy-Guide.pdf.

24

For a longer explanation of this point, see John J. Clegg & Adaner Usmani, The Economic Origins of Mass Incarceration, 3 Catalyst: J. Theory & Strategy 9 (2019).

25

John J. Donohue & Peter Siegelman, Allocating Resources Among Prisons and Social Programs in the Battle Against Crime, 27 J. Legal Stud. 1 (1998); James J. Heckman & Ganesh Karapakula, The Perry Preschoolers at Late Midlife: A Study in Design-Specific Inference (Nat’l Bureau of Econ. Rsch., Working Paper No. 25888, 2019), https://www.nber.org/papers/w25888.

26

“And yet this misses the point of the prison crisis: you cannot relieve the suffering of the prison population without increasing safety risks for the rest of us. And increasing those risks, from a moral standpoint, is the right thing to do.” Christopher Glazek, Raise the Crime Rate, n+1 (2012), https://www.nplusonemag.com/issue-13/politics/raise-the-crime-rate/.

27

See, e.g., Todd Clear, Imprisoning Communities: How Mass Incarceration Makes Disadvantaged Communities Worse (2007); Sandra Smith, Lone Pursuit: Distrust and Defensive Individualism Among the Black Poor (2010).

28

See, e.g., Monica Bell, Police Reform & the Dismantling of Legal Estrangement, 126 Yale L.J. 2054 (2017); Monica Bell, Anti-Segregation Policing, 95 N.Y.U. L. Rev. 650 (2020); Monica Bell, Located Institutions: Neighborhood Frames, Residential Preferences, and the Case of Policing, 125 Am. J. Socio. 917 (2020); Amy E. Lerman & Vesla M. Weaver, Arresting Citizenship: The Democratic Consequences of American Crime Control (2014).

29

See, e.g., Eisha Jain, Arrests as Regulation, 67 Stan. L. Rev. 809, 826–44 (2015).

30

See, e.g., Gerard Torrats-Espinosa & Patrick Sharkey, The Effect of Violent Crime on Economic Mobility, 102 J. Urban Econ. 22 (2017).

31

See, e.g., David Harding, Living the Drama: Community, Conflict, and Culture Among Inner-City Boys 41–42 (2010); Susan Clampet-Lundquist et al., Moving Teenagers Out of High-Risk Neighborhoods: How Girls Fare Better than Boys, 116 Am. J. Socio. 1154 (2011).

32

See, e.g., Gerard Torrats-Espinosa, Crime and Inequality in Academic Achievement Across School Districts in the United States, 57 Demography 123 (2020); Seth Gershenson & Erdal Tekin, The Effect of Community Traumatic Events on Student Achievement: Evidence from the Beltway Sniper Attacks, 13 Educ. Fin. & Poly 513 (2018); Patrick T. Sharkey, The Acute Effect of Local Homicides on Children’s Cognitive Performance, 107 Proc. Natl Acad. Scis. 11733 (2010); Patrick T. Sharkey et al., The Effect of Local Violence on Children’s Attention and Impulse Control, 102 Am. J. Pub. Health 2287 (2012); Lauren Hale et al., Perceived Neighborhood Quality, Sleep Quality, and Health Status: Evidence from the Survey of the Health of Wisconsin, 79 Soc. Sci. & Med. 16 (2013); Jennifer A. Heissel et al., Violence and Vigilance: The Acute Effects of Community Violent Crime on Sleep and Cortisol, 89 Child Dev. 323 (2018).

33

See William Julius Wilson, The Declining Significance of Race : Blacks and Changing American Institutions (2d ed. 1980); William Julius Wilson, The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy (1987).

34

Elijah Anderson, Code of the Street: Decency, Violence, and the Moral Life of the Inner City (2000).

35

Aaron Chalfin & Justin McCrary, Criminal Deterrence: A Review of the Literature, 55 J. Econ. Literature 5 (2017).

36

Amos Tversky & Daniel Kahneman, Availability: A Heuristic for Judging Frequency and Probability, 5 Cognitive Psych. 207 (1973).

37

We base these calculations on the elasticities reported by Aaron Chalfin and Justin McCary (who reported the elasticity of homicide with respect to policing at −0.67) and John Donohue (who estimated an elasticity of homicide with respect to incarceration of −0.05 to −0.15). See Chalfin & McCrary, supranote 14; John J. Donohue, Assessing the Relative Benefits of Incarceration, in Do Prisons Make Us Safer? (Steven Raphael & Michael A. Stoll eds., 2009). Because David Roodman has argued that the -0.15 estimate is too optimistic about the crime-reducing effect of incarceration at today’s levels, we choose the lower estimate from Donohue. David Roodman, The Impacts of Incarceration on Crime (Sept. 25, 2017), https://papers.ssrn.com/abstract=3635864. Given the fact that the marginal dollar today in the United States purchases a slightly larger percentage increase in policing than in incarceration (1.17 times), these elasticities imply that a dollar spent on policing is 15.7 times more efficient than a dollar spent on incarceration (−0.67/−0.05 * 1.17 = 15.7). Full calculations and code to replicate all results discussed in this section are available at https://github.com/ausmani23/miwhatswrong.

38

Aaron Chalfin et al., Police Force Size and Civilian Race, Am. Econ. Rev.: Insights 139, https://www.aeaweb.org/articles?id=10.1257/aeri.20200792. To calculate this, we take the mean of the two types of estimates they report, which implies that each additional police officer adds 15.6 arrests.

39

Note that if we are wrong and police violence were to increase linearly with the number of police officers (i.e., an increase of about 1,300 killed), the welfare costs are still substantially smaller than the welfare benefits of the decline in crime. And thus, as we note again below, our conclusions are unaffected.

40

Derek Parfit, Equality or Priority (The Lindley Lecture, University of Kansas, 1995).

41

Id. at 19.

42

The ANES is missing data on about sixteen percent of respondents. We ignore these respondents, but it would be better to impute these missing observations, since they are unlikely to be missing at random. But even if one assumes that every missing Black respondent has been arrested, the proportion of the “ever arrested” who are Black barely changes (10.6% rather than 10.3%).

43

This conclusion follows from (1)–(3), regardless of whether one thinks police violence is likely to increase linearly with police force size, because of premise (2).

44

See, e.g., Paul Butler, Chokehold: Policing Black Men 131 (2018) (writing of “a crucial difference between the violence that the police does to Black people versus the violence that African Americans do to each other”).

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits you to copy and redistribute in any medium or format, for non-commercial use only, provided that the original work is not remixed, transformed, or built upon, and that appropriate credit to the original source is given. For a full description of the license, please visit https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode.