The root cause hypothesis and arguments for abolition

There are different ways to make root cause arguments for police abolition, and disambiguating them is important for evaluating them. One might argue as follows:

This is a bad argument, and obviously so. The reason is that the conclusion clearly does not follow from the premises. It requires an additional, false premise: we should only address problems at their roots. Consider the argument described above in the medical context. We want to address the root causes of hypertension, but this gives us no reason at all not to address proximate causes or the symptoms of hypertension. No serious theorist of the social determinants of health is a physician abolitionist. No police abolitionist who holds capitalism to be the root cause of homelessness thinks we should abolish homeless shelters to redouble our efforts to achieve the socialist revolution.

There is a serious argument worth considering. It is my interpretation of the obsolescence argument found in the abolitionist texts cited above:

On this way of arguing, we don’t need to think that only addressing root causes is effective. To get to an abolitionist conclusion, we need to think only that making the police obsolete is feasible and preferable to making them more effective.

Of course, police abolitionists are concerned with the costs of policing, something not represented in these arguments. The observation that policing has high and unfair social costs is an important part of the police abolitionist perspective. It is represented in probably all abolitionist thought, but perhaps made most forcefully by Paul Butler (Reference Butler2016, Reference Butler2018). The reason I omit it in these reconstructions is to focus on the positive or constructive aspect of the view, which holds that we should experiment with policy to find ways of eliminating crime that do not rely on police and thus render them obsolete.Footnote ³

Premises 1 and 2 in our second argument entail a commitment to the root cause of crime hypothesis, according to which we can determine the original cause of [some crime] and eliminate it, solving [some crime problem] from its source. We can only render the police obsolete if root cause reforms succeed in eliminating the problematic behavior we use police to deter.Footnote ⁴ The hypothesis confronts two questions. One is an empirical question: which are the true root causes? There is also a conceptual question: are the causes of the problem in question uproot-able?

Perhaps we can graphically represent the problems of gun violence or public health in a tree-like manner, with several cases leading back to fewer and more fundamental causes. But this is not the same as settling the empirical question (the true graphical representation of causation in a complex system will almost certainly not be tree-like and unidirectional). Policy interventions tend not to have large and lasting impacts as measured by randomized control trials, something that plausibly results from the complexity of the social world being resistant to intervention. If this is right, then we should not think of the world in a “mechanical” fashion, and so we should reject an “engineering” view of society encouraged by root cause analysis (Stevenson Reference Stevenson2023).

Nor does graphically representing crime as a tree settle the conceptual question. It merely represents the problem in a way that suggests, given the metaphorical content, such causes can be uprooted. Difficulty on the empirical question is not necessarily a problem for abolitionists in the so-called “experimental” phase (though it will be if such difficulty leads us to reject an engineering view). Yet, the root cause hypothesis requires an affirmative answer to the conceptual question. If we cannot identify and eliminate the true, original cause of crime, then the root cause argument for abolition fails.

The empirical question has been live for at least a century. Some doubt the very idea of a science of the causes of crime: “the “causes” of crime listed in criminological literature are wise guesses as to several outstanding conditions generating crime,” (Cantor Reference Cantor1931, p. 863). Given the complexity of the social world, identifying specific root causes is daunting. Rejecting the left-wing diagnosis of root causes, some conservatives have thought we must fix the family to uproot crime (Heritage Foundation 1995). James Q. Wilson criticized the root cause approach to crime in part by arguing that were it true, the poverty reduction programs of the 1960s would have reduced crime but did not (Reference Sampson, Winship and Knight2013).Footnote ⁵ Robert Sampson and coauthors have pointed out that “the causal turn in the social sciences” has refocused the field on the causal analysis of policy and away from the root cause analysis of crime (Sampson et al. Reference Sampson, Winship and Knight2013).

