Why is there more crime in larger cities?

This week, we are moving back to the topic of scaling in cities, and we are looking specifically at the question, why do larger cities tend to have more crime than smaller cities? But first, we will consider whether larger cities actually do have more crime.

The material in this post will eventually be on my Scaling in Human Societies project, but it is not yet.

What are we Measuring?

Before we proceed further, and at the risk of being pedantic, let us distinguish between overall crime levels and the crime rate. We should certainly expect that a large city will have a higher overall crime level than a small city. But most statistics on the subject focus on the crime rate, typically expressed as the number of offenses per year per 100,000 residents. For an ordinary person who is concerned about their safety, the rate is the relevant metric. Crime rates also tend to be greater in larger cities, and it is this fact that more cries out for an explanation.

If we are comparing cities by crime rate, we obviously have to know the population of a city to have any meaningful data. This is not so straightforward, since crimes are not necessarily tied to a city’s residential population. A city with a high ratio of offices and business to residences will have an inflated crime rate when going by standard population metrics. Of the papers I will review today, many are aware of this issue, but they don’t make a serious effort to account for it.

Crime and City Size

The relationship between crime and city size is well-studied, and we will look at a handful of the studies that have examined this relationship.

As Friendly (2007) explains, major work on crime and urbanization goes back at least to André-Michel Guerry (1833) and Babli and Guerry (1829) (in French). The latter work examines property crimes and personal crimes across different départements (administrative divisions) in France and finds that incidence of the two types of crimes are inversely related to each other, but both rates tend to be higher in more urbanized areas.

Hanley, Lewis, and Riberio (2016) consider the relationship between crime levels and population density in the 537 Parliamentary Constituencies in England and Wales. In assessing the quality of the prediction, they find that population density is a better predictor than overall population of a constituency, and that crime density yields better predictions that overall crime. This shouldn’t be too surprising. They find a scaling exponent of 1.16 ± 0.01, meaning that crime density grows with the 1.16 power of population density. In other words, if the population density doubles, then the crime density should increase by a factor of about 2.23.

For many crimes, Hanley, Lewis, and Riberio (2016) find "urban effects", or discontinuities in the scaling rates at certain population density thresholds. For the majority of crimes, including bike theft, drugs, order, other theft, robbery, theft from the person, and violence, an accelerated urban effect was found, by which the scaling factor increases above a population density threshold. Inhibited urban scaling, indicating that the scaling coefficient decreases above a population density threshold, was found for anti-social behavior, criminal damage and arson, and shoplifting. For several crime categories, including total crime, neither urban effect was found. All data in the analysis is publicly available, with crime data taken from the UK Home Office and population data taken from Ordnance Survey and the Office of National Statistics.

Considering cities in 12 countries, Oliveira (2016) examines rates of burglary and theft across cities of different sizes where statistics are available. He finds that in most countries for which data is available, theft increases superlinearly in city population size, meaning that as cities grow, the number of thefts per person increases. By contrast, he finds that burglary generally scales linearly with city size, meaning that as cities grow, the number of burglaries per person generally remains constant. However, in both cases, there is substantial variance across countries in the growth rate.

Nolan (2004) uses the the Uniform Crime Reports, published by the American Federal Bureau of Investigation, to establish a link between crime rates and city size. The UCR reported crimes rates for 1294 cities in the United States with a population of at least 25,000. The paper divides the cities into four groups, delineated by population size, and finds that crime rates are indeed higher on average for groupings of larger cities. Within most groups, the crime rate and population are also positively correlated. However, in the largest group--population at least 250,000--a negative correlation between city size and crime is found, with a negative correlation specifically for property crimes and no statistically significant correlation for violent crimes. This suggests that the crime/size relationship might break down for sufficiently large cities.

I’ve mentioned Bettencourt et al. (2007) in the past. Along with Bettencourt (2013), this paper takes a physics-inspired view of urban scaling, and it estimates how many different kinds of quantities scale with city size. It finds that crime scales approximately with the 1.16 power of population, which is very close to the 7/6 exponent that they predict from theory for the growth of various socioeconomic quantities, such as crime, GDP, and the spread of infectious disease. Bettencourt (2013) collects some additional data and finds a scaling exponent of 1.20 for cities in Japan in 2008. The source is cited as “research notes”, for which the author can be contacted for more information.

Gomez-Lievano, Youn, and Bettencourt (2012) consider national statistics on homicides in Colombia, Brazil, and Mexico, three countries with high crime rates. In 2007, the scaling exponents for these three countries were, respectively, 1.06, 1.12, and 1.35.

