Street-Level Crime Patterns Diverge Sharply Across City Sizes, Landmark Study Finds

Crime researchers have long known that a city’s criminal activity is not spread evenly across its map. Instead, a small fraction of street segments—sometimes just one or two percent—hosts a disproportionate share of robberies, shootings, and burglaries, and this concentration remains remarkably stable over years. What has been far less clear is whether the forces that drive these micro-level patterns in a metropolis like Seattle or Chicago operate the same way in a small city like Reading, Pennsylvania, or Asheville, North Carolina. A sweeping new longitudinal study published in the American Journal of Criminal Justice tackles that question directly, and its answer challenges decades of assumptions built on big-city research.

Xiaoshuang Iris Luo of the University of Akron analyzed official violent and property crime data from 297,199 street segments across 12 U.S. cities between 2010 and 2018, drawing on the National Incident Crime Study compiled by the Irvine Laboratory for the Study of Space and Crime. The cities—ranging from Los Angeles and Seattle to Scottsdale, Akron, and Diamond Bar—were selected through stratified random sampling to form a balanced design across three dimensions: city size, population growth, and employment presence. This deliberate variation is the study’s methodological heart. Where nearly all prior micro-place research examined a single large city, Luo’s design allows the effects of neighborhood characteristics to be compared systematically across urban environments rather than inferred from convenience-selected cases.

The analytical engine of the study is group-based trajectory modeling, a statistical technique that clusters individual units—in this case, street segments—into groups sharing similar longitudinal patterns of crime. Rather than assigning segments to groups and then modeling the assignments separately, a common two-step practice that treats group membership as known and understates uncertainty, Luo estimated trajectory parameters and predictors of group membership jointly in a single likelihood. Each segment’s contribution is weighted by its full posterior probability distribution across groups, so classification uncertainty propagates into the estimated effects. Violent crime was modeled with a logit specification because roughly 90 percent of segments recorded no incidents, while property crime used a censored normal specification to handle skewed distributions.

The trajectory solutions revealed five distinct patterns for each crime type. For violent crime, 54.4 percent of street segments were crime-free throughout the nine-year period, 33.4 percent followed a low stable path, and only small minorities followed medium-high decreasing, medium decreasing-increasing, or high stable-decreasing trajectories. Property crime was more widely dispersed: 28.3 percent of segments were crime-free, 37.5 percent showed minimal stable activity, and 1.4 percent sat on a high stable trajectory. The low-crime groups dominate both outcomes, echoing the well-established law of crime concentration at places, yet the aggregate stability conceals substantial street-to-street heterogeneity even among adjacent segments.

Turning to explanations, the study confirms the central prediction of social disorganization theory: concentrated disadvantage—a composite of poverty, unemployment, single-parent households, and the absence of educational and economic resources—is strongly associated with elevated crime trajectories. A one standard deviation increase in disadvantage nearly doubles the odds that a street segment follows a low stable violent crime trajectory rather than remaining crime free, and more than triples the odds of following the highest violent crime trajectory. Consumer-facing employment, capturing the retail and service establishments that generate routine activity and draw visitors, showed among the strongest and most consistent gradients of all, with the odds of high crime membership rising dramatically as consumer-facing employment increased.

But the study’s most striking findings emerge when city context enters the picture. The criminogenic effect of disadvantage is not uniform across urban environments. For violent crime, a one standard deviation increase in disadvantage roughly doubles the odds of high crime trajectory membership in small cities, but increases those odds more than fivefold in medium cities and more than threefold in large ones. For property crime the contrast is sharper still: the association actually reverses direction in small cities, where greater disadvantage is modestly associated with lower odds of crime-active trajectories, while remaining clearly criminogenic in medium and large cities. Disadvantage, in other words, is not a fixed risk factor but one whose consequences depend on the scale of the surrounding urban environment.

Residential stability tells an equally context-dependent story. In small cities, stability is strongly protective: higher stability substantially reduces the odds that street segments fall into any crime-active violent crime trajectory, with the protective effect growing stronger as trajectories become more severe. In medium and large cities, this protective association is largely absent, and for property crime it reverses, with stability positively associated with membership in the lower crime trajectories. City growth conditions these relationships as well. Concentrated disadvantage predicts severe violent crime trajectories more strongly in stagnant cities than in growing ones, while residential stability is more protective against medium and high violent crime trajectories in growing cities. Growing cities may benefit from economic prosperity, new investment, and an influx of residents and businesses that curb violent crime, whereas stagnant cities face persistent disinvestment, declining municipal capacity, and limited resources for prevention.

These results carry a substantial implication: the neighborhood-crime relationships documented in the large-city literature may describe a special case rather than a general process. Social disorganization theory holds that structural disadvantage weakens informal social control, but the capacity for such control also depends on the density of local institutions, the responsiveness of municipal services, and the stability of the surrounding economy—all of which vary across cities. In a small city with an intact institutional fabric, disadvantage may carry different consequences than the same level of disadvantage in a large city where residents are more anonymous and criminal opportunities more abundant. Similarly, residential stability in a large city may reflect immobility and constrained housing options rather than the attachment and investment that generate protective social ties.

The study is candid about its limitations. The 12 cities were drawn through stratified random sampling from 108 eligible cities in the NICS database, ensuring balanced variation across the dimensions of interest but not constituting a simple random sample of U.S. cities, and the sample skews somewhat toward Southern California. Neighborhood measures are drawn from 2010 and treated as static over the nine-year observation period, so the design captures the association between baseline conditions and subsequent trajectories but cannot speak to whether changing neighborhood conditions produce changing crime patterns. Spatial dependence among nearby segments is not explicitly modeled, and collective efficacy—the shared willingness of residents to intervene that is central to contemporary social disorganization theory—requires survey data unavailable at the street-segment level across 12 cities, leaving the connecting mechanism unobserved.

Even with those caveats, the findings point toward a clear policy conclusion: one-size-fits-all crime prevention is unlikely to work. Efforts to reduce concentrated disadvantage may yield larger returns in medium and large cities, while programs that sustain residential stability appear most consequential in smaller and growing ones. Place-based interventions calibrated to a single metropolitan context cannot simply be transplanted elsewhere. As micro-level crime research moves beyond its traditional focus on large, well-resourced cities, the message of this study is that the street corner and the city around it form a single interacting system—and understanding crime where most Americans actually live requires studying both at once.

Subject of Research: Longitudinal trajectories of violent and property crime at street segments across varied U.S. city contexts

Article Title: Micro-Level Crime Trajectories Across Urban Contexts: A Multi-City Longitudinal Study in the U.S.

Article References: Luo, X. I. (2026). Micro-Level Crime Trajectories Across Urban Contexts: A Multi-City Longitudinal Study in the U.S.. American Journal of Criminal Justice. https://doi.org/10.1007/s12103-026-09952-w

Image Credits: AI Generated

DOI: 10.1007/s12103-026-09952-w

Keywords: crime trajectories, street segments, social disorganization theory, concentrated disadvantage, residential stability, city context, group-based trajectory modeling, crime concentration, urban criminology, property crime, violent crime, crime prevention