The conceptual question has, I think, tended to get less attention. That is not to say it has not been raised. One possible interpretation or extension of Cantor’s skepticism is conceptual: given our inability to single out individual root causes to eliminate, the causes of crime cannot be uprooted in practice. There are more direct critiques of the root cause hypothesis available. A useful interpretation of Thomas Hobbes in Leviathan is that the causes of conflict run deep and cannot be eliminated, even in principle. To simplify the argument, bees and ants get along so well because they lack our intelligence, our sense of dignity, and thus the possibility for conflict between “social” or common interests and personal interests. They are fully integrated. These qualities are essential to humans, and thus so is the conflict they produce. Hobbes thinks that the horror of conflict in the state of nature requires us to form civil society, which for him was a central and unlimited authority. He nevertheless recognized our right to fight back against the state if we were being punished, showing that he did not think our life in civil society would be free from conflict (Hobbes Reference Hobbes and Malcolm2014, p. 22; Sreedhar Reference Sreedhar2010). The causes of conflict in human society, we can interpret Hobbes to argue, cannot be uprooted.

Some proponents of the root cause hypothesis, it is worth pointing out, appear not to think it produces an argument for police abolition. After claiming that “the party poses the task of assuring the uprooting of all violations of law and order, the elimination of crime, and the removal of all its causes. Socialist society is, as a matter of fact, in a position to resolve this problem,” Shargorodskii sounds a curious note, approvingly citing Khrushchev:

This way of thinking takes capitalism to cause crime. But it rejects the idea that capitalism is the only thing that causes crime. Achieving the socialist revolution will be insufficient for eliminating crime and conflict even if it reduces it substantially. Not all proponents of the root cause hypothesis, then, think that we can get beyond the need for active law enforcement. Yet, as we’ve seen, it is a feature of many arguments for police abolition.

On the liberal view, the criminal code and the criminal legal system more broadly exist to deter certain kinds of conflict, especially public nuisance, property crime, and physical violence. In the next three sections, taking a cue from Hobbes, I’ll offer simple models or recipes for each kind of conflict. I’ll argue that they undermine the root cause hypothesis by showing that crime has some causes are not uproot-able.

Nuisance

The recipe for nuisance is simple. We need agents in a sufficiently dense environment. Both variables or elements of the model require some specification. I will specify them in such a way as to minimize conflict to show how easily nuisances can arise. If we can get nuisance conflict in a highly idealized society, then we can get nuisance conflict (and more of it) in our own realized society and nearby alternatives.

Take our agents. They are non-tuistic, or neither perfectly altruistic nor perfectly selfish. They are homogenous, or share the same basic preferences about how to live a good life (this is the strong idealizing assumption). Importantly, they are non-integrated. That is, though they share basic preferences, they do not always want to do the same things at the same time.

These specifications are inspired by the way Hobbes thinks about conflict. Again, ants and bees do not see a difference in the common good and their personal good. Instead, “by nature enclined to their private, they procure thereby the common benefit,” (Leviathan 17.20). Ant preferences are homogenous and integrated.

Instead of ants, consider Hobbits. Hobbits are non-tuistic and have homogenous preferences, sharing a love of food, drink, pipeweed, and things that grow. As a result, they will not have conflict over things like what is permissible to eat, whether pipeweed is vicious and smelly and should be illegal, or whether a plot of land should be used for a commercial gym or agriculture. Still, it is possible that one Hobbit gets hungry and decides to cook some crispy bacon, and this interferes with the goals of another Hobbit. This can be so even though they both love crispy bacon. This is what it is for homogenous preferences to be non-integrated.

Next we need a specification of the environment’s density. If the density is very low, say one person per square mile, most activities will not affect other agents. To use some economics jargon, with sufficiently low density, most of what we do will have little to no negative externalities, or costs that accrue to those not involved in the relevant action. So if one Hobbit cooks crispy bacon, the other Hobbits won’t even notice. Yet if we crank up the density, the externalities become more serious. One Hobbit’s pipe smoke might irritate another Hobbit who does not want their young child yet exposed to smoke. Or the waste and smell of one Hobbit slaughtering pigs for crispy bacon runs off onto another Hobbit’s vegetable garden.

These kinds of conflicts are nuisances. Nuisances can arise even with homogeneous preferences, so long as we stipulate non-tuism, non-integration, and sufficient density. Further, these conflicts are exactly the ones that city governments and their general police power exist to manage. We need rules for where we can put waste, where we can graze our livestock, how much water we can take out of the water system, how much noise we can make and when, how to construct dwellings that don’t threaten citywide fires, and so on. More density generally means we need more rules.