That’s enough for now. My takeaway is this. Within an urban system—generally taken to be a single country—crime rates are generally higher in large cities than in small cities. We consider an urban system as that in which cities are connected by a unified culture, free travel, common national policies, etc.; the relationship breaks down if we look beyond a single urban system and try, for instance, to compare a large city in Japan with a small city in Brazil. There are exceptions to this pattern, but the pattern seems to be robust. Now we consider why this is the case.

Why do large cities have more crime?

This is obviously a complicated question. As Chamlin and Cochran (2004) discuss, there are three main theories in criminology about the crime/city size relationship.

Structuralism

Unfortunately, the word “structuralism” is overloaded in academia, with meanings in linguistics, biology, mathematics, sociology, psychology, anthropology, and elsewhere. But this is the term that we have.

The structuralist perspective holds that crime is the opportunistic product of interpersonal interactions. Insofar as larger and denser cities foster more interactions in general, they foster criminal interactions in particular. This is very much the perspective of the Bettencourt papers. In them, crime is modeled in very much the same way as other socioeconomic phenomena, such as gross domestic product, patenting, and infectious disease.

Chamlin and Cochran (2004) observe that the structuralist perspective does not necessarily imply that larger cities have higher crime rates, because larger cities might promote criminal interactions as well as integrative (i.e. reduces crime) interactions. However, my read on the literature is the quantitative effect of these integrative interactions is not well defined, and that they would fully cancel out the criminal interactions is not something that should not be expected.

Social Control

The social control perspective considers crime to be antisocial behavior that occurs when the mechanisms of social control fail. Social control can divided into two forms: formal social control includes institutions such as the police, and informal social control consists of ordinary people who, in various ways, enforce norms.

Groff (2014) points out that the vast majority of social control is informal, since only a small fraction of the population can serve in formal roles such as police officers. Consequently, in larger cities, the personal knowledge that is necessary to institute informal social control is weaker, and thus there is less deterrent to crime and crime is more prevalent. She further divides informal social control into two posited mechanisms.

The “eyes on the street” perspective, associated with Jane Jacobs, holds that tight-knit neighborhoods help control crime by insuring that residents, as well as business owners and workers in the neighborhood, watch for criminal activity. The human territorial functioning (HTF) perspective holds that residents, especially homeowners, are incentivized to watch their neighborhoods through territorial instinct. “Eyes on the street” and HTF do offer some differing predictions on what will reduce crime. For example, the former perspective holds that mixed use development will reduce crime by putting more eyes on the street at all times, while the latter perspective holds that mixed use development will increase crime by creating holes in the neighborhood fabric and inviting people into the neighborhood who are less invested in its safety. However, Groff (2014) finds that there is little empirical evidence to assess which, if either, of these is right.

Subculturalism

Finally, the subcultural perspective posits that large cities will have more crime because large cities offer more opportunities for subcultures to emerge, including subcultures with deviant practices.

Blackman (2014) offers some historical perspective on the development of the notions of subculture, delinquency, and deviancy. The particular theory that large cities promote the development of deviant subculture was developed in Fischer (1975). However, Tittle (1989) argues that the theory can explain at most a fraction of the crime in large cities.

It is not easy to test these theories empirically. D’Alessio and Stolzenberg (2010) have done so by analyzing co-offending behavior across cities in the United States. Co-offense is when multiple people who are socially linked participate in a crime. If the subculturalism theory is sound, then co-offending should be more prevalent in larger cities, indicating the presence of deviant subcultures. However, this paper finds the opposite: co-offending is less common in larger cities. The authors conclude that this instead is evidence for the social breakdown model of crime, which is a failure of the mechanisms of social control.

Size Isn’t Everything

Criminology is a deep and complex topic about which I do not know much, and I only wanted to consider the specific question of the influence of city size on the crime rate. However, the following chart of the homicide rate in New York City over more than 200 years clearly demonstrates that there are important factors other than size at work.

The homicide rate in New York City from 1800 to 2023. Image from Vital City.

Obviously, changing population levels cannot explain the spike in homicides in the 1960s and fall in the 1990s, and I will not attempt to comment further.

I think the main takeaway from a policy perspective is that the issue of crime in large cities needs to be taken seriously. Cities are engines of economic activity and invention, and there is much evidence that cities perform better environmentally than suburban and rural living. However, large cities come with substantial drawbacks, and higher rates of crime is one of them. There is therefore not much more important to good urbanism than the effective management of crime.

Quick Hits

Open Philanthropy recently announced a $120 million Abundance and Growth Fund. The goal of this investment is to support economic growth, particularly by confronting various forms of institutional sclerosis. This is an exciting and badly needed venture, and I encourage any readers who have an interest in working in this topic to explore further. My Scaling in Human Societies project, which includes today’s topic of crime and city size, is supported by Open Philanthropy, though nothing I have written should be construed as officially representing OpenPhil’s views.