Importantly, this model omits the kinds of disagreements and diversity that produce the more interesting conflicts in moral and political philosophy. The result is that we can produce conflict that requires some kind of social or political power to resolve with extremely minimal assumptions.

Property crime

We can generate property crime by adding scarcity to the model. In this model, the essential elements are scarcity, non-tuism, and an opportunity to maximize preference satisfaction by theft, often a result of density. Suppose that there is not enough pipeweed to go around. A Hobbit might thus decide to clear some of his potato field to plant a larger pipeweed crop. But this means fewer potatoes. It also means more farm labor. Though Hobbits love things that grow, it does not mean they want to spend all their time tending to their field. Recall that the agents in our model are non-tuistic; they are not moral angels. So another option would be to take someone else’s pipeweed when they aren’t looking and continue to enjoy one’s larger potato crop.

What would lead one to take another’s pipeweed rather than make other sacrifices like reducing one’s consumption? The answer, naturally, is that one thinks that, given all their interests and probable outcomes, taking the pipeweed produces the most satisfaction. This is, of course, to put things in a rational choice perspective. On one way of understanding the world, people are trying to maximize their preference satisfaction, and they generally behave accordingly. So, the way to keep this from happening is to make the action not worth it, or to make crime not pay.

Gary Becker defended this view, proposing that we think of the task of the criminal legal system to be “pricing” crime and responding with sanctions accordingly (Reference Becker1968). This approach is highly controversial, and there is a large literature on rational choice generally and rational choice theories of crime specifically. For my purposes, there are a few things to note.

Rational choice theory likely cannot explain all human choices. According to Loughran et al., one of the primary criticisms of rational choice approaches to crime is that it

In the present context, though, we can simply grant the objection. All we need is the idea that a significant proportion of property crime can be explained in this way. Rational choice theories of crime need not explain all kinds of crime to be good theories.

Relatedly, behavioral economists (rational choice theory’s primary opponents in economics) working on crime have taken great pains to show that standard rational choice cannot explain violent crime because violence is generally not rational. Yet some behavioral economists claim not that rational choice models are outright false, but that they can only explain a subset of crimes. Jens Ludwig, for instance, notes that “it’s not that Becker’s view is wrong; it’s that it’s incomplete. One might say incentives + predictable mistakes = behavioral economics,” (Reference Ludwig2025, p. 28). We often, but not always, make predictable mistakes.

Recent empirical research on rational choice theory indeed finds some evidence in its favor, and not just for property crimes. Loughran et al. evaluate individual panel-level data to find “general support that individuals seem to be responsive to rational choice perceptions and that this held across different types of crimes, including aggressive crimes.” They also find that “adolescents with a history of serious criminal offending act in accordance with the anticipated rewards and costs of offending and that this is as true for drug and violent offenses as it is for property crimes,” (Reference Loughran, Paternoster, Chalfin and Wilson2016). Further, rational choice theory helps to ground criminological theories like “routine activities theory” (Felson and Cohen Reference Felson and Cohen1980) and related “situational crime prevention” strategies that have often been effective (Clarke Reference Clarke1995).

An alternative explanation of property crime comes from sociology and focuses on inequality rather than scarcity: strain theory, according to which some deviant behavior can be explained by the inability to achieve culturally defined goals. The focus on material success in the United States, combined with some individuals’ inability to achieve those goals, produces property crime. On this approach, some deviant behavior, some property crime, is “normal” or “culturally oriented” in light of the social structure (Merton Reference Merton1938). Like rational choice theory, the literature on strain theory is large, and I lack space for a comprehensive discussion. One thing to note, though, is that on this conception of strain theory, deviance follows not from scarcity, but from a kind of inequality. This is easily incorporated into the simple model, and also makes the prospects of eliminating property crime even dimmer. Eliminating scarcity is insufficient. We would have to eliminate the sorts of activities that produce inequality, including zero-sum competition and positional goods.

Non-tuism, density sufficient to put us in close proximity that generates opportunities for theft, and scarcity (or inequality producing frustrated expectations) that causes some to take theft to be their best option thus seems sufficient to produce some property crime. Maybe policymakers can reduce crime, property crime in particular, by increasing (e.g.) cash transfers from the wealthy to the poor. This does not mean we can eliminate the police. As police abolitionists point out, white collar or financial crime exists (and receives insufficient attention from police). In fact, this kind of root cause strategy entails the need for an additional kind of policing: policing tax evasion. Rational choice theory does not explain crime completely, but financial crime suggests that if someone can maximize their preference satisfaction through theft, there is a risk of theft and predation.

Violent crime

To whatever extent we engage in reasoning that is properly characterized by rational choice theory, we will have the possibility of property crime. Abolitionists might then want to push back on the rationality assumption: “we’re not H. economicus,” the argument often goes. It is not obvious that this will help the case for the root cause hypothesis. The reason is that, should one reject rational choice theory in favor of a more psychologically accurate understanding where we have “bounded” rationality and make routine errors of reasoning, this can be sufficient to generate violent conflict and crime. Our recipe for violent crime requires a disagreement, and bad, irrational thinking, added to the foregoing model.

My focus here will be on gun violence. The focus on gun violence, and so the addition of guns to the model is, in a sense, a departure from the minimalism of the nuisance and property crime models. Part of the reason to focus on gun violence is because of how much there is in the US. Another is because it has been the subject of much scholarly discussion. Many police abolitionists are, to be fair, proponents of gun bans. But note two things. This is another case of a possible root cause strategy replacing one type of policing with another. Additionally, nothing here will turn on the guns in particular; knife violence is a problem in some societies, for example.

Behavioral economists distinguish “system one” and “system two” thinking. System one is “thinking fast” and system two is “thinking slow,” (Kahneman Reference Kahneman2013). To simplify, system two thinking is the domain of deliberation, and system one is the domain of instinct or quick reaction. Jens Ludwig argues that most gun violence is a result of system one thinking gone bad. This is in contrast to the received wisdom on the political Left (gun violence is rational and can be eliminated with poverty relief) and the Right (gun violence is a result of bad people rationally satisfying their bad preferences) (Ludwig Reference Ludwig2025).

Two examples will be useful. In 1996, 18-year-old Brian Willis had a dispute with 23-year-old Alexander Clair. The dispute involved a property crime: Clair sold Willis a car and Willis refused to pay. Learning this, Clair told Willis he would burn the car. After more arguing, Willis shot Clair in the stomach and head with a shotgun. Moments later, Willis shot Clair’s girlfriend, Jewel Washington, three times, killing her. Willis was sentenced to life in prison (Ludwig Reference Ludwig2025, p. 2).

Our second example involves a nuisance-like dispute between Jeremy Brown and Carlishia Hood in a busy take-out restaurant (Ludwig Reference Ludwig2025, pp. 148–52). On Brown’s understanding, there was an important informal norm at the restaurant: don’t ask for special orders (it is too busy). Hood, in Brown’s view, violated this norm by asking for modifications to her order. Brown commented, and this escalated into an argument. The argument further escalated into Brown punching Hood. Standing in the entrance to the restaurant by this time and witnessing the punch was Hood’s 14-year-old son (an honor student), carrying her legally owned gun. He shot Brown in the back. Brown left the restaurant, pursued by Hood and her son who continued to fire, killing Brown. Prosecutors filed and ultimately dropped first-degree murder charges.

What we have here are two cases of violence instigated by arguments, one following nuisance-like conflict and another property crime. This is ordinary, says Ludwig: “most homicides are arguments that end in tragedy, usually because someone has a gun,” (Reference Ludwig2025, p. 104). 70 to 80 percent of the homicides in Chicago are classified by the Chicago Police as having stemmed from an argument (119). A “bit less than 10 percent of all murders” occur in the course of robberies (112).

Why do they escalate? They escalate because of problems of system one thinking, problems like egocentric construal, fundamental attribution error, the objectivity illusion, and catastrophizing, among others (Reference Ludwig2025, p. 125–47). We are likely to construe events in such a way that we focus on ourselves, thinking that (e.g.,) we do most of the work in a group project, or think people are paying more attention to us than they actually are. When someone inconveniences us, we are likely to think it is an attack on us, rather than a simple mistake. We are likely to downplay situational factors and focus on personal factors causing another’s behavior. So not only was the inconvenience because of their lack of respect for us and not a mistake, it reflects something bad about them. We are also prone to catastrophizing, or thinking in the moment that things are worse than they actually are. The objectivity illusion involves us mistaking these quick impressions with the objective truth.

So in the Willis case, it is easy to understand the situation as one in which system one thinking caused Willis to take the dispute personally, to understand it as a personal failing of Clair, to mistake this as the hard truth instead of a quick impression, to think of losing the car and one’s respect as the worst possible outcome, and to react accordingly. Likewise for Hood, her son, and Brown. Hood’s norm violation might have been interpreted as a personal affront to Brown, indicating to Brown that Hood is a terrible person. This caused Brown to overreact, prompting an escalation from Hood’s son who in turn might have catastrophized, causing the pursuit and additional gunshots. Each one of these conclusions is a mistake produced by thinking fast. At least, this is Ludwig’s explanation.

If this explanation is right, then we need to interrupt violence and slow down thinking instead of trying to make crime “not pay.” These decisions are not rational, and so cannot be modified by changing the expected utility calculation. I won’t cover all the reasons Ludwig thinks the behavioral economics approach best explains the data, but it is worth mentioning one powerful reason. He notes that the neighborhoods of Greater Grand Crossing and South Shore have significantly different rates of gun violence (in 2023, the rates were 423.6 vs. 263.3 per 100,000, respectively). Yet the two neighborhoods are adjacent, subject to the same local, state, and federal policies, and with similar sociodemographics. Like Wilson, Ludwig argues that changes to purported root causes don’t result in changes to crime:

Instead, Ludwig argues that the learning environments in each neighborhood are different, and one’s learning environment influences their system one thinking. Some neighborhoods are gentle learning environments for social interactions, and others are unforgiving (Reference Ludwig2025, pp. 163–75). The unforgiving places have more “decision-making complexity,” (Reference Ludwig2025, p. 177).

Roots: Deep and valuable

According to the models or simple recipes for nuisance, property, and violent crime outlined above, important ingredients include non-tuistic agents with non-integrated preferences, density, scarcity or inequality (producing incentive and opportunity to maximize one’s preference satisfaction via theft), and system one thinking gone bad in the wake of disputes occasioned by the former sources of conflict. These roots, if they are roots, are deep and often valuable. They are essential to who we are, and they are responsible for much in society we want to retain. We cannot or do not want to uproot them. I’ll start with the essentiality claim and then move to the desirability claim.

In Vonnegut’s Galápagos, human intelligence is responsible for the downfall of human society and eventually the devolution of the species into lizard creatures. This devolution is presented as an improvement; our big brains actually make us “dumb” and unable to live well together. The novel’s assessment of human intelligence is clearly congenial with the Hobbesian explanation of conflict. We fight because we are smart enough to see that our personal good often conflicts with the good of others, or the good of the group.

To uproot this cause, we’d have to eliminate or substantially reduce human rationality. But that, of course, would be impossible without changing who we are, fundamentally or essentially (as in Galápagos). In this way, the root cause response to crimes of rational choice would be to change the topic. It would no longer be a way of developing policies for human society. It would be a way of eliminating humans as we are now, perhaps by turning us into lizard creatures, instead of eliminating a root cause. No one seriously recommends this. But if it is true that human rationality explains crimes of rational choice like property crime, then there is no uprooting the cause of property crime.

We can say the same for the system one thinking that can produce gun violence. We might find our departures from the rationality of rational choice theory to be disappointing and regrettable. Still, it might just be how the brain works. And rational choice crimes suggest perfect rationality is no cure either.

Human cognition has features that predictably produce conflict that societies will want to prohibit. These features can be represented or interpreted as root causes. Yet though we might be able to graphically represent the situation in a tree-like structure, distinguishing more from less fundamental causes, we cannot uproot them. They are, to retain the metaphor, too deep. But as a result, the metaphor breaks down.

Consider now the desirability of potentially uproot-able causes. Density can produce nuisance conflicts that we must manage. In most cases, we have a basic conflict of interest; one person wants quiet or clean air and another person wants to operate their noisy factory or engage in activities that produce emissions. Yet we’ve seen that even if we abstract away the philosophical disputes underlying these policy disagreements in the real world, non-integrated preferences are enough to produce nuisance conflict. To manage such conflict is to exercise the basic police power of the state. Unlike ordinary human cognition, density can clearly be uprooted. Nevertheless, we have good reason not to. Density produces conflict because our actions have negative externalities. Yet our interactions also have positive externalities. For this reason, density is a major source of human flourishing. Uprooting density would eliminate one of the most significant sources of material and cultural progress.

To emphasize: the claim is not that there is a direct, necessary link between density and crime. Crime is an extremely complex social phenomenon. New Orleans has a higher murder rate than New York City. Part of the explanation is likely to be that there are important differences in the social technologies that the two cities have developed to reduce crime. And as abolitionists are keen to point out, much of a city’s crime is dependent on things the police have no control over, like the number of community organizations in an area (Sharkey et al. Reference Sharkey, Torrats-Espinosa and Takyar2017). But more people mean more opportunities for conflict and crime.

Let’s briefly return to some major criminological theories. According to strain theory, inequality and frustrated expectations cause crime. But there is good reason to think these are difficult or even impossible to eliminate. The most general problem is that we should likely anticipate that some people will at least sometimes struggle to achieve socially or culturally desirable goals. This might sometimes have nothing to do with one’s place in the social hierarchy, as when talented athletes resort to cheating. There is the more specific problem that equalizing policies all require a theory of what it is we want to equalize. Consensus on such a theory appears not to be forthcoming.Footnote ⁶ Moreover, as Rawls argues, we might have good reason to prefer a system with inequality: inequality can redound to the benefit of the least well-off (Reference Rawls1971, pp. 75–78). But some inequality that redounds to the benefit of the least well-off might exist alongside some people having a preference for more money than they can earn through employment opportunities and the welfare state, producing some risk of property and violent crime. As strain theorists have argued, the central variable in the theory is an individual’s monetary satisfaction (cf. Agnew et al. Reference Agnew, Cullen, Burton, Evans and Dunaway1996). However we deal with the resulting crime, it is not uproot-able.

If we reject this kind of inequality, and instead favor leveling down, the problem of scarcity becomes more acute. Of course, strain theory has come under criticism, but it has been revised and extended and still has its adherents (Agnew Reference Agnew1985, Reference Agnew2012). The important point here is that even if we can eliminate scarcity, there is at least some reason to think that doing so will not necessarily eliminate crime, and that strain theory is a clear example of how some theories of crime rely on features of social structures that might be worth retaining.

Likewise, consider rational choice theory and the related routine activity theory. On this view, crime occurs when we have an offender, a target, and we lack a guardian (Felson and Cohen Reference Felson and Cohen1980). One way to think about these simple models of crime is that we have no reason to think we can modify society such that there are no motivated offenders and no vulnerable targets. In such cases, our remaining option is to hire a guardian.

Something else worth emphasizing: the claim is not that conflict can never be resolved voluntarily or without today’s professional police departments. As anarchists argue, social organization is frequently possible without a state. Support comes from Elinor Ostrom, who shows that communities can manage their common pool resources collectively. But this requires a system of (mutual) monitoring and (graduated) sanctioning that ensures obedience to the acceptable terms of resource management (Reference Ostrom1990, p. 90, 186). Ostrom also notes that the system requires “minimal recognition” by the government to ensure that it is not undermined or appropriated in the future (Reference Ostrom1990, p. 90). Such recognition allows for the social enforcement of rules; without it, sustaining the system is more difficult (Reference Ostrom1990, p. 101). Community policing is still policing, and it might require the backstop of political legitimacy.

The claim I am making is that crime issues from inevitable conflict in dense human society, even ones extraordinarily more homogenous than our own. The conflict is inevitable because its sources – density, non-tuism, non-integrated preferences, and human cognition – are here to stay or should be here to stay. Therefore, social technology, including monitoring and sanctioning, can reduce but not eliminate the possibility of escalating conflict.

Our choice is how to provide such monitoring and sanctioning. Often, we can get along with just informal control. And where we need professional enforcement, non-carceral options (like parking meter enforcement) sometimes suffice. In some cases, they might need to be carceral (Shelby Reference Shelby2022). The question is about the nature and scope of policing. In these arguments, a certain sense of efficiency, or whether policing is worth the costs, will loom large (Jones and Lim Reference Jones and Lim2025). A defense of full-blown police abolition requires more than the claim that we often don’t want police to be the tools of social control. The root cause argument, to succeed, needs elimination to be a real possibility. It appears not to be.

Two kinds of root cause explanations

If root cause arguments fail when deployed in the abolitionist debate, does that mean root cause analysis fails more generally? If so, that would appear to be a compelling reason to reject the foregoing arguments. How can something so effective in contexts like engineering and accident investigation fail in other contexts? To address this obvious concern, I want to draw a distinction between two ways of thinking about root cause analysis. In the quality engineering or accident investigation context we see a genealogical form of root cause analysis. In the social context, we see a modeling form. Focusing on how root cause analysis operates in these two different spaces highlights some further difficulties for the root cause hypothesis as deployed by police abolitionists.

When investigating an accident, or the source of problems in a manufacturing facility, it is often possible to work backwards towards a root cause. The so-called “five whys” method (developed by an engineer at Toyota) encourages us to ask of a problem, “why does this happen?,” five times. By the fifth why, we’re at least much closer to the root of the problem. This method has come under criticism, but it highlights an important point: we can often identify very specific causes of problems to eliminate. If cars are coming off the assembly line with bubbles in the paint job, it will usually be possible to trace the problem back to a more fundamental problem. Maybe the painting machines are operating at the wrong temperature. And maybe the temperature is wrong because its section of the factory is too hot. In turn, this might be because some other nearby manufacturing process produces a lot of heat. Rather than fiddling with the painting machines, the manufacturer might need to rearrange the factory. This is the genealogical approach.Footnote ⁷

The modeling approach identifies type, rather than token, problems. The models of nuisance, property crime, and violence identify general problems: our non-integrated preferences conflict, or the opportunity to steal is the way to maximize our preference satisfaction, or we thought too fast and violence broke out. The modeling approach identifies risks or possibilities for conflict, rather than the specific genealogy of a particular problem. This is true of the social scientific theories of crime discussed here as well.

These are simply different tools for analysis. The important point is that the root cause argument for abolition has aspirations to fully uproot the causes of crime to abolish the police. The genealogical approach can, in principle, make good on this kind of root cause hypothesis. The modeling approach, on the other hand, cannot. The basic reason is that, especially when it comes to complex social phenomena, many of the contributing factors cannot be eliminated. They can only be managed. Our tools for management will naturally include some form of monitoring and sanctioning.

Moreover, social complexity means that the conceptual question might not be so easily disentangled from the empirical question about root causes. Recall that the conceptual question concerns whether root causes can be uprooted, whereas the empirical question concerns which cause is the true root cause. Forget the conceptual question for a moment and focus on the empirical: what, for example, are the true root causes of gun violence? It turns out that gun violence makes it much harder to improve economic and educational opportunities. These are classic candidates for root cause explanations: capitalism or economic exploitation or low levels of government funding. Yet not only do we see substantial variations in gun violence without substantial variations in socioeconomic or political conditions (Ludwig Reference Ludwig2025), there is a plausible perspective on which the relationship goes in the other direction. We might have to fix gun violence before we can fix the other, purported root causes (Abt Reference Abt2019).

It is difficult for one to pay attention in school when one routinely witnesses or is involved in traumatic events like shootings. Exposure to a homicide reduces children’s performance on reading tests by half a standard deviation or more (Sharkey Reference Sharkey2010). There is some evidence that violent crime reduces the quality of sleep in adolescents (Chung et al. Reference Chung, Lorenzo, Chae, El-Sheikh and Yip2025). Nuisances and property and violent crime quite plausibly make it less likely for a neighborhood to develop robust economic opportunities. Reducing violence increases economic mobility, slows the rise of urban inequality and reduces concentrated poverty (Sharkey and Torrats-Espinosa Reference Sharkey and Torrats-Espinosa2017; Torrats-Espinosa and Sharkey Reference Torrats-Espinosa and Sharkey2025).

Put differently, the social complexity of the world casts some doubt on even the possibility of a non-controversial tree-like model of crime in which some causes are more fundamental than others (never mind whether they can be uprooted). This is a function of the macro nature of social explanation represented in what I’ve called modeling approaches to root cause explanations. In other words, we’re back to where Cantor was (Reference Cantor1